This work is written in the philosophy that a university research laboratory has to play an active role in the study, and implementation of quality control procedures.; The first chapter of this work consists of a short literature study about quality and the implementation of quality assurance in the analytical laboratory. At the end of the chapter, the role of chemometrics and the importance of it in a chemical analysis is explained.; In the second chapter, the importance of the use of standard operating procedures (SOP) is explained. The combination of SOPs and fault trees is made, for troubleshooting purposes. The use of fault trees is important for the development of the troubleshooting SOP. On the other hand, the fault tree gives a visual aid to understand and implement the SOP.; In the third chapter, principal component analysis is applied on two data sets. During calibration of the EDXRF instrument, it was noticed that sometimes an increase in the number of counts of the elements appeared. A short time stability study was performed. This results in the first data set which constitutes subsequent measurements of a standard (GaP). From the univariate approach, one concludes that an increase in the Ga intensity appeared. From the multivariate approach, (PCA), it was concluded that this effect was compensated by a decrease of the spectrum background in the channels located before the Ga peaks. A compensation effect was present. The second data set consists of EDXRF measurements of Chinese porcelains. This data set is also used in the fifth chapter on partial least squares. There we will explain the relation found between the age of an object and its elemental composition.; In chapter IV we describe a new technique, in order to use whole spectra, to control the sensitivity stability of spectroscopy equipment. This theory, named the theory of control boxes, has some advantages over the univariate approach. Besides sensitivity changes, one is also able to detect gain or zero channel changes. This is important if one wants to implement full spectrum calibration procedures.; When spectra are complex, spectrum evaluation using least squares fitting fails. Partial least squares (PLS) can then be used. It will be shown in this chapter, that a number of parameters that normally degrade the quality of the analytical results are automatically taken care of when using this multivariate calibration method. This opens interesting possibilities for integrating quality control procedures into the analysis process. The second data set consists of the Chinese porcelains studied in chapter III. A relation between period of fabrication and the elemental composition is found. The resulting calibration coefficients are on close approximation with the results found in the literature. The advantage of the multivariate approximation is the robustness of the system. (Abstract shortened by UMI.)
展开▼
