Model-based methodology coupled with a Laser-based experiment for non-destructive thermal parameter estimation of an Epoxy nanocomposite

ALTUNA F.I.; ANTONACCI, J.; PONTIS, A.; HOPPE, C.E.; CHIURO, C.; FRONTINI, G.; OTERO, F.; ARENAS, G. F.; PUIG, J.; ELIÇABE, G.

Resistencia

Congreso; IEEE ARGENCON 2020. V Congreso Bienal de la Sección Argentina del IEEE; 2020

Facultad Regional Resistencia, Universidad Tecnológica Nacional y Sección Argentina del IEEE

This article addresses the thermal characterization, i.e., the estimation of the thermal  conductivity k and the specific heat capacity Cp of a synthetic nanocomposite, by means of coupling a laser-based experiment and a reduced model-based inverse problem solved via a hybrid Finite Element Method (FEM) computational approach. The associated multiphysics problem can be described as an optical-radiative-thermal coupled process where a laser source is irradiating the sample corresponding to an Epoxy-Based Vitrimer (EV) withdispersed gold nanoparticles (NPs). The so-called photothermal effect occurs at the  electronic level of the embedded NPs in the vitrimer and it is responsible for the remote heating of the material, which turns the original radiation into an equivalent thermal source Q. Since the complexity of the problem, a simplification of the complete process is proposed. A 1D reduced model for an infinite slab (having both boundaries subjected to natural convection and under certain conditions on both experiment and material properties) has been used to solve the forward radiative transfer problem involved, in order to achieve an initial approximation for Q. This approximation is then used for the source term in the heat transfer equation, whose solution is computed using the finite element method (FEM). Time-varying and space-varying temperature measurements are simulated for a 3D-real geometry with additive Gaussian noise considering partial and total illumination from the beam. The corresponding statistical solution for the thermal conductivity k and the heat capacity Cp is calculated by the inversion of the T-thermocouple temperature time series measured at the center of the irradiated specimen, using a stochastic version of the Levenberg-Marquardt algorithm. Reduced models have been analyzed for 1D and 2D reduced geometries as well as approximations for Q in the 3D case. Results show a confidence interval for the achieved estimates of k and Cp in a good agreement to the Differential Scanning Calorimetry (DSC) referential values. As a conclusion, even when the performed laser remote heating experiment has been developed as a self-healing process, italso appears to be a very promising NDT methodology for retrieving the thermal parameters along with the proposed computational approach in some thermosetting polymers, such as the EV studied here. As a final remark, a modified scheme using Deep Learning (DL) is proposed in order to improve both equivalent model and results in a future work.