An Improved Self-Adaptive Constraint Sequencing approach for constrained optimization problems

Real life optimization problems involve a number of constraints arising out of user requirements, physical laws, statutory requirements, resource limitations etc. Such constraints are routinely evaluated using computationally expensive analysis i.e., solvers relying on finite element methods, computational fluid dynamics, computational electro magnetic, etc. Existing optimization approaches adopt a full evaluation policy, i.e., all the constraints corresponding to a solution are evaluated throughout the course of search. Furthermore, a common sequence of constraint evaluation is used for all the solutions. In this paper, we introduce a novel scheme for constraint handling, wherein every solution is assigned a random sequence of constraints and the evaluation process is aborted whenever a constraint is violated. The solutions are sorted based on two measures i.e., the number of satisfied constraints and the violation measure. The number of satisfied constraints takes precedence over the amount of violation and the most efficient sequence of constraint evaluation is evolved during the course of search. The performance of the proposed scheme is rigorously compared with other state of the art constraint handling methods using single objective inequality constrained test problems of CEC-2006 and CEC-2010. The constraint handling approach is generic and can be easily used for the solution of constrained multiobjective optimization problems or even problems with equality constraints. A bi-objective welded beam design problem, a tri-objective car side impact problem and an equality constrained optimization problem of CEC-2006 is solved as an illustration. The results clearly highlight potential savings offered by the proposed strategy and more importantly the paper provides a detailed insight on why such a strategy performs better than others.