论文部分内容阅读
模糊神经网络能够以任意精度逼近任意复杂的非线性关系,具有高度的自适应和自组织性,在解决高度非线性和严重不确定系统的控制方面具有巨大的潜力,然而基于BP训练算法易陷入局部极小点的缺点,提出了控制器以三角型隶属度函数的BP神经网络结构,利用改进的遗传算法(GA)对结构和参数进行同步优化,改进适应度函数指导搜索过程,保证稳定情况下大大加快了收敛的速度。最后采用Matlab7.0的Simulink工具以轧机张力为对象进行仿真试验,结果证明了其有效性。
Fuzzy neural network can approach arbitrary complex nonlinear relations with arbitrary precision with high degree of self-adaptability and self-organization. It has great potential to solve the problem of highly nonlinear and severe uncertain system control. However, BP-based training algorithm is easy to fall into This paper proposes a BP neural network structure with triangular membership function as the controller. The improved genetic algorithm (GA) is used to synchronously optimize the structure and parameters, and the fitness function is improved to guide the search process and ensure the stability Under the greatly accelerated convergence speed. Finally, Simulink tool of Matlab7.0 was used to simulate the tension of the rolling mill. The results proved its effectiveness.