Diagnosis and classification of gear hydraulic pump defects using vibration signal processing by continuous wavelet transform and convolutional neural network

Subject Areas :

Received: 2023-06-13 Accepted : 2023-06-13 Published : 2022-08-23

Keywords: Fault Detection, Hydraulic Pump, Scalogram, CNN,

Abstract :

Hydraulic pumps are the heart of a hydraulic system, and any failure that occurs in them disrupts the performance of the hydraulic system, so it is essential to monitor the operation of hydraulic pumps to ensure that hydraulic systems are ready for operation. In this article, an intelligent method based on vibration signal processing and deep learning method was used to detect faults in gear hydraulic pump. In this research, an internal gear hydraulic pump was used. Data acquisition tests were performed on the pump at a speed of 1200 rpm using the Global Test AP 98-100 vibration sensor and the Advantech USB-4704 data acquisition set. For each situation, 100 signals were recorded from the pump, and then each signal was processed by the continuous wavelet transform method, and finally, these images were used to create a deep learning model for pump fault classification. A convolutional neural network with 13 layers was created to classify faults in the pump, and after 70 repetitions of training, the training model was optimized and achieved an accuracy of 96.67%. Evaluation of the model showed that the accuracy of this model to detect healthy, wear in the inner gear and wear in the outer gear was equal to 88.33%.