Pub Date : 2026-01-01Epub Date: 2025-11-21DOI: 10.1016/j.advengsoft.2025.104070
Binqi Xiao , Biao Wei , Jun Chen , Ruimin Zhang , Mingyu Chen , Xianglin Zheng , Zhixing Yang
Near-fault earthquakes seriously endanger the structural safety and operational performance of high-speed railway track-bridge systems (HSRT-BS). To address this issue, this study proposes a multi-component multi-level seismic design (MMSD) method and develops a reduced-order model for parameter design. Using a CRTS Ⅲ track-continuous beam bridge as a case study, a finite element model is established based on the OpenSEES engineering seismic software to implement the MMSD and conduct numerical analyses. The seismic responses of key components in the MMSD system and the ordinary system are compared, while operational safety is evaluated using the velocity-related spectral intensity (VSI) index. Results indicate that the MMSD markedly reduces seismic responses of the track, girder, rail, bearings, and piers, showing stable behavior under earthquake, and lowers the VSI index by nearly 50 %, demonstrating its effectiveness and feasibility for HSRT-BS.
{"title":"Damage control of CRTS Ⅲ track-bridge systems using multi-component multi-level seismic design under near-fault earthquakes","authors":"Binqi Xiao , Biao Wei , Jun Chen , Ruimin Zhang , Mingyu Chen , Xianglin Zheng , Zhixing Yang","doi":"10.1016/j.advengsoft.2025.104070","DOIUrl":"10.1016/j.advengsoft.2025.104070","url":null,"abstract":"<div><div>Near-fault earthquakes seriously endanger the structural safety and operational performance of high-speed railway track-bridge systems (HSRT-BS). To address this issue, this study proposes a multi-component multi-level seismic design (MMSD) method and develops a reduced-order model for parameter design. Using a CRTS Ⅲ track-continuous beam bridge as a case study, a finite element model is established based on the OpenSEES engineering seismic software to implement the MMSD and conduct numerical analyses. The seismic responses of key components in the MMSD system and the ordinary system are compared, while operational safety is evaluated using the velocity-related spectral intensity (VSI) index. Results indicate that the MMSD markedly reduces seismic responses of the track, girder, rail, bearings, and piers, showing stable behavior under earthquake, and lowers the VSI index by nearly 50 %, demonstrating its effectiveness and feasibility for HSRT-BS.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104070"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145571700","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}
Pub Date : 2026-01-01Epub Date: 2025-11-08DOI: 10.1016/j.advengsoft.2025.104050
Ahmed Slimen , Rabï Ben Sghaier
Accurate quantification of residual stresses (RS) is essential to maintaining the structural integrity, durability, and performance of engineering components. Conventional approaches—including experimental techniques and process modeling—often suffer from limitations such as sparse data availability, high computational expense, and demanding material characterization requirements. In contrast, the eigenstrain method has emerged as a powerful alternative, enabling efficient RS reconstruction via linear elastic finite element analysis (FEA), while inherently satisfying equilibrium and compatibility conditions with minimal experimental input.
Despite its theoretical appeal, the practical application of eigenstrain-based methods—particularly for large-scale engineering components with complex geometries—has been limited by computational demands, lack of native implementation in commercial FEA platforms, and dependence on third-party software. These constraints fragment workflows, increase susceptibility to errors, and hinder broader adoption, highlighting the need for a unified computational framework.
EigenRec3D addresses this gap by providing a fully integrated platform for reconstructing residual stress fields in arbitrary two- and three-dimensional geometries via the eigenstrain method. Implemented entirely within the ANSYS® APDL environment through advanced scripting, it eliminates external dependencies while ensuring computational robustness. Its modular design and intuitive graphical interface streamline setup, minimize user intervention, and enhance accessibility for both research and industrial applications.
The tool’s capability is validated through case studies involving large-scale, surface-treated components of arbitrary shape, demonstrating accuracy, scalability, and readiness for deployment. EigenRec3D offers a pathway for integration into advanced manufacturing workflows, including additive manufacturing.
