Advances in Engineering Software最新文献

A computationally efficient algorithm to calculate the weighting coefficients required to evaluate derivatives for arbitrary multi-dimensional distributions of points is presented. The iterative algorithm guarantees IEEE 754 64-bit precision (at least 15 significant decimal digits) for the weighting coefficients. Convergence acceleration is achieved through the use of a Taylor series of up to third order, and hyper-dual numbers to obtain the derivatives required for the Taylor series. The method is applied as part of a finite point solution for three test examples, a Poisson equation, creeping flow around a cylinder, and heat conduction in a triangular annulus. The open source FORTRAN-90 implementation has been optimised for random distributions of points in 1 to 3 dimensions.

Deep learning (DL) methods have been widely applied for structural response prediction. However, classical DL methods rely heavily on training data with no consideration to the information at the structural level. They generally show poor generalization performance for unknown structural forms. To address this issue, a graph representation is proposed in this study to abstractly represent the actual structure as a graph structure, which is subsequently processed using the graph isomorphic network (GIN). Due to the unique self-evolutionary mechanism of the graph structure, the GIN model is able to disentangle from the training data, leading to excellent generalization performance on the task of response analysis with unknown structural forms. Taking train-bridge coupled (TBC) systems as examples, for different working conditions, the test results show that the prediction accuracy and generalization performance of the GIN model reach an extremely high level. Moreover, a GIN-based iterative system is proposed in this study. It exhibits significantly better generalization performance than classical DL methods for unknown structural forms, indicating its high potential for practical applications in various engineering fields. The content and findings of this study contribute to the future development of a new generation of DL methods with advanced performance.

This study addresses the challenge of multi-objective optimization of a composite shell structure while adhering to constraints on the number of calls to a pseudo-experimental model, simulating real experiments. Two considered objective functions are defined to determine the investigated structure’s dynamic properties and material costs; the optimization involves genetic algorithms, neural surrogate model and multi-fidelity finite-element models. The results of multi-objective optimization were presented as Pareto fronts. A new strategy for preliminary result verification is proposed, significantly reducing the need for a computationally intensive complete verification that requires complex models or experimental investigations. Two different indicators are applied to assess the quality of the obtained Pareto fronts; one is a new one proposed in the paper. Moreover, a multi-fidelity approach is discussed, and three finite element models with different mesh densities are employed, together with a pseudo-experimental model constructed using high-fidelity results and incorporating a nonlinear transformation. However, challenges arise due to the arbitrarily constrained number of pseudo-experiments, limiting future experiments is crucial. The study highlights the need for further analysis of Pareto front indicators and statistical analysis of applied tools like deep neural networks and genetic algorithms. Future research directions include exploring ensemble learning in surrogate models for potential optimization benefits.

Multi-objective optimization (MOO) for structural design is addressed. A new MOO algorithm, named MOEA/D-EpDE, which combines the advantages of a recently developed pbest-based differential evolution method (EpDE) and the multi-objective evolutionary algorithm based on decomposition with dynamical resource allocation (MOEA/D_DRA), is proposed to solve such challenging MOO problems effectively. In MOEA/D-EpDE, a decomposition approach is performed using MOEA/D_DRA to convert a problem of approximation of the Pareto front (PF) into many scalar optimization problems, in which a dynamic computational resource allocation strategy is used to optimize the computational efforts. The EpDE algorithm, a robust single objective optimization (SOO) algorithm, is improved for MOO to solve the scalar optimization problems effectively. A simple technique for integrating an external archive to MOEA/D-EpDE is also developed to save good Pareto optimal solutions during the optimization process. The performance of MOEA/D-EpDE is first evaluated through 5 bi-objectives (ZDT1–4 and ZDT6) and 7 tri-objectives unconstrained benchmark functions. Numerical results revealed that the proposed method outperformed several MOO algorithms given the inverted generational distance (IGD) indicator. In the end, MOEA/D-EpDE is applied to solve three real-world design problems, including a welded-beam and two nonlinear inelastic truss structures. The effectiveness of the proposed algorithm is confirmed through comparison with some recently developed algorithms regarding several indicators: generational distance (GD), GD+, IGD, IGD+, and Hypervolume (HV).

The advancement in high-order computational methods is reshaping the landscape of mesh generation in Computational Fluid Dynamics (CFD), steering the focus towards curvilinear mesh techniques to meet the escalating accuracy demands. Gridder-HO, the software designed to generate high-order curvilinear mesh efficiently and rapidly, has been developed. Gridder-HO supports the elevation of meshes to P2 (quadratic-order) or P3 (cubic-order). It features a layered architecture and utilizes the concurrent hash table and the Alternating Digital Tree (ADT) data structure, supporting thread-level parallelism to convert straight-edge mesh into high-order curvilinear mesh seamlessly. Gridder-HO utilizes the projection method based on a thread pool to precisely preserve geometry, and employs a novel localized RBF method with ADT for volume node interpolation to untangle the mesh, which aims to achieve a satisfactory balance between efficiency and accuracy. Validated through CFD simulations using the GPU-accelerated Python Flux Reconstruction (PyFR) solver, the practicality of Gridder-HO is demonstrated across various Reynolds numbers in typical cases such as sphere, cylinder, and SD7003 airfoil. These results confirm the high-order curvilinear meshes generated by Gridder-HO meet the high-order requirements of emerging computational methods. Moreover, Gridder-HO exemplifies its effectiveness in generating large-scale, high-order curvilinear meshes for the DLR-F6 transport aircraft configuration standard test cases. It elevates a mesh with 5 million elements to P2 in 3 min 39 sec at 68% parallel efficiency on 16 threads, and another with 14 million elements to P3 in 52 min 39 sec at 60% efficiency, illustrating its efficiency and potential in satisfying the demands of complex geometries in engineering applications.

