Defects are an inevitable occurrence in the manufacturing of composite components, leading to deviations from intended specifications and impacting performance. Nondestructive testing (NDT) techniques provide a noninvasive alternative to destructive methods for detecting and quantifying defects in CFRP materials. However, challenges remain in applying NDT effectively, particularly in sizing defects. This study evaluates the potential of simulation for predicting defect sizing accuracy using front and back step-heating thermography on thin CFRP materials. Finite-element-based software is used to simulate temperature distribution on the model surface, with defects consisting of voids and polyethylene inserts. The derivative and full-width half maximum (FWHM) methods are employed to measure defect size. For an accuracy threshold of 20% error, the minimum detectable void size in front heating is 4 mm in simulations and 6 mm in experiments. For polyethylene defects, the minimum detectable size is 6 mm in simulations, while no defect sizes meet the threshold in experiments. In back heating, both void and polyethylene defects have a minimum detectable size of 4 mm in both simulations and experiments. Collected thermal images are further analyzed using an artificial neural network (ANN). Robust principal component analysis (RCPA) summarizes the thermography data, and image segmentation is then applied to the resulting image. Initial results will be presented in the conference.