Pub Date : 2025-12-20DOI: 10.1016/j.enganabound.2025.106608
Yang Li , Detao Wan , Rongdong Wang , Zhonghua Wang , Dean Hu , Chao Jiang
Accurate and rapid prediction of thermal–hydraulic behavior and multi-physical field distribution is critical for the safety and efficiency of sodium-cooled fast reactors (SFR). This work presents a nonlinear Gaussian kernel-based reduced-order modeling (ROM) framework, which combines kernel eigen-decomposition with a deep neural network (DNN) to map varying boundary conditions to reduced-order coefficients, enabling reliable and efficient reconstruction of high-dimensional CFD fields while capturing nonlinear flow structures. The proposed framework effectively leverages the physical interpretability of kernel methods, overcoming limitations of purely black-box models such as autoencoders. The framework is applied to CFD snapshots of wire-wrapped fuel assemblies and printed circuit heat exchangers (PCHE) in SFR, demonstrating its capability to capture complex nonlinear flow and heat transfer phenomena. For the wire-wrapped fuel assembly case, 95 modes retain over 99.5% of the flow energy, with maximum normalized absolute error (NAE) below 0.3 for temperature and velocity fields. For the PCHE case, the ROM accurately reconstructs temperature, axial velocity, and pressure fields with NAE below 0.15 across 200 sampled operating conditions. The proposed framework enables efficient, high-fidelity predictions of nonlinear thermal-hydraulic fields, providing a practical tool for design optimization, uncertainty quantification, and real-time monitoring in SFR systems.
{"title":"Nonlinear Gaussian kernel-based ROM integrated with DNN for thermal-hydraulic prediction in sodium-cooled fast reactors","authors":"Yang Li , Detao Wan , Rongdong Wang , Zhonghua Wang , Dean Hu , Chao Jiang","doi":"10.1016/j.enganabound.2025.106608","DOIUrl":"10.1016/j.enganabound.2025.106608","url":null,"abstract":"<div><div>Accurate and rapid prediction of thermal–hydraulic behavior and multi-physical field distribution is critical for the safety and efficiency of sodium-cooled fast reactors (SFR). This work presents a nonlinear Gaussian kernel-based reduced-order modeling (ROM) framework, which combines kernel eigen-decomposition with a deep neural network (DNN) to map varying boundary conditions to reduced-order coefficients, enabling reliable and efficient reconstruction of high-dimensional CFD fields while capturing nonlinear flow structures. The proposed framework effectively leverages the physical interpretability of kernel methods, overcoming limitations of purely black-box models such as autoencoders. The framework is applied to CFD snapshots of wire-wrapped fuel assemblies and printed circuit heat exchangers (PCHE) in SFR, demonstrating its capability to capture complex nonlinear flow and heat transfer phenomena. For the wire-wrapped fuel assembly case, 95 modes retain over 99.5% of the flow energy, with maximum normalized absolute error (NAE) below 0.3 for temperature and velocity fields. For the PCHE case, the ROM accurately reconstructs temperature, axial velocity, and pressure fields with NAE below 0.15 across 200 sampled operating conditions. The proposed framework enables efficient, high-fidelity predictions of nonlinear thermal-hydraulic fields, providing a practical tool for design optimization, uncertainty quantification, and real-time monitoring in SFR systems.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106608"},"PeriodicalIF":4.1,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145791016","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 : 2025-12-19DOI: 10.1016/j.enganabound.2025.106606
Ye Ouyang , Wei Jiang , Wei Du , Jintao Zhang , Yong Chen , Hong Zheng
The deep energy method provides a variational framework for solving mechanical problems governed by higher-order partial differential equations. In this framework, the imposition of boundary conditions is a crucial step. Boundary conditions are usually enforced using the classical penalty function method and its variants. However, the penalty factor directly affects the computational accuracy and efficiency. If the boundary conditions are imposed using the Lagrangian multiplier method, the resulting solution may not coincide with the optimal solution of the original constrained problem, especially when the Lagrangian is not strictly convex in the primal variables. To overcome this limitation, the generalized multiplier method is used to impose the boundary conditions. The neural-network loss function is constructed within the generalized multiplier framework. Finally, gradient-based optimization is employed to update the network parameters until the loss satisfies a prescribed tolerance. Numerical results show that the proposed method achieves higher accuracy and better solution efficiency than neural networks based on the classical penalty method, the L1 exact penalty method, and the Lagrange multiplier method. The training process is also more stable.
