基于近红外光谱技术的钢结构防火涂料品牌鉴别方法研究

发布时间：2019-01-06 13:47

【摘要】：为了实现钢结构防火涂料在流通使用领域中不同品牌的现场快速鉴别,提出了一种用近红外光谱技术快速鉴别钢结构防火涂料品牌的方法。运用光栅扫描型近红外光谱仪器,通过近红外漫反射光谱获取不同品牌钢结构防火涂料的光谱曲线,并对光谱数据进行标准正态变量变换(standard normal variate transformation,SNV)、Norris二阶求导等优化处理。利用主成分分析法(principal component analysis,PCA)对钢结构防火涂料品牌进行聚类分析,前五个主成分的累积方差贡献率已达到99.791%,以PC1,PC2和PC3×10的得分值对所有建模样品在三维空间作图,对不同品牌的钢结构防火涂料具有很好的聚类作用。利用5个品牌的各25个样品建立校正模型,用余下5个品牌的各5个样品,共计25个样品进行外部验证,通过未知样品光谱的主成分得分值计算其与校正模型中每个品牌的马氏距离值,实现未知样品的品牌鉴别。建立的定性分析模型对未知样品的外部验证正确率达到100%。说明该分析方法能够快速准确的鉴别钢结构防火涂料品牌,并为市场规范提供技术参考。
[Abstract]:In order to quickly identify different brands of fire resistant coatings for steel structure in circulation, a method of rapid identification of fire resistant coatings for steel structures by near infrared spectroscopy was proposed. The spectral curves of fire resistant coatings of different brands of steel structures were obtained by using raster scanning near infrared spectrometer and diffuse reflectance spectroscopy of near infrared, and the spectral data were transformed by standard normal variable (standard normal variate transformation,SNV. Norris second order derivation and other optimization treatment. Using principal component analysis (principal component analysis,PCA) to cluster the steel structure fire resistant paint brand, the cumulative variance contribution rate of the first five principal components has reached 99.791.The contribution rate of the first five principal components is 99.791g. The scores of PC2 and PC3 脳 10 can be used to map all the modeling samples in three dimensional space, and they have good clustering effect on different brands of fire resistant coatings of steel structure. A calibration model was established using 25 samples from each of the 5 brands, and 25 samples from each of the remaining 5 brands were used for external validation. The Mahalanobis distance of each brand in the calibration model is calculated by the principal component score of the unknown sample spectrum, and the brand identification of the unknown sample is realized. The external validation accuracy of the qualitative analysis model for unknown samples is 100%. This analysis method can quickly and accurately identify the steel structure fire resistant paint brand and provide technical reference for the market specification.
【作者单位】：公安部天津消防研究所;
【基金】：公安部消防局重点攻关项目(2013XFGG05) 公安部天津消防研究所科研计划项目(2013SJ-A-06)资助
【分类号】：O657.33;TQ637

【共引文献】