Health State Estimation of Hydraulic System Based on Generalized Multiclass Support Vector Machine

Abstract

Health status estimation plays an important role in hydraulic system for ensuring safety and efficient operation. Accurate health status estimation can effectively avoid sudden failure of hydraulic system and greatly reduce maintenance costs. Support vector machine (SVM) is a common and effective method for health status estimation. However, traditional extensions of the binary support vector machine (SVM) to multistate classification are either heuristics or require solving a large dual optimization problem. This paper presents a generalized multiclass support vector machine (GenSVM)-based method for health status estimation of hydraulic system. First of all, the multisensory monitoring signals are preprocessed and sensor data which are highly correlated with health status are selected. Then, feature extraction is implemented using massive multiple sensor signals. Meanwhile, a new and flexible GenSVM model is constructed to identify the health status of hydraulic system, which integrates simplex encoding for input data high-dimensional transformation. In addition, grid search is used to determine the optimal model parameters. A practical application case study is implemented to verify the effectiveness of the proposed method. Health state estimation for the four main components of the hydraulic system is achieved including cooler, valve, pump and accumulator. The results show the superiority performance of the proposed method compared with other standard methods.