Please use this identifier to cite or link to this item:
http://hdl.handle.net/11375/29988
Title: | Finite Element Modeling of Plastic Deformation and Damage in Synthetic 3d Microstructure of Aluminum AA7075 Created Using Generative AI |
Authors: | Al toyuri, Amro |
Advisor: | Jain, Mukesh |
Department: | Mechanical Engineering |
Publication Date: | 2024 |
Abstract: | 3D microstructures provide valuable insight into material behavior which is essential in elucidating microstructural phenomena, such as particle morphology and void damage, and the consequent macroscopic material response at large plastic strains. However, experimentally obtaining 3D microstructures is extremely laborious and expensive, requiring complex microstructural imaging setups and layer-by-layer microstructural characterization using techniques such as Focussed-Ion Beam based Scanning Electron Microscopy (FIB-SEM). Nonetheless, a new method based on generative Artificial Intelligence (AI) has now become available to expedite this process, provided sufficient statistical information about the real microstructure is available. To this end, SliceGAN, a generative AI tool, was used to rapidly create statistically equivalent 3D microstructures from three orthogonal 2D microstructural SEM images, allowing for a practical and cost-effective method to generate synthetic 3D microstructures. Initially, a proof-of-concept study was performed to assess the capability of SliceGAN in generating synthetic 3D microstructures and the feasibility of using an integrated Artificial Intelligence and Finite Element (AI-FE) approach to study 3D microstructural response to plastic loading. To this end, Dream.3D was used to create multiple idealized microstructures of varying complexity and random orthogonal images were then used to train SliceGAN models. Both sets of models were assessed and validated by comparing particle fraction, morphology, and the microstructure response by performing FEA, where SliceGAN microstructure showed comparable performance to the Dream.3D-created microstructure. Based on the success of the initial work, SliceGAN was used to generate multiple synthetic 3D microstructures of AA7075-O aluminum sheet material from the three 2D microstructural SEM images. AA7075 sheet material has a complex microstructure with various strengthening precipitates within a softer aluminum matrix produced by hot and cold rolling processes. The synthetic 3D microstructures were post-processed, meshed, and modeled for large strain plasticity and particle-induced void damage behavior by finite element analysis using advanced constitutive material models. Subsequently, the synthetic and real 3D microstructures were qualitatively and quantitatively analyzed and compared for their elastoplastic deformation and ductile void damage. This work illustrates the viability of an integrated AI-FE methodology to study microstructural phenomena, demonstrating that synthetic microstructures can indeed reveal a comparable performance to real microstructures in terms of macroscopic stress-strain response, local stress and strain distribution, and void damage, albeit with some discrepancies. Also, it emphasizes the influence of particle morphology on strength and damage, where highly irregular particles play a dual role in increasing strain hardening by restricting matrix flow at the cost of increased ductile damage induced by decohered particles. Lastly, the more advanced FE models, with multiple voiding mechanisms further reduced the discrepancy between real and synthetic microstructures compared to the simpler models. |
URI: | http://hdl.handle.net/11375/29988 |
Appears in Collections: | Open Access Dissertations and Theses |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Al-Toyuri_Amro_HH_2024July_MASc.pdf | 25.7 MB | Adobe PDF | View/Open |
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.