Prediction of leakage of a pressure-relief valve based on support vector regression with auxiliary input information

It is significant if the gas leakage of pressure-relief valve (PRV) can be accurately predicted in advance, because gas leakage may destroy the PRV's stability and result in a potential damage to the entire pneumatic circuit. This article puts forward a model based on support vector regression (SVR) to predict the gas leakage of PRV. The experimental data of three identical PRVs are employed to evaluate the performance of the leakage prediction model with different input information. The experimental results demonstrate the superiority of the proposed model, proved especially when the pressure difference between inlet and outlet pressures of PRV is chosen as the auxiliary input information. The SVR prediction model which only uses leakage data as input information can perform better than the artificial neural network prediction model, but the prediction accuracy can be further improved if the proper auxiliary input information is taken.