Computer-Aided Design最新文献

The Geometry-Independent Field approximaTion (GIFT) technique, an extension of isogeometric analysis (IGA), allows for separate spaces to parameterize the computational domain and approximate solution field. Based on the GIFT approach, this paper proposes a novel IGA methodology that incorporates toric surface patches for multi-sided geometry representation, while utilizing B-spline or truncated hierarchical B-spline (THB-spline) basis for analysis. By creating an appropriate bijection between the parametric domains of distinct bases for modeling and approximation, our method ensures smoothness within the computational domain and combines the compact support of B-splines or the local refinement potential of THB-splines, resulting in more efficient and precise solutions. To enhance the quality of parameterization and consequently boost the accuracy of downstream analysis, we suggest optimizing the composite toric parameterization. Numerical examples validate the effectiveness and superiority of our suggested approach.

Structural shape optimization plays a significant role in structural design, as it can find an appropriate layout and shape to improve structural performance. Isogeometric analysis provides a promising framework for structural shape optimization, unifying the design model and analysis model in the optimization process. This paper presents an adjoint-based analytical sensitivity for isogeometric shape optimization of Reissner–Mindlin shell structures. The shell structures are modeled by multiple NURBS surfaces and design variables are associated with the position of control points. A multilevel approach is performed with a coarse mesh for the design model and a dense mesh for the analysis model. The sensitivity propagation is achieved through a transformation matrix between the design and analysis models. Structural compliance minimization problems with and without constraints are studied and the optimization history shows that the optimization can converge quickly within fewer iterations. The developed formulations are validated through several numerical examples and applied to the optimization of cellular sandwich structures, which are widely used in engineering applications. Numerical results show that optimized sandwich panels can achieve better performance in bending resistance.

Rapid reduction in the number of quad-strips, to accommodate narrower surface passages or reduced shape fluctuation, leads to configurations that challenge existing spline surface constructions. A new spline surface construction for fast contracting polyhedral control-nets delivers good shape. A nestedly refinable construction of piecewise degree (2,4) is compared with a uniform degree (3,3) spline construction.

Collision detection is a crucial part of CNC machining, however, many state-of-the-art algorithms test collisions as a post-process, after the path-planning stage, or use conservative approaches that result in low machining accuracy in the neighborhood of the cutter’s contact paths. We propose a fast collision detection test that does not require a costly construction of the configuration space nor high-resolution sampling of the cutter’s axis and uses the information of the neighboring points to efficiently prune away points of the axis that cannot cause collisions. The proposed collision detection test is incorporated directly as a part of the tool motion-planning stage, enabling design of highly-accurate motions of a toroidal cutting tool along free-form geometries. We validate our algorithm on a variety of benchmark surfaces, showing that our results provide high-quality approximations with provably non-colliding motions.

Textures of woven fabrics are usually designed and produced according to geometric laws in the 2D plane. Physically Based Rendering (PBR) can further optimize and enrich the texture effect, but its application to the more complex 3D structures has been limited. This work reports a method that uses PBR and parametric modeling to construct woven textured materials with centimeter and millimeter level 3D structures. The method can design the structures of various woven fabrics without the need for analyzing the fabric structure details and transfer the inherently iterative work of fabric design to the digital space. The design can be directly applied to mainstream 3D modeling software for virtual presentations in different applications, hence improving the efficiency of woven fabric design and the fidelity of virtual presentation of fabric materials.

A method of generating a continuous toolpath that can be biased in a user-specified direction of travel is proposed for the fabrication of density-based functionally graded parts through Additive Manufacturing (AM). The methodology utilizes Lin Kernighan's (LK) Travelling Salesman Problem (TSP) solver over a digitized grid within the contour domain to generate a toolpath with minimal lifts and a common start and end point. Three force-based methods of digitization, namely rectangular, circular, and contour adaptive, are proposed in this work. Each of these methods initialize from a structured or an unstructured grid, where the grid points are assumed to be connected with either linear (rectangular digitization) or a combination of linear and torsional springs (circular and contour adaptive digitization). Enforcing an equilibrium amongst the spring forces and appropriately selecting the ideal spring length, the necessary configuration of grid points can be generated for a desired toolpath.

The density of grid points (consequently, part density) can be varied through the user-defined input function or an image-based density map imposed on the ideal spring length over the contour domain. The proposed toolpath, as a case study, was implemented for printing a bone with density prescribed through a CT scan image stack. The CT scan of the printed part qualitatively establishes the conformity of the toolpath to the user-specified density gradient.

We introduce a progressive and iterative method for B-spline curve and surface approximation, incorporating parameter correction based on the Newton iterative method. While parameter corrections have been used in existing Geometric Approximation (GA) methods to enhance approximation quality, they suffer from low computational efficiency. Our approach unifies control point updates and parameter corrections in a progressive and iterative procedure, employing a one-step strategy for parameter correction. We provide a theoretical proof of convergence for the algorithm, demonstrating its superior computational efficiency compared to current GA methods. Furthermore, the provided convergence proof offers a methodology for proving the convergence of existing GA methods with location parameter correction.

Neural network-based approaches for solving partial differential equations (PDEs) have recently received special attention. However, most neural PDE solvers only apply to rectilinear domains and do not systematically address the imposition of boundary conditions over irregular domain boundaries. In this paper, we present a neural framework to solve partial differential equations over domains with irregularly shaped (non-rectilinear) geometric boundaries. Given the shape of the domain as an input (represented as a binary mask), our network is able to predict the solution field, and can generalize to novel (unseen) irregular domains; the key technical ingredient to realizing this model is a physics-informed loss function that directly incorporates the interior-exterior information of the geometry. We also perform a careful error analysis which reveals theoretical insights into several sources of error incurred in the model-building process. Finally, we showcase various applications in 2D and 3D, along with favorable comparisons with ground truth solutions.

Point cloud denoising is a crucial task in the field of geometric processing. Recent years have witnessed significant advancements in deep learning-based point cloud denoising algorithms. These methods, compared to traditional techniques, demonstrate enhanced robustness against noise and produce point cloud data of higher fidelity. Despite their impressive performance, achieving a balance between denoising efficacy and computational efficiency remains a formidable challenge in learning-based methods. To solve this problem, we introduce LPCDNet, a novel lightweight point cloud denoising network. LPCDNet consists of three main components: a lightweight feature extraction module utilizing trigonometric functions for relative position encoding; a non-parametric feature aggregation module to leverage semantic similarities for global context comprehension; and a decoder module designed to realign noise points with the underlying surface. The network is designed to capture both local details and non-local structures, thereby ensuring high-quality denoising outcomes with a minimal parameter count. Extensive experimental evaluations demonstrate that LPCDNet achieves comparable or superior performance to state-of-the-art methods, while significantly reducing the number of learnable parameters and the necessary running time.

Recent advances in generative modeling, namely Diffusion models, have revolutionized generative modeling, enabling high-quality image generation tailored to user needs. This paper proposes a framework for the generative design of structural components. Specifically, we employ a Latent Diffusion model to generate potential designs of a component that can satisfy a set of problem-specific loading conditions. One of the distinct advantages our approach offers over other generative approaches is the editing of existing designs. We train our model using a dataset of geometries obtained from structural topology optimization utilizing the SIMP algorithm. Consequently, our framework generates inherently near-optimal designs. Our work presents quantitative results that support the structural performance of the generated designs and the variability in potential candidate designs. Furthermore, we provide evidence of the scalability of our framework by operating over voxel domains with resolutions varying from $32^3$ to $128^3$. Our framework can be used as a starting point for generating novel near-optimal designs similar to topology-optimized designs.