Performance degradation prediction for a hydraulic servo system based on Elman network observer and GMM–SVR

Abstract A hydraulic servo system often exhibits the characteristics of high nonlinearity and non-stationarity, which result in difficulties in establishing an accurate model for performance degradation prediction. Thus, this study investigates an approach for the performance degradation prediction of a hydraulic servo system based on Elman network observer and support vector regression (SVR). First, an Elman neural network is used to establish a fault observer, and the residual error is obtained by comparing the estimated output and the actual output. Second, time domain features of the residual error are extracted for performance degradation assessment, in which a Gaussian mixed model (GMM) is utilized to calculate the overlap ratio of the residual errors between the normal state and the most recent state, thus obtaining a health confidence value that represents the performance degradation degree of the hydraulic servo system. Third, an SVR model is established to forecast the health state of the hydraulic servo system over time. Finally, experimental results demonstrate the effectiveness of the proposed method based on residual error and GMM–SVR.