Keywords

PINN, Heat Transfer, Thermal Modeling, TIMs, Interfacial Thermal Resistance, PDEs

Abstract

Silicon/Carbon Nanotube (Si/CNT) composites have shown great performance as Thermal Interface Materials (TIMs) for cooling in advanced electronics. The heat transfer in these materials is highly influenced by the interfacial thermal resistance (ITR) at the interfaces between the CNT filler and the silicon matrix. The ITR poses a critical challenge that hinders the design and performance of the TIM. This research designs a Physics Informed Neural Network (PINN) that predicts the thermal behavior of Si/CNT through a forward and inverse solver since they are known networks in solving complex partial and ordinary differential equations. However, despite the advantages of PINNs, standard PINNs are known to produce models with continuous derivatives. There is an issue seen in standard PINNs when used in predicting heterogenous materials because it struggles with sharp gradients and non-smooth interfaces across the discontinuous materials. This research explores the domain decomposed PINN approach of solving a 2D steady state heat conduction problem of a composite material used in semiconductor packaging. A model of 1µm x 1µm Si/CNT square matrix is sub-divided into four vertical rectangular strips of varying thermal conductivity () to represent the composite material thermal properties. A fine mesh is generated using MATLAB’S PDE Toolbox to solve the Poisson equation under Dirichlet and Neumann boundary conditions. Abaqus FEA simulation results are also used to compare results from PINNs and MATLAB. A good correlation is achieved through fine-tuning model parameters. This research shows that the forward and inverse PINN framework is a powerful data driven tool for material modeling. It successfully implements the flexibility PINNs governed by laws of physics to predict the hidden parameters in next generation TIMs.

Completion Date

2025

Semester

Fall

Committee Chair

Bai, Yuanli

Degree

Master of Science in Mechanical Engineering (M.S.M.E.)

College

College of Engineering and Computer Science

Department

Mechanical and Aerospace Engineering

Format

PDF

Identifier

DP0029809

Document Type

Thesis

Campus Location

Orlando (Main) Campus

Download

Share

COinS