A hybrid intelligent diagnostic approach for spool jamming faults of hydraulic directional valves

Abstract

Directional valves are critical functional components in hydraulic systems for diverting the flow of hydraulic oil and controlling oil pressure. Faults of the valves could lead to failure and even serious accidents in hydraulic systems. Fault diagnostic accuracy is directly influenced by the effectiveness of signal processing. Due to the harsh operating environment of hydraulic systems, the condition monitoring signal acquired from the systems contains a large amount of redundant information and environmental noise. This makes the training processes of fault diagnostic approaches time-consuming and inefficient. In this paper, a new hybrid intelligent approach to spool jamming fault diagnostics for hydraulic directional valves is presented. The approach is designed based on the combined strengths of wavelet packet decomposition (WPD), empirical mode decomposition (EMD), variable mode decomposition (VMD), and the multi-layer perceptron (MLP) to improve the efficiency and accuracy of the fault diagnostic process. Five evaluation indices, including processing time (PT), local average accuracy (LAA), stability of accuracy (SA), weighted average of the neuron number (WANN), and robustness of WANN (ROB), are applied to benchmark the approach using ablation experiments. The experimental results show that the hybrid approach has a strong capability to extract fault-related features from the signal and perform fault diagnostics with high accuracy, efficiency, and robustness.