Engineering Fracture Mechanics最新文献_第9页

Experimental study and machine learning-based fatigue life prediction of thermoplastic laminated composites after low-velocity impact 热塑性层合复合材料低速冲击疲劳寿命的实验研究与机器学习预测

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2025-12-31 DOI: 10.1016/j.engfracmech.2025.111821

Zhen Yue , Chi Zhan , Hanming Yang , Yifang Qin , Ningge Fan , Shunhua Chen

Fiber-reinforced thermoplastic laminated composites are highly sensitive to low-velocity impacts, which induces barely visible damage and accelerates fatigue failure under cyclic loading, thereby reducing structural service life. Conventional approaches for predicting post-impact fatigue behavior rely heavily on experimental testing and numerical simulations, which are often time-consuming and costly. Moreover, existing machine learning studies pay limited attention to the effects of initial impact-induced damage. To address these limitations, this study combines experimental and machine learning-based approaches for accurate fatigue life prediction of laminated composites after low-velocity impacts. Low-velocity impact tests are performed on composite specimens, and their impact responses are recorded. The induced damage is characterized using non-destructive techniques. The impacted specimens are then subjected to tensile–tensile fatigue tests to determine residual fatigue life and construct the corresponding S–N curves. The experimental results show that higher energy impacts significantly reduce the fatigue life of laminated composites. To improve model robustness, a fatigue knowledge-based data augmentation strategy via S–N curves is presented to expand the fatigue life dataset. Multiple machine learning algorithms, including Support Vector Machines (SVM), Random Forests (RF), Back-Propagation Neural Networks (BPNN), and Bayesian Neural Networks (BNNs), are introduced, trained, and optimized through hyperparameter tuning. The predictive results indicate that all employed models estimate post-impact fatigue life with reasonable accuracy, with BPNN and BNNs achieving the best overall performance.

纤维增强热塑性层合复合材料对低速冲击非常敏感，在循环载荷下产生几乎不可见的损伤并加速疲劳破坏，从而降低结构的使用寿命。传统的预测冲击后疲劳行为的方法严重依赖于实验测试和数值模拟，这些方法通常既耗时又昂贵。此外，现有的机器学习研究对初始冲击引起的损伤的影响关注有限。为了解决这些限制，本研究将实验和基于机器学习的方法结合起来，对层压复合材料在低速冲击后的疲劳寿命进行准确预测。对复合材料试样进行了低速冲击试验，并记录了其冲击响应。诱导损伤采用非破坏性技术表征。然后对冲击试样进行拉伸-拉伸疲劳试验，确定残余疲劳寿命并构建相应的S-N曲线。实验结果表明，高能量冲击会显著降低层合复合材料的疲劳寿命。为了提高模型的鲁棒性，提出了一种基于疲劳知识的S-N曲线数据增强策略，对疲劳寿命数据集进行扩展。多种机器学习算法，包括支持向量机（SVM），随机森林（RF），反向传播神经网络（BPNN）和贝叶斯神经网络（bnn），介绍，训练，并通过超参数调优。预测结果表明，所采用的模型对冲击后疲劳寿命的预测精度合理，其中bp神经网络和bnn神经网络的综合性能最好。

{"title":"Experimental study and machine learning-based fatigue life prediction of thermoplastic laminated composites after low-velocity impact","authors":"Zhen Yue , Chi Zhan , Hanming Yang , Yifang Qin , Ningge Fan , Shunhua Chen","doi":"10.1016/j.engfracmech.2025.111821","DOIUrl":"10.1016/j.engfracmech.2025.111821","url":null,"abstract":"<div><div>Fiber-reinforced thermoplastic laminated composites are highly sensitive to low-velocity impacts, which induces barely visible damage and accelerates fatigue failure under cyclic loading, thereby reducing structural service life. Conventional approaches for predicting post-impact fatigue behavior rely heavily on experimental testing and numerical simulations, which are often time-consuming and costly. Moreover, existing machine learning studies pay limited attention to the effects of initial impact-induced damage. To address these limitations, this study combines experimental and machine learning-based approaches for accurate fatigue life prediction of laminated composites after low-velocity impacts. Low-velocity impact tests are performed on composite specimens, and their impact responses are recorded. The induced damage is characterized using non-destructive techniques. The impacted specimens are then subjected to tensile–tensile fatigue tests to determine residual fatigue life and construct the corresponding S–N curves. The experimental results show that higher energy impacts significantly reduce the fatigue life of laminated composites. To improve model robustness, a fatigue knowledge-based data augmentation strategy via S–N curves is presented to expand the fatigue life dataset. Multiple machine learning algorithms, including Support Vector Machines (SVM), Random Forests (RF), Back-Propagation Neural Networks (BPNN), and Bayesian Neural Networks (BNNs), are introduced, trained, and optimized through hyperparameter tuning. The predictive results indicate that all employed models estimate post-impact fatigue life with reasonable accuracy, with BPNN and BNNs achieving the best overall performance.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111821"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145881615","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Towards a new methodology to characterize the fracture energies of the woven composite/copper interface in mode I and mode II: Application to printed circuit boards 一种表征I型和II型编织复合材料/铜界面断裂能的新方法：在印刷电路板上的应用

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2025-12-26 DOI: 10.1016/j.engfracmech.2025.111819

