Fuzzy Adaptive Genetic Algorithm for Improving the Solution of Industrial Optimization Problems

Abstract In the industrial and manufacturing field, many problems require the tuning of the parameters of complex empirical or theoretical models by means of the exploitation of data. In some cases the use of analytical methods for the determination of such parameters is not applicable, thus heuristic methods are employed. One of the main disadvantages of the use of these latter approaches is the risk of converging to a sub-optimal solution due to characteristics of the search surface which is determined both by the model and the data available for the tuning. In this paper the use of a novel type of genetic algorithms is proposed to overcome this drawback. This approach exploits a fuzzy inference system that controls the search strategies of the genetic algorithm on the basis of the real-time status of the optimization process. In this paper this method is tested on three problems drawn from the steel-making industry that put into evidence the improvement of the capability of avoiding the local minima of the optimization problem and the acceleration of the search process.