Modelo de regressão por processos gaussianos aplicado a problemas de otimização estrutural via meta-heurísticas

Abstract

The structural optimization is an important enginnering area that is being explored, specially in the industrial sector, because of its great gamma of problems that impact directly the economy and the production process as a whole. The creation of methods that solve such problems in a more effective manner and/or produce more trustworthy results has gained importance in the most diverse knowledge areas. Following this thought, we propose a Gaussian Process Regression Model study and its use as a surrogate model in the structural optimization process via Diferential Evolution (DE). From the construction of the theoretical framework that promotes the interdisciplinarity between the two areas (Statistics and Computer Science), we apply it in a real problem of enginnering, in five different cases of structural arrangements. Furthermore, through the R software, used in the previous statistical analysis and tests, and the programming language Python, used for the development of optimization process through Diferential Evolution assisted by the statistical surrogate model, we present the test results and compare them to the classical method of the DE.