Fault Diagnosis for Hydraulic Loading System of Altitude Simulation Test Facility Based on Multichannel 1D-CNN

The hydraulic loading system is the system that simulates the operation of the aircraft airborne pump by extracting the power of the aero-engine during the altitude simulation test, and its safety and stability is very important during operation. Aiming at the shortcomings of traditional fault diagnosis in feature extraction, the intelligent method based on deep learning combines feature extraction and classification to automatically isolate faults. In this paper, a method based on multi-channel one-dimensional convolutional neural network(1D-CNN) for fault diagnosis of hydraulic loading system is proposed. First, the four sequential sensor signals are preprocessed to form the input of the multi-channel 1D-CNN. Then, Multi-channel CNN extracts complex features from sensor signals and fuses features deeply. Finally, fully connected layer uses extracted features for fault classification. This paper uses indicators such as precision rate, recall rate and confusion matrix to analyze the effect of fault diagnosis. Compared with the multi-layer perceptron-based fault diagnosis method, the multi-channel 1D-CNN has higher accuracy and solves the problem of misjudging the health state as a fault. Experiments show that the method based on multi-channel 1D-CNN can effectively and accurately identify faults of the hydraulic loading system of altitude simulation test platform (HLS-ASTF).