UAV Propulsion Health Monitor

A Physics-Based Prognostics and Health Management (PHM) system for UAV propulsion systems.

This system replaces simple hard-coded thresholds with a baseline-driven, physics-based health engine. It continuously compares real-time telemetry against a calibrated "healthy" propulsion baseline to detect mechanical, thermal, and electrical degradation before catastrophic failure occurs.

Hardware Configuration

The current architecture is designed for a single-motor propulsion unit monitored by the following I2C sensors:

• 1x INA219 (or INA226) Current/Voltage Monitor:
  • Battery & ESC Monitor (0x40): Measures total pack voltage and current (motor draws ~99% of total current).
• 1x MPU6050 IMU (0x68): Mounted on the motor base to measure high-frequency vibration.

Core Concepts: Physics-Based Health Indices

Instead of raising alerts when a sensor hits an arbitrary limit, the system computes four core Health Indices based on the deviations from the calibrated baseline at the current throttle level:

1. Propeller Balance Index (PBI): Correlates excessive vibration (from the IMU) with electrical ripple (from the INA219) to detect mechanical imbalance, damaged propellers, or bent shafts.
2. Propulsion Load Index (PLI): Compares actual current draw to expected current draw. Positive PLI indicates excess drag (bearing friction, obstruction); negative PLI indicates loss of load (missing or loose propeller).
3. ESC Thermal Index (ETI): Evaluates temperature rise relative to the expected thermal load for the current power draw. Detects cooling path failures or internal resistance degradation before hard thermal limits are reached.
4. Battery Stress Index (BSI): Correlates voltage sag with current demand. Identifies aging batteries, high internal resistance, or excessive load.

Software Setup & Execution

# Install dependencies
pip3 install -r requirements.txt

# Verify I2C devices are visible
i2cdetect -y 1

# Run the primary monitor
python3 monitor.py

Dashboard

The system runs a responsive, high-frequency web dashboard accessible at: http://<raspberry-pi-ip>:5000

Calibration Procedure (Crucial Step!)

Because the PHM engine relies on detecting deviations, it must first learn what "healthy" looks like.

1. Install a perfectly balanced, undamaged propeller.
2. Open the web dashboard and navigate to the Health tab.
3. Click ▶ Start Calibration.
4. Follow the on-screen instructions: Use the throttle slider to slowly sweep from 0% up to 100%, pausing for about 5 seconds at each 10% interval.
5. The system automatically collects 3 baseline sweeps, computes the statistical normal bounds, and saves them to config/baseline.json.
6. Once calibrated, the system will begin emitting active health scores and fault predictions.

Project Structure

• core/: The core processing pipeline.
  • baseline_manager.py: Manages the throttle-indexed baseline and interpolates expected values.
  • deviation_engine.py: Computes normalized deviations of real-time telemetry vs. the baseline.
  • health_engine.py: Computes the 4 core health indices (PBI, PLI, ETI, BSI).
  • diagnostics.py: Classifies abstract conditions (e.g., rotational_imbalance, thermal_anomaly) based on index severity, complete with a debounce filter.
• web/: The Flask web server and dashboard UI.
• calibrate.py: Operator-driven calibration routine.
• monitor.py: The main entry point that wires all components together.

Extension & Development

• To run the test suite: pytest tests/ -v
• Shunt resistor configuration should not be hardcoded; it is dynamically loaded from hardware.json.
• Cross-flight health trends are logged to local SQLite databases per flight (flight_health_indices table).