A CUSTOMIZABLE COMPUTATIONAL STRATEGY FOR EVALUATION OF ENGINEERING PROBLEMS USING SURFACE INTERPOLATION TECHNIQUES
Bruna L. B. Santos1; Catarina N. de Araújo1; Ricardo A. Fernandes1; William W. M. Lira1
1 LCCV/UFAL
doi:10.20906/CPS/CILAMCE2017-0985
Resumo
This work presents the development of a configurable computational strategy for evaluation of engineering problems using multivariable interpolation techniques. Technological advances intensify the search for numerical methods to solve increasingly complex problems. Depending on the complexity involved, numerical solutions can demand high computational costs, requiring much time to execute robust analyzes. The adopted strategy is presented as an alternative numerical technique to quickly estimate the solution of these problems. Polynomial response surfaces are generated from certain known configurations, and can be evaluated to provide instant approximations of the behavior of other configurations. It's assumed that exists a relation between the variables and the interested result, which can be approximated by a polynomial function. The methodology adopted is divided in four steps: a) data generation relating several variables combinations and their respective responses, which are obtained through experimental or numerical tests (known configurations); b) response surface generation applying multivariable interpolation techniques to the data obtained in the previous step; c) development of a computational tool that is configured according to the problem data and d) evaluation of unknown configurations using the computational tool developed in the previous step. Thus, the strategy allows the user to configure the computational environment with a specific problem data from numerical and/or experimental tests, allowing rapid estimates for situations not considered during the generation of the original data (used to generate the interpolation surfaces). In order to validate and illustrate the proposed strategy, case studies of engineering problems are presented.
Palavras-chave: Response surface; Customizable computational tool; Multivariable interpolation