Physics-Based Sonic IR Crack Length Estimation using Thermal Images Alone

The absence of analytical solutions describing a crack frictional heat generation and diffusion in sonic infrared (IR) inspection technology makes it rather difficult to estimate a crack length from thermal images alone. This study presents the recent development in theoretical-based techniques assuming uniform, arbitrary and point frictional heat generation functions along the crack and how they lend themselves for crack length estimation. The different forward heat diffusion models are validated in close comparison with finite element (FE) simulations for different heat generation functions. Moreover, the capability in retrieving a crack arbitrary heat generation function and estimating a crack length from simulated thermal images alone is validated with and without the virtual addition of thermal noise. It demonstrates the benefits of applying the principle of superposition of predefined heat generation functions due to the linearity in the governing heat diffusion model while retrieving the heat generation function from thermal images. Targeting a peak temperature change between 0.2 and 1.6 K with the addition of different random noise levels for different crack lengths can deliver up to 20% uncertainty in crack length estimation at 95% confidence level. The application of the proposed point heat generation function along the crack underestimated a crack length by 10% over independently measured sonic IR thermal images. This illustrates the benefits and potential capabilities from advancing this approach in the future.