A Two-Step Sensor Fusion Methodology to Assess Damage on Drone Propellers by Audio and Radar Measurements

Area of Science:

• Aerospace Engineering
• Mechanical Engineering
• Artificial Intelligence

Background:

• Growing use of Unmanned Aerial Vehicles (UAVs) necessitates enhanced operational safety.
• Detecting and classifying component damage is crucial for timely intervention and accident prevention.
• Propeller damage poses a significant risk to UAV stability and flight integrity.

Purpose of the Study:

• To propose a novel two-step methodology for detecting and classifying UAV propeller damage.
• To develop a real-time, onboard system for initial damage assessment.
• To integrate a secondary system for accurate damage severity and location determination.

Main Methods:

• Onboard, real-time damage detection using acoustic emissions and edge processing with a Machine Learning (ML) classifier.
• A secondary, ground-station-based radar system for contactless vibration and frequency measurements.
• Combining acoustic and radar-based methods for a comprehensive diagnostic system.

Main Results:

• The embedded audio-based ML model achieved over 99% performance in damage classification.
• The radar-based system accurately differentiates and measures the location of propeller damage.
• The integrated system provides a time-responsive and accurate diagnosis of propeller faults.

Conclusions:

• The proposed two-step methodology effectively detects and classifies UAV propeller damage.
• The system enhances UAV safety by enabling timely and appropriate countermeasures.
• This approach significantly improves the reliability and safety of Unmanned Aerial Vehicle operations.