The Processing of Active Infrared Thermography Data by a Hybrid Neural Algorithm for the Evaluation of Thermal Barrier Coating Thicknesses

Today, with the rising cost of fossil combustible, thermal efficiency is a necessity in the design and the manufacture of all gears. In gas turbines, the thermal barrier coatings increase the thermal efficiency by providing a thermal protection to the parts operating at very high temperatures and avoids also accidents that have serious consequences, especially for airplanes and ships. Controlling the thickness uniformity of these thermal barrier coatings is very important to have a good thermal efficiency of the alloys and a good performance. In this work, using the data of pulsed and lock-in infrared thermography controls, a neural algorithm is proposed to evaluate the thicknesses of thermal barrier coatings irregularly deposited on alloys. The neural algorithm combines the neural network quality and the genetic algorithm advantages. The neural network is trained using the phases calculated by the Fourier transforms of the temperatures. The genetic algorithm is used to optimize the neural network by searching the initial weights matrix inducing a rapid convergence to the optimal solution. This method has improved the network learning by minimizing the mean squared error and the number of iterations. The obtained results by the neural algorithm have shown that both thermal control methods are very effective in estimating the thermal barrier coatings. The thicknesses have been estimated with uncertainties less than 5%.