Multivariate Korrektur des Temperatureinflusses in der NIR-spektroskopischen Materialfeuchtebestimmung

The market of process measurement technology grows continuously and the need of universal measurement technologies for quality control increases steadily. The most important parameter of an industrial process control is moisture content. It plays an important role in processes like drying, mixing, production of flour, roasting of coffee, etc. The best measurement technolgy for determination of moisture content of a product allready in the runing process is the NIR-spectroscopy. The analysis of NIR-spectra is very complex and needs a well-founded knowledge about the proceeding and the interpretation of an analysis. The biggest and not yet solved problem coming along with this technology is the variance of the product temperature during the process. These temperature variances can not only lead to wrong measurement results but also can even reduce the whole analysis to absurdity. This work has lead to a solution of the temperature problem in the NIR-spectrocopic measurement technology. Therefor the most important side effects which can lead to an erroneous measurement were investigated and characterised. Based on this knowledge it was shown which influences the temperature variance has on the prediction of a chemometric model for the analysis of a NIR-spectrum. For this characterisation models were developed whose reaction to the temperature variance were compared. The up to now well established but still very inaccurate approaches for the correction of sample temperature were investigated and it was shown that the results of these methods and there conditions should be improved. A novel solution of the temperature problem was found. Therefor the Piecewise Direct Standardization was integrated in a method such that spectra of the calibration data set were changed in there spectral shape by the tranfer algorithm as if they were allready measured at the measurement temperature to be corrected. The novel method yields on the one hand results as if the variance of the temperature never appeared. On the other hand this new developed universal method has the big advantage that the user of this technology does not has to change his established way of correcting the temperature effect. Contrawise he can still use his well established working procedures which he created for his process for correcting the temperature effect. Just another intelligent an highly complex analysis step will be integrated in the analysis process. The user will only realise the better results, the maximised unversality and the much higher reliability of his process measurement technolgy.