摘要
由于传统方法没有对模型噪声实现有效处理,导致辨识精度与辨识速度较低,为此提出基于调制函数法的动力学系统参数辨识算法。对模型噪声进行表示与替换,通过调制函数法构造辨识模型,通过递推算法对辨识模型进行处理以实现噪声的处理。为构造出与动力学系统模型相一致的参数辨识模型,以动力学系统模型为基础对参数辨识模型的神经网络拓扑结构进行构造。由于动力学系统存在输出误差,需要对参数辨识模型的权值调整算法进行设置,最终完成对动力学系统的参数辨识。在仿真实验中,利用设计算法对某伺服系统进行参数辨识,辨识的参数包括阻尼、刚度、惯性、死区四种物理参数。实验结果表明该算法实现了辨识误差的降低与辨识时间的缩短,使辨识精度与辨识速度得到了大幅提升。
Generally,the traditional methods have the disadvantages of low identification accuracy and speed.Therefore,we report a dynamic system parameter identification algorithm based on modulation function method.The model noise was represented and replaced,the identification model was constructed by the modulation function method,and the identification model was processed by the recursive method to realize the noise processing.In order to construct a parameter identification model consistent with the dynamic system model,the neural network topology of the parameter identification model was constructed based on the dynamic system model.Due to the output error of the dynamic system,the weight adjustment algorithm of the parameter identification model needs to be set to finally complete the parameter identification of the dynamic system.In the simulation experiment,the design algorithm is used to identify the parameters of a servo system.The identified parameters include four physical parameters:damping,stiffness,inertia and dead zone.The experimental results show that the algorithm not only reduces the identification error and identification time,but also improves the identification accuracy and speed.
作者
房怀媛
杨中艳
FANG Huai-yuan YANG Zhong-yan(Linyi University,Shandong Linyi 273400,China)
出处
《计算机仿真》
北大核心
2022年第3期362-366,共5页
Computer Simulation
关键词
调制函数法
白噪声
动力学系统
递推算法
参数辨识算法
modulation function method
White noise
Dynamic system
Recursive algorithm
Parameter identification algorithm
