Date of Award


Publication Type


Degree Name



Mechanical, Automotive, and Materials Engineering


Direct Enrgy Deposition;Finite Element Analysis;Junction structures;Machine Learning;Residual stress;Thin-wall


Jill Urbanic



Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


This dissertation focuses on the investigation of residual stress in metal additive manufacturing of thin-walled components. Residual stress is a critical issue in additive manufacturing processes as it can lead to structural deformations, reduced mechanical properties, and potential failure of the manufactured parts. The objective of this research is to develop a fast and accurate prediction model for residual stress using numerical and machine learning techniques, with a specific focus on path planning. To achieve the research objectives, the dissertation employs numerical simulations to understand the underlying physics and mechanics of residual stress formation in the additive manufacturing process. These simulations serve as a basis for generating training data and identifying the key factors influencing residual stress. The study explores the application of three machine learning algorithms: Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Random Forest Regressor. These algorithms are trained and evaluated using numerical data obtained from simulation of laser metal deposition (LMD), considering various thin-walled configuration components. The residual stress predictions obtained from the numerical models are validated through experimental measurements using the X-ray diffraction method. To begin, numerical simulations are performed to gain insights into the underlying physics and mechanics of residual stress formation in metal additive manufacturing. These simulations serve as a basis for generating training data and identifying the effects of path planning including deposition sequence and direction and geometrical features that influence residual stress. Furthermore, the dissertation investigates the role of path planning in mitigating residual stress in junction structures and thin-walled components. Various path planning strategies are explored using numerical models to identify optimal printing paths that minimize residual stress. The results of this research provide valuable insights into the prediction of residual stress in thin-wall metal additive manufacturing. The machine learning models, including ANN, ANFIS, and Random Forest Regressor, demonstrate their effectiveness in fast and accurate prediction of residual stress. Notably, the Random Forest algorithm excels in accuracy owing to its robust ensemble learning capacity. The validation of the numerical models using experimental measurements enhances their reliability and ensures their practical applicability. Overall, this dissertation contributes to the advancement of metal additive manufacturing by providing engineers with valuable tools and strategies for predicting and mitigating residual stress in thin-walled components. The combination of numerical and experimental approaches, along with the utilization of machine learning algorithms, enables fast and accurate prediction of residual stress, leading to improved part quality and reduced risk of failure in metal additive manufacturing applications.