Method, system and medium for converting mesh geometry to watertight boundary representation
By generating a quadrilateral patch network in a computer-aided design program and using hyperlimit interpolation, the mesh and smooth boundary curve are combined, and the accuracy problem of converting mesh geometry into watertight boundary representation is solved, achieving efficient generation design and topology optimization.
Patent Information
- Application Number
- CN201980072338.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2019-04-18
- Filing Date
- 2019-11-08
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2039-11-08
AI Technical Summary
Existing computer-aided design (CAD) software is difficult to achieve high-precision and seamless integration when converting mesh geometry into watertight boundary representations, especially in the process of generative design and topological optimization. Mesh boundary approximation is not accurate enough, which affects the The efficiency of physical simulation and material synthesis is achieved.
By generating a quadrilateral patch network from a polygon mesh in a computer-aided design program, combining the mesh with a smooth boundary curve using hyperlimit interpolation, defining a locally refineable representation of smooth surfaces, and using the model through stitching operations Combining entities to form a watertight three-dimensional model.
It realizes high-precision watertight boundary representation, which can be effectively processed and edited in CAD programs, improves the efficiency of generative design and topological optimization, and ensures the accuracy and economicality of physical design.
Smart Images

Figure CN113228115B_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims U.S. patent application No. 62 / 758,053, filed on November 9, 2018, entitled "Conversion of Generative Design Geometry to Editable and Watertight Boundary Representation in Computer-Aided Design," and U.S. patent application No. 16 / 388,771, filed on April 18, 2019, entitled "Conversion of Mesh Geometry to Watertight Boundary Representation," both of which are incorporated herein by reference in their entirety. Technical Field
[0003] The present application relates to computer-aided design of physical structures that can be manufactured using additive manufacturing, subtractive manufacturing, and / or other manufacturing systems and techniques. Background Art
[0004] Computer-aided design (CAD) software has been developed and used to generate three-dimensional (3D) representations of objects, and computer-aided manufacturing (CAM) software has been developed and used to manufacture the physical structure of those objects, for example using computer numerical control (CNC) manufacturing techniques. Typically, CAD software uses a boundary representation (B-Rep) format to store a 3D representation of the geometry of the object to be modeled. A B-Rep model is a set of connected surface elements that specify the boundaries between the solid and non-solid parts of the modeled 3D object. In a B-Rep model (commonly referred to as B-Rep), the geometry is stored in the computer using smooth and precise mathematical surfaces, as opposed to the discrete and approximated surfaces of mesh model geometry, which are difficult to use in CAD programs.
[0005] In addition, CAD programs have been used in conjunction with subtractive manufacturing systems and techniques. Subtractive manufacturing refers to any manufacturing process that creates a 3D object from stock material (usually a "blank" or "workpiece" that is larger than the 3D object) by cutting away multiple portions of the stock material. Such a manufacturing process typically involves the use of multiple CNC machine tools in a series of operations that begin with roughing, optional semi-finishing, and finishing. In addition to CNC machining, other subtractive manufacturing techniques include electrode discharge machining, chemical machining, water jet machining, and the like. In contrast, additive manufacturing (also known as non-solid form manufacturing or 3D printing) refers to any manufacturing process that builds a 3D object from a raw material (usually a powder, liquid, suspension, or molten solid) in a series of layers or cross-sections. Examples of additive manufacturing include fused filament fabrication (FFF) and selective laser sintering (SLS). Other manufacturing techniques for building 3D objects from raw materials include casting and forging (hot and cold).
[0006] In addition, CAD software has been designed to perform automatic generation of 3D geometry (generative design) for a part or one or more parts in a larger system of parts to be manufactured. This automatic generation of 3D geometry is usually limited to the design space specified by the CAD software user, and the 3D geometry generation is usually subject to design goals and constraints, which can be defined by the CAD software user or others and imported into the CAD software. Design goals (such as minimizing material waste or the weight of the designed part) can be used to drive the geometry generation process towards better designs. Design constraints can include structural integrity constraints of individual parts (i.e., the requirement that the part must not fail under the expected structural load during the use of the part), physical constraints imposed by the larger system (i.e., the requirement that the part will not interfere with another part in the system during use), and aesthetic and manufacturability constraints. In addition, the input to the generative design process can include a set of input entities (B-Rep inputs) that specify the boundary conditions of the generative design process, but many modern generative design solvers do not directly operate on the exact surface boundary representation of their input entities. Instead, the B-Rep is sampled and replaced with a volume representation (eg, level set or tetrahedral or hexahedral mesh), which is significantly more convenient and efficient for physical simulation and material synthesis computed by the solver. Summary of the invention
[0007] This description relates to computer-aided design of physical structures that can be manufactured using additive manufacturing, subtractive manufacturing, and / or other manufacturing systems and techniques.
[0008] Generally, one or more aspects of the subject matter described in this specification may be embodied in one or more methods (and may also be embodied in one or more non-transitory computer-readable media, the non-transitory computer-readable media tangibly encoding a computer program operable to cause a data processing device to perform operations), the method comprising: obtaining a first three-dimensional model of an object to be manufactured in a computer-aided design program, wherein the first three-dimensional model includes a polygonal mesh associated with one or more modeled entities, the one or more modeled entities being in a boundary representation format; generating, by the computer-aided design program, a quadrilateral patch network from the polygonal mesh, the quadrilateral patch network combining the polygonal mesh with one or more smooth boundary curves corresponding to the one or more modeled entities using a transfinite interpolation method; defining, by the computer-aided design program, one or more locally refinable smooth surface representations using the quadrilateral patch network as input and based at least in part on a minimum size representable by a geometric modeling kernel of the computer-aided design program; and combining, by the computer-aided design program, the one or more locally refinable smooth surface representations with the one or more modeled entities to form a second three-dimensional model of the object, the second three-dimensional model being watertight at the one or more smooth boundary curves.
[0009] The polygonal mesh may be generated by a generative design process that employs the one or more modeled entities as one or more subspaces of an optimization domain of the generative design process. The generative design process may employ a volume meshing method or a level set method to create a generative design of the object, and the polygonal mesh may include an original or converted triangular mesh.
[0010] The generating may include: extracting a quadrilateral patch layout of the quadrilateral patch network from the polygonal mesh, wherein the quadrilateral patch layout represents a quadrilateral partitioning of a domain of a global continuous graph of the polygonal mesh; mapping at least one boundary of the quadrilateral patch layout to the one or more smooth boundary curves corresponding to the one or more modeled entities; and constructing a surface patch using the global continuous graph of the polygonal mesh, the mapping of the at least one boundary of the quadrilateral patch layout to the one or more smooth boundary curves, and transfinite interpolation of the one or more smooth boundary curves. The method may include receiving user input specifying a quadrilateral size, a quadrilateral shape, or both for the quadrilateral patch network.
[0011] The definition may include: constructing one or more initial locally refinable smooth surface representations based on the quadrilateral patch network; refining the one or more initial locally refinable smooth surface representations at one or more boundaries of the one or more initial locally refinable smooth surface representations to approximate the one or more smooth boundary curves within a first tolerance value, the first tolerance value being equal to or greater than the minimum size representable by the geometric modeling kernel of the computer-aided design program and within a tolerance range supported by the geometric modeling kernel of the computer-aided design program; freezing control points of the one or more locally refinable smooth surface representations at the one or more boundaries; and modifying the remaining internal portion of the one or more locally refinable smooth surface representations to approximate the polygonal mesh within a second tolerance value, the second tolerance value being equal to or greater than the accuracy of the method used to obtain the polygonal mesh and being at least one order of magnitude greater than the first tolerance value.
[0012] The method may include receiving user input specifying the second tolerance value. The method may include: identifying at least one negative boundary cross-field singularity of the polygonal mesh; and relocating the at least one negative boundary cross-field singularity to an interior vertex of the polygonal mesh. The combining may include using a stitching operation to connect the one or more locally refinable smooth surface representations to the one or more modeled entities in the boundary representation format.
[0013] The computer-aided design program may include two or more programs operating in concert on two or more independent computer processors. The method may include providing, by the computer-aided design program, the second three-dimensional model for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems. The providing may include saving the second three-dimensional model to a permanent storage device for use when using the one or more computer-controlled manufacturing systems to manufacture the physical structure corresponding to the object. The one or more computer-controlled manufacturing systems may include an additive manufacturing machine or a subtractive manufacturing machine, and the providing may include: using the second three-dimensional model to generate a tool path specification for the additive manufacturing machine or the subtractive manufacturing machine; and using the tool path specification generated for the additive manufacturing machine or the subtractive manufacturing machine, using the additive manufacturing machine or the subtractive manufacturing machine to manufacture at least a portion of the physical structure corresponding to the object.
[0014] In addition, one or more aspects of the subject matter described in this specification can be embodied in one or more systems, the systems employing one or more methods and one or more non-transitory computer-readable media (e.g., non-transitory storage media) having stored thereon instructions of a computer-aided design program; and one or more data processing devices, the one or more data processing devices being capable of executing the instructions of the computer-aided design program to perform operations specified by the instructions of the computer-aided design program (as described herein). The mesh can be generated by a generative design process that employs the one or more modeled entities as one or more subspaces of an optimization domain of the generative design process, and the second tolerance value can be set based on a representation accuracy of the generative design process. The system may include one or more computer-controlled manufacturing systems, the manufacturing systems including an additive manufacturing machine or a subtractive manufacturing machine, wherein the one or more data processing devices are capable of running the instructions of the computer-aided design program to: generate a tool path specification for the additive manufacturing machine or the subtractive manufacturing machine from the three-dimensional model, and use the tool path specification generated for the additive manufacturing machine or the subtractive manufacturing machine to manufacture at least a portion of the physical structure corresponding to the object with the additive manufacturing machine or the subtractive manufacturing machine.