{"title":"A unified computational tool for residual stress reconstruction in surface-treated, large-scale components with arbitrary geometries using the eigenstrain method","authors":"Ahmed Slimen , Rabï Ben Sghaier","doi":"10.1016/j.advengsoft.2025.104050","DOIUrl":"10.1016/j.advengsoft.2025.104050","url":null,"abstract":"<div><div>Accurate quantification of residual stresses (RS) is essential to maintaining the structural integrity, durability, and performance of engineering components. Conventional approaches—including experimental techniques and process modeling—often suffer from limitations such as sparse data availability, high computational expense, and demanding material characterization requirements. In contrast, the eigenstrain method has emerged as a powerful alternative, enabling efficient RS reconstruction via linear elastic finite element analysis (FEA), while inherently satisfying equilibrium and compatibility conditions with minimal experimental input.</div><div>Despite its theoretical appeal, the practical application of eigenstrain-based methods—particularly for large-scale engineering components with complex geometries—has been limited by computational demands, lack of native implementation in commercial FEA platforms, and dependence on third-party software. These constraints fragment workflows, increase susceptibility to errors, and hinder broader adoption, highlighting the need for a unified computational framework.</div><div><em>EigenRec3D</em> addresses this gap by providing a fully integrated platform for reconstructing residual stress fields in arbitrary two- and three-dimensional geometries via the eigenstrain method. Implemented entirely within the ANSYS® APDL environment through advanced scripting, it eliminates external dependencies while ensuring computational robustness. Its modular design and intuitive graphical interface streamline setup, minimize user intervention, and enhance accessibility for both research and industrial applications.</div><div>The tool’s capability is validated through case studies involving large-scale, surface-treated components of arbitrary shape, demonstrating accuracy, scalability, and readiness for deployment. EigenRec3D offers a pathway for integration into advanced manufacturing workflows, including additive manufacturing.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104050"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145468231","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}
Pub Date : 2026-01-01Epub Date: 2025-11-13DOI: 10.1016/j.advengsoft.2025.104059
Zhou Huang , Xianjie Shi , Peng Zuo
A dynamic analysis model is developed to investigate the free vibration characteristics of a spherical-conical-cylindrical shell-circular plate coupling structure (SCCCCS). First, within the framework of the first-order shear deformation theory, the structural displacement function for a unified analysis model of revolving plate-shell structures is derived using spectro-geometric method. The artificial virtual spring technique is then applied to equivalently simulate the boundary and coupling conditions. The Ritz method is employed to solve the energy functional, resulting in the dynamic equation governing the SCCCCS analytical model. Numerical verification of the model's reliability and accuracy is performed by comparing its results with those obtained from the finite element method over a wide frequency range. A parameterized study on the dynamic characteristics of the SCCCCS under arbitrary boundary conditions is also conducted, considering various relevant parameters. The results indicate that both the semi-vertex angle of the conical shell and the coupling position of the circular plate significantly influence the structural stiffness of the SCCCCS, thereby affecting the variation of its frequency characteristics.
{"title":"Research on vibrational characteristics of joined spherical- conical-cylindrical shells with multiple annular plates","authors":"Zhou Huang , Xianjie Shi , Peng Zuo","doi":"10.1016/j.advengsoft.2025.104059","DOIUrl":"10.1016/j.advengsoft.2025.104059","url":null,"abstract":"<div><div>A dynamic analysis model is developed to investigate the free vibration characteristics of a spherical-conical-cylindrical shell-circular plate coupling structure (SCCCCS). First, within the framework of the first-order shear deformation theory, the structural displacement function for a unified analysis model of revolving plate-shell structures is derived using spectro-geometric method. The artificial virtual spring technique is then applied to equivalently simulate the boundary and coupling conditions. The Ritz method is employed to solve the energy functional, resulting in the dynamic equation governing the SCCCCS analytical model. Numerical verification of the model's reliability and accuracy is performed by comparing its results with those obtained from the finite element method over a wide frequency range. A parameterized study on the dynamic characteristics of the SCCCCS under arbitrary boundary conditions is also conducted, considering various relevant parameters. The results indicate that both the semi-vertex angle of the conical shell and the coupling position of the circular plate significantly influence the structural stiffness of the SCCCCS, thereby affecting the variation of its frequency characteristics.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104059"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145520353","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}
Pub Date : 2026-01-01Epub Date: 2025-11-04DOI: 10.1016/j.advengsoft.2025.104062
Judy P. Yang, Yu-Hui Kao
A nonlinear collocation method incorporating numerical boundary treatment is developed to solve lid-driven cavity flow problems governed by the Navier-Stokes equations in the stream function-vorticity form. In contrast to approaches that rely solely on the vorticity formulation, the present method avoids the computational challenges associated with evaluating fourth-order derivatives of reproducing kernel shape functions. To address the difficulties posed by non-physical vorticity boundary conditions, a higher-order finite difference-based numerical boundary scheme is introduced, in which Neumann boundary conditions for the stream function are implicitly enforced. The effectiveness and accuracy of the method are validated through a series of benchmark investigations, demonstrating its robustness and capability to handle a wide range of Reynolds numbers in lid-driven cavity flow problems.