Charaf-Eddine Ziouani , Gautier Girard , Sébastien Mercier , François Lechleiter

The continuous trend towards miniaturization in electronic devices has stimulated the development of a new generation of printed circuit boards (PCBs) with embedded components. Throughout their lifespan, PCBs are subjected to thermal loads generated by heat from active components or the surrounding environment. In particular, mismatches in thermal expansion coefficients between materials are a leading cause of thermal stresses, often resulting in layer delamination, either between insulating substrates or at the copper-substrate interface. Traditionally, the peel test has been the dominant method for evaluating interfacial energy within PCBs, offering an estimate of the interface energy based on the IPC standard. During peeling, the copper layer often undergoes significant plastic deformation, complicating the precise determination of the fracture energy. Thus, achieving an accurate assessment of the mechanical response at the interface remains a challenging task. To overcome these limitations, we have designed a new specimen and adapted the Double Cantilever Beam (DCB) and End Notched Flexure (ENF) tests to the PCB context where layer thickness is significantly constrained (with copper layers ranging from 17 to

70 μ m

). Prior to experimentation, simulations demonstrate that, unlike the peel test, the DCB and ENF configurations exhibit minimal plastic dissipation. One of the main outcomes of the work is that a precise description of the plastic behavior of copper is not necessary to determine accurate estimations of the critical strain energy release rates in mode I and mode II. Furthermore, a notable advantage of these methods is their ability to maintain a controlled fracture mode, whereas the peel test inherently involves a mixed-mode (I and II) fracture process. The synergy between finite element analysis and experimental testing provides critical insights about the framework of application of the methods.

电子器件小型化的持续趋势刺激了新一代嵌入式元件印刷电路板（pcb）的发展。在其整个使用寿命中，pcb都受到来自有效元件或周围环境的热量产生的热负荷的影响。特别是，材料之间热膨胀系数的不匹配是热应力的主要原因，通常导致绝缘衬底之间或铜衬底界面处的层脱层。传统上，剥离测试一直是评估pcb内部界面能的主要方法，提供基于IPC标准的界面能估计。在剥离过程中，铜层经常发生显著的塑性变形，使断裂能的精确测定变得复杂。因此，实现在界面上的机械响应的准确评估仍然是一个具有挑战性的任务。为了克服这些限制，我们设计了一种新的试样，并将双悬臂梁（DCB）和端缺口弯曲（ENF）测试适应于层厚度明显受限的PCB环境（铜层范围从17到70μm）。在实验之前，模拟表明，与剥离试验不同，DCB和ENF结构表现出最小的塑性耗散。这项工作的主要结果之一是，铜的塑性行为的精确描述是没有必要的，以确定准确的估计临界应变能释放率在I型和II型。此外，这些方法的一个显著优势是它们能够保持可控的断裂模式，而剥离试验本质上涉及混合模式（I和II）断裂过程。有限元分析和实验测试之间的协同作用为方法的应用框架提供了关键的见解。

{"title":"Towards a new methodology to characterize the fracture energies of the woven composite/copper interface in mode I and mode II: Application to printed circuit boards","authors":"Charaf-Eddine Ziouani , Gautier Girard , Sébastien Mercier , François Lechleiter","doi":"10.1016/j.engfracmech.2025.111819","DOIUrl":"10.1016/j.engfracmech.2025.111819","url":null,"abstract":"<div><div>The continuous trend towards miniaturization in electronic devices has stimulated the development of a new generation of printed circuit boards (PCBs) with embedded components. Throughout their lifespan, PCBs are subjected to thermal loads generated by heat from active components or the surrounding environment. In particular, mismatches in thermal expansion coefficients between materials are a leading cause of thermal stresses, often resulting in layer delamination, either between insulating substrates or at the copper-substrate interface. Traditionally, the peel test has been the dominant method for evaluating interfacial energy within PCBs, offering an estimate of the interface energy based on the IPC standard. During peeling, the copper layer often undergoes significant plastic deformation, complicating the precise determination of the fracture energy. Thus, achieving an accurate assessment of the mechanical response at the interface remains a challenging task. To overcome these limitations, we have designed a new specimen and adapted the Double Cantilever Beam (DCB) and End Notched Flexure (ENF) tests to the PCB context where layer thickness is significantly constrained (with copper layers ranging from 17 to <span><math><mrow><mn>70</mn><mspace></mspace><mi>μ</mi><mi>m</mi></mrow></math></span>). Prior to experimentation, simulations demonstrate that, unlike the peel test, the DCB and ENF configurations exhibit minimal plastic dissipation. One of the main outcomes of the work is that a precise description of the plastic behavior of copper is not necessary to determine accurate estimations of the critical strain energy release rates in mode I and mode II. Furthermore, a notable advantage of these methods is their ability to maintain a controlled fracture mode, whereas the peel test inherently involves a mixed-mode (I and II) fracture process. The synergy between finite element analysis and experimental testing provides critical insights about the framework of application of the methods.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111819"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145882174","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Multi-crack competition induced by differential sintering in strain-tolerant thermal barrier coatings 耐热热障涂层中差异烧结引起的多裂纹竞争

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2026-01-02 DOI: 10.1016/j.engfracmech.2025.111837

Bowen Lv , Dingjun Li , Jie Mao , Chunming Deng , Changguang Deng , Min Liu , Kesong Zhou

Strain-tolerant thermal barrier coatings achieve superior thermal shock resistance by incorporating ceramic top coats with vertically cracked, segmented or columnar structures. In hydrogen-fueled gas turbines, the elevated operating temperatures intensify the sintering process, rendering the interactions among multiple cracks in these complex architectures particularly pronounced but still insufficiently understood. In this work, mechanisms underlying multi-crack competition driven by differential sintering were investigated through a combined experimental–numerical approach. Experimental characterization under both uniform/nonuniform temperature fields was conducted to capture sintering-induced structural and mechanical evolution. Based on these findings, a temperature-dependent constitutive model was developed within a variational principle framework and implemented in finite element simulations for fracture analyses. The model predictions were validated by thermal shock and sintering experiments under various thermomechanical boundary conditions. The results show that enhanced interfacial strength and differential sintering promote branching crack propagation in different regions of the ceramic top coat. Although interfacial delamination remains the dominant fracture mode, this failure mechanism can be mitigated through controlled interfacial strengthening and sintering gradients. A three-dimensional fracture mechanism map is further proposed to elucidate the relationships among sintering behavior, crack competition, and fracture modes in strain-tolerant ceramic coatings.

