TinyML Vibration Monitoring FYP: Predictive Maintenance Demo with ESP32
Create a TinyML vibration monitoring FYP using accelerometer data, feature extraction, simple classification, and honest model evaluation.
Rectronx
2026-07-14
Quick Summary
TinyML vibration monitoring is a strong FYP when it focuses on classifying simple machine states such as normal, loose, imbalance, or high vibration. It combines sensors, embedded systems, and machine learning without needing a huge cloud system.
Project Scope
Use an accelerometer to collect vibration data from a small DC motor or fan. Train a lightweight model or threshold-based classifier, then deploy the result to ESP32 where possible.
Keep the claim narrow: the prototype detects selected vibration patterns in a controlled demo.
Suggested Components
Use ESP32, MPU6050 or ADXL345 accelerometer, small DC motor, adjustable load, microSD card or serial logging, Edge Impulse or Python training workflow, and dashboard or display.
Build Steps
- Collect labeled vibration data for each condition.
- Clean the dataset and split training and testing data.
- Extract features such as RMS, peak, or frequency-related values.
- Train a simple model or rule-based classifier.
- Test the model on new data not used during training.
Accuracy and Safety Notes
Predictive maintenance is a serious engineering field. For FYP, say "vibration state classification" unless you have long-term failure data. A controlled prototype is still valuable when the evaluation is honest.
Common Mistakes
- Training and testing on the same samples.
- No labels or unclear class definitions.
- Claiming failure prediction without failure data.
- Mounting the sensor loosely.
- Ignoring motor speed changes.