基于复合正交神经网络的自适应逆控制系统

目前，在自适应逆控制系统中常采用BP神经网络，而BP网络存在算法复杂、易陷入局部极小解等不足。而正交神经网络能克服BP网络的不足，但由于正交神经网络学习算法存在某些局限性，提出了一种复合正交神经网络，该正交网络结构与三层前向正交网络相同，不同的是正交网络的隐单元处理函数采用带参数的Sigmoid函数的复合正交函数，该神经网络算法简单，学习收敛速度快，并能对网络的函数参数进行优化，为非线性系统的动态建模提供了一种方法。仿真实验表明，网络在用于过程的自适应逆控制中具有很高的控制精度和自适应学习能力。该动态神经网络比其它神经网络具有更强的建模能力与学习适应性，有线性、非线性逼近精度高等优异特性，非常适合于实时控制系统。

Adaptive Inverse Control System Based on Compound Orthogonal Neural Network

Current neural networks usually use BP network in the adaptive control system. BP network's algorithm is complex and cannot avoid false local minima. The orthogonal neural network can overcome the insufficiency of BP network. But there are some limitations in the learning algorithm of orthogonal neural network. A kind of compound orthogonal neural network is presented oriented to the existing insufficiencies. The orthogonal neural network's structure is the same as that of the three layered feedforward neural network. The difference is that the processing function of hidden unit of the orthogonal neural network is the compound orthogonal function using a sigmoid function with a parameter. The neural network is a simple algorithm and a high-speed convergence of learning process, and optimizes the function parameter of the neural network. A kind of method is provided in dynamic modeling for the nonlinear systems. The simulation experiment shows that the network has higher control accuracy and adaptive learning ability in the adaptive inverse control system of the process. The dynamic neural network has stronger modeling ability and learning adaptation than other neural networks, and excellent characteristics in the linear and nonlinear accurate approximation. The compound orthogonal neural network is very suited for the real-time control system.