Extension of a hybrid Genetic Algorithm for nonlinear programming problems with equality and inequality constraints

As an extension of the hybrid Genetic Algorithm-HGA proposed by Tang et al. (Comput. Math. Appl. 36 (1998) 11), this paper focuses on the critical techniques in the application of the GA to nonlinear programming (NLP) problems with equality and inequality constraints. Taking into account the equality constraints and embedding the information of infeasible points/chromosomes into the evaluation function, an extended fuzzy-based methodology and three new evaluation functions are proposed to formulate and evaluate the infeasible chromosomes. The extended version of concepts of dominated semi-feasible direction (DSFD), feasibility degree (FD₁) of semi-feasible direction, feasibility degree (FD₂) of infeasible points ‘belonging to’ feasible domain are introduced. Combining the new evaluation functions and weighted gradient direction search into the Genetic Algorithm, an extended hybrid Genetic Algorithm (EHGA) is developed to solve nonlinear programming (NLP) problems with equality and inequality constraints. Simulation shows that this new algorithm is efficient.Scope and purposeNon-linear Programming (NLP) problems with equality and inequality constraints is an important type of constrained optimization problems. Genetic Algorithm (GA) is one of the well known evolutionary computation techniques. In the application of GA to NLP problems, chromosomes randomly generated at the beginning and/or generated by genetic operators during the evolutionary process usually violate the constraints, resulting in infeasible chromosomes. Therefore, the handling of system constraints, particularly the nonlinear equation constraints, and the measurement and evaluation of infeasible chromosomes, are major concerns in GA. Penalty strategy in the construction of fitness function is commonly used to evaluate the infeasible chromosomes in some traditional AG methods. However, this approach essentially narrows down the search space by eliminating all infeasible chromosomes from the evolutionary process, and it may reduce the chances of finding better candidates for the global optimization. In particular, it absolutely ignores the information carried by the infeasible chromosomes itself. Therefore, formulating the infeasible chromosomes by embedding the relevant information into the evaluation function are important when applying GA to NLP.As an extension of the Hybrid Genetic Algorithm-HGA proposed by Tang et al. (1998), this paper focuses on the critical techniques in the application of GA to NLP problems with equality and inequality constraints. Taking into account the equality constraints and embedding the information of infeasible chromosomes into the evaluation function, an extended fuzzy-based methodology and three new evaluation functions are designed to formulate and evaluate the infeasible chromosomes. By introducing an extended version of the concepts of dominated semi-feasible direction (DSFD), feasibility degree (FD₁) of semi-feasible direction, feasibility degree (FD₂) of infeasible points ‘belonging to’ feasible domain, an extended hybrid Genetic Algorithm (EHGA) is developed for solving NLP problems with equality and inequality constraints.