基于SA-BP算法的小波神经网络在喷射器性能预测中应用  

Application of Wavelet Neural Networks Based on SA-BP Algorithm in Forecasting the Performance of Ejectors

作　　者：黄亮亮[1] 曹家枞[1] 汤小亮[1] 

机构地区：[1]东华大学环境科学与工程学院

出　　处：《建筑热能通风空调》2009年第6期1-4,共4页Building Energy & Environment

基　　金：上海市重点学科建设项目资助(B604)

摘　　要：以现有的喷射器实验结果数据集作为样本,用三种小波神经网络分别预测喷射器的性能,网络的训练分别采用模拟退火BP算法和BP算法,得出了最适合本模型的小波函数。数值实验结果显示,这种小波神经网络预测喷射器性能的精度能够满足实际工程的要求,而采用模拟退火BP算法比采用BP算法训练的效果略好。In this paper, models of 3 kinds of wavelet neural network were established to forecast the performance of questioned ejectors, of which a great number of experimental data had been obtained. SA-BP and BP algorithms were adopted, respectively, for network training based on the obtained dataset. According to the results of training, the most suitable wavelet base was selected from the 3 wavelet bases to be the activation function of the network. Numerical experiments show that the forecasting of the ejector performance can meet the requirement of ejector in practice, and the application of SA-BP training algorithm is somewhat better than of BP training algorithm.

关 键 词：喷射器 小波神经网络 模拟退火BP算法 性能预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP18[自动化与计算机技术—控制科学与工程]