{"title":"Numerical boundary treatment in meshfree collocation for lid-driven cavity flow in stream function-vorticity form","authors":"Judy P. Yang, Yu-Hui Kao","doi":"10.1016/j.advengsoft.2025.104062","DOIUrl":"10.1016/j.advengsoft.2025.104062","url":null,"abstract":"<div><div>A nonlinear collocation method incorporating numerical boundary treatment is developed to solve lid-driven cavity flow problems governed by the Navier-Stokes equations in the stream function-vorticity form. In contrast to approaches that rely solely on the vorticity formulation, the present method avoids the computational challenges associated with evaluating fourth-order derivatives of reproducing kernel shape functions. To address the difficulties posed by non-physical vorticity boundary conditions, a higher-order finite difference-based numerical boundary scheme is introduced, in which Neumann boundary conditions for the stream function are implicitly enforced. The effectiveness and accuracy of the method are validated through a series of benchmark investigations, demonstrating its robustness and capability to handle a wide range of Reynolds numbers in lid-driven cavity flow problems.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104062"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145468232","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}
Pub Date : 2026-01-01Epub Date: 2025-10-24DOI: 10.1016/j.advengsoft.2025.104047
Feifei Chen , Xiaoting Rui , Hehua Ju , Kaimeng Wang
Dynamic simulation is critical for the design and control of complex multi-DOF rigid body systems. Existing general-purpose dynamics software often relies on pre-built toolkits or open source libraries, which provide versatile functions but face challenges such as functional constraints and difficulty in error tracking. This study presents a self-developed, modular simulation platform based on an explicit joint space canonical dynamics formulation. The proposed approach extends the explicit dynamics theory by exploiting its block matrix structure: rotational and prismatic joints are classified, and the joint space mass matrix and bias force vector are assembled explicitly at the block level, enabling parallel computation of only half of the symmetric matrix, thus allowing real-time performance for multi-DOF systems. The entire numerical pipeline from topology initialization to forward and inverse dynamics solving is transparent, lightweight, and implemented in C++, ensuring full controllability and solver traceability. The platform demonstrates real-time simulation in millisecond level for a 6-DOF robotic arm, a 25-DOF Mars rover, and a 48-DOF multi-satellite system on standard CPUs, validating its accuracy, stability, and engineering applicability. This work highlights a way from traditional recursive solvers to an explicit dynamics framework for robotic and aerospace systems.
