1994

Vainionpää, Jukka

The extrusion cooking of cereals and cereal-based mixtures was modelled and examined by combining the graphical and numerical methods in the analysis of the response surface models. The study included the diversified application of extrusion cooking, such as the processing of intermediate model food systems, as well as the study of the extrusion of carrot pulp in the mixture with wheat flour. The screw wear was modelled regarding new and worn screw elements as a qualitative variable. Correspondingly, the behaviour of wheat, rye, barley and oat flour in the extrusion cooking was modelled with both the cereal type and the screw configuration as qualitative variables. In parallel to graphical presentation, numerical methods were introduced to analyze the response surface models. The graphical approach and the inherent contour plots provide the detailed and complete view on the problems with two independent variables only. However, with increasing numbers of independent variables the analysis easily becomes laborious and increasingly complex, especially when multiple responses are introduced. Accordingly, the advantages of the mathematical methods become evident. A new approach to canonical analysis eliminating the need for the stationary point to be the clear optimum was successfully employed to determine an operation point of extrusion processes without extensive graphical work. Similarly, the simultaneous solution of second-order models was used in finding operation points numerically, starting from a product profile defined by several properties. Both proposed approaches are quite straight-forward procedures that save a lot of work in selecting an appropriate coordinate system also if graphical presentation is of importance. Nevertheless, the mathematical methods do not render the graphical methods unnecessary. Instead, together with graphical presentation, the mathematical tools provide an integrated technique for the analysis of the response surface models. This technique greatly improves the exploitation of the extensive amount of information compiled on the models.