Search

International audience; The main goal of this paper is the estimation of thermal resistive fields in multilayer samples using the classical front face flash method as excitation and InfRared Thermography (IRT) as a monitoring sensor. The complete inverse processing of a multilayer analytical model can be difficult due to a lack of sensitivity to certain parameters (layer thickness, depth of thermal resistance, etc.) or processing time. For these reasons, our present strategy proposes a Bayesian inference approach. Using the analytical quadrupole method, a reference model can be calculated for a set of parameters. Then, the Bayesian probabilistic method is used to determine the maximum likelihood probability between the measured data and the reference model. To keep the processing method robust and fast, an automatic selection of the calculation range is proposed. Finally, in the case of a bilayer sample, both the thickness and resistive 3D layers are estimated in less than 2 min for a space and time matrix of 50,000 pixels by 4000 time steps with a reasonable relative error of less than 5%.

In this paper, the problem of the quantitative characterization of thermal resistance fields in a multilayer sample is addressed by using the classical front face flash method as the thermal excitation and infrared thermography (IRT) as the monitoring sensor. In this challenging problem, the complete inverse processing of a multilayer analytical model is difficult due to the lack of sensitivity of some parameters (layer thickness, depth of thermal resistance, etc.) and the expansive computational iterative processing. For these reasons, the proposed strategy is to use a simple multilayer problem where only one resistive layer is estimated. Moreover, to simplify the inverse processing often based on iterative methods, an asymptotic development method is proposed here. Regarding the thermal signal reconstruction (TSR) methods, the drawback of these methods is the inability to be quantitative. To overcome this problem, the method incorporates a calibration process originating from the complete analytical quadrupole solution to the thermal problem. This analytical knowledge allows self-calibration of the asymptotic method. From this calibration, the quantitative thermal resistance field of a sample can be retrieved with a reasonable accuracy lower than 5%.