关键词: 270 V HVDC系统, 深度残差收缩网络, 线电压幅值分析, 故障诊断, 故障模块识别

Abstract:

The fault diagnosis of the airborne 270 V high-voltage direct current （HVDC） system has always been a difficult problem in the field of avionics. Therefore， a fault module identification algorithm based on deep residual contraction network （DRSN） and a fault component localization algorithm based on line voltage amplitude analysis have been proposed. Firstly， the total current of the system is collected， and the difference is standardized to obtain characteristic data. Based on the feature data， the Flatten layer is used to improve the original DRSN structure， so as to improve the algorithm recognition accuracy for fault modules. After determining the fault of the system inverter module， the fault phase is determined using the ratio of the two phase line-voltages， and then the fault device is determined using the average line-voltage model. Compared to existing methods， the proposed method only uses one current sensor and two voltage sensors to achieve system fault diagnosis， meeting the weight limitation requirements of aircraft. The experiment proves that the proposed method has good practicality in identifying fault modules and locating fault components with an accuracy of over 97%.

Key words: 270 V HVDC system, deep residual shrinkage network, line voltage amplitude analysis, fault diagnosis, fault module identification