基于GA-BP神经网络的气压伺服系统辨识研究

气压伺服系统控制器的优化设计依赖于准确的系统模型。针对系统的非线性问题,研究采用神经网络进行系统辨识的原理和结构;考虑传统BP算法存在局部收敛、学习速度慢的问题,采用遗传算法对神经网络的初值和权值进行优化,并采用LM算法进行网络学习,最终建立系统的神经网络辨识模型。通过仿真对比神经网络辨识结果与传统线性模型辨识结果,结果表明:基于GA-BP神经网络的辨识模型精度较高,适用于非线性系统辨识。

Research on Identification of Pneumatic Servo System Based on GA-BP Neural Network

The optimal design of the controller of pneumatic servo system depends on the accurate system model.For the nonlinear problem of the system,the principle and structure of the system identification using neural network were studied.In view of the shortcomings of local convergence and slow learning speed of traditional BP algorithm,genetic algorithm was used to optimize weights and thresholds of neural networks,and LM algorithm was used for network learning.The neural network identification model of the system was established finally.The identification results of the neural network and traditional linear model were compared by simulation.It shows that the identification model based on GA-BP neural network is more accurate and suitable for nonlinear system identification.