Session: 03-03-01: Integrated Computational Materials Engineering (ICME)

Paper Number: 99091

99091 - Ai-Enhanced Advanced Algorithms for the Micromechanical Modeling and Design of Materials With Complex Microstructures 

We present a set of integrated computational and deep learning algorithms for the automated modeling of materials with complex microstructures, including various composite materials systems and biomaterials. The proposed framework relies on a set virtual microstructure reconstruction techniques [1,2] and a parallel mesh generation algorithms coined CISAMR [3,4] for creating high-fidelity finite elements models of such materials. In order to synthesize the material microstructure, we introduce three classes of  algorithms: (a) virtual packing/optimization approach for the virtual reconstruction of heterogenous material  microstructures, capable of packing arbitrary-shaped particles and replicating any number of target statistical microstructural descriptors, such as the volume fraction, size distribution, spatial  arrangement, and orientation of embedded heterogeneities; (b) physics-based approach for synthesizing woven composite microstructures, starting with generating a loosely-woven initial  microstructure followed by a reduced-order FE simulation to replicate the final model; (c) we have  developed a modelling framework coined ReconGAN, which relies on Deep Convolutional Generative Adversarial Networks (DCGAN) and shape optimization to generate digital twins of the human vertebral body with a highly realistic characterization of the trabecular bone microarchitecture.

The parallel CISAMR algorithm is a unique non-iterative meshing algorithm capable of discretizing massive problems (tens or hundreds of elements) with complex geometries. In this presentation, we show several capabilities recently added to CISASMR, including automated modeling of crack  growth problems and complex woven composite microstructures. Further,  we show how CISAMR is expanded to handle domains with sharp edges/corners,  such as polycrystalline material microtructures.

We demonstrate the application of this integrated reconstruction-meshing framework for predicting the failure response and fatigue life of a variety of materials systems, including particulate and fiber-reinforced composites. We will also show how this  framework can be implemented as an automated modeling engine for generating the training data for predating the strength of steel pipes subjected to pitting corrosion using a squeeze-and-excitation residual network  (SE-ResNet).

 

[1]       M. Yang, B. Lian, A. Nagarajan, and S. Soghrati “New algorithms for virtual reconstruction of heterogeneous microstructures.” Computer Methods in Applied Mechanics and Engineering, 338, 275-298 (2018).

[2]       H. Ahmadian, P. Mageswaren, B. Walter, D.M. Blakaj, E. Bourekas, E. Mendel, W.S. Marras, S. Soghrati, “Towards an AI‐assisted framework for reconstructing the digital twin of vertebra and predicting its fracture response.” International Journal for Numerical Methods in Biomedical Engineering, In press (2022).

[3]       A. Nagarajan and S. Soghrati, “Conforming to interface structured adaptive mesh refinement: 3D algorithm and implementation.” Computational Mechanics, 62(5), 1213-1238 (2018).

[4]       B. Liang, A. Nagarajan, and S. Soghrati, “Scalable parallel implementation of CISAMR: A non-iterative mesh generation algorithm.” Computational Mechanics, 64, 173-195 (2019).

Presenting Author: Soheil Soghrati Ohio State University

Presenting Author Biography: Dr. Soheil Soghrati is an Associate professor of Mechanical and Aerospace Engineering & Materials Science and Engineering at The Ohio State University. He earned his PhD in Structural Engineering with Minor in Computational Science in Engineering from the University of Illinois at Urbana-Champaign, during which he held a graduate research assistantship at the Beckman Institute for Advanced Science and Technology. He received both his Masters and Bachelor degrees in Civil Engineering from Isfahan University of Technology and a certification of Advanced Structural Engineering from Bauhaus University in Germany. Dr. Soghrati joined the Department of Mechanical and Aerospace Engineering at OSU in June 2013 with a joint appointment in the Department of Materials Science and Engineering. He is also one of the steering board faculty members in the Simulation Innovation and Modeling Center (SIMCenter) at OSU. Dr. Soghrati’s research interests lay in the area of computational solid mechanics with especial focus on advanced finite element and meshfree methods for the automated modeling of problems with complex and/or evolving morphologies. He has established the Automated Computational Mechanics Laboratory (ACML) at OSU. Some of the problems investigated in Dr. Soghrati's research group include cancer engineering, microstructure reconstruction and mesh generation algorithms, deep learning algorithms and their applications in computational mechanics, simulating the corrosion assisted damage processes, digital manufacturing, Investigating the multiscale failure response of composite materials, computational biomechanics, and computational design of Lithium-ion battery electrodes. Current projects in ACML are supported by the National Science Foundation, Air Force Office of Scientific Research, Department of Defense, Honda R&D Americas, Ford, and Center for Cancer Engineering.

Authors:

Soheil Soghrati Ohio State University
Mingshi Ji The Ohio State University
Pengfei Zhang The Ohio State University
Salil Pai The Ohio State University
Balavignesh Vemparala The Ohio State University