{"title":"A deep energy method for solid mechanics based on a generalized multiplier approach","authors":"Ye Ouyang , Wei Jiang , Wei Du , Jintao Zhang , Yong Chen , Hong Zheng","doi":"10.1016/j.enganabound.2025.106606","DOIUrl":"10.1016/j.enganabound.2025.106606","url":null,"abstract":"<div><div>The deep energy method provides a variational framework for solving mechanical problems governed by higher-order partial differential equations. In this framework, the imposition of boundary conditions is a crucial step. Boundary conditions are usually enforced using the classical penalty function method and its variants. However, the penalty factor directly affects the computational accuracy and efficiency. If the boundary conditions are imposed using the Lagrangian multiplier method, the resulting solution may not coincide with the optimal solution of the original constrained problem, especially when the Lagrangian is not strictly convex in the primal variables. To overcome this limitation, the generalized multiplier method is used to impose the boundary conditions. The neural-network loss function is constructed within the generalized multiplier framework. Finally, gradient-based optimization is employed to update the network parameters until the loss satisfies a prescribed tolerance. Numerical results show that the proposed method achieves higher accuracy and better solution efficiency than neural networks based on the classical penalty method, the <em>L</em><sub>1</sub> exact penalty method, and the Lagrange multiplier method. The training process is also more stable.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106606"},"PeriodicalIF":4.1,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784765","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 : 2025-12-19DOI: 10.1016/j.enganabound.2025.106604
Qinghua Li , Shenshen Chen , Xing Wei , Yan Gu
A novel hybrid approach for transient heat conduction analysis is proposed in this paper, which couples a fully smoothed finite element formulation with a spectral integration technique to enhance computational efficiency and accuracy. Starting from an initial triangular mesh, a smoothing domain is constructed for each edge by connecting its two vertices to the centroids of adjacent triangular elements. Unlike the conventional gradient smoothing technique, which is limited to domain integrals involving shape function derivatives, the quasi-weak form of the smoothed integral handles domain integrals of the shape functions themselves. This transformation converts all domain integrals in the heat conduction and heat capacity matrices into boundary integrals over the smoothing domains, eliminating the need for coordinate mapping and Jacobian matrix calculations. The semi-discrete heat conduction equation is solved using a spectral integration technique, which achieves arbitrary orders of accuracy while significantly improving computational efficiency and stability. Numerical examples demonstrate the capability and accuracy of the proposed method in solving transient heat conduction problems.
