Mechanistic artificial intelligence (mechanistic-AI) for modeling, design, and control of advanced manufacturing processes: Current state and perspectives

Journal of Materials Processing Technology, Journal Year: 2021, Volume and Issue: 302, P. 117485 - 117485

Published: Dec. 30, 2021

Language: Английский

---

Exploration of optimal microstructure and mechanical properties in continuous microstructure space using a variational autoencoder

Materials & Design, Journal Year: 2021, Volume and Issue: 202, P. 109544 - 109544

Published: Feb. 5, 2021

Data-driven approaches enable a deep understanding of microstructure and mechanical properties materials greatly promote one's capability in designing new advanced materials. Deep learning-based image processing outperforms conventional techniques with unsupervised learning. This study employs variational autoencoder (VAE) to generate continuous space based on synthetic microstructural images. The structure-property relationships are explored using computational approach quantification, dimensionality reduction, finite element method (FEM) simulations. FEM representative volume (RVE) microstructure-based constitutive model is proposed for predicting the overall stress-strain behavior investigated dual-phase steels. Then, Gaussian process regression (GPR) used make connections between latent point ferrite grain size as inputs outputs. GPR VAE successfully predicts newly generated microstructures target high accuracy. work demonstrates that variety can be candidates optimal material manner.

Language: Английский

---

Inverse design of truss lattice materials with superior buckling resistance

npj Computational Materials, Journal Year: 2022, Volume and Issue: 8(1)

Published: Nov. 29, 2022

Abstract Manipulating the architecture of materials to achieve optimal combinations properties (inverse design) has always been dream scientists and engineers. Lattices represent an efficient way obtain lightweight yet strong materials, providing a high degree tailorability. Despite massive research done on lattice architectures, inverse design problem complex phenomena (such as structural instability) remained elusive. Via deep neural network genetic algorithm, we provide machine-learning-based approach inverse-design non-uniformly assembled lattices. Combining basic building blocks, our allows us independently control geometry topology periodic aperiodic structures. As example, architectures with superior buckling performance, outperforming traditional reinforced grid-like bio-inspired lattices by ~30–90% 10–30%, respectively. Our results insights into behavior beam-based lattices, opening avenue for possible applications in modern structures infrastructures.

Language: Английский

---

A deep learning energy method for hyperelasticity and viscoelasticity

European Journal of Mechanics - A/Solids, Journal Year: 2022, Volume and Issue: 95, P. 104639 - 104639

Published: May 4, 2022

Language: Английский

---

Dispersion relation prediction and structure inverse design of elastic metamaterials via deep learning

Materials Today Physics, Journal Year: 2022, Volume and Issue: 22, P. 100616 - 100616

Published: Jan. 1, 2022

Language: Английский

---