YAN Dan2, ZHANG Ji-long1, WANG Zhi-bin1, et al. Prediction model of gas volume fraction based on golden section and BP network[J]. 2012, 31(1): 4-6. DOI: 10.13873/j.1000-97872012.01.003.
基于黄金分割和BP网络的气体体积分数预测模型
摘要
研究基于黄金分割法的神经网络在红外光谱定量中的应用。通过黄金分割法对神经网络隐含层节点数进行优化
再将4组分混合气体预测集样本通过LM学习算法建立模型
用所得到的神经网络模型对10组混合气体光谱数据进行检验
平均相对误差为0.63%。使用该方法建立多组分污染气体定量分析模型的时间短
而且能够较精确地预测各组分的体积分数。
Abstract
The application of the quantitative analysis of the infrared spectra based on neural network with golden section method is studied.Golden section method optimizes the number of nodes in the hidden layer of neural networks.Using the LM learning algorithm models of the sample of the four-component mixture gas are set up.Using the model of the neural network
tests on spectral data of 10 groups of mixture gas and the average relative error is 0.63 %.By this method
the time to build multi-component model for quantitative analysis on pollutional gas is short
and this model can accurately predict the volume fraction of each component.