Featured temporal segmentation method and AdaBoost-BP detector for internal leakage evaluation of a hydraulic cylinder

摘要

Internal leakage detection is a vital technique for the maintenance of hydraulic systems. In this paper, we proposed a featured temporal segmentation method and an AdaBoost-BP detector to automatically evaluate the internal leakage faults in hydraulic actuators. The internal leakage detection model based on the proposed featured temporal segmentation (FTS) combined with AdaBoost-BP method consists of three main steps. First, the pressure signals are divided into several segments using wavelet analysis, with the assumption that the temporal segments may contain different sensitive features related to the internal leakage. Second, for every segment, several statistics from the time domain are extracted to describe the working conditions of the hydraulic cylinders. All these variables from different temporal segments are combined to form the feature vectors. Third, an AdaBoost-BP neural network classifier is applied as the evaluator to differentiate the levels of the internal leakage of the hydraulic cylinders. The proposed method was validated on a test bench of hydraulic cylinders with different levels of the internal leakage. The comparison with the traditional features extracted from the wavelet coefficients and the FTS method are conducted. In addition, the accuracy of the classifier in terms of the AdaBoost-BP method is compared with that of the classifier based on Random Forest. The results show that the FTS and AdaBoost-BP method can not only effectively identify the leakage levels but also improve the accuracy of evaluating the internal leakage. Thus, the proposed method of internal leakage evaluation may enable the condition based monitoring of the hydraulic system by using the built-in pressure sensors, which is feasible in most practical industrial applications. (C) 2018 Elsevier Ltd. All rights reserved.