Showing papers in "Computer Methods in Applied Mechanics and Engineering in 2021"

TL;DR: Experimental results show that the AOA provides very promising results in solving challenging optimization problems compared with eleven other well-known optimization algorithms.

TL;DR: It is found that honoring the physics leads to improved robustness: when trained only on a few parameters, the PINN model can accurately predict the solution for a wide range of parameters new to the network—thus pointing to an important application of this framework to sensitivity analysis and surrogate modeling.

TL;DR: A general framework for hp-variational physics-informed neural networks (hp-VPINNs) based on the nonlinear approximation of shallow and deep neural networks and hp-refinement via domain decomposition and projection onto space of high-order polynomials is formulated.

TL;DR: SciANN is designed to abstract neural network construction for scientific computations and solution and discovery of partial differential equations (PDE) using the physics-informed neural networks (PINN) architecture, therefore providing the flexibility to set up complex functional forms.

TL;DR: A high-accuracy multi-fidelity surrogate model correlating the configuration parameters of an aircraft and its aerodynamic performance by blending different fidelity information and adaptively learning their linear or nonlinear correlation without any prior assumption is constructed.

TL;DR: A Physics-Informed Neural Network (PINN) is presented to simulate the thermochemical evolution of a composite material on a tool undergoing cure in an autoclave by optimizing the parameters of a deep neural network using a physics-based loss function.

TL;DR: In this paper, the authors investigate how PINNs are biased towards learning functions along the dominant eigen-directions of their limiting NTK and construct novel architectures that employ spatio-temporal and multi-scale random Fourier features, and justify how such coordinate embedding layers can lead to robust and accurate PINN models.

TL;DR: These findings indicate that SCLs can dramatically boost the computational efficiency and scalability of computational homogenization for nonlinear and history-dependent materials with arbitrary microstructures, enabling the automatic and systematic generation of microstructurally-informed constitutive laws that can be adopted for the solution of macro-scale complex structures.

TL;DR: The results show that PINNs can infer the material properties from noisy synthetic data, and thus they have great potential for inferring these properties from experimental multi-modality and multi-fidelity data.

TL;DR: P-DEM does not need any classical discretization and requires only a definition of the potential energy, which simplifies the implementation and leads to much faster convergence compared to the original DEM.

TL;DR: A deep learning framework designed to train smoothed elastoplasticity models with interpretable components, such as the stored elastic energy function, yield surface, and plastic flow that evolve based on a set of deep neural network predictions is introduced.

TL;DR: The proposed approach is unsupervised, it requires no stress data but only displacement and global force data, and it delivers interpretable models, i.e., models that are embodied by parsimonious mathematical expressions discovered through sparse regression of a large catalogue of candidate functions.

TL;DR: A unified AI-framework named HiDeNN-AI is proposed to solve challenging computational science and engineering problems with little or no available physics as well as with extreme computational demand to show the flexibility of the framework for diverse problems from disparate fields.

TL;DR: Nonlocal PDDO-PINN is applied to the solution and identification of material parameters in solid mechanics and, specifically, to elastoplastic deformation in a domain subjected to indentation by a rigid punch, for which the mixed displacement--traction boundary condition leads to localized deformation and sharp gradients in the solution.

TL;DR: A neural network-based computational framework is established to characterize the finite deformation of elastic plates, which in classic theories is described by the Foppl--von Karman equations with a set of boundary conditions (BCs).

TL;DR: This work provides a unified framework, based on LDRBMs, for generating full heart muscle fibers and proposes, for the first time, a LDRBM to be used for generating atrial fibers.

TL;DR: In this paper, a double-phase-field formulation is proposed to describe cohesive tensile fracture and frictional shear fracture individually, and the formulation rigorously combines the two phase fields through three approaches: (i) crack-direction-based decomposition of the strain energy into the tensile, shear, and pure compression parts, (ii) contact-dependent calculation of the potential energy, and (iii) energy-based determination of the dominant fracturing mode in each contact condition.

TL;DR: The ML-Solver is demonstrated to solve the steady, incompressible Navier-Stokes equations in 3-D for several cases such as, lid-driven cavity, flow past a cylinder and conjugate heat transfer.

TL;DR: In this paper, a 3D recurrent residual U-Net (referred to as recurrent R-U-Net) is proposed to capture the spatial-temporal information associated with dynamic subsurface flow systems.

TL;DR: In this paper, a new phase field framework for modeling fracture and fatigue in shape memory alloys (SMAs) is presented, which is implemented in an implicit time integration scheme, with both monolithic and staggered solution strategies.

TL;DR: In this paper, a fully-decoupled and second-order time-accurate scheme of the flow-coupled phase-field type model was proposed, where the key idea achieving the full decoupling structure is to introduce an ordinary differential equation to deal with the nonlinear coupling terms that satisfy the zero-energy-contribution property.

TL;DR: A general machine learning-based topology optimization framework, which greatly accelerates the design process of large-scale problems, without sacrifice in accuracy, is put forward, which can efficiently handle design problems with a wide range of discretization levels, different load and boundary conditions, and various design considerations.

TL;DR: An algorithm to learn the relevant latent variables of a large-scale discretized physical system and predict its time evolution using thermodynamically-consistent deep neural networks using sparse autoencoders.

TL;DR: A neural network-based method for solving linear and nonlinear partial differential equations, by combining the ideas of extreme learning machines, domain decomposition and local neural networks, which exhibits a clear sense of convergence with respect to the degrees of freedom in the neural network.

TL;DR: In this paper, the phase-field regularized cohesive zone model (PF-CZM) was extended with a rational degradation function dependent on elasticity and fracture related material parameters, to solve the three-field (displacements, phase field and temperature) coupling equations altogether.

TL;DR: In this paper, a new data-driven reduced order model (ROM) framework is proposed, which is based on the hierarchical structure of the variational multiscale methodology and utilizes data to increase the ROM accuracy at a modest computational cost.

TL;DR: A multi-objective version of the recently proposed marine predator algorithm (MPA) is presented, which selects the effective solutions from the archive as the top predators to simulate the predator’s foraging behavior.

TL;DR: Several 3D benchmark problems involving mode I, I+II or I+III failure in brittle and quasi-brittle solids is addressed based on recent theoretical and numerical progresses on the unified phase-field theory for damage and fracture.

TL;DR: In this article, a two-scale topology optimization framework for the design of macroscopic bodies with an optimized elastic response is presented by means of a spatially-variant cellular architecture on the microscale.

TL;DR: The direct computation of the third-order normal form for a geometrically nonlinear structure discretised with the finite element (FE) method, is detailed, allowing to define a nonlinear mapping in order to derive accurate reduced-order models (ROM) relying on invariant manifold theory.