Optimization of Extrusion Process Variables Using a Genetic Algorithm

Abstract

Optimization of process variables during single screw extrusion cooking of a fish and rice flour blend was carried out using a genetic algorithm (GA). Second-degree regression equations were developed for the extrudate properties such as expansion ratio (ER), bulk density (BD), hardness (H) and water solubility index (WSI) in terms of the independent variables: barrel temperature (°C), screw speed (rpm), fish content (%) and feed moisture content (%). These equations were used as the objective function to find the optimum process conditions separately for each (individual) and commonly for all (common) extrudate properties. The regression equations of ER and WSI were maximized and BD and H were minimized for both types of optimum process conditions. A maximum population of 100, and 100 iterations, was sufficient for successful convergence. A high barrel temperature, of about 200°C, and screw speed of 110 rpm were identified as both individual and common optimum process conditions for all the extrudate properties except for WSI. Fish content and feed moisture content were identified as two interacting process variables. Under individual optimum process conditions, maximum ER and minimum hardness required a low fish content of 5% (w/w) and feed moisture contents of 60 and 40%, respectively, and minimum BD and maximum WSI required high fish content of 41–45% and medium moisture content of about 40%, respectively. However under common optimum process conditions all four extrudate properties were optimized at a high fish content of 41–45% and medium moisture content of 40%. The experimental extrudate properties matched the values predicted for ‘common optimum’ conditions more closely than those for individual optimum process conditions.