[0015] Specific embodiments of the subject matter described in this specification can be implemented to achieve one or more of the following advantages. A general topological mesh (e.g., an organically shaped mesh produced by a physics solver of a generative design process) can be approximated with one or more locally refinable smooth surface representations (e.g., non-uniform rational Catmull-Clark surfaces with T-junctions (T-NURCC)), while also fitting the smooth surface representations to a set of smooth boundary curves used to refine the mesh boundaries to a level of accuracy that produces a watertight boundary representation (e.g., B-Rep) model. The smooth boundary curves can be specified by the application to ensure that the generated surface accurately conforms to existing geometry, such as patch boundaries, intersections, and trim curves. The generated smooth surface can be fitted to mathematically accurate smooth boundary curves using tolerance values different from the tolerance values used to fit the interior portion of the smooth surface to the mesh.
[0016] The approximation tolerance of the generated smooth surface can be guaranteed to within a limit that ensures that the surface will adhere to the rest of the 3D model while also seamlessly approximating a mesh with a looser tolerance away from the smooth boundary curve. For example, the generated surface (e.g., T-NURCC) can approximate the smooth boundary curve to within 3 to 6 orders of magnitude smaller than the size of the 3D model itself, so the smooth boundary curve is essentially as good as an exact intersection, and the smooth boundary curve replaces the exact intersection when stitching the generated surface to the entity in the 3D model. Thus, a water-tight synthesis of a locally refinable smooth surface representation (e.g., T-NURCC) to a modeled entity (e.g., B-Rep) is achieved, where free-form organic-like shapes are accompanied on both sides by classical CAD surfaces, such as planes, cylinders, and non-uniform rational basis spline (NURBS) patches. Applications of the described systems and techniques include reverse engineering (object scanning) as well as topology optimization and generative design, where meshes often mix and match organic and non-organic forms.
[0017] The described systems and techniques promote the use of generative design and topology optimization, which provides a fast and cost-effective way to obtain optimized geometry that meets physical design requirements. Most physical solvers operate on volume representations such as level sets and tetrahedral meshes, and optimized designs are usually exported as meshes. In order to integrate these designs into CAD programs, rather than CAD engineers having to spend a lot of time reshaping the composite shapes from scratch using traditional modeling techniques, the systems and techniques described herein can be used to easily convert meshes into smooth surfaces that can be effectively processed in CAD programs (e.g., easily edited without a lot of processing time) and seamlessly integrated with the rest of the CAD model. In addition, the availability of accurate geometry allows the final accurate simulation and verification of the mechanical / aesthetic properties of the body to be performed. This verification is usually not accurate for meshes because the meshes define the shape in an approximate manner. For example, on a mesh, it is impossible to check whether the curvature is continuous.
[0018] The details of one or more embodiments of the subject matter described in this specification are set forth in the accompanying drawings and the detailed description that follows. Other features, aspects, and advantages of the invention will become apparent from the detailed description, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 An example of a system that can be used to design and manufacture a physical structure is shown.
[0020] Figure 2A An example of a process for converting mesh geometry into a watertight boundary representation of an object to be manufactured is shown.
[0021] Figure 2B Shows Figure 2A Example of the conversion process applied to an organic mesh synthesized by topology optimization of a CAD solid model.
[0022] Figure 3A is a graphical diagram showing an overview of an example of a process of generating a quadrilateral patch network and using the quadrilateral patch network to define one or more locally refinable smooth surface representations.
[0023] Figure 3B is a graphical plot comparing the areas affected by different methods with the transfinite interpolation method.
[0024] Figure 3C is a graphical diagram illustrating an example of the effect of a graph size field on the construction of a transfinite interpolation.
[0025] Figure 4A is a flow chart illustrating an example of a process for generating a quadrilateral patch network.
[0026] Figure 4B is a graphical diagram illustrating an example of transfinite patch construction.
[0027] Figure 5A is a flow chart illustrating an example of a process for defining one or more locally refinable smooth surface representations using a quadrilateral patch network.
[0028] Figure 5B is a graphical diagram showing four boundary conditions.
[0029] Figure 5C is a graphical diagram showing an example of boundary singularity relocation.
[0030] Figure 5D is a graphical diagram showing an example of local knot insertion during boundary refinement and local refinement propagation for balanced surface domains.
[0031] Figure 5E is a graphical diagram showing an example of a quadrilateral domain sample mesh, nesting, and refinement.
[0032] FIG. 6A to FIG. 6C Examples of results of the process applied to organic meshes produced by topology optimization and generative design applications are shown.
[0033] Figure 7 is a schematic diagram of a data processing system including a data processing device that can be programmed as a client or a server.
[0034] Like reference numbers and designations in the various drawings indicate like elements. DETAILED DESCRIPTION
[0035] Figure 1 An example of a system 100 that can be used to design and manufacture a physical structure is shown. A computer 110 includes a processor 112 and a memory 114, and the computer 110 can be connected to a network 140, which can be a private network, a public network, a virtual private network, etc. The processor 112 can be one or more hardware processors, each of which can include multiple processor cores. The memory 114 can include both volatile and non-volatile memory, such as random access memory (RAM) and flash RAM. The computer 110 can include various types of computer storage media and devices (the computer can include the memory 114) to store instructions for programs running on the processor 112, including a computer-aided design (CAD) program 116 that implements the following functions: three-dimensional (3D) modeling functions and conversion of mesh geometry to a watertight boundary representation of a part to be manufactured. The mesh geometry can be obtained from different sources, but typically, the mesh geometry will be associated with one or more modeled entities in a boundary representation format.
[0036] For example, the system 100 may include a 3D scanner (not shown) that scans the physical object 180 to generate a mesh representation of the physical object, and the CAD program 116 may use this input mesh to generate one or more modeled entities (e.g., B-Rep) by fitting regular geometry to the mesh, and the CAD program 116 may use the systems and techniques described herein to convert the remainder of the mesh (those portions that are too complex to be modeled using regular geometry) to a boundary representation format (e.g., B-Rep) and connect the remainder to the one or more modeled entities. In this example, the CAD program 116 includes both scanning and modeling functionality.
[0037] As another example, the CAD program 116 may implement one or more generative design processes that use material or microstructural techniques and / or geometric or macrostructural techniques for generative design (e.g., solid isotropic material with penalty (SIMP) and / or level set methods for topology optimization). One or more modeled entities may be generated in (or loaded into) the CAD program(s) using conventional 3D modeling techniques, and these modeled entities may be used as one or more subspaces of the optimization domain of the generative design process. Thus, the generative design process may output a mesh representing the automatically generated geometry that will be connected to the modeled entities used as input. Additionally, in various embodiments, the CAD program 116 may implement physical simulation (finite element analysis (FEA) or other) as well as manufacturing control functions.
[0038] As used herein, CAD refers to any suitable program for designing a physical structure that meets specified design requirements, whether or not the program is capable of interfacing with and / or controlling a particular manufacturing device. Thus, the CAD program 116 may include a computer-aided engineering (CAE) program, a computer-aided manufacturing (CAM) program, or the like. The program 116 may be run locally on the computer 110, remotely on a computer of one or more remote computer systems 150 (e.g., one or more server systems of one or more third-party suppliers accessible by the computer 110 via a network 140), or both locally and remotely. Thus, the CAD program 116 may be two or more programs operating in concert on two or more separate computer processors, because the program 116 operating locally on the computer 110 may offload processing operations (e.g., mesh conversion, generative design, and / or physical simulation operations) “to the cloud” by causing one or more programs 116 on one or more computers 150 to perform the offloaded processing operations.
[0039] CAD program 116 presents a user interface (UI) 122 on display device 120 of computer 110, which can be operated using one or more input devices 118 (e.g., keyboard and mouse) of computer 110. Figure 1 1 as separate devices, the display device 120 and / or the input device 118 may also be integrated with each other and / or with the computer 110, such as in a tablet computer (e.g., a touch screen may be the input / output device 118, 120). In addition, the computer 110 may include a virtual reality (VR) or augmented reality (AR) system, or be part of the system. For example, the input / output devices 118, 120 may include VR / AR input gloves 118a and / or VR / AR headsets 120a.
[0040] As described above, the CAD program 116 implements 3D modeling capabilities, which means that the CAD program 116 can be used to build the 3D model 132. The CAD program 116 can implement physical simulation (locally and / or through remote procedure calls) to help build the 3D model 132. Physical simulations such as FEA, computational fluid dynamics (CFD), acoustics / noise control, heat transfer, and / or computational injection molding simulations are often an integral part of CAD-based product development. The CAD program 116 can be used to build an accurate geometric description of the design model, and physical simulation can improve performance without time-consuming physical testing.
[0041] The CAD program 116 may provide user interface elements that enable a user to specify inputs for the physical simulation, such as materials and load states for the 3D model 132, wherein the load states define the loads in different directions to which the designed part will be subjected during use of the part. Thus, the user 190 may interact with the UI 122 of the CAD program 116, including generating a complete mechanical problem definition for a part to be manufactured, in order to construct and modify the 3D model 132 that may be stored in the 3D model document 130. Figure 1 In the example of , 3D model 132 has a particular bracket, but this is just one of many possible 3D models that can be designed using the systems and techniques described herein.
[0042] In addition, the CAD program 116 may implement at least one generative design method (locally and / or through remote procedure calls) that enables the CAD program 116 to automatically generate one or more portions of the 3D model 132 (or the entirety of the 3D model) based on design criteria, wherein the geometric design is iteratively optimized based on simulation feedback. Note that as used herein, "optimization" (or "optimal") does not mean that the best of all possible designs is achieved in all cases, but rather that the best (or near-optimal) design is selected from a finite set of possible designs that can be generated within an allotted time given available processing resources. The design criteria may be defined by the user 190 or another party and imported into the CAD program 116. For example, the design space, load states, and design criteria may be obtained for an object modeled by the CAD program 116 by receiving user input, for example, via the UI 122 on the display device 120.
