遗传神经网络滑模控制在交流伺服控制中的应用研究

Application Research of GA Optimized RBF Network-Sliding Model Controller for AC Servo System

基于永磁同步电机(PMSM)的交流伺服系统能够有效提高机床加工能力,但是PMSM具有时变非线性,很难建立它的精确数学模型,因此采用传统PID控制器难以得到期望的控制效果。为保证控制精度,研究了一种遗传神经网络-滑模变结构矢量控制交流伺服系统。该控制系统采用id=0矢量控制策略,应用滑模控制器来准确跟踪伺服位置指令,并在速度环利用径向基神经网络(RBF)来优化滑模控制律以改善与消除滑模抖振;同时为了提高RBF工作效率,采用改进遗传算法(GA)离线优化RBF初始结构。通过dSPACE半实物仿真平台对PMSM进行实验测试,结果表明,所设计的遗传-RBF网络-滑模控制器能够在外界干扰下稳定工作,精确跟踪伺服位置指令,并且控制性能比只采用滑模控制好。

The application of permanent magnet synchronous motor（PMSM）can enhance the working ability of the machine tools.However,in the PMSM based servo system,it is difficult to establish accurate control mathematical model due to of the strong nonlinearity of the PMSM.Hence,traditional PID controller is hard to get satisfied control performance.In order to improve the control efficiency of the AC servo system,the artificial neural network（ANN）based sliding model control（SMC）method has been presented in this paper.The id=0 vector control scheme was adopted for the servo control,and the SMC was employed to track the servo position.In addition,the radial basis function（RBF）neural network was used to eliminate the buffeting problem of the sliding model.Moreover,to enhance the RBF efficiency,the improved genetic algorithm（GA）was applied to the structure optimization of the RBF.The control experiments using the dSPACE simulation platform has been established to evaluate and validate the proposed control method.The test results show that the proposed controller is effective for the AC servo system control under severe disturbance,and can follow the given position inference precisely.Additionally,the control performance of the proposed GA-RBF-SMC is superior to the SMC.