A strategy for design optimization of nonlinearly constrained problems is presented. The strategy combines techniques used in production rule systems with an optimization procedure dealing with local monotonicity and with sequential quadratic programming techniques. The production rule system is based on the observations obtained by applying the optimization program to different classes of test problems. The observations made are incorporated in the rule-based system in such a way that if its premise is true, then the action part of the rule is concluded. This is the first attempt at developing such a microcomputer rule-based system for design optimization.