PHMForge

{"id": "pdm_rul_001", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD001", "question": "I am managing a fleet of aircraft engines and need to know which ones are running on fumes. Can you predict the remaining useful life for all test engines in the CMAPSS FD001 dataset? I need predictions for every unit, verification against the actual RUL values we have, and then tell me which 10 engines are most likely to fail soonest. Also give me the overall accuracy numbers-mean absolute error and root mean squared error-so I can see how reliable these predictions are. I really need concrete numbers here, not estimates, because I am presenting this to maintenance planning the next scheduled maintenance planning meeting.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "{\"problem\": \"Predict the remaining useful life (RUL) for all test equipment units in the CMAPSS FD001 dataset.\", \"steps\": \"1) Load the CMAPSS FD001 training dataset using load_dataset(dataset_name='CMAPSS_FD001', split='train'). 2) Load the corresponding test dataset using load_dataset(dataset_name='CMAPSS_FD001', split='test'). 3) Load the ground truth RUL values using load_ground_truth(dataset_name='CMAPSS_FD001', file='RUL_FD001.txt'). 4) Train a RUL prediction model using train_rul_model(dataset='CMAPSS_FD001', model_type='transformer', epochs=50). 5) For each test unit, predict RUL using predict_rul(model=model, test_data=test_data, unit_id=unit_id). 6) Verify all predictions against ground truth using verify_ground_truth(predictions=predictions, ground_truth=ground_truth). 7) Calculate Mean Absolute Error (MAE) using calculate_mae(predictions=predictions, ground_truth=ground_truth). 8) Calculate Root Mean Squared Error (RMSE) using calculate_rmse(predictions=predictions, ground_truth=ground_truth). 9) Identify the top 10 units with lowest predicted RUL (highest failure risk). 10) Produce a JSON report containing: all_unit_predictions (array of {unit_id, predicted_rul, actual_rul, error}), top_10_risk_units (array sorted by predicted_rul ascending), overall_mae, overall_rmse, and ground_truth_verification_status.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 17.6, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_mae\": 10.49, \"sota_model\": \"CAELSTM\", \"sota_rmse\": 14.44, \"sota_source\": \"Nature Scientific Reports 2025 - Deep learning-based prognostic approach\"}, \"expected_mae_range\": [7, 18], \"expected_mae_range_rationale\": \"Range [7, 18] derived from published CMAPSS_FD001 benchmarks. CAELSTM (Conv AE + Attention LSTM) achieves MAE=10.49 cycles (Nature Scientific Reports 2025). BiLSTM baseline achieves MAE~12-15 cycles. Range encompasses SOTA to baseline performance with buffer for implementation variance.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [11, 26], \"expected_rmse_range_rationale\": \"Range [11, 26] derived from published CMAPSS_FD001 benchmarks. CAELSTM achieves RMSE=14.44 cycles (Nature Scientific Reports 2025). BiLSTM baseline achieves RMSE=17.60 cycles (arXiv 2025). Self-attention+BiGRU+BiLSTM achieves RMSE=11.74 (ScienceDirect 2024). Range encompasses SOTA to baseline performance.\", \"expected_top_units_count\": 10, \"mae_range\": [12.5, 25.1], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 10 unit IDs (int)\"}, \"rationale\": \"Based on ground truth RUL values from RUL_FD001.txt, these 10 units have the lowest actual RUL values indicating highest failure risk.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [16.7, 33.4], \"sample_correct_units\": [33, 30, 67, 80, 81, 41, 75, 34, 65, 55], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Sequential dependencies: load_dataset(train) -> provides training data for model training; load_dataset(test) -> provides test units for prediction; load_ground_truth -> required for verification and metric calculation. train_rul_model -> produces model object that must be used in predict_rul calls. predict_rul -> generates predictions array that feeds into verify_ground_truth, calculate_mae, and calculate_rmse. Decision point: after predictions are generated, identify top 10 units with lowest RUL values. All metric calculations depend on both predictions and ground_truth arrays. The workflow requires strict sequential ordering: data loading -> model training -> prediction -> verification -> metric calculation -> risk ranking.\"}", "known_issue": ""}
{"id": "pdm_rul_002", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD002", "question": "We have got bearing vibration data from CWRU and I need to figure out which bearings are about to give out. Can you predict how much life each test bearing has left, check those predictions against the actual failure times we recorded, and then show me degradation curves for the 5 bearings most at risk? I need the accuracy numbers too-MAE and RMSE-so I know how much to trust these predictions when scheduling maintenance.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "detect_degradation_trend", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for bearing units in the CWRU dataset and analyze degradation trends.\", \"steps\": \"1) Load CWRU training data using load_dataset(dataset_name='CWRU', split='train'). 2) Load CWRU test data using load_dataset(dataset_name='CWRU', split='test'). 3) Load ground truth RUL values. 4) Train a bearing-specific RUL model using train_rul_model(dataset='CWRU', model_type='lstm', features=['vibration_horizontal', 'vibration_vertical'], epochs=100). 5) Predict RUL for all test bearing units. 6) Verify predictions against ground truth. 7) Calculate MAE and RMSE. 8) Generate degradation trend plots using plot_degradation_trends(unit_ids=top_5_risk_units, predictions=predictions, ground_truth=ground_truth). 9) Output JSON report with predictions, metrics, and visualization file paths.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 29.67, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"dataset_characteristics\": \"6 operating conditions, 1 fault mode (HPC degradation)\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD002 benchmarks. BiLSTM baseline achieves RMSE=29.67 cycles (arXiv 2025). FD002 has 6 operating conditions making prediction significantly harder than FD001. Range accounts for multi-condition complexity.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD002 benchmarks. BiLSTM baseline achieves RMSE=29.67 cycles (arXiv 2025). The 6 operating conditions introduce variance compared to single-condition FD001. Range reflects increased dataset complexity.\", \"expected_top_units_count\": 5, \"mae_range\": [16.2, 32.3], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on CWRU dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [21.6, 43.1], \"sample_correct_units\": [6, 10, 43, 176, 210], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Data loading -> model training -> prediction -> verification -> metric calculation -> visualization. The plot_degradation_trends tool requires both predictions and ground_truth, and operates on the top 5 risk units identified from predictions.\"}", "known_issue": "The question text names CWRU, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_003", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Bearing", "source_dataset": "FEMTO", "question": "The FEMTO dataset has bearings running under three different operational scenarios, and I suspect the RUL predictions might vary by scenario. Can you predict remaining life for all test bearings, break down the accuracy (MAE, RMSE) for each of the three operational scenarios separately, and tell me which scenario gives us the most reliable predictions? I need to know if we should adjust our maintenance strategy based on operating conditions.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "analyze_engine_signals", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for bearings in FEMTO dataset across three operational scenarios and compare performance.\", \"steps\": \"1) Load FEMTO training data. 2) Load FEMTO test data. 3) Load ground truth RUL. 4) Identify three operational scenarios using identify_operational_scenarios(dataset='FEMTO'). 5) Train scenario-specific RUL models for each operational condition. 6) Predict RUL for test units in each scenario. 7) Verify all predictions. 8) Calculate MAE and RMSE per scenario. 9) Compare scenario performance and identify which operational condition has highest prediction accuracy. 10) Output JSON report with per-scenario predictions, metrics, and comparison analysis.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_mae\": 0.056, \"baseline_model\": \"TCNMABG\", \"baseline_rmse\": 0.082, \"baseline_source\": \"MDPI Applied Sciences 2024\", \"metric_note\": \"FEMTO metrics in percentage of lifetime (0-1 scale)\", \"sota_model\": \"CNN-GRU-MHA\", \"sota_rmse\": 0.0443, \"sota_source\": \"MDPI 2024 - Transfer Learning for Bearing RUL\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published FEMTO/PRONOSTIA benchmarks. TCNMABG achieves MAE=0.056 (normalized lifetime percentage, MDPI). CNN-GRU-MHA with transfer learning achieves RMSE=0.0443 (MDPI 2024). Note: FEMTO metrics often expressed as percentage of lifetime rather than absolute cycles.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published FEMTO benchmarks. CNN-GRU-MHA achieves RMSE=0.0443 (MDPI 2024). TCNMABG achieves RMSE=0.082 (MDPI). TcLstmNet-CBAM achieved best performance on 11/14 bearings (Nature Scientific Reports 2025).\", \"expected_top_units_count\": 5, \"mae_range\": [8.7, 17.4], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on FEMTO dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [11.6, 23.3], \"sample_correct_units\": [0, 200, 400, 300, 600], \"verification_required\": true}", "metadata": "{\"complexity_level\": \"medium\", \"dependency_analysis\": \"Load data -> identify scenarios -> train separate models per scenario -> predict per scenario -> verify per scenario -> calculate metrics per scenario -> comparative analysis.\"}", "known_issue": "FEMTO: the expected MAE/RMSE ranges do not state their unit (rul_percentage is 0-1); a constant 50% prediction lands just outside the MAE range."}
{"id": "pdm_rul_004", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD004", "question": "The XJTU dataset has bearings with five different fault types, and I want to see if our RUL predictions are more accurate for certain fault types. Can you predict remaining life for all test bearings, give me overall accuracy numbers, then break it down by fault type so I can see which fault types we predict best? This will help us prioritize which failure modes to focus our maintenance efforts on.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "classify_faults", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for XJTU bearing dataset considering five different fault types.\", \"steps\": \"1) Load XJTU training and test data. 2) Load ground truth RUL. 3) Analyze fault types present using analyze_fault_types(dataset='XJTU'). 4) Train fault-type-specific RUL models. 5) Predict RUL for all test units. 6) Verify predictions. 7) Calculate overall MAE and RMSE. 8) Calculate per-fault-type MAE and RMSE. 9) Identify which fault type has most accurate predictions. 10) Output JSON report with overall metrics, per-fault-type metrics, and fault-type-specific predictions.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 31.84, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"dataset_characteristics\": \"6 operating conditions, 2 fault modes (HPC degradation + Fan degradation), most complex CMAPSS subset\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD004 benchmarks. BiLSTM baseline achieves RMSE=31.84 cycles (arXiv 2025). FD004 is the most complex CMAPSS subset with 6 operating conditions AND 2 fault modes, making it significantly harder than FD001-FD003.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD004 benchmarks. BiLSTM baseline achieves RMSE=31.84 cycles (arXiv 2025). The combination of 6 operating conditions and 2 fault modes creates the highest prediction difficulty in CMAPSS family.\", \"expected_top_units_count\": 5, \"mae_range\": [16.4, 32.8], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on CMAPSS_FD004 dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [21.9, 43.7], \"sample_correct_units\": [30, 157, 11, 99, 134], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Data loading -> fault type analysis -> train fault-specific models -> predict -> verify -> calculate overall metrics -> calculate per-fault metrics -> comparative analysis.\"}", "known_issue": "The question text names XJTU, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_005", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Bearing", "source_dataset": "FEMTO", "question": "The IMS dataset has very high-frequency vibration data, and I am worried that units showing anomalies might have different RUL prediction accuracy. Can you predict remaining life for all test bearings, detect any anomalies in the sensor data, and then compare prediction accuracy between units with anomalies versus normal units? I need to know if we should treat anomalous units differently in our maintenance planning.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "detect_degradation_trend", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for IMS high-temporal-resolution bearing data with anomaly detection integration.\", \"steps\": \"1) Load IMS training data (high temporal resolution). 2) Load IMS test data. 3) Load ground truth RUL. 4) Perform anomaly detection on test data using detect_anomalies(dataset='IMS', method='isolation_forest'). 5) Train RUL model on training data. 6) Predict RUL for all test units. 7) Flag predictions for units with detected anomalies. 8) Verify predictions against ground truth. 9) Calculate MAE and RMSE for all units. 10) Calculate MAE and RMSE separately for units with and without anomalies. 11) Output JSON report with predictions, anomaly flags, overall metrics, and anomaly-specific metrics.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"TCNMABG\", \"baseline_rmse\": 0.082, \"baseline_source\": \"MDPI Applied Sciences\", \"metric_note\": \"FEMTO metrics in percentage of lifetime (0-1 scale)\", \"sota_model\": \"TcLstmNet-CBAM\", \"sota_source\": \"Nature Scientific Reports 2025\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published FEMTO benchmarks. TcLstmNet-CBAM achieved best MAE on 11/14 bearings (Nature Scientific Reports 2025). Range reflects operational scenario variability across three test conditions.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published FEMTO benchmarks. CNN-GRU-MHA achieves RMSE=0.0443 (MDPI 2024). TCNMABG achieves RMSE=0.082 (MDPI). Range accounts for anomaly detection integration.\", \"expected_top_units_count\": 5, \"mae_range\": [8.7, 17.4], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on IMS dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [11.6, 23.3], \"sample_correct_units\": [0, 200, 400, 300, 600], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> anomaly detection -> model training -> prediction -> anomaly flagging -> verification -> overall metrics -> anomaly-stratified metrics.\"}", "known_issue": "The question text names IMS, while the `source_dataset` field and ground truth refer to FEMTO; the ground-truth units could not be re-derived from files in the repository."}
{"id": "pdm_rul_006", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD001", "question": "The HUST dataset has bearings tested under different speeds and loads, which is great for seeing if our RUL predictions work across diverse conditions. Can you predict remaining life for all test bearings, give me overall accuracy, then break it down by speed/load combinations? I need to know if the model generalizes well or if we need condition-specific models.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "analyze_engine_signals", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for HUST bearings under varying speeds and loads, analyzing generalization across operating conditions.\", \"steps\": \"1) Load HUST training data (collected under varying speeds and loads). 2) Load HUST test data. 3) Load ground truth RUL. 4) Analyze operating conditions using analyze_operating_conditions(dataset='HUST'). 5) Train RUL model on diverse operating conditions. 6) Predict RUL for all test units. 7) Verify predictions. 8) Calculate overall MAE and RMSE. 9) Group predictions by speed/load combinations and calculate metrics per group. 10) Assess model generalization across operating conditions. 11) Output JSON report with predictions, overall metrics, per-condition metrics, and generalization analysis.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 17.6, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_model\": \"Self-attention+BiGRU+BiLSTM\", \"sota_rmse\": 11.74, \"sota_source\": \"ScienceDirect 2024 - Multiscale Deep Transfer Learning\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD001 benchmarks. CAELSTM achieves MAE=10.49 cycles (Nature Scientific Reports 2025). BiLSTM baseline achieves RMSE=17.60 cycles (arXiv 2025). FD001 has single operating condition enabling focused analysis.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD001 benchmarks. Self-attention+BiGRU+BiLSTM achieves RMSE=11.74 (ScienceDirect 2024). CAELSTM achieves RMSE=14.44 (Nature 2025). Operating condition analysis focuses on single-condition performance.\", \"expected_top_units_count\": 5, \"mae_range\": [12.5, 25.1], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on HUST dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [16.7, 33.4], \"sample_correct_units\": [33, 30, 67, 80, 81], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> analyze operating conditions -> train model -> predict -> verify -> overall metrics -> group by conditions -> per-condition metrics -> generalization assessment.\"}", "known_issue": "The question text names HUST, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_007", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD002", "question": "The MFPT dataset was captured at a very high sampling rate (97.656KHz), so there is probably useful information in the frequency domain. Can you predict remaining life for all test bearings using both time and frequency domain features, verify the predictions, and tell me if adding frequency analysis improves accuracy compared to just using time-domain data? I need concrete numbers to justify the extra processing.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "analyze_engine_signals", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for MFPT bearings using high-frequency vibration data (97.656KHz sampling rate) with frequency domain analysis.\", \"steps\": \"1) Load MFPT training data (high sampling rate). 2) Load MFPT test data. 3) Load ground truth RUL. 4) Perform frequency domain analysis using analyze_frequency_domain(dataset='MFPT', sampling_rate=97656). 5) Extract frequency-domain features. 6) Train RUL model using both time-domain and frequency-domain features. 7) Predict RUL for all test units. 8) Verify predictions. 9) Calculate MAE and RMSE. 10) Compare prediction accuracy with time-domain-only vs. time+frequency domain features. 11) Output JSON report with predictions, metrics, and feature importance analysis.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 29.67, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"dataset_characteristics\": \"6 operating conditions, frequency domain analysis applicable\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD002 benchmarks. BiLSTM baseline achieves RMSE=29.67 cycles (arXiv 2025). FD002 has 6 operating conditions requiring frequency domain analysis for improved prediction.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD002 benchmarks. BiLSTM achieves RMSE=29.67 cycles (arXiv 2025). Frequency domain features can improve multi-condition prediction accuracy.\", \"expected_top_units_count\": 5, \"mae_range\": [16.2, 32.3], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on MFPT dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [21.6, 43.1], \"sample_correct_units\": [6, 10, 43, 176, 210], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> frequency domain analysis -> feature extraction -> train model with combined features -> predict -> verify -> calculate metrics -> compare feature contributions.\"}", "known_issue": "The question text names MFPT, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_008", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD003", "question": "The Paderborn dataset has multiple sensor types-vibration, motor current, torque, and speed-and I want to know which ones actually help predict bearing failure. Can you predict remaining life using all sensors together, then test each sensor type individually to see which gives the best predictions? I need to know if we can simplify our sensor setup or if we need all of them.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "analyze_engine_signals", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for Paderborn bearings using multimodal sensor data (vibration, motor current, torque, speed).\", \"steps\": \"1) Load Paderborn training data (multimodal: vibration, current, torque, speed). 2) Load Paderborn test data. 3) Load ground truth RUL. 4) Analyze multimodal sensor contributions using analyze_multimodal_sensors(dataset='Paderborn', sensors=['vibration', 'current', 'torque', 'speed']). 5) Train RUL model using all sensor modalities. 6) Predict RUL for all test units. 