[0043] In the example shown, the 3D model 132 includes a modeled solid portion 134 and a mesh portion 136. The modeled solid 134 is used as an input to a generative design process (e.g., a B-Rep as a volume-retaining input), which generates a mesh 136 taking into account boundary conditions set by the input geometry. Various generative design processes can be used to optimize the shape and topology of the mesh portion 136 of the 3D model 132. The input solid 134 can be converted to a mesh geometry for use as an input to the generative design process, and the mesh output from the generative design process can include mesh portions corresponding to the input solid 134, and these portions can be removed to generate the mesh portion 136. This mesh portion 136 is then converted to a boundary representation format (e.g., B-Rep, T-spline, subdivision surface, and / or other computer 3D modeling formats) and connected to the modeled solid portion 134 to form a watertight model 132 at the boundary between the two portions 134, 136.
[0044] Additionally, in some embodiments, the CAD program 116 implements manufacturing control functionality. Once the user 190 is satisfied with the 3D model 132, the 3D model 132 may be stored as a 3D model document 130 and / or another representation of the model (e.g., a .STL file for additive manufacturing). This may be done at the request of the user 190, or at the user's request for another action, such as sending the 3D model 132 to an additive manufacturing (AM) machine 170 or other manufacturing machine, which may be directly connected to the computer 110, as shown, or connected via a network 140. This may involve post-processing performed on the local computer 110, or exporting the 3D model 132 to a cloud service for electronic documents that are based on manufacturing. Note that an electronic document (referred to as a document for simplicity) may be a file, but does not necessarily correspond to a file. A document may be stored in a portion of a file that holds other documents, in a single file dedicated to related documents, or in multiple coordinated files.
[0045] In any case, the CAD program 116 can provide a document 160 (with tool path instructions in an appropriate format) to an AM machine 170 to produce a physical structure 180. The AM machine 170 can use one or more additive manufacturing technologies, such as particle technologies (e.g., powder bed fusion (PBF), selective laser sintering (SLS), and direct metal laser sintering (DMLS)), extrusion technologies (e.g., fused deposition modeling (FDM), which may include metal deposition AM). In some cases, the AM machine 170 directly builds the physical structure 180, while in some cases, the AM machine 170 builds a mold for use in casting or forging the physical structure 180. In addition, the user 190 can save or send the 3D model 132 for later use. For example, the CAD program 116 can store the document 130 including the 3D model 132.
[0046] In some embodiments, the CAD program 116 implements manufacturing control functions for a subtractive manufacturing (SM) machine 174 (e.g., a computer numerically controlled (CNC) milling machine, such as a multi-axis multi-tool milling machine) to be used in the manufacturing process. For example, the CAD program 116 can generate CNC instructions for a machine tool system 174, which includes a plurality of tools 174A that can be used for various machining operations (e.g., solid carbide fillet tools of different sizes and shapes, and insert tools of different sizes that receive metal inserts to create different cutting surfaces). Thus, in some implementations, the CAD program 116 can provide a corresponding document 160 (with a tool path specification in an appropriately formatted format, such as a CNC numerical control (NC) program) to the SM machine 174 for use in manufacturing the physical device 180 using various cutting tools, etc.
[0047] In some implementations, the CAD program 116 provides instructions for building the workpiece 180 in a first manufacturing stage, for example, using direct AM and / or using forging and / or casting methods, and the CAD program 116 also provides instructions for machining the workpiece 180 (e.g., roughing and / or finishing operations) in a second manufacturing stage, for example, in a 3-axis CNC milling system to form a completed structure. In general, a variety of different manufacturing systems and techniques may be employed, alone or in combination, to produce the final structure, and the CAD program 116 may include suitable algorithms to generate tool path specifications 160 for one or more of these various systems to manufacture a part that has been designed using the systems and techniques described in this application.
[0048] Figure 2A An example of a process of converting a mesh geometry into a watertight boundary representation of an object to be manufactured is shown. A first 3D model of an object is obtained 205, for example, by a CAD program 116. The first 3D model includes a polygonal mesh associated with one or more modeled entities. The polygonal mesh may be an original or converted triangle mesh. Various types of mesh formats may be used as input, including tetrahedral meshes or hexahedral meshes. In addition, as described above, one or more modeled entities are in a boundary representation format (e.g., B-Rep) and may be regular geometry that has been adapted for a scanned mesh, or may be (multiple) modeled entities used as input to a generative design process that employs a volume mesh method or a level set method to create a generative design for the object. Thus, the polygonal mesh may be an output from one or more generative design processes that employ the one or more modeled entities as one or more subspaces of an optimization domain of the generative design process.
[0049] In some implementations, the input to the process includes an open general topological mesh M, where Constrained to be a set of smooth (analytical, process, spline) curves {C i These curves represent boundary edges on the residual B-Rep surface and may include intersections, trim curves, patch boundaries, etc. In some implementations, user input may be received 210, such as by the CAD program 116, to define one or more input parameters for the mesh to boundary representation conversion process.
[0050] For example, user input may be received 210 to define one or more smooth boundary curves (eg, curve {C i}), for example, the CAD program 116 may provide tools in the UI 122 to define smooth boundary curves. These smooth boundary curves may be obtained 215 based on this user input and represent boundary edges on the residual B-Rep surface. The boundary curves are smooth because they are mathematically specified to have continuous tangent directions at each point of the curve, in contrast to mesh boundaries, where the tangents are discontinuous at each vertex (usually).
[0051] In some implementations, one or more smooth boundary curves corresponding to one or more modeled entities are automatically obtained 215 without user input, such as by the CAD program 116 (e.g., the process proceeds from directly obtaining 205 the 3D model to obtaining 215 boundary curves and then potentially directly generating 230 a quadrilateral patch mesh). For example, known algorithms can be used to generate smooth boundary curves on the modeled entities from the mesh and the modeled entities. In some implementations, such as by the CAD program 116, the smooth boundary curves are generated by projecting the boundary of the mesh onto each corresponding B-Rep and interpolating the curves on the same face of the B-Rep using the sequence of projected points on that face. Thus, based on the received mesh, a set of one or more smooth curves is generated for the modeled entity.
[0052] In some implementations, user input may be received 210, such as by the CAD program 116, to specify a quadrilateral size, a quadrilateral shape, or both for use when generating 230 a quadrilateral patch network from the mesh. In these cases, the quadrilateral size, the quadrilateral shape (e.g., the angle variation of the quadrilaterals), or both may be set 220, such as by the CAD program 116, based on the CAD input. Additionally, in some implementations, user input may be received 210, such as by the CAD program 116, to specify a mesh internal tolerance value for use when defining 235 one or more locally refinable smooth surface representations using the quadrilateral patch network. In these cases, the mesh internal tolerance value may be set 225, such as by the CAD program 116, based on the user input.
[0053] A quadrilateral patch network is generated 230 from a polygonal mesh, for example, by the CAD program 116. This may include converting the received mesh to a triangular mesh, as well as additional operations, such as computing a global continuous quadrilateral parameterization map of the mesh, and blending the global continuous quadrilateral parameterization map with procedural quadrilateral patches that interpolate one or more smooth boundaries, as described in more detail below. The quadrilateral patch network combines the mesh with one or more smooth boundary curves corresponding to one or more modeled entities (e.g., using transfinite interpolation). The transfinite interpolation construct may define C curves that interpolate the curves at the boundaries. 0The patch network is internally coincident with the mesh. Thus, a globally continuous quadrilateral parameterized model of the mesh can be generated, where the mesh is divided by a four-parameter domain on the mesh.
[0054] One or more locally refinable smooth surface representations are defined 235, for example, by the CAD program 116 using a quadrilateral mesh as input and based at least in part on a minimum size representable by a geometric modeling kernel of a computer-aided design program (e.g., the CAD program 116). 0 The mesh network is converted to a T-NURCC surface, and the approximation algorithm is designed to produce very accurate boundary fits and accurately reproduce all boundary angles. These properties allow the B-Rep modeling kernel to stitch the output T-NURCC with adjacent B-Rep faces into a watertight, low-tolerance B-Rep CAD model, which means that the resulting geometry can be processed by downstream CAD modeling operations without the risk of separation of organic and non-organic surfaces. The conversion output may include a T-NURCC surface S, where Approximation i} to a very tight tolerance ξ>0, and using G 1 Boundary discontinuities reproduce all corners accurately. Therefore, the maximum tolerance for stitching the transformed surface to the residual B-Rep is <ξ. In addition, the algorithm can also limit the approximation error between M and S to a user-specified tolerance δ: δ≥ξ>0. In some implementations, the tolerance δ is related to the coarse geometric representation of the physical solver that synthesizes M. In any case, the tolerance ξ depends on the B-Rep model resolution, and in general, δ>>ξ.
[0055] Further details of definition 235 are provided below, but note that one or more locally refinable smooth surface representations may be T-NURCC, T-Splines, LR B-Splines, or Layered B-Splines. Many suitable variations of the geometry format may be used in various implementations. In any event, the boundary conforming mesh conversion method allows the tolerance to be made very small: if ∈>0 is the underlying B-Rep resolution, i.e., the minimum representable distance in the model, then the gap from the conversion to the corresponding B-Rep boundary may be made <10∈. In downstream modeling operations, this small tolerance is effectively equivalent to an exact model because the tolerance allows the computer-aided design program to treat the B-Rep as a valid watertight model, i.e., any suitable downstream operation is valid for the model.
