Prestressed Tension Hydraulic System Identification Using Linear Programming Support Vector Regression

Abstract

Prestressed tension hydraulic system has been widely applied in many construction engineering fields. Getting an accurate mathematical model for this system could improve its control properties largely. For the common derivation model, many parameters related with oil properties or elements’ characters are ascertained by experiments which finally lead to the model deviation. To obtain a relatively accurate system model, the mathematical model of the system is firstly deduced to determine the identified model with unknown parameters to serve as the priori knowledge for the identified model structure construction. Then the linear programming support vector regression method (LP-SVR) is designed. After choosing the motivated signal, the optimized regularization parameter and the insensitive loss functions, the real system is identified and the acquired model’s generalization performance is tested. Test results shows that the identified model could represent the real system well, which means a good identification ability of the LP-SVR for the prestressed tension hydraulic system.