Sponsor
Portland State University. Department of Mechanical and Materials Engineering
First Advisor
Sung Yi
Term of Graduation
Winter 2023
Date of Publication
1-18-2023
Document Type
Thesis
Degree Name
Master of Science (M.S.) in Mechanical Engineering
Department
Mechanical and Materials Engineering
Language
English
Subjects
fatigue, gradient boosting, machine learning, neural network, random forest, solder
DOI
10.15760/etd.3425
Physical Description
1 online resource (ix, 105 pages)
Abstract
Predicting the reliability of board-level solder joints is a challenging process for the designer because the fatigue life of solder is influenced by a large variety of design parameters and many nonlinear, coupled phenomena. Machine learning has shown promise as a way of predicting the fatigue life of board-level solder joints. In the present work, the performance of various machine learning models to predict the fatigue life of board-level solder joints is discussed. Experimental data from many different solder joint thermal fatigue tests are used to train the different machine learning models. A web-based database for storing, sharing, and uploading data related to the performance of electronics materials, the Electronics Packaging Materials Database (EPMD), has been developed and used to store and serve the training data for the present work. Data regression is performed using artificial neural networks, random forests, gradient boosting, extreme gradient boosting (XGBoost), and adaptive boosting with neural networks (AdaBoost). While previous works have studied artificial neural networks as a way to predict the fatigue life of board-level solder joints, the results in this paper suggest that machine learning techniques based on regression trees may also be useful in predicting the fatigue life of board-level solder joints. This paper also demonstrates the need for a large collection of curated data related to board-level solder joint reliability, and presents the Electronics Packaging Materials Database to meet that need.
Rights
©2022 Jason Scott Ross
In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/ 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).
Persistent Identifier
https://archives.pdx.edu/ds/psu/39843
Recommended Citation
Ross, Jason Scott, "Comparing the Performance of Different Machine Learning Models in the Evaluation of Solder Joint Fatigue Life Under Thermal Cycling" (2023). Dissertations and Theses. Paper 6358.
https://doi.org/10.15760/etd.3425