Title: Wear defect detection of hydraulic pump using a hybrid method of VGG and LSTM

Authors: Shengnan Tang; Yixuan Jiang; Hong Su; Kian Meng Lim; Zhijian Zheng; Yong Zhu

Addresses: Institute of Advanced Manufacturing and Modern Equipment Technology, School of Mechanical Engineering, Jiangsu University, Zhenjiang, 212013, China; State Key Laboratory of Mechanical Transmission for Advanced Equipment, Chongqing University, Chongqing, 400044, China ' School of Mechanical Engineering, Jiangsu University, Zhenjiang, 212013, China ' National Research Center of Pumps, Jiangsu University, Zhenjiang, 212013, China ' Department of Mechanical Engineering, National University of Singapore, 119077, Singapore ' Ningbo Fiber Inspection Institute, Ningbo Institute for Product and Food Quality Inspection, Ningbo, 315000, China ' National Research Center of Pumps, Jiangsu University, Zhenjiang, 212013, China

Abstract: Hydraulic pump is an important power element among the core components of the hydraulic transmission system in mechanical equipment. The concealment of typical faults in hydraulic pumps makes the accurate fault detection challenging. Fault signals are typically temporal, and this paper proposes an intelligent method based on synchrosqueezing wavelet transform (SWT) and an improved fusion model. The method fully incorporates: 1) SWT to improve the time-frequency separation of fault information; 2) deep feature extraction from image information of VGG; 3) long short-term memory (LSTM) to control network information transmission and optimise memory stacking methods of the network to achieve high-precision fault detection. The proposed hybrid VGG-LSTM is evaluated by analysing pressure signals from the hydraulic pump. Compared with other models, the proposed model has excellent fault recognition ability with minimum error and standard deviation, which proves that the theory of fusion reconfiguration network is reasonable.

Keywords: hydraulic pump; convolutional neural network; defect detection; long short-term memory; LSTM.

DOI: 10.1504/IJHM.2025.150808

International Journal of Hydromechatronics, 2025 Vol.8 No.4, pp.412 - 443

Received: 18 Apr 2025
Accepted: 01 Aug 2025
Published online: 23 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article