A Method of Multiobjective Optimization Using a Genetic Algorithm and an Artificial Immune System


The immune system has pattern recognition capabilities based on reinforced learning, memory, and affinity maturation interacting between antigens (Ags) and antibodies (Abs). This article deals with an adaptation of artificial immune system (AIS) into genetic-algorithm (GA)-based multi-objective optimization. The present study utilizes the pattern recognition from an AIS and the evolution from a GA. Using affinity measures between Ags and Abs, GA-based immune simulation discovers a generalist Ab that represents the common pattern among Ags. Non-dominated Pareto-optimal solutions are obtained via GA-based immune simulation in which dominated designs are considered as Ags, whereas non-dominated designs are assigned to Abs. This article discusses the procedure of identifying Pareto-oplimal solutions through the immune system-based pattern recognition. A number of mathematical function problems that are described by discontinuity or disconnection in the shape of Pareto surface are first examined as test examples. Subsequently, engineering optimization problems such as rotating flywheel disc and ten-bar planar truss are explored to support the present study.