Predictive Maintenance Platform

This project is motivated by the emergence of Industry 4.0 (I4.0), where smart systems, machine learning (ML), and predictive maintenance techniques are increasingly adopted to monitor and manage the health of industrial equipment, read more.

Thanks to modern information technologies and communication networks, it is now possible to collect large volumes of operational and process data from industrial assets. These data enable automated fault detection, diagnostics, and Remaining Useful Life (RUL) estimation, with the goal of:

• Minimizing unplanned downtime
• Increasing equipment utilization
• Extending component lifetimes

Scope of This Work

The objective of this project is to solve a regression problem: predicting the Remaining Useful Life (RUL) of turbofan engines using a Long Short-Term Memory (LSTM) neural network.

The model is trained and evaluated using the NASA C-MAPSS Turbofan Engine Degradation Dataset, a widely used benchmark in predictive maintenance research.

Key Features

• Deep learning–based RUL prediction using LSTM networks
• FastAPI backend for real-time inference
• Dockerized deployment for reproducibility
• Prometheus metrics for monitoring predictions and errors
• Grafana dashboards for visualization and alerting
• Modular codebase following MLOps-friendly structure

Runtime View

sequenceDiagram
    participant Client
    participant API as FastAPI Service
    participant Model as LSTM Model
    participant Prometheus
    participant Grafana

    Note over Client,API: Prediction Request
    Client->>+API: POST /predict
    API->>+Model: RUL inference
    Model-->>-API: RUL value
    API-->>-Client: Response
    
    API->>Prometheus: expose metrics
    Prometheus->>Grafana: scrape & visualize

Loading

Example

Usage

1. Build and run services

docker-compose up -d

or for running just the api locally (development)

uvicorn src.api.main:app --reload

2. Access services

• API Docs: http://localhost:8000/docs
• Prometheus: http://localhost:9090
• Grafana: http://localhost:3000

(Default Grafana credentials: admin / admin)

A Grafana dashboard is used to visualize prediction volume, error rates, and the latest RUL estimates.

Dataset

This project is based on the dataset provided by Saxena et. al (2008).The NASA Turbofan Engine Degradation Simulation:

Saxena, A., Goebel, K., Simon, D., & Eklund, N. (2008). Damage Propagation Modeling for Aircraft Engine Run-to-Failure Simulation.

---

Notes for Reviewers

• This project focuses on backend, ML inference, and monitoring, not frontend development
• The monitoring stack (Prometheus + Grafana) is included for demonstration and observability, not production-scale deployment
• The architecture is intentionally designed to be simple, interpretable, and extensible