Computer Methods in Applied Mechanics and Engineering最新文献_第2页

Concurrent topology optimization of two-scale structures considering high-cycle fatigue damage 考虑高周疲劳损伤的双尺度结构并行拓扑优化

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-13 DOI: 10.1016/j.cma.2026.118820

Xiaopeng Zhang , Zheng Ni , Yaguang Wang , Junling Fan

Two-scale structures demonstrate great potential in engineering due to their superior mechanical performance. However, under variable-amplitude loading, the analysis of structural fatigue response is complex, which makes the fatigue design of two-scale structures a challenge. In this study, we propose a concurrent topology optimization method considering high-cycle fatigue damage under variable-amplitude loading, which controls the maximum fatigue damage by designing the microstructure and its distribution at the macro scale under given volume constraints at both scales. To facilitate fatigue analysis under complex loading conditions, the rainflow counting method is employed to convert load history into analyzable cyclic loads. By incorporating the Palmgren-Miner linear cumulative damage rule into the microscale homogenization method, the fatigue damage at the microscale can be effectively analyzed. In fatigue damage analysis, three damage models signed von Mises, Brown-Miller, and Dang Van are considered. To address the challenge of microscale fatigue localization caused by highly nonlinear damage distribution, penalized fatigue damage constraints are defined by scaling the fatigue damage values. Based on the adjoint variable method, sensitivity analysis for the fatigue damage constraints is performed to update the design variables through the Method of Moving Asymptotes (MMA). Numerical examples demonstrate that the optimized design can effectively control fatigue damage. The results confirm that fatigue damage is more severe under tensile than compressive loading, a fact that directly leads to differing optimal designs.

双尺度结构由于其优越的力学性能，在工程上显示出巨大的潜力。然而，在变幅荷载作用下，结构的疲劳响应分析比较复杂，这给双尺度结构的疲劳设计带来了挑战。本文提出了一种考虑变幅载荷下高周疲劳损伤的并行拓扑优化方法，该方法通过在给定体积约束下设计宏观尺度上的微观结构及其分布来控制两尺度下的最大疲劳损伤。为了便于复杂载荷条件下的疲劳分析，采用雨流计数法将载荷历史转换为可分析的循环载荷。将Palmgren-Miner线性累积损伤规律引入到微尺度均匀化方法中，可以有效地分析微尺度下的疲劳损伤。在疲劳损伤分析中，考虑了von Mises、Brown-Miller和Dang Van三种损伤模型。为了解决由高度非线性损伤分布引起的微尺度疲劳局部化问题，通过对疲劳损伤值进行缩放来定义惩罚疲劳损伤约束。基于伴随变量法，对疲劳损伤约束进行敏感性分析，通过移动渐近线法更新设计变量。数值算例表明，优化设计能有效地控制疲劳损伤。结果证实，拉伸载荷下的疲劳损伤比压缩载荷更严重，这一事实直接导致了不同的优化设计。

{"title":"Concurrent topology optimization of two-scale structures considering high-cycle fatigue damage","authors":"Xiaopeng Zhang , Zheng Ni , Yaguang Wang , Junling Fan","doi":"10.1016/j.cma.2026.118820","DOIUrl":"10.1016/j.cma.2026.118820","url":null,"abstract":"<div><div>Two-scale structures demonstrate great potential in engineering due to their superior mechanical performance. However, under variable-amplitude loading, the analysis of structural fatigue response is complex, which makes the fatigue design of two-scale structures a challenge. In this study, we propose a concurrent topology optimization method considering high-cycle fatigue damage under variable-amplitude loading, which controls the maximum fatigue damage by designing the microstructure and its distribution at the macro scale under given volume constraints at both scales. To facilitate fatigue analysis under complex loading conditions, the rainflow counting method is employed to convert load history into analyzable cyclic loads. By incorporating the Palmgren-Miner linear cumulative damage rule into the microscale homogenization method, the fatigue damage at the microscale can be effectively analyzed. In fatigue damage analysis, three damage models signed von Mises, Brown-Miller, and Dang Van are considered. To address the challenge of microscale fatigue localization caused by highly nonlinear damage distribution, penalized fatigue damage constraints are defined by scaling the fatigue damage values. Based on the adjoint variable method, sensitivity analysis for the fatigue damage constraints is performed to update the design variables through the Method of Moving Asymptotes (MMA). Numerical examples demonstrate that the optimized design can effectively control fatigue damage. The results confirm that fatigue damage is more severe under tensile than compressive loading, a fact that directly leads to differing optimal designs.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118820"},"PeriodicalIF":7.3,"publicationDate":"2026-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MultiLevel variational MultiScale (ML-VMS) framework for large-scale simulation 大规模模拟的多级变分多尺度（ML-VMS）框架

