Metastructure-enhanced single-sensor method for improving UAV propeller fault localization

Propeller fault localization in multirotor UAVs is essential for flight safety, especially under limited sensor resources and complex environmental conditions. However, traditional single-sensor strategies often fail to distinguish fault-induced vibration features that are highly similar due to the structural symmetry of multirotor UAVs, thereby limiting fault localization accuracy. To address this limitation, this study proposes a single-sensor propeller fault localization method enhanced by metastructure racks. These racks are optimally designed using an effective mass random distribution strategy to modulate vibration transmission and improve feature distinguishability. Theoretical analysis, numerical simulations, and comparative experiments confirm that the metastructure racks substantially reduce the cross-correlation of vibration responses. The t-SNE visualization further presents good distinguishability in signal feature space. In single-sensor fault localization experiments involving both single-type and multi-type damage datasets, the Meta-group equipped with metastructure racks achieves average localization accuracies exceeding 98% in both single- and dual-propeller damage scenarios. These results are obtained with a lightweight Single-Layer Perceptron model requiring approximately 30 to 60 s of training and represent improvements of up to 23.95% over the Ctrl-group. Furthermore, even under data sparsity or noise interference, the proposed method maintains at least 90% localization accuracy, demonstrating excellent robustness and adaptability. This study offers a promising structure–algorithm co-design paradigm for achieving high-precision, low-power, and interference-resilient UAV fault localization.