Recently, two students (21’ Electrical Engineering), LU Haofan and YU Yi, published their paper "Deep learning techniques elucidate and modify the shape factor to extend the effective medium theory beyond its original formulation" in the International Journal of Heat and Mass Transfer under the supervision of ZJUI Associate Professor Weeliat Ong. Their paper proposes a universal optimization method based on neural networks, which realizes the rapid estimation of the thermal conductivity of composite materials with new particle shapes and expands the application range of the original effective medium theories(EMTs).

There are two widely-used methods for approximating the thermal conductivity of composite materials - one is the Analytical EMT Model, the other is the Finite Element Model. They both have advantages and disadvantages- while it is faster to use the Analytical EMT Model, a re-derivation is required for a new particle shape； and the Finite Element Model requires much time and computing resources.

In this regard, the two students were inspired by the transfer learning of neural networks. They applied the known EMT formulation and limited stimulation data of finite elements, used deep learning techniques to correct the mathematically derived physical model, and extended the original EMT for new thermal transport problems.

▲  Group photo of LU Haofan(Left), Wee-liat Ong(Middle), YU Yi(Right)

"I am very happy to see that they started from the SRTP in their sophomore year and insisted on this topic. Without help from other master students, doctoral students, and postdoctoral fellows, they went through constantly self-denying, exploring, and improving, and then finally, the result is published in the International Journal of Heat and Mass Transfer.

This is very rare and valuable for undergraduate students. I am also convinced that this valuable experience of spending two years on a research project during undergraduate years will be beneficial to their future research and career. ” Although it takes more time and energy to guide undergraduates to do scientific research, Professor Ong is still dedicated to guiding the students and is proud for their growth and transformation.