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-13 DOI: 10.1016/j.cma.2026.118807

Lei Zhang , Jiachen Guo , Shaoqiang Tang , Thomas J.R. Hughes , Wing Kam Liu

In this paper, we propose the MultiLevel Variational MultiScale (ML-VMS) method, a novel approach that seamlessly integrates a multilevel mesh strategy into the Variational Multiscale (VMS) framework. A key feature of the ML-VMS method is the use of the Convolution Hierarchical Deep-learning Neural Network (C-HiDeNN) as the approximation basis, which enables fine-grained control over the trade-off between computational efficiency and interpolation accuracy. The framework employs a coarse mesh throughout the domain, with localized fine meshes placed only in subdomains of high interest, such as those surrounding a source. Solutions at different resolutions are robustly coupled through the variational weak form and interface conditions. Crucially, our method departs from existing VMS-based multilevel approaches by approximating the fine-scale solution directly using the fine-scale basis functions. Compared to existing multilevel methods, ML-VMS (1) can couple an arbitrary number of mesh levels across different scales using variational multiscale framework; (2) allows approximating functions with arbitrary orders with linear finite element mesh due to the C-HiDeNN basis; (3) is supported by a rigorous theoretical error analysis; (4) features several tunable hyperparameters (e.g., order p, patch size s) with a systematic guide for their selection. We first show the theoretical error estimates of ML-VMS. Then through numerical examples, we demonstrate that ML-VMS with the C-HiDeNN takes less computational time than the FEM basis given comparable accuracy. Furthermore, we incorporate a space-time reduced-order model (ROM) based on C-HiDeNN-Tensor Decomposition (TD) into the ML-VMS framework. For a large-scale single-track laser powder bed fusion (LPBF) transient heat transfer problem that is equivalent to a full-order finite element model with 10¹⁰ spatial degrees of freedom (DoFs), our three-level ML-VMS C-HiDeNN-TD demonstrates a promising speedup of approximately 5,000x speedup on a single CPU over a single-level linear FEM-TD ROM. We further validate the generality of ML-VMS through a 3D elasticity case study. Compared to the linear FEM-TD ROM, our approach achieves theoretical convergence rates and provides significant speedups with higher precision.

本文提出了多层次变分多尺度（ML-VMS）方法，这是一种将多层次网格策略无缝集成到变分多尺度（VMS）框架中的新方法。ML-VMS方法的一个关键特征是使用卷积层次深度学习神经网络（C-HiDeNN）作为近似基础，它可以对计算效率和插值精度之间的权衡进行细粒度控制。该框架在整个域中使用粗网格，而局部细网格仅放置在高兴趣的子域中，例如源周围的子域。不同分辨率下的解通过变分弱形式和界面条件进行鲁棒耦合。关键是，我们的方法与现有的基于vms的多层方法不同，它直接使用精细尺度基函数逼近精细尺度解。与现有的多尺度方法相比，ML-VMS(1)使用变分多尺度框架，可以在不同尺度上耦合任意数量的网格层；(2)由于C-HiDeNN基，允许用线性有限元网格逼近任意阶函数；(3)有严格的理论误差分析支持；(4)具有几个可调的超参数（例如，顺序p，补丁大小s），并具有系统的选择指南。我们首先展示了ML-VMS的理论误差估计。通过数值算例表明，在精度相当的情况下，基于C-HiDeNN的ML-VMS比基于FEM的计算时间更少。此外，我们将基于c - hidenn张量分解（TD）的时空降阶模型（ROM）整合到ML-VMS框架中。对于相当于1010空间自由度（dfs）的全阶有限元模型的大规模单轨道激光粉末床熔合（LPBF）瞬态传热问题，我们的三级ML-VMS C-HiDeNN-TD在单级线性FEM-TD ROM上的单个CPU上显示了大约5,000倍的加速。我们通过三维弹性案例研究进一步验证了ML-VMS的通用性。与线性FEM-TD ROM相比，我们的方法实现了理论收敛率，并提供了更高精度的显着加速。