耐应变热障涂层通过结合具有垂直裂纹、分段或柱状结构的陶瓷面涂层来实现卓越的抗热震性。在氢燃料燃气轮机中，升高的工作温度加剧了烧结过程，使得这些复杂结构中多个裂缝之间的相互作用特别明显，但仍然没有充分了解。在这项工作中，通过结合实验-数值方法研究了由差异烧结驱动的多裂纹竞争机制。在均匀和非均匀温度场下进行了实验表征，以捕捉烧结引起的结构和力学演化。基于这些发现，在变分原理框架内建立了温度相关的本构模型，并将其应用于断裂分析的有限元模拟中。在不同的热力学边界条件下进行了热冲击和烧结实验，验证了模型的预测。结果表明，界面强度的增强和差异烧结促进了陶瓷面涂层不同区域的分支裂纹扩展。虽然界面分层仍然是主要的断裂模式，但这种破坏机制可以通过控制界面强化和烧结梯度来缓解。进一步提出了三维断裂机理图，以阐明耐应变陶瓷涂层的烧结行为、裂纹竞争和断裂模式之间的关系。

{"title":"Multi-crack competition induced by differential sintering in strain-tolerant thermal barrier coatings","authors":"Bowen Lv , Dingjun Li , Jie Mao , Chunming Deng , Changguang Deng , Min Liu , Kesong Zhou","doi":"10.1016/j.engfracmech.2025.111837","DOIUrl":"10.1016/j.engfracmech.2025.111837","url":null,"abstract":"<div><div>Strain-tolerant thermal barrier coatings achieve superior thermal shock resistance by incorporating ceramic top coats with vertically cracked, segmented or columnar structures. In hydrogen-fueled gas turbines, the elevated operating temperatures intensify the sintering process, rendering the interactions among multiple cracks in these complex architectures particularly pronounced but still insufficiently understood. In this work, mechanisms underlying multi-crack competition driven by differential sintering were investigated through a combined experimental–numerical approach. Experimental characterization under both uniform/nonuniform temperature fields was conducted to capture sintering-induced structural and mechanical evolution. Based on these findings, a temperature-dependent constitutive model was developed within a variational principle framework and implemented in finite element simulations for fracture analyses. The model predictions were validated by thermal shock and sintering experiments under various thermomechanical boundary conditions. The results show that enhanced interfacial strength and differential sintering promote branching crack propagation in different regions of the ceramic top coat. Although interfacial delamination remains the dominant fracture mode, this failure mechanism can be mitigated through controlled interfacial strengthening and sintering gradients. A three-dimensional fracture mechanism map is further proposed to elucidate the relationships among sintering behavior, crack competition, and fracture modes in strain-tolerant ceramic coatings.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111837"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145922182","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Dynamic response behavior and damage evolution mechanism of additively manufactured porous structures of composites under high strain rates 高应变率下复合材料增材多孔结构的动态响应行为及损伤演化机制

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2026-01-05 DOI: 10.1016/j.engfracmech.2026.111840

Xuanming Cai , Penglei Wang , Yang Hou , Zhiyong Wang , Wei Zhang , Zhongcheng Mu , Anxiao Guo , Linzhuang Han , Yunhao Yang , Yalin He , Bin Liu , Wenbo Xie

Short fiber-reinforced polymer-matrix composites (SFRPC), renowned for high specific strength, are widely employed in high strain-rate structures. Elucidating their dynamic response and failure mechanisms under high strain-rate loading is crucial for safety design and performance optimization. A three-dimensional multiscale constitutive model and failure criterion for SFRPC were developed using micro–macro mechanics, whose validity was verified by comparing quasi-static uniaxial tensile simulations of representative volume element (RVE) cells with experimental results at the macroscopic level. The RVE model extracted effective elastic constants under various loadings, acting as key dynamic parameters for high strain-rate simulations. Three types of SFRPC porous structures with different volume fractions were designed via triply periodic minimal surface (TPMS) equations and fabricated into specimens by 3D printing. Multiscale simulations and high strain-rate impact experiments investigated the dynamic response and damage evolution. Results show that the SFRPC structures exhibit strain-rate sensitivity under dynamic loading, with dynamic strength rising as strain rate increases. At similar strain rates, peak stress, specific energy absorption (SEA), and energy absorption efficiency (EAE) rise with higher volume fractions. SEA and EAE both increase with the strain rate, with EAE of higher volume fraction structures more influenced by strain rate effects. Microscopic damage analysis showed volume fraction strongly affects shear failure: 25 % and 35 % fractions show dominant fiber pull-out, while 45 % shows brittle fracture and plastic deformation. Multiscale simulations reproduced experimental damage patterns, and their multi-directional modes clarify internal damage evolution under high strain rate conditions.

