伺服电机NCS神经网络PID趋近律滑模控制

Neural Network PID Reaching Law Sliding Mode Control of Servo Motor NCS

针对直流伺服驱动电机的网络控制系统的非线性控制系统特性和神经网络多包传输的特性,提出一种基于滑动窗口策略的多核LS-SVM神经网络PID趋近滑模控制器。该控制器可以在线控制和预测丢包补偿,并将其控制系统实现为一种具有延迟和丢包的多包数据传输直流控制器的伺服驱动电机神经网络自动控制补偿系统。其主要方法为,首先基于等效变换、无延迟和滑动窗口相结合的LS-SVM在线数据包损耗预测,建立系统的延迟补偿模型。其后通过神经网络的非线性映射对PID参数进行在线调整,实现稳态并进行分析。仿真结果表明,组合内核LS-SVM预测策略可以提高数据包损失补偿的准确性,减少系统抖振,在响应速度较快的情况下完成整定。

In view of the nonlinear control system characteristics of DC servo-driven motor and the characteristics of neural network multi-pack transmission,a multi-core LS-SVM neural network PID based on sliding window strategy is proposed.The controller can control and predict packet loss compensation online and realize its control system as a servo-driven motor neural network automatic control compensation system with a multi-pack data transmission DC controller with delay and packet loss.The main method is to establish the system’s delay compensation model based on the LS-SVM online packet loss prediction combining equivalent transformation,no delay and sliding window.The PID parameters are then adjusted online by the nonlinear mapping of neural networks to achieve steady state and analysis.The results show that the combined kernel LS-SVM prediction strategy can improve the accuracy of packet loss compensation,reduce system vibration,and complete the correction in the case of faster response speed.