Machine Learning-Based Vibration Analysis for Multi-Class Damage Detection in Quadrotor UAVs

Abstract

Quadrotor unmanned aerial vehicles (UAVs) are increasingly deployed in industrial and agricultural operations, where structural integrity and flight stability are critical. Minor mechanical defects, such as small propeller bends and chips, can cause subtle vibration irregularities that are difficult to detect using conventional inspection techniques. This study investigates the application of supervised machine learning to classify vibration signals collected from a custom low-cost acquisition system for the early identification of structural damage. Vibration data were acquired under normal conditions and for multiple damage cases, stored in Excel format, and processed using time-domain feature extraction methods. Four classifiers, namely K-Nearest Neighbours, Support Vector Machine, Decision Tree, and Random Forest, were trained and evaluated to differentiate between healthy, bent, and chipped propeller states. The models achieved pooled classification accuracies ranging from 80.0% to 93.3%, with Random Forest achieving the highest pooled accuracy of 93.3%. These findings support the feasibility of machine-learning-based vibration analysis as a proof-of-concept approach for pre-flight UAV fault screening using low-cost sensing hardware.