Performance Assessment and Fault Diagnosis for Hydraulic Pump Based on WPT and SOM

Hydraulic pump is the heart of hydraulic system, therefore a real-time condition monitoring for hydraulic pump is crucial to the reliability of the entire system. In this study, a method for performance assessment and fault diagnosis to hydraulic pump based on wavelet packet transform (WPT) and Self-organizing mapping (SOM) neural network is proposed. First, WPT is utilized to decompose the vibration signal into components, energy of each component is extracted and normalized to form the feature vector. Second, SOM neural network, trained only by normal data, is used to map feature vectors into Minimum Quantization Error (MQE), which is then normalized into confidence values (CV). Performance assessment is accomplished by tracking the trends of CVs. Finally, when faults occur, SOM, trained by both normal and faulty samples, is employed to classify the faults into different groups, which delegates different fault modes of the hydraulic pump. In addition, Taguchi method is used to reduce the redundant features and extract the principal components to ensure the effectiveness of the approach. A case study based on the vibration dataset of test plunger pump rig is conducted to demonstrate that the proposed method is able to assess the performance of hydraulic pump and diagnose faults suitably.