POD preprocessing of IR thermal data to assess heat source distributions

Abstract

Infrared thermography is a useful imaging technique for analyzing the thermomechanical behaviour of materials. It allows, under certain conditions, surface temperature monitoring and, via a diffusion model, estimation of heat sources induced by dissipative and/or thermally coupled deformation mechanisms. However, the noisy and discrete character of thermal data, the regularizing effect of heat diffusion and heat exchanges with the surroundings complicate the passage from temperature to heat source. The aim of this paper is to show that the prior use of reduced-basis projection of thermal data improves the signal-to-noise ratio before estimating the heat source distributions. The reduced basis is generated by proper orthogonal decomposition (POD) of physically-admissible thermal fields. These fields are solutions of ideal diffusion problems related to a set of putative heat sources. preprocessing is applied to different direct methods (finite differences, spectral solution, local least-squares fitting) already used in the past. The gain of this preprocessing is determined using a numerical penalizing benchmark test. The methods are finally compared using data extracted from a dynamic cyclic test on a pure copper specimen.