Optimization of Aqua Feed Pellet Properties using Genetic Algorithm

Abstract

Aqua feed properties like water stability (WS), expansion ratio (ER) and true density (TD) play a vital role in establishing the quality. These properties are dependent on the extrusion process variables like screw speed (rpm), L/D ratio, barrel temperature (°C) and feed moisture content (%).The second degree regressions equations were developed for the feed pellet properties WS, ER and TD in terms of process variables. The analysis of variance (ANOVA) indicated that among the process variables screw speed was least significant for WS and TD whereas for ER, L/D ratio was found to be non-significant. Further the process variables were optimized for maximum WS, ER and TD using genetic algorithm (GA). The regression equations of WS, ER and TD with only significant terms were used as objective functions to find the optimum process conditions separately for each (individual) and commonly for all (common) for the maximization of WS, ER and TD using genetic algorithm (GA). A maximum population of 100 and 100 iterations was found to be sufficient for successful convergence of the both optimums. As both individual and common optimum process conditions an L/D ratio of 12-13.5 and medium feed moisture content of 30-33 % were necessary for maximization of WS, ER and TD. Barrel temperature (°C) and screw speed of (rev min-1) were found to be the two interacting variables. The common optimum process condition of barrel temperature of about 90°C has corroborated with the individual optimum process condition for WS but not for ER and TD. A common optimum process condition of screw speed of 105.84 rev min-1has not corroborated with the individual optimum for WS, ER and TD where a screw speed of 90-95 rev min-1 was necessary for WS and TD and for ER it was about 60.33 rev min-1. The screw speed followed by barrel temperature had influenced the aqua pellet properties the most. There was a marginal difference between the predicted values of the ER and TD (4.79 and 1.30 %) as both individual and common optimum process condition but for WS there was a difference of 10.28 %.