短纤维增强聚合物基复合材料（SFRPC）以其高比强度而闻名，广泛应用于高应变率结构。阐明其在高应变率载荷下的动态响应和破坏机制对其安全设计和性能优化至关重要。采用微宏观力学方法建立了SFRPC的三维多尺度本构模型和破坏准则，并通过比较具有代表性的体积单元（RVE）单元的准静态单轴拉伸模拟与宏观水平的试验结果，验证了其有效性。RVE模型提取了不同载荷下的有效弹性常数，作为高应变率模拟的关键动力学参数。利用三周期最小表面（TPMS）方程设计了三种不同体积分数的SFRPC多孔结构，并通过3D打印将其制作成试样。多尺度模拟和高应变率冲击试验研究了其动态响应和损伤演化规律。结果表明：钢纤维混凝土结构在动荷载作用下表现出应变率敏感性，动强度随应变率的增大而增大；在相同应变速率下，峰值应力、比能吸收（SEA）和能量吸收效率（EAE）随体积分数的增加而增加。SEA和EAE均随应变速率的增大而增大，且体积分数越高的EAE受应变速率效应的影响越大。细观损伤分析表明，体积分数对剪切破坏影响较大，25%和35%的体积分数以纤维拔出为主，45%的体积分数以脆性断裂和塑性变形为主。多尺度模拟再现了实验损伤模式，其多向模式阐明了高应变速率条件下内部损伤的演化规律。

{"title":"Dynamic response behavior and damage evolution mechanism of additively manufactured porous structures of composites under high strain rates","authors":"Xuanming Cai , Penglei Wang , Yang Hou , Zhiyong Wang , Wei Zhang , Zhongcheng Mu , Anxiao Guo , Linzhuang Han , Yunhao Yang , Yalin He , Bin Liu , Wenbo Xie","doi":"10.1016/j.engfracmech.2026.111840","DOIUrl":"10.1016/j.engfracmech.2026.111840","url":null,"abstract":"<div><div>Short fiber-reinforced polymer-matrix composites (SFRPC), renowned for high specific strength, are widely employed in high strain-rate structures. Elucidating their dynamic response and failure mechanisms under high strain-rate loading is crucial for safety design and performance optimization. A three-dimensional multiscale constitutive model and failure criterion for SFRPC were developed using micro–macro mechanics, whose validity was verified by comparing quasi-static uniaxial tensile simulations of representative volume element (RVE) cells with experimental results at the macroscopic level. The RVE model extracted effective elastic constants under various loadings, acting as key dynamic parameters for high strain-rate simulations. Three types of SFRPC porous structures with different volume fractions were designed via triply periodic minimal surface (TPMS) equations and fabricated into specimens by 3D printing. Multiscale simulations and high strain-rate impact experiments investigated the dynamic response and damage evolution. Results show that the SFRPC structures exhibit strain-rate sensitivity under dynamic loading, with dynamic strength rising as strain rate increases. At similar strain rates, peak stress, specific energy absorption (SEA), and energy absorption efficiency (EAE) rise with higher volume fractions. SEA and EAE both increase with the strain rate, with EAE of higher volume fraction structures more influenced by strain rate effects. Microscopic damage analysis showed volume fraction strongly affects shear failure: 25 % and 35 % fractions show dominant fiber pull-out, while 45 % shows brittle fracture and plastic deformation. Multiscale simulations reproduced experimental damage patterns, and their multi-directional modes clarify internal damage evolution under high strain rate conditions.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111840"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145922254","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Dynamic damage evolution and fracture initiation in finite deformation ductile materials 有限变形韧性材料的动态损伤演化与断裂起裂

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2025-12-31 DOI: 10.1016/j.engfracmech.2025.111829

Zhongpan Li , Yan Li , Boumediene Nedjar , Ling Tao , Huijian Chen , Zhiqiang Feng

This paper presents a semi-explicit algorithm for modeling dynamic damage and fracture in ductile materials under finite deformation. The algorithm combines the efficiency of explicit methods with the stability of implicit schemes, enabling robust simulations in large deformation and contact scenarios. To further enhance numerical stability, a rotational stress update scheme based on Kirchhoff stress is implemented, which effectively handles rigid-body rotations and mitigates artificial stress artifacts. Frictional contact is addressed using an implicit algorithm based on the bi-potential method, ensuring stable and efficient contact resolution. The damage model is formulated within the continuum damage mechanics (CDM) framework, following the damage evolution theory of Chaboche and Lemaitre. Material nonlinearity is captured using an isotropic von Mises yield criterion. The proposed method is implemented in the plastic finite element program CCMPF and verified through a series of numerical examples. Two quasi-static simulations are first conducted to evaluate the mesh sensitivity of the local damage model and to verify the accuracy of the constitutive integration scheme. A dynamic Taylor impact, including both 2D and 3D cases, is performed to validate the algorithm under high strain-rate conditions. The results demonstrate the method’s accuracy, efficiency, and robustness in simulating dynamic failure in ductile materials.

