Hugging Face's logo

Spaces:
deropxyz
/
AC02-ML
Sleeping

App Files Community
AC02-ML / api /README.md
deropxyz's picture
deropxyz
init
03bcd34
preview code
|
raw
history blame contribute delete
9.3 kB

Predictive Maintenance FastAPI

🚀 Quick Start

1. Install Dependencies 

pip install -r requirements.txt

2. Run API Server 

# Development mode with auto-reload
python main.py

# Or using uvicorn directly
uvicorn main:app --reload --host 0.0.0.0 --port 8000

3. Access API

4. Quick Testing (Windows) 

# Install dependencies
install_requirements.bat

# Start server
start_server.bat

# Open test interface
open_test_interface.bat

📡 API Endpoints

1. Single Prediction: POST /predict

Request Body:

{
  "machine_id": "M_L_01",
  "air_temperature": 298.1,
  "process_temperature": 308.6,
  "rotational_speed": 1551,
  "torque": 42.8,
  "tool_wear": 0,
  "type": "M"
}

Response (Enhanced with Features & Anomaly Detection):

{
  "machine_id": "M_L_01",
  "timestamp": "2024-11-26T10:30:00.123456",
  "prediction": "HEALTHY",
  "confidence": 0.9234,
  "diagnostics": {
    "primary_cause": "Normal Operation",
    "sensor_alert": "All parameters within optimal range. Health score: 100%.",
    "recommended_action": "CONTINUE NORMAL OPERATION. Next scheduled maintenance as planned.",
    "severity": "LOW"
  },
  "features": {
    "Air_temp_K": 298.1,
    "Process_temp_K": 308.6,
    "Speed_rpm": 1551,
    "Torque_Nm": 42.8,
    "Tool_wear_min": 0,
    "Tool_wear_hours": 0.0,
    "Temperature_Diff": 10.5,
    "Power_W": 6951.2,
    "Type_Encoded": 1
  },
  "anomalies": [],
  "overall_health": "🟢 EXCELLENT HEALTH - All systems operating optimally within normal parameters."
}

2. Batch Prediction (CSV Download): POST /predict/batch

Request:

  • Upload CSV file (any size)
  • Flexible CSV Format - accepts columns with:
    • UPPERCASE/lowercase/MixedCase names
    • Units in brackets: air_temperature_[k], Air Temperature [K], etc.
    • Extra columns (will be ignored)
    • Any column order

Required Columns (flexible naming):

  • air_temperature or air_temperature_[k]
  • process_temperature or process_temperature_[k]
  • rotational_speed or rotational_speed_[rpm]
  • torque or torque_[nm]
  • tool_wear or tool_wear_[min]
  • type
  • Optional: machine_id

Response:

  • CSV file download with added columns:
    • Prediction: HEALTHY or FAILURE
    • Confidence: 0.0 to 1.0
    • Primary_Cause: Root cause analysis
    • Sensor_Alert: Specific sensor readings
    • Recommended_Action: Actionable maintenance advice
    • Severity: LOW, MEDIUM, HIGH, CRITICAL
    • Processed_At: Timestamp

3. Batch Prediction (JSON for UI): POST /predict/batch/json

Request:

  • Upload CSV file (max 500 rows)
  • Same flexible format as /predict/batch

Response:

{
  "summary": {
    "total_records": 5,
    "predictions": {
      "failure": 2,
      "healthy": 3,
      "errors": 0
    },
    "failure_rate": 40.0
  },
  "results": [
    {
      "row_number": 1,
      "machine_id": "M_001",
      "prediction": "HEALTHY",
      "confidence": 0.9234,
      "diagnostics": {...},
      "features": {...},
      "anomalies": [...],
      "overall_health": "🟢 EXCELLENT HEALTH",
      "input_data": {...}
    }
  ],
  "processed_at": "2024-11-26T10:30:00"
}

🧠 Diagnostic Engine Rules

Priority 1: CRITICAL

  • Tool Wear > 200 min: Immediate tool replacement required

Priority 2: HIGH

  • Power > 9000 W or Torque > 60 Nm: Mechanical overload detected
  • ML Anomaly (Confidence > 80%): Unusual pattern detected

Priority 3: MEDIUM

  • Temperature Diff < 8.0 K at Speed > 1300 RPM: Cooling system issue
  • Temperature Diff > 15.0 K: Excessive thermal stress
  • Speed > 2500 RPM: High-speed operation risk

Priority 4: LOW

  • All parameters normal: Continue operation

🔬 Feature Engineering (Physics-Based)

The API automatically calculates:

  1. Temperature_Diff = process_temperature - air_temperature

    • Indicates thermal stress
    • Optimal range: 8-12 K

  2. Power_W = torque × rotational_speed × (2π / 60)

    • Mechanical power output
    • Safety limit: 9000 W

  3. Type_Encoded = {L: 0, M: 1, H: 2}

    • Machine type categorization

📊 Example cURL Commands

Single Prediction 

curl -X POST "http://localhost:8000/predict" \
  -H "Content-Type: application/json" \
  -d '{
    "machine_id": "M_L_01",
    "air_temperature": 298.1,
    "process_temperature": 308.6,
    "rotational_speed": 1551,
    "torque": 42.8,
    "tool_wear": 0,
    "type": "M"
  }'