{"title":"A time spectral fully smoothed finite element method for transient heat conduction analysis","authors":"Qinghua Li , Shenshen Chen , Xing Wei , Yan Gu","doi":"10.1016/j.enganabound.2025.106604","DOIUrl":"10.1016/j.enganabound.2025.106604","url":null,"abstract":"<div><div>A novel hybrid approach for transient heat conduction analysis is proposed in this paper, which couples a fully smoothed finite element formulation with a spectral integration technique to enhance computational efficiency and accuracy. Starting from an initial triangular mesh, a smoothing domain is constructed for each edge by connecting its two vertices to the centroids of adjacent triangular elements. Unlike the conventional gradient smoothing technique, which is limited to domain integrals involving shape function derivatives, the quasi-weak form of the smoothed integral handles domain integrals of the shape functions themselves. This transformation converts all domain integrals in the heat conduction and heat capacity matrices into boundary integrals over the smoothing domains, eliminating the need for coordinate mapping and Jacobian matrix calculations. The semi-discrete heat conduction equation is solved using a spectral integration technique, which achieves arbitrary orders of accuracy while significantly improving computational efficiency and stability. Numerical examples demonstrate the capability and accuracy of the proposed method in solving transient heat conduction problems.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106604"},"PeriodicalIF":4.1,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784766","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 : 2025-12-19DOI: 10.1016/j.enganabound.2025.106610
Farui Shi , Minghui Li , Nicholas Fantuzzi , Bozhi Deng , Delei Shang , Jun Lu , Heping Xie
The microstructural characteristics (e.g., joints and interfaces) and their scale effects can be crucial determinants of mechanical behavior in microstructured composites such as rocks, advanced materials, and construction structures. In recent years, the physics-informed neural network (PINN) has undergone rapid development for solving problems in computational solid mechanics. However, the application of PINN to modeling multi-scale mechanical behavior in microstructured composites remains largely unexplored. One reason is probably that the existence of microstructure in materials is inherently ignored in the classical Cauchy continuum that has been extensively adopted as the foundational continuum theory in previous PINN studies for computational solid mechanics. In the current work, physical laws and equations of a non-local model, i.e., Cosserat (or micropolar) continuum, are employed to design the loss function of a fully connected artificial neural network, establishing a PINN architecture capable of capturing mechanical behavior in three hexagon-structured composites (termed regular, hourglass, and skew) with distinct microstructural length scales. The results show that the PINN method can successfully model the mechanical behavior of the microstructured composites. A quantitative comparison with finite element method (FEM) solutions reveals excellent agreement, with relative errors in the predicted displacement fields maintained within the range of , thereby validating the accuracy and reliability of the PINN for mechanical analysis. Further, the results also demonstrate the capability of the PINN for simulating the multiscale mechanical behavior of microstructured composites by considering the Cosserat continuum. As the microstructure’s scale increases, the Cosserat mechanical responses of composites show varying characteristics and more significant deviation from the results of the Cauchy continuum. This study demonstrates a potential application of PINN in the context of computational multiscale mechanics by the Cosserat continuum, providing an essential framework for accurately capturing the realistic mechanical behavior of microstructured materials.
{"title":"Physics-informed neural network for multiscale mechanical behavior of microstructured composite materials as Cosserat continuum","authors":"Farui Shi , Minghui Li , Nicholas Fantuzzi , Bozhi Deng , Delei Shang , Jun Lu , Heping Xie","doi":"10.1016/j.enganabound.2025.106610","DOIUrl":"10.1016/j.enganabound.2025.106610","url":null,"abstract":"<div><div>The microstructural characteristics (e.g., joints and interfaces) and their scale effects can be crucial determinants of mechanical behavior in microstructured composites such as rocks, advanced materials, and construction structures. In recent years, the physics-informed neural network (PINN) has undergone rapid development for solving problems in computational solid mechanics. However, the application of PINN to modeling multi-scale mechanical behavior in microstructured composites remains largely unexplored. One reason is probably that the existence of microstructure in materials is inherently ignored in the classical Cauchy continuum that has been extensively adopted as the foundational continuum theory in previous PINN studies for computational solid mechanics. In the current work, physical laws and equations of a non-local model, i.e., Cosserat (or micropolar) continuum, are employed to design the loss function of a fully connected artificial neural network, establishing a PINN architecture capable of capturing mechanical behavior in three hexagon-structured composites (termed regular, hourglass, and skew) with distinct microstructural length scales. The results show that the PINN method can successfully model the mechanical behavior of the microstructured composites. A quantitative comparison with finite element method (FEM) solutions reveals excellent agreement, with relative errors in the predicted displacement fields maintained within the range of <span><math><mrow><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>4</mn></mrow></msup><mo>∼</mo><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>8</mn></mrow></msup></mrow></math></span>, thereby validating the accuracy and reliability of the PINN for mechanical analysis. Further, the results also demonstrate the capability of the PINN for simulating the multiscale mechanical behavior of microstructured composites by considering the Cosserat continuum. As the microstructure’s scale increases, the Cosserat mechanical responses of composites show varying characteristics and more significant deviation from the results of the Cauchy continuum. This study demonstrates a potential application of PINN in the context of computational multiscale mechanics by the Cosserat continuum, providing an essential framework for accurately capturing the realistic mechanical behavior of microstructured materials.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106610"},"PeriodicalIF":4.1,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784767","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 : 2025-12-19DOI: 10.1016/j.enganabound.2025.106589
Yida Mao , Jingxu Hao , Tao Zhang , Zhenyu Chen , Fulong Shi
This paper presents a novel numerical approach based on the inverse differential quadrature method (iDQM) for analyzing free vibrations of functionally graded material (FGM) cylindrical shells. Utilizing Flügge classical shell theory, the mathematical similarities between FGM and uniform material cylindrical shells are examined. Free oscillation characteristics are further investigated through a thickness-based homogenization procedure. The iDQM is applied to discretize the governing ordinary differential equations into an algebraic system, yielding a generalized eigenvalue problem for linear vibrations. A new restriction technique employing nullspace decomposition is additionally developed to address internal continuity and boundary conditions. The convergence and accuracy of the new method are validated against benchmark results from the literature. Finally, the segmented strategy for FGM cylindrical shells is formulated by varying material parameters and interpolation point numbers.
{"title":"The inverse differential quadrature method for free vibration analysis of segmented functionally graded cylindrical shells","authors":"Yida Mao , Jingxu Hao , Tao Zhang , Zhenyu Chen , Fulong Shi","doi":"10.1016/j.enganabound.2025.106589","DOIUrl":"10.1016/j.enganabound.2025.106589","url":null,"abstract":"<div><div>This paper presents a novel numerical approach based on the inverse differential quadrature method (iDQM) for analyzing free vibrations of functionally graded material (FGM) cylindrical shells. Utilizing Flügge classical shell theory, the mathematical similarities between FGM and uniform material cylindrical shells are examined. Free oscillation characteristics are further investigated through a thickness-based homogenization procedure. The iDQM is applied to discretize the governing ordinary differential equations into an algebraic system, yielding a generalized eigenvalue problem for linear vibrations. A new restriction technique employing nullspace decomposition is additionally developed to address internal continuity and boundary conditions. The convergence and accuracy of the new method are validated against benchmark results from the literature. Finally, the segmented strategy for FGM cylindrical shells is formulated by varying material parameters and interpolation point numbers.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106589"},"PeriodicalIF":4.1,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784764","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 : 2025-12-19DOI: 10.1016/j.enganabound.2025.106603
Pelin Senel , Daniel Lesnic , Andreas Karageorghis
We consider a geometric inverse problem which requires detecting an unknown obstacle, e.g. a submarine or an aquatic mine, submerged in a Stokes slow viscous stationary flow of an incompressible fluid. In two-dimensions, the problem is formulated in terms of the biharmonic streamfunction in an unbounded domain which is approximated using the Trefftz method and the method of fundamental solutions (MFS). This is, apparently, the first time the Trefftz method and the MFS are applied for the solution of the biharmonic equation in an unbounded domain. We first examine direct problems and then consider inverse problems. The unknown obstacle is determined by employing a nonlinear Tikhonov regularization procedure. Numerical results are presented and discussed.
