Model Based Inversion for Pulse Thermography

Abstract

We present a model-based inversion approach to the interpretation of flash thermography data. Traditionally, flash thermography has been image sequences have been hard to interpret in the presence of lateral heat flows. This is a particularly significant problem for composite laminate materials where heat conducts readily in the fiber planes. We use a linear inversion approach to represent the observed surface temperature as a superposition of contributions from buried reflectors. We investigate the additional complexities of considering cases where the specimen is curved, changing the nature of the heat flow. By solving a large linear inversion problem we are able to determine the strengths of the buried reflectors and create a concrete representation of observed defects. Results from this approach are compared to more traditional analysis approaches for flash thermography.