Physics-based thermal noise effect reduction in sonic IR crack length estimation

Using heat diffusion models in sonic infrared (IR) inspection technology that captures arbitrary frictional heat generation function along a crack lends itself to estimation of crack length using thermal images alone. This study introduces novel physics-based approach that relies on the governing heat diffusion model linearity and superposition of point heat sources along crack to reduce thermal noise effect in crack length estimation. It utilises thermal images and their timestamps in sonic IR inspection to deliver spatiotemporal average representation of the aforementioned point heat sources to reduce thermal noise effect in crack length estimation. The proposed methodology is tested against two-dimensional and three-dimensional finite element simulations with an arbitrary frictional heat generation function along crack. Virtual random thermal noise levels ranging between 50 and 200 mK, which are higher than IR cameras Noise Equivalent Temperature Difference values, are added to simulated results to reflect realistic and drastic thermal noise levels. Spatiotemporal averaging using combined sets of different thermal images captured at various times in single sonic IR inspection delivered about 16% total uncertainty in the estimation of a crack length at 95% confidence level based on 60 experiments. Similar uncertainty level is obtained using independently reported sonic IR experimental data.