Date Approved
5-14-2025
Graduate Degree Type
Thesis
Degree Name
Engineering (M.S.E.)
Degree Program
School of Engineering
First Advisor
Abishek Balsamy-Kamaraj, Ph.D.
Academic Year
2024/2025
Abstract
Electrodeposition plays a vital role in processes such as electrochemical additive manufacturing (ECAM) and electroplating, where precise control over operating conditions, such as current density, deposit thickness and electrolyte concentration is crucial to achieving desirable mechanical and structural properties. Given the vast number of independent variables available for the process of electrodeposition, traditional methods of experimentation and simulation are often time-consuming, costly, and prone to inaccuracies, limiting their usefulness in optimizing electrodeposition parameters. This study investigates the application of machine learning (ML) models, specifically data-driven artificial neural networks (ANNs), simulation-based physics-informed neural networks (PINNs), and hybrid PINNs (combining experimental data with simulation parameters and physics laws), to accelerate and enhance the accuracy of electrodeposition parameter predictions. Baseline data was generated from a potentiostatic electrodeposition experiment and simulation models incorporating Fick’s Second Law and Faraday’s Law. Data-driven ANNs utilized experimental data, simulation-based PINNs integrated simulation parameters with physics laws while hybrid PINNs combined the strengths of both approaches. The results show that ML models, particularly hybrid PINNs, can achieve up to 120 times increase in computational speed compared to traditional simulations, with superior accuracy. These models accurately captured key trends- including current density (a measure of ionic flux), deposit growth over time, and the dynamic behavior of electrolyte ion concentration- with their robustness validated through comparisons with experimental and simulation results. This study highlights the potential of ML models, especially hybrid PINNs, to provide rapid, reliable, and scalable solutions for optimizing process parameters. These advancements pave the way for more efficient and cost-effective innovations in electrochemical technologies.
ScholarWorks Citation
Ngaruiya, Allan Mwangi, "Accelerating Electrodeposition Simulation Using Physics Informed Neural Networks" (2025). Masters Theses. 1142.
https://scholarworks.gvsu.edu/theses/1142