[0056] Note that tolerance modeling provides operations to modify CAD models (also called tolerance models) where the gap between adjacent face surfaces is larger than the geometric kernel resolution ∈>0. This requirement means that these operations are less reliable and more difficult to implement than operations performed on the exact CAD model where the maximum gap is less than or equal to ∈. In addition, the reliability of tolerance modeling operations depends largely on the gap size: the larger the gap, the less likely the operation will produce valid results. Therefore, minimizing the gap size when generating the tolerance model is very important to ensure that the model can be used for downstream modeling. In practice, keeping the gap size <10∈ makes tolerance modeling operations as reliable as exact model operations. From the user's perspective, this means that a tolerance model with a maximum gap size <10∈ is as good as an exact model.
[0057] For example, one or more locally refinable smooth surface representations are combined 240 with one or more modeled entities by the CAD program 116 to form a second 3D model of the object, which is watertight at one or more smooth boundary curves. Please note that the second 3D model does not have to be a separate model, but can be an updated version of the first 3D model. The combination may include using a stitching operation to connect the one or more locally refinable smooth surface representations to the one or more modeled entities in the boundary representation format. Other operations, such as Boolean operations or cellular modeling operations, can be used, but stitching operations can be faster and simpler. In any case, the result after the combination 240 is a watertight 3D model, which is conducive to further processing because accurate mathematical surfaces are easier to process using CAD modeling operations (e.g., geometric modeling operations, such as holes, fillets or shells, and creating assembly relationships, such as mechanical joints), physical simulation operations, and CAM operations.
[0058] Figure 2B Shows Figure 2A An example of the conversion process applied to an organic mesh synthesized by topology optimization of a CAD solid model. Figure 1132. The initial version of the 3D model 132 includes the modeled solid 134 and the mesh 136. As shown in the enlarged view 260, the mesh boundary vertices 262 are constrained to trimming curves 264, which limit the residual trimmed solid B-Rep surface 134. Note the gap between the boundary mesh edge 262 and the trimming curve 264. T-NURRC surfaces 274 are constructed, which seamlessly approximate the trimming curve of the solid B-Rep surface 134 to a very tight tolerance and the interior of the mesh to a larger tolerance (e.g., ten, one hundred, or one thousand times the tight tolerance). As shown in the enlarged view 270, the T-NURRC surface 274 uses more patches 272 close to the trimming curve 264 to achieve the tight tolerance. In some implementations, the maximum boundary deviation is only 4.9x10 of the mesh bounding sphere diameter. -6 In addition, Figure 2B As shown, the control T mesh complexity (eg, 5287 patches) gradually decreases away from the boundary.
[0059] Therefore, if Figure 2B As shown, the conversion method can seamlessly approximate {C i} and M, and produces an adaptive structured control T-mesh that defines a surface that satisfies these different boundary and interior approximation requirements. In addition, a watertight solid B-rep model can be stitched together from the transformed output and the trimmed solid surfaces; as shown in the enlarged view 280, the resulting surface 282 intersects the trimming curve 264 in a watertight connection. In some implementations, the T-spline patch 284 is merged with the modeled solid 134 to produce a simpler final B-Rep surface layout. In some implementations, the maximum edge tolerance in the stitched B-Rep is 3∈, where ∈=10 -6 is the model resolution (smallest representable distance).
[0060] return Figure 2A, a watertight 3D model may be provided 250, for example, by the CAD program 116, for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems (e.g., the AM machine 170, the SM machine 174, and / or other manufacturing machines). Providing 250 may involve saving the watertight 3D model to a permanent storage device for use in manufacturing the physical structure corresponding to the object using one or more computer-controlled manufacturing systems (e.g., the AM machine 170 and / or the SM machine 174). In some implementations, providing 250 involves generating 252 a tool path specification for a computer-controlled manufacturing system (e.g., the AM machine 170 and / or the SM machine 174) using the watertight 3D model, for example, by the CAD program 116, and manufacturing 254 at least a portion of the physical structure corresponding to the object using the tool path specification generated for the additive manufacturing machine (e.g., the AM machine 170 and / or the SM machine 174) using the computer-controlled manufacturing system (e.g., the AM machine 170 and / or the SM machine 174), for example, by the CAD program 116.
[0061] Thus, a boundary conforming mesh conversion method can be designed to meet the requirements of CAD workflows including generative design and topology optimization CAD workflows. As described in more detail below, this goal can be achieved using a transfinite interpolation construction that seamlessly combines the input mesh with the B-Rep boundary curves. Note that a simple but flawed solution is to replace each triangle with a boundary edge with a triangular patch that interpolates the corresponding boundary curve segment and the internal mesh edge. Figure 3B is a graphical diagram showing an example of input for interpolation.
[0062] As shown, interpolating a smooth curve 342 using triangle patches 340 that replace input triangles adjacent to the boundary affects the surface (grey) with a bumpy tessellation-dependent shape. This results in a region of bumpy tessellation offset where {C i} affects the output surface in a limited way; in fact, only the region of triangles adjacent to the boundary is affected by the boundary curve. Instead, the global quadrilateral graph of M is computed and the image of the graph is interpolated with {C i} process quad patch blending aligns the area affected by the interpolated smooth curve with the boundary shape. Figure 3B As shown, the transfinite interpolation of the smooth curve 342 using the procedural quadrilateral patch 344 results in a surface region (light grey) of the boundary shape that is affected by the interpolated curve 342 and the affected region is independent of the mesh tessellation shown by the dashed line. Note that all the procedural quadrilateral patches 344 are adjacent to the smooth curve 342.
[0063] Furthermore, the quadrilateral parameterization of the mesh is not a replacement for the mesh, and can have different shapes (e.g., more rectangular or more square, or more angled (anisotropic)) and sizes of the quadrilaterals, thereby changing the area of influence of the smooth curve on the resulting smooth surface. The quadrilateral parameterization of the mesh is used to transfinitely interpolate the smooth curve, but the quadrilaterals are procedural, not actual 3D geometric quadrilaterals. Rather, the quad patch 344 is a parameter domain on the mesh. The quad patch 344 associates the mesh with the smooth surface to be generated, and maintains the mesh as the source of the generated geometry.
[0064] Figure 3A is a graphical diagram showing an overview of an example of a process for generating a quadrilateral patch network and using the quadrilateral patch network to define one or more locally refinable smooth surface representations. The process includes two main stages. In the first stage, C 0 Quad Patch Network The network combines the mesh M with the B-Rep boundary curve {C i}Combines seamlessly. Figure 3A The input data and the transfinite interpolation method are shown in row 300. The integer grid map (IGM) φ can be calculated by: To extract The quadrilateral patch layout of -1 , you can Mapped to input curve {C i}, and the displacement field can be defined Add these patches to M = φ -1 (Ω) Get C 0 Surface The surface is Interpolation i} and is internally consistent with M.
[0065] In the second stage, the T-NURCC surface S(Ω) can be defined and the approximation problem can be solved and Where δ≥ξ>0. Figure 3A The flow of the approximation phase is shown in row 310 of FIG. , which can first be solved using iterations of least squares optimization and error-driven adaptive refinement. Refinement may also include a balancing component that propagates new knots to the interior of the surface when necessary to ensure overall process stability. Once boundary conditions are met, the boundary control points are frozen and a similar process is carried out internally.
[0066] Note that the overall tolerance δ>L is used ∞(S,M) to limit the approximation error is not feasible in typical CAD environments, because even for moderately complex instances of CAD models, e.g. Figure 2B The CAD model shown is 2.8×10 5 This cannot meet the accuracy requirements of typical CAD applications without introducing a large number of facets. Therefore, the concept of discrete boundary error bound ξ: δ≥ξ>0 is introduced. Since in practice δ>>ξ, the proposed technique can be used to solve nonlinear problems. and L ∞ (S,M)<δ while balancing surface refinement to maintain stability. This strategy achieves the desired boundary accuracy (e.g., Figure 2B The converted surface of the CAD model shown in consists of only 5287 patches.) In addition, several modifications to the mesh parameterization at the boundaries are proposed to more efficiently represent corners (G1 discontinuities) between the input curves.
[0067] Finally, as mentioned above, user input can be used to influence the quad parameterization. Figure 3C is a graphical diagram showing an example of the effect of a graph size field on the construction of a transfinite interpolation. In this example, the input to the process is a dome-shaped mesh 360 having four boundary edges constrained to smooth curves 362. Figure 3C Also shown in are three different transformation outputs, where the target quadrilateral size is halved for each subsequent instance. The first instance 372 shows the result produced using a first (large) quadrilateral size, the second instance 374 shows the result produced using a second (medium) quadrilateral size, and the third instance 376 shows the result produced using a third (small) quadrilateral size. As shown, the quadrilateral parameterization produces a quadrilateral shape that has curved edges and interpolates a smooth curve 362 at the boundary and a mesh 360 inside. Also as shown, the surface area affected by the curve 362 gradually decreases but remains aligned with the curve shape (note that error-driven refinement of the surface domain is disabled for these transformation instances to illustrate the transfinite interpolation construction isolated from the rest of the process). Therefore, the extent of the curve influence can be controlled by a user-specified size field that is independent of the mesh tessellation, and the final result is still a seamless, continuous surface that combines the curve 362 and mesh 360 geometry.
[0068] Figure 4A is an example showing the process of generating a quadrilateral patch network (e.g., Figure 2A230 ). In general, a continuous model is defined, wherein the continuous model combines curves and meshes; the transfinite interpolation construction of the patch network enables the definition of seamless continuous surfaces combining meshes with smooth curves, which facilitates seamless integration of the meshes with CAD models in a watertight manner that easily supports downstream 3D modeling operations. Automatically converting meshes into smooth surfaces that can be integrated in watertight, low-tolerance B-Rep models is a key enabler of generative design and topology optimization CAD workflows.
