Model-Based Inversion of Flash Thermography Nondestructive Evaluation Measurements of Composites

Abstract

Flash thermography is a nondestructive evaluation (NDE) technique that finds flaws such as delaminations in composites through the blockage of heat flow and resulting change in surface temperature. In this presentation we will discuss a model-based inversion approach to identifying and localizing damage from a thermal image sequence in a composite material. The approach is built on the observation that the influence of defects in blocking heat flows on surface temperature can be approximated by replacing the defect with a heat source, and using the well known thermal Green's function to predict the resulting surface temperature. Linear inversion is then used to represent the measured thermal image sequence as a sum of these Green's functions. The method has been tested in both simulation and experiment. In general, this method seems to give reasonable approximations of the first subsurface insulating layer. Inversion results behind that layer are not meaningful and often not physical (for example, negative source intensity). As would be expected for any inversion process, the inversion process also results in noise gain. Nevertheless, this process offers for the first time a direct inversion of flash thermography measurement data to localize the source of observed contrast in regimes where 3D heat flow is significant.