{"title":"A real-time dynamic simulation platform for multi-DOF rigid body systems based on a novel explicit modelling method","authors":"Feifei Chen , Xiaoting Rui , Hehua Ju , Kaimeng Wang","doi":"10.1016/j.advengsoft.2025.104047","DOIUrl":"10.1016/j.advengsoft.2025.104047","url":null,"abstract":"<div><div>Dynamic simulation is critical for the design and control of complex multi-DOF rigid body systems. Existing general-purpose dynamics software often relies on pre-built toolkits or open source libraries, which provide versatile functions but face challenges such as functional constraints and difficulty in error tracking. This study presents a self-developed, modular simulation platform based on an explicit joint space canonical dynamics formulation. The proposed approach extends the explicit dynamics theory by exploiting its block matrix structure: rotational and prismatic joints are classified, and the joint space mass matrix and bias force vector are assembled explicitly at the block level, enabling parallel computation of only half of the symmetric matrix, thus allowing real-time performance for multi-DOF systems. The entire numerical pipeline from topology initialization to forward and inverse dynamics solving is transparent, lightweight, and implemented in C++, ensuring full controllability and solver traceability. The platform demonstrates real-time simulation in millisecond level for a 6-DOF robotic arm, a 25-DOF Mars rover, and a 48-DOF multi-satellite system on standard CPUs, validating its accuracy, stability, and engineering applicability. This work highlights a way from traditional recursive solvers to an explicit dynamics framework for robotic and aerospace systems.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104047"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145365679","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}
Pub Date : 2026-01-01Epub Date: 2025-10-29DOI: 10.1016/j.advengsoft.2025.104060
Peng Zuo , Zhao Du , Xianjie Shi , Huiyong Feng , Zhengyang Gao , Bing Hu
This study proposed a dynamic analysis model to systematically investigate the free vibration and stationary random vibration problems of functionally graded graphene platelet reinforced porous composite (FG-GPLRPC) coupled conical-cylindrical-conical shells with annular plates under stationary random excitation. Based on the closed-cell theory, the effective material properties with respect to FG-GPLRPC are characterized using the Halpin–Tsai micromechanics model combined with rule of mixture. Also, the artificial spring approach is adopted to describe different boundary restraints and coupling conditions that existed in the model. Within the shell theoretical framework of the first-order shear deformation theory (FSDT), the dynamic model for analyzing the free vibration and random vibration responses of the studied FG-GPLRPC coupled structures is established based on the Rayleigh-Ritz method. The spectro-geometric method (SGM) and pseudo-excitation method (PEM) are employed to calculate the vibration response results of the FG-GPLRPC coupled structures. The validity of the presented model is verified through implementing several numerical cases associated with the comparative analysis of free vibration and random vibration response results. Furthermore, some physical mechanisms regarding the influences of porosity coefficient, weight fraction, and the length of cylindrical shell, etc., on the model frequencies and random response behaviors of the FG-GPLRPC coupled structures are revealed.
{"title":"Dynamic modeling and vibration analysis for functionally graded graphene platelet reinforced porous composite coupled conical-cylindrical-conical shells with annular plates","authors":"Peng Zuo , Zhao Du , Xianjie Shi , Huiyong Feng , Zhengyang Gao , Bing Hu","doi":"10.1016/j.advengsoft.2025.104060","DOIUrl":"10.1016/j.advengsoft.2025.104060","url":null,"abstract":"<div><div>This study proposed a dynamic analysis model to systematically investigate the free vibration and stationary random vibration problems of functionally graded graphene platelet reinforced porous composite (FG-GPLRPC) coupled conical-cylindrical-conical shells with annular plates under stationary random excitation. Based on the closed-cell theory, the effective material properties with respect to FG-GPLRPC are characterized using the Halpin–Tsai micromechanics model combined with rule of mixture. Also, the artificial spring approach is adopted to describe different boundary restraints and coupling conditions that existed in the model. Within the shell theoretical framework of the first-order shear deformation theory (FSDT), the dynamic model for analyzing the free vibration and random vibration responses of the studied FG-GPLRPC coupled structures is established based on the Rayleigh-Ritz method. The spectro-geometric method (SGM) and pseudo-excitation method (PEM) are employed to calculate the vibration response results of the FG-GPLRPC coupled structures. The validity of the presented model is verified through implementing several numerical cases associated with the comparative analysis of free vibration and random vibration response results. Furthermore, some physical mechanisms regarding the influences of porosity coefficient, weight fraction, and the length of cylindrical shell, etc., on the model frequencies and random response behaviors of the FG-GPLRPC coupled structures are revealed.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104060"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145419534","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}
Pub Date : 2026-01-01Epub Date: 2025-11-01DOI: 10.1016/j.advengsoft.2025.104049
A. Idesman, W. Ajwad, M. Mobin
This study expands the optimal local truncation error method (OLTEM) with unfitted Cartesian meshes, designed for 2-D elastodynamics with interfaces (wave propagation and structural dynamics), to the general 3-D case and non-homogeneous interface conditions. The technique employs compact 27-point stencils, similar to those used in linear finite elements, while avoiding the introduction of additional unknowns at material interfaces. Importantly, the global semi-discrete equations retain a similar structure for homogeneous and heterogeneous materials. OLTEM with the diagonal mass matrix, suitable for explicit time-integration schemes, represents a subset of the broader formulation using the non-diagonal mass matrix.
