⚙️ Physics-Guided ML Fault Detection for Induction Motors

A System-Level Digital Twin + Hardware-in-the-Loop (HIL) project that uses physics-informed machine learning to detect outer-race bearing faults in induction motors in real time.

This repository contains the complete production-ready source code, integrating structural vibration simulation, SVM machine learning intelligence, and a live serial bridge to an physical Arduino LCD dashboard.

---

📖 Overview

Bearing failures account for nearly 40% of induction motor breakdowns. This project builds a fault detection system that:

1. Generates synthetic vibration data simulating mechanical fault characteristics.
2. Extracts critical spectral features, primarily the Ball-Pass Frequency (Outer Race) — f_BPFO.
3. Trains a 3-Dimensional Support Vector Machine (SVM) classifier using Statistical/Signal features (RMS, Kurtosis, BPFO Peaks).
4. Bridges MATLAB & Hardware connecting the ML intelligence directly to an Arduino-powered physical LCD for real-time Remaining Useful Life (RUL) forecasting.

---

📂 Project Structure

File / Folder	Role in System
generate_motor_data.m	Data Engine: Simulates 12kHz motor vibrations and physics conditions; builds root .mat Datasets.
train_svm_model.m	AI Training: Reads generated data, extracts the 3D IEEE features, and trains mySVM.mat using 5-Fold Validation.
matlab_simulation_IM.slx	Digital Twin: A Simulink canvas modeling the physical block systems (if visual modeling is preferred).
Arduino_Hardware_Integration/	HIL Connectors: Houses the .ino Arduino LCD Sketch and the vital run_live_svm_bridge.m real-time simulation script.
assets/	Contains reference charts of the expected output vibration mappings.

---

🚀 Getting Started

📋 Prerequisites

• MATLAB R2023b (or later) with Statistics & Machine Learning Toolbox.
• Arduino IDE with a compatible Microcontroller connected via USB (Any standard board).
• Simple 16x2 I2C LCD connected to the Arduino (for the HIL Dashboard display).

---

Step 1: Pre-Train the Machine Learning Model

The simulation requires the SVM model to be compiled locally first.

1. Open MATLAB and navigate to this repository's root folder.
2. Open and hit Run on generate_motor_data.m (this simulates 100 datasets of physics data).
3. Once completed, Open and hit Run on train_svm_model.m.

   This extracts the IEEE features and compiles a highly robust mySVM.mat brain.

Step 2: Flash the Physical Hardware

Before starting the live diagnostic bridge, the peripheral must be ready.

1. Navigate into the Arduino_Hardware_Integration/Phase4_LCD_HIL_I2C folder.
2. Open Phase4_LCD_HIL_I2C.ino using the standard Arduino IDE.
3. Check your COM Port, compile/upload it to your board.
4. Open the Arduino Serial Monitor and wait until it successfully prints: OK.

   Leave the Serial Monitor open or the Arduino plugged in.

Step 3: Launch the Live Dashboard (HIL)

1. Head back into MATLAB and navigate into the Arduino_Hardware_Integration/Phase4_LCD_HIL_I2C/ folder.
2. Open the run_live_svm_bridge.m script.
3. Ensure the arduinoPort variable matches your exact COM Port (e.g., "COM3").
4. Click RUN!

   A live Diagnostic Dashboard UI will open inside MATLAB charting the frequency/time domains while simultaneously broadcasting predictive statuses & Remaining Useful Life (RUL) strings dynamically to your Arduino's LCD Screen!

---

Created for Academic/Industrial Demonstration built upon realistic CWRU-equivalent kinematics.