Advanced Fault Diagnosis of DC-DC Power Modules Using Wavelet Packet Decomposition and Few-Shot Learning

In order to address the problem of limited samples in the fault diagnosis of DC-DC (Direct Current to Direct Current, DC-DC) power modules, this paper proposes a method combining wavelet analysis and Few-Shot Learning. Wavelet packet decomposition is used to extract multifrequency features from output voltage signals, using Euclidean distance (L2 metric) to measure similarity between samples. The experimental results show that the proposed method can efficiently identify and classify faults under various degraded states of the power devices. Compared to the baseline model, the proposed method has been shown to improve fault detection accuracy by 18.12% using only six samples per class and by 6.37% using 100 samples per class. This approach is well-suited for real-world applications where data is limited, as it not only addresses the challenge of scarce data but also enhances the robustness and reliability of the fault diagnosis system.