基于BP神经网络的非晶相外标定量分析  

Amorphous Phase Quantitative Analysis Using External Standard Method Based on BP Neural Network

作　　者：郑毅[1] 郭培民[2] 王海风[2] 方建锋[1] 柳春兰[1] 

机构地区：[1]钢铁研究总院粉末冶金研究所,北京100081 [2]钢铁研究总院先进钢铁流程及材料国家重点实验室,北京100081

出　　处：《钢铁研究学报》2008年第2期57-59,共3页Journal of Iron and Steel Research

基　　金：国家自然科学基金资助项目(50474006)

摘　　要：说明了非晶相不能用Rietveld方法来定量的原因。BP神经网络可以实现非线性的映射关系,非晶相的含量与其本底的高度存在一一对应关系,将两者联系了起来。用本底和衍射峰面积的数据,通过神经网络计算来求得非晶相的含量,所得结果优于多项式及S函数计算所得结果。同时指出了BP神经网络计算方法的适用范围。The quantification of amorphous phase can not be obtained by Rietveld. As the nonlinear mapping relationship can be carry out by BP neural network, the content and the background of the amorphous phase have corresponding relation, and their relationship is determined. The data of the background and the diffraction peak area was calculated with BP neural network to acquire the content of the amorphous phase. The calculated result of BP neural networks is better than that of polynomial and sigmoidal function. At the same time, the applicability of BP neural network calculation is also given.

关 键 词：外标法 BP神经网络 非晶 

分 类 号：TG111[金属学及工艺—物理冶金]