本文提出了一种半显式的模拟有限变形韧性材料动态损伤和断裂的算法。该算法结合了显式方法的效率和隐式方法的稳定性，能够在大变形和接触场景下进行鲁棒模拟。为了进一步提高数值稳定性，实现了基于基尔霍夫应力的旋转应力更新方案，该方案有效地处理了刚体旋转并减轻了人为应力伪像。采用基于双势法的隐式算法求解摩擦接触，保证了稳定、高效的接触分辨率。根据Chaboche和Lemaitre的损伤演化理论，在连续损伤力学（CDM）框架下建立损伤模型。材料非线性被捕获使用各向同性冯米塞斯屈服准则。在塑性有限元程序CCMPF中实现了该方法，并通过一系列数值算例进行了验证。首先进行了两个准静态仿真，以评估局部损伤模型的网格敏感性，并验证本构积分方案的准确性。为了验证该算法在高应变率条件下的有效性，进行了二维和三维动态泰勒冲击实验。结果表明，该方法在模拟延性材料动态破坏方面具有准确性、有效性和鲁棒性。

{"title":"Dynamic damage evolution and fracture initiation in finite deformation ductile materials","authors":"Zhongpan Li , Yan Li , Boumediene Nedjar , Ling Tao , Huijian Chen , Zhiqiang Feng","doi":"10.1016/j.engfracmech.2025.111829","DOIUrl":"10.1016/j.engfracmech.2025.111829","url":null,"abstract":"<div><div>This paper presents a semi-explicit algorithm for modeling dynamic damage and fracture in ductile materials under finite deformation. The algorithm combines the efficiency of explicit methods with the stability of implicit schemes, enabling robust simulations in large deformation and contact scenarios. To further enhance numerical stability, a rotational stress update scheme based on Kirchhoff stress is implemented, which effectively handles rigid-body rotations and mitigates artificial stress artifacts. Frictional contact is addressed using an implicit algorithm based on the bi-potential method, ensuring stable and efficient contact resolution. The damage model is formulated within the continuum damage mechanics (CDM) framework, following the damage evolution theory of Chaboche and Lemaitre. Material nonlinearity is captured using an isotropic von Mises yield criterion. The proposed method is implemented in the plastic finite element program CCMPF and verified through a series of numerical examples. Two quasi-static simulations are first conducted to evaluate the mesh sensitivity of the local damage model and to verify the accuracy of the constitutive integration scheme. A dynamic Taylor impact, including both 2D and 3D cases, is performed to validate the algorithm under high strain-rate conditions. The results demonstrate the method’s accuracy, efficiency, and robustness in simulating dynamic failure in ductile materials.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111829"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145922262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 novel boundary displacement decomposition method for stress intensity factor calculation in 2D inclined crack 一种计算二维倾斜裂纹应力强度因子的边界位移分解方法

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2026-01-10 DOI: 10.1016/j.engfracmech.2026.111859

Sanshao Zhuang , Tao Hu , Junfeng Zhang , Wenqing Zheng , Miaolin Feng

For a 2D inclined crack, a novel boundary displacement decomposition method based on the submodel technique and using circular harmonic (CH) as basis functions is proposed to calculate the stress intensity factor (SIF). This method computes the SIF by linearly superimposing the precomputed SIF values of the basis functions and the decomposition coefficients of the submodel boundary displacement, which are obtained from finite element analysis (FEA). The approach circumvents the computationally intensive task of generating crack meshes in the global model during analysis. Specifically, CH functions are employed as the basis functions for submodel boundary displacement decomposition. A dimensionless stress intensity factor (s-SIF) is derived, and a workflow based on FEA is established for the precomputation of the s-SIFs of the basis functions. The s-SIFs for the first eight orders of the CH basis functions are precomputed. Finally, the relationship between the SIF prediction accuracy and the submodel domain size is examined, and a numerical example from NASGRO is used to validate the proposed method.

针对二维倾斜裂纹，提出了一种基于子模型技术、以圆谐波为基函数的边界位移分解方法来计算应力强度因子。该方法通过将有限元分析得到的基函数和子模型边界位移分解系数的预先计算的SIF值线性叠加来计算SIF。该方法避免了在分析过程中在全局模型中生成裂纹网格的计算量大的任务。具体而言，采用CH函数作为子模型边界位移分解的基函数。推导了无量纲应力强度因子，建立了基于有限元分析的无量纲应力强度因子预计算流程。CH基函数的前八阶s- sif是预先计算的。最后，分析了SIF预测精度与子模型域大小之间的关系，并用NASGRO的一个数值算例验证了该方法的有效性。

{"title":"A novel boundary displacement decomposition method for stress intensity factor calculation in 2D inclined crack","authors":"Sanshao Zhuang , Tao Hu , Junfeng Zhang , Wenqing Zheng , Miaolin Feng","doi":"10.1016/j.engfracmech.2026.111859","DOIUrl":"10.1016/j.engfracmech.2026.111859","url":null,"abstract":"<div><div>For a 2D inclined crack, a novel boundary displacement decomposition method based on the submodel technique and using circular harmonic (CH) as basis functions is proposed to calculate the stress intensity factor (SIF). This method computes the SIF by linearly superimposing the precomputed SIF values of the basis functions and the decomposition coefficients of the submodel boundary displacement, which are obtained from finite element analysis (FEA). The approach circumvents the computationally intensive task of generating crack meshes in the global model during analysis. Specifically, CH functions are employed as the basis functions for submodel boundary displacement decomposition. A dimensionless stress intensity factor (s-SIF) is derived, and a workflow based on FEA is established for the precomputation of the s-SIFs of the basis functions. The s-SIFs for the first eight orders of the CH basis functions are precomputed. Finally, the relationship between the SIF prediction accuracy and the submodel domain size is examined, and a numerical example from NASGRO is used to validate the proposed method.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111859"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145973635","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Information mining-assisted fatigue life prediction of aluminum alloys 信息挖掘辅助铝合金疲劳寿命预测

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2025-12-20 DOI: 10.1016/j.engfracmech.2025.111801

