Use of multivariate techniques in quality control of kerosene production

The aim of this work is to present a new strategy for daily quality control in kerosene production. Exploratory data analysis of kerosene batches monitored during 5 years allowed extraction of nine 'essential' variables from a total set of 26. A new 5 year period was used for confirmatory purposes, revealing good agreement between conclusions from both data sets. Once essential variables were defined (including chemical considerations), alternative analytical methodologies were developed based on combining multivariate regression approaches (PLS, PCR, MLR) and FT-i.r. spectroscopy. Promising results are presented so that progress can be made from sequential univariate tests to multivariate approaches allowing better throughput, lower delay time and increased laboratory performance. Copyright (C) 1996 Elsevier Science Ltd.