摘要
本文介绍了国内迄今未见报道的偏最小二乘方法(PLS)在痕量多组分分析中的基本原理和方法。以笔者新近合成的新有机试剂4-(2-苯并噻唑偶氮)邻笨二酚(BTAPC)为显色剂,在二苯胍存在下,与痕量钼(VI)、钨(VI)在最佳条件下形成三元有色配合物。根据PLS法、共轭梯度法(CG)和线性规划法(LP)的计算原理,用BASIC语言编制程序,通过电子计算机可同时分光光度测得二组分的含量。用PLS法、CG法和LP法处理十组含痕量Mo(VI)、W(VI)的模拟试样,相对标准偏差对Mo(VI)而言分别为2.6%、3.6%和4.0%,对W(VI)而言分别为5.1%、8.6%和10.7%。与CG法和LP法相比,PLS法的运行时间短,结果更准确可靠,完全满足一般分光光度法测定痕量组分的误差要求,尤其适用于处理成批试样,为带微处理机的分光光度计提供了一种新的计算测定方法。
The basic principle and method of partial least squares (PLS) in the analysis of traces of multicomponents, of which no information has ever been available in domestic literature, are presented in this paper. 4-(2-Benzothiazolylazo)pyrocatechol (BTAPC), a new chromogenic reagent recently synthesized by the authors, was used to form the colored ternary complexes under optimum conditions with traces of Mo(VI) and W(VI) in the presence of diphenyl guanidine. Based on the computation principles of PLS, conjugate gradient method (CG) and linear programming method (LP),. a computer was applied and a program designed using BASIC language to determine spectrophotometrically the content of two components at the same time. Relative standard deviations were 2.6%, 3.6% and 4.0% respectively for Mo(VI) and 5.1%, 8.6% and 10.7% respectively for W(VI), when PLS, CG and LP were used to process ten sets of analogue samples respectively. Compared with CG and LP, PLS has shorter operating time, and more accurate and reliable results, which is consistent with the error requirement for normal spectrophotometry of traces of components, and is especially applicable to processing samples in batches. PLS provides a new method of computing determination for the spectrophotometers with the microprocessors.
出处
《南京工业大学学报(自然科学版)》
CAS
1989年第1期15-21,共7页
Journal of Nanjing Tech University(Natural Science Edition)
关键词
偏最小二乘方法(PLS)
多组分同时分光光度测定
4-(2-笨并噻唑偶氮)邻苯二酚
钼钨
二苯胍
共轭梯度法
线性规划法
partial least squares method (PLS)
simultaneous spectrophotometric determination of multicomponents
4-(2-benzothiazolylazo)pyrocatechol
molybdnum
tungsten
diphenyl guanidine
conjugate gradient method
linear programming method