{"title":"Meshless methods for the detection of an obstacle submerged in a two-dimensional Stokes flow","authors":"Pelin Senel , Daniel Lesnic , Andreas Karageorghis","doi":"10.1016/j.enganabound.2025.106603","DOIUrl":"10.1016/j.enganabound.2025.106603","url":null,"abstract":"<div><div>We consider a geometric inverse problem which requires detecting an unknown obstacle, e.g. a submarine or an aquatic mine, submerged in a Stokes slow viscous stationary flow of an incompressible fluid. In two-dimensions, the problem is formulated in terms of the biharmonic streamfunction in an unbounded domain which is approximated using the Trefftz method and the method of fundamental solutions (MFS). This is, apparently, the first time the Trefftz method and the MFS are applied for the solution of the biharmonic equation in an unbounded domain. We first examine direct problems and then consider inverse problems. The unknown obstacle is determined by employing a nonlinear Tikhonov regularization procedure. Numerical results are presented and discussed.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106603"},"PeriodicalIF":4.1,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784768","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 : 2025-12-16DOI: 10.1016/j.enganabound.2025.106601
Chunquan Li, Runxin Cao, Yuanhao Zheng, Hongyan Huang, Ming Zhang
The microchannel heat sink with gradient array of ribs and pin fins (MCHS-GDRPF) integrates two types of thermal microstructures, rib and pin fin, which effectively enhances the thermal performance of the microchannel and improves the temperature uniformity. To enhance the thermal performance of the MCHS-GDRPF, machine learning methods are employed to predict and optimize performance based on structural design parameters. Four machine learning methods were evaluated to construct the thermal performance prediction model. The Support Vector Regression (SVR) model, demonstrating the highest accuracy, was then integrated with the fast Non-dominated Sorting Genetic Algorithm-II (NSGA-II) to perform multi-objective optimization of thermal resistance, temperature inhomogeneity, and friction coefficient. Subsequently, the Pareto front solution set was derived, and the optimal structural parameters were determined via the TOPSIS method. Finally, a thermal performance analysis was conducted for MCHS-GDRPF with four distinct structural parameter configurations. Comparative evaluation at Reynolds number 622 indicates three key advantages of the MCHS-GDRPF: (1) 28 K lower maximum surface temperature, (2) 72% improvement in temperature uniformity, and (3) 1.397 PEC value, surpassing conventional microchannel performance.
{"title":"A machine learning-driven multi-objective optimization framework for advanced microchannel heat sinks with gradient rib-pin fin arrays","authors":"Chunquan Li, Runxin Cao, Yuanhao Zheng, Hongyan Huang, Ming Zhang","doi":"10.1016/j.enganabound.2025.106601","DOIUrl":"10.1016/j.enganabound.2025.106601","url":null,"abstract":"<div><div>The microchannel heat sink with gradient array of ribs and pin fins (MCHS-GDRPF) integrates two types of thermal microstructures, rib and pin fin, which effectively enhances the thermal performance of the microchannel and improves the temperature uniformity. To enhance the thermal performance of the MCHS-GDRPF, machine learning methods are employed to predict and optimize performance based on structural design parameters. Four machine learning methods were evaluated to construct the thermal performance prediction model. The Support Vector Regression (SVR) model, demonstrating the highest accuracy, was then integrated with the fast Non-dominated Sorting Genetic Algorithm-II (NSGA-II) to perform multi-objective optimization of thermal resistance, temperature inhomogeneity, and friction coefficient. Subsequently, the Pareto front solution set was derived, and the optimal structural parameters were determined via the TOPSIS method. Finally, a thermal performance analysis was conducted for MCHS-GDRPF with four distinct structural parameter configurations. Comparative evaluation at Reynolds number 622 indicates three key advantages of the MCHS-GDRPF: (1) 28 K lower maximum surface temperature, (2) 72% improvement in temperature uniformity, and (3) 1.397 PEC value, surpassing conventional microchannel performance.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106601"},"PeriodicalIF":4.1,"publicationDate":"2025-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145784769","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 : 2025-12-13DOI: 10.1016/j.enganabound.2025.106599
Kai Chen , Haitao Guo , Xu Luo , Degao Zou , Shanlin Tian
Addressing the challenge of predicting deformation in concrete-faced rockfill dams (CFRDs) due to the nonlinear relationship between dam deformation and material parameters, this paper proposes a rapid deformation field prediction method integrating Proper Orthogonal Decomposition (POD) and Multi-Layer Perceptron (MLP). The Latin Hypercube Sampling (LHS) method is used to sample the parameter space of the Duncan-Chang E-B model. A finite element simulation dataset is created using the SBFEM-FEM coupled efficient analysis algorithm, assembling a deformation field snapshot matrix. The POD algorithm reduces the high-dimensional deformation field, extracting dominant modes and calculating corresponding modal coefficients. A regression model integrates material parameters and modal coefficients via MLP theory, enabling rapid prediction of the global displacement field with millisecond-level precision. The method is validated through cantilever beam bending, single-zone, and multi-zone dam body deformation analyses. The findings suggest that the proposed method can achieve high-precision reconstruction of the deformation field with fewer modes, offering advantages such as low prediction error, reduced computation time, strong generalization capability, and good engineering applicability. This method provides an efficient and reliable research tool for response analysis and prediction of geotechnical structures such as CFRDs, demonstrating promising application prospects and promotional value.