{"title":"MultiLevel variational MultiScale (ML-VMS) framework for large-scale simulation","authors":"Lei Zhang , Jiachen Guo , Shaoqiang Tang , Thomas J.R. Hughes , Wing Kam Liu","doi":"10.1016/j.cma.2026.118807","DOIUrl":"10.1016/j.cma.2026.118807","url":null,"abstract":"<div><div>In this paper, we propose the MultiLevel Variational MultiScale (ML-VMS) method, a novel approach that seamlessly integrates a multilevel mesh strategy into the Variational Multiscale (VMS) framework. A key feature of the ML-VMS method is the use of the Convolution Hierarchical Deep-learning Neural Network (C-HiDeNN) as the approximation basis, which enables fine-grained control over the trade-off between computational efficiency and interpolation accuracy. The framework employs a coarse mesh throughout the domain, with localized fine meshes placed only in subdomains of high interest, such as those surrounding a source. Solutions at different resolutions are robustly coupled through the variational weak form and interface conditions. Crucially, our method departs from existing VMS-based multilevel approaches by approximating the fine-scale solution directly using the fine-scale basis functions. Compared to existing multilevel methods, ML-VMS (1) can couple an arbitrary number of mesh levels across different scales using variational multiscale framework; (2) allows approximating functions with arbitrary orders with linear finite element mesh due to the C-HiDeNN basis; (3) is supported by a rigorous theoretical error analysis; (4) features several tunable hyperparameters (e.g., order <em>p</em>, patch size <em>s</em>) with a systematic guide for their selection. We first show the theoretical error estimates of ML-VMS. Then through numerical examples, we demonstrate that ML-VMS with the C-HiDeNN takes less computational time than the FEM basis given comparable accuracy. Furthermore, we incorporate a space-time reduced-order model (ROM) based on C-HiDeNN-Tensor Decomposition (TD) into the ML-VMS framework. For a large-scale single-track laser powder bed fusion (LPBF) transient heat transfer problem that is equivalent to a full-order finite element model with 10<sup>10</sup> spatial degrees of freedom (DoFs), our three-level ML-VMS C-HiDeNN-TD demonstrates a promising speedup of approximately 5,000x speedup on a single CPU over a single-level linear FEM-TD ROM. We further validate the generality of ML-VMS through a 3D elasticity case study. Compared to the linear FEM-TD ROM, our approach achieves theoretical convergence rates and provides significant speedups with higher precision.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118807"},"PeriodicalIF":7.3,"publicationDate":"2026-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191974","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A naturally sharpened level-set formulation for incompressible free-surface flows 不可压缩自由表面流的自然锐化水平集公式

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-13 DOI: 10.1016/j.cma.2026.118798

Jongmin Rim , Jinhui Yan , Yuri Bazilevs

The level set method is widely employed in two-phase flow simulations due to its robustness in handling complex interface topological changes. However, it suffers from two main limitations. First, the method is not inherently mass conservative. Second, the signed-distance property of the level set field can deteriorate under strong convection, particularly in high Reynolds-number flows. Consequently, conventional level set methods often require auxiliary procedures such as sharpening (or re-distancing) and mass correction, which rely on and are sensitive to user-defined parameters and also increase implementation complexity and computational cost. Here, we present a naturally sharpened level-set formulation for incompressible air-water flows that is mass conservative and eliminates the need for these additional algorithmic steps. The resulting free-surface flow modeling and simulation framework is more efficient and robust as demonstrated through several challenging numerical test cases.

水平集方法在处理复杂界面拓扑变化方面具有鲁棒性，在两相流模拟中得到了广泛的应用。然而，它有两个主要的限制。首先，该方法本身并不是质量保守的。其次，在强对流条件下，特别是在高雷诺数流动条件下，水平集场的带符号距离性质会变差。因此，传统的水平集方法通常需要辅助过程，如锐化（或重新距离）和质量校正，这些过程依赖于用户定义的参数，并且对用户定义的参数很敏感，也增加了实现的复杂性和计算成本。在这里，我们提出了一个自然锐化的不可压缩空气-水流动的水平集公式，它是质量保守的，并且消除了这些额外算法步骤的需要。通过几个具有挑战性的数值测试案例证明，由此产生的自由表面流建模和仿真框架更加高效和鲁棒。

引用次数: 0

Variational data-consistent assimilation 变分数据一致同化

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-13 DOI: 10.1016/j.cma.2026.118804

Rylan Spence , Troy Butler , Clint Dawson

This work introduces a new class of four-dimensional variational data assimilation (4D-Var) methods grounded in data-consistent inversion (DCI) theory. The methods extend classical 4D-Var by incorporating a predictability-aware regularization term. The first method formulated is referred to as Data-Consistent 4D-Var (DC-4DVar), which is then enhanced using a Weighted Mean Error (WME) quantity-of-interest map to construct the DC-WME 4D-Var method. While the DC and DC-WME cost functions both involve a predictability-aware regularization term, the DC-WME function includes a modification to the model-data misfit, thereby improving estimation accuracy, robustness, and theoretical consistency in nonlinear and partially observed dynamical systems. Proofs are provided that establish the existence and uniqueness of the minimizer and analyze how a predictability assumption that is common within the DCI framework helps to promote solution stability. Numerical experiments are presented on benchmark dynamical systems (Lorenz-63 and Lorenz-96) as well as for the shallow water equations (SWE). In the benchmark dynamical systems, the DC-WME 4D-Var formulation is shown to consistently outperform standard 4D-Var in reducing both error and bias while maintaining robustness under high observation noise and short assimilation windows. Despite introducing modest computational overhead, DC-WME 4D-Var delivers improvements in estimation performance and forecast skill, demonstrating its potential practicality and scalability for high-dimensional data assimilation problems.