[0069] A quadrilateral patch layout is extracted 405 from a polygonal mesh (e.g., a triangle mesh). The quadrilateral patch layout represents a quadrilateral partitioning of a parameter domain Ω of a global continuous graph φ:M→Ω of the mesh M. In some implementations, the input includes an open general topology, an orientable 2-manifold mesh M=({V i},{E i},{T i}) and C assigned to the border edge 1 Curve segment {C i (t):t∈[t i ,0,t i ,1]}. {V i},{E i},{T i} are mesh vertices, edges, and triangles. Without loss of generality, we can assume that {T i} is oriented counterclockwise. is constrained to the curve segment, that is, for each E i =V i,0 , We all have C i (t i,0 )=V i,0 and C i (t i,1 )=V i,1 .
[0070] In some implementations, the graph φ is computed by applying an integer grid map (IGM) method, as described in D. Bommes, M. Campen, H.-C. Ebke, P. Alliez, L. Kobbelt, Integer-grid maps for reliable quad meshing (ACM Transactions on Graphics 32 (2013)), and efficiently quantized as proposed in M. Campen, D. Bommes, L. Kobbelt, Quantized global parametrization (ACM Transactions on Graphics 34 (2015)). The graph image Ω can be divided into unit quadrilateral layouts Q i :Ψ i +[0,1] 2 ,Ψ i ∈Z 2 ,{Q i |∪Q i =Ω}Ψ i , and 1 Represent the vertices and edges of the quadrilateral layout respectively. IGM inverse mapping φ -1 :Ω→M plays an important role in the construction: given Q i and (u,v)∈[0,1] 2 ,φ -1 (Q i ,u,v) to retrieve the corresponding points on M. By i Store a set of overlapping triangles Θ i ={T j |dim(φ(T j )∩Q i )=2}, by positioning T j ∈Θ i : It is very effective in evaluating φ -1 .
[0071] In some implementations, the process includes an additional step to resolve boundary singularities. Any negative boundary cross-field singularities of the mesh may be identified 410. This may occur during extraction 405, since at least some quadrilateral parameterization is required to identify negative boundary cross-field singularities. For example, IGM parameterization includes several stages, and after the direction field and cross-field establishment stages, the locations of parameterization singularities have already been identified. Therefore, it is not necessary to calculate a complete parameterization before adjusting the singularities. If present, one or more identified negative boundary cross-field singularities may be relocated 415 to interior vertices of the mesh. More details on this process will be provided in conjunction with Figure 5B and Figure 5C This is described below.
[0072] In any case, at least one boundary of the quadrilateral patch layout is mapped 420 to one or more smooth boundary curves corresponding to one or more modeled entities. For example, since φ: Through the Synthesize the curve Γ to achieve mapping make in And E 0 (t A )=φ -1 (Ψ A ),E N (t B )=φ -1 (Ψ B ),t A ∈[t 0,0 ,t 0,1 ),t B ∈(t N,0 , t N,1 ]. 0 [t A ,t 0,1 ],C 1 [t 1,0 ,t 1,1 ],…,C N [t N,0 ,t B ] are joined together to obtain the geometry Γ. In order to complete the mapping Υ→Γ and avoid relying on {C i}, Γ can be parameterized by normalizing its arc length to [0,1]. In cases where numerical integration is required to evaluate, for example, the arc length function of a spline In the case of , s can be recalculated using the techniques described in the following items iExact B-spline approximations: RJ Sharpe, RW Thorne, Numerical method for extracting an arc length parameterization from parametric curves (Computer-Aided Design 14(2)(1982)79–81), and JW Peterson, Arc Length Parameterization of Spline Curves (Journal of Computer-Aided Design 14(2)(2006)).
[0073] The surface patch is constructed 425 using a globally continuous patch of the polygonal mesh, a mapping of at least one boundary of the quadrilateral patch layout to one or more smooth boundary curves, and transfinite interpolation of the one or more smooth boundary curves. Figure 4B is a graphical diagram illustrating an example 450 of a transfinite patch construction. For each Constructing patches The patch will input the mesh geometry φ -1 (Q i ) and assigned to Q i The composite curve {Γ j}Mix. Let be the IGM edges sorted counterclockwise along the contour lines v=0,u=1,v=1,u=0. Also let if but Defining a Vector Field Curve It means that from φ -1 (Υ j ) to Γ j The displacement is as follows:
[0074]
[0075] Otherwise, define Then, a displacement Coons patch is constructed (see SACoons, Surfaces for Computer-Aided Design of Space Forms, Tech. rep. (1967)). i (u,v):
[0076] Ξ i (u,v)=(1-v)·μ 0 (u)+u·μ 1 (v)+v·μ 2 (u)+(1-u)·μ 3 (v)
[0077] Finally, construct the transfinite interpolation patch:
[0078]
[0079] By the dough Composition, if but C 0 Continuity is directly derived from φ -1 C 0 Continuity and {Γ i} concluded.
[0080] Figure 5A is an example showing a process for defining one or more locally refinable smooth surface representations using a quadrilateral patch network (e.g., Figure 2A For example, this second stage of the process can approximate a patch network with a T-NURCC surface S(Ω) One or more initial locally refinable smooth surface representations (eg, T-NURCC, T-spline, LR B-spline, or layered B-spline) are constructed 505 based on a quadrilateral patch network.
[0081] For example, the above IGM quadrilateral can be divided into {Q i} Extract the initial quadrilateral control mesh for the output surface S. Since all IGM singularities are located at the input mesh vertices, no local optimization is required to accommodate "in-triangle" singularities. The initial quadrilateral layout of S is completely uniform, so for each quadrilateral edge Υ i Set the knot interval value d i = 1. Since S is refined using local edge and quadrilateral partitioning (see T-spline simplification and local refinement by TWSederberg, DLCardon, GTFinnigan, NSNorth, J.Zheng, T.Lyche, ACM SIG-GRAPH 2004 paper on SIGGRAPH'04, 2004, p.276), the knot spacing values will be updated accordingly. Split Y i→ j ,Υ k} Create a new knot interval d j +d k =d i . Accordingly, Q is divided into four equal parts i →{Q i,j,k} j,k=0,1 Split four side knot interval d i,u and d i,v , which is equal to the sum of the edge parameter intervals along the u and v sides.
[0082] In the following description, {Q i} to mark the quadrilateral domain partition of the refined T-NURCC surface, and use To mark the initial quadrilateral partition. Initially, each face Corresponding to the patch When When refining, use parametric translation and scale diagrams To keep from each Therefore, In the subsequent description of this article, this detail is omitted and the
[0083] To control The approximation at can be obtained by applying The standard way to set up Bézier boundary conditions is to add a 0 channel for quad faces. Figure 5B is a graphical diagram showing the four boundary conditions. Figure 5B In the figure, the solid line represents the quadrilateral domain {Q i}, the dotted line indicates the edge of channel 0. Point 550 indicates the control The larger point 550 represents the boundary control point. 1 Discontinuity. Point 552 indicates that the control is along The control points of the tangent plane.
[0084] Additional vertices are added along the boundary, if necessary, to interpolate the boundary first-order derivatives, and the additional vertices accurately represent the input boundary angles. At regular boundary vertices and index 1 singularities (see regular vertex instance 560 and convex vertex instance 570 corresponding to index 1 singularities), this process is equivalent to inserting boundary knots in the tensor product NURBS surface.
[0085] In addition, using local knot refinement, G can be created on regular vertices. 1 Discontinuity (see the representation with C 1The local node insertion can be used to create a boundary angle that defines the boundary angle along the discontinuity. The boundary control points of two adjacent first derivatives of . T-junctions can be replicated on internal 0-channel edges to allow for tangent plane control.
[0086] However, concave boundary vertices on negative index singularities are significantly more complicated to handle, as discussed in Appendix A of J. Shen, J. Kosinka, M. Sabin, N. Dodgson, Converting a CAD model into a non-uniform subdivision surface (Computer Aided Geometric Design 48 (C) (2016) 17–35). If C 1 The internal propagation method of discontinuities is used to represent concave boundary vertices (see Example 575 for concave corner vertices corresponding to index-1 singularities). Then, in solving the internal minimum C 1 L should be added when the problem 2 Constraints (as described in detail below in conjunction with the modification 520 operation). As with example 565, local knot insertion may be applied to provide control over the first-order derivatives in example 575. However, in this case, the added boundary knots are propagated 577 into the surface to prevent the internal basis functions from affecting This configuration can also be applied to higher priced border vertices.
[0087] However, in some implementations, an alternative approach may be used for concave corner vertices. By limiting the cross-field index of all non-corner boundary vertices to 0 when computing the mesh cross-field, the G of the curve segment can be prevented from being 0 Boundary singularities anywhere except at joints. Therefore, is C far away from any corner point 2 , which reduces the surface complexity because C 1 There are no tangent control points on the joint.
[0088] In order to avoid C originating from the concave edge corner 1Internal discontinuities can be represented by adjacent internal singularities using a strategy similar to that described in Watertighttrimmed NURBS by TW Sederberg, G.T. Finnigan, X. Li, H. Lin, H. Ipson, ACM SIGGRAPH 2008 paper (ACM Press (New York, New York, USA, 2008, p. 1) vol. 27) to rewrite such corners. In practice, this is done with Figure 5B The type in instance 565 of Figure 5B Boundary conditions of the type in the example 575 in FIG. Figure 3A and Figure 4A During (or after) the computation of the IGM discussed above, negative boundary cross-field singularities can be relocated to nearby interior vertices as described below.
[0089] Figure 5C 5 is a graphical diagram showing an example of boundary singularity relocation. A negative boundary singularity 594 (singularity index is -1) is shown on dashed line 592, which represents a local quadrilateral partition overlaid on mesh 590. As shown, negative boundary singularity 594 is placed at a large interior angle α>π (concave) boundary vertex V i Relocate the singular point 594A to the adjacent internal vertex V j , so that the connection V i 、V j The edge of is as close to the bisector of α as possible. This results in the replacement of the negative singular vertex 594 at the boundary by the regular vertex 596.
