基于遗传算法和神经网络的油气弹簧非参数化建模

Nonparametric Modeling of Hydro-pneumatic Spring Based on Genetic Algorithm and Artifical Neural Network

将遗传算法应用于油气弹簧神经网络模型的优化,首先利用遗传算法的全局搜索能力得到神经网络权值的次优解,然后利用BP算法精确搜索到权值的最优解,从而克服了传统BP算法易陷入局部最小点的缺点。与采用传统BP算法的神经网络比对结果表明,遗传算法能显著地提高神经网络的精度,建立的油气弹簧人工神经网络模型可以对油气弹簧的输出特性进行可靠地预测。

Genetic algorithm（GA） was used to optimize the artifical neural network（ANN） model of hydro--pneumatic spring herein. With GA, the suboptimal weight matrices of the ANN model were firstly obtained, then the optimal weight matrices were farther achieved with back propagation algorithm（BPA）. Thus, the shortcoming of plunging into local optimum of BPA was overcome. The results of the ANN model with GA and without GA show that GA can improve the precision of ANN obviously and the ANN model of hydro--pneumatic spring can preestimate the output characteristics of hydro--pneumatic spring reliably.