Lu Xiao , Jingli Ren

We propose an information mining-assisted machine learning framework to predict the fatigue life of aluminum alloys. This framework aims to improve model’s performance by integrating mined information into modeling process. Specifically, it employs genetic programming-based symbolic regression (SR) to mine underlying information, which describes a relationship between key material parameters (stress amplitude, maximum stress, ultimate tensile strength) and fatigue life. The mined relationship is then integrated into modeling for fatigue life prediction. Experimental datasets of various aluminum alloys, multi-principal element alloys, and steels are utilized to evaluate the proposed framework. The results demonstrate that the SR-assisted models achieve superior accuracy and generalization (R

^{2}

>

0.8) to the black-box models (R

^{2}

<

0.8). Moreover, interpretability analysis revealed that high concentrations of Mg, Zn, Zr, Cr, Mn, and Cu are beneficial to the fatigue strength of aluminum alloys. The proposed framework provides a more general and accurate approach for fatigue life prediction than conventional methods, thereby offering more reliable support for the risk assessment of structural components.

提出了一种基于信息挖掘的机器学习框架来预测铝合金的疲劳寿命。该框架旨在通过将挖掘的信息集成到建模过程中来提高模型的性能。具体来说，它采用基于遗传规划的符号回归（SR）来挖掘潜在信息，这些信息描述了关键材料参数（应力振幅、最大应力、极限抗拉强度）与疲劳寿命之间的关系。然后将挖掘的关系集成到疲劳寿命预测的建模中。利用各种铝合金、多主元素合金和钢的实验数据集来评估所提出的框架。结果表明，sr辅助模型的精度和泛化程度（R2 > 0.8）优于黑盒模型（R2 < 0.8）。此外，可解释性分析表明，高浓度的Mg、Zn、Zr、Cr、Mn和Cu有利于铝合金的疲劳强度。该框架为疲劳寿命预测提供了一种比传统方法更通用、更准确的方法，从而为结构构件的风险评估提供了更可靠的支持。

{"title":"Information mining-assisted fatigue life prediction of aluminum alloys","authors":"Lu Xiao , Jingli Ren","doi":"10.1016/j.engfracmech.2025.111801","DOIUrl":"10.1016/j.engfracmech.2025.111801","url":null,"abstract":"<div><div>We propose an information mining-assisted machine learning framework to predict the fatigue life of aluminum alloys. This framework aims to improve model’s performance by integrating mined information into modeling process. Specifically, it employs genetic programming-based symbolic regression (SR) to mine underlying information, which describes a relationship between key material parameters (stress amplitude, maximum stress, ultimate tensile strength) and fatigue life. The mined relationship is then integrated into modeling for fatigue life prediction. Experimental datasets of various aluminum alloys, multi-principal element alloys, and steels are utilized to evaluate the proposed framework. The results demonstrate that the SR-assisted models achieve superior accuracy and generalization (R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span> <span><math><mo>></mo></math></span> 0.8) to the black-box models (R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span> <span><math><mo><</mo></math></span> 0.8). Moreover, interpretability analysis revealed that high concentrations of Mg, Zn, Zr, Cr, Mn, and Cu are beneficial to the fatigue strength of aluminum alloys. The proposed framework provides a more general and accurate approach for fatigue life prediction than conventional methods, thereby offering more reliable support for the risk assessment of structural components.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111801"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145838695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 regularized phase-field model for dynamic fracture in bi-material structures: Influence of interface and geometric characteristics 双材料结构动态断裂的正则相场模型：界面和几何特性的影响

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2026-01-03 DOI: 10.1016/j.engfracmech.2025.111832

Krishnendu Sivadas , Amol Vuppuluri , Chandu Parimi , Raghu Piska , Hirshikesh

In this work, we employed a phase field fracture model for understanding the crack-interface interactions in the presence of material heterogeneities under dynamic loading. The crack-interface interaction is explored for the system with a sharp interface (zero thickness) as well as a regularized interface (finite thickness) having tan-hyperbolic regularization. The interface is regularized in order to capture the realistic condition across the interface. Within this framework, the length-scale parameters of the bulk material

ℓ_{s}

and the interface

ℓ_{i}

are not treated independently. Instead, they are expressed in terms of a tuning parameter called the diffusivity parameter

k

, as a length-scale diffusivity ratio

ℓ_{s} / k

. This enables optimization of the length-scale, balancing accuracy and computational efficiency while maintaining the physical relevance. Moreover, the existing numerical complexity in capturing interface mechanics, the need for conforming interface, and homogenization methods are no longer required, making the implementation straightforward. Crack propagation, branching/penetration, and crack arrest are readily simulated, highlighting the capability of the model to reproduce the complex dynamic fracture mechanisms. The comparison with sharp interface results confirms the accuracy of predictions. The in-depth fracture studies are carried out by considering different fracture toughness ratios between the constituent materials, varying the nature, location of the interfaces, and the inclination of the interface as well. All these factors have found to play a vital role in governing the dynamic fracture characteristics.

