利用支持向量机的摩擦模型参数辨识

Research on the Parameter Identification of Friction Model Based on Support Vector Machine

以Tustin摩擦模型为参数辨识对象,提出一种基于支持向量机算法的摩擦模型参数辨识的方法.构建训练样本并选取适当的支持向量机模型,选择具有较好泛化能力的径向基核函数和具有稀疏性特点的ε不敏感损失函数,以求解最优化问题,得到最优解.以某直流电机高精度位置伺服系统为对象,用辨识得到的参数估计值设计摩擦力矩的补偿环节,对系统进行补偿,仿真结果表明,算法的辨识精度比较高.

A method for the parameter identification of the friction model based on support vector machine is proposed with Tustion friction models as the object for parameter identification. The optimum solutions are obtained by solving the optimization problem where training samples are constructed, the appropriate model of support vector machine （SVM for short） is selected, and .the radial kernal function with better generalization ability and E-insensitive loss function with the sparse characteristics are selected as well. With a DC motor high-precision position servo system as the research object, the system is compensated by using the estimated value of parameters to design the compensation aspect of friction torque. The simulation results show that the algorithm has high recognition accuracy.