Session: 01-04: System Modelling and Fault Diagnosis

Paper Number: 139566

139566 - Fault Diagnosis of Pump Rotor-Bearing System Based on Frequency Temporal Series Graph Under Time-Varying Speed Conditions 

Abstract: 

The pump serves as a crucial power component in the hydraulic system. In the actual working process, it often runs in time-varying speed conditions such as startup, shutdown, acceleration, and deceleration, which makes the rotor-bearing system in the pump often work in an unstable environment and prone to failure. Most of the existing fault diagnosis researches are based on vibration signals. However, variations in rotational speed over time will also induce changes in vibration signal, which will cover the signal change caused by the fault of the system itself, resulting in confusion of signal characteristics and misjudgment of fault diagnosis. To perform precise fault diagnosis under time-varying speed conditions, a diagnosis method based on the Frequency Temporal Series (FTS) graph is proposed. The time domain vibration signal is converted into the frequency domain signal by the Fast Fourier Transform (FFT). The frequency values corresponding to the top  maximum amplitudes in the spectrum are selected as the node features, and the nodes are connected in chronological order through edges. The fault signal is constructed as the FTS graphs to map the time-varying fault information. Subsequently, a synchronous extraction model of spatio-temporal information is built, and the fault feature information embedded in graph nodes and the time dependence information among internal nodes of the graph are extracted at the same time to generate multi-scale fusion features. Finally, the fully connected layer and  are employed as the classifier to categorize the multi-scale fusion features that have been extracted spatio-temporal information. Through experiments under time-varying speed conditions, it is verified that the proposed method can effectively identify the speed change during the operation of the system. The diagnostic accuracy reaches 99.17%, outperforming other deep learning diagnostic methods.

Presenting Author: Sichao Sun State Key Laboratory of Fluid Power Components and Mechatronic Systems, Zhejiang University

Presenting Author Biography: Sichao Sun received a B.S. degree in mechanical engineering from Jilin University, Changchun, China, in 2019. He is currently pursuing a Ph.D. degree in the State Key Laboratory of Fluid Power and Mechatronic Systems, Zhejiang University, Hangzhou, China. His current research interests include signal processing, prognostic and health management (PHM), and fault diagnosis of fluid machinery.

Authors:

Sichao Sun State Key Laboratory of Fluid Power Components and Mechatronic Systems, Zhejiang University
Xinyu Xia State Key Laboratory of Fluid Power Components and Mechatronic Systems, Zhejiang University
Jiale Yang State Key Laboratory of Fluid Power Components and Mechatronic Systems, Zhejiang University
Hua Zhou State Key Laboratory of Fluid Power Components and Mechatronic Systems, Zhejiang University