A hybrid fault diagnosis method for mechanic-electronic-hydraulic control system based on simulated knowledge from virtual prototyping

Abstract

For the fault diagnosis of mechanic-electronic-hydraulic control system (MEHCS), the main barrier that restricts the application of knowledge-based methods is the lack of historical fault data. Aiming at this problem, this paper proposed a hybrid fault diagnosis method based on simulated knowledge from virtual prototyping. As a special form of mathematical model, virtual prototyping of MEHCS under faulty and nominal condition was established, validated, fault-injected and simulated to obtain simulation data. Fault features of different fault types were extracted, which were then trained by three pattern recognition methods to build the knowledge database for diagnosis. Threshold test and ensemble classifier constituted by the three pattern recognition methods were employed respectively to realize fault detection and isolation. To verify the proposed methodology, a case study of vessel steering system was presented. Fault types of stuck rudder and steady state error were studied. Probabilistic neural network (PNN), naive Bayes (NB), and k-nearest neighbor (kNN) were employed to constitute ensemble classifier based on majority voting. The diagnosis results showed that the accuracy of fault detection and isolation of both fault types were highly acceptable. The ensemble classifier performed better on comprehensiveness and smoothness than any individual pattern recognition method for the overall diagnosis. The proposed method might be an available choice for the fault diagnosis of MEHCS, especially for large-scale and complicated cases.

Approach for Hydraulic Pump Fault Diagnosis Based on WPT-SVD and SVM

“…Hydraulic pump, as one crucial component of the hydraulic system, has a significant impact on the re-liable operation of the entire hydraulic system, be-cause of which the methods to diagnose faults of hydraulic pumps are urgent needed. Generally, the faults of hydraulic pump are accompanied with changes in vibration signals [1], that is, the faults can be detected by vibration signals, for which, diagnosis based on vibration signals is of great importance and also becomes the hottest topic discussed at home and abroad [2]. However, in practice, it is difficult to obtain effective fault vibration signals along with interference signals because of the complexity of the structure of hydraulic pump and the severe working conditions [3], considering which, data-driven fault detection methods are usually applied to fault diagnosis.…”