Frequency-domain diverse modulation for single-channel UAV propeller damage diagnosis

Multirotor UAVs are vulnerable to propeller damage, which can seriously compromise flight safety. However, conventional fault diagnosis methods still face a trade-off between lightweight sensing and high-precision diagnosis. This research proposes a frequency-domain diverse modulation (FDM) strategy for single-channel UAV propeller damage localization and identification. The core idea is to transfer damage-related enhancement of vibration-feature differentiation from the algorithmic back-end to the sensing front-end through an engineering frame with frequency-domain modulation capability. Theoretical analysis reveals that the proposed strategy is governed by the non-uniform distribution of frequency-dependent effective mass and the diversified modulation of local resonance frequencies. Guided by this mechanism, an engineering frame incorporating L-shaped resonators is designed, and an optimization algorithm is employed to maximize vibration transmission differentiation among UAV arms under constraints. Experiments show that the average correlation coefficient among vibration transmissions decreases from 0.54 to 0.10, confirming effective frequency-domain decorrelation. With only a single-channel vibration response and a lightweight single-layer perceptron, the proposed strategy achieves localization and identification accuracies above 97%, with the maximum exceeding 99%. The model size remains below 0.2 MB, with a single-sample inference time below 0.1 ms. Moreover, the diagnostic accuracy remains above 90% under noise interference or limited-data conditions. Real hovering flight validation further achieves localization and identification accuracies of 92.5% and 96.2%, respectively. Aforementioned experiments confirm the effectiveness of the proposed FDM strategy and its engineering feasibility in balancing lightweight UAV sensing with high-precision diagnosis, showing promising potential for engineering applications.