Estimation of thermal properties at high temperatures through the application of radial basis function interpolation in an inverse heat transfer problem

Abstract

© 2024 Elsevier LtdThe thermal characterization of materials at high temperatures is crucial to various modern engineering applications. However, direct experimental measurements under severe conditions can be complex, expensive and offer several other disadvantages. Hence, in this work, a novel Radial Basis Function (RBF) based inverse method is proposed as an alternative to solve nonlinear Inverse Heat Transfer Problems (IHTPs). Here, a proof of concept is performed by estimating a two parameters exponential function describing the specific heat of an AISI 1020 steel submitted to LASER Beam Welding (LBW). An inverse algorithm combined with an RBF interpolation algorithm enables an enhanced search domain scan. A least squares objective function with Future Time Regularization (FTR) is implemented to govern the estimation. The algorithms are sequentially run and refeed to refine the minimization region through adjustable search factors. A Finite Volume Method (FVM) thermal model was implemented through a highly parallelized inhouse CUDA-C code, run on an Nvidia Geforce® RTX™ 3090. A verification was performed using three commercial solutions. The method's efficiency was demonstrated with both noiseless and variable standard deviation data. The approach is less sensitive to local minima than previous Quadrilateral Optimization Method (QOM), with estimation errors below 1.0 % in nearly all cases.

Keywords