基于遗传算法优化小波网络的柔性喷管力矩特性辨识方法  

Identification of flexible nozzle torque properties based on wavelet neural network optimized by genetic algorithm

作　　者：杨弘枨 刘山[1] 靳广在 焦玮玮[1] 姜玉峰[1] YANG Hongcheng;LIU Shan;JIN Guangzai;JIAO Weiwei;JIANG Yufeng(Laboratory of Aerospace Servo Actuation and Transmission,Beijing Institute of Precision Mechatronics and Controls,Beijing 100076,China)

机构地区：[1]北京精密机电控制设备研究所航天伺服驱动与传动技术实验室,北京100076

出　　处：《航空动力学报》2022年第9期1936-1945,共10页Journal of Aerospace Power

摘　　要：为准确辨识负载力矩并提高主动加载负载模拟的真实度,使用了一种基于遗传算法优化的神经网络辨识方法。使用小波分析方法对测试信号进行预处理,将消噪与分解后得到的信息作为神经网络训练的扩充样本,提高了辨识精度。使用遗传算法选择最优输入信息、网络结构和隐含层规模,加快网络收敛速度并简化计算过程,实现对柔性喷管力矩的快速准确辨识。仿真结果表明该辨识方法可以准确地描述柔性喷管在典型测试信号激励下的力矩特性,平均辨识误差为2%,对于实现精确主动加载控制和验证伺服控制性能具有重要意义。In order to identify the load torque accurately and improve the authenticity of active load simulation,a neural network identification method based on genetic algorithm optimization was used. The wavelet analysis method was used to preprocess the test signal,and the information obtained after de-noising and decomposition was used as the expanded sample of neural network training,which improved the identification accuracy. The genetic algorithm was used to select the optimal input information,network structure and hidden layer scale,which can speed up the network convergence speed and simplify the calculation process,so as to realize rapid and accurate identification of the characteristics of flexible nozzle. The simulate results showed that this identification method can accurately reflect the torque characteristics of the flexible nozzle excited by typical test signal,with the average identification error of 2%,proving that it is of great significance to realize precise active load control and accurately verify the servo control performance.

关 键 词：伺服系统 柔性喷管 系统辨识 小波网络 小波分析 遗传算法 

分 类 号：V448[航空宇航科学与技术—飞行器设计]