Batch Prediction 

curl -X POST "http://localhost:8000/predict/batch" \
  -F "file=@test_data.csv" \
  -o predictions_output.csv

Health Check 

curl http://localhost:8000/health

Model Info 

curl http://localhost:8000/model/info

🛠️ Python Client Example 

import requests
import pandas as pd

# Single prediction
url = "http://localhost:8000/predict"
data = {
    "machine_id": "M_L_01",
    "air_temperature": 298.1,
    "process_temperature": 308.6,
    "rotational_speed": 1551,
    "torque": 42.8,
    "tool_wear": 0,
    "type": "M"
}

response = requests.post(url, json=data)
result = response.json()
print(f"Prediction: {result['prediction']}")
print(f"Confidence: {result['confidence']}")
print(f"Action: {result['diagnostics']['recommended_action']}")

# Batch prediction
url_batch = "http://localhost:8000/predict/batch"
files = {'file': open('test_data.csv', 'rb')}
response = requests.post(url_batch, files=files)

with open('predictions.csv', 'wb') as f:
    f.write(response.content)

📂 CSV Format for Batch Processing (FLEXIBLE!)

✅ Format 1: Simple (Recommended) 

machine_id,air_temperature,process_temperature,rotational_speed,torque,tool_wear,type
M_L_01,298.1,308.6,1551,42.8,0,M
M_L_02,299.2,310.1,1450,38.5,150,L
M_H_01,297.5,312.8,1800,65.2,220,H

✅ Format 2: With Units (Dataset Format) 

machine_id,air_temperature_[k],process_temperature_[k],rotational_speed_[rpm],torque_[nm],tool_wear_[min],type
M_L_01,298.1,308.6,1551,42.8,0,M
M_L_02,299.2,310.1,1450,38.5,150,L

✅ Format 3: Mixed Case with Spaces 

Machine ID,Air Temperature [K],Process Temperature [K],Rotational Speed [RPM],Torque [Nm],Tool Wear [min],TYPE
M_L_01,298.1,308.6,1551,42.8,0,M

✅ Format 4: Extra Columns (Will be Ignored) 

machine_id,timestamp,air_temperature,process_temperature,rotational_speed,torque,tool_wear,type,location,operator
M_L_01,2024-11-26,298.1,308.6,1551,42.8,0,M,Factory_A,John

All formats above are automatically handled by the API!

Output CSV (additional columns):

...,Prediction,Confidence,Primary_Cause,Sensor_Alert,Recommended_Action,Severity,Processed_At
...,HEALTHY,0.9234,Normal Operation,All parameters within optimal range...,CONTINUE NORMAL OPERATION...,LOW,2024-11-26T10:30:00
...,HEALTHY,0.8876,Normal Operation,All parameters within optimal range...,CONTINUE NORMAL OPERATION...,LOW,2024-11-26T10:30:01
...,FAILURE,0.9567,Tool End of Life,Tool wear (220 min) exceeds safety threshold...,IMMEDIATE ACTION: Replace cutting tool...,CRITICAL,2024-11-26T10:30:02

🔐 Production Deployment

Environment Variables 

export MODEL_PATH="/path/to/model_pipeline.joblib"
export API_HOST="0.0.0.0"
export API_PORT="8000"
export LOG_LEVEL="info"

Docker Deployment 

FROM python:3.11-slim

WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY . .
EXPOSE 8000

CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]

Run with Gunicorn (Production) 

pip install gunicorn
gunicorn main:app -w 4 -k uvicorn.workers.UvicornWorker --bind 0.0.0.0:8000

📈 Performance Metrics

  • Model: RandomForest Classifier (Optimized with Ensemble)
  • Features: 9 engineered features (including Tool_wear_hours)
  • F1-Score: ~75-85% (after optimization)
  • Inference Time: ~10-20ms per prediction
  • Batch Processing:
    • CSV Download: Unlimited rows
    • JSON Response: Max 500 rows (for UI display)
    • Throughput: ~100-200 records/second
  • Anomaly Detection: 7 parameter checks (Real-time)
  • Overall Health Assessment: 4 levels (EXCELLENT/GOOD/FAIR/POOR)

🐛 Troubleshooting

Model Not Loading 

# Check if model file exists
ls ../models/model_pipeline_improved.joblib
ls ../models/model_pipeline.joblib

# Verify model structure
python -c "import joblib; m = joblib.load('../models/model_pipeline.joblib'); print(m.keys())"

CORS Issues

  • Already configured for allow_origins=["*"]
  • Modify in main.py if needed for specific domains

Port Already in Use 

# Change port in main.py or use:
uvicorn main:app --port 8001

📞 API Support

For issues or questions:

  1. Check /health endpoint for model status
  2. Review logs for detailed error messages
  3. Validate input data matches schema in /docs

Built with FastAPI 🚀 | Production-Ready | ML + Physics-Based Diagnostics