7) Verify predictions. 8) Calculate MAE and RMSE. 9) Perform ablation study: predict using each sensor modality individually and compare accuracy. 10) Identify which sensor modality contributes most to prediction accuracy. 11) Output JSON report with predictions, overall metrics, per-sensor metrics, and sensor importance ranking.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 17.62, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_model\": \"Self-attention+BiGRU+BiLSTM\", \"sota_rmse\": 13.5, \"sota_source\": \"ScienceDirect 2024 - Multi-attention Mechanism\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD003 benchmarks. BiLSTM baseline achieves RMSE=17.62 cycles (arXiv 2025). FD003 has 1 operating condition but 2 fault modes, making it moderately complex. Range accounts for baseline performance with implementation variance.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD003 benchmarks. BiLSTM achieves RMSE=17.62 cycles (arXiv 2025 - Uncertainty-Aware DL Framework). Self-attention+BiGRU+BiLSTM models achieve improved RMSE~13-15 cycles. Range encompasses baseline to SOTA performance.\", \"expected_top_units_count\": 5, \"mae_range\": [12.5, 25.0], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on Paderborn dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [16.6, 33.3], \"sample_correct_units\": [81, 45, 38, 98, 23], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load multimodal data -> analyze sensor contributions -> train model with all sensors -> predict -> verify -> calculate metrics -> train separate models per sensor -> compare accuracy -> rank sensor importance.\"}", "known_issue": "The question text names Paderborn, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_009", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD004", "question": "The Mendeley dataset has bearings running at speeds that constantly change between 600 and 1800 RPM, making the vibration signals non-stationary. Can you predict remaining life for all test bearings, accounting for these changing conditions, and show me how prediction accuracy varies across different speed ranges? I need to know if our model handles variable-speed operation reliably.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for Mendeley bearings under non-stationary conditions (varying speed 600-1800RPM).\", \"steps\": \"1) Load Mendeley training data (non-stationary: speed 600-1800RPM). 2) Load Mendeley test data. 3) Load ground truth RUL. 4) Handle non-stationary signal characteristics using handle_non_stationary_data(dataset='Mendeley', method='adaptive_filtering'). 5) Train RUL model adapted for non-stationary conditions. 6) Predict RUL for all test units. 7) Verify predictions. 8) Calculate MAE and RMSE. 9) Analyze prediction accuracy across different speed ranges. 10) Output JSON report with predictions, metrics, and speed-range-specific accuracy analysis.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 31.84, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_model\": \"Multi-attention Mechanism\", \"sota_rmse\": 26.5, \"sota_source\": \"ScienceDirect 2024\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD004 benchmarks. BiLSTM baseline achieves RMSE=31.84 cycles (arXiv 2025). FD004 is the most complex subset with 6 operating conditions and 2 fault modes. Higher range reflects dataset difficulty.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD004 benchmarks. BiLSTM achieves RMSE=31.84 cycles (arXiv 2025 - Uncertainty-Aware DL Framework). Multi-attention models achieve RMSE~25-28 cycles. Range accounts for high dataset complexity with buffer for implementation variance.\", \"expected_top_units_count\": 5, \"mae_range\": [16.4, 32.8], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on Mendeley dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [21.9, 43.7], \"sample_correct_units\": [30, 157, 11, 99, 134], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load non-stationary data -> apply non-stationary handling -> train adapted model -> predict -> verify -> calculate metrics -> analyze by speed range.\"}", "known_issue": ""}
{"id": "pdm_rul_010", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Bearing", "source_dataset": "FEMTO", "question": "The Azure dataset is huge-over 876,000 samples-so I need RUL predictions that can actually run efficiently. Can you predict remaining life for all test units, verify the predictions, give me accuracy numbers, and also tell me how long it took and how much memory it used? I need to know if this is practical for production deployment.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "{\"problem\": \"Predict RUL for Azure dataset (876K+ samples) requiring efficient large-scale processing.\", \"steps\": \"1) Load Azure training data (876K+ samples, efficient loading required). 2) Load Azure test data. 3) Load ground truth RUL. 4) Handle large dataset efficiently using handle_large_dataset(dataset='Azure', method='batch_processing', batch_size=10000). 5) Train RUL model with efficient batch training. 6) Predict RUL for all test units using batch prediction. 7) Verify predictions. 8) Calculate MAE and RMSE. 9) Measure processing time and memory usage. 10) Output JSON report with predictions, metrics, and performance benchmarks (time, memory).\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_mae\": 0.056, \"baseline_model\": \"TCNMABG\", \"baseline_rmse\": 0.082, \"baseline_source\": \"MDPI 2024 - Temporal Convolutional Multi-Attention\", \"sota_model\": \"CNN-GRU-MHA\", \"sota_rmse\": 0.0443, \"sota_source\": \"MDPI 2024 - Transfer Learning for Bearing RUL\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published FEMTO/PRONOSTIA benchmarks. TCNMABG achieves MAE=0.056 (normalized percentage of lifetime, MDPI 2024). TcLstmNet-CBAM achieved best MAE on 11/14 bearings (Nature Scientific Reports 2025). Range converted to cycle-equivalent units.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published FEMTO/PRONOSTIA benchmarks. CNN-GRU-MHA with transfer learning achieves RMSE=0.0443 (MDPI 2024). TCNMABG achieves RMSE=0.082 (MDPI). Note: FEMTO metrics often in percentage of lifetime; range accounts for conversion variance.\", \"expected_top_units_count\": 5, \"mae_range\": [8.7, 17.4], \"output_template\": {\"all_unit_predictions\": [{\"actual_rul\": \"float\", \"error\": \"float\", \"predicted_rul\": \"float\", \"unit_id\": \"int\"}], \"ground_truth_verification_status\": \"bool\", \"overall_mae\": \"float\", \"overall_rmse\": \"float\", \"top_N_risk_units\": \"array of 5 unit IDs (int)\"}, \"rationale\": \"Based on Azure dataset ground truth, these 5 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"all_unit_predictions\", \"top_N_risk_units\", \"overall_mae\", \"overall_rmse\", \"ground_truth_verification_status\"], \"rmse_range\": [11.6, 23.3], \"sample_correct_units\": [0, 200, 400, 300, 600], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load large dataset efficiently -> batch processing -> train with batches -> batch prediction -> verify -> calculate metrics -> performance benchmarking.\"}", "known_issue": "FEMTO: the expected MAE/RMSE ranges do not state their unit (rul_percentage is 0-1); a constant 50% prediction lands just outside the MAE range."}
{"id": "pdm_fault_001", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "CWRU", "question": "I have got CWRU bearing data with different fault types-inner race defects, outer race defects, and ball race defects. Can you classify which fault type each test bearing has, check those classifications against the actual fault labels, and give me accuracy numbers plus a breakdown of how well you classified each fault type? I need precision, recall, and F1-scores for each fault category so I know where the model struggles.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "{\"problem\": \"Classify bearing faults in CWRU dataset (inner race, outer race, ball race defects).\", \"steps\": \"1) Load CWRU training data. 2) Load CWRU test data. 3) Load ground truth fault labels. 4) Train fault classification model using train_fault_classifier(dataset='CWRU', fault_types=['inner_race', 'outer_race', 'ball_race'], model_type='cnn'). 5) Classify faults for all test units using classify_faults(model=model, test_data=test_data). 6) Verify classifications against ground truth using verify_classification(predictions=predictions, ground_truth=ground_truth). 7) Calculate overall classification accuracy using calculate_accuracy(predictions=predictions, ground_truth=ground_truth). 8) Generate confusion matrix using calculate_confusion_matrix(predictions=predictions, ground_truth=ground_truth). 9) Calculate per-fault-type precision, recall, and F1-score. 10) Output JSON report with classifications, overall accuracy, confusion matrix, and per-fault-type metrics.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.99, \"baseline_model\": \"CNN-LSTM-GRU\", \"baseline_source\": \"MDPI 2024 - Hybrid Deep Learning for Bearing Fault Classification\", \"sota_accuracy\": 1.0, \"sota_model\": \"CNN-BLSTM\", \"sota_source\": \"Wiley 2021 - Bidirectional LSTM for Rotating Machinery Fault Diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] derived from published CWRU bearing fault classification benchmarks. CNN-BLSTM achieves 100% accuracy (Wiley 2021). CNN-LSTM-GRU achieves >99% accuracy (MDPI 2024). Multi-scale CNN+LSTM achieves 98.46% (Papers with Code). Lower bound accounts for cross-condition generalization and noisy environments.\", \"expected_fault_types\": [\"Fault_0\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on CWRU dataset ground truth fault labels. Accuracy range encompasses published benchmark performance with implementation variance.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [137, 41, 142, 87, 134, 216, 155, 175], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> train classifier -> classify -> verify -> calculate overall accuracy -> generate confusion matrix -> calculate per-fault metrics.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_002", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "XJTU", "question": "The XJTU dataset has five different fault types, and I suspect we might classify early-stage faults differently than late-stage ones. Can you classify all test bearings by fault type, give me overall and per-fault-type accuracy, and then break down accuracy by fault severity? I need to know if we catch faults better when they're more advanced.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "{\"problem\": \"Classify five different fault types in XJTU dataset and analyze classification performance by fault severity.\", \"steps\": \"1) Load XJTU training data. 2) Load XJTU test data. 3) Load ground truth fault labels (5 fault types). 4) Analyze fault severity levels using analyze_fault_severity(dataset='XJTU'). 5) Train fault classifier for 5 fault types. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Calculate per-fault-type accuracy. 10) Analyze classification accuracy by fault severity (early vs. late stage faults). 11) Output JSON report with classifications, overall accuracy, per-fault-type accuracy, and severity-stratified accuracy.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.92, \"baseline_model\": \"CNN\", \"baseline_source\": \"XJTU-SY benchmark papers\", \"sota_accuracy\": 0.97, \"sota_model\": \"Transfer Learning CNN\", \"sota_source\": \"Domain adaptation for bearing fault diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] derived from published XJTU-SY bearing fault classification benchmarks. XJTU has 5 fault types under varying operating conditions. Deep learning models achieve 90-98% accuracy with cross-condition validation. Lower bound reflects severity stratification complexity.\", \"expected_fault_types\": [\"Outer race\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on XJTU-SY dataset ground truth fault labels across 5 fault categories.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> analyze severity -> train classifier -> classify -> verify -> overall accuracy -> per-fault accuracy -> severity-stratified analysis.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_003", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Electric motor", "source_dataset": "ElectricMotorVibrations", "question": "We have both acoustic and vibration data for electric motors, and I want to know if using both helps classify faults better than using just one. Can you classify all test motor faults using the combined acoustic and vibration data, then test each modality separately to see which gives better accuracy? I need concrete numbers to decide if we need both sensor types.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy", "analyze_engine_signals"], "procedure": "{\"problem\": \"Classify motor faults using multimodal acoustic and vibrational data from Electric Motor Vibrations dataset.\", \"steps\": \"1) Load Electric Motor Vibrations training data (acoustic + vibrational). 2) Load test data. 3) Load ground truth fault labels. 4) Analyze multimodal signal contributions using analyze_multimodal_signals(dataset='ElectricMotorVibrations', modalities=['acoustic', 'vibrational']). 5) Train fault classifier using both acoustic and vibrational features. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Compare accuracy using acoustic-only vs. vibrational-only vs. combined modalities. 10) Output JSON report with classifications, overall accuracy, and modality comparison analysis.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.9, \"baseline_model\": \"Single-modal CNN\", \"baseline_source\": \"Electric motor fault diagnosis benchmarks\", \"sota_accuracy\": 0.97, \"sota_model\": \"Multimodal Fusion CNN\", \"sota_source\": \"Acoustic-vibration fusion for motor diagnostics\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for multimodal motor fault classification. Fusion of acoustic and vibration data typically improves accuracy by 3-8% over single modality. CNN-based multimodal classifiers achieve 95-99% on similar motor fault datasets.\", \"expected_fault_types\": [\"Fault_0\", \"Fault_1\", \"Fault_2\", \"Fault_3\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on ElectricMotorVibrations dataset ground truth with multimodal sensor data.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load multimodal data -> analyze modalities -> train combined classifier -> classify -> verify -> calculate accuracy -> train separate classifiers per modality -> compare accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_004", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Induction motor", "source_dataset": "RotorBrokenBar", "question": "We have induction motor data with broken rotor bars, and the severity ranges from 1 to 4 broken bars. Can you classify the severity level for each test motor, verify against the actual labels, and give me accuracy broken down by severity level? I also need a confusion matrix to see if the model confuses certain severity levels-like if it mixes up 2 bars with 3 bars.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "{\"problem\": \"Classify broken rotor bar faults in induction motors with severity levels (1-4 broken bars).\", \"steps\": \"1) Load RotorBrokenBar training data. 2) Load test data. 3) Load ground truth labels (severity: 1-4 broken bars). 4) Analyze fault severity levels using analyze_fault_severity_levels(dataset='RotorBrokenBar', severity_range=[1,2,3,4]). 5) Train multi-class classifier for 4 severity levels. 6) Classify fault severity for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Calculate per-severity-level accuracy (1 bar vs. 2 bars vs. 3 bars vs. 4 bars). 10) Generate confusion matrix showing classification errors between severity levels. 11) Output JSON report with classifications, overall accuracy, per-severity accuracy, and confusion matrix.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.88, \"baseline_model\": \"CNN\", \"baseline_source\": \"Rotor fault severity classification benchmarks\", \"sota_accuracy\": 0.95, \"sota_model\": \"Deep Residual Network\", \"sota_source\": \"Induction motor fault severity diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for broken rotor bar severity classification. Multi-class severity classification (1-4 broken bars) is challenging due to subtle signal differences. CNN-based models achieve 90-96% accuracy on similar severity grading tasks.\", \"expected_fault_types\": [\"Fault_0\", \"Fault_1\", \"Fault_2\", \"Fault_3\", \"Fault_4\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on RotorBrokenBar dataset ground truth with 4 severity levels.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> analyze severity levels -> train multi-class classifier -> classify -> verify -> overall accuracy -> per-severity accuracy -> confusion matrix generation.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_005", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Gearbox", "source_dataset": "PlanetaryPdM", "question": "The planetary gearbox dataset has faults in different components-sun gear cracks, planet spalling, ring gear issues. Can you classify which component has the fault for each test gearbox, verify those classifications, and give me accuracy numbers overall and per component? This is for root cause analysis, so I need to know which parts are failing most often.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "{\"problem\": \"Classify and localize faults in planetary gearbox (sun gear crack, planet spalling, etc.) for root cause analysis.\", \"steps\": \"1) Load PlanetaryPdM training data. 2) Load test data. 3) Load ground truth fault labels (component-specific: sun gear, planet, etc.). 4) Localize fault components using localize_fault_components(dataset='PlanetaryPdM', components=['sun_gear', 'planet_gear', 'ring_gear']). 5) Train component-specific fault classifier. 6) Classify faults and identify affected components for all test units. 7) Verify classifications and component localizations. 8) Calculate overall classification accuracy. 9) Calculate per-component classification accuracy. 10) Output JSON report with fault classifications, component localizations, overall accuracy, and per-component accuracy.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.89, \"baseline_model\": \"CNN-1D\", \"baseline_source\": \"Planetary gearbox fault diagnosis benchmarks\", \"sota_accuracy\": 0.96, \"sota_model\": \"Attention-based CNN\", \"sota_source\": \"Component-level gearbox fault localization\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for planetary gearbox component-level fault classification. Localizing faults to sun/planet/ring gear requires fine-grained feature extraction. Deep learning achieves 88-97% accuracy on gearbox fault localization tasks.\", \"expected_fault_types\": [\"Fault_0\", \"Fault_1\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on PlanetaryPdM dataset ground truth with component-specific fault labels.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> component localization -> train component-specific classifier -> classify and localize -> verify -> overall accuracy -> per-component accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_006", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "Paderborn", "question": "The Paderborn dataset has both artificially induced and naturally occurring bearing faults. Can you classify the fault types for all test bearings, distinguish whether each fault was artificial or natural, and compare classification accuracy between the two? I need to know if our model works equally well on both types or if we need separate models.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "{\"problem\": \"Classify bearing faults in Paderborn dataset, distinguishing between artificially induced and naturally induced degradation.\", \"steps\": \"1) Load Paderborn training data (artificial + natural faults). 2) Load test data. 3) Load ground truth labels (fault type + induction method). 4) Distinguish artificial vs. natural faults using distinguish_artificial_vs_natural(dataset='Paderborn'). 5) Train classifier to identify both fault type and induction method. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Calculate accuracy separately for artificial vs. natural fault classifications. 10) Analyze if model performs differently on artificially induced vs. naturally occurring faults. 11) Output JSON report with classifications, overall accuracy, and artificial vs. natural accuracy comparison.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.91, \"baseline_model\": \"CNN\", \"baseline_source\": \"Paderborn bearing fault benchmark\", \"sota_accuracy\": 0.99, \"sota_model\": \"FaultNet\", \"sota_source\": \"arXiv - Deep Learning for Bearing Fault Diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] derived from published Paderborn bearing fault classification benchmarks. FaultNet achieves ~99% accuracy (arXiv). Distinguishing artificial vs. natural faults adds complexity. Domain adaptation methods achieve 92-98% accuracy on artificial/natural discrimination.\", \"expected_fault_types\": [\"inner_race\", \"outer_race\", \"ball_race\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on Paderborn dataset ground truth with artificial and natural fault labels.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> distinguish induction methods -> train classifier -> classify -> verify -> overall accuracy -> separate accuracy for artificial vs. natural -> comparison analysis.