[0090] make is the subset of the input mesh boundary vertices assigned to negative index singularities. Figure 5C As shown, select The corresponding internal vertex Then, for all pairs The cross-field indicators are swapped and the cross-fields are updated. During subsequent IGM recalculations, this can be enforced by parameter quantization. Apply local knot insertion to regular boundary control mesh vertices {φ(V i )} to set Figure 5B The boundary conditions in Example 565 allow The corresponding angle at is represented.
[0091] This technique has some limitations that are worth noting. The nearby input mesh may be too coarse, so suitable relocation vertices may not exist or are too far from the bisector of the interior angle α. This problem can be solved by locally optimizing the input mesh to introduce more interior vertices. Alternatively, if the accumulated interior angle is too large, e.g. α>2π, and the singularity index <-1, relocation can introduce large parameter distortions, which can lead to poor surface quality. In this case, it is better to accept the interior C 1 Discontinuity.
[0092] return Figure 5A , one or more initial locally refinable smooth surface representations are refined 510 at one or more boundaries of the one or more initial locally refinable smooth surface representations (e.g., iteratively approximating and refining boundaries of the refinable smooth surface until curves on the B-rep are approximated by the boundaries within a tolerance that the CAD program can consider these curves and boundaries to be coincident). This refinement causes the smooth surface representation to approximate one or more smooth boundary curves within a first tolerance value (boundary tolerance value). The boundary approximation tolerance value controls the deviation from the boundary during conversion to the boundary representation format. Therefore, the boundary approximation tolerance should be set to a sufficiently small value to ensure that the computer-aided design program (e.g., CAD program 116) will consider the smooth surface representation to satisfy the modeled entity at its smooth boundary curves.
[0093] The boundary approximation tolerance value may be set to be equal to or greater than the minimum size representable by the geometric modeling kernel of the computer-aided design program, and may also be within the tolerance range supported by the geometric modeling kernel of the computer-aided design program, e.g., small enough to allow the geometric modeling kernel of the CAD program 116 to consider curves within the tolerance range to be coincident. For example, the boundary tolerance may be set to a value that is always within 100 times (usually less than) the resolution of the geometric kernel to maintain acceptable performance, and this tolerance value is small enough to allow the geometric kernel to consider the surface boundary to be completely on the B-rep. Note that setting the boundary approximation tolerance to a value less than the resolution of the geometric modeling kernel increases complexity without gaining additional useful accuracy. In addition, the boundary approximation tolerance value may be set to be substantially higher than the minimum resolution (size) of the geometric modeling kernel while still achieving the purpose of ensuring surface connectivity. Setting a larger value provides a less accurate model, but results in a more compact representation and shorter run time. In some implementations, ξ is set to 10 or even 100 times the model resolution, because the resulting B-Rep model is still sufficiently accurate and much more accurate than can be achieved using existing methods. In some implementations, the upper limit of the tolerance range is as high as 1000 times the kernel resolution, because tolerance values within this range can still be small enough for the geometric kernel to consider the curves as coincident.
[0094] In some implementations, once S(Ω) is constructed, The solution includes the following iterative boundary approximation operations: solving the minimum L 2 , estimate L ∞ and refine where appropriate. When the process converges. For the boundary L 2 Fitting, please note, By C 2 NURBS curves, except at corners. The discretization can be performed as follows:
[0095]
[0096] {(w k ,t k )} is the approximation for each In the interval [0,d Υ ] on the integral L 2 The order of the formulas is based on the continuity of the corresponding composite curve Γ. Standard least squares system techniques can be applied to solve along The minimum control point question.
[0097] For the boundary angle, let Ψ be expressed as Figure 5B C shown in Examples 556, 570, 575 1 The vertices of the boundary quadrilateral of the discontinuity. is the corresponding input mesh corner vertex. By considering the position of V and the first-order derivative Γ′ i (d i and Γ′ i (0) Adding interpolation constraints to the least squares system can Constrained in Ψ, where (Υ i ,Υ j ) are the edges of its neighboring quadrilaterals. These constraints force Accurately represent the boundary shape at V.
[0098] Regarding the boundary error, for each quadrilateral edge Nonlinear numerical optimization can be applied to calculate If the interval subdivision is restricted to bisection and the binary parameter sample is shared with the Newton-Cotes quadrature formula (described above), then can be effectively evaluated. However, due to depends on φ and thus may lead to unnecessary refinement when parameter error bounds are not needed. A useful but more expensive alternative is to compute the Hausdorff distance between corresponding boundary loop curves. because This geometric metric may therefore lead to reduced refinement.
[0099] Figure 5D is a graphical diagram showing an example of local knot insertion during boundary refinement. For boundary refinement, all It can be equally divided by inserting local knots at intermediate parameters. Figure 5D As shown in 530 in FIG. 1 , the interior unit quadrilateral is adjacent to the boundary 0-channel quadrilateral with equally divided boundary edges. This selection does not require L 2 and L ∞ Discretization allows integration because the samples evaluated in these calculations can be reused. To prevent highly non-uniform knot sequences that could cause surface quality issues, the boundaries are balanced by dividing all edges equally according to
[0100]
[0101] The boundary edge split will be replicated on the corresponding channel 0 interior edge, thus dividing the incident channel 0 quadrilateral into equal parts. Figure 5D As shown in 532 in FIG. 5 , the division on the internal 0-channel edge is replicated, thereby dividing the 0-channel quadrilateral into equal parts. The additional vertex controls the first-order derivative across the boundary and can be represented by G 1 Boundary conditions are used. As the boundaries are refined, configurations can be created where small supporting boundary basis functions overlap with larger interior basis functions. To avoid the potential adverse effects of this non-uniform parameterization on surface quality, any quadrilateral {Q i} Quarter (if d i >2d j ) and adaptively propagates the boundary refinement to the interior of the surface, where d i is the length of the 0-channel quadrilateral side junction, d j is the junction length of the shortest 0-channel edge on the side of this quadrilateral. Figure 5D As shown in 534 in FIG. 1 , because bisecting one (or two) smaller 0-channel quadrilaterals during subsequent refinement may cause the configuration to become unbalanced, the inner quadrilateral may be bisected to rebalance the layout. Figure 5E As discussed in more detail, further refinement of the internals may be required to rebalance the parameterization.
[0102] return Figure 5A , freezing 515 the control points of one or more locally refinable smooth surface representations at one or more boundaries. Fixing (freezing) the control points and restricting all modifications to the interior of S ensures that the resulting surface will not break contact established with the modeled solid and the smooth curve on it. This is true even though much higher tolerance values can be used from this point on when approximating the interior of the mesh, e.g., iteratively solving for the minimum L 2 , estimate L ∞ And refine until until.
[0103] The remaining interior portions of the one or more locally refinable smooth surface representations are modified 520 to approximate the mesh. This modification causes the smooth surface representation to approximate the mesh within a second tolerance value (the interior tolerance value). The interior approximation tolerance value controls the deviation from the mesh during conversion to the boundary representation format. Thus, the interior approximation tolerance can be set much larger than the boundary approximation tolerance, and can even be an input parameter, e.g., an input to the CAD program 116, thereby providing user control over how closely the new surface should match the mesh.
[0104] The internal approximation tolerance value may be set equal to or greater than the accuracy of the method used to obtain the mesh. For example, when the mesh is obtained using one or more generative design processes, the internal approximation tolerance value may be set based on the generative design representation accuracy, e.g., greater than or equal to the generative design representation accuracy and also small enough to preserve relevant features of the mesh. As another example, when the mesh is obtained by scanning (e.g., laser scanning), the internal approximation tolerance value may be set based on the accuracy of the scanning device, e.g., greater than or equal to the representation accuracy of the scanning device.
[0105] Note that setting the internal approximation tolerance to a value less than the accuracy of the method used to obtain the mesh geometry increases complexity without achieving significant benefits. Additionally, values much larger than those used for the boundaries may be used. In some implementations, the second (inner) approximation tolerance is set to a value that is at least one order of magnitude (10 times) of the first (boundary) tolerance value. In some implementations, the second (inner) approximation tolerance is set to a value that is at least two orders of magnitude (100 times) of the first (boundary) tolerance value. In some implementations, the second (inner) approximation tolerance is set to a value that is at least three orders of magnitude (1000 times) of the first (boundary) tolerance value.
[0106] For internal L 2 Fitting, let d Q,u and d Q,v is the parameter space interval along the u and v directions of the quadrilateral Q, and We can use Newton-Cotes to find the volume and change L 2 Metric discretization:
[0107]
[0108] {(u k ,v l )} defines a uniform sample grid on Q, and {w k,l} is the corresponding volume weight. The interior is C 0 , so the trapezoidal rule can be used for a 5x5 grid.
[0109] Figure 5E is a graphical diagram showing an example of a quadrilateral domain sample grid, nesting, and refinement. As shown at 540, the L 2 The discretized samples of the measurement. Each sample is composed of The edge and vertex samples are reused for each incident quadrilateral. Except that the control point positions are fixed and no fairing term needs to be added to the energy, the least squares system is composed identically to that in Li, Wan-Chiu, Ray, Nicolas and Lévy, Bruno, "Automatic and Interactive Mesh to T-spline Conversion", in Proceedings of the 4th European Symposium on Geometry Processing in Graphics (2006), pp. 191-200.
[0110] If there are any internal C propagated from the boundary condition settings 1 discontinuities (as discussed above in connection with paragraphs
[0076] to
[0079] ), derivative interpolation constraints should be added across these discontinuities to achieve C 1 Internal continuity These constraints can be applied to Figure 5B Internal control point 559 is shown.
