Computational Mechanics最新文献

A curved non-isoparametric Reissner–Mindlin shell element is developed for analyzing thin-walled structures. The standard kinematic description of the element requires the calculation of the director vector. To address this demand accurately, similar to isogeometric analysis (IGA), the geometry is defined by utilization of the non-uniform rational B-splines (NURBS) imported directly from computer-aided design (CAD) files. Then, shape functions of the Legendre spectral element method (SEM) are used to interpolate the displacements. Consequently, the shell director vector and Jacobian of the transformation are calculated properly according to the presented formulation. On the other hand, in Legendre SEM combined with Gauss–Lobatto–Legendre quadrature, the integration points and the element nodes coincide. Thus, the easily computable local coordinate systems at the integration points can be used directly as nodal basis systems. A separate calculation of nodal basis systems at control points, which is the source of either complexity or error in IGA shells, is not required. Given the condition number of the stiffness matrix in the developed method, super high-order elements can also be used. Very high order p-refined elements are used in addition to h-refinement of the mesh to show the capability of higher order elements to analyze problems without mesh refinement. The validity and convergence rate of the method are investigated and verified through various cases of h- and p-refinement in challenging obstacle course problems.

Abstract

This paper introduces a metamodelling technique that employs gradient-enhanced Gaussian process regression (GPR) to emulate diverse internal energy densities based on the deformation gradient tensor (varvec{F}) and electric displacement field (varvec{D}_0) . The approach integrates principal invariants as inputs for the surrogate internal energy density, enforcing physical constraints like material frame indifference and symmetry. This technique enables accurate interpolation of energy and its derivatives, including the first Piola-Kirchhoff stress tensor and material electric field. The method ensures stress and electric field-free conditions at the origin, which is challenging with regression-based methods like neural networks. The paper highlights that using invariants of the dual potential of internal energy density, i.e., the free energy density dependent on the material electric field (varvec{E}_0) , is inappropriate. The saddle point nature of the latter contrasts with the convexity of the internal energy density, creating challenges for GPR or Gradient Enhanced GPR models using invariants of (varvec{F}) and (varvec{E}_0) (free energy-based GPR), compared to those involving (varvec{F}) and (varvec{D}_0) (internal energy-based GPR). Numerical examples within a 3D Finite Element framework assess surrogate model accuracy across challenging scenarios, comparing displacement and stress fields with ground-truth analytical models. Cases include extreme twisting and electrically induced wrinkles, demonstrating practical applicability and robustness of the proposed approach.

The aim of this work is the derivation and examination of a material model, accounting for large elastic deformations, coupled with species diffusion and thermal effects. This chemo-thermo-mechanical material model shows three key aspects regarding its numerical formulation. Firstly, a multiplicative split of the deformation gradient into a mechanical, a swelling and a thermal part. Secondly, temperature-scaled gradients for a numerical design comprising symmetric tangents and, thirdly, dissipation potentials for the modelling of dissipative effects. Additionally, the derived general material model is specialised to thermoresponsive hydrogels to study its predictive capabilities for a relevant example material class. An appropriate finite element formulation is established and its implementation discussed. Numerical examples are investigated, including phase transition and stability phenomena, to verify the ability of the derived chemo-thermo-mechanical material model to predict relevant physical effects properly. We compare our results to established models in the literature and discuss emerging deviations.

