Computational Mechanics最新文献

Many aerospace applications involve complex multiphysics in compressible flow regimes that are challenging to model and analyze. Fluid–structure interaction (FSI) simulations offer a promising approach to effectively examine these complex systems. In this work, a fully coupled FSI formulation for compressible flows is summarized. The formulation is developed based on an augmented Lagrangian approach and is capable of handling problems that involve nonmatching fluid–structure interface discretizations. The fluid is modeled with a stabilized finite element method for the Navier–Stokes equations of compressible flows and is coupled to the structure formulated using isogeometric Kirchhoff–Love shells. To solve the fully coupled system, a block-iterative approach is used. To demonstrate the framework’s effectiveness for modeling industrial-scale applications, the FSI methodology is applied to the NASA Common Research Model (CRM) aircraft to study buffeting phenomena by performing an aircraft pitching simulation based on a prescribed time-dependent angle of attack.

An enhanced virtual element formulation for large displacement analyses is presented. Relying on the corotational approach, the nonlinear geometric effects are introduced by assuming nodal large displacements but small strains in the element. The element deformable behavior is analyzed with reference to the local system, corotating with the element during its motion. Then, the large displacement-induced nonlinearity is accounted for through the transformation matrices relating the local and global quantities. At the local level, the Virtual Element Method is adopted, proposing an enhanced procedure for strain interpolation within the element. The reliability of the proposed approach is explored through several benchmark tests by comparing the results with those evaluated by standard virtual elements, finite element formulations, and analytical solutions. The results prove that: (i) the corotational formulation can be efficiently used within the virtual element framework to account for geometric nonlinearity in the presence of large displacements and small strains; (ii) the adoption of enhanced polynomial approximation for the strain field in the virtual element avoids, in many cases, the need for ad-hoc stabilization procedures also in the nonlinear geometric framework.

The rapid growth of deep learning research, including within the field of computational mechanics, has resulted in an extensive and diverse body of literature. To help researchers identify key concepts and promising methodologies within this field, we provide an overview of deep learning in deterministic computational mechanics. Five main categories are identified and explored: simulation substitution, simulation enhancement, discretizations as neural networks, generative approaches, and deep reinforcement learning. This review focuses on deep learning methods rather than applications for computational mechanics, thereby enabling researchers to explore this field more effectively. As such, the review is not necessarily aimed at researchers with extensive knowledge of deep learning—instead, the primary audience is researchers on the verge of entering this field or those attempting to gain an overview of deep learning in computational mechanics. The discussed concepts are, therefore, explained as simple as possible.

Computational modeling of heterogeneous materials is increasingly relying on multiscale simulations which typically leverage the homogenization theory for scale coupling. Such simulations are prohibitively expensive and memory-intensive especially when modeling damage and fracture in large 3D components such as cast metallic alloys. To address these challenges, we develop a physics-constrained deep learning model that surrogates the microscale simulations. We build this model within a mechanistic data-driven framework such that it accurately predicts the effective microstructural responses under irreversible elasto-plastic hardening and softening deformations. To achieve high accuracy while reducing the reliance on labeled data, we design the architecture of our deep learning model based on damage mechanics and introduce a new loss component that increases the thermodynamical consistency of the model. We use mechanistic reduced-order models to generate the training data of the deep learning model and demonstrate that, in addition to achieving high accuracy on unseen deformation paths that include severe softening, our model can be embedded in 3D multiscale simulations with fracture. With this embedding, we also demonstrate that state-of-the-art techniques such as teacher forcing result in deep learning models that cause divergence in multiscale simulations. Our numerical experiments indicate that our model is more accurate than pure data-driven models and is much more efficient than mechanistic reduced-order models.

Elastomeric mechanical metamaterials exhibit unconventional behaviour, emerging from their microstructures often deforming in a highly nonlinear and unstable manner. Such microstructural pattern transformations lead to non-local behaviour and induce abrupt changes in the effective properties, beneficial for engineering applications. To avoid expensive simulations fully resolving the underlying microstructure, homogenization methods are employed. In this contribution, a systematic comparative study is performed, assessing the predictive capability of several computational homogenization schemes in the realm of two-dimensional elastomeric metamaterials with a square stacking of circular holes. In particular, classical first-order and two enriched schemes of second-order and micromorphic cmoputational homogenziation type are compared with ensemble-averaged full direct numerical simulations on three examples: uniform compression and bending of an infinite specimen, and compression of a finite specimen. It is shown that although the second-order scheme provides good qualitative predictions, it fails in accurately capturing bifurcation strains and slightly over-predicts the homogenized response. The micromorphic method provides the most accurate prediction for tested examples, although soft boundary layers induce large errors at small scale ratios. The first-order scheme yields good predictions for high separations of scales, but suffers from convergence issues, especially when localization occurs.

