Fault Diagnosis of Hydraulic Seal Wear and Internal Leakage Using Wavelets and Wavelet Neural Network

Abstract

The piston seal wear in hydraulic cylinder is one of the main factors that give rise to an internal leakage. This paper focuses on diagnosing piston seal wear and subsequent internal leakage from a double acting seal combination seal used in the support oil cylinder of a QY110 mobile crane. Wavelet transform is applied as a feature extractor to transform the raw oil pressure data into a feature vector consisting of wavelet packet subband energy, energy entropy, energy variance, and root mean square of the wavelet detailed coefficient $d_{4}$ . This feature vector feeds into the wavelet neural network serving as a pattern recognizer for automatically classifying the fault patterns. We demonstrate with the leakage experiment and simulation data that the proposed fault detection and identification (FDI) scheme is capable of effectively detecting and classifying the piston seal wear with excellent accuracy. Our comparison studies reveal that the proposed FDI tandem produces much more accurate result than that from back-propagation neural network. This paper is supplement to and enrichment of existing studies on fault simulation and diagnosis associated with hydraulic cylinder leakage problems.