The Smoothed Particle Finite Element Method (SPFEM) has gained popularity as an effective numerical method for modelling geotechnical problems involving large deformations. To promote the research and application of SPFEM in geotechnical engineering, we present ESPFEM2D, an open-source two-dimensional SPFEM solver developed using MATLAB. ESPFEM2D discretizes the problem domain into computable particle clouds and generates the finite element mesh using Delaunay triangulation and the ( alpha )-shape technique to resolve mesh distortion issues. Additionally, it incorporates a nodal integration technique based on strain smoothing, effectively eliminating defects associated with the state variable mapping after remeshing. Furthermore, the solver adopts a simple yet robust approach to prevent the rank-deficiency problem due to under-integration by using only nodes as integration points. The Drucker-Prager model is adopted to describe the soil’s constitutive behavior as a demonstration. Implemented in MATLAB, this open-source solver ensures easy accessibility and readability for researchers interested in utilizing SPFEM. ESPFEM2D can be easily extended and effectively coupled with other existing codes, enabling its application to simulate a wide range of large geomechanical deformation problems. Through rigorous validation using four numerical examples, namely the oscillation of an elastic cantilever beam, non-cohesive soil collapse, cohesive soil collapse, and slope stability analysis, the accuracy, effectiveness and stability of this open-source solver have been thoroughly confirmed.

The Carrier-Domain Method was introduced for high-resolution computation of time-periodic long-wake flows. The cost-effectiveness of the method makes such computations practical in 3D. A short segment of the wake domain, the carrier domain, moves in the free-stream direction, from the beginning of the long wake domain to the end. The data at the moving inflow plane comes from the time-periodic data computed at an earlier position of the carrier domain. With the high mesh resolution that can easily be afforded over the short domain segment, the wake flow patterns can be carried, with superior accuracy, far downstream. Computing the long-wake flow with a high-resolution moving mesh that covers a short segment of the wake domain at any instant during the computation would certainly be far more cost-effective than computing it with a high-resolution fixed mesh that covers the entire length. We present high-resolution 3D computation of time-periodic long-wake flow for a cylinder and a wind turbine, both computed with isogeometric discretization and the Space–Time Variational Multiscale method. In the isogeometric discretization, the basis functions are quadratic NURBS in space and linear in time. The cylinder flow is at Reynolds number 100. At this Reynolds number, the flow has an easily discernible vortex shedding period. The wake flow is computed up to 350 diameters downstream of the cylinder, far enough to see the secondary vortex street. In the wind turbine long-wake flow computation, the velocity data at the inflow boundary of the wake domain comes from an earlier wind turbine computation, with the turbine rotor having a diameter of ({126},hbox {m}), extracted by projection from a plane located ({10},hbox {m}) downstream of the turbine. The wake flow is computed up to ({482},hbox {m}) downstream of the wind turbine. In both the cylinder and wind turbine wake flow computations, the flow patterns obtained with the full domain and carrier domain show a near-perfect match, clearly demonstrating the effectiveness and practicality of the Carrier-Domain Method in high-resolution 3D computation of time-periodic long-wake flows.

In the field of materials engineering, the accurate prediction of material behavior under various loading conditions is crucial. Machine Learning (ML) methods have emerged as promising tools for generating constitutive models straight from data, capable of describing complex material behavior in a more flexible way than classical constitutive models. Yield functions, which serve as foundation of constitutive models for plasticity, can be properly described in a data-oriented manner using ML methods. However, the quality of these descriptions heavily relies on the availability of sufficient high-quality and representative training data that needs to be generated by fundamental numerical simulations, experiments, or a combination of both. The present paper addresses the issue of data selection, by introducing an active learning approach for Support Vector Classification (SVC) and its application in training an ML yield function with suitable data. In this regard, the Query-By-Committee (QBC) algorithm was employed, guiding the selection of new training data points in regions of the feature space where a committee of models shows significant disagreement. This approach resulted in a marked reduction in the variance of model predictions throughout the active learning process. It was also shown that the rate of decrease in the variance went along with an increase in the quality of the trained model, quantified by the Matthews Correlation Coefficient (MCC). This demonstrated the effectiveness of the approach and offered us the possibility to define a dynamic stopping criterion based on the variance in the committee results.

