Simultaneous identification of thermophysical properties of semitransparent media using an artificial neural network trained by a 2-D axisymmetric direct model

Abstract

In this article, a multilayer artificial neural network (ANN) identification model is developed to simultaneously identify the thermal conductivity and the effective absorption coefficient of semitransparent materials from flash-type experimental measurements. Firstly, the ANN is trained by means of data generated by a 2-D axisymmetric heat transfer model whose radiative part is treated via the P1 approximation. A sensitivity study is then used to prove the theoretical feasibility of the identification strategy. Several training data distributions (uniform or Gaussian types) are tested on synthetic data, and on noisy ones for checking the robustness. Finally, the efficiency of this estimation approach is investigated using experimental data obtained by flash method on a PMMA sample. The estimated thermal conductivity and the effective absorption coefficient are compared with values obtained from the literature and other measurements.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was partially supported by the China Scholarship Council Foundation. The authors would also thank CPER [2015-2020] and ERDF [2014-2020] grants for supporting this work through the funding of experimental equipment.