在这项工作中，我们采用相场断裂模型来理解在动态加载下材料非均质存在时裂纹-界面的相互作用。研究了具有尖锐界面（零厚度）和具有棕双曲正则化的正则界面（有限厚度）系统的裂纹-界面相互作用。为了捕获跨接口的真实情况，对接口进行了正则化。在此框架内，块体材料的长度尺度参数和界面的长度尺度参数并不是单独处理的。相反，它们是用一个叫做扩散系数参数k的调节参数来表示的，作为长度尺度的扩散系数比（s/k）。这使得长度尺度的优化，平衡精度和计算效率，同时保持物理相关性。此外，不再需要现有的捕获界面力学的数值复杂性，对一致性接口的需求以及均质化方法，使实现变得简单。该模型可以很容易地模拟裂纹扩展、分支/渗透和裂纹止裂，突出了该模型重现复杂动态断裂机制的能力。与锐界面结果的比较证实了预测的准确性。考虑不同组成材料之间的断裂韧性比，改变界面的性质、界面的位置和界面的倾角，进行了深入的断裂研究。这些因素对动态断裂特性的控制起着至关重要的作用。

{"title":"A regularized phase-field model for dynamic fracture in bi-material structures: Influence of interface and geometric characteristics","authors":"Krishnendu Sivadas , Amol Vuppuluri , Chandu Parimi , Raghu Piska , Hirshikesh","doi":"10.1016/j.engfracmech.2025.111832","DOIUrl":"10.1016/j.engfracmech.2025.111832","url":null,"abstract":"<div><div>In this work, we employed a phase field fracture model for understanding the crack-interface interactions in the presence of material heterogeneities under dynamic loading. The crack-interface interaction is explored for the system with a sharp interface (zero thickness) as well as a regularized interface (finite thickness) having tan-hyperbolic regularization. The interface is regularized in order to capture the realistic condition across the interface. Within this framework, the length-scale parameters of the bulk material <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mi>s</mi></mrow></msub></math></span> and the interface <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mi>i</mi></mrow></msub></math></span> are not treated independently. Instead, they are expressed in terms of a tuning parameter called the diffusivity parameter <span><math><mi>k</mi></math></span>, as a length-scale diffusivity ratio <span><math><mrow><msub><mrow><mi>ℓ</mi></mrow><mrow><mi>s</mi></mrow></msub><mo>/</mo><mi>k</mi></mrow></math></span>. This enables optimization of the length-scale, balancing accuracy and computational efficiency while maintaining the physical relevance. Moreover, the existing numerical complexity in capturing interface mechanics, the need for conforming interface, and homogenization methods are no longer required, making the implementation straightforward. Crack propagation, branching/penetration, and crack arrest are readily simulated, highlighting the capability of the model to reproduce the complex dynamic fracture mechanisms. The comparison with sharp interface results confirms the accuracy of predictions. The in-depth fracture studies are carried out by considering different fracture toughness ratios between the constituent materials, varying the nature, location of the interfaces, and the inclination of the interface as well. All these factors have found to play a vital role in governing the dynamic fracture characteristics.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111832"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145922252","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Finite element modelling of cracking behaviour of reinforced concrete tensile members using a phase field approach 用相场法模拟钢筋混凝土受拉构件的开裂行为

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2026-01-02 DOI: 10.1016/j.engfracmech.2025.111835

Mario D. Barahona, Laura Carreras, Cristina Barris

Modelling the cracking behaviour of reinforced concrete (RC) elements remains a major challenge due to the inherent heterogeneity of concrete and the complex interaction with steel reinforcement. Existing finite element (FE) approaches are restricted to simplified 2D representations, depend on predefined crack paths, or do not incorporate the material heterogeneity of RC in three dimensions. This study presents a 3D FE framework in Abaqus to model the cracking behaviour of RC tie elements, combining a phase field formulation with stochastic random fields (RF) to represent spatial variability in tensile strength and fracture toughness. Parametric studies demonstrate the influence of key modelling parameters, including the phase field length scale, solution scheme, and correlation length of the RF. The numerical results are validated against experimental data from RC tie tests in the literature, and demonstrate good agreement in the global load–displacement response and localised crack patterns. The study shows that the proposed approach is a robust predictive tool able to capture the uncertainty arising from local material heterogeneity, and can simulate diverse crack initiation and propagation scenarios in RC.

由于混凝土固有的非均质性和与钢筋的复杂相互作用，钢筋混凝土（RC）构件的开裂行为建模仍然是一个主要挑战。现有的有限元（FE）方法仅限于简化的二维表示，依赖于预定义的裂纹路径，或者没有在三维中纳入RC的材料非均质性。本研究在Abaqus中提出了一个三维有限元框架来模拟RC连接单元的开裂行为，结合相场公式和随机随机场（RF）来表示拉伸强度和断裂韧性的空间变异性。参数化研究证明了关键建模参数的影响，包括相场长度尺度、求解方案和射频的相关长度。数值结果与文献中的钢筋混凝土拉杆试验数据进行了验证，并证明了整体荷载-位移响应和局部裂缝模式的良好一致性。研究表明，该方法是一种鲁棒的预测工具，能够捕捉由局部材料非均质性引起的不确定性，并能模拟RC中不同的裂纹萌生和扩展场景。

{"title":"Finite element modelling of cracking behaviour of reinforced concrete tensile members using a phase field approach","authors":"Mario D. Barahona, Laura Carreras, Cristina Barris","doi":"10.1016/j.engfracmech.2025.111835","DOIUrl":"10.1016/j.engfracmech.2025.111835","url":null,"abstract":"<div><div>Modelling the cracking behaviour of reinforced concrete (RC) elements remains a major challenge due to the inherent heterogeneity of concrete and the complex interaction with steel reinforcement. Existing finite element (FE) approaches are restricted to simplified 2D representations, depend on predefined crack paths, or do not incorporate the material heterogeneity of RC in three dimensions. This study presents a 3D FE framework in Abaqus to model the cracking behaviour of RC tie elements, combining a phase field formulation with stochastic random fields (RF) to represent spatial variability in tensile strength and fracture toughness. Parametric studies demonstrate the influence of key modelling parameters, including the phase field length scale, solution scheme, and correlation length of the RF. The numerical results are validated against experimental data from RC tie tests in the literature, and demonstrate good agreement in the global load–displacement response and localised crack patterns. The study shows that the proposed approach is a robust predictive tool able to capture the uncertainty arising from local material heterogeneity, and can simulate diverse crack initiation and propagation scenarios in RC.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111835"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145922253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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

