Modern engines emit distinct sounds when operating normally or experiencing faults. Detecting anomalies early is critical for preventive maintenance, reducing costly repairs, and ensuring safety. Traditional monitoring solutions are often expensive, bulky, or complex, making them less practical for small workshops or DIY applications.
This project leverages Google Teachable Machine and the IndusBoard Coin to build an AI-powered predictive maintenance system. The system continuously monitors engine sounds in real time, detects anomalies, and triggers alerts both visually (webpage) and physically (LED/buzzer), providing a low-cost, efficient, and automated solution for early fault detection.
Key Features
- AI-based engine sound classification (normal vs faulty) using Teachable Machine
- Real-time alerts via web interface hosted on IndusBoard Coin
- Visual indicators (webpage color change) and LED/buzzer notifications
- Easy integration with laptops or other devices for monitoring
- Uses a lightweight, browser-based TensorFlow.js model for continuous audio monitoring
Applications
- Predictive maintenance for cars, motorcycles, and industrial motors
- Automotive workshops for early fault diagnostics
- Fleet management and monitoring systems
- DIY vehicle monitoring and hobbyist projects
This low-cost, AI-powered solution demonstrates how machine learning can be applied to real-world automotive maintenance. By detecting engine anomalies early, it reduces the risk of breakdowns, improves safety, and empowers users with actionable insights—all through an accessible and compact system using the IndusBoard Coin.