本文介绍了一种基于数据一致性反演（DCI）理论的新型四维变分数据同化（4D-Var）方法。该方法通过加入可预测性感知正则化项扩展了经典的4D-Var。制定的第一种方法被称为数据一致性4D-Var (DC-4DVar)，然后使用加权平均误差（WME）兴趣量图对其进行增强，以构建DC-WME 4D-Var方法。虽然DC和DC- wme代价函数都涉及可预测性感知的正则化项，但DC- wme函数包括对模型数据不拟合的修正，从而提高了非线性和部分观测动力系统的估计精度、鲁棒性和理论一致性。提供了证明，建立了最小化的存在性和唯一性，并分析了DCI框架中常见的可预测性假设如何有助于促进解决方案的稳定性。给出了Lorenz-63和Lorenz-96两种基准动力系统以及浅水方程（SWE）的数值实验。在基准动力系统中，DC-WME 4D-Var公式在降低误差和偏差方面始终优于标准4D-Var，同时在高观测噪声和短同化窗口下保持鲁棒性。尽管引入了适度的计算开销，DC-WME 4D-Var在估计性能和预测技能方面提供了改进，展示了其在高维数据同化问题上的潜在实用性和可扩展性。

{"title":"Variational data-consistent assimilation","authors":"Rylan Spence , Troy Butler , Clint Dawson","doi":"10.1016/j.cma.2026.118804","DOIUrl":"10.1016/j.cma.2026.118804","url":null,"abstract":"<div><div>This work introduces a new class of four-dimensional variational data assimilation (4D-Var) methods grounded in data-consistent inversion (DCI) theory. The methods extend classical 4D-Var by incorporating a predictability-aware regularization term. The first method formulated is referred to as Data-Consistent 4D-Var (DC-4DVar), which is then enhanced using a Weighted Mean Error (WME) quantity-of-interest map to construct the DC-WME 4D-Var method. While the DC and DC-WME cost functions both involve a predictability-aware regularization term, the DC-WME function includes a modification to the model-data misfit, thereby improving estimation accuracy, robustness, and theoretical consistency in nonlinear and partially observed dynamical systems. Proofs are provided that establish the existence and uniqueness of the minimizer and analyze how a predictability assumption that is common within the DCI framework helps to promote solution stability. Numerical experiments are presented on benchmark dynamical systems (Lorenz-63 and Lorenz-96) as well as for the shallow water equations (SWE). In the benchmark dynamical systems, the DC-WME 4D-Var formulation is shown to consistently outperform standard 4D-Var in reducing both error and bias while maintaining robustness under high observation noise and short assimilation windows. Despite introducing modest computational overhead, DC-WME 4D-Var delivers improvements in estimation performance and forecast skill, demonstrating its potential practicality and scalability for high-dimensional data assimilation problems.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118804"},"PeriodicalIF":7.3,"publicationDate":"2026-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191976","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Polynomial chaos expansion for operator learning 算子学习的多项式混沌展开

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-12 DOI: 10.1016/j.cma.2026.118796

Himanshu Sharma , Lukáš Novák , Michael Shields

Operator learning (OL) has emerged as a powerful tool in scientific machine learning (SciML) for approximating mappings between infinite-dimensional functional spaces. One of its main applications is learning the solution operator of partial differential equations (PDEs). While much of the progress in this area has been driven by deep neural network-based approaches such as Deep Operator Networks (DeepONet) and Fourier Neural Operator (FNO), recent work has begun to explore traditional machine learning methods for OL. In this work, we introduce polynomial chaos expansion (PCE) as an OL method. PCE has been widely used for uncertainty quantification (UQ) and has recently gained attention in the context of SciML. For OL, we establish a mathematical framework that enables PCE to approximate operators in both purely data-driven and physics-informed settings. The proposed framework reduces the task of learning the operator to solving a system of equations for the PCE coefficients. Moreover, the framework provides UQ by simply post-processing the PCE coefficients, without any additional computational cost. We apply the proposed method to a diverse set of PDE problems to demonstrate its capabilities. Numerical results demonstrate the strong performance of the proposed method in both OL and UQ tasks, achieving excellent numerical accuracy and computational efficiency.

