Modelagem e otimização de experimentos para o tratamento térmico de recozimento: um estudo com o algoritmo simulated annealing

Abstract

This dissertation presents a study for experiments modeling applied for a heat treatment industrial process. The driving force to this work have raised from usual difficulties while low-alloy steels annealing industrial processes get on, trying to find out temperatures which the superficial hardness measurements reaches low values enough and suitable for the following stages, in special focus, the machining. At the first time, some experiments have been done with several steel grades, where superficial hardness is achieved as a function by annealing temperature, percent carbon content and percent silicon content of the samples. There was purposed a quadratic model for modeling the hardness as a function of that three parameters. The parameter estimation of the purposed model was done by the simulated annealing algorithm, a metaheuristic for global optimization which can be as the same as annealing of a solid material. Starting at the purposed model, the inverse problem was solved, where the estimated annealing temperature was achieved by the carbon and silicon percent contents and the suitable hardness.