Authors:

Luke S

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Thermal grease is an important component of electronics and computer hardware and is the medium between a heat source and a heat sink. Thermal grease breaks down after being subjected to stress in a particular limit. We sought to predict and graph the stress breakdown or the relaxation regime of the thermal grease (DOWSIL TC-5622) through physics-informed neural networks that use a Thixotropic-elasto-visco-plastic (TEVP) model for the stress. The material behavior of thermal greases follows the TEVP model, which is a certain set of ordinary differential equations with certain unknown parameters. First, we created a model to solve stress as a forward problem with the unknown parameters given, so we can observe the relaxation regime. Then we take the experimental data and turn the model into an inverse problem. In this way we can use it to find the unknown parameters of the model. Specifically, we used the DeepXDE library implemented in Python and experimental data of the stress-strain to solve the problem. The goal of this study is to find the optimum of the six unknown parameters of the TEVP model. The unknown parameters are elastic shear modulus, background viscosity, plastic viscosity, yield stress, structure build-up coefficient, and structure breakage coefficient. Once the parameters are optimized, we can then predict how the test data shear stress should change over time. By comparing this predicted data to the actual test data, we will learn how well our physics-informed neural network is predicting the thermal grease behavior.

Source:

Purdue University / 2023

Topics:

No topics listed

Co-authors:

Luke S