Acoustic signal-based fault detection of hydraulic piston pump using a particle swarm optimization enhancement CNN. (April 2022)

• Record Type: Journal Article
• Authors: Zhu, Yong; Li, Guangpeng; Tang, Shengnan; Wang, Rui; Su, Hong; Wang, Chuan
• Is Part Of: Applied acoustics. Volume 192(2022)
• Publication Date: 2022-04
• Subjects: Hydraulic piston pump -- Fault identification -- Acoustic signal -- Convolutional neural network -- Particle swarm optimization; Acoustical engineering -- Periodicals; Periodicals; 620.2
• DOI: 10.1016/j.apacoust.2022.108718
• Languages: English
• ISSNs: 0003-682X

Abstract

Highlights: The structure of the standard LeNet model is adjusted. Model hyperparameters are automatically optimized instead of manually adjusting. PSO-LeNet model has strong stability and high identification accuracy. The optimized model can accurately identify the typical faults of piston pump. Abstract: As the heart of a fluid power system, hydraulic piston pumps are widely used in many critical applications, such as for marine, aerospace, and engineering equipment. The health status of a pump is important for the safety and reliability of the mechanical equipment. Hence, it is necessary to develop intelligent fault diagnosis for a hydraulic piston pump. In this research, the particle swarm optimization (PSO) algorithm is introduced to automatically select the hyperparameters of diagnosis model. A convolutional neural network (CNN) model optimized by PSO is constructed based on the standard LeNet. The PSO-LeNet model is applied to identify five common states of a hydraulic piston pump using an acoustic signal: normal state, swash plate wear, center spring failure, loose slipper, and slipper wear. Many typical deeper CNN models are compared and used for the verification of the performance of the proposed model, such as AlexNet, VGG11, VGG13, VGG16, and GoogleNet. Results indicate that the PSO-LeNet has the best stability and the highest identification accuracy. Thus, the proposed model has the laudable overall performance.