How m-shas/railway-track-fault-detection-and-predictive-maintenance works
The product is a railway-track fault detection and predictive-maintenance system. Its main workflow loads PLC and vibration sensor data, preprocesses it, runs anomaly and fault/RUL models, then produces maintenance alerts and dashboard outputs. The README also documents optional sequence models and SHAP explainability. The pipeline excerpt confirms alert and dashboard-data generation; other orchestration links reflect the README’s phase description, while uncertain wiring is omitted.
Main entry points: pipeline.py Pipeline Orchestrator · preprocess.py Preprocessing · classifier.py Fault Classifier · alerts.py Alert & Action Rules
Components 16
Sensor Inputs
- Preprocessing
src/preprocess.py
Data & Models
- Pipeline Orchestrator
src/pipeline.py - Anomaly Detection
src/anomaly_model.py - RUL Predictor
src/rul_model.py - Fault Classifier
src/classifier.py - BiLSTM RUL
src/lstm_rul_model.py - CNN-LSTM Classifier
src/cnn_lstm_model.py - SHAP Explainability
src/xai_explainer.py - Data Relabeling
src/relabel_data.py
Maintenance Decisions
- Alert & Action Rules
src/alerts.py - Alert Log
src/alerts.py
Dashboards & Outputs
- HTML Dashboard
src/pipeline.py - Dashboard Data
src/pipeline.py - Streamlit Dashboard
app.py
Other components
- PLC & Vibration Data
- Maintenance User
Connections 17
- PLC & Vibration Data → Pipeline Orchestrator: starts run
- Pipeline Orchestrator → Preprocessing: runs preprocessing
- Preprocessing → Pipeline Orchestrator: returns features
- Pipeline Orchestrator → Anomaly Detection: trains detector
- Pipeline Orchestrator → RUL Predictor: trains RUL model
- Pipeline Orchestrator → Fault Classifier: trains classifier
- Pipeline Orchestrator → BiLSTM RUL: trains optional model
- Pipeline Orchestrator → CNN-LSTM Classifier: trains optional model
- Pipeline Orchestrator → Alert & Action Rules: generates alerts
- Alert & Action Rules → Pipeline Orchestrator: returns alert log
- Pipeline Orchestrator → Alert Log: exports CSV
- Pipeline Orchestrator → Dashboard Data: builds and saves
- Pipeline Orchestrator → HTML Dashboard: builds dashboard
- Dashboard Data → Streamlit Dashboard: displayed by
- Maintenance User → HTML Dashboard: views dashboard
- Maintenance User → Streamlit Dashboard: uses dashboard
- Fault Classifier → SHAP Explainability: explains predictions