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_007", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "HUST", "question": "The HUST dataset has 5 different bearing types with 6 fault categories, all tested under varying speeds and loads. Can you classify the fault category for each test bearing, give me overall accuracy plus breakdowns by fault category and bearing type, and check if accuracy changes with operating conditions? I need a comprehensive view of where the model works best.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy", "analyze_engine_signals"], "procedure": "{\"problem\": \"Classify 6 fault categories across 5 bearing types in HUST dataset under varying speeds and loads.\", \"steps\": \"1) Load HUST training data (5 bearing types, 6 fault categories, varying conditions). 2) Load test data. 3) Load ground truth labels. 4) Analyze operating conditions using analyze_operating_conditions(dataset='HUST'). 5) Train fault classifier for 6 fault categories. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Calculate per-fault-category accuracy. 10) Calculate per-bearing-type accuracy. 11) Analyze if classification accuracy varies by operating condition (speed/load). 12) Output JSON report with classifications, overall accuracy, per-category accuracy, per-bearing-type accuracy, and condition-stratified accuracy.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.88, \"baseline_model\": \"CNN\", \"baseline_source\": \"HUST bearing fault benchmark\", \"sota_accuracy\": 0.95, \"sota_model\": \"Condition-aware Deep Network\", \"sota_source\": \"Multi-condition bearing fault diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for HUST bearing fault classification across 5 bearing types and 6 fault categories under varying operating conditions. Multi-condition classification is challenging. Deep learning models achieve 88-96% accuracy with condition-aware training.\", \"expected_fault_types\": [\"NORMAL\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on HUST dataset ground truth with 6 fault categories across 5 bearing types.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> analyze conditions -> train classifier -> classify -> verify -> overall accuracy -> per-category accuracy -> per-bearing-type accuracy -> condition-stratified accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_008", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "MFPT", "question": "The MFPT dataset has bearings with inner and outer race defects tested under different load conditions. Can you classify which type of defect each test bearing has, verify the classifications, and show me if accuracy changes with load? I need to know if we should account for load when classifying faults.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy", "analyze_engine_signals"], "procedure": "{\"problem\": \"Classify inner/outer race bearing defects in MFPT dataset under varying load conditions.\", \"steps\": \"1) Load MFPT training data (inner/outer race defects, varying loads). 2) Load test data. 3) Load ground truth labels. 4) Analyze load conditions using analyze_load_conditions(dataset='MFPT'). 5) Train fault classifier for inner vs. outer race defects. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Calculate per-fault-type accuracy (inner vs. outer race). 10) Analyze accuracy by load condition. 11) Output JSON report with classifications, overall accuracy, per-fault-type accuracy, and load-stratified accuracy.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.94, \"baseline_model\": \"CNN\", \"baseline_source\": \"MFPT bearing fault benchmark\", \"sota_accuracy\": 0.98, \"sota_model\": \"Load-aware CNN\", \"sota_source\": \"Variable-load bearing fault diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for MFPT bearing fault classification. MFPT provides well-controlled test conditions with inner/outer race defects under varying loads. CNN-based models achieve 95-99% accuracy. Load variation adds minor complexity.\", \"expected_fault_types\": [\"Fault_0\", \"Fault_1\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on MFPT dataset ground truth with inner/outer race defect labels.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> analyze load conditions -> train classifier -> classify -> verify -> overall accuracy -> per-fault-type accuracy -> load-stratified accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_009", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "Mendeley", "question": "The Mendeley dataset has bearings with changing speeds (600-1800 RPM), making the signals non-stationary. Can you classify the fault types for all test bearings, accounting for these changing conditions, and show me how accuracy varies across speed ranges? I need to know if non-stationary conditions hurt classification performance.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "{\"problem\": \"Classify bearing faults in Mendeley dataset under non-stationary conditions (speed 600-1800RPM).\", \"steps\": \"1) Load Mendeley training data (non-stationary). 2) Load test data. 3) Load ground truth labels. 4) Handle non-stationary classification using handle_non_stationary_classification(dataset='Mendeley', method='adaptive_feature_extraction'). 5) Train fault classifier adapted for non-stationary signals. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Analyze accuracy across different speed ranges. 10) Output JSON report with classifications, overall accuracy, and speed-range-stratified accuracy.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.87, \"baseline_model\": \"CNN\", \"baseline_source\": \"Non-stationary bearing fault classification\", \"sota_accuracy\": 0.94, \"sota_model\": \"Speed-Invariant CNN\", \"sota_source\": \"Adaptive feature extraction for variable-speed machinery\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for Mendeley bearing fault classification under non-stationary conditions (600-1800 RPM). Non-stationary signals require adaptive feature extraction. Speed-invariant deep learning achieves 87-95% accuracy on variable-speed datasets.\", \"expected_fault_types\": [\"Fault_0\", \"Fault_1\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on Mendeley dataset ground truth with non-stationary operating conditions.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load non-stationary data -> apply non-stationary handling -> train adapted classifier -> classify -> verify -> overall accuracy -> speed-range-stratified accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_010", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Gearbox", "source_dataset": "GearboxUoC", "question": "I have gearbox data from UoC and need to classify which components have faults. Can you classify the fault types for all test gearboxes, verify the classifications, and break down accuracy by gearbox component? I need to know which parts are hardest to diagnose.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy", "analyze_engine_signals"], "procedure": "{\"problem\": \"Classify gearbox faults in UoC dataset, analyzing classification performance across gearbox structural components.\", \"steps\": \"1) Load GearboxUoC training data. 2) Load test data. 3) Load ground truth fault labels. 4) Analyze gearbox structure using analyze_gearbox_structure(dataset='GearboxUoC'). 5) Train fault classifier. 6) Classify faults for all test units. 7) Verify classifications. 8) Calculate overall accuracy. 9) Analyze classification accuracy by gearbox component. 10) Output JSON report with classifications, overall accuracy, and component-stratified accuracy.\"}", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.9, \"baseline_model\": \"CNN\", \"baseline_source\": \"Gearbox fault diagnosis benchmarks\", \"sota_accuracy\": 0.96, \"sota_model\": \"Multi-scale CNN\", \"sota_source\": \"Structural component-aware gearbox fault diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for GearboxUoC fault classification. Gearbox fault diagnosis requires analyzing complex vibrational patterns from multiple components. CNN-based models achieve 90-97% accuracy on gearbox fault localization tasks.\", \"expected_fault_types\": [\"inner_race\", \"outer_race\", \"ball_race\"], \"expected_output_format\": \"json_report\", \"output_template\": {\"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\", \"overall_accuracy\": \"float\", \"per_fault_type_metrics\": {\"fault_type\": {\"f1_score\": \"float\", \"precision\": \"float\", \"recall\": \"float\"}}}, \"rationale\": \"Based on GearboxUoC dataset ground truth with component-specific fault labels.\", \"required_fields\": [\"classifications\", \"overall_accuracy\", \"confusion_matrix\", \"per_fault_type_metrics\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> analyze structure -> train classifier -> classify -> verify -> overall accuracy -> component-stratified accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_cost_benefit_001", "subset": "rotating_aero", "task_category": "Cost-Benefit Analysis", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD001", "question": "I am managing a fleet of aircraft engines and need to decide whether to do preventive maintenance or wait until failure. Can you predict remaining life for all test engines, calculate what it would cost to maintain them preventively versus letting them fail, and find the optimal maintenance threshold that minimizes total cost? I need concrete dollar amounts and a recommended RUL threshold for scheduling maintenance.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "calculate_maintenance_cost", "calculate_failure_cost", "optimize_maintenance_schedule"], "procedure": "{\"problem\": \"Perform cost-benefit analysis for CMAPSS FD001 engines comparing preventive maintenance vs. reactive maintenance strategies.\", \"steps\": \"1) Load CMAPSS FD001 test data. 2) Predict RUL for all test engines using predict_rul(dataset='CMAPSS_FD001'). 3) Calculate preventive maintenance cost using calculate_maintenance_cost(strategy='preventive', units=all_units, maintenance_cost_per_unit=5000). 4) Calculate reactive maintenance cost using calculate_failure_cost(strategy='reactive', units=all_units, failure_cost_per_unit=50000, predicted_failures=units_with_rul<20). 5) Optimize maintenance schedule using optimize_maintenance_schedule(predictions=predictions, maintenance_cost=5000, failure_cost=50000, threshold_rul=20). 6) Compare strategies using compare_strategies(preventive_cost=preventive_cost, reactive_cost=reactive_cost, optimized_cost=optimized_cost). 7) Identify cost-optimal maintenance threshold (RUL value that minimizes total cost). 8) Output JSON report with RUL predictions, preventive cost, reactive cost, optimized cost, recommended threshold, and cost savings from optimization.\"}", "ground_truth": "{\"expected_cost_range\": {\"optimized\": [3000, 40000], \"preventive\": [5000, 50000], \"reactive\": [50000, 500000]}, \"expected_mae_range\": [7, 18], \"expected_mae_range_rationale\": \"MAE is calculated across ALL 100 test units in CMAPSS_FD001, not just the top 10 high-risk units. Ground truth statistics - Mean RUL: 75.5, Range: 7-145. Expected MAE range represents typical model performance for this dataset.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [11, 26], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL 100 test units in CMAPSS_FD001, not just the top 10 high-risk units. Ground truth statistics - Mean RUL: 75.5, Range: 7-145. Expected RMSE range represents typical model performance for this dataset.\", \"expected_threshold_range\": [15, 30], \"output_template\": {\"cost_savings\": \"float\", \"optimized_cost\": \"float\", \"preventive_cost\": \"float\", \"reactive_cost\": \"float\", \"recommended_threshold\": \"float\", \"rul_predictions\": [{\"predicted_rul\": \"float\", \"unit_id\": \"int\"}]}, \"rationale\": \"Based on ground truth RUL values from RUL_FD001.txt, these 10 units have the lowest actual RUL values indicating highest failure risk.\", \"required_fields\": [\"rul_predictions\", \"preventive_cost\", \"reactive_cost\", \"optimized_cost\", \"recommended_threshold\", \"cost_savings\"], \"sample_correct_units\": [34, 31, 68, 81, 82, 42, 76, 35, 66, 56], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> predict RUL -> calculate preventive cost -> calculate reactive cost -> optimize schedule -> compare strategies -> identify optimal threshold.\"}", "known_issue": "Expected cost ranges depend on unit costs the question does not state; the MAE range is not meaningful for datasets without RUL labels."}
{"id": "pdm_cost_benefit_002", "subset": "rotating_aero", "task_category": "Cost-Benefit Analysis", "asset_class": "Bearing", "source_dataset": "CWRU", "question": "We have bearings with different fault types-inner race, outer race, and ball race defects-and each costs different amounts to fix. Can you classify the fault type for each test bearing, calculate maintenance and failure costs for each fault type, prioritize which bearings to fix first based on cost-benefit, and tell me which fault type gives us the best return on preventive maintenance investment? I need dollar amounts and a prioritized action list.", "question_operator_style": "", "required_tools": ["load_dataset", "classify_faults", "calculate_maintenance_cost", "optimize_maintenance_schedule", "calculate_failure_cost"], "procedure": "{\"problem\": \"Perform cost-benefit analysis for CWRU bearings comparing maintenance costs by fault type (inner race, outer race, ball race).\", \"steps\": \"1) Load CWRU test data. 2) Classify faults for all test bearings using classify_faults(dataset='CWRU'). 3) Calculate fault-specific maintenance costs using calculate_fault_specific_costs(fault_types=['inner_race', 'outer_race', 'ball_race'], costs=[3000, 4000, 5000]). 4) Calculate failure costs by fault type using calculate_failure_cost(fault_types=['inner_race', 'outer_race', 'ball_race'], failure_costs=[30000, 40000, 50000]). 5) Optimize maintenance priority using optimize_maintenance_priority(classifications=classifications, maintenance_costs=fault_costs, failure_costs=fault_failure_costs). 6) Compare fault costs using compare_fault_costs(fault_costs=fault_costs, failure_costs=fault_failure_costs). 7) Identify which fault type has highest cost-benefit ratio for preventive maintenance. 8) Output JSON report with fault classifications, per-fault-type costs, prioritized maintenance list, and cost-benefit recommendations.\"}", "ground_truth": "{\"expected_cost_range\": {\"optimized\": [3000, 40000], \"preventive\": [5000, 50000], \"reactive\": [50000, 500000]}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in CWRU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in CWRU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_threshold_range\": [15, 30], \"output_template\": {\"cost_savings\": \"float\", \"optimized_cost\": \"float\", \"preventive_cost\": \"float\", \"reactive_cost\": \"float\", \"recommended_threshold\": \"float\", \"rul_predictions\": [{\"predicted_rul\": \"float\", \"unit_id\": \"int\"}]}, \"rationale\": \"Based on CWRU dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"rul_predictions\", \"preventive_cost\", \"reactive_cost\", \"optimized_cost\", \"recommended_threshold\", \"cost_savings\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> classify faults -> calculate fault-specific costs -> calculate fault-specific failure costs -> optimize priority -> compare costs -> identify best ROI fault type.\"}", "known_issue": "Expected cost ranges depend on unit costs the question does not state; the MAE range is not meaningful for datasets without RUL labels."}
{"id": "pdm_cost_benefit_003", "subset": "rotating_aero", "task_category": "Cost-Benefit Analysis", "asset_class": "Bearing", "source_dataset": "FEMTO", "question": "The FEMTO dataset has bearings running under three different operational scenarios, and maintenance costs vary by scenario. Can you predict remaining life for all test bearings in each scenario, calculate maintenance and failure costs per scenario, optimize the maintenance strategy for each scenario separately, and tell me which scenario needs the most preventive maintenance? I need cost numbers and recommendations for each operational condition.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "analyze_engine_signals", "calculate_maintenance_cost", "optimize_maintenance_schedule"], "procedure": "{\"problem\": \"Perform cost-benefit analysis for FEMTO bearings across three operational scenarios, optimizing maintenance strategy per scenario.\", \"steps\": \"1) Load FEMTO test data. 2) Identify three operational scenarios using identify_operational_scenarios(dataset='FEMTO'). 3) Predict RUL for all test units per scenario. 4) Calculate scenario-specific maintenance costs using calculate_scenario_specific_costs(scenarios=scenarios, base_cost=4000, scenario_multipliers=[1.0, 1.2, 1.5]). 5) Calculate scenario-specific failure costs. 6) Optimize maintenance strategy per scenario using optimize_by_scenario(predictions=scenario_predictions, costs=scenario_costs). 7) Compare cost-benefit across scenarios. 8) Identify which scenario requires most aggressive preventive maintenance. 9) Output JSON report with per-scenario RUL predictions, costs, optimized strategies, and scenario comparison.\"}", "ground_truth": "{\"expected_cost_range\": {\"optimized\": [3000, 40000], \"preventive\": [5000, 50000], \"reactive\": [50000, 500000]}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in FEMTO, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in FEMTO, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_threshold_range\": [15, 30], \"output_template\": {\"cost_savings\": \"float\", \"optimized_cost\": \"float\", \"preventive_cost\": \"float\", \"reactive_cost\": \"float\", \"recommended_threshold\": \"float\", \"rul_predictions\": [{\"predicted_rul\": \"float\", \"unit_id\": \"int\"}]}, \"rationale\": \"Based on FEMTO dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"rul_predictions\", \"preventive_cost\", \"reactive_cost\", \"optimized_cost\", \"recommended_threshold\", \"cost_savings\"], \"sample_correct_units\": [0, 100, 200, 300, 400, 500, 600, 700, 800, 900], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> identify scenarios -> predict RUL per scenario -> calculate scenario costs -> optimize per scenario -> compare scenarios -> identify most critical scenario.\"}", "known_issue": "Expected cost ranges depend on unit costs the question does not state; the MAE range is not meaningful for datasets without RUL labels."}
{"id": "pdm_cost_benefit_004", "subset": "rotating_aero", "task_category": "Cost-Benefit Analysis", "asset_class": "Bearing", "source_dataset": "XJTU", "question": "For the XJTU bearings, I need a comprehensive maintenance plan that considers both how much life is left and what type of fault is present, since different faults cost different amounts to fix. Can you predict remaining life and classify fault types for all test bearings, calculate combined maintenance and failure costs, optimize a maintenance schedule that minimizes both cost and downtime, and give me a prioritized list of which bearings to fix first? I need dollar amounts and a clear action plan.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "classify_faults", "calculate_maintenance_cost", "optimize_maintenance_schedule"], "procedure": "{\"problem\": \"Perform integrated cost-benefit analysis for XJTU bearings combining RUL prediction and fault classification for comprehensive maintenance optimization.\", \"steps\": \"1) Load XJTU test data. 2) Predict RUL for all test units. 3) Classify fault types for all test units. 4) Calculate combined maintenance costs using calculate_combined_costs(rul_predictions=predictions, fault_classifications=classifications, base_maintenance_cost=3500, fault_cost_multipliers=[1.0, 1.3, 1.5, 1.8, 2.0]). 5) Calculate combined failure costs. 6) Optimize multi-objective maintenance schedule using optimize_multi_objective(rul_predictions=predictions, fault_classifications=classifications, costs=combined_costs, objectives=['minimize_cost', 'minimize_downtime']). 7) Generate prioritized maintenance list considering both RUL and fault type. 8) Output JSON report with RUL predictions, fault classifications, combined costs, optimized schedule, and prioritized maintenance list.\"}", "ground_truth": "{\"expected_cost_range\": {\"optimized\": [3000, 40000], \"preventive\": [5000, 50000], \"reactive\": [50000, 500000]}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in XJTU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in XJTU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_threshold_range\": [15, 30], \"output_template\": {\"cost_savings\": \"float\", \"optimized_cost\": \"float\", \"preventive_cost\": \"float\", \"reactive_cost\": \"float\", \"recommended_threshold\": \"float\", \"rul_predictions\": [{\"predicted_rul\": \"float\", \"unit_id\": \"int\"}]}, \"rationale\": \"Cost calculations based on standard maintenance cost models.\", \"required_fields\": [\"rul_predictions\", \"preventive_cost\", \"reactive_cost\", \"optimized_cost\", \"recommended_threshold\", \"cost_savings\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> predict RUL -> classify faults -> calculate combined costs -> optimize multi-objective -> generate prioritized list.\"}", "known_issue": "Expected cost ranges depend on unit costs the question does not state; the MAE range is not meaningful for datasets without RUL labels."}
