基于PLSDA对中红外透射法压片过程中溴化钾与样品质量配比的研究

The application of PLSDA to investigate the choice of the tabletting process in the mid-infrared transmission method of KBr and the ratio of the quality of the sample

在中红外透射法测量样品时,基质溴化钾(KBr)与样品的合理压片过程能有效避免量测光谱出现克里斯坦森效应。但由于缺乏足够的量测光谱的先验信息,压片时KBr和样品之间合理质量配比的选择在很大程度上取决于经验值,可能造成样品的真实光谱存在信息差异。为解决该问题,本文应用偏最小二乘判别分析(PLSDA)的模式识别方法在没有先验信息的情况下,对样品原始光谱及处理光谱数据进行差异信息提取建立模型,探讨中红外透射法中KBr和样品之间合理质量配比。对四组实验数据的分析结果均表明,在样品与KBr质量配比的范围为3:194～3:204时,KBr的量对实验结果影响较小,为压片中较好的质量配比。本文为中红外透射法测定样品提供一定参考意义,同时也是PLSDA在实验条件优化方面的新应用。

Reasonable tabletting process of KBr and sample in the mid-infrared transmission technology is indispensable when the measured spectra are subject to Christensen effect. However, due to the lack of sufficient prior information of the raw spectra, choice of reasonable ration between quality of KBr and sample depends largely on experience value. This might degrade certain qualities of spectra. In order to tackle this problem, in this paper, no prior information of the data and expertise value are required, prtial least squares discriminant analysis （PLSDA） pattern recognition method was adopted to extract inherent information differences from the original spectral data and preprocessed specrtra of measured sample with different quality ration of KBr and sample, hereby exploring the optimum ratio in the mid-infrared transmission technology. The results of the four real experiment data sets showed that quality ratio range of the sample and KBr as 3：194 to 3：204 has little effect to the spectral information, which also provided a reference for future experiment. Meanwhile, it was a new application of PLSDA method on optimization of experimental conditions.