Fault Diagnosis of Electric City Bus High-Voltage Load System Based on Multidomain Sparse Representation

This article presents a system-level fault diagnosis scheme for the high-voltage load system of electric city bus (ECB). First, a predesigned excitation signal generated by the battery system is injected into the high-voltage load system during parking, and the response signal is captured by high-speed data acquisition device. Then, time domain features, frequency domain features, and time-frequency domain features are extracted from the response signal, in addition, novel geometry features are also extracted from the response signal, which are composed of shape features (SF), dynamic performance (DP) indices, and frequency spectrum (FS) features. Further, in time domain, frequency domain, time-frequency domain, and geometry feature domain, the above features extracted from labeled samples are utilized to construct dictionaries, and sparse representation are conducted for testing samples to obtain sparse vectors. Finally, based on AdaBoost, the sparse vectors obtained from the four domains are fused, and the fault diagnosis is realized by analyzing the nonzero elements distribution of the fused sparse vector. Validations for the proposed method are conducted based on datasets obtained from AMESim simulation, ECB test rig, and real ECB, the diagnosis accuracy are 98.33%, 94.00%, and 92.50%, respectively.