{"title":"A rapid prediction approach for the global deformation field of concrete-faced rockfill dams based on POD–MLP","authors":"Kai Chen , Haitao Guo , Xu Luo , Degao Zou , Shanlin Tian","doi":"10.1016/j.enganabound.2025.106599","DOIUrl":"10.1016/j.enganabound.2025.106599","url":null,"abstract":"<div><div>Addressing the challenge of predicting deformation in concrete-faced rockfill dams (CFRDs) due to the nonlinear relationship between dam deformation and material parameters, this paper proposes a rapid deformation field prediction method integrating Proper Orthogonal Decomposition (POD) and Multi-Layer Perceptron (MLP). The Latin Hypercube Sampling (LHS) method is used to sample the parameter space of the Duncan-Chang E-B model. A finite element simulation dataset is created using the SBFEM-FEM coupled efficient analysis algorithm, assembling a deformation field snapshot matrix. The POD algorithm reduces the high-dimensional deformation field, extracting dominant modes and calculating corresponding modal coefficients. A regression model integrates material parameters and modal coefficients via MLP theory, enabling rapid prediction of the global displacement field with millisecond-level precision. The method is validated through cantilever beam bending, single-zone, and multi-zone dam body deformation analyses. The findings suggest that the proposed method can achieve high-precision reconstruction of the deformation field with fewer modes, offering advantages such as low prediction error, reduced computation time, strong generalization capability, and good engineering applicability. This method provides an efficient and reliable research tool for response analysis and prediction of geotechnical structures such as CFRDs, demonstrating promising application prospects and promotional value.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106599"},"PeriodicalIF":4.1,"publicationDate":"2025-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145738144","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 : 2025-12-13DOI: 10.1016/j.enganabound.2025.106600
Haoran Yan , Yunpeng Lu , Hong Song , Yan Wang , Guiyong Zhang
This study develops a multiple-relaxation-time multiphase lattice Boltzmann flux solver (MRT-MLBFS) for the simulation of incompressible multiphase flows involving Newtonian and non-Newtonian fluids with large density ratio and rheological contrasts. The method integrates a phase-field model for interface capturing and a finite-volume-based LBFS framework to solve the macroscopic Navier–Stokes equations. By combining a MRT formulation with a flux reconstruction strategy, the method effectively suppresses numerical instabilities while reducing artificial dissipation, thereby ensuring robust simulations under extreme density and viscosity contrasts. Additionally, the algorithm is implemented in a fully explicit scheme and accelerated on GPU, which leads to a significant gain in computational efficiency, achieving speedups exceeding two orders of magnitude compared to CPU implementations. Through a series of benchmark cases, including droplet on plate, Rayleigh–Taylor instability, and droplet spreading on thin film, MRT-MLBFS is shown to capture complex interfacial evolution and nonlinear rheological effects with high accuracy and stability. This work establishes MRT-MLBFS as a reliable solution strategy for non-Newtonian multiphase flows with broad applicability.
