A Comparison of Machine Learning Algorithms for The Prediction of Thermal-Expansion of Spherical Shells

Search for this publication on Google Scholar
Bumgarner, K. (2022). A Comparison of Machine Learning Algorithms for The Prediction of Thermal-Expansion of Spherical Shells. Unc Charlotte Electronic Theses And Dissertations.
Download PDF

Analytics

32 views ◎
8 downloads ⇓


Abstract

Machine Learning (ML) applications in material science is an active area of research due to its accurate predictive capabilities while also being generalizable. Experimentation where a physical model of the problem is constructed has been widely used in the past and is still used today but is very costly in time and capital. Simulation using virtual models can be much more efficient than experimental modeling but still takes time to construct and solve computationally. ML uses statistics and data science methods to predict a variety of different problem configurations quickly while remaining accurate. While becoming an increasingly popular method, different ML model types have their own advantages and disadvantages. In this work, ML models are trained to predict the thermal expansion of a hollow sphere so that these advantages and disadvantages may be studied. Six ML model types were explored by training on instances of data generated from Finite Element Analysis (FEA) data then compared using error metrics. Sample size analysis was completed to determine required amounts of input data to produce an accurate model for each ML type. A noise study was conducted to observe how each model type would react to varying levels of noise added to the data. Sensitivity analysis was carried out on the optimal model to test how each predictor variable affected the prediction value. Finally, the optimal model was tested against new data that had not been encountered before to simulate how a production ready model would work in the real world.

Details
Author

Bumgarner, Kolby

Title

A Comparison of Machine Learning Algorithms for The Prediction of Thermal-Expansion of Spherical Shells

Physical Description

1 online resource (61 pages) : PDF

Date

2022

Degree Granting Institution

University of North Carolina at Charlotte

Abstract

Machine Learning (ML) applications in material science is an active area of research due to its accurate predictive capabilities while also being generalizable. Experimentation where a physical model of the problem is constructed has been widely used in the past and is still used today but is very costly in time and capital. Simulation using virtual models can be much more efficient than experimental modeling but still takes time to construct and solve computationally. ML uses statistics and data science methods to predict a variety of different problem configurations quickly while remaining accurate. While becoming an increasingly popular method, different ML model types have their own advantages and disadvantages. In this work, ML models are trained to predict the thermal expansion of a hollow sphere so that these advantages and disadvantages may be studied. Six ML model types were explored by training on instances of data generated from Finite Element Analysis (FEA) data then compared using error metrics. Sample size analysis was completed to determine required amounts of input data to produce an accurate model for each ML type. A noise study was conducted to observe how each model type would react to varying levels of noise added to the data. Sensitivity analysis was carried out on the optimal model to test how each predictor variable affected the prediction value. Finally, the optimal model was tested against new data that had not been encountered before to simulate how a production ready model would work in the real world.

Genre

masters theses

Subjects--Topics

Mechanical engineering

Degree

M.S.

Subject Area

Mechanical Engineering

Advisor(s)

Cherukuri, Harish

Committee Members

Amengonu, Yawo
Keanini, Russell

Degree Note

Thesis (M.S.)--University of North Carolina at Charlotte, 2022.

Rights Statement

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.

Rights Holder Information

Copyright is held by the author unless otherwise indicated.

Identifier

Bumgarner_uncc_0694N_13314

Permalink

http://hdl.handle.net/20.500.13093/etd:3070