How webex042/coffee-predictive-maintenance works
This is a simulated coffee-line predictive-maintenance prototype. Its main workflow turns sensor history—either the bundled simulation or an uploaded CSV—into 24-hour failure-risk predictions, then presents machine health, explanations, recommendations, and production-impact priorities in a Streamlit control room. Training, simulation, scoring, and operator-facing outputs are distinct stages. The graph reflects the supplied excerpts and README; some scoring internals and dashboard capabilities are documented but not fully sampled.
Main entry points: scoring.py History scoring · app.py Control room · simulate.py Sensor simulator · train_model.py Model training
Components 15
Sensor data
- CSV upload
product maintance/dashboard/app.py - Sensor simulator
product maintance/src/simulate.py - Sensor history
product maintance/data/raw/sensor_data.csv - Machine baselines
product maintance/config.py - Input validation
product maintance/src/scoring.py
Prediction
- Feature engineering
product maintance/src/feature_engineering.py - Model training
product maintance/src/train_model.py - Calibrated model
product maintance/models/model.joblib - History scoring
product maintance/src/scoring.py
Maintenance insights
- Health scoring
product maintance/src/health.py - Maintenance advice
product maintance/src/maintenance_rules.py - Impact ranking
product maintance/src/priority.py
Operator dashboard
- Control room
product maintance/dashboard/app.py - Arcade presentation
product maintance/dashboard/theme.py
Other components
- Plant operator
Connections 17
- Sensor simulator → Machine baselines: uses baselines
- Sensor simulator → Sensor history: writes history
- Plant operator → CSV upload: provides CSV
- CSV upload → Input validation: checks data
- Sensor history → History scoring: loads history
- Input validation → History scoring: validates input
- Feature engineering → History scoring: builds features
- Calibrated model → History scoring: predicts risk
- Model training → Feature engineering: uses features
- Model training → Calibrated model: saves model
- History scoring → Control room: returns scores
- Control room → History scoring: requests snapshot
- History scoring → Health scoring: computes health
- History scoring → Maintenance advice: requests advice
- History scoring → Impact ranking: ranks machines
- Control room → Arcade presentation: styles views
- Plant operator → Control room: views results