Fatigue life prediction of metal rubber under small-sample conditions via data augmentation and physics-informed fusion 基于数据增强和物理信息融合的小样本条件下金属橡胶疲劳寿命预测

IF 5.3 2区工程技术 Q1 MECHANICS

Engineering Fracture Mechanics

Pub Date : 2026-02-21 Epub Date: 2026-01-13 DOI: 10.1016/j.engfracmech.2026.111865

Zihao Huang , Zhiying Ren , Han Lin , Yuedan Li , Chunhui Zhou , Xianbao Xiang , Jinhui Chen

Metal rubber (MR), as a novel elastic porous material, has been widely applied in critical fields such as aerospace, and its fatigue life prediction is of great significance for ensuring equipment safety. To address the dual challenges of data scarcity and complex physical mechanisms in life modeling of MR, this paper proposes a fatigue life prediction method that integrates data augmentation with physics-informed modeling—namely, the GAN-PILM-TCN model. Specifically, a Wasserstein Generative Adversarial Network (W-GAN) is first employed to augment the limited fatigue test data, effectively alleviating the problem of insufficient samples. Then, stiffness and damping damage factors are incorporated into the modeling process to enhance the physical interpretability of the model. In addition, Principal Component Analysis (PCA) is applied to reduce the dimensionality of multi-dimensional features and extract key information on damage evolution. Furthermore, a hybrid loss function is constructed based on the Weibull distribution to guide the model in better capturing the relationship between fatigue damage and life during training. Experimental results demonstrate that the proposed method achieves a significant improvement in prediction performance under small-sample conditions, with a maximum R² of 0.9336. To further evaluate model stability, 10 independent training runs were conducted, yielding an average mean squared error of MSE_avg = 0.0142 and a standard deviation of Std = 0.0011, indicating good stability and reproducibility. These findings provide a new methodological framework and technical support for fatigue life prediction and reliability design of MR under complex operating conditions.

金属橡胶作为一种新型的弹性多孔材料，已广泛应用于航空航天等关键领域，其疲劳寿命预测对保障设备安全具有重要意义。为了解决MR寿命建模中数据稀缺和物理机制复杂的双重挑战，本文提出了一种将数据增强与物理信息建模相结合的疲劳寿命预测方法，即GAN-PILM-TCN模型。具体而言，首先采用Wasserstein生成对抗网络（W-GAN）对有限的疲劳试验数据进行扩充，有效缓解了样本不足的问题。然后，将刚度和阻尼损伤因素纳入建模过程，以增强模型的物理可解释性。此外，应用主成分分析（PCA）对多维特征进行降维，提取损伤演化的关键信息。此外，基于威布尔分布构造了混合损失函数，以指导模型更好地捕捉训练过程中疲劳损伤与寿命的关系。实验结果表明，该方法在小样本条件下的预测性能有显著提高，最大R2为0.9336。为了进一步评价模型的稳定性，我们进行了10次独立训练，平均均方误差为MSEavg = 0.0142，标准差为Std = 0.0011，表明模型具有良好的稳定性和重复性。研究结果为复杂工况下磁流变器的疲劳寿命预测和可靠性设计提供了新的方法框架和技术支持。

{"title":"Fatigue life prediction of metal rubber under small-sample conditions via data augmentation and physics-informed fusion","authors":"Zihao Huang , Zhiying Ren , Han Lin , Yuedan Li , Chunhui Zhou , Xianbao Xiang , Jinhui Chen","doi":"10.1016/j.engfracmech.2026.111865","DOIUrl":"10.1016/j.engfracmech.2026.111865","url":null,"abstract":"<div><div>Metal rubber (MR), as a novel elastic porous material, has been widely applied in critical fields such as aerospace, and its fatigue life prediction is of great significance for ensuring equipment safety. To address the dual challenges of data scarcity and complex physical mechanisms in life modeling of MR, this paper proposes a fatigue life prediction method that integrates data augmentation with physics-informed modeling—namely, the GAN-PILM-TCN model. Specifically, a Wasserstein Generative Adversarial Network (W-GAN) is first employed to augment the limited fatigue test data, effectively alleviating the problem of insufficient samples. Then, stiffness and damping damage factors are incorporated into the modeling process to enhance the physical interpretability of the model. In addition, Principal Component Analysis (PCA) is applied to reduce the dimensionality of multi-dimensional features and extract key information on damage evolution. Furthermore, a hybrid loss function is constructed based on the Weibull distribution to guide the model in better capturing the relationship between fatigue damage and life during training. Experimental results demonstrate that the proposed method achieves a significant improvement in prediction performance under small-sample conditions, with a maximum R<sup>2</sup> of 0.9336. To further evaluate model stability, 10 independent training runs were conducted, yielding an average mean squared error of MSE<sub>avg</sub> = 0.0142 and a standard deviation of Std = 0.0011, indicating good stability and reproducibility. These findings provide a new methodological framework and technical support for fatigue life prediction and reliability design of MR under complex operating conditions.</div></div>","PeriodicalId":11576,"journal":{"name":"Engineering Fracture Mechanics","volume":"333 ","pages":"Article 111865"},"PeriodicalIF":5.3,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145973634","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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