AI-based Predictive Maintenance System for Vehicles Overview This repository contains the code and model implementation for an AI-based predictive maintenance system aimed at vehicles. The project was developed for the VW Hackathon 'i.mobileathon 4.0'. The system uses machine learning algorithms to predict potential part failures based on real-time sensor data from various vehicle components.

Problem Statement Vehicle breakdowns are costly and reduce vehicle lifespan, causing inconvenience for owners and losses for OEMs. Existing maintenance systems rely on scheduled checks, which do not consider the unique, real-time usage patterns of each vehicle. This project aims to address this issue by leveraging AI to provide predictive maintenance, minimizing unplanned downtimes and repair costs.

Proposed Solution An AI-driven system that uses real-time data from vehicle sensors (engine, transmission, brakes, suspension) to predict potential failures. By analyzing sensor data continuously, the model can notify both the vehicle owner and OEMs about upcoming maintenance needs before a failure occurs.

Features Real-time Monitoring: Continuously collects sensor data from the vehicle. AI-based Predictions: Utilizes neural networks and KNN models to predict part failures. User Notifications: Alerts users and OEMs about maintenance needs. Interactive Dashboard: Visual representation of sensor data and health status of vehicle components. Historical Data Analysis: Stores data for model training and continuous learning. Tech Stack Frontend: React, Tailwind CSS Backend: Node.js, Express Machine Learning: Python, PyTorch, scikit-learn Database: MongoDB Data Visualization: Plotly, Dash Architecture Data Collection: Real-time data from vehicle sensors is collected and stored. Data Preprocessing: The data is cleaned and standardized. Model Inference: Neural network and KNN models analyze the data to make predictions. Notification System: Alerts are sent to users based on the model predictions. Dashboard: Displays vehicle health and predictive alerts to the user. Models Used Neural Network (PyTorch): A deep learning model trained on historical sensor data. Captures complex relationships and generalizes well for different vehicle conditions. K-Nearest Neighbors (KNN) Model (scikit-learn): A simpler model used as a baseline. Classifies new data points based on similarity to previous data. Setup Instructions Clone the repository:

bash Copy code git clone https://github.com/shaurya2412/i.mobilothon.git cd i.mobilothon Install dependencies:

bash Copy code npm install pip install -r requirements.txt Run the frontend:

bash Copy code npm run dev Run the backend:

bash Copy code node server.js Train the models (Optional):

bash Copy code python model_training.py Usage Access the dashboard at http://localhost:3000. View real-time sensor data and health status. Receive predictive maintenance alerts based on the analysis. Data Sources Simulated real-time vehicle sensor data. Historical data for model training. Contributing Feel free to open issues or submit pull requests for improvements.

License This project is licensed under the MIT License. See LICENSE for more details.

You can modify or add more details if needed. Let me know if you want to include anything else!