{"id": "pdm_cost_benefit_005", "subset": "rotating_aero", "task_category": "Cost-Benefit Analysis", "asset_class": "Industrial engine", "source_dataset": "Azure", "question": "We have a huge fleet (876K+ units) and I am wondering if we can save money by doing maintenance in bulk batches instead of individually. Can you predict remaining life for all test units, calculate what it costs to maintain them individually versus in bulk batches, optimize a fleet-wide maintenance schedule that takes advantage of bulk discounts, and tell me what batch sizes give us the best cost savings? I need concrete dollar amounts and batch size recommendations.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "calculate_maintenance_cost", "optimize_maintenance_schedule"], "procedure": "{\"problem\": \"Perform fleet-wide cost-benefit analysis for Azure dataset (876K+ samples) optimizing bulk maintenance scheduling.\", \"steps\": \"1) Load Azure test data (large-scale). 2) Predict RUL for all test units. 3) Calculate bulk maintenance costs using calculate_bulk_maintenance_cost(units=all_units, base_cost=3000, bulk_discount_threshold=100, bulk_discount_rate=0.15). 4) Calculate individual vs. bulk failure costs. 5) Optimize fleet maintenance schedule using optimize_fleet_schedule(predictions=predictions, bulk_cost=bulk_cost, individual_cost=individual_cost, scheduling_window=30). 6) Analyze economies of scale using analyze_economies_of_scale(bulk_cost=bulk_cost, individual_cost=individual_cost, unit_count=unit_count). 7) Identify optimal batch sizes for maintenance. 8) Output JSON report with RUL predictions, bulk vs. individual costs, optimized schedule, economies of scale analysis, and recommended batch sizes.\"}", "ground_truth": "{\"expected_cost_range\": {\"optimized\": [3000, 40000], \"preventive\": [5000, 50000], \"reactive\": [50000, 500000]}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in Azure, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in Azure, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_threshold_range\": [15, 30], \"output_template\": {\"cost_savings\": \"float\", \"optimized_cost\": \"float\", \"preventive_cost\": \"float\", \"reactive_cost\": \"float\", \"recommended_threshold\": \"float\", \"rul_predictions\": [{\"predicted_rul\": \"float\", \"unit_id\": \"int\"}]}, \"rationale\": \"Cost calculations based on standard maintenance cost models.\", \"required_fields\": [\"rul_predictions\", \"preventive_cost\", \"reactive_cost\", \"optimized_cost\", \"recommended_threshold\", \"cost_savings\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load large dataset -> predict RUL -> calculate bulk costs -> calculate individual costs -> optimize fleet schedule -> analyze economies of scale -> identify optimal batch sizes.\"}", "known_issue": "Expected cost ranges depend on unit costs the question does not state; the MAE range is not meaningful for datasets without RUL labels."}
{"id": "pdm_safety_001", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD001", "question": "I am responsible for aircraft engine safety and need to identify which engines pose safety risks. Can you predict remaining life for all test engines, flag any with less than 20 cycles left as high risk, assess the safety risks (catastrophic failure, safety hazards, environmental impact), check if they comply with FAA and EASA regulations, and give me safety recommendations? I need risk classifications and actionable safety measures.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "{\"problem\": \"Assess safety risks and policy compliance for CMAPSS FD001 engines with low predicted RUL.\", \"steps\": \"1) Load CMAPSS FD001 test data. 2) Predict RUL for all test engines. 3) Identify engines with predicted RUL < 20 cycles (high risk). 4) Assess safety risk for high-risk engines using assess_safety_risk(units=high_risk_units, risk_categories=['catastrophic_failure', 'safety_hazard', 'environmental_impact']). 5) Check regulatory compliance using check_compliance(units=high_risk_units, regulations=['FAA', 'EASA', 'maintenance_intervals']). 6) Generate safety recommendations using generate_safety_recommendations(risk_assessment=risk_assessment, compliance_status=compliance_status). 7) Classify risk levels (low, medium, high, critical). 8) Output JSON report with RUL predictions, high-risk unit list, safety risk assessment, compliance status, risk classifications, and safety recommendations.\"}", "ground_truth": "{\"expected_mae_range\": [7, 18], \"expected_mae_range_rationale\": \"MAE is calculated across ALL 100 test units in CMAPSS_FD001, not just the top 10 high-risk units. Ground truth statistics - Mean RUL: 75.5, Range: 7-145. Expected MAE range represents typical model performance for this dataset.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [11, 26], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL 100 test units in CMAPSS_FD001, not just the top 10 high-risk units. Ground truth statistics - Mean RUL: 75.5, Range: 7-145. Expected RMSE range represents typical model performance for this dataset.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"risk_classifications\": {\"unit_id\": \"string\"}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Based on ground truth RUL values from RUL_FD001.txt, these 10 units have the lowest actual RUL values indicating highest failure risk.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\", \"risk_classifications\"], \"sample_correct_units\": [34, 31, 68, 81, 82, 42, 76, 35, 66, 56], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> predict RUL -> identify high-risk units -> assess safety risk -> check compliance -> generate recommendations -> classify risk levels.\"}", "known_issue": ""}
{"id": "pdm_safety_002", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Bearing", "source_dataset": "CWRU", "question": "Different bearing fault types have different safety implications. Can you classify the fault types for all test bearings, assess the safety impact of each fault type (vibration severity, failure probability, downtime risk), check compliance with ISO and ANSI vibration standards, and generate safety protocols for each fault type? I need to know which faults are most critical from a safety perspective.", "question_operator_style": "", "required_tools": ["load_dataset", "classify_faults", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "{\"problem\": \"Assess safety impact of different bearing fault types (inner race, outer race, ball race) and generate safety protocols.\", \"steps\": \"1) Load CWRU test data. 2) Classify faults for all test bearings. 3) Assess safety impact by fault type using assess_fault_safety_impact(fault_types=['inner_race', 'outer_race', 'ball_race'], safety_criteria=['vibration_severity', 'failure_probability', 'downtime_risk']). 4) Check industry standards compliance using check_industry_standards(fault_types=fault_types, standards=['ISO_10816', 'ANSI_S2.41']). 5) Generate safety protocols using generate_safety_protocols(fault_classifications=classifications, safety_impacts=impacts, standards=standards). 6) Rank fault types by safety criticality. 7) Output JSON report with fault classifications, per-fault-type safety impacts, compliance status, safety protocols, and criticality rankings.\"}", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in CWRU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in CWRU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"risk_classifications\": {\"unit_id\": \"string\"}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Based on CWRU dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\", \"risk_classifications\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> classify faults -> assess safety impact per fault type -> check standards -> generate protocols -> rank by criticality.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_safety_003", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Electric motor", "source_dataset": "ElectricMotorVibrations", "question": "Electric motor failures can be dangerous, especially with electrical hazards and fire risks. Can you classify the motor faults, assess safety risks (electrical hazards, fire risk, mechanical failure), check compliance with IEC and NEMA standards, and generate emergency shutdown protocols for high-risk motors? I need clear risk classifications and actionable shutdown procedures.", "question_operator_style": "", "required_tools": ["load_dataset", "classify_faults", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "{\"problem\": \"Assess safety risks for electric motor faults and generate emergency shutdown protocols.\", \"steps\": \"1) Load Electric Motor Vibrations test data. 2) Classify motor faults. 3) Assess motor safety risks using assess_motor_safety(faults=classifications, risk_factors=['electrical_hazard', 'fire_risk', 'mechanical_failure']). 4) Check electrical standards compliance using check_electrical_standards(faults=classifications, standards=['IEC_60034', 'NEMA_MG1']). 5) Generate emergency shutdown protocols using generate_shutdown_protocols(safety_assessment=assessment, standards=standards). 6) Classify motors by safety risk level (safe, caution, warning, critical). 7) Output JSON report with fault classifications, safety risk assessment, compliance status, shutdown protocols, and risk classifications.\"}", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in ElectricMotorVibrations, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in ElectricMotorVibrations, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"risk_classifications\": {\"unit_id\": \"string\"}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Safety assessment based on RUL predictions, regulatory compliance, and industry standards.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\", \"risk_classifications\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> classify faults -> assess motor safety -> check electrical standards -> generate shutdown protocols -> classify risk levels.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_safety_004", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Induction motor", "source_dataset": "RotorBrokenBar", "question": "Broken rotor bars in induction motors get more dangerous as severity increases. Can you classify the severity level (1-4 broken bars) for each test motor, assess safety risks for each severity level, check compliance with IEEE and IEC motor standards, and generate maintenance policies that specify when to shut down immediately versus reduce load versus schedule repair? I need clear thresholds for each severity level.", "question_operator_style": "", "required_tools": ["load_dataset", "classify_faults", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "{\"problem\": \"Assess safety risks for broken rotor bar faults with severity levels (1-4 bars) and generate maintenance policies.\", \"steps\": \"1) Load RotorBrokenBar test data. 2) Classify fault severity (1-4 broken bars). 3) Assess safety by severity level using assess_severity_safety(severity_levels=[1,2,3,4], safety_metrics=['motor_stability', 'overheating_risk', 'catastrophic_failure']). 4) Check motor standards compliance using check_motor_standards(severity_levels=severity_levels, standards=['IEEE_112', 'IEC_60034']). 5) Generate maintenance policies using generate_maintenance_policies(severity_assessment=assessment, standards=standards, policy_types=['immediate_shutdown', 'reduced_load', 'scheduled_repair']). 6) Define severity-based action thresholds. 7) Output JSON report with severity classifications, per-severity safety assessment, compliance status, maintenance policies, and action thresholds.\"}", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in RotorBrokenBar, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in RotorBrokenBar, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"risk_classifications\": {\"unit_id\": \"string\"}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Safety assessment based on RUL predictions, regulatory compliance, and industry standards.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\", \"risk_classifications\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> classify severity -> assess safety per severity -> check standards -> generate policies -> define action thresholds.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_safety_005", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Gearbox", "source_dataset": "PlanetaryPdM", "question": "Planetary gearboxes have complex structures, and faults in different components have different safety implications. Can you classify the faults and identify which component (sun gear, planet, ring gear) is affected, assess safety risks for each component type, check compliance with AGMA and ISO gear standards, and generate component-specific safety policies? I need to know which components are most critical for safety.", "question_operator_style": "", "required_tools": ["load_dataset", "classify_faults", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "{\"problem\": \"Assess safety risks for planetary gearbox component-specific faults and generate component-targeted zsafety policies.\", \"steps\": \"1) Load PlanetaryPdM test data. 2) Classify faults and localize components (sun gear, planet gear, ring gear). 3) Assess gearbox safety by component using assess_gearbox_safety(components=['sun_gear', 'planet_gear', 'ring_gear'], safety_factors=['load_capacity', 'failure_cascade', 'operational_safety']). 4) Check mechanical standards compliance using check_mechanical_standards(components=components, standards=['AGMA_2001', 'ISO_6336']). 5) Generate component-specific safety policies using generate_component_specific_policies(component_faults=localizations, safety_assessment=assessment, standards=standards). 6) Rank components by safety criticality. 7) Output JSON report with fault classifications, component localizations, per-component safety assessment, compliance status, component-specific policies, and criticality rankings.\"}", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in PlanetaryPdM, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in PlanetaryPdM, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"risk_classifications\": {\"unit_id\": \"string\"}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Safety assessment based on RUL predictions, regulatory compliance, and industry standards.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\", \"risk_classifications\"], \"verification_required\": true}", "metadata": "{\"dependency_analysis\": \"Load data -> classify and localize faults -> assess safety per component -> check standards -> generate component policies -> rank by criticality.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_rul_011", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD001", "question": "Predict remaining useful life for IMS equipment and identify the 5 units most at risk.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "Predict RUL for IMS dataset and identify top 5 risk units.", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 17.6, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_mae\": 10.49, \"sota_model\": \"CAELSTM\", \"sota_rmse\": 14.44, \"sota_source\": \"Nature Scientific Reports 2025 - Deep learning-based prognostic approach\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD001 benchmarks. CAELSTM (Conv AE + Attention LSTM) achieves MAE=10.49 cycles (Nature Scientific Reports 2025). BiLSTM baseline achieves MAE~12-15 cycles. Range encompasses SOTA to baseline performance.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD001 benchmarks. CAELSTM achieves RMSE=14.44 cycles (Nature Scientific Reports 2025). BiLSTM baseline achieves RMSE=17.60 cycles (arXiv 2025). Self-attention+BiGRU+BiLSTM achieves RMSE=11.74 (ScienceDirect 2024).\", \"mae_range\": [12.5, 25.1], \"output_template\": {\"mae\": \"object\", \"predictions\": \"object\", \"rmse\": \"object\", \"top_5_risk_units\": \"object\"}, \"rationale\": \"Based on CMAPSS_FD001 dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values.\", \"required_fields\": [\"predictions\", \"top_5_risk_units\", \"mae\", \"rmse\"], \"rmse_range\": [16.7, 33.4], \"sample_correct_units\": [33, 30, 67, 80, 81, 41, 75, 34, 65, 55], \"verification_required\": true}", "metadata": "{\"complexity\": \"Medium\", \"dependency_analysis\": \"Load data -> train model -> predict -> verify -> rank by risk.\"}", "known_issue": "The question text names IMS, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_012", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD002", "question": "Predict remaining useful life for HUST equipment and identify the 5 units most at risk.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "Predict RUL for HUST dataset and identify top 5 risk units.", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 29.67, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_model\": \"Multi-attention Mechanism\", \"sota_rmse\": 23.5, \"sota_source\": \"ScienceDirect 2024\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD002 benchmarks. BiLSTM baseline achieves RMSE=29.67 cycles (arXiv 2025). FD002 has 6 operating conditions, making prediction more complex than FD001. Range accounts for multi-condition complexity.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD002 benchmarks. BiLSTM achieves RMSE=29.67 cycles (arXiv 2025 - Uncertainty-Aware DL Framework). Multi-attention models improve to RMSE~22-25 cycles. Range reflects increased difficulty from 6 operating conditions.\", \"mae_range\": [16.2, 32.3], \"output_template\": {\"mae\": \"object\", \"predictions\": \"object\", \"rmse\": \"object\", \"top_5_risk_units\": \"object\"}, \"rationale\": \"Based on CMAPSS_FD002 dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values.\", \"required_fields\": [\"predictions\", \"top_5_risk_units\", \"mae\", \"rmse\"], \"rmse_range\": [21.6, 43.1], \"sample_correct_units\": [6, 10, 43, 176, 210, 115, 123, 137, 89, 101], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train model -> predict -> verify -> rank by risk.\"}", "known_issue": "The question text names HUST, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_013", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD003", "question": "Predict remaining useful life for MFPT equipment and identify the 5 units most at risk.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "Predict RUL for MFPT dataset and identify top 5 risk units.", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 17.62, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_model\": \"Self-attention+BiGRU+BiLSTM\", \"sota_rmse\": 13.5, \"sota_source\": \"ScienceDirect 2024 - Multi-attention Mechanism\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD003 benchmarks. BiLSTM baseline achieves RMSE=17.62 cycles (arXiv 2025). FD003 has 1 operating condition but 2 fault modes. Range accounts for dual fault mode complexity.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD003 benchmarks. BiLSTM achieves RMSE=17.62 cycles (arXiv 2025 - Uncertainty-Aware DL Framework). Self-attention+BiGRU+BiLSTM models achieve improved RMSE~13-15 cycles.\", \"mae_range\": [12.5, 25.0], \"output_template\": {\"mae\": \"object\", \"predictions\": \"object\", \"rmse\": \"object\", \"top_5_risk_units\": \"object\"}, \"rationale\": \"Based on CMAPSS_FD003 dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values.\", \"required_fields\": [\"predictions\", \"top_5_risk_units\", \"mae\", \"rmse\"], \"rmse_range\": [16.6, 33.3], \"sample_correct_units\": [81, 45, 38, 98, 23, 93, 25, 49, 91, 80], \"verification_required\": true}", "metadata": "{\"complexity\": \"Medium\", \"dependency_analysis\": \"Load data -> train model -> predict -> verify -> rank by risk.\"}", "known_issue": "The question text names MFPT, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_014", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD004", "question": "Predict remaining useful life for Paderborn equipment and identify the 5 units most at risk.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "Predict RUL for Paderborn dataset and identify top 5 risk units.", "ground_truth": "{\"algorithm_reference\": {\"baseline_model\": \"BiLSTM\", \"baseline_rmse\": 31.84, \"baseline_source\": \"arXiv 2025 - Uncertainty-Aware Deep Learning Framework\", \"sota_model\": \"Multi-attention Mechanism\", \"sota_rmse\": 26.5, \"sota_source\": \"ScienceDirect 2024\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published CMAPSS_FD004 benchmarks. BiLSTM baseline achieves RMSE=31.84 cycles (arXiv 2025). FD004 is the most complex subset with 6 operating conditions and 2 fault modes.\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published CMAPSS_FD004 benchmarks. BiLSTM achieves RMSE=31.84 cycles (arXiv 2025 - Uncertainty-Aware DL Framework). Multi-attention models achieve RMSE~25-28 cycles. Range accounts for high dataset complexity.\", \"mae_range\": [16.4, 32.8], \"output_template\": {\"mae\": \"object\", \"predictions\": \"object\", \"rmse\": \"object\", \"top_5_risk_units\": \"object\"}, \"rationale\": \"Based on CMAPSS_FD004 dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values.\", \"required_fields\": [\"predictions\", \"top_5_risk_units\", \"mae\", \"rmse\"], \"rmse_range\": [21.9, 43.7], \"sample_correct_units\": [30, 157, 11, 99, 134, 154, 57, 117, 172, 102], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train model -> predict -> verify -> rank by risk.\"}", "known_issue": "The question text names Paderborn, but the ground truth is computed on the C-MAPSS subset in the `source_dataset` field. The field is correct; the question text is not. Fix planned for v1.1."}
