Cavitation intensity recognition for high-speed axial piston pumps using 1-D convolutional neural networks with multi-channel inputs of vibration signals

Abstract

Raising rotational speed is an effective way to improve power density of axial piston pumps, but high rotational speed tends to cause undesirable cavitation in the pump. Although some machine learning methods have been successfully applied to detect the cavitation with high accuracy, these conventional methods suffer from the drawback of time-consuming and experience-dependent manual feature extraction. In this paper, a new model based on 1-D convolutional neural network (CNN) is proposed to recognize the cavitation intensity of axial piston pumps. To improve the recognition accuracy under noisy environment, the 1-D CNN receives multi-channel vibration data instead of single-channel data. The experimental results show that the proposed anti-noise 1-D CNN model with multi-channel inputs can achieve 15% higher recognition accuracy than its counterpart with single-channel input on a testing set with SNR = 5 dB.