SunnyShaurya1981 commited on
Commit
99df695
·
verified ·
1 Parent(s): 268ba2c

Add app.py for Streamlit app deployment

Browse files
Files changed (1) hide show
  1. app.py +102 -1
app.py CHANGED
@@ -1 +1,102 @@
1
-
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ import pandas as pd
3
+ import joblib
4
+ from huggingface_hub import hf_hub_download
5
+
6
+ # ---------------------------------------
7
+ # CONFIG
8
+ # ---------------------------------------
9
+ MODEL_REPO_ID = "SunnyShaurya1981/engine-predictive-maintenance-model"
10
+
11
+ FILES = {
12
+ "model": "random_forest_model.joblib",
13
+ # No scaler, cat_cols, num_cols, processed_cols for this model as it's a simple RF without complex preprocessing artifacts
14
+ # If the chosen model requires them, these paths should be uncommented and defined.
15
+ }
16
+
17
+ # ---------------------------------------
18
+ # LOAD ARTIFACTS
19
+ # ---------------------------------------
20
+ @st.cache_resource
21
+ def load_artifacts():
22
+ model = joblib.load(hf_hub_download(MODEL_REPO_ID, FILES["model"]))
23
+ # Uncomment and load other artifacts if they were part of your model pipeline
24
+ # scaler = joblib.load(hf_hub_download(MODEL_REPO_ID, FILES["scaler"]))
25
+ # categorical_cols = joblib.load(hf_hub_download(MODEL_REPO_ID, FILES["cat_cols"]))
26
+ # numerical_cols = joblib.load(hf_hub_download(MODEL_REPO_ID, FILES["num_cols"]))
27
+ # processed_columns = joblib.load(hf_hub_download(MODEL_REPO_ID, FILES["processed_cols"]))
28
+ return model # , scaler, categorical_cols, numerical_cols, processed_columns
29
+
30
+
31
+ model = load_artifacts() # Unpack other artifacts if returned
32
+
33
+ # ---------------------------------------
34
+ # STREAMLIT UI
35
+ # ---------------------------------------
36
+ st.set_page_config(page_title="Engine Predictive Maintenance", layout="centered")
37
+ st.title("⚙️ Engine Predictive Maintenance")
38
+ st.write("Predict if an engine is in a healthy or unhealthy condition based on sensor readings.")
39
+
40
+ with st.form("prediction_form"):
41
+ col1, col2 = st.columns(2)
42
+
43
+ with col1:
44
+ engine_rpm = st.number_input("Engine RPM", min_value=0, max_value=3000, value=750)
45
+ lub_oil_pressure = st.number_input("Lub Oil Pressure (MPa)", min_value=0.0, max_value=10.0, value=3.0, format="%.6f")
46
+ fuel_pressure = st.number_input("Fuel Pressure (MPa)", min_value=0.0, max_value=30.0, value=6.0, format="%.6f")
47
+
48
+ with col2:
49
+ coolant_pressure = st.number_input("Coolant Pressure (MPa)", min_value=0.0, max_value=10.0, value=2.5, format="%.6f")
50
+ lub_oil_temp = st.number_input("Lub Oil Temperature (C)", min_value=50.0, max_value=150.0, value=75.0, format="%.6f")
51
+ coolant_temp = st.number_input("Coolant Temperature (C)", min_value=50.0, max_value=150.0, value=80.0, format="%.6f")
52
+
53
+ submitted = st.form_submit_button("Predict Engine Condition")
54
+
55
+ # ---------------------------------------
56
+ # PREDICTION
57
+ # ---------------------------------------
58
+ if submitted:
59
+ # Create a DataFrame from the input values
60
+ input_data = pd.DataFrame([{
61
+ 'Engine rpm': engine_rpm,
62
+ 'Lub oil pressure': lub_oil_pressure,
63
+ 'Fuel pressure': fuel_pressure,
64
+ 'Coolant pressure': coolant_pressure,
65
+ 'lub oil temp': lub_oil_temp,
66
+ 'Coolant temp': coolant_temp
67
+ }])
68
+
69
+ # Note: For this specific Random Forest model, if no scaling or one-hot encoding was applied in training,
70
+ # then these steps are not needed here. Otherwise, include them based on your model_build_eval.py.
71
+ # Example of how you would apply preprocessing if needed:
72
+ # if scaler is not None:
73
+ # input_data[numerical_cols] = scaler.transform(input_data[numerical_cols])
74
+ # if categorical_cols is not None:
75
+ # input_data_encoded = pd.get_dummies(input_data, columns=categorical_cols, drop_first=True)
76
+ # input_data_processed = input_data_encoded.reindex(columns=processed_columns, fill_value=0)
77
+ # else:
78
+ # input_data_processed = input_data
79
+
80
+ # In this case, assuming the RF model takes raw numerical input directly (after data preparation steps)
81
+ # If your model_build_eval.py involved scaling, you would need to load and apply the scaler here.
82
+ # For this simplified Random Forest, we assume direct input without further preprocessing artifacts.
83
+
84
+ try:
85
+ prediction = model.predict(input_data)[0]
86
+ prediction_proba = model.predict_proba(input_data)[0]
87
+
88
+ st.subheader("Prediction Result")
89
+ if prediction == 0:
90
+ st.success("✅ The engine is predicted to be in **Healthy** condition.")
91
+ st.metric("Confidence (Healthy)", f"{prediction_proba[0]*100:.2f}%")
92
+ st.metric("Confidence (Unhealthy)", f"{prediction_proba[1]*100:.2f}%")
93
+ else:
94
+ st.warning("⚠️ The engine is predicted to be in **Unhealthy** condition.")
95
+ st.metric("Confidence (Unhealthy)", f"{prediction_proba[1]*100:.2f}%")
96
+ st.metric("Confidence (Healthy)", f"{prediction_proba[0]*100:.2f}%")
97
+
98
+ except Exception as e:
99
+ st.error(f"An error occurred during prediction: {e}")
100
+ st.write("Please ensure all model artifacts are correctly loaded and input features match the model's expectations.")
101
+
102
+ st.caption("⚠️ ML-based prediction for decision support only.")