算子学习（Operator learning， OL）已成为科学机器学习（SciML）中用于逼近无限维函数空间之间映射的强大工具。它的主要应用之一是学习偏微分方程（PDEs）的解算子。虽然该领域的大部分进展是由基于深度神经网络的方法（如深度算子网络（DeepONet）和傅立叶神经算子（FNO））推动的，但最近的工作已经开始探索传统的机器学习方法。在这项工作中，我们引入多项式混沌展开（PCE）作为一种OL方法。PCE在不确定度量化（UQ）中得到了广泛的应用，近年来在scil中得到了广泛的关注。对于OL，我们建立了一个数学框架，使PCE能够在纯数据驱动和物理知情的设置中近似操作符。该框架将学习算子的任务简化为求解PCE系数方程组。此外，该框架通过简单地后处理PCE系数来提供UQ，而不需要任何额外的计算成本。我们将提出的方法应用于不同的PDE问题集来证明其能力。数值结果表明，该方法在OL和UQ任务中都具有较强的性能，具有较高的数值精度和计算效率。

{"title":"Polynomial chaos expansion for operator learning","authors":"Himanshu Sharma , Lukáš Novák , Michael Shields","doi":"10.1016/j.cma.2026.118796","DOIUrl":"10.1016/j.cma.2026.118796","url":null,"abstract":"<div><div>Operator learning (OL) has emerged as a powerful tool in scientific machine learning (SciML) for approximating mappings between infinite-dimensional functional spaces. One of its main applications is learning the solution operator of partial differential equations (PDEs). While much of the progress in this area has been driven by deep neural network-based approaches such as Deep Operator Networks (DeepONet) and Fourier Neural Operator (FNO), recent work has begun to explore traditional machine learning methods for OL. In this work, we introduce polynomial chaos expansion (PCE) as an OL method. PCE has been widely used for uncertainty quantification (UQ) and has recently gained attention in the context of SciML. For OL, we establish a mathematical framework that enables PCE to approximate operators in both purely data-driven and physics-informed settings. The proposed framework reduces the task of learning the operator to solving a system of equations for the PCE coefficients. Moreover, the framework provides UQ by simply post-processing the PCE coefficients, without any additional computational cost. We apply the proposed method to a diverse set of PDE problems to demonstrate its capabilities. Numerical results demonstrate the strong performance of the proposed method in both OL and UQ tasks, achieving excellent numerical accuracy and computational efficiency.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118796"},"PeriodicalIF":7.3,"publicationDate":"2026-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146161061","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A learning-based domain decomposition method 一种基于学习的领域分解方法

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-11 DOI: 10.1016/j.cma.2026.118799

Rui Wu , Nikola Kovachki , Burigede Liu

Recent developments in mechanical, aerospace, and structural engineering have driven a growing need for efficient ways to model and analyze structures at much larger and more complex scales than before. While established numerical methods like the Finite Element Method remain reliable, they often struggle with computational cost and scalability when dealing with large and geometrically intricate problems. In recent years, neural network-based methods have shown promise because of their ability to efficiently approximate nonlinear mappings. However, most existing neural approaches are still largely limited to simple domains, which makes it difficult to apply to real-world partial differential equations (PDEs) involving complex geometries. In this paper, we propose a learning-based domain decomposition method (L-DDM) that addresses this gap. Our approach uses a single, pre-trained neural operator-originally trained on simple domains-as a surrogate model within a domain decomposition scheme, allowing us to tackle large and complicated domains efficiently. We provide a general theoretical result on the existence of neural operator approximations in the context of domain decomposition solution of abstract PDEs. We then demonstrate our method by accurately approximating solutions to elliptic PDEs with discontinuous microstructures in complex geometries, using a physics-pretrained neural operator (PPNO). Our results show that this approach not only outperforms current state-of-the-art methods on these challenging problems, but also offers resolution-invariance and strong generalization to microstructural patterns unseen during training.