Vertically aligned carbon nanotube (VACNT) arrays are moderately dense ensembles of nominally vertical carbon nanotubes (CNT) tethered to a rigid substrate. Variations in their synthesis protocols translate to largely unpredictable fluctuations in height, density, tortuosity and stiffness of the individual CNTs. Consequently, experimental studies on compression of these VACNT arrays exhibit a variety of responses. Moreover, many experimental studies report concerted buckling behaviour of the CNTs under compression. Numerical modelling of such coordinated behaviour in VACNT arrays poses many challenges. Each CNT can be modelled as a flexible beam capable of large deformations, allowing for tortuous initial shapes, mutual and/or self interactions that can be repulsive or attractive and periodic boundary conditions. Confining ourselves to a set of minimally realistic 2-dimensional parametric studies, we attempt to address how geometry/property fluctuations in an array of interacting columns leads to different types of collective compressive responses. We model each CNT as a geometrically exact beam using an established framework. A novel contact formulation is employed to model their mutual van der Waals interactions. In all cases, we capture coordinated buckling and are able to negotiate the response in the post-buckling stages. We first model ideal vertical arrays of defect-free CNTs and then discuss the effects of fluctuations in height, density, stiffness and tortuosity on their compressive behaviour.

Abstract

A nonparametric surrogate model for ductile failure is developed from simulation results on cells with a random distribution of voids. This model fully takes into account the anisotropy induced by the simulation conditions. The metamodeling strategy uses Gaussian Process Regression coupled with a multifidelity approach involving simulations on a cell with a single void. Through cokriging and metamodel parameter transfer, information can be transferred from the unit cell simulations to the model on random cells. This allows an increased accuracy, for a given computational capacity. Strategies for adaptive experimental design are also investigated.

We present an isogeometric analysis (IGA) framework for structural vibrations involving complex geometries. The framework is based on the Complex-Geometry IGA Mesh Generation (CGIMG) method. The CGIMG process is flexible and can accommodate, without a major effort, challenging complex-geometry applications in computational mechanics. To demonstrate how the new IGA framework significantly increases the computational effectiveness, in a set of structural-vibration test computations, we compare the accuracies attained by the IGA and finite element (FE) method as the number of degrees-of-freedom is increased. The results show that the NURBS meshes lead to faster convergence and higher accuracy compared to both linear and quadratic FE meshes. The clearly defined IGA mesh generation process and significant per-degree-of-freedom accuracy advantages of IGA over FE discretization make IGA more accessible, reliable, and attractive in applications of both academic and industrial interest. We note that the accuracy of a structural mechanics discretization, which may be assessed through eigenfrequency analysis, plays an important role in the overall accuracy of fluid–structure interaction computations.

Abstract

Data-driven models utilizing powerful artificial intelligence (AI) algorithms have been implemented over the past two decades in different fields of simulation-based engineering science. Most numerical procedures involve processing data sets developed from physical or numerical experiments to create closed-form formulae to predict the corresponding systems’ mechanical response. Efficient AI methodologies that will allow the development and use of accurate predictive models for solving computational intensive engineering problems remain an open issue. In this research work, high-performance machine learning (ML) algorithms are proposed for modeling structural mechanics-related problems, which are implemented in parallel and distributed computing environments to address extremely computationally demanding problems. Four machine learning algorithms are proposed in this work and their performance is investigated in three different structural engineering problems. According to the parametric investigation of the prediction accuracy, the extreme gradient boosting with extended hyper-parameter optimization (XGBoost-HYT-CV) was found to be more efficient regarding the generalization errors deriving a 4.54% residual error for all test cases considered. Furthermore, a comprehensive statistical analysis of the residual errors and a sensitivity analysis of the predictors concerning the target variable are reported. Overall, the proposed models were found to outperform the existing ML methods, where in one case the residual error was decreased by 3-fold. Furthermore, the proposed algorithms demonstrated the generic characteristic of the proposed ML framework for structural mechanics problems.

We applied physics-informed neural networks to solve the constitutive relations for nonlinear, path-dependent material behavior. As a result, the trained network not only satisfies all thermodynamic constraints but also instantly provides information about the current material state (i.e., free energy, stress, and the evolution of internal variables) under any given loading scenario without requiring initial data. One advantage of this work is that it bypasses the repetitive Newton iterations needed to solve nonlinear equations in complex material models. Furthermore, after training, the proposed approach requires significantly less effort in terms of implementation and computing time compared to the traditional methods. The trained model can be directly used in any finite element package (or other numerical methods) as a user-defined material model. We tested this methodology on rate-independent processes such as the classical von Mises plasticity model with a nonlinear hardening law, as well as local damage models for interface cracking behavior with a nonlinear softening law. In order to demonstrate the applicability of the methodology in handling complex path dependency in a three-dimensional (3D) scenario, we tested the approach using the equations governing a damage model for a three-dimensional interface model. Such models are frequently employed for intergranular fracture at grain boundaries. However, challenges remain in the proper definition of collocation points and in integrating several non-equality constraints that become active or non-active simultaneously. As long as we are in the training regime, we have observed a perfect agreement between the results obtained through the proposed methodology and those obtained using the classical approach. Finally, we compare this new approach against available standard methods and discuss the potential and remaining challenges for future developments.