A significant innovation in this work is a new 3-D post-processing procedure for stress calculations. It improves accuracy by incorporating accelerations and the governing elastodynamics equations into the analysis. Like the primary computations, this post-processing technique utilizes compact 27-point stencils. The new post-processing procedure outperforms traditional methods that depend solely on displacements.
OLTEM with unfitted Cartesian meshes shows superior accuracy compared to linear finite elements with equivalent stencils and conformal meshes, while requiring significantly fewer degrees of freedom (DOF). For instance, at an accuracy of 0.1% for the displacements, OLTEM with the non-diagonal mass matrix reduces the number of DOF by more than times; at an accuracy of 0.1% for the stresses, OLTEM with the new post-processing procedure reduces the number of DOF about times compared to linear finite elements. OLTEM also provides increased computational efficiency compared to high-order finite elements, despite their wider stencils and conformal meshes.
{"title":"Optimal local truncation error method for 3-D elastodynamics interface problems on unfitted Cartesian meshes","authors":"A. Idesman, W. Ajwad, M. Mobin","doi":"10.1016/j.advengsoft.2025.104049","DOIUrl":"10.1016/j.advengsoft.2025.104049","url":null,"abstract":"<div><div>This study expands the optimal local truncation error method (OLTEM) with unfitted Cartesian meshes, designed for 2-D elastodynamics with interfaces (wave propagation and structural dynamics), to the general 3-D case and non-homogeneous interface conditions. The technique employs compact 27-point stencils, similar to those used in linear finite elements, while avoiding the introduction of additional unknowns at material interfaces. Importantly, the global semi-discrete equations retain a similar structure for homogeneous and heterogeneous materials. OLTEM with the diagonal mass matrix, suitable for explicit time-integration schemes, represents a subset of the broader formulation using the non-diagonal mass matrix.</div><div>A significant innovation in this work is a new 3-D post-processing procedure for stress calculations. It improves accuracy by incorporating accelerations and the governing elastodynamics equations into the analysis. Like the primary computations, this post-processing technique utilizes compact 27-point stencils. The new post-processing procedure outperforms traditional methods that depend solely on displacements.</div><div>OLTEM with unfitted Cartesian meshes shows superior accuracy compared to linear finite elements with equivalent stencils and conformal meshes, while requiring significantly fewer degrees of freedom (DOF). For instance, at an accuracy of 0.1% for the displacements, OLTEM with the non-diagonal mass matrix reduces the number of DOF by more than <span><math><mrow><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mn>4</mn></mrow></msup></mrow></math></span> times; at an accuracy of 0.1% for the stresses, OLTEM with the new post-processing procedure reduces the number of DOF about <span><math><mrow><mn>2</mn><mo>×</mo><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mn>6</mn></mrow></msup></mrow></math></span> times compared to linear finite elements. OLTEM also provides increased computational efficiency compared to high-order finite elements, despite their wider stencils and conformal meshes.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104049"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145419536","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}
Pub Date : 2026-01-01Epub Date: 2025-11-08DOI: 10.1016/j.advengsoft.2025.104064
Geonwoo Lee , Mingyu Lee , Ikjin Lee
Recently, the rapid advancement of deep learning technology has led to the development of numerous topology optimization approaches, significantly reducing computational costs. However, conventional deep learning-based methods inherently suffer from a chronic limitation. They require large-scale data to extract features from the data itself. In particular, in the worst-case scenario where the data is insufficient, these methods may fail to capture the physical characteristics of the structure accurately, potentially leading to physically meaningless and unrealistic results. To solve this problem, this paper proposes an enhanced deep learning model suitable for topology optimization. The main novelty of this study is embedding the feature of topology optimization into a deep learning model. To effectively embed the topology domain, the proposed method introduces three key strategies. Firstly, topology convolutional neural network (CNN) filter layers are incorporated into the neural network model. A CNN is a specialized deep learning architecture designed for grid-structured data such as images, and the topology CNN filter layers are specifically designed to enhance structural connectivity by considering the influence of neighboring elements. Secondly, the pixel-based loss function is augmented with physics-informed loss functions that encapsulate the physical knowledge of topology optimization. Thirdly, a modified output layer is added to prevent zero values in the structure, thereby enhancing numerical stability. Numerical experiments demonstrate that the proposed deep learning approach successfully overcomes the limitations of conventional deep learning methods in data-scarce environments. Furthermore, the results confirm that the proposed method produces designs comparable to the traditional SIMP method.