最近在机械、航空航天和结构工程方面的发展推动了对比以前更大、更复杂尺度的结构建模和分析的有效方法的需求。虽然现有的数值方法，如有限元法仍然是可靠的，但在处理大型和几何上复杂的问题时，它们经常受到计算成本和可扩展性的困扰。近年来，基于神经网络的方法因其有效逼近非线性映射的能力而显示出前景。然而，大多数现有的神经方法仍然很大程度上局限于简单的领域，这使得难以应用于涉及复杂几何的现实世界的偏微分方程（PDEs）。在本文中，我们提出了一种基于学习的领域分解方法（L-DDM）来解决这一差距。我们的方法使用一个单独的、预先训练的神经算子（最初是在简单的域上训练的）作为域分解方案中的代理模型，使我们能够有效地处理大型和复杂的域。给出了抽象偏微分方程域分解解中神经算子近似存在性的一般理论结果。然后，我们通过使用物理预训练的神经算子（PPNO）精确逼近具有复杂几何形状的不连续微结构的椭圆偏微分方程的解来证明我们的方法。我们的研究结果表明，该方法不仅在这些具有挑战性的问题上优于当前最先进的方法，而且还提供了分辨率不变性和对训练中看不到的微观结构模式的强泛化。

{"title":"A learning-based domain decomposition method","authors":"Rui Wu , Nikola Kovachki , Burigede Liu","doi":"10.1016/j.cma.2026.118799","DOIUrl":"10.1016/j.cma.2026.118799","url":null,"abstract":"<div><div>Recent developments in mechanical, aerospace, and structural engineering have driven a growing need for efficient ways to model and analyze structures at much larger and more complex scales than before. While established numerical methods like the Finite Element Method remain reliable, they often struggle with computational cost and scalability when dealing with large and geometrically intricate problems. In recent years, neural network-based methods have shown promise because of their ability to efficiently approximate nonlinear mappings. However, most existing neural approaches are still largely limited to simple domains, which makes it difficult to apply to real-world partial differential equations (PDEs) involving complex geometries. In this paper, we propose a learning-based domain decomposition method (L-DDM) that addresses this gap. Our approach uses a single, pre-trained neural operator-originally trained on simple domains-as a surrogate model within a domain decomposition scheme, allowing us to tackle large and complicated domains efficiently. We provide a general theoretical result on the existence of neural operator approximations in the context of domain decomposition solution of abstract PDEs. We then demonstrate our method by accurately approximating solutions to elliptic PDEs with discontinuous microstructures in complex geometries, using a physics-pretrained neural operator (PPNO). Our results show that this approach not only outperforms current state-of-the-art methods on these challenging problems, but also offers resolution-invariance and strong generalization to microstructural patterns unseen during training.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118799"},"PeriodicalIF":7.3,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146161065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Gap-SBM: A new conceptualization of the shifted boundary method with optimal convergence for the Neumann and Dirichlet problems Gap-SBM: Neumann和Dirichlet问题的最优收敛移动边界法的新概念

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-11 DOI: 10.1016/j.cma.2026.118793

J. Haydel Collins , Kangan Li , Alexei Lozinski , Guglielmo Scovazzi

We propose and mathematically analyze a new Shifted Boundary Method for the treatment of Dirichlet and Neumann boundary conditions, with provable optimal accuracy in the L²- and H¹-norms of the error. The proposed method is built on three stages. First, the distance map between the SBM surrogate boundary and the true boundary is used to construct an approximation to the geometry of the gap between the two. Then, the representations of the numerical solution and test functions are extended from the surrogate domain to the such gap. Finally, approximate quadrature formulas and specific shift operators are applied to integrate a variational formulation that also involves the fields extended in the gap. An extensive set of two- and three-dimensional tests demonstrates the theoretical findings and the overall optimal performance of the proposed method.

我们提出并分析了一种新的移位边界法来处理Dirichlet和Neumann边界条件，并证明了在误差的L2-和h1 -范数下的最优精度。该方法分为三个阶段。首先，使用SBM代理边界和真边界之间的距离图来构建两者之间间隙的几何形状的近似值。然后，将数值解和测试函数的表示从代理域扩展到该间隙。最后，应用近似正交公式和特定的移位算子来积分一个变分公式，该变分公式也涉及在间隙中扩展的域。一组广泛的二维和三维试验证明了理论发现和所提出的方法的整体最佳性能。

引用次数: 0

A rotation-based approach to third medium contact regularization 基于旋转的第三介质接触正则化方法

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-11 DOI: 10.1016/j.cma.2026.118801

Vilmer Dahlberg, Filip Sjövall, Anna Dalklint, Mathias Wallin

The third medium contact method utilizes a fictitious “third” medium to implicitly model contact interactions. When contact occurs, the fictitious medium is severely deformed which necessitates numerical regularization to ensure numerical stability. Ultimately, this regularization should promote good element quality but otherwise not interfere with the modeling of the contact mechanics. One approach penalizes both stretch and rotational deformation modes using the displacement Hessian which requires higher order elements. Another approach introduces additional degrees of freedom to penalize an approximation of the rotation gradient which drastically increases the system size. We propose a new regularization based on an approximation of the rotation gradient in the fictitious medium, which does not penalize stretch deformation modes and can be used with first-order elements. The efficacy of our method is exemplified using several numerical examples including benchmark tests, an investigation of parasitic forces in the third medium and a novel application to general loading condition.