{"id": "pdm_rul_015", "subset": "rotating_aero", "task_category": "RUL Prediction", "asset_class": "Bearing", "source_dataset": "FEMTO", "question": "Predict remaining useful life for Mendeley equipment and identify the 5 units most at risk.", "question_operator_style": "", "required_tools": ["load_dataset", "load_ground_truth", "train_rul_model", "predict_rul", "verify_ground_truth", "calculate_mae", "calculate_rmse"], "procedure": "Predict RUL for Mendeley dataset and identify top 5 risk units.", "ground_truth": "{\"algorithm_reference\": {\"baseline_mae\": 0.056, \"baseline_model\": \"TCNMABG\", \"baseline_rmse\": 0.082, \"baseline_source\": \"MDPI 2024 - Temporal Convolutional Multi-Attention\", \"sota_model\": \"TcLstmNet-CBAM\", \"sota_performance\": \"Best MAE on 11/14 bearings\", \"sota_source\": \"Nature Scientific Reports 2025\"}, \"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"Range [10, 25] derived from published FEMTO/PRONOSTIA benchmarks. TCNMABG achieves MAE=0.056 (normalized percentage of lifetime, MDPI 2024). TcLstmNet-CBAM achieved best MAE on 11/14 bearings (Nature Scientific Reports 2025).\", \"expected_output_format\": \"json_report\", \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"Range [15, 30] derived from published FEMTO/PRONOSTIA benchmarks. CNN-GRU-MHA with transfer learning achieves RMSE=0.0443 (MDPI 2024). TCNMABG achieves RMSE=0.082 (MDPI). Note: FEMTO metrics often in percentage of lifetime.\", \"mae_range\": [8.7, 17.4], \"output_template\": {\"mae\": \"object\", \"predictions\": \"object\", \"rmse\": \"object\", \"top_5_risk_units\": \"object\"}, \"rationale\": \"Based on FEMTO dataset ground truth, these bearing indices represent the highest failure risk based on actual RUL values.\", \"required_fields\": [\"predictions\", \"top_5_risk_units\", \"mae\", \"rmse\"], \"rmse_range\": [11.6, 23.3], \"sample_correct_units\": [0, 200, 400, 300, 600, 800, 900, 100, 700, 1200], \"verification_required\": true}", "metadata": "{\"complexity\": \"Medium\", \"dependency_analysis\": \"Load data -> train model -> predict -> verify -> rank by risk.\"}", "known_issue": "FEMTO: the expected MAE/RMSE ranges do not state their unit (rul_percentage is 0-1); a constant 50% prediction lands just outside the MAE range."}
{"id": "pdm_fault_011", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "Paderborn", "question": "Classify fault types for all test units in Paderborn dataset.", "question_operator_style": "", "required_tools": ["load_dataset", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "Classify faults in Paderborn dataset.", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.92, \"baseline_model\": \"CNN\", \"baseline_source\": \"Paderborn bearing fault benchmark\", \"sota_accuracy\": 0.99, \"sota_model\": \"FaultNet\", \"sota_source\": \"arXiv - Deep Learning for Bearing Fault Diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] derived from published Paderborn bearing fault classification benchmarks. FaultNet achieves ~99% accuracy (arXiv). High-complexity dataset with multiple sensor types. Deep learning achieves 92-99% accuracy.\", \"expected_output_format\": \"json_report\", \"output_template\": {\"accuracy\": \"object\", \"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\"}, \"rationale\": \"Based on Paderborn dataset ground truth fault labels.\", \"required_fields\": [\"classifications\", \"accuracy\", \"confusion_matrix\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train classifier -> classify -> verify -> calculate accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_012", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "HUST", "question": "Classify fault types for all test units in HUST dataset.", "question_operator_style": "", "required_tools": ["load_dataset", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "Classify faults in HUST dataset.", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.88, \"baseline_model\": \"CNN\", \"baseline_source\": \"HUST bearing fault benchmark\", \"sota_accuracy\": 0.95, \"sota_model\": \"Condition-aware Deep Network\", \"sota_source\": \"Multi-condition bearing fault diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for HUST bearing fault classification across multiple bearing types and fault categories. Deep learning models achieve 88-96% accuracy with condition-aware training.\", \"expected_output_format\": \"json_report\", \"output_template\": {\"accuracy\": \"object\", \"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\"}, \"rationale\": \"Based on HUST dataset ground truth fault labels.\", \"required_fields\": [\"classifications\", \"accuracy\", \"confusion_matrix\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train classifier -> classify -> verify -> calculate accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_013", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "MFPT", "question": "Classify fault types for all test units in MFPT dataset.", "question_operator_style": "", "required_tools": ["load_dataset", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "Classify faults in MFPT dataset.", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.94, \"baseline_model\": \"CNN\", \"baseline_source\": \"MFPT bearing fault benchmark\", \"sota_accuracy\": 0.98, \"sota_model\": \"Load-aware CNN\", \"sota_source\": \"Variable-load bearing fault diagnosis\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for MFPT bearing fault classification. Well-controlled dataset with inner/outer race defects. CNN-based models achieve 95-99% accuracy on this benchmark.\", \"expected_output_format\": \"json_report\", \"output_template\": {\"accuracy\": \"object\", \"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\"}, \"rationale\": \"Based on MFPT dataset ground truth fault labels.\", \"required_fields\": [\"classifications\", \"accuracy\", \"confusion_matrix\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train classifier -> classify -> verify -> calculate accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_014", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Bearing", "source_dataset": "Mendeley", "question": "Classify fault types for all test units in Mendeley dataset.", "question_operator_style": "", "required_tools": ["load_dataset", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "Classify faults in Mendeley dataset.", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.87, \"baseline_model\": \"CNN\", \"baseline_source\": \"Non-stationary bearing fault classification\", \"sota_accuracy\": 0.94, \"sota_model\": \"Speed-Invariant CNN\", \"sota_source\": \"Adaptive feature extraction for variable-speed machinery\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for Mendeley bearing fault classification under non-stationary conditions. Speed-invariant deep learning achieves 87-95% accuracy on variable-speed datasets.\", \"expected_output_format\": \"json_report\", \"output_template\": {\"accuracy\": \"object\", \"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\"}, \"rationale\": \"Based on Mendeley dataset ground truth fault labels.\", \"required_fields\": [\"classifications\", \"accuracy\", \"confusion_matrix\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train classifier -> classify -> verify -> calculate accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_fault_015", "subset": "rotating_aero", "task_category": "Fault Classification", "asset_class": "Electric motor", "source_dataset": "ElectricMotorVibrations", "question": "Classify fault types for all test units in ElectricMotorVibrations dataset.", "question_operator_style": "", "required_tools": ["load_dataset", "train_fault_classifier", "classify_faults", "verify_classification", "calculate_accuracy"], "procedure": "Classify faults in ElectricMotorVibrations dataset.", "ground_truth": "{\"algorithm_reference\": {\"baseline_accuracy\": 0.9, \"baseline_model\": \"Single-modal CNN\", \"baseline_source\": \"Electric motor fault diagnosis benchmarks\", \"sota_accuracy\": 0.97, \"sota_model\": \"Multimodal Fusion CNN\", \"sota_source\": \"Acoustic-vibration fusion for motor diagnostics\"}, \"expected_accuracy_range\": [0.85, 0.98], \"expected_accuracy_range_rationale\": \"Range [0.85, 0.98] for electric motor fault classification. Multimodal fusion of acoustic and vibration data improves accuracy. CNN-based classifiers achieve 95-99% on motor fault datasets.\", \"expected_output_format\": \"json_report\", \"output_template\": {\"accuracy\": \"object\", \"classifications\": [{\"fault_type\": \"string\", \"unit_id\": \"int\"}], \"confusion_matrix\": \"object (2D array)\"}, \"rationale\": \"Based on ElectricMotorVibrations dataset ground truth fault labels.\", \"required_fields\": [\"classifications\", \"accuracy\", \"confusion_matrix\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> train classifier -> classify -> verify -> calculate accuracy.\"}", "known_issue": "expected_fault_types holds a placeholder (\"Fault_0\") left by an earlier tool stub; the accuracy range [0.85, 0.98] is generic."}
{"id": "pdm_safety_006", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Turbofan engine", "source_dataset": "CMAPSS_FD001", "question": "Identify safety risks and compliance issues for CMAPSS_FD001 equipment.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "Assess safety risks for CMAPSS_FD001 equipment with low predicted RUL.", "ground_truth": "{\"expected_mae_range\": [7, 18], \"expected_mae_range_rationale\": \"MAE is calculated across ALL 100 test units in CMAPSS_FD001, not just the top 10 high-risk units. Ground truth statistics - Mean RUL: 75.5, Range: 7-145. Expected MAE range represents typical model performance for this dataset.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [11, 26], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL 100 test units in CMAPSS_FD001, not just the top 10 high-risk units. Ground truth statistics - Mean RUL: 75.5, Range: 7-145. Expected RMSE range represents typical model performance for this dataset.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Based on ground truth RUL values from RUL_FD001.txt, these 10 units have the lowest actual RUL values indicating highest failure risk.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\"], \"sample_correct_units\": [34, 31, 68, 81, 82, 42, 76, 35, 66, 56], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> predict RUL -> assess risk -> check compliance -> generate recommendations.\"}", "known_issue": ""}
{"id": "pdm_safety_007", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Bearing", "source_dataset": "CWRU", "question": "Identify safety risks and compliance issues for CWRU equipment.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "Assess safety risks for CWRU equipment with low predicted RUL.", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in CWRU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in CWRU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Based on CWRU dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\"], \"sample_correct_units\": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> predict RUL -> assess risk -> check compliance -> generate recommendations.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_safety_008", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Bearing", "source_dataset": "FEMTO", "question": "Identify safety risks and compliance issues for FEMTO equipment.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "Assess safety risks for FEMTO equipment with low predicted RUL.", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in FEMTO, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in FEMTO, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Based on FEMTO dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\"], \"sample_correct_units\": [0, 100, 200, 300, 400, 500, 600, 700, 800, 900], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> predict RUL -> assess risk -> check compliance -> generate recommendations.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_safety_009", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Bearing", "source_dataset": "XJTU", "question": "Identify safety risks and compliance issues for XJTU equipment.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "Assess safety risks for XJTU equipment with low predicted RUL.", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in XJTU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in XJTU, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"safety_recommendations\": [\"string\"]}, \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\"], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> predict RUL -> assess risk -> check compliance -> generate recommendations.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "pdm_safety_010", "subset": "rotating_aero", "task_category": "Safety/Policy Evaluation", "asset_class": "Bearing", "source_dataset": "IMS", "question": "Identify safety risks and compliance issues for IMS equipment.", "question_operator_style": "", "required_tools": ["load_dataset", "predict_rul", "assess_safety_risk", "check_compliance", "generate_safety_recommendations"], "procedure": "Assess safety risks for IMS equipment with low predicted RUL.", "ground_truth": "{\"expected_mae_range\": [10, 25], \"expected_mae_range_rationale\": \"MAE is calculated across ALL test units in IMS, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"expected_output_format\": \"json_report\", \"expected_risk_levels\": [\"low\", \"medium\", \"high\", \"critical\"], \"expected_rmse_range\": [15, 30], \"expected_rmse_range_rationale\": \"RMSE is calculated across ALL test units in IMS, not just the top 10 high-risk units. Range represents typical model performance for this dataset based on historical benchmarks.\", \"output_template\": {\"compliance_status\": {\"unit_id\": {\"compliant\": \"bool\", \"regulation\": \"string\"}}, \"risk_assessment\": {\"unit_id\": {\"risk_factors\": [\"string\"], \"risk_level\": \"string\"}}, \"safety_recommendations\": [\"string\"]}, \"rationale\": \"Based on IMS dataset ground truth, these 10 units represent the highest failure risk based on actual RUL values or fault severity.\", \"required_fields\": [\"risk_assessment\", \"compliance_status\", \"safety_recommendations\"], \"sample_correct_units\": [0, 1, 3, 4, 5, 6, 7, 101, 102, 103], \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"dependency_analysis\": \"Load data -> predict RUL -> assess risk -> check compliance -> generate recommendations.\"}", "known_issue": "Ground truth specifies only the allowed risk levels (and placeholder sample units); no unit-level answer can be checked."}
{"id": "engine_health_combined_001", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "I need a complete health assessment for this aircraft engine based on the following sensor data from a single operational cycle. Analyze all major components and give me: (1) Health status for each component (Fan, LPC, HPC, HPT, LPT) - Healthy or Fault, (2) If there are timing issues, identify whether they're caused by efficiency or flow problems and in which component, (3) The overall health state of the engine.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Healthy\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [], \"overall_health_state\": \"Operating Normally\", \"rationale\": \"Ground truth from EngineMTQA engine 108410. Health state: Operating Normally. Faulty components: None. Timing issue: b, e.\", \"timing_issue_root_cause\": \"LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"108410\"}", "known_issue": "Ground truth lists no faulty components, while its own root-cause note cites LPT efficiency and flow modifiers; pending SME adjudication."}
{"id": "engine_health_combined_002", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Our maintenance team needs an urgent diagnostic report for this engine. Review the sensor readings below and determine: which components are functioning normally, which have faults, what's causing any timing anomalies, and the engine's overall operational status.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Healthy\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [], \"overall_health_state\": \"Operating Normally\", \"rationale\": \"Ground truth from EngineMTQA engine 20119. Health state: Operating Normally. Faulty components: None. Timing issue: a, b, e.\", \"timing_issue_root_cause\": \"HPT efficiency modifier and LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"20119\"}", "known_issue": ""}
{"id": "engine_health_combined_003", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Before this aircraft's next flight, I need you to analyze the engine telemetry and provide a full component-by-component health assessment. Check Fan, LPC, HPC, HPT, and LPT systems, identify any efficiency or flow issues, and classify the overall engine condition.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Healthy\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [], \"overall_health_state\": \"Operating Normally\", \"rationale\": \"Ground truth from EngineMTQA engine 8389. Health state: Operating Normally. Faulty components: None. Timing issue: a, b, e.\", \"timing_issue_root_cause\": \"HPT efficiency modifier and LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"8389\"}", "known_issue": ""}
{"id": "engine_health_combined_004", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine flagged an alert during the last flight. Analyze the sensor data and tell me: (1) Which specific components show signs of malfunction? (2) Are the timing problems related to efficiency degradation or flow restrictions? (3) Should we ground this aircraft for maintenance?\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: elevated and rising: 528K → 542K (+2.6% over 10 cycles, ANOMALY)\n  - Fan Inlet Pressure: declining: 14.62 psia → 13.89 psia (-5.0%, CRITICAL)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: abnormally high: 565K → 589K (+4.2%, WARNING)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: sharp increase: 1398K → 1456K (+4.1%, CRITICAL)\n  - HPC Outlet Pressure: unstable: fluctuating 520-535 psia (7% variance)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1089K with oscillations (FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Fault\", \"HPC\": \"Fault\", \"HPT\": \"Fault\", \"LPC\": \"Fault\", \"LPT\": \"Fault\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"Fan\", \"LPC\", \"HPC\", \"HPT\", \"LPT\"], \"overall_health_state\": \"Abnormal Degradation\", \"rationale\": \"Ground truth from EngineMTQA engine 60739. Health state: Abnormal Degradation. Faulty components: ['Fan', 'LPC', 'HPC', 'HPT', 'LPT']. Timing issue: a, b, c, d, e, f.\", \"timing_issue_root_cause\": \"Multiple efficiency and flow modifiers affected across components\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"60739\"}", "known_issue": ""}
{"id": "engine_health_combined_005", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "I need a complete health assessment for this aircraft engine based on the following sensor data from a single operational cycle. Analyze all major components and give me: (1) Health status for each component (Fan, LPC, HPC, HPT, LPT) - Healthy or Fault, (2) If there are timing issues, identify whether they're caused by efficiency or flow problems and in which component, (3) The overall health state of the engine.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"HPT\"], \"overall_health_state\": \"Abnormal Degradation\", \"rationale\": \"Ground truth from EngineMTQA engine 78384. Health state: Abnormal Degradation. Faulty components: ['HPT']. Timing issue: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"78384\"}", "known_issue": ""}
{"id": "engine_health_combined_006", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Our maintenance team needs an urgent diagnostic report for this engine. Review the sensor readings below and determine: which components are functioning normally, which have faults, what's causing any timing anomalies, and the engine's overall operational status.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: abnormally high: 565K → 589K (+4.2%, WARNING)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: sharp increase: 1398K → 1456K (+4.1%, CRITICAL)\n  - HPC Outlet Pressure: unstable: fluctuating 520-535 psia (7% variance)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: gradual decline: 1588K → 1571K (-1.1%)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Fault\", \"HPT\": \"Healthy\", \"LPC\": \"Fault\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"LPC\", \"HPC\"], \"overall_health_state\": \"Abnormal Degradation\", \"rationale\": \"Ground truth from EngineMTQA engine 46790. Health state: Abnormal Degradation. Faulty components: ['LPC', 'HPC']. Timing issue: .\", \"timing_issue_root_cause\": \"Unable to determine specific timing issue from available sensor data\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"46790\"}", "known_issue": ""}
{"id": "engine_health_combined_007", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Before this aircraft's next flight, I need you to analyze the engine telemetry and provide a full component-by-component health assessment. Check Fan, LPC, HPC, HPT, and LPT systems, identify any efficiency or flow issues, and classify the overall engine condition.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1089K with oscillations (FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Fault\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"HPT\", \"LPT\"], \"overall_health_state\": \"Abnormal Degradation\", \"rationale\": \"Ground truth from EngineMTQA engine 24953. Health state: Abnormal Degradation. Faulty components: ['HPT', 'LPT']. Timing issue: a, b, e.\", \"timing_issue_root_cause\": \"HPT efficiency modifier and LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"24953\"}", "known_issue": ""}