{"title":"Domain-embedded deep learning frameworks for topology optimization: Enhancing structural performance under data scarce environments","authors":"Geonwoo Lee , Mingyu Lee , Ikjin Lee","doi":"10.1016/j.advengsoft.2025.104064","DOIUrl":"10.1016/j.advengsoft.2025.104064","url":null,"abstract":"<div><div>Recently, the rapid advancement of deep learning technology has led to the development of numerous topology optimization approaches, significantly reducing computational costs. However, conventional deep learning-based methods inherently suffer from a chronic limitation. They require large-scale data to extract features from the data itself. In particular, in the worst-case scenario where the data is insufficient, these methods may fail to capture the physical characteristics of the structure accurately, potentially leading to physically meaningless and unrealistic results. To solve this problem, this paper proposes an enhanced deep learning model suitable for topology optimization. The main novelty of this study is embedding the feature of topology optimization into a deep learning model. To effectively embed the topology domain, the proposed method introduces three key strategies. Firstly, topology convolutional neural network (CNN) filter layers are incorporated into the neural network model. A CNN is a specialized deep learning architecture designed for grid-structured data such as images, and the topology CNN filter layers are specifically designed to enhance structural connectivity by considering the influence of neighboring elements. Secondly, the pixel-based loss function is augmented with physics-informed loss functions that encapsulate the physical knowledge of topology optimization. Thirdly, a modified output layer is added to prevent zero values in the structure, thereby enhancing numerical stability. Numerical experiments demonstrate that the proposed deep learning approach successfully overcomes the limitations of conventional deep learning methods in data-scarce environments. Furthermore, the results confirm that the proposed method produces designs comparable to the traditional SIMP method.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104064"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145520354","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}
Pub Date : 2026-01-01Epub Date: 2025-11-25DOI: 10.1016/j.advengsoft.2025.104071
Ondřej Ježek , Ján Kopačka , Martin Isoz , Dušan Gabriel , Pavel Maršálek , Martin Šotola , Radim Halama
This paper presents a novel post-processing methodology for extracting high-quality geometries from density-based topology optimization results. Current post-processing approaches often struggle to simultaneously achieve smooth boundaries, preserve volume fraction, and maintain topological features. We propose a robust method based on a signed distance function (SDF) that addresses these challenges through a two-stage process: first, an SDF representation of density isocontours is constructed, which is followed by geometry refinement using radial basis functions (RBFs). The method generates smooth boundary representations that appear to originate from much finer discretization, while maintaining the computational efficiency of coarse mesh optimization. Our approach can reduce maximum equivalent stress values compared to conventional methods. This reduction is achieved through continuous geometric transitions at boundaries. The resulting implicit boundary representation facilitates seamless export to standard manufacturing formats without intermediate reconstruction steps, providing a robust foundation for practical engineering applications where high-quality geometric representations are essential.
