In this paper, buildings are composed as three-dimensional "form" by using a new systems technique of intelligent artificial life. Cubic cells which are basic spaces of three-dimensional forms are proliferated to specified directions according to surrounding conditions. At every proliferation step various architectural evaluations are performed, and degrees of adaptation to environments are obtained. The purpose of this paper is to compose a multi-optimization-system with evolutionary evaluation in which optimal forms of architectural SD-structures and optimal membership functions of evaluations for phenotypes can be obtained by using Genetic Algorithm(GA) originated by Holland and Goldberg. First, a model of GA is employed, and surrounding conditions are coded as genotype. Next, it's phenotypes are estimated, and employed genetic processing of crossover and mutation is performed, so that genotypes with more excellent degrees of adaptation emerge. Finally optimal forms with optimal membership functions for evaluations can be obtained.