{"title":"Multiphase lattice Boltzmann flux solver for non-Newtonian power-law fluid flows with high efficiency and stability","authors":"Haoran Yan , Yunpeng Lu , Hong Song , Yan Wang , Guiyong Zhang","doi":"10.1016/j.enganabound.2025.106600","DOIUrl":"10.1016/j.enganabound.2025.106600","url":null,"abstract":"<div><div>This study develops a multiple-relaxation-time multiphase lattice Boltzmann flux solver (MRT-MLBFS) for the simulation of incompressible multiphase flows involving Newtonian and non-Newtonian fluids with large density ratio and rheological contrasts. The method integrates a phase-field model for interface capturing and a finite-volume-based LBFS framework to solve the macroscopic Navier–Stokes equations. By combining a MRT formulation with a flux reconstruction strategy, the method effectively suppresses numerical instabilities while reducing artificial dissipation, thereby ensuring robust simulations under extreme density and viscosity contrasts. Additionally, the algorithm is implemented in a fully explicit scheme and accelerated on GPU, which leads to a significant gain in computational efficiency, achieving speedups exceeding two orders of magnitude compared to CPU implementations. Through a series of benchmark cases, including droplet on plate, Rayleigh–Taylor instability, and droplet spreading on thin film, MRT-MLBFS is shown to capture complex interfacial evolution and nonlinear rheological effects with high accuracy and stability. This work establishes MRT-MLBFS as a reliable solution strategy for non-Newtonian multiphase flows with broad applicability.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106600"},"PeriodicalIF":4.1,"publicationDate":"2025-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145730708","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 : 2025-12-12DOI: 10.1016/j.enganabound.2025.106597
Liying Wang, Hongtao Ye, Jianhao Cui, Yihan Xie
Aiming at the problems of nonlinearity and non-stationarity of vibration signals of hydraulic turbines under complex conditions, this paper proposes a condition recognition method based on the improved sparrow search algorithm (ISSA) to optimize the CNN-BiLSTM-Attention model. By introducing three strategies, namely generalized quadratic interpolation (GQI), adaptive sinusoidal perturbation and cooperative learning mechanism, the SSA algorithm is improved, and its convergence accuracy, ability to escape local optimum and global search stability are significantly enhanced. The parameters of the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) are optimized by the ISSA algorithm, and the vibration signals are decomposed to extract the dominant modal components as the model input. By integrating the spatial extraction ability of the convolutional neural network (CNN), the time series modeling ability of the bidirectional long short-term memory network (BiLSTM), and the SE attention mechanism, the CNN-BiLSTM-Attention model is established, and the hyperparameters of this model are optimized by the ISSA algorithm to enhance the model performance. The results show that the proposed model has an accuracy rate of 97.21 % in the condition recognition of hydraulic turbines, which verifies the effectiveness and superiority of this method in the recognition of complex vibration signals.
{"title":"A novel CNN-BiLSTM-attention framework based on improved sparrow search algorithm for hydraulic turbine condition recognition","authors":"Liying Wang, Hongtao Ye, Jianhao Cui, Yihan Xie","doi":"10.1016/j.enganabound.2025.106597","DOIUrl":"10.1016/j.enganabound.2025.106597","url":null,"abstract":"<div><div>Aiming at the problems of nonlinearity and non-stationarity of vibration signals of hydraulic turbines under complex conditions, this paper proposes a condition recognition method based on the improved sparrow search algorithm (ISSA) to optimize the CNN-BiLSTM-Attention model. By introducing three strategies, namely generalized quadratic interpolation (GQI), adaptive sinusoidal perturbation and cooperative learning mechanism, the SSA algorithm is improved, and its convergence accuracy, ability to escape local optimum and global search stability are significantly enhanced. The parameters of the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) are optimized by the ISSA algorithm, and the vibration signals are decomposed to extract the dominant modal components as the model input. By integrating the spatial extraction ability of the convolutional neural network (CNN), the time series modeling ability of the bidirectional long short-term memory network (BiLSTM), and the SE attention mechanism, the CNN-BiLSTM-Attention model is established, and the hyperparameters of this model are optimized by the ISSA algorithm to enhance the model performance. The results show that the proposed model has an accuracy rate of 97.21 % in the condition recognition of hydraulic turbines, which verifies the effectiveness and superiority of this method in the recognition of complex vibration signals.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":"183 ","pages":"Article 106597"},"PeriodicalIF":4.1,"publicationDate":"2025-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145731607","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