Huang et al.'s framework diagnoses drone propeller damage using only IMU data

A new paper from researchers Shinan Huang, Jingxi Zhu, and Fotis Kopsaftopoulos presents a unified statistical framework for diagnosing propeller damage on multicopters. The key innovation is that it relies solely on standard inertial measurement unit (IMU) data—specifically three-axis acceleration and angular velocity from six channels—without requiring additional sensors or specialized hardware. The framework uses functionally pooled autoregressive (FP-AR) models to capture system dynamics across varying operating conditions from short data records. Damage detection is achieved through statistical testing of prediction residuals, while motor-level identification uses model selection. For damage magnitude estimation, the framework employs a Bayesian inference scheme that yields posterior estimates and explicit uncertainty bounds, offering a statistically rigorous alternative to conventional batch methods.

The framework was experimentally evaluated on a custom-built hexacopter performing figure-eight trajectories outdoors under ambient wind disturbances. Results showed consistent cross-flight performance without case-specific retuning, and the Bayesian quantification approach produced more stable estimates and clearer uncertainty characterization compared to traditional batch methods. While not a consumer product, this research has immediate applications in drone structural health monitoring. By using existing IMU sensors already present on most multicopters, it enables cost-effective damage diagnosis without hardware modifications, potentially improving safety and maintenance scheduling in commercial and industrial drone operations.