Abstract: A novel algorithm of neural network structure optimization is described in this paper. The capabilities of the algorithm are demonstrated by application in steady-state process, dynamic process and industrial process. The recognized results of neural network structure show that the optimal neural network can be widely used for modelling and simulation of chemical processes.

摘要： 利用修正的反传神经网络(BP)算法分别对稳态过程和动态过程以及两个实际化工生产过程进行模型结构辨识,所确定的最佳神经网络结构经验证表明具有较好的收敛性和稳定性,模拟精度高、适用性强.