铝合金光电光谱分析的径向基神经网络方法

RBF method of photoelectric spectral analysis of aluminium alloy

在铝合金光电光谱分析的定量分析阶段,确定光谱强度与元素含量的对应关系,将直接影响定量分析的质量。根据标准铝合金成份的确定性,将单一径向基神经网络,改为由每一元素对应一子径向基神经网络,再将这些子径向基神经网络组合成一完整神经网络,以完成铝合金的定量分析,并利用Matlab中的径向基网络,构建函数newrb()返回误差,使每个径向基网络的均方误差减到最小。在Matlab中,模拟实验证明用该方法训练的组合径向基网络所得均方误差,是单一径向基网络均方误差的1/20。

At the photoelectric quantitative analysis stage for spectral analysis of aluminium alloy, the certain intensity of spectrum and corresponding relation of the content of element will influence quantitative analysis quality directly. According to standard aluminium alloy determinacy of composition, creating a sub radial basis function（RBF） neural network by every element in aluminium alloy to replace single RBF neural network, make these sub RBF neural network up one intact neural network finishing the quantitative analysis of the aluminium alloy,then utilize characteristic unexposed of RBF network struction function newrb （ ）of Matlab, make square error minimum for every radial base network. The simulation experiment proves the association RBF neural network,training with this method,it＇s the square error is 20 times less than the single radial base network in Matlab.