摘要
采用傅里叶变换近红外光谱法测定发射药中外挥发分和内挥发分的含量。本文提出了一种混合算法,该算法将偏最小二乘法(PLS)和人工神经网络(ANN)结合起来,同时利用马氏距离(Maha-lanobis)法对异常样品进行剔除。与传统的多元校正算法PLS和主成分回归(PCR)相比,该算法所建模型的预测精度有明显的提高。结果表明,该算法可以满足发射药成分含量的快速分析的需要。
In this paper,a mixed algorithm was developed based on the combination of Partial l.east Square （PLS） with artificial neural network （ANN） and mahalanobis-distance method to eliminate the outlier samples,and applied to the determination of contents of outer-volatile matter and inner-volatile matter in detonator by Fourier transform near infrared （FT-NIR ） spectroscopy. Compared with the classical multivariate calibration methods such as principle component regression （PCR） and PLS,the proposed algorithm performcd much better,and can be used for fast analyzing of the contents of components in detonator.
出处
《光谱实验室》
CAS
CSCD
2006年第2期187-190,共4页
Chinese Journal of Spectroscopy Laboratory
关键词
近红外光谱
偏最小二乘法
人工神经网络
马氏距离
混合算法.
Near-Infrared Spectroscopy, Partial Least Square, Artificial Neural Network,Mahalanobis Distance,Mixed Algorithm.
作者简介
郭志强（1981-）,男,北京市人,硕士研究生,从事过程自动化和在线检测技术研究工作.
任芊（1953-）,女,太原市人,副教授,从事传感器与检测技术研究工作.联系人,手机：（0）13810715935;E-mail：wiserguo@163.com