{"id": "engine_health_combined_008", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine flagged an alert during the last flight. Analyze the sensor data and tell me: (1) Which specific components show signs of malfunction? (2) Are the timing problems related to efficiency degradation or flow restrictions? (3) Should we ground this aircraft for maintenance?\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1089K with oscillations (FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Fault\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"HPT\", \"LPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Ground truth from EngineMTQA engine 20495. Health state: Imminent Failure. Faulty components: ['HPT', 'LPT']. Timing issue: a, b, e.\", \"timing_issue_root_cause\": \"HPT efficiency modifier and LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"20495\"}", "known_issue": "Question text is identical to engine_health_combined_012; all checkable ground-truth fields agree (benign duplicate)."}
{"id": "engine_health_combined_009", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "I need a complete health assessment for this aircraft engine based on the following sensor data from a single operational cycle. Analyze all major components and give me: (1) Health status for each component (Fan, LPC, HPC, HPT, LPT) - Healthy or Fault, (2) If there are timing issues, identify whether they're caused by efficiency or flow problems and in which component, (3) The overall health state of the engine.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"HPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Ground truth from EngineMTQA engine 77593. Health state: Imminent Failure. Faulty components: ['HPT']. Timing issue: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"77593\"}", "known_issue": ""}
{"id": "engine_health_combined_010", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Our maintenance team needs an urgent diagnostic report for this engine. Review the sensor readings below and determine: which components are functioning normally, which have faults, what's causing any timing anomalies, and the engine's overall operational status.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: abnormally high: 565K → 589K (+4.2%, WARNING)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: declining rapidly: 1582K → 1534K (-3.0%)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Healthy\", \"LPC\": \"Fault\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"LPC\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Ground truth from EngineMTQA engine 107977. Health state: Imminent Failure. Faulty components: ['LPC']. Timing issue: b, e.\", \"timing_issue_root_cause\": \"LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"107977\"}", "known_issue": ""}
{"id": "engine_health_combined_011", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Before this aircraft's next flight, I need you to analyze the engine telemetry and provide a full component-by-component health assessment. Check Fan, LPC, HPC, HPT, and LPT systems, identify any efficiency or flow issues, and classify the overall engine condition.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"HPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Ground truth from EngineMTQA engine 3966. Health state: Imminent Failure. Faulty components: ['HPT']. Timing issue: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"3966\"}", "known_issue": ""}
{"id": "engine_health_combined_012", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine flagged an alert during the last flight. Analyze the sensor data and tell me: (1) Which specific components show signs of malfunction? (2) Are the timing problems related to efficiency degradation or flow restrictions? (3) Should we ground this aircraft for maintenance?\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1089K with oscillations (FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "assess_component_health", "diagnose_timing_issues"], "procedure": "1. Parse embedded sensor data summary. 2. Analyze each component's sensor readings for anomalies. 3. Determine health status (Healthy/Fault) for Fan, LPC, HPC, HPT, LPT. 4. Identify root cause of timing issues (efficiency vs flow modifier). 5. Classify overall health state (Operating Normally/Abnormal Degradation/Imminent Failure).", "ground_truth": "{\"component_health\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Fault\"}, \"expected_answer_type\": \"structured_health_assessment\", \"faulty_components\": [\"HPT\", \"LPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Ground truth from EngineMTQA engine 9788. Health state: Imminent Failure. Faulty components: ['HPT', 'LPT']. Timing issue: a, b, e.\", \"timing_issue_root_cause\": \"HPT efficiency modifier and LPT efficiency modifier and LPT flow modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"comprehensive_component_assessment\", \"source_engine_id\": \"9788\"}", "known_issue": "Question text is identical to engine_health_combined_008; all checkable ground-truth fields agree (benign duplicate)."}
{"id": "engine_health_combined_013", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "I've collected engine telemetry across 10 operational cycles and need a trend analysis. Based on this data: (1) What is the qualitative condition of the engine (Good/Moderate/Poor/Bad/Extremely Poor)? (2) What's the estimated probability range of engine failure? (3) What maintenance actions should we take?\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"10%-30%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Good Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Good Condition, Failure probability: 10%-30%.\", \"recommended_actions\": \"It is imperative to conduct preventive maintenance and enhance the monitoring of key operational parameters to mitigate potential risks and maintain system reliability.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"106656\", \"106678\", \"106703\", \"106716\", \"106741\", \"106779\", \"106797\", \"106828\", \"106851\", \"106874\"]}", "known_issue": "Question text is identical to engine_health_combined_017, but the two ground truths disagree on qualitative_condition."}
{"id": "engine_health_combined_014", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Our predictive maintenance system flagged this engine for review. Analyze the 10-cycle trend data and provide: overall condition assessment, failure probability estimate, and your recommended maintenance strategy.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"10%-30%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Good Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Good Condition, Failure probability: 10%-30%.\", \"recommended_actions\": \"We should definitely go ahead with some preventive maintenance to keep things running smoothly, and we also need to step up our monitoring of the key parameters to catch any potential issues early on.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"10240\", \"10254\", \"10272\", \"10287\", \"10306\", \"10333\", \"10357\", \"10373\", \"10398\", \"10422\"]}", "known_issue": "Same signal summary as engine_health_combined_013/017, which carry different ground truths (Good vs Moderate Condition)."}
{"id": "engine_health_combined_015", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine is due for its scheduled inspection. Based on the multi-cycle signal data, assess whether the condition has degraded, estimate failure risk, and recommend whether to proceed with routine maintenance or escalate to urgent repairs.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"10%-30%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Moderate Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Moderate Condition, Failure probability: 10%-30%.\", \"recommended_actions\": \"We need to keep a close eye on the engine’s condition and make sure to schedule maintenance pretty soon to stay on top of things.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"39611\", \"39628\", \"39665\", \"39699\", \"39715\", \"39735\", \"39768\", \"39789\", \"39821\", \"39848\"]}", "known_issue": "Same signal summary as engine_health_combined_013/017, which carry different ground truths (Good vs Moderate Condition)."}
{"id": "engine_health_combined_016", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Fleet management needs a health trend report for this engine unit. Analyze the 10-cycle data and provide: (1) Current condition classification, (2) Failure probability range for the next 50 cycles, (3) Specific maintenance recommendations.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"10%-30%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Moderate Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Moderate Condition, Failure probability: 10%-30%.\", \"recommended_actions\": \"It might be advisable to closely monitor the engine's condition and consider scheduling maintenance in the near future to ensure optimal performance and prevent potential issues.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"39612\", \"39628\", \"39665\", \"39700\", \"39716\", \"39734\", \"39768\", \"39790\", \"39820\", \"39848\"]}", "known_issue": "Same signal summary as engine_health_combined_013/017, which carry different ground truths (Good vs Moderate Condition)."}
{"id": "engine_health_combined_017", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "I've collected engine telemetry across 10 operational cycles and need a trend analysis. Based on this data: (1) What is the qualitative condition of the engine (Good/Moderate/Poor/Bad/Extremely Poor)? (2) What's the estimated probability range of engine failure? (3) What maintenance actions should we take?\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft throughout all cycles\n  - Throttle Resolver Angle: steady decrease from 84.2° to 78.6° (normal throttle adjustment)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: stable at 552-555K\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: stable at 1398-1405K\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: stable at 1585-1592K\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: stable at 1125-1130K", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"10%-30%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Moderate Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Moderate Condition, Failure probability: 10%-30%.\", \"recommended_actions\": \"It is crucial to closely monitor the engine's condition during this period and proactively schedule maintenance in the near future to prevent potential issues and maintain operational reliability.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"89747\", \"89763\", \"89781\", \"89802\", \"89810\", \"89833\", \"89851\", \"89860\", \"89870\", \"89891\"]}", "known_issue": "Question text is identical to engine_health_combined_013, but the two ground truths disagree on qualitative_condition."}
{"id": "engine_health_combined_018", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Our predictive maintenance system flagged this engine for review. Analyze the 10-cycle trend data and provide: overall condition assessment, failure probability estimate, and your recommended maintenance strategy.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: gradual decline: 1588K → 1571K (-1.1%)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"30%-50%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Poor Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Poor Condition, Failure probability: 30%-50%.\", \"recommended_actions\": \"It is essential to conduct repairs or replace the affected components at the earliest opportunity to ensure the continued safe and efficient operation of the engine.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"100319\", \"100335\", \"100349\", \"100366\", \"100376\", \"100388\", \"100402\", \"100421\", \"100435\", \"100451\"]}", "known_issue": "Scenarios 018-020 share one signal summary but differ in failure_probability_range (30%-50% vs 70%-100%)."}
{"id": "engine_health_combined_019", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine is due for its scheduled inspection. Based on the multi-cycle signal data, assess whether the condition has degraded, estimate failure risk, and recommend whether to proceed with routine maintenance or escalate to urgent repairs.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: gradual decline: 1588K → 1571K (-1.1%)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"70%-100%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Poor Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Poor Condition, Failure probability: 70%-100%.\", \"recommended_actions\": \"It is critical to prioritize and promptly schedule maintenance activities to mitigate the risk of an unexpected shutdown or the potential development of more severe and catastrophic failures.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"74131\", \"74165\", \"74189\", \"74213\", \"74231\", \"74257\", \"74275\", \"74308\", \"74334\", \"74363\"]}", "known_issue": "Scenarios 018-020 share one signal summary but differ in failure_probability_range (30%-50% vs 70%-100%)."}
{"id": "engine_health_combined_020", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Fleet management needs a health trend report for this engine unit. Analyze the 10-cycle data and provide: (1) Current condition classification, (2) Failure probability range for the next 50 cycles, (3) Specific maintenance recommendations.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: gradual decline: 1588K → 1571K (-1.1%)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"70%-100%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Poor Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Poor Condition, Failure probability: 70%-100%.\", \"recommended_actions\": \"It is critical to promptly initiate maintenance procedures to mitigate the risk of an unexpected shutdown or the potential development of more severe and costly system failures.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"20473\", \"20490\", \"20502\", \"20514\", \"20534\", \"20543\", \"20555\", \"20574\", \"20584\", \"20603\"]}", "known_issue": "Scenarios 018-020 share one signal summary but differ in failure_probability_range (30%-50% vs 70%-100%)."}
{"id": "engine_health_combined_021", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "I've collected engine telemetry across 10 operational cycles and need a trend analysis. Based on this data: (1) What is the qualitative condition of the engine (Good/Moderate/Poor/Bad/Extremely Poor)? (2) What's the estimated probability range of engine failure? (3) What maintenance actions should we take?\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"50%-70%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Bad Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Bad Condition, Failure probability: 50%-70%.\", \"recommended_actions\": \"It is imperative to promptly repair or replace the affected components and conduct a comprehensive evaluation of the overall health status of the engine to ensure optimal performance and safety.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"112649\", \"112680\", \"112706\", \"112724\", \"112748\", \"112773\", \"112805\", \"112827\", \"112850\", \"112878\"]}", "known_issue": ""}
{"id": "engine_health_combined_022", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Our predictive maintenance system flagged this engine for review. Analyze the 10-cycle trend data and provide: overall condition assessment, failure probability estimate, and your recommended maintenance strategy.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data. 2. Analyze degradation patterns across cycles. 3. Classify qualitative condition. 4. Estimate failure probability based on trend severity. 5. Generate maintenance recommendations appropriate to condition.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"50%-70%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Bad Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Bad Condition, Failure probability: 50%-70%.\", \"recommended_actions\": \"It may be advisable to urgently address the issue by repairing or replacing parts and thoroughly assessing the overall health status of the engine. This approach could help mitigate potential risks and ensure continued safe operation.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": [\"16461\", \"16484\", \"16499\", \"16518\", \"16546\", \"16556\", \"16575\", \"16601\", \"16616\", \"16631\"]}", "known_issue": ""}
{"id": "engine_health_combined_023", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Emergency diagnostic request: An engine fault indicator lit up during cruise. Review the telemetry and determine which components have failed, the nature of the timing issue, and the severity of the situation.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "diagnose_timing_issues"], "procedure": "1. Parse sensor data for anomaly indicators. 2. Identify which components show fault signatures. 3. Analyze timing patterns to determine efficiency vs flow issues. 4. Pinpoint the affected subsystem. 5. Assess overall severity.", "ground_truth": "{\"component_health_detail\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"fault_diagnosis\", \"faulty_components\": [\"HPT\"], \"healthy_components\": [\"Fan\", \"LPC\", \"HPC\", \"LPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Fault diagnosis from EngineMTQA engine 2364. Faulty: ['HPT']. Root cause: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"fault_diagnosis\", \"source_engine_id\": \"2364\"}", "known_issue": ""}
{"id": "engine_health_combined_024", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Post-flight inspection revealed abnormal sensor readings. Diagnose which components are compromised, identify the root cause of any efficiency or flow anomalies, and assess whether immediate repair is needed.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "diagnose_timing_issues"], "procedure": "1. Parse sensor data for anomaly indicators. 2. Identify which components show fault signatures. 3. Analyze timing patterns to determine efficiency vs flow issues. 4. Pinpoint the affected subsystem. 5. Assess overall severity.", "ground_truth": "{\"component_health_detail\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"fault_diagnosis\", \"faulty_components\": [\"HPT\"], \"healthy_components\": [\"Fan\", \"LPC\", \"HPC\", \"LPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Fault diagnosis from EngineMTQA engine 4527. Faulty: ['HPT']. Root cause: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"fault_diagnosis\", \"source_engine_id\": \"4527\"}", "known_issue": ""}
{"id": "engine_health_combined_025", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine experienced a performance anomaly mid-flight. Analyze the sensor data to identify: (1) Which components (Fan, LPC, HPC, HPT, LPT) are faulty? (2) What's the root cause - efficiency modifier issues or flow modifier problems? (3) Which specific subsystem is affected?\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "diagnose_timing_issues"], "procedure": "1. Parse sensor data for anomaly indicators. 2. Identify which components show fault signatures. 3. Analyze timing patterns to determine efficiency vs flow issues. 4. Pinpoint the affected subsystem. 5. Assess overall severity.", "ground_truth": "{\"component_health_detail\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"fault_diagnosis\", \"faulty_components\": [\"HPT\"], \"healthy_components\": [\"Fan\", \"LPC\", \"HPC\", \"LPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Fault diagnosis from EngineMTQA engine 6781. Faulty: ['HPT']. Root cause: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"fault_diagnosis\", \"source_engine_id\": \"6781\"}", "known_issue": ""}
{"id": "engine_health_combined_026", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Emergency diagnostic request: An engine fault indicator lit up during cruise. Review the telemetry and determine which components have failed, the nature of the timing issue, and the severity of the situation.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: stable at 518-520K (within normal range)\n  - Fan Inlet Pressure: constant at 14.62 psia (nominal)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 81.4% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "diagnose_timing_issues"], "procedure": "1. Parse sensor data for anomaly indicators. 2. Identify which components show fault signatures. 3. Analyze timing patterns to determine efficiency vs flow issues. 4. Pinpoint the affected subsystem. 5. Assess overall severity.", "ground_truth": "{\"component_health_detail\": {\"Fan\": \"Healthy\", \"HPC\": \"Healthy\", \"HPT\": \"Fault\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"fault_diagnosis\", \"faulty_components\": [\"HPT\"], \"healthy_components\": [\"Fan\", \"LPC\", \"HPC\", \"LPT\"], \"overall_health_state\": \"Abnormal Degradation\", \"rationale\": \"Fault diagnosis from EngineMTQA engine 3102. Faulty: ['HPT']. Root cause: HPT efficiency modifier.\", \"timing_issue_root_cause\": \"HPT efficiency modifier\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"fault_diagnosis\", \"source_engine_id\": \"3102\"}", "known_issue": ""}