第三种媒介接触方法利用虚拟的“第三”媒介来隐式地模拟接触相互作用。当接触发生时，虚拟介质会发生严重变形，因此需要进行数值正则化以保证数值稳定性。最终，这种正则化应该促进良好的元素质量，但否则不会干扰接触力学的建模。一种方法使用位移Hessian来惩罚拉伸和旋转变形模式，这需要高阶元素。另一种方法引入了额外的自由度来惩罚旋转梯度的近似值，这极大地增加了系统大小。我们提出了一种基于虚拟介质中旋转梯度近似的新正则化方法，该方法不惩罚拉伸变形模式，可用于一阶元素。通过基准测试、第三介质中寄生力的研究以及在一般加载条件下的新应用，验证了该方法的有效性。

{"title":"A rotation-based approach to third medium contact regularization","authors":"Vilmer Dahlberg, Filip Sjövall, Anna Dalklint, Mathias Wallin","doi":"10.1016/j.cma.2026.118801","DOIUrl":"10.1016/j.cma.2026.118801","url":null,"abstract":"<div><div>The third medium contact method utilizes a fictitious “third” medium to implicitly model contact interactions. When contact occurs, the fictitious medium is severely deformed which necessitates numerical regularization to ensure numerical stability. Ultimately, this regularization should promote good element quality but otherwise not interfere with the modeling of the contact mechanics. One approach penalizes both stretch and rotational deformation modes using the displacement Hessian which requires higher order elements. Another approach introduces additional degrees of freedom to penalize an approximation of the rotation gradient which drastically increases the system size. We propose a new regularization based on an approximation of the rotation gradient in the fictitious medium, which does not penalize stretch deformation modes and can be used with first-order elements. The efficacy of our method is exemplified using several numerical examples including benchmark tests, an investigation of parasitic forces in the third medium and a novel application to general loading condition.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118801"},"PeriodicalIF":7.3,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146152926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Monotone peridynamic neural operator for nonlinear material modeling with conditionally unique solutions 具有条件唯一解的非线性材料建模的单调周动力神经算子

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-11 DOI: 10.1016/j.cma.2026.118792

Jihong Wang , Xiaochuan Tian , Zhongqiang Zhang , Stewart Silling , Siavash Jafarzadeh , Yue Yu

Nonlocal continuum mechanics models, including peridynamics, have emerged as powerful tools for describing the mechanical responses of complex nonlinear materials. In typical applications of peridynamics, the functional form of the material model is prescribed in advance, based on the analyst’s preferences and insight, creating the need for time-consuming calibration and validation for the particular material at hand. Although data-driven methods were proposed to streamline the modeling process, the well-posedness of these learned peridynamic models is generally not guaranteed, which creates the possibility of non-physical solutions in downstream simulation tasks.

In this study, we address this challenge of developing an accurate data-driven model with known uniqueness properties. To do this, we introduce the monotone peridynamic neural operator (MPNO), a novel approach for learning a data-driven nonlocal constitutive model with guaranteed well-posedness for certain classes of problems. Our approach learns a nonlocal kernel together with a nonlinear constitutive relation, while ensuring solution uniqueness through a monotone gradient network. This architectural constraint on the gradient induces the convexity of the learnt energy density function. This guarantees the uniqueness of solutions in the small deformation regime. To validate our approach, we evaluate MPNO’s performance on both synthetic and real-world datasets. On synthetic datasets generated with a manufactured kernel and constitutive relation, we show, both theoretically and numerically, that the learnt model converges to the ground-truth as the measurement grid size decreases. Additionally, our MPNO exhibits superior generalization capabilities in comparison with conventional neural networks. It yields smaller displacement solution errors in down-stream tasks with unseen loadings that are outside of the distribution of training samples. Finally, we showcase the practical utility of our approach through applications in learning a homogenized model from molecular dynamics data, highlighting the model’s expressivity and physical interpretability in real-world scenarios.

