Solving constrained global optimization problems without penalty parameters

Abstract

Purpose

To propose a new method for solving constrained global optimization problems using a method that consists of transforming a constrained global optimization problem into an unconstrained one without using any penalty coefficients.

Design/methodology/approach

Use of an unconstrained global optimization method such as the Alienor method which has been adapted for several variables.

Findings

Use of the adapted Alienor method allowed the solution of the transformed problem with little difficulty.

Research limitations/implications

Transforms the original objective function into a new one involves the introduction of some extra parameters. Cannot guarantee the convergence to a global solution of the original problem. The simple described approach, provides new possibilities.

Practical implications

No further parameters introduced in this new approach, and no conditions or hypotheses are imposed on the objective function or on the constraints.

Originality/value

New method of transforming a constrained problem into an unconstrained one, with use of proven Alienor method adapted to several variables.

Citation

Konfe, B.O., Cherruault, Y. and Some, B. (2005), "Solving constrained global optimization problems without penalty parameters", Kybernetes, Vol. 34 No. 7/8, pp. 1090-1103. https://doi.org/10.1108/03684920510605902