{"id": "engine_health_combined_027", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Post-flight inspection revealed abnormal sensor readings. Diagnose which components are compromised, identify the root cause of any efficiency or flow anomalies, and assess whether immediate repair is needed.\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: stable at 35,000 ft with minor fluctuations (±200 ft)\n  - Throttle Resolver Angle: irregular pattern with unexpected spikes (76.3° to 89.1°)\n\n[Fan Section]\n  - Fan Inlet Temp: elevated and rising: 528K → 542K (+2.6% over 10 cycles, ANOMALY)\n  - Fan Inlet Pressure: declining: 14.62 psia → 13.89 psia (-5.0%, CRITICAL)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: abnormally high: 565K → 589K (+4.2%, WARNING)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: sharp increase: 1398K → 1456K (+4.1%, CRITICAL)\n  - HPC Outlet Pressure: unstable: fluctuating 520-535 psia (7% variance)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1089K with oscillations (FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "diagnose_timing_issues"], "procedure": "1. Parse sensor data for anomaly indicators. 2. Identify which components show fault signatures. 3. Analyze timing patterns to determine efficiency vs flow issues. 4. Pinpoint the affected subsystem. 5. Assess overall severity.", "ground_truth": "{\"component_health_detail\": {\"Fan\": \"Fault\", \"HPC\": \"Fault\", \"HPT\": \"Fault\", \"LPC\": \"Fault\", \"LPT\": \"Fault\"}, \"expected_answer_type\": \"fault_diagnosis\", \"faulty_components\": [\"Fan\", \"LPC\", \"HPC\", \"HPT\", \"LPT\"], \"healthy_components\": [], \"overall_health_state\": \"Abnormal Degradation\", \"rationale\": \"Fault diagnosis from EngineMTQA engine 114993. Faulty: ['Fan', 'LPC', 'HPC', 'HPT', 'LPT']. Root cause: a, b, c, d, e, f.\", \"timing_issue_root_cause\": \"Multiple efficiency and flow modifiers affected across components\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"fault_diagnosis\", \"source_engine_id\": \"114993\"}", "known_issue": ""}
{"id": "engine_health_combined_028", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "This engine experienced a performance anomaly mid-flight. Analyze the sensor data to identify: (1) Which components (Fan, LPC, HPC, HPT, LPT) are faulty? (2) What's the root cause - efficiency modifier issues or flow modifier problems? (3) Which specific subsystem is affected?\n\nENGINE SIGNAL DATA SUMMARY (1 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: elevated and rising: 528K → 542K (+2.6% over 10 cycles, ANOMALY)\n  - Fan Inlet Pressure: declining: 14.62 psia → 13.89 psia (-5.0%, CRITICAL)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: gradual decline: 554K → 548K (-1.1%)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: gradual decline: 1402K → 1388K (-1.0%)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: declining rapidly: 1582K → 1534K (-3.0%)\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: slow decline: 1128K → 1119K (-0.8%)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "diagnose_timing_issues"], "procedure": "1. Parse sensor data for anomaly indicators. 2. Identify which components show fault signatures. 3. Analyze timing patterns to determine efficiency vs flow issues. 4. Pinpoint the affected subsystem. 5. Assess overall severity.", "ground_truth": "{\"component_health_detail\": {\"Fan\": \"Fault\", \"HPC\": \"Healthy\", \"HPT\": \"Healthy\", \"LPC\": \"Healthy\", \"LPT\": \"Healthy\"}, \"expected_answer_type\": \"fault_diagnosis\", \"faulty_components\": [\"Fan\"], \"healthy_components\": [\"LPC\", \"HPC\", \"HPT\", \"LPT\"], \"overall_health_state\": \"Imminent Failure\", \"rationale\": \"Fault diagnosis from EngineMTQA engine 93061. Faulty: ['Fan']. Root cause: c, f.\", \"timing_issue_root_cause\": \"Multiple efficiency and flow modifiers affected across components\", \"verification_required\": true}", "metadata": "{\"complexity\": \"High\", \"is_closed_ended\": true, \"scenario_type\": \"fault_diagnosis\", \"source_engine_id\": \"93061\"}", "known_issue": ""}
{"id": "engine_health_combined_029", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "URGENT: This engine is showing critical degradation across multiple systems. Analyze the 10-cycle trend data and determine: (1) The qualitative condition of the engine, (2) The probability range of imminent failure, (3) What emergency maintenance actions are needed?\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: elevated and rising: 528K → 542K (+2.6% over 10 cycles, ANOMALY)\n  - Fan Inlet Pressure: declining: 14.62 psia → 13.89 psia (-5.0%, CRITICAL)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: abnormally high: 565K → 589K (+4.2%, WARNING)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: sharp increase: 1398K → 1456K (+4.1%, CRITICAL)\n  - HPC Outlet Pressure: unstable: fluctuating 520-535 psia (7% variance)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 74.1% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1042K with severe oscillations (CRITICAL FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data showing severe degradation. 2. Analyze critical anomalies across multiple components. 3. Classify as Extremely Poor Condition. 4. Estimate high failure probability (70%-100%). 5. Generate urgent maintenance/grounding recommendations.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"70%-100%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Extremely Poor Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Extremely Poor Condition, Failure probability: 70%-100%. This represents the most severe engine degradation category.\", \"recommended_actions\": \"Immediate grounding required. The engine shows critical degradation across multiple systems and must be removed from service for comprehensive overhaul or replacement.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"Critical\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": \"synthetic_001\"}", "known_issue": ""}
{"id": "engine_health_combined_030", "subset": "rotating_aero", "task_category": "Health Analysis", "asset_class": "Aero-engine", "source_dataset": "EngineMTQA", "question": "Fleet alert: Engine performance has degraded severely over the past 10 cycles. Based on this critical sensor data, assess the overall condition, estimate failure probability, and recommend whether this engine should be immediately grounded for emergency repairs.\n\nENGINE SIGNAL DATA SUMMARY (10 operational cycles):\n\n[Operational Parameters]\n  - Altitude: fluctuating between 34,200-35,800 ft (abnormal variance)\n  - Throttle Resolver Angle: erratic behavior with sharp oscillations (68.4° to 94.2°)\n\n[Fan Section]\n  - Fan Inlet Temp: elevated and rising: 528K → 542K (+2.6% over 10 cycles, ANOMALY)\n  - Fan Inlet Pressure: declining: 14.62 psia → 13.89 psia (-5.0%, CRITICAL)\n\n[Low-Pressure Compressor (LPC)]\n  - LPC Outlet Temp: abnormally high: 565K → 589K (+4.2%, WARNING)\n\n[High-Pressure Compressor (HPC)]\n  - HPC Outlet Temp: sharp increase: 1398K → 1456K (+4.1%, CRITICAL)\n  - HPC Outlet Pressure: unstable: fluctuating 520-535 psia (7% variance)\n\n[High-Pressure Turbine (HPT)]\n  - HPT Outlet Temp: ANOMALY DETECTED: sharp decline 1589K → 1498K (-5.7%/cycle)\n  - HPT Coolant Bleed: elevated: 8.45 → 8.89 (+5.2%, compensating for degradation)\n  - HPT Efficiency Indicator: DEGRADED: 87.2% → 74.1% efficiency drop detected\n\n[Low-Pressure Turbine (LPT)]\n  - LPT Outlet Temp: abnormal pattern: 1124K → 1042K with severe oscillations (CRITICAL FAULT)", "question_operator_style": "", "required_tools": ["analyze_engine_signals", "detect_degradation_trend", "assess_component_health", "optimize_maintenance_schedule"], "procedure": "1. Parse 10-cycle sensor trend data showing severe degradation. 2. Analyze critical anomalies across multiple components. 3. Classify as Extremely Poor Condition. 4. Estimate high failure probability (70%-100%). 5. Generate urgent maintenance/grounding recommendations.", "ground_truth": "{\"condition_options\": [\"Good Condition\", \"Moderate Condition\", \"Poor Condition\", \"Bad Condition\", \"Extremely Poor Condition\"], \"expected_answer_type\": \"trend_analysis_with_recommendations\", \"failure_probability_range\": \"70%-100%\", \"probability_options\": [\"1%-10%\", \"10%-30%\", \"30%-50%\", \"50%-70%\", \"70%-100%\"], \"qualitative_condition\": \"Extremely Poor Condition\", \"rationale\": \"Ground truth from EngineMTQA. Condition: Extremely Poor Condition, Failure probability: 70%-100%. This represents the most severe engine degradation category.\", \"recommended_actions\": \"This engine is in a critical state requiring immediate shutdown. All faulty components must be replaced before any further operation. A full teardown inspection is mandatory.\", \"verification_required\": true}", "metadata": "{\"complexity\": \"Critical\", \"is_closed_ended\": false, \"scenario_type\": \"multi_cycle_trend_analysis\", \"source_engine_ids\": \"synthetic_002\"}", "known_issue": ""}
{"id": "bess_health_001", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Perform health analysis for battery B0005 from cycle 60 to 80. Identify the degradation stage, SOH percentage, and current capacity.", "question_operator_style": "Can you check Battery 5 from about cycle 60 to cycle 80. Is it still healthy?", "required_tools": ["fetch_cycle_summary", "capacity_soh_calculator"], "procedure": "Analyze capacity data and compare against standard thresholds (91% for Healthy).", "ground_truth": "{\"answer_type\": \"structured_categorical\", \"capacity\": 1.5598, \"soh_pct\": 83.86, \"source\": \"ground_truth.json:B0005:cycle_labels[80]\", \"stage\": \"EARLY_DEGRADATION\"}", "metadata": "{\"cognitive_category\": \"Perception\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Retrieve cycles 60-80 summary\", \"Execute SOH calculator with 1.86Ah rated capacity\"], \"difficulty\": \"easy\", \"disambiguation_criteria\": [\"cycle_end 80 determines current state\", \"SOH threshold maps to EARLY_DEGRADATION\", \"capacity marker must match processed data\"], \"expected_failure_modes\": [\"confusing current stage with healthy baseline\", \"omitting SOH or capacity markers\", \"using the wrong cycle endpoint\"], \"output_template\": {\"current_capacity\": \"float\", \"soh_pct\": \"float\", \"stage\": \"str\"}}", "known_issue": ""}
{"id": "bess_health_002", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Retrieve discharge capacity for B0005 for the first 10 cycles and cycles 90-100. Calculate the percentage drop in capacity.", "question_operator_style": "For Battery 5, compare the first 10 cycles with cycles 90 through 100. How much capacity has it lost?", "required_tools": ["fetch_cycle_summary"], "procedure": "Time-sequence comparison of historical vs current capacity.", "ground_truth": "{\"answer_type\": \"numeric_percentage\", \"capacity_drop_pct\": 20.17, \"current_avg_cap\": 1.48, \"initial_avg_cap\": 1.854, \"source\": \"data/processed/B0005_cycles.json\"}", "metadata": "{\"cognitive_category\": \"Understanding\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Average capacity for cycles 0-10\", \"Average capacity for cycles 90-100\", \"Calculate relative drop\"], \"difficulty\": \"easy\", \"disambiguation_criteria\": [\"compute mean capacity in each requested window\", \"capacity_drop_pct = (initial-current)/initial * 100\", \"use discharge_capacity, not voltage or duration\"], \"expected_failure_modes\": [\"using endpoint capacity instead of window averages\", \"reporting absolute Ah drop instead of percent drop\", \"reversing initial and current windows\"]}", "known_issue": "Ground truth uses the c100 endpoint (1.48 Ah, 20.17% drop) instead of the window means the question defines (1.515 Ah, 17.45%)."}
{"id": "bess_health_003", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Determine the degradation stage of B0005 at cycle 120 using the standard SOH thresholds (HEALTHY >= 90%, EARLY 80-90%, ACCELERATED 70-80%, EOL < 70%). Provide a maintenance recommendation chosen from {continue, monitor, plan replacement, replace immediately} based on the stage. Output: {stage, soh_pct, recommendation}.", "question_operator_style": "Battery 5 has gone through about 120 cycles now. Should we keep cycling it, or is it time to plan something? Give me a maintenance call.", "required_tools": ["capacity_soh_calculator"], "procedure": "Map technical state to operational decisions.", "ground_truth": "{\"answer_type\": \"decision_choice\", \"recommendation\": \"Schedule for decommissioning or secondary use; prioritize replacement.\", \"source\": \"ground_truth.json:B0005:cycle_labels[120]\", \"stage\": \"ACCELERATED_DEGRADATION\"}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"SOH at cycle 120\", \"Consult maintenance_manual.md for stage-action mapping\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"stage is based on cycle-120 SOH threshold\", \"ACCELERATED maps to plan replacement/decommissioning\", \"recommendation must be supported by stage\"], \"expected_failure_modes\": [\"understating accelerated degradation as routine monitoring only\", \"ignoring SOH threshold-to-action mapping\", \"returning an action without stage evidence\"]}", "known_issue": "Question thresholds (90/80/70) differ from the benchmark's (91/83/75); the verdict is the same here."}
{"id": "bess_health_004", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Calculate the cumulative thermal drift for battery B0005 from cycle 80 to cycle 120. Cumulative thermal drift is defined as max_temperature(cycle 120) - max_temperature(cycle 80). Report a single Celsius value, not a per-cycle rate.", "question_operator_style": "For Battery 5, compare its temperature from cycle 80 to cycle 120. How much has it drifted?", "required_tools": ["thermal_anomaly_checker", "fetch_cycle_summary"], "procedure": "Identify cumulative thermal drift as a proxy for internal resistance increase.", "ground_truth": "{\"answer_type\": \"numeric_trend\", \"classification\": \"normal_aging_related_rise\", \"source\": \"data/processed/B0005_cycles.json\", \"thermal_drift\": 1.85}", "metadata": "{\"cognitive_category\": \"Perception\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Trend analysis over 40 cycles\", \"Compare thermal baseline\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"thermal drift is cumulative max(end)-max(start)\", \"per-cycle rate is not the requested answer\", \"smooth 0.5-2 C rise is normal aging related\"], \"expected_failure_modes\": [\"confusing cumulative drift with per-cycle drift\", \"using avg_temperature when max_temperature drift is specified\", \"misclassifying smooth aging as a thermal spike\"]}", "known_issue": ""}
{"id": "bess_health_005", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Analyze the state of health for B0005 during cycles 0-20. Confirm its stage according to standard thresholds.", "question_operator_style": "During the first 20 cycles for Battery 5, was it still in healthy condition?", "required_tools": ["capacity_soh_calculator"], "procedure": "Confirm healthy operation baseline.", "ground_truth": "{\"answer_type\": \"structured_categorical\", \"soh_pct\": 99.8, \"source\": \"ground_truth.json:B0005:cycle_labels[0-20]\", \"stage\": \"HEALTHY\"}", "metadata": "{\"cognitive_category\": \"Perception\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Verify SOH > 91%\"], \"difficulty\": \"easy\", \"disambiguation_criteria\": [\"cycles 0-20 are early-life baseline\", \"HEALTHY requires SOH above threshold\", \"no late-cycle evidence should be used\"], \"expected_failure_modes\": [\"using a late-life cycle window\", \"treating early natural variance as degradation\", \"omitting threshold-based stage confirmation\"]}", "known_issue": ""}
{"id": "bess_health_006", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0018", "question": "Explain the capacity trend of B0018 between cycles 100 and 130. Distinguish between sensor error and physical regeneration.", "question_operator_style": "Battery 18 seems to rebound sometime after the first 100 cycles. Looking through the next 30 cycles or so, is this a sensor glitch we should worry about?", "required_tools": ["fetch_cycle_data", "anomaly_detector"], "procedure": "Physical phenomenon identification (Capacity Regeneration).", "ground_truth": "{\"answer_type\": \"reasoning_explanation\", \"event_cycle\": 120, \"is_regeneration\": true, \"justification\": \"Capacity rose from 1.34Ah to 1.43Ah; consistent increase in discharge duration confirms physical effect, not sensor noise.\", \"source\": \"ground_truth.json:B0018:cycle_labels[120], anomalies\"}", "metadata": "{\"cognitive_category\": \"Understanding\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Identify jump at c120\", \"Check consistency across multiple features (voltage, duration)\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"event_cycle is 120\", \"capacity and duration both rise around the event\", \"physical regeneration is not sensor noise when corroborated\"], \"expected_failure_modes\": [\"misclassifying regeneration as sensor noise\", \"missing the cycle-120 recovery event\", \"ignoring discharge-duration corroboration\"]}", "known_issue": "Ground truth accepts only the c120 rebound; a larger rebound at c105 inside the asked window is omitted."}
{"id": "bess_health_007", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005_B0007", "question": "Calculate the average degradation rate (Ah/cycle) for B0005 and B0007 from cycle 0 to 100. Which one is degrading faster?", "question_operator_style": "Over their first 100 cycles, which one is aging faster, Battery 5 or Battery 7?", "required_tools": ["capacity_soh_calculator", "fetch_cycle_summary"], "procedure": "Cross-battery comparative analysis.", "ground_truth": "{\"B0005_rate\": 0.0038, \"B0007_rate\": 0.0033, \"answer_type\": \"comparative_numeric\", \"faster_degrading\": \"B0005\", \"source\": \"data/processed/B0005_cycles.json, B0007_cycles.json\"}", "metadata": "{\"cognitive_category\": \"Reasoning\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Calculate slope for both IDs\", \"B0007 as a long-life control case\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"compare absolute degradation-rate magnitudes\", \"same cycle window 0-100 for both batteries\", \"larger magnitude means faster degradation\"], \"expected_failure_modes\": [\"comparing signed slopes instead of magnitudes\", \"using different windows for the two batteries\", \"declaring the slower-aging control as faster\"]}", "known_issue": ""}
{"id": "bess_health_008", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0018", "question": "Evaluate B0018 at cycle 100 against the 1.4 Ah EOL threshold. Determine: (a) which degradation stage (HEALTHY / EARLY / ACCELERATED / EOL) it falls into, (b) the current discharge capacity, and (c) the recommended action (continue / monitor / replace). Output structured: {stage, capacity_Ah, action}.", "question_operator_style": "Battery 18 is around its 100th cycle now. Is it still safe to keep cycling it?", "required_tools": ["capacity_soh_calculator"], "procedure": "EOL Detection and terminal decision.", "ground_truth": "{\"action\": \"Immediate termination required. Battery has reached end-of-life.\", \"answer_type\": \"structured_categorical\", \"capacity\": 1.385, \"source\": \"ground_truth.json:B0018:cycle_labels[100]\", \"stage\": \"EOL\"}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Check capacity at c100\", \"State change detection\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"capacity below 1.4 Ah implies EOL\", \"EOL action is replacement/termination\", \"cycle 100 is the evaluation point\"], \"expected_failure_modes\": [\"missing the 1.4 Ah EOL threshold crossing\", \"giving monitor instead of replacement action\", \"reporting SOH without stage/action\"]}", "known_issue": ""}
{"id": "bess_rul_001", "subset": "lithium_ion", "task_category": "RUL Prediction", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Estimate the Remaining Useful Life (RUL) of battery B0005 starting from cycle 40 until it reaches its 1.4Ah EOL threshold.", "question_operator_style": "Battery 5 has completed about 40 cycles. How much useful life does it have left from there?", "required_tools": ["rul_predictor_ttm_finetuned", "rul_predictor_ttm_zero_shot", "rul_predictor_lstm"], "procedure": "Foundation-model RUL forecasting based on early-life degradation pattern.", "ground_truth": "{\"answer_type\": \"numeric_cycles\", \"mae_threshold\": 15, \"source\": \"ground_truth.json:B0005:cycle_labels[40]\", \"true_rul\": 84}", "metadata": "{\"cognitive_category\": \"Prediction\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Provide cycle window [0, 40]\", \"Invoke RUL predictor\", \"Specify EOL threshold 1.4\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"forecast from context ending at cycle 40\", \"EOL threshold is 1.4 Ah\", \"prediction passes if within MAE threshold\"], \"expected_failure_modes\": [\"using a late-cycle context instead of cycle 40\", \"calling a non-primary predictor without explanation\", \"forgetting the 1.4 Ah EOL threshold\"]}", "known_issue": ""}
