Application of state parameter learning for fault diagnosis on the large reciprocating compressor
Xiaoran Li,Tianshuo Guo,Weifeng Wu,Chengyi Li,Jie Li,Zhongqing Zhu
DOI: https://doi.org/10.1177/09544089221149824
2023-01-11
Abstract:Proceedings of the Institution of Mechanical Engineers, Part E: Journal of Process Mechanical Engineering, Ahead of Print. According to the statistical results of the reciprocating compressor maintenance in chemical enterprises, the probability of faults caused by wear or damage of vulnerable parts inside the cylinder is close to 80%. Now, fault diagnosis of vulnerable parts inside the cylinder is relied on vibration signal, acoustic emission signal, or thermal parameters. However, it is difficult to extract eigenvalues from vibration signals and acoustic emission signals, which can be disturbed by noise easily. Thermal parameters are relatively stable and less affected by noise. The measurement of thermal parameters requires drilling testing holes through the cylinder wall, which will decrease the strength of the cylinder. According to the working principle of the compressor, fault of the vulnerable parts inside the cylinder would change the pressure and temperature distribution at the suction and the discharge port of the cylinder, while these state parameters are monitored by the parameter monitoring system on most of the compressors. For these reasons, a fault prediction model based on state parameters learning is proposed in this paper aiming to fault diagnosis of vulnerable parts inside the cylinder, which could make fault prediction without adding new hardware cost, such as sensors. Optimized back propagation neural network method by genetic algorithm (GA-BP) is applied to establish the fault prediction model to describe normal working process of the compressor. When fault of vulnerable parts inside the cylinder occurs, monitored pressure and temperature would deviate from the predicted value by the fault prediction. The degree of the deviation is adopted to the fault diagnosis. Then, experiments simulating the fault condition were carried out, and it shows that the accuracy of fault diagnosis using this method could exceed 95%. Verification test shows that the proposed method could successfully predict the fault before the unplanned shutdown.
engineering, mechanical