🔧 Predictive Maintenance System with MLOps Pipeline

Industrial IoT Analytics • Machine Learning • MLOps • Production Deployment

A complete ML system to predict equipment failures before they happen, built with modern MLOps practices: experiment tracking, model versioning, containerized deployment, and automated workflows.

---

🎯 Highlights

• 88–92% accuracy across multiple equipment types
• <100ms latency for real-time predictions
• End-to-end pipeline: data ingestion → model training → API deployment
• Production-ready FastAPI service with health endpoints and Pydantic validation
• Fully containerized via Docker for local & cloud use

---

📊 Project Overview

Predictive maintenance allows manufacturers to schedule repairs before breakdowns, reducing downtime and costs. This project processes sensor data (temperature, torque, speed, tool wear) to predict failures using ML models and serves predictions via a web API.

---

🛠 Technical Stack

ML: scikit-learn, pandas, numpy, SMOTE

API: FastAPI, Pydantic

MLOps: MLflow (experiment tracking & model registry)

Containerization: Docker, Docker Compose

CI/CD: GitHub Actions

Deployment: AWS ECR + ECS

---

📁 Architecture

Data Pipeline → Model Training (MLflow) → FastAPI API → Docker → AWS ECS

---

📸 Screenshots

---

📈 Model Performance

Model	Accuracy	Precision	Recall	F1-Score
Random Forest	91.2%	89.4%	92.1%	90.7%
Gradient Boosting	89.8%	87.3%	91.5%	89.3%
Logistic Regression	86.4%	84.1%	88.7%	86.3%
SVM	88.1%	85.9%	90.2%	88.0%

Feature Importance:

1. Tool Wear (32%)
2. Temperature Differential (24%)
3. Torque Variance (21%)
4. Rotational Speed (15%)
5. Equipment Type (8%)

---

🚀 Quick Start

git clone https://github.com/Sa1f27/predictive-maintenance-mlops.git
cd predictive-maintenance-mlops

# Setup environment
python -m venv venv
venv\Scripts\activate  # Windows
pip install -r requirements.txt

# Train model
python run_pipeline.py --mode train

# Start API
python app.py

Docker Deployment

docker-compose up -d --build

---

🔮 Future Improvements

• Advanced feature engineering (rolling stats, lag features)
• Ensemble/stacking models
• Real-time data streaming with automated retraining
• Advanced monitoring with dashboards

---