Date of Award
9-27-2023
Publication Type
Dissertation
Degree Name
Ph.D.
Department
Mechanical, Automotive, and Materials Engineering
Supervisor
Jill Urbanic
Supervisor
Ofelia Jianu
Rights
info:eu-repo/semantics/embargoedAccess
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Abstract
This dissertation explores additive manufacturing (AM) processes, specifically directed energy deposition (DED), and investigates the prediction of mechanical and physical properties using different approaches. The analysis of thermal effects and scanning patterns, data clustering for post-fabrication properties, and the assessment of residual stresses in multi-layer geometries have been performed. A part of the study has investigated the geometrical and mechanical properties of single bead clads using multi-physics finite element models. Feed-forward back-propagation artificial neural network (ANN) were employed in 3D domains, with ANN demonstrating a good potential for single 3D bead data predictions. However, single bead analyses provide limited information. Therefore, thermo-mechanical characteristics of scanning patterns for different geometries have been investigated. One-way, zigzag, spiral, and raster-angled tool patterns were analyzed to understand residual stresses and distortions in one-layer multi-shape geometries employing a combination of unsteady finite volume and finite element methods. The impact of geometry is evident as the best tool path for one component is the worse for another. The study also investigates the application of clustering methods in analyzing additive manufactured components' data for improved understanding and quality enhancement. Multiple geometries with varying heat histories are categorized based on post-fabrication properties using self-organizing map, k-means clustering, and fuzzy c-means clustering. The local-based and global-based clusters are discussed. The research has also examined the effects of heat accumulation and melting on residual stresses in multi-track, and multi-layer geometries. Numerical, experimental, and machine learning (Long Short-Term Memory) methods are explored to provide insightful discussion for wire-laser direct energy deposition. Overall, this dissertation contributes to the understanding and improvement of additive manufacturing processes by addressing key aspects such as prediction of properties, analysis of scanning patterns, data clustering, and evaluation of residual stresses. Future research directions are suggested to further enhance the accuracy and efficiency of these processes.
Recommended Citation
Mirazimzadeh, Seyedeh Elnaz, "Thermo-mechanical analysis of complex geometries manufactured by directed energy deposition processes using machine learning approaches" (2023). Electronic Theses and Dissertations. 9287.
https://scholar.uwindsor.ca/etd/9287