ON THE METAHEURISTIC OPTIMIZATION ALGORITHMS IN THE STRUGGLE FOR THE HOT FLOW CURVE APPROXIMATION ACCURACY

Abstract

A hot flow curve approximation performed via flow stress models as well as artificial neural networks requires precisely estimated constants. This estimation is in the case of highly-nonlinear issues often solved via gradient optimization algorithms. Nevertheless, by natural processes or physical laws inspired approaches (metaheuristic algorithms) are also of high interest. In the submitted manuscript, three selected metaheuristic algorithms were compared under the approximation of an experimental hot flow curve dataset via the wellknown Hensel-Spittel relationship. One often used gradient algorithm was also included into this comparison. Results have showed that the metaheuristic algorithms are useful if such complex approximation model is applied and no estimate of material constants from a previous approximation issue is used. On the other hand, if this estimation exists, the gradient algorithms should provide a better solution.