Fault prediction of Power-Shift Steering Transmission based on support vector regression

Spectrometric oil analysis technology is an important method in condition monitoring. This method has been applied to study the state of Power-shift Steering Transmission (PSST) in this paper. But, how to predict the future state of the PSST using existing data is a difficult work. In order to solve this problem, a support vector regression method is applied. The building process of this method is offered. Radial Basis Function (RBF) is selected as the kernel function. This method is applied to study the spectrometric oil analysis data. During the process, the values of parameters γ and σ are studied using grid search method. And the prediction of spectrometric oil analysis data for PSST is done. A comparative analysis is made between predictive and actual values. The method has been proved that it has better accuracy in prediction, and any possible problem in PSST can be found through a comparative analysis which has important significance for preventing faults.