Thermographic Signal Reconstruction (TSR) is a processing technique used in thermal nondestructive testing. TSR provides good qualitative results allowing the detection of hidden defects, the compression of data for processing, and the filtration of high frequency noise. To improve the quantitative characterization capabilities of TSR, we use Artificial Neural Networks given their easiness of implementation, low sensibility to noise, and abilities for learning and generalization. To illustrate this, we analyze the results of several Multilayer Perceptron Artificial Neural Networks that were trained with the coefficients acquired after the application of the TSR to infrared sequences. The latter sequences were obtained from simulations of nondestructive experiments on glass reinforced plastic fiber samples containing air defects.

The CVSL seminars are held on Fridays at 11:30 a.m. in room PLT-2700.