Author ORCID Identifier

https://orcid.org/0000-0003-2392-9747

Date Available

8-20-2025

Year of Publication

2025

Document Type

Doctoral Dissertation

Degree Name

Doctor of Philosophy (PhD)

College

Engineering

Department/School/Program

Mechanical Engineering

Faculty

Dr. Julius Schoop

Faculty

Dr. Jonathan Wenk

Abstract

Process-induced residual stress (RS) plays a critical role in determining the structural integrity, fatigue life, corrosion resistance, and dimensional stability of manufactured components. Tensile residual stress can be harmful to the machined workpiece by lowering fatigue strength, while compressive residual stress helps to prevent such issues. Therefore, accurate prediction and control of machining-induced residual stresses are essential for optimizing manufacturing processes and enhancing component performance. However, existing modeling approaches—empirical, analytical, and numerical—often have limited predictive accuracy, demand extensive calibration, or involve high computational costs. Traditional finite element analysis (FEA) provides modeling insights but is constrained by its high computational cost and complexity. To address these challenges, this dissertation develops a hybrid physics-informed, data-driven framework that integrates machine learning, semianalytical contact mechanics, and uncertainty quantification (UQ) to predict and optimize RS in machining and burnishing processes with with greater accuracy and efficiency.

First, a semi-analytical model extends prior 2D machining RS predictions to 3D processes (turning, milling, drilling) by incorporating geometric, kinematic, and size-effect constraints. Unlike brute-force 3D-FEM or purely empirical methods, the model leverages in-situ digital image correlation (DIC) to calibrate Hertzian contact pressures, friction coefficients, and multi-pass shakedown effects. Validation against published RS data for Ti64 and Inconel 718 demonstrates strong agreement within experimental error margins, accurately capturing near-surface stress, peak stress magnitude/location, and profile depth. The model’s efficiency enables rapid parameter sweeps for industrial process design.

Next, beyond deterministic modeling, machining processes exhibit inherent uncertainty due to material microstructure variations, process variability, and tool wear. This dissertation incorporates stochastic simulations and uncertainty quantification (UQ) to evaluate the probabilistic nature of residual stress formation. By integrating a physics informed, data-driven framework, we assess the impact of material property fluctuations on stress distributions, revealing that microstructural variations significantly influence local residual stress profiles even when macroscale variability remains low.

Finally, this research introduces a Physics-Informed Neural Network (PINN) framework to address both forward and inverse problems in contact mechanics, embedding the Prandtl-Reuss constitutive equations into the loss function to ensure physically consistent solutions. The forward PINN model predicts machining induced residual stress distributions by solving governing partial differential equations, while the inverse PINN approach estimates key material and process parameters critical for residual stress modeling. This method accelerates computations compared to conventional numerical techniques while maintaining high accuracy in simulating complex plasticity phenomena.

The results of this research provide a computationally efficient, experimentally validated, and physics-informed modeling framework for machining-induced residual stress prediction. This approach facilitates the development of digital process twins (DPT), offering manufacturers a robust tool to optimize machining parameters and improve surface integrity. The integration of machine learning, physics-based modeling, and uncertainty quantification advances the field of computational mechanics, making significant contributions to residual stress modeling. Ultimately, this research enhances the predictive capabilities of residual stress modeling, paving the way for improved machining process optimization, fatigue life prediction, and component reliability in high-performance applications.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2025.420

Funding Information

1. U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE) under the Advanced Manufacturing Office’s (AMO) DE-FOA-0001980, Award Number DE-EE0009121/0000, project title “AI-Enabled Discovery and Physics-Based Optimization of Energy Efficient Processing Strategies for Advanced Turbine Alloys”.

2. National Science Foundation, grant number 2143806, project title “CAREER: Thermomechanical Response and Fatigue Performance of Surface Layers Engineered by Finish Machining: In-situ Characterization and Digital Process Twin”.

Download

Share

COinS