[0111] Although multivariable optimization can be used to accurately calculate the parameter errors But the computational cost is unacceptable in many applications. As an alternative, Figure 5E As shown in 542 in FIG. 5 , the current uniform binary grid of distance samples can be refined, and can be calculated on a finer 9×9 grid By doubling the sample grid resolution in both parameter directions, as shown in 542, when evaluating L ∞ This will nest the refinement samples within the existing grid when errors are present.
[0112] Then quartering can be applied, e.g. Figure 5E 544 in order to refine all the quadrilaterals by dividing them into four equal parts. And the balance parameterization
[0113]
[0114] If the quad is subsequently refined, a finer quad mesh sample is immediately available and only the basis function combination needs to be updated to account for the added control points. The local edge segmentation used may already exist on the boundary, or the boundary edges may need to be segmented. In the latter case, any added boundary control points can be used to solve the minimum solution again along all affected boundaries. (discussed above in paragraph
[0087] ) to further improve This is done after the inner iterations are complete.
[0115] In practice, improved output surface quality is observed if more quadrilaterals are refined as necessary to maintain the "standard" T-spline basic properties, as described in T-splines and T-NURCCs by T. W. Sederberg, J. Zheng, A. Bakenov, A. Nasri, ACM Transactions on Graphics (Volume 22, 2003, page 477). The method described in Local refinement of analysis-suitable T-splines by MA Scott, X. Li, T. W. Sederberg, T. J. Hughes, Computer Methods in Applied Mechanics and Engineering 213-216 (2012) 206–222 can keep the number of such quadrilaterals low and prevent excessive output complexity. In some implementations, a greedy approach can be used, which may occasionally result in excessive internal refinement.
[0116] FIG. 6A to FIG. 6C Examples of results of the process applied to organic meshes produced by topology optimization and generative design applications are shown. Fig. 6AAs shown, the generative design produces three organic meshes 600, 602, 604. All mesh boundaries are constrained to process trim curves (e.g., curves 606, 608 shown in the zoomed-in area of the 3D model), which are constructed to lie exactly on existing non-organic surfaces 610, 612, 614, 616, 618, 620, 622 in the optimized B-Rep model. In this example, ξ is set to 100×B-Rep resolution. In order for the tolerance value to make sense in the context of the conversion example, the value of ξ is given relative to the bounding sphere diameter of the input mesh (e.g., ξ=3x10 -6 ). In this example, δ = 100ξ.
[0117] Figure 6B The resulting T-NURCC surfaces 630, 632, 634 generated for the organic meshes 600, 602, 604 are shown. As shown in the zoomed-in area of the 3D model, as the surface approaches the procedural trim curves 606, 608, the quadrilateral patches become more detailed 636, 638. Additionally, Figure 6C is a graph showing an approximation metric at iterations indicated on the horizontal axis. Graphs 650, 652, 654 track the error metric and The error values are shown on the left vertical axis. Since the boundary metrics are measured only for boundary iterations and Therefore these graphs may not cover the entire horizontal range. In addition, the curve graph 656 tracks the number of output patches after each iteration | {Q i}|, and the values are shown on the right vertical axis. In this example, the final number of output patches |{Q i}| is 34,532 and the running time is 42 minutes (the running time here includes IGM calculation).
[0118] Figure 7 7 is a schematic diagram of a data processing system including a data processing device 700, which can be programmed as a client or a server. The data processing device 700 is connected to one or more computers 790 via a network 780. Figure 7 Only one computer is shown as data processing device 700, and multiple computers can be used. Data processing device 700 includes various software modules that can be distributed between the application layer and the operating system. These software modules may include executable and / or interpretable software programs or libraries, including tools and services of one or more 3D modeling programs 704 that implement the systems and techniques described herein. Therefore, (multiple) 3D modeling programs 704 can be (multiple) CAD programs 704 that implement 3D modeling functions and conversion of mesh geometry to watertight boundary representations of parts to be manufactured.
[0119] In addition, program(s) 704 may implement physical simulation operations (finite element analysis (FEA) or other), generative design operations (e.g., using level set-based methods for generative design), and / or manufacturing control operations (e.g., generating and / or applying tool path specifications to affect the manufacture of a design object). In some implementations, program(s) 704 are CAD program(s) 704 with generative design functionality that uses conversion of mesh output to editable, watertight B-Reps. The number of software modules used may vary from one implementation to another. In addition, the software modules may be distributed across one or more data processing devices connected via one or more computer networks or other suitable communications networks.
[0120] The mesh geometry conversion function of the 3D modeling program(s) 704 (and the similar function of the CAD program(s) 116) can be designed to operate on boundary constrained meshes and approximate smooth boundary curve constraints very closely. This can be achieved by using transfinite interpolation to reduce the total number of quadrilateral patches required to accurately match the smooth curve boundary to the modeled entity. The mesh parameterization method used can be IGM, which adds new control points to all patches that violate the error bound at the same time, and does not require adding new control points to the L mesh. 2 Energy smoothing terms are added. These choices complement well the components that conform to the boundaries and do not affect the output surface quality internally. Furthermore, the described process converges and performs stably even for very detailed and complex organic meshes synthesized through topology optimization and generative design processes. Since the converted surface boundaries conform to the boundary edges of the trimmed residual B-Rep, a watertight CAD model with very tight tolerances can be produced to meet the requirements of downstream modeling operations.
[0121] The data processing device 700 also includes hardware or firmware devices, which include one or more processors 712, one or more additional devices 714, a computer-readable medium 716, a communication interface 718, and one or more user interface devices 720. Each processor 712 is capable of processing instructions for execution within the data processing device 700. In some implementations, the processor 712 is a single-threaded or multi-threaded processor. Each processor 712 is capable of processing instructions stored on a computer-readable medium 716 or stored on a storage device such as one of the additional devices 714. The data processing device 700 uses a communication interface 718 to communicate with one or more computers 790, for example, via a network 780. Examples of user interface devices 720 include displays, cameras, speakers, microphones, tactile feedback devices, keyboards, mice, and VR and / or AR devices. The data processing device 700 may store instructions for implementing operations associated with the above-mentioned programs, for example, on a computer-readable medium 716 or one or more additional devices 714 (e.g., one or more of a hard disk device, an optical disk device, a tape device, and a solid-state storage device).
[0122] The embodiments of the subject matter and functional operations described in this specification can be implemented with digital electronic circuits or with computer software, firmware or hardware (including the structures disclosed in this specification and their equivalent structures) or with any combination of one or more of the foregoing. The embodiments of the subject matter described in this specification can be implemented using one or more modules of computer program instructions, which are encoded on a non-transitory computer-readable medium for data processing equipment to perform or control the operation of data processing equipment. Computer-readable media can be manufactured products, such as hard disk drives in computer systems, or optical disks sold through retail channels, or embedded systems. Computer-readable media can be obtained separately and then encoded with one or more modules of computer program instructions, for example, after transmitting one or more modules of computer program instructions through a wired or wireless network. Computer-readable media can be machine-readable storage devices, machine-readable storage substrates, memory devices, or a combination of one or more of them.
[0123] The term "data processing apparatus" encompasses all devices, means and machines for processing data, including, for example, a programmable processor, a computer or a plurality of processors or computers. In addition to hardware, the apparatus may also include code that generates an execution environment for an associated computer program, for example, code constituting processor firmware, a protocol stack, a database management system, an operating system, a runtime environment or a combination of one or more of them. In addition, the apparatus may employ a variety of different computing model infrastructures, such as World Wide Web services, distributed computing and grid computing infrastructures.
[0124] A computer program (also referred to as a program, software, software application, script, or code) may be written in any suitable form of a programming language (including compiled or interpreted languages, declarative or procedural languages), and the computer program may be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for a computing environment. A computer program does not necessarily correspond to a file in a file system. A program may be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to a related program, or in multiple collaborative files (e.g., files storing one or more modules, subroutines, or code portions). A computer program may be deployed to execute on one computer or on multiple computers located at one location or distributed at multiple locations and interconnected by a communication network.
[0125] The processes and logic flows described in this specification may be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows may also be performed by special purpose logic circuits, and the apparatus may also be implemented as special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application specific integrated circuits).
[0126] Processors suitable for executing computer programs include, for example, general-purpose microprocessors and special-purpose microprocessors, and any one or more processors of any kind of digital computer. Typically, the processor will receive instructions and data from a read-only memory or a random access memory or both. The basic elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more large-capacity storage devices for storing data, such as a disk, a magneto-optical disk, or an optical disk, or be operably coupled to receive data from the large-capacity storage device or transfer data to the large-capacity storage device, or perform these two operations. However, a computer does not necessarily have such a device. In addition, a computer may be embedded in another device such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device (e.g., a universal serial bus (USB) flash drive), to name a few. Devices suitable for storing computer program instructions and data include all forms of nonvolatile memory, media and storage devices, including, for example, semiconductor memory devices, such as EPROM (erasable programmable read-only memory), EEPROM (electrically erasable programmable read-only memory), and flash memory devices; magnetic disks, such as internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The processor and memory may be supplemented by, or incorporated in, special purpose logic circuits.
[0127] To provide interaction with a user, embodiments of the subject matter described in this specification may be implemented on a computer having: a display device for displaying information to a user, such as an LCD (liquid crystal display) display device, an OLED (organic light emitting diode) display device, or another monitor, and a keyboard and pointing device (e.g., a mouse or trackball) through which a user can provide input to the computer. Other types of devices may also be used to provide interaction with a user; for example, feedback provided to a user may be any suitable form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback; and input from a user may be received in any suitable form, including sound, voice, or tactile input.
