Performance assessment of hydraulic servo system based on bi-step neural network and autoregressive model

In recent years, condition monitoring and fault diagnosis of hydraulic servo systems has attracted increasing attention. However, few studies have focused on the performance assessment of these systems. This study proposes a performance assessment method based on a bi-step neural network and an autoregressive model for a hydraulic servo system; the performance is quantized by the performance confidence value (CV). First, a fault observer based on a radial basis function (RBF) neural network is designed to estimate the output of the system and calculate the residual error. Second, the corresponding adaptive threshold is generated by using another RBF neural network during system operation. Third, the difference value between the coefficients of the autoregressive model for the generated residual error and the adaptive threshold is obtained, and the Mahalanobis distance (MD) between the most recent difference (unknown conditions) and the constructed Mahalanobis space by using samples under normal conditions is calculated. Then, the condition of the system can be determined by normalizing the MD into a CV. The proposed method was further validated for three types of faults, and data were obtained using a simulation model. The experimental analysis results show that the performance of hydraulic servo systems can be assessed effectively by the proposed method.