Enhancing nuclear power plant (NPP) safety is demanded because of the recent beyond-design-basis earthquake near a NPP. Therefore, research on improving the seismic performance of the electrical cabinet, which ensures the safe operation of NPPs, is needed. In this paper, a tuned mass damper (TMD) is employed to control the seismic response of cabinet. To design the TMD, we employ existing design equations or perform numerical model–based optimization. However, limitations, such as inconsistencies with targeted control of the load and structure, the possibility of converging a local solution, and the high cost of numerical analysis. Therefore, this paper proposes a two-step machine learning and optimization method. Such an approach is utilized to find the optimal global design solution and reduce numerical analysis costs. Each step involves the design of experiment (DOE), response surface, and optimization. Notably, range setting in the DOE accounts for the difference between each step. In the first step, the sampling range is widened to determine the relationship between the design variables and the cabinet's response, and in the second step, the sampling range is narrowed depending on the result of the first step. Consequently, the proposed method reduced the cabinet's response by 35.4 % on average and numerical analysis cost declined by 1/3.

The seismic safety of arch dams has long been a focal point of research. Due to the complexity of modeling and computation, analyzing the seismic response of arch dams using traditional finite element methods requires a considerable amount of time. In the event of a sudden earthquake, it is challenging to quickly obtain stress analysis results or conduct a safety assessment. To address these issues, a rapid and automated analysis procedure is proposed in this paper, providing seismic response of arch dams within hours after an earthquake. The procedure includes a pre-processing program, a computing program EACD-3D-2008, and a post-processing program, achieving a fully automated process from generating non-uniform earthquakes to analyzing dam dynamic responses and visualizing computation results. As a case study, the 294.5 m high Xiaowan arch dam in southwest China is analyzed, which is equipped with strong motion instruments that have recorded several small earthquakes. The case study revealed that accounting for the non-uniformity of the earthquake significantly improves simulation results, with maximum principal stresses typically occurring near the dam-foundation rock interface. Additionally, the procedure effectively compensates for missing data, allowing for the successful supplementation of the missing acceleration records. For stronger earthquakes, high-stress regions are clearly displayed in the result visualization, providing an effective reference for safety assessment. The case study validated the accuracy and wide applicability of the procedure, demonstrating its potential to offer valuable insights for similar analyses in various engineering projects.

This paper describes the design and implementation of a digital twin model in the open-source MuPIF simulation platform. MuPIF enables a user-defined data model based on an ontology or schema to be created. A representation of the data model is generated in a target data management system. The data model, integrated with MuPIF, lets model entities to be linked, and model attributes can be assigned to simulation workflows inputs and outputs. The model is semantically-defined, provides full traceability, and has a web-based API for data discovery.

The existing data indicate that two-thirds of engineering accidents occur during construction among which engineering accidents caused by scaffold collapse account for a large proportion. Due to the complex mechanical behavior of connection and random nature of scaffold system caused by random geometrical imperfection, the reliability of scaffold system is lower than other kinds of building structures. However, the method considering the random geometrical imperfection is limited. To facilitate the analysis of random geometrical imperfection, the original numerical algorithm is proposed based on ANSYS Parametric Design Language. Through proposed method, two types of geometrical imperfections, i.e., the nodal location error and initial curvature can be automatically considered. The randomness in initial curvature includes random magnitude and random direction. The established numerical model is as close to reality as possible and the process of establishing stochastic numerical model can be automatically finished. The only work that needs to be done is to enter the dimensions of the scaffold. Except the propose of numerical algorithm, the objective of this study is to reveal the influence of geometrical imperfection on random distribution of loading capacity of scaffold system under different load conditions. The influence of random geometrical imperfection on probabilistic distribution of loading capacity is systematically investigated. The results indicated that there may be several buckling modes exist and the buckling mode occurred in actual condition is closely related to the random distribution of geometrical imperfection. The load factor of internal post (point 3) is 8 %–12 % larger than that of corner post. The load factor of side post is 4.7 %–7.2 % larger than that of corner post. The ultimate bending capacity M_u has little influence on the loading capacity of scaffold system when the initial bending stiffness k_o is small enough.

This paper provides an insight into the development of a state-of-the-art video processing system to address limitations within Durham University’s ‘Encore’ lecture capture solution. The aim of the research described in this paper is to digitally remove the persons presenting from the view of a whiteboard to provide students with a more effective online learning experience. This work enlists a ‘human entity detection module’, which uses a remodelled version of the Fast Segmentation Neural Network to perform efficient binary image segmentation, and a ‘background restoration module’, which introduces a novel procedure to retain only background pixels in consecutive video frames. The segmentation network is trained from the outset with a Tversky loss function on a dataset of images extracted from various Tik-Tok dance videos. The most effective training techniques are described in detail, and it is found that these produce asymptotic convergence to within 5% of the final loss in only 40 training epochs. A cross-validation study then concludes that a Tversky parameter of 0.9 is optimal for balancing recall and precision in the context of this work. Finally, it is demonstrated that the system successfully removes the human form from the view of the whiteboard in a real lecture video. Whilst the system is believed to have the potential for real-time usage, it is not possible to prove this owing to hardware limitations. In the conclusions, wider application of this work is also suggested.