基于小波神经网络的化工安全评估

鉴于传统神经网络方法解决非线性问题收敛速度慢,易陷入局部最优解的缺陷,本文通过对小波神经网络的结构及学习算法的简要介绍,结合神经网络的自学习能力,提出一种充分利用小波变换时频局部化性质的小波神经网络安全评价方法,通过用小波神经网络评价方法与BP神经网络评价方法对某大型炼油化工厂相应原始数据进行分析、对比,表明该小波神经网络评价方法较BP神经网络评价方法收敛迅速,绝对误差小,预测精度高。

Safety Assessment of Chemical Industry Based on Wavelet Neural Network

BP neural network has the shortcomings of slow convergence and is prone to fall into local optimums flows.In this paper,structure and learning algorithm of wavelet neural network are briefly introduced,then a new method of wavelet neural network which can make full use of part characteristic of wavelet time-frequent and combine the ability of self-study of neural network is presented to form a model for safety assessment.Based on analyzing relevant data from a large-scale petroleum refinery company with wavelet neural network assessment model and BP network,the results show what the new model can achieve faster convergence and lower absolute error and more accurate effect compared with that of BP network.