[0128] A computing system may include a client and a server. The client and the computer are usually far away from each other and usually interact through a communication network. The relationship between the client and the server is generated by means of a computer program that runs on a corresponding computer and has a client-server relationship between each other. The embodiments of the subject matter described in this specification may be implemented in a computing system, the computing system including, for example, a back-end component as a data server, or including, for example, a middleware component of an application server, or including a front-end component (for example, a client computer with a graphical user interface or a browser user interface, a user can interact with the implementation of the subject matter described in this specification through a graphical user interface or a browser user interface), or including any combination of one or more such back-end components, middleware components, or front-end components. The components of the system can be interconnected by digital data communication (for example, a communication network) of any suitable form or medium. Examples of communication networks include local area networks ("LAN") and wide area networks ("WAN"), internetworks (for example, the Internet), and peer-to-peer networks (for example, dedicated peer-to-peer networks).
[0129] Although this specification contains many implementation details, these details should not be interpreted as limitations on the scope of what is claimed or can be claimed, but should be understood as descriptions of features that are peculiar to a particular embodiment of the disclosed subject matter. Certain features described in this specification with a separate embodiment as the background can also be implemented in a single embodiment in combination. On the contrary, each feature described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination. In addition, although the features are described above as acting in certain combinations and even initially required, one or more features from the required combination can be separated from the combination in some cases, and the required combination can be directed to a sub-combination or a change in the sub-combination.
[0130] Similarly, when operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring the operations to be performed in the particular order shown or in a sequential order, or to perform all of the illustrated operations to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.
[0131] Thus, certain embodiments of the invention have been described. Other embodiments are within the scope of the following claims. Additionally, the actions recited in the claims can be performed in a different order and still achieve desirable results.
Claims
1. A method for converting a mesh geometry to a watertight boundary representation, the method comprising: obtaining a first three-dimensional model of an object to be manufactured in a computer-aided design program, wherein the first three-dimensional model includes a polygonal mesh associated with one or more modeled entities, the one or more modeled entities being in a boundary representation format; generating, by the computer-aided design program, a quadrilateral patch network from the polygonal mesh, the quadrilateral patch network combining the polygonal mesh with one or more smooth boundary curves corresponding to the one or more modeled entities using a transfinite interpolation method; defining, by the computer-aided design program, one or more locally refinable smooth surface representations using the quadrilateral patch network as input and based at least in part on a minimum size representable by a geometric modeling kernel of the computer-aided design program; as well as The one or more locally refinable smooth surface representations are combined with the one or more modeled solids by the computer-aided design program to form a second three-dimensional model of the object, the second three-dimensional model being watertight at the one or more smooth boundary curves.
2. The method of claim 1, wherein the polygonal mesh is generated by a generative design process that employs the one or more modeled entities as one or more subspaces of an optimization domain of the generative design process.
3. The method of claim 2, wherein the generative design process employs a volume mesh method or a level set method to create a generative design of the object, and the polygonal mesh comprises an original or converted triangle mesh.
4. The method of claim 1, wherein said generating comprises: extracting a quadrilateral patch layout of the quadrilateral patch network from the polygonal mesh, wherein the quadrilateral patch layout represents a quadrilateral partitioning of a domain of a global continuous graph of the polygonal mesh; mapping at least one boundary of the quadrilateral patch layout to the one or more smooth boundary curves corresponding to the one or more modeled entities; and A surface patch is constructed using the global continuity graph of the polygonal mesh, the mapping of the at least one boundary of the quadrilateral patch layout to the one or more smooth boundary curves, and transfinite interpolation of the one or more smooth boundary curves. 5 . The method of claim 4 , comprising receiving user input specifying a quad size, a quad shape, or both for the quad patch network.
6. The method of claim 4, wherein the definition comprises: constructing one or more initial locally refinable smooth surface representations based on the quadrilateral patch network; Refining the one or more initial locally refinable smooth surface representations at one or more boundaries of the one or more initial locally refinable smooth surface representations to approximate the one or more smooth boundary curves within a first tolerance value, the first tolerance value being equal to or greater than the minimum size representable by the geometric modeling kernel of the computer-aided design program and within a supported tolerance range of the geometric modeling kernel of the computer-aided design program; freezing control points of the one or more locally refinable smooth surface representations at the one or more boundaries; as well as The remaining interior portions of the one or more locally refinable smooth surface representations are modified to approximate the polygonal mesh within a second tolerance value that is equal to or greater than an accuracy of a method used to obtain the polygonal mesh and that is at least an order of magnitude greater than the first tolerance value.
7. The method of claim 6, comprising receiving user input specifying the second tolerance value.
8. The method of claim 6, comprising: identifying at least one negative boundary-crossing field singularity of the polygonal mesh; as well as The at least one negative boundary cross-field singularity is relocated to an interior vertex of the polygonal mesh.
9. The method of claim 1, wherein the combining comprises using a stitching operation to connect the one or more locally refinable smooth surface representations to the one or more modeled entities in the boundary representation format.
10. The method of claim 1, wherein the computer-aided design program comprises two or more programs operating in coordination on two or more separate computer processors.
11. The method of claim 1 , further comprising providing, by the computer-aided design program, the second three-dimensional model for use in fabricating a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
12. The method of claim 11, wherein said providing comprises saving said second three-dimensional model to a permanent storage device for use in fabricating said physical structure corresponding to said object using said one or more computer-controlled fabrication systems.
13. The method of claim 11, wherein the one or more computer-controlled manufacturing systems include an additive manufacturing machine or a subtractive manufacturing machine, and the providing includes: generating a tool path specification for the additive manufacturing machine or the subtractive manufacturing machine using the second three-dimensional model; as well as At least a portion of the physical structure corresponding to the object is manufactured with the additive manufacturing machine or the subtractive manufacturing machine using the tool path specification generated for the additive manufacturing machine or the subtractive manufacturing machine.
14. A system for conversion of mesh geometry to a watertight boundary representation, the system comprising: a non-transitory storage medium having stored thereon instructions of a computer-aided design program; as well as one or more data processing devices capable of executing the instructions of the computer-aided design program to perform the operations specified by the instructions of the computer-aided design program; wherein the instructions of the computer-aided design program include means for constructing a quadrilateral patch network from a mesh, the quadrilateral patch network seamlessly combining the mesh with one or more smooth boundary curves on one or more modeled entities in a three-dimensional model of an object to be manufactured, means for generating one or more smooth surfaces from the quadrilateral patch network using a first tolerance value at the boundaries of the mesh and a second tolerance value within the interior of the mesh, the first tolerance value being set based on a minimum three-dimensional model resolution supported by the computer-aided design program and the second tolerance value being at least three orders of magnitude greater than the first tolerance value, and Instructions for combining the one or more smooth surfaces with the one or more modeled entities in the three-dimensional model.
15. The system of claim 14, wherein the mesh is generated by a generative design process that employs the one or more modeled entities as one or more subspaces of an optimization domain of the generative design process, and the second tolerance value is set based on a representation accuracy of the generative design process.
16. The system of claim 15, comprising means for removing any negative boundary crossing field singularities of the mesh prior to generating the one or more smooth surfaces.
17. The system of claim 16, wherein the system comprises one or more computer-controlled manufacturing systems, the manufacturing systems comprising an additive manufacturing machine or a subtractive manufacturing machine, wherein the one or more data processing devices are capable of running the instructions of the computer-aided design program to: generate a tool path specification for the additive manufacturing machine or the subtractive manufacturing machine from the three-dimensional model, and use the tool path specification generated for the additive manufacturing machine or the subtractive manufacturing machine to manufacture at least a portion of a physical structure corresponding to the object using the additive manufacturing machine or the subtractive manufacturing machine.
18. A non-transitory computer-readable medium encoding instructions operable to cause a data processing apparatus to perform operations comprising: obtaining a first three-dimensional model of an object to be manufactured in a computer-aided design program, wherein the first three-dimensional model includes a polygonal mesh associated with one or more modeled entities, the one or more modeled entities being in a boundary representation format; generating, by the computer-aided design program, a quadrilateral patch network from the polygonal mesh, the quadrilateral patch network combining the polygonal mesh with one or more smooth boundary curves corresponding to the one or more modeled entities using a transfinite interpolation method; defining, by the computer-aided design program, one or more locally refinable smooth surface representations using the quadrilateral patch network as input and based at least in part on a minimum size representable by a geometric modeling kernel of the computer-aided design program; and The one or more locally refinable smooth surface representations are combined with the one or more modeled solids by the computer-aided design program to form a second three-dimensional model of the object, the second three-dimensional model being watertight at the one or more smooth boundary curves.
19. The non-transitory computer readable medium of claim 18, wherein said generating comprises: extracting a quadrilateral patch layout of the quadrilateral patch network from the polygonal mesh, wherein the quadrilateral patch layout represents a quadrilateral partitioning of a domain of a global continuous graph of the polygonal mesh; mapping at least one boundary of the quadrilateral patch layout to the one or more smooth boundary curves corresponding to the one or more modeled entities; and A surface patch is constructed using the global continuity graph of the polygonal mesh, the mapping of the at least one boundary of the quadrilateral patch layout to the one or more smooth boundary curves, and transfinite interpolation of the one or more smooth boundary curves.
20. The non-transitory computer readable medium of claim 19, wherein the definition comprises: constructing one or more initial locally refinable smooth surface representations based on the quadrilateral patch network; Refining the one or more initial locally refinable smooth surface representations at one or more boundaries of the one or more initial locally refinable smooth surface representations to approximate the one or more smooth boundary curves within a first tolerance value, the first tolerance value being equal to or greater than the minimum size representable by the geometric modeling kernel of the computer-aided design program and within a supported tolerance range of the geometric modeling kernel of the computer-aided design program; freezing control points of the one or more locally refinable smooth surface representations at the one or more boundaries; as well as The remaining interior portions of the one or more locally refinable smooth surface representations are modified to approximate the polygonal mesh within a second tolerance value that is equal to or greater than an accuracy of a method used to obtain the polygonal mesh and that is at least an order of magnitude greater than the first tolerance value.