{"title":"Smooth geometry extraction from SIMP topology optimization: Signed distance function approach with volume preservation","authors":"Ondřej Ježek , Ján Kopačka , Martin Isoz , Dušan Gabriel , Pavel Maršálek , Martin Šotola , Radim Halama","doi":"10.1016/j.advengsoft.2025.104071","DOIUrl":"10.1016/j.advengsoft.2025.104071","url":null,"abstract":"<div><div>This paper presents a novel post-processing methodology for extracting high-quality geometries from density-based topology optimization results. Current post-processing approaches often struggle to simultaneously achieve smooth boundaries, preserve volume fraction, and maintain topological features. We propose a robust method based on a signed distance function (SDF) that addresses these challenges through a two-stage process: first, an SDF representation of density isocontours is constructed, which is followed by geometry refinement using radial basis functions (RBFs). The method generates smooth boundary representations that appear to originate from much finer discretization, while maintaining the computational efficiency of coarse mesh optimization. Our approach can reduce maximum equivalent stress values compared to conventional methods. This reduction is achieved through continuous geometric transitions at boundaries. The resulting implicit boundary representation facilitates seamless export to standard manufacturing formats without intermediate reconstruction steps, providing a robust foundation for practical engineering applications where high-quality geometric representations are essential.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104071"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145618140","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}
Pub Date : 2026-01-01Epub Date: 2025-11-06DOI: 10.1016/j.advengsoft.2025.104061
Ran Yuan , Xi-Long Huang , Yong Fang , Kiyonobu Kasama , Yin Cheng , Yi He
This paper develops a Physics-Informed Neural Network (PINN) framework for solving Mohr-Coulomb (M-C) plasticity in geomechanics, and the plane strain layered perforated soils subjected to surface compression pressure are employed to validate the PINN solutions, through comparisons with parallel numerical experiments conducted in OptumG2. To incorporate the physical information for the elasto-plastic problem into neural networks (NNs), two modified multi-objective loss functions, respectively known as the collocation loss function and the Least Squares Weighted Residual (LSWR) loss function, are constructed through coarse data-driven information and physical constrains, consisting of M-C constitutive relations, associated/non-associated flow rules, Karush-Kuhn-Tucker (KKT) conditions, equilibrium conditions, and boundary conditions. The total loss function incorporates terms obtained from Finite Element Method (FEM) solutions for a range of elastoplastic field variables, i.e., stress and displacement, to inform the physical knowledge fitting. By employing several independently operating and densely connected artificial neural networks (ANNs), the PINN framework achieves the M-C plastic solutions by minimizing the designed total loss functions. Furthermore, influences of sample size, sampling strategy, and the loss function, on performances of the proposed PINN framework, are investigated for parametric analysis. In all cases, the PINN predictions were compared with finite element solutions at 145,023 mesh points, showing that over 90% of points had relative errors within 10%. The proposed PINN model is effective for data-scarce geotechnical problems, though its performance in regions with significant rates of change in physical quantity still requires further improvement.
{"title":"A physics-informed machine learning computational framework for solving Mohr-Coulomb plasticity in geomechanics","authors":"Ran Yuan , Xi-Long Huang , Yong Fang , Kiyonobu Kasama , Yin Cheng , Yi He","doi":"10.1016/j.advengsoft.2025.104061","DOIUrl":"10.1016/j.advengsoft.2025.104061","url":null,"abstract":"<div><div>This paper develops a Physics-Informed Neural Network (PINN) framework for solving Mohr-Coulomb (M-C) plasticity in geomechanics, and the plane strain layered perforated soils subjected to surface compression pressure are employed to validate the PINN solutions, through comparisons with parallel numerical experiments conducted in OptumG2. To incorporate the physical information for the elasto-plastic problem into neural networks (NNs), two modified multi-objective loss functions, respectively known as the collocation loss function and the Least Squares Weighted Residual (LSWR) loss function, are constructed through coarse data-driven information and physical constrains, consisting of M-C constitutive relations, associated/non-associated flow rules, Karush-Kuhn-Tucker (KKT) conditions, equilibrium conditions, and boundary conditions. The total loss function incorporates terms obtained from Finite Element Method (FEM) solutions for a range of elastoplastic field variables, i.e., stress and displacement, to inform the physical knowledge fitting. By employing several independently operating and densely connected artificial neural networks (ANNs), the PINN framework achieves the M-C plastic solutions by minimizing the designed total loss functions. Furthermore, influences of sample size, sampling strategy, and the loss function, on performances of the proposed PINN framework, are investigated for parametric analysis. In all cases, the PINN predictions were compared with finite element solutions at 145,023 mesh points, showing that over 90% of points had relative errors within 10%. The proposed PINN model is effective for data-scarce geotechnical problems, though its performance in regions with significant rates of change in physical quantity still requires further improvement.</div></div>","PeriodicalId":50866,"journal":{"name":"Advances in Engineering Software","volume":"212 ","pages":"Article 104061"},"PeriodicalIF":5.7,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145468230","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号