包括周动力学在内的非局部连续介质力学模型已经成为描述复杂非线性材料力学响应的有力工具。在周动力学的典型应用中，材料模型的功能形式是根据分析人员的偏好和见解预先规定的，这就需要对手头的特定材料进行耗时的校准和验证。虽然提出了数据驱动的方法来简化建模过程，但通常不能保证这些学习到的周动力学模型的适定性，这就为下游仿真任务中的非物理解创造了可能性。在本研究中，我们解决了开发具有已知唯一性属性的准确数据驱动模型的挑战。为了做到这一点，我们引入了单调周期动态神经算子（MPNO），这是一种学习数据驱动的非局部本构模型的新方法，对于某些类别的问题具有保证的适定性。该方法学习非局部核和非线性本构关系，同时通过单调梯度网络保证解的唯一性。梯度上的结构约束导致了学习到的能量密度函数的凸性。这保证了小变形区解的唯一性。为了验证我们的方法，我们在合成数据集和真实数据集上评估了MPNO的性能。在使用人造核和本构关系生成的合成数据集上，我们从理论上和数值上表明，随着测量网格大小的减小，学习模型收敛于基本事实。此外，与传统神经网络相比，我们的MPNO表现出优越的泛化能力。在训练样本分布之外的未见负载的下游任务中，它产生更小的位移解误差。最后，我们通过从分子动力学数据中学习均匀化模型的应用，展示了我们方法的实际效用，突出了模型在现实世界场景中的表现力和物理可解释性。

{"title":"Monotone peridynamic neural operator for nonlinear material modeling with conditionally unique solutions","authors":"Jihong Wang , Xiaochuan Tian , Zhongqiang Zhang , Stewart Silling , Siavash Jafarzadeh , Yue Yu","doi":"10.1016/j.cma.2026.118792","DOIUrl":"10.1016/j.cma.2026.118792","url":null,"abstract":"<div><div>Nonlocal continuum mechanics models, including peridynamics, have emerged as powerful tools for describing the mechanical responses of complex nonlinear materials. In typical applications of peridynamics, the functional form of the material model is prescribed in advance, based on the analyst’s preferences and insight, creating the need for time-consuming calibration and validation for the particular material at hand. Although data-driven methods were proposed to streamline the modeling process, the well-posedness of these learned peridynamic models is generally not guaranteed, which creates the possibility of non-physical solutions in downstream simulation tasks.</div><div>In this study, we address this challenge of developing an accurate data-driven model with known uniqueness properties. To do this, we introduce the monotone peridynamic neural operator (MPNO), a novel approach for learning a data-driven nonlocal constitutive model with guaranteed well-posedness for certain classes of problems. Our approach learns a nonlocal kernel together with a nonlinear constitutive relation, while ensuring solution uniqueness through a monotone gradient network. This architectural constraint on the gradient induces the convexity of the learnt energy density function. This guarantees the uniqueness of solutions in the small deformation regime. To validate our approach, we evaluate MPNO’s performance on both synthetic and real-world datasets. On synthetic datasets generated with a manufactured kernel and constitutive relation, we show, both theoretically and numerically, that the learnt model converges to the ground-truth as the measurement grid size decreases. Additionally, our MPNO exhibits superior generalization capabilities in comparison with conventional neural networks. It yields smaller displacement solution errors in down-stream tasks with unseen loadings that are outside of the distribution of training samples. Finally, we showcase the practical utility of our approach through applications in learning a homogenized model from molecular dynamics data, highlighting the model’s expressivity and physical interpretability in real-world scenarios.</div></div>","PeriodicalId":55222,"journal":{"name":"Computer Methods in Applied Mechanics and Engineering","volume":"453 ","pages":"Article 118792"},"PeriodicalIF":7.3,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146152928","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Isogeometric fluid-structure interaction using a mixed continuous/discontinuous Galerkin scheme 使用混合连续/不连续伽辽金格式的等几何流固相互作用

IF 7.3 1区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

Computer Methods in Applied Mechanics and Engineering

Pub Date : 2026-02-11 DOI: 10.1016/j.cma.2026.118795

Régis Duvigneau

A mixed continuous / discontinuous Galerkin scheme is introduced for the simulation of fluid-structure interaction problems in an isogeometric analysis framework. The properties of Non-Uniform Rational B-Spline basis functions are leveraged to enable an exact transfer of the structural displacement to the fluid domain, while using different discretizations and refinements on the two sides of the coupling interface. The proposed approach is applied to the simulation of a compressible flow around an elastic wing membrane and to a classical fluid-structure benchmark involving the flow around a cylinder equipped with a hyper-elastic bar. For both cases, the results obtained are compared to those found in the literature to assess the accuracy of the proposed method.

采用连续/不连续混合Galerkin格式在等几何分析框架下模拟流固耦合问题。利用非均匀有理b样条基函数的性质，在耦合界面两侧采用不同的离散化和细化方法，使结构位移精确地传递到流体域。将该方法应用于弹性翼膜周围可压缩流动的模拟和一个经典的流体结构基准，该基准涉及装有超弹性杆的圆柱体周围的流动。对于这两种情况，所获得的结果将与文献中发现的结果进行比较，以评估所提出方法的准确性。

引用次数: 0