Improving accuracy of cavitation severity recognition in axial piston pumps by denoising time-frequency images
Abstract Vibration signal is a good indicator of cavitation in axial piston pumps. Some vibration-based machine learning methods have been developed for recognizing the pump cavitation. However, their fault diagnostic performance is often unsatisfactory in industrial applications due to the sensitivity of the vibration signal to noise. In this paper, we presented an intelligent method to recognize the cavitation severity of an axial piston pump under noisy environment. First, we adopted short-time Fourier transformation to convert the raw vibration data into spectrograms that acted as input images of a modified LeNet-5 convolutional neural network (CNN). Second, we proposed a denoising method for the converted spectrograms based on frequency spectrum characteristics. Finally, we verified the proposed method on the dataset from a test rig of high-speed axial piston pump. The experimental results indicate that the denoising method significantly improves the diagnostic performance of the CNN model under noisy environment. For example, the accuracy rate of the cavitation recognition increases from 0.52 to 0.92 at SNR of 4 dB by the denoising method.