基于GABP神经网络的液压互联悬架建模研究

Hydraulic Interconnection Suspension Modeling Based on GABP Neural Network

液压互联悬架(hydraulically interconnected suspension,HIS)是一种非线性系统,运用机理分析法建模存在建模精度和速度不可兼得的缺点。为解决上述矛盾,提出了一种基于遗传算法(genetic algorithm,GA)优化的反向传播(back propagation,BP)神经网络对HIS系统进行建模的方法。首先,通过Simulink建立的液压互联悬架模型仿真获取网络的训练数据。其次,使用遗传算法优化BP神经网络的初始权值和阈值;然后,两种建模方法对比验证GABP建模方法优点;最后,通过液压互联悬架台架实验获取实验数据,与神经网络训练结果进行比较分析。结果表明:在垂向模态下,低、中、高3种频率下相对误差百分数分别为4.12%、2.27%、1.51%;在侧倾模态下,低、中、高3种频率下相对误差百分数分别为7.64%、4.07%、4.35%。与机理建模法相比,GABP建模方法兼具较好的建模精度和速度。

Hydraulic interconnected suspension(HIS)is a nonlinear system,whose modeling method by mechanism can’t both have modeling accuracy and speed.In order to solve the contradiction,a modeling method for HIS system based on genetic algorithm(GA)optimized back propagation(BP)neural network was proposed.Firstly,the training data was obtained by the simulation of hydraulic interconnected suspension in Simulink.Secondly,genetic algorithm was used to optimize the initial weights and thresholds of the BP neural network.Thirdly,the two modeling methods were compared to verify the advantages of GABP modeling method.Finally,the experimental data obtained by the experiment of bench and the results simulated by the neural network were compared and analyzed.The results show that the relative error percentages of the low,medium and high are 4.12%,2.27%,and 1.51%in the vertical mode respectively.The relative error percentages of the low,medium and high are 7.64%,4.07%and 4.35%in the roll made respectively.It is concluded that Compared with the mechanism modeling method,the GABP modeling method has better modeling accuracy and speed at the same time.