AI4Manufacturing/SEUG-perception

SEU gearset — perception representations (visual grounding)

The same SEU gearset windows rendered as perception images — one HF config per representation. Unlike the SEUG (modulation-spectrum) repo, these are not for compute-then-check CoT (reasoning stays empty).

Configs

load_dataset("AI4Manufacturing/SEUG-perception", "spectrogram")

config	records	splits
spectrogram	391	{'train': 311, 'test': 80}
scalogram	391	{'train': 311, 'test': 80}
waveform	391	{'train': 311, 'test': 80}
reshaped	391	{'train': 311, 'test': 80}

Schema (7-field unified record)

field	meaning
query	the classification instruction (one of 30 deterministic paraphrases per representation)
image	the rendered signal image (bytes embedded)
annot	gold gear condition: health / chipped / miss / root / surface
reasoning	chain-of-thought (empty here; filled in the -annotated sibling)
cate / task	C / T-C1 (signal fault classification)
metadata	JSON string: representation, condition, file, window_idx, start_sample, channel, fs, fr_nominal, fr_used, fr_source, planetary, gear_lines, computed_verdict, computed_score, integer_score, family_obs, evidence_tier, image_sha256, split

Provenance & reproducibility

Generated deterministically by forge_agent/examples/seu/convert.py (a990b2ef69) → forge_model/SEUG/convert_seug.py (8892ffb2db); see provenance.json.

Gold = filenames (the files' internal Title fields are provably stale operator templates); the five gear conditions are physically implanted on the stage-1 sun gear of the DDS planetary gearbox [evidenced: every fault class modulates the mesh at the sun-fault order 5/6·fr] and are steady-state, so every window carries its file's condition. The gear-train constants (2-stage planetary 20/40×4/100 → 24/30×3/84, 27:1) were derived from this dataset's own spectra and validated against the manufacturer's published 27:1 ratio — tooth counts are not published anywhere. Confidence grades: stage 1 high (carrier line at exactly fr/6, sun-fault line at 5/6·fr, GMF₁ = 16.665 orders with dominant 2×/4× harmonics, valid 4-planet assembly), stage 2 moderate (GMF₂ = 3.111 orders at both speeds; sole assembly-valid candidate). Full chain + grades in provenance.json (planetary_derivation).

Caveats

• The evidence tier is BINARY. The label-independent detector (mesh_modulation) attests that a gear fault is visibly present (sun-fault-family modulation beating integer-order modulation) — it cannot name which of the four implanted subtypes, because all four share the same modulation signature. confirmed = binary agreement with the gold; subtype discrimination is learnable from these signals (deep-learning literature) but not physics-nameable.
• Conflict rule (binary): weak records are dropped only when the detector claims a fault on a health record; a quiet detector on a fault record is benign non-detection (kept in perception).
• Split is time-stratified per file (first 80% of each recording → train, last 20% → test): the rig has ONE physical specimen per (condition, speed-load) cell, so no unit-wise split exists. Cross-specimen generalization cannot be evaluated from this dataset.
• Two operating conditions (20 Hz-0 V, 30 Hz-2 V motor speed-load) are both included with condition metadata.

Source & license

Source: SEU gearbox dataset — Southeast University, Drivetrain Dynamics Simulator (SpectraQuest/Sumyoung DDS). Authors' research release: github.com/cathysiyu/Mechanical-datasets (no LICENSE file — cite the paper): S. Shao, S. McAleer, R. Yan, P. Baldi, IEEE Trans. Industrial Informatics 15(4):2446–2455, 2019 (DOI 10.1109/TII.2018.2864759). fs = 5120 Hz [evidenced: DAQ header × 2.56 convention + shaft combs at nominal in both conditions]. The release's dataset/ folder (CWRU fan-end copies) is excluded — CWRU is published separately from its original source.