Development of Multiobjective Design Simulator for Livestock Feed Formulation Using a GA

作为利用遗传算法开发畜禽饲料配方多目标设计模拟器的初步步骤，对禽、猪、牛三种饲料进行了同时优化设计试验。在规定营养成分含量的约束条件下，设计了玉米、谷物高粱、脱脂米糠等12种不同饲料原料的日粮。利用的营养成分包括粗蛋白质、可消化粗蛋白质、总消化营养和其他。对遗传算法进行了改进，使上述三种饲料配方的原料成本最小化。采用12种原料在3种条件下对不同饲料配方进行了7次模拟。首先，对禽、猪、牛饲料的各种原料进行充分准备，不加限制地进行混合设计。其次，对玉米和/或玉米面筋饲料等指定原料的总量及其用于三种饲料的可用量进行了限制。第三，对玉米和/或玉米蛋白粉饲料等特定原料的总量进行了限制，但对三种饲料的可用量没有限制。研究表明，本研究开发的多目标设计模拟器具有实现三种饲料配方在不同条件下同时优化设计的可行性。

As a preliminary step to develop a multiobjective design simulator for livestock feed formulations using a genetic algorithm, simultaneous optimum design of three kinds of feed, such as poultry feed, swine feed and cattle feed were tried. Twelve different rations of feed ingredients such as corn, grain sorghum, defatted rice bran, and others were designed under constraint conditions of specified nutrient contents. Examples of nutrient contents utilized were crude protein, digestible crude protein, total digestive nutrition and others. Genetic algorithm has been improved upon to minimize the raw material cost of three feed formulations mentioned above.Seven simulations were carried out for different feed formulation using twelve raw materials under three conditions. Firstly, the volume of each raw material for poultry, swine and cattle feed was sufficiently prepared for mix design without restriction. Secondly, the total volume of specified raw materials such as corn and/or corn gluten feed and their usable volumes for three types of feed were restricted. Thirdly, the total volume of specified raw materials such as corn and/or corn gluten feed was restricted but their usable volumes for three types of feed were not restricted.Based on this study, it was cleared that the multiobjective design simulator which was developed in this research has the feasibility to attain the simultaneous optimum design of three kinds of feed formulation under various conditions.