基于神经网络的复杂过程系统模型辨识

针对复杂生产过程中的一阶和二阶液位系统，利用MATLAB软件的神经网络工具箱，分别应用BP和径向基两种神经网络模型进行系统辨识，得到系统模型．通过结果比较，得出两种神经网络的应用特点：对于一阶非线性液位过程，径向基神经网络创建的数学模型性能较好；对于二阶线性液位过程，BP神经网络的建模效果较好；尽管BP神经网络的模型训练过程有学习收敛慢、局部最小点、层数和单元数不易确定的缺点，但其函数逼近的精确度对二阶线性的辨识具有独特优势．

Model Identification of Complex Process Systems based on Neural Networks

The models of the first and second order liquid level system of the complex production process are identified by using the BP and RBF neural networks with the neural network toolbox of MATLAB. Based on the comparison result, the application properties of the two neural networks are obtained. To the first order nonlinear liquid level process, the RBF network model has the better performance. To the second order linear liquid level process, the BP network model has the better effectiveness. The BP network possesses a distinct advantage in identification of second order linear systems with good accuracy of function approximation, even though it has some disadvantage of low convergence speed, etc.