Novel method for performance degradation assessment and prediction of hydraulic servo system

Performance degradation assessment and prediction of hydraulic servo systems has attracted increasing attention in recent years. This study proposes a performance degradation assessment and prediction method based on Mean Impact Value (MIV), Mahalanobis Distance (MD), and Elman neural network. First, a state observer based on Radial Basis Function (RBF) is designed to calculate the residual error between the actual and estimated outputs, and typical time-domain features, such as Root Mean Square (RMS), peak value and kurtosis, are extracted. Second, the MIV analysis based on BP neural network is applied to evaluate the sensitivity of each extracted feature, and the selected optimal features are employed to construct the Mahalanobis space for normal states. Third, the MD between the most recent state and the constructed space of normal state is calculated, which can be normalized into a condence value so as to assess the performance. Finally, an Elman neural network is used to predict the degradation trend. The proposed method is proven to be eective by a simulation model with the commonly occurring faults.