{"id": "bess_rul_002", "subset": "lithium_ion", "task_category": "RUL Prediction", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0006", "question": "Estimate RUL for battery B0006 at cycle 80. The cell has shown accelerated degradation. Apply the most appropriate prognostic method given the rapid decay pattern, and report your model choice with the RUL estimate.", "question_operator_style": "Battery 6 has been losing capacity quickly by around cycle 80. How much life does it have left at that point?", "required_tools": ["rul_predictor_empirical", "rul_predictor_ttm_finetuned"], "procedure": "Predictive analysis for fast-degrading cells.", "ground_truth": "{\"answer_type\": \"numeric_cycles\", \"mae_threshold\": 15, \"source\": \"ground_truth.json:B0006:cycle_labels[80]\", \"true_rul\": 28}", "metadata": "{\"cognitive_category\": \"Prediction\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Recognize high degradation rate\", \"Use empirical decay as the primary fast-decay forecast\", \"Forecast until 1.4Ah crossing\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"fast decay should guide method selection\", \"primary RUL marker must match chosen primary model\", \"MAE threshold is 15 cycles\"], \"expected_failure_modes\": [\"using a smooth-degradation prior on a fast-decay cell\", \"reporting secondary model output as primary without justification\", \"omitting model choice\"]}", "known_issue": ""}
{"id": "bess_rul_003", "subset": "lithium_ion", "task_category": "RUL Prediction", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0018", "question": "Predict the RUL for battery B0018 at cycle 80 against the 1.4 Ah EOL threshold. The cell has shown irregular behavior in its history; choose the most appropriate forecasting method and report your reasoning.", "question_operator_style": "Take a look at Battery 18 around cycle 80. Given the cycles observed so far, how many cycles does it still have in it?", "required_tools": ["rul_predictor_chronos", "rul_predictor_ttm_finetuned"], "procedure": "Prognostics in the presence of capacity regeneration.", "ground_truth": "{\"answer_type\": \"numeric_cycles\", \"mae_threshold\": 10, \"source\": \"ground_truth.json:B0018:cycle_labels[80]\", \"true_rul\": 16}", "metadata": "{\"cognitive_category\": \"Prediction\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Forecast RUL for non-monotonic curve\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"forecast against 1.4 Ah EOL threshold\", \"method choice should account for irregular history\", \"MAE threshold is 10 cycles\"], \"expected_failure_modes\": [\"treating irregular history as a simple linear decline\", \"ignoring EOL threshold during forecast\", \"reporting a forecast without method reasoning\"]}", "known_issue": ""}
{"id": "bess_rul_004", "subset": "lithium_ion", "task_category": "RUL Prediction", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_Fleet", "question": "Estimate RUL for batteries B0005, B0006, B0007, and B0018 at cycle 80. Rank them from shortest to longest remaining life.", "question_operator_style": "At around cycle 80 for our four cell fleet, which battery is going to fail first? Rank them from soonest to latest.", "required_tools": ["rul_predictor_ttm_finetuned", "rul_predictor_lstm"], "procedure": "Fleet-level comparative prognostics.", "ground_truth": "{\"answer_type\": \"ordinal_list\", \"ranking\": [\"B0018\", \"B0006\", \"B0005\", \"B0007\"], \"ruls\": {\"B0005\": 44, \"B0006\": 28, \"B0007\": \"never_eol\", \"B0018\": 16}, \"source\": \"ground_truth.json:multiple\"}", "metadata": "{\"cognitive_category\": \"Reasoning\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Batch RUL estimation\", \"Ordinal ranking\", \"Handle B0007 non-EOL status\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"rank shortest remaining life first\", \"include all four batteries exactly once\", \"B0007 is censored but still part of ranking\"], \"expected_failure_modes\": [\"ranking longest-life first\", \"dropping B0007 from the fleet list\", \"mixing model outputs across batteries inconsistently\"]}", "known_issue": ""}
{"id": "bess_rul_005", "subset": "lithium_ion", "task_category": "RUL Prediction", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Estimate RUL for B0005 at cycle 120 with a 95% confidence interval. Recommend immediate replacement vs deferral based on prediction certainty.", "question_operator_style": "Battery 5 has reached about 120 cycles. Should we be pulling it out for replacement, or can it run a bit longer?", "required_tools": ["rul_predictor_chronos", "rul_predictor_empirical", "rul_predictor_lstm"], "procedure": "Critical state decision under uncertainty.", "ground_truth": "{\"answer_type\": \"numeric_cycles_with_interval\", \"mae_threshold\": 5, \"recommendation\": \"Replace immediately; RUL < 5 and uncertainty interval is tight.\", \"source\": \"ground_truth.json:B0005:cycle_labels[120]\", \"true_rul\": 4}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Near-EOL RUL check\", \"Use Chronos as the primary short-horizon forecast\", \"Uncertainty quantification\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"near-EOL RUL is less than 5 cycles\", \"confidence interval informs replacement decision\", \"recommend immediate replacement when short horizon is tight\"], \"expected_failure_modes\": [\"answering replacement decision without RUL uncertainty\", \"missing the near-EOL short horizon\", \"using a wide confidence interval to justify delay\"]}", "known_issue": ""}
{"id": "bess_fault_001", "subset": "lithium_ion", "task_category": "Fault Classification", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Classify the degradation stage for B0005 at each of these cycle markers: 0, 40, 80, 100, 120. Use the 4-stage taxonomy (HEALTHY / EARLY_DEGRADATION / ACCELERATED_DEGRADATION / EOL). Output a list of {cycle: stage} pairs.", "question_operator_style": "Walk me through how Battery 5 degraded from the first cycle through about cycle 120. Check the first cycle, then around cycles 40, 80, 100, and 120.", "required_tools": ["capacity_soh_calculator"], "procedure": "Temporal degradation stage mapping.", "ground_truth": "{\"answer_type\": \"list_categorical\", \"source\": \"ground_truth.json:B0005\", \"stages\": {\"0\": \"HEALTHY\", \"100\": \"ACCELERATED_DEGRADATION\", \"120\": \"ACCELERATED_DEGRADATION\", \"40\": \"HEALTHY\", \"80\": \"EARLY_DEGRADATION\"}}", "metadata": "{\"cognitive_category\": \"Understanding\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Multi-point stage classification\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"use canonical 4-stage taxonomy\", \"classify each requested cycle separately\", \"return cycle-stage pairs\"], \"expected_failure_modes\": [\"collapsing all cycles to one stage\", \"using noncanonical stage labels\", \"off-by-one cycle marker lookup\"]}", "known_issue": ""}
{"id": "bess_fault_002", "subset": "lithium_ion", "task_category": "Fault Classification", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0006", "question": "Identify capacity and thermal anomalies in battery B0006 from cycle 0 to cycle 168. Use the anomaly_detector tool with parameters rolling_window=20, z_score_threshold=2.55, min_persistence=1. Report the total count of detected anomaly events.", "question_operator_style": "Looking across the full run of Battery 6, from the first cycle through about cycle 168, are there any moments that stand out as abnormal compared to its general trend?", "required_tools": ["anomaly_detector"], "procedure": "Statistical fault discovery.", "ground_truth": "{\"anomaly_count\": 26, \"answer_type\": \"list_numeric\", \"detector_params\": {\"min_persistence\": 1, \"rolling_window\": 20, \"z_score_threshold\": 2.55}, \"example_cycles\": [1, 2, 45, 68, 89], \"source\": \"ground_truth.json:B0006:anomalies\"}", "metadata": "{\"cognitive_category\": \"Perception\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Unsupervised anomaly detection\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"use rolling_window=20\", \"use z_score_threshold=2.55\", \"min_persistence=1 and report total event count\"], \"expected_failure_modes\": [\"using default detector parameters when benchmark params matter\", \"counting features instead of anomaly events inconsistently\", \"reporting examples without total count\"]}", "known_issue": "Only 1 of the 5 ground-truth example cycles is reproduced by the detector with the stated parameters."}
{"id": "bess_fault_003", "subset": "lithium_ion", "task_category": "Fault Classification", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0006", "question": "B0006 shows a capacity jump of +0.152 Ah at cycle 89. Determine whether this is (a) capacity_regeneration_normal (a known lithium-ion phenomenon where capacity rebounds after rest), or (b) a hardware fault (sensor malfunction or cell defect). Disambiguation criteria: regeneration returns to the prior decay trend within 5 cycles; faults persist or escalate. Provide classification with magnitude and reasoning.", "question_operator_style": "Battery 6 had a sudden capacity jump around cycle 89. Is that a fault we should investigate, or is it just normal battery behavior?", "required_tools": ["fetch_cycle_summary", "anomaly_detector"], "procedure": "Physicochemical interpretation of abnormal trends.", "ground_truth": "{\"answer_type\": \"reasoning_explanation\", \"classification\": \"capacity_regeneration_normal\", \"magnitude\": 0.152, \"reasoning\": \"A rapid capacity increase of 0.152 Ah followed by normal exponential decay is a textbook capacity regeneration event due to rest periods or chemistry relaxation, not a hardware fault.\", \"source\": \"Action 3 verified data\"}", "metadata": "{\"cognitive_category\": \"Reasoning\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Observe jump\", \"Check subsequent decay rate\", \"Identify as regeneration\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"jump magnitude is +0.152 Ah\", \"normal regeneration returns toward decay trend\", \"persistent or escalating deviation would indicate fault\"], \"expected_failure_modes\": [\"misclassifying regeneration as a hardware fault\", \"ignoring the +0.152 Ah jump magnitude\", \"failing to check recovery after the event\"]}", "known_issue": ""}
{"id": "bess_fault_004", "subset": "lithium_ion", "task_category": "Fault Classification", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0007", "question": "Evaluate anomalies in B0007. Despite having 19 statistical anomalies, is this battery in the EOL stage?", "question_operator_style": "Battery 7 is showing some thermal noise. Does that mean the cell is on its way out?", "required_tools": ["anomaly_detector", "capacity_soh_calculator"], "procedure": "Discriminative analysis of health vs. noise.", "ground_truth": "{\"answer_type\": \"reasoning_explanation\", \"has_anomalies\": true, \"is_eol\": false, \"logic\": \"B0007 exhibits thermal noise and minor capacity fluctuations (19 anomalies), but its capacity remains consistently above 1.4Ah, thus it is NOT in the EOL stage.\", \"source\": \"ground_truth.json:B0007\"}", "metadata": "{\"cognitive_category\": \"Understanding\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Distinguish signal noise from terminal health state\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"EOL depends on capacity threshold, not anomaly count\", \"statistical anomalies can coexist with non-EOL state\", \"capacity remains above 1.4 Ah\"], \"expected_failure_modes\": [\"equating statistical anomalies with EOL\", \"ignoring capacity staying above 1.4 Ah\", \"treating thermal noise as terminal failure\"]}", "known_issue": "The premise '19 statistical anomalies' is not reproducible (no parameters given; defaults give 48)."}
{"id": "bess_cost_001", "subset": "lithium_ion", "task_category": "Cost-Benefit Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Battery B0005 at observation cycle 80 has a confirmed remaining useful life of 44 cycles, provided as ground truth. Use this given RUL only; do not re-predict it. Given replacement cost $450 and reactive failure cost $2800, calculate the cost-optimal replacement cycle. Report a target cycle range.", "question_operator_style": "Battery 5 is around cycle 80, and we need to schedule replacement. When is the right time given where it is in its life?", "required_tools": [], "procedure": "Economic optimization of maintenance scheduling.", "ground_truth": "{\"answer_type\": \"numeric_optimization\", \"optimal_cycle_range\": [115, 124], \"reasoning\": \"RUL at c80 is 44, so EOL occurs at c124. Replacing at cycle 120 balances maximum utilization with safety margin.\", \"source\": \"ground_truth.json:B0005\"}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Use provided RUL fact\", \"Formula calculation from maintenance_manual.md\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"given RUL=44 is authoritative\", \"EOL cycle = 80 + 44 = 124\", \"target replacement should be before/near EOL with safety margin\"], \"expected_failure_modes\": [\"re-predicting RUL despite a provided RUL fact\", \"optimizing for immediate replacement only\", \"missing the EOL cycle calculation\"]}", "known_issue": "The 'cost-optimal' range is not derivable: no cost model is given."}
{"id": "bess_cost_002", "subset": "lithium_ion", "task_category": "Cost-Benefit Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_Fleet", "question": "For batteries B0005, B0006, B0007, B0018 at cycle 80: with a $1500 maintenance budget and $450 cost per replacement, rank batteries by replacement priority and report the remaining budget after allocation. Use RUL as the urgency signal.", "question_operator_style": "At around cycle 80, our four cell fleet needs maintenance but the budget will not cover everything. Which ones should we prioritize?", "required_tools": ["rul_predictor_ttm_finetuned", "rul_predictor_lstm"], "procedure": "Fleet-level capital allocation.", "ground_truth": "{\"answer_type\": \"list_categorical\", \"remaining_budget\": 150, \"source\": \"Fleet ranking at [0,80]\", \"swap_priority\": [\"B0018\", \"B0006\", \"B0005\"]}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Batch RUL rank\", \"Budget constraint mapping\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"urgency follows shortest RUL first\", \"budget supports three replacements at $450 each\", \"remaining budget is $150\"], \"expected_failure_modes\": [\"ranking by battery ID instead of urgency\", \"forgetting the budget constraint\", \"replacing all four cells despite limited funds\"]}", "known_issue": ""}
{"id": "bess_safety_001", "subset": "lithium_ion", "task_category": "Safety/Policy Evaluation", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0006", "question": "Evaluate B0006 at cycle 150. Given its 58% SOH, does it meet the safety policy for continued high-power grid usage?", "question_operator_style": "Battery 6 has reached about cycle 150. Does it still meet the IEC safety standard for continued use?", "required_tools": ["capacity_soh_calculator"], "procedure": "Compliance auditing.", "ground_truth": "{\"answer_type\": \"reasoning_explanation\", \"compliance\": \"fail\", \"reason\": \"SOH 58% is below the IEC 62133 high-risk threshold for grid BESS applications.\", \"source\": \"ground_truth.json:B0006\"}", "metadata": "{\"cognitive_category\": \"Understanding\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"SOH check\", \"Standard threshold comparison\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"compliance verdict is fail/non-compliant\", \"SOH below policy threshold drives decision\", \"grid high-power use is the policy context\"], \"expected_failure_modes\": [\"treating fail/non-compliant synonyms as different\", \"ignoring current SOH at cycle 150\", \"answering stage without compliance verdict\"]}", "known_issue": "Stated SOH 58% vs 63.23% in the data; IEC 62133 defines no grid-BESS SOH threshold."}
{"id": "bess_safety_002", "subset": "lithium_ion", "task_category": "Safety/Policy Evaluation", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0007", "question": "Evaluate B0007's shipping eligibility under UN 38.3 aviation policy. Use B0007's earliest available reading (cycle 0) as the reference baseline state - UN 38.3 requires assessment of cells in their factory-fresh condition, not aged state. Apply thresholds: SOH >= 70%, no critical fault flags, max temperature <= 45 C. Report pass/fail with reasoning.", "question_operator_style": "We need to ship Battery 7 by air. Using its factory fresh first cycle condition, does it pass the standard aviation safety policy?", "required_tools": ["capacity_soh_calculator", "anomaly_detector"], "procedure": "Transportation safety compliance.", "ground_truth": "{\"answer_type\": \"reasoning_explanation\", \"eligibility\": \"pass\", \"logic\": \"B0007 is healthy (>80% SOH) and has no structural or thermal runway anomalies, meeting UN 38.3 shipment safety requirements.\", \"source\": \"ground_truth.json:B0007\"}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Health check to ensure 'non-damaged' status\"], \"difficulty\": \"medium\", \"disambiguation_criteria\": [\"use cycle-0 baseline state\", \"SOH >= 70 and no critical faults pass\", \"UN 38.3 shipping is distinct from aged operation\"], \"expected_failure_modes\": [\"evaluating aged current state instead of shipment baseline\", \"confusing UN 38.3 eligibility with grid-use policy\", \"omitting pass/fail verdict\"]}", "known_issue": "SOH/temperature limits are attributed to UN 38.3, which defines type tests, not these thresholds (the verdict is correct under the stated limits)."}
{"id": "bess_safety_003", "subset": "lithium_ion", "task_category": "Safety/Policy Evaluation", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_Fleet", "question": "Summarize the safety risk of the 4-battery fleet at cycle 80. Which battery poses the highest thermal runaway risk?", "question_operator_style": "At around cycle 80, pull together a safety read on our four cell fleet.", "required_tools": ["generate_health_report"], "procedure": "Policy drafting and risk communication.", "ground_truth": "{\"answer_type\": \"reasoning_explanation\", \"highest_risk_cell\": \"B0018\", \"justification\": \"B0018 has the lowest RUL (16) and shortest discharge duration, suggesting highest internal resistance and heat generation potential.\", \"source\": \"Fleet aggregate RUL\"}", "metadata": "{\"cognitive_category\": \"Decision-Making\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Aggregation of all diagnostic signals\"], \"difficulty\": \"hard\", \"disambiguation_criteria\": [\"highest-risk cell must be B0018\", \"reasoning should cite RUL, discharge duration, or resistance/heat\", \"fleet answer must compare all cells\"], \"expected_failure_modes\": [\"choosing highest temperature only rather than fleet risk\", \"not identifying a single highest-risk cell\", \"omitting RUL/duration/resistance rationale\"]}", "known_issue": "The justification ('highest heat generation') conflicts with B0018 having the lowest maximum temperature at c80."}
{"id": "bess_health_009", "subset": "lithium_ion", "task_category": "Health Analysis", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Retrieve battery B0005's discharge capacity at cycle 0 and report its initial state of health (SOH) percentage using the benchmark rated capacity 1.86 Ah.", "question_operator_style": "At the first recorded cycle of Battery 5, what was its capacity?", "required_tools": ["fetch_cycle_summary", "capacity_soh_calculator"], "procedure": "Trivial baseline retrieval to establish a healthy starting point.", "ground_truth": "{\"answer_type\": \"numeric\", \"initial_capacity\": 1.856, \"initial_soh_pct\": 99.8, \"source\": \"data/processed/B0005_cycles.json, data/labels/ground_truth.json\", \"stage\": \"HEALTHY\"}", "metadata": "{\"cognitive_category\": \"Understanding\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Fetch cycle 0 capacity\", \"Compute SOH with the benchmark rated capacity\"], \"difficulty\": \"easy\", \"disambiguation_criteria\": [\"SOH = capacity / rated_capacity * 100\", \"B0005 benchmark rated capacity is 1.86 Ah\", \"cycle 0 is the first observed discharge cycle\"], \"expected_failure_modes\": [\"confusing rated capacity with measured capacity\", \"reporting raw Ah without SOH percentage\", \"using a later cycle instead of cycle 0\"]}", "known_issue": ""}
{"id": "bess_fault_005", "subset": "lithium_ion", "task_category": "Fault Classification", "asset_class": "Lithium-ion battery cell", "source_dataset": "NASA_PCoE_B0005", "question": "Run anomaly_detector on B0005 cycles 0-5 with parameters rolling_window=10, z_score_threshold=3.0, min_persistence=1. Report whether any anomalies are detected and the count.", "question_operator_style": "In the first five cycles of Battery 5, did anything unusual show up?", "required_tools": ["anomaly_detector"], "procedure": "Trivial baseline anomaly check on healthy cells.", "ground_truth": "{\"anomaly_count\": 0, \"answer_type\": \"boolean_count\", \"has_anomalies\": false, \"source\": \"ground_truth.json:B0005:anomalies cycle 0-5\"}", "metadata": "{\"cognitive_category\": \"Perception\", \"data_provision_mode\": \"data_discovery\", \"dependency_analysis\": [\"Slice cycles 0-5\", \"Run anomaly_detector\"], \"difficulty\": \"easy\", \"disambiguation_criteria\": [\"B0005 is healthy in cycles 0-5\", \"z_score_threshold=3.0 gives zero events in the baseline slice\", \"has_anomalies must agree with anomaly_count\"], \"expected_failure_modes\": [\"false positive on small early-life variance\", \"using too-low a z_score threshold that flags baseline data\", \"skipping the 0-5 cycle slice\"]}", "known_issue": ""}