Abstract

Standard non-local gradient damage methodology for fatigue analysis has an intrinsic drawback of unusual widening of the damage zone. This causes a rapid growth of crack in the simulations which often violate experimental evidences. In order to tackle this undesirable behaviour, the localizing gradient damage methodology has been formulated for high cycle fatigue crack growth simulations. The framework comprises of coupling damage and elasticity through continuum mechanics, a fatigue damage law and an interaction function which reduces the influence of damaged regions on the surrounding locality. The present scheme prevents the spurious widening of the damage-band around the critically damaged area and therefore the non-physical growth of fatigue crack in the simulations is successfully countered. The developed framework is tested on various standard specimens under mode-I and mixed-mode high cycle fatigue loads. Nonlinear finite element analysis is used for this purpose. The discretized form of solver equations for the localizing framework is mathematically derived. Numerical examples show that the simulated crack-growth curves using proposed localizing framework agree closely with the experimental data and has a higher accuracy than the standard non-local framework.

ELDIRK methods are defined to have an Explicit Last stage in the general Butcher array of Diagonal Implicit Runge-Kutta methods, with the consequence, that no additional system of equations must be solved, compared to the embedded RK method. Two general formulations for second- and third-order ELDIRK methods have been obtained recently in Mahnken [21] with specific schemes, e.g. for the embedded implicit Euler method, the embedded trapezoidal-rule and the embedded Ellsiepen method. In the first part of this paper, we investigate some general stability characteristics of ELDIRK methods, and it will be shown that the above specific RK schemes are not A-stable. Therefore, in the second part, the above-mentioned general formulations are used for further stability investigations, with the aim to construct new second- and third-order ELDIRK methods which simultaneously are A-stable. Two numerical examples are concerned with the curing for a thermosetting material and phase-field RVE modeling for crystallinity and orientation. The numerical results confirm the theoretical results on convergence order and stability.

Abstract

A recent mixed formulation of the Virtual Element Method in 2D elastostatics, based on the Hu-Washizu variational principle, is here extended to 2D elastodynamics. The independent modeling of the strain field, allowed by the mixed formulation, is exploited to derive first order quadrilateral Virtual Elements (VEs) not requiring a stabilization (namely, self-stabilized VEs), in contrast to the standard VEs, where an artificial stabilization is always required for first order quads. Lumped mass matrices are derived using a novel approach, based on an integration scheme that makes use of nodal values only, preserving the correct mass in the case of rigid-body modes. In the case of implicit time integration, it is shown how the combination of a self-stabilized stiffness matrix with a self-stabilized lumped mass matrix can produce excellent performances both in the compressible and quasi-incompressible regimes with almost negligible sensitivity to element distortion. Finally, in the case of explicit dynamics, the performances of the different types of derived VEs are analyzed in terms of their critical time-step size.

In this study, a novel approach is proposed by integrating the finite element tearing and interconnecting (FETI) method into the B-differentiable equations (BDEs) method for the analysis of 3D elastic frictional contact problem with small deformations. The contact blocks are divided into several nonoverlapping substructures with nonconforming meshes on the contact surface and the interface between two adjacent substructures. The enforcement of contact conditions and interface continuity conditions is achieved by using dual Lagrange multipliers discretized on the slave surface, typically defined with fine meshes. The modified Boolean transformation matrix is utilized to convert the contact stress into the equivalent nodal force. For large-scale elastic contact problems, the equilibrium equations for substructures and the relationship between the relative displacements and contact stresses on the contact surfaces and interfaces (i.e., the contact flexibility matrix) are efficiently computed using the FETI method. Subsequently, the governing equations consisting of the contact equations, interface continuity equations, and equilibrium equations for each floating substructure are uniformly formulated as the BDEs. These BDEs can be solved using the B-differentiable damped Newton method (BDNM). The proposed method harnesses the parallel scalability of the FETI method and extends the applicability of the BDEs algorithm, benefiting from its ability to precisely satisfy the contact constraints and theoretically ensure convergence when solving large-scale contact problems. The Hilber/Hughes/Taylor (HHT) time integration scheme is employed to investigate elastic dynamic contact problems. Numerical examples demonstrate the accuracy, convergence rate, and parallel scalability of the proposed algorithm.