Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List

Full text release has been delayed at the author's request until May 08, 2025

ETD Abstract Container

Abstract Header

Advanced Computational Frameworks for Predicting the Mechanical Response of Materials with Complex Microstructures

Ji, Mingshi
http://orcid.org/0000-0001-9918-4126
http://rave.ohiolink.edu/etdc/view?acc_num=osu1670955270979843

Abstract Details

2023, Doctor of Philosophy, Ohio State University, Mechanical Engineering.
Finite element method (FEM) is a numerical method that is widely used for obtaining approximate solutions to various engineering and research problems. In material science, FEM is used to simulate the mechanical response of materials under di erent loadings and to guide the design of materials with targeted properties. Generating high- delity nite element (FE) models of materials with complex microstructures, such as composite materials, requires generating realistic virtual microstructures of materials that are statistically equivalent to the actual microstructure and converting them to high- delity FE meshes. Choosing appropriate mechanical models and material properties for materials is also a fundamental challenge to obtaining an accurate simulation result. In this thesis, advanced computational frameworks are introduced to overcome these challenges, including creating realistic virtual microstructures of materials by algorithm, developing new damage models, and applying arti cial intelligence (AI) techniques to the model of materials with complex microstructures. Several numerical techniques are presented to model materials and analyze their micromechanical behaviors. Two new high-cycle fatigue damage models are introduced for modeling mechanical performance under cyclic loading to predict the fatigue life of heterogeneous adhesives, one for the matrix and the other for the particle-matrix interfaces. An automated computational framework is developed to simulate the performance of adhesives under cyclic loading. High- delity nite element models of this adhesive's representative volume elements (RVEs) are generated using an automated computational framework, enabling the virtual reconstruction of the microstructure and mesh generation. These 3D FE models are used to calibrate the fatigue damage model parameters with fatigue test data under di erent loading conditions. Another example presents a high- delity FE model, constructed based on in-plane micrograph images of beam cross sections, to simulate the failure response of L-Shape cross-ply composite beams under bending forces. Explicit FE failure simulations are performed to study the e ects of wrinkles and di erentiate among di erent failure modes. Numerical techniques such as the nite element method can accurately predict the mechanical response of materials with complex microstructures on the microscale, however the labor and computational cost associated with these methods render their application unfeasible for modeling a pipeline with over hundreds of miles. We introduced a computational framework combining the nite element simulation results with a deep learning model relying on the squeeze-and-excitation residual network (SE-ResNet) to predict the failure response of a statistical volume element (SVEs) of a corroded pipe. An automated microstructure reconstruction and mesh generation framework is utilized to synthesize the training data for this model by simulating the failure response of 10000 SVEs subject to a tensile load (hoop stress). A Bayesian optimization (BO) approach is utilized to determine the optimal combination of hyperparameters for the SE-ResNet model, followed by a k-fold cross-validation of the model. We also introduce another computational framework to predict the bone remodeling process using deep learning models. To synthesize the training data, a large virtual trabecular bone is generated by another deep convolutional generative adversarial network (DCGAN), followed by implementing the bond remodeling simulation of this trabecular bone and splitting it into multiple statistical volume elements (SVEs). A deep learning approach relying on a conditional generative adversarial network using Pix2pix architecture is presented to predict the morphological change of the trabecular bone.
Soheil Soghrati (Advisor)
William Marras (Committee Member)
Alok Sutradhar (Committee Member)
150 p.
Mechanical Engineering
Fatigue simulation, Image processing, Failure analysis, Deep learning model, Bone remodeling

Recommended Citations

Citations

  • Ji, M. (2023). Advanced Computational Frameworks for Predicting the Mechanical Response of Materials with Complex Microstructures [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1670955270979843

    APA Style (7th edition)

  • Ji, Mingshi. Advanced Computational Frameworks for Predicting the Mechanical Response of Materials with Complex Microstructures. 2023. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1670955270979843.

    MLA Style (8th edition)

  • Ji, Mingshi. "Advanced Computational Frameworks for Predicting the Mechanical Response of Materials with Complex Microstructures." Doctoral dissertation, Ohio State University, 2023. http://rave.ohiolink.edu/etdc/view?acc_num=osu1670955270979843

    Chicago Manual of Style (17th edition)

osu1670955270979843
© 2023, some rights reserved.Creative Commons License Advanced Computational Frameworks for Predicting the Mechanical Response of Materials with Complex Microstructures by Mingshi Ji is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by The Ohio State University and OhioLINK.
Release 3.2.12