System and method for computer-aided design of physical structures
By optimizing 3D topology and shape through a generative design process, combined with physical simulation and manufacturing technology, the efficiency and performance problems of complex structure design in existing technologies are solved, achieving efficient topology optimization and shape improvement to meet design goals and manufacturing requirements.
Patent Information
- Application Number
- CN201980073660.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2018-11-09
- Filing Date
- 2019-11-06
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2039-11-06
AI Technical Summary
Existing computer-aided design software struggles to effectively combine additive and subtractive manufacturing when generating 3D geometric models, especially when dealing with complex structures such as lattices and hollow components. It cannot efficiently optimize topology and shape to meet design goals and constraints.
A generative design process is adopted, which iteratively modifies the three-dimensional topology and external shape through computer-aided design programs, and optimizes the design of lattice and hollow structure by combining physical simulations such as finite element analysis. Additive and subtractive manufacturing technologies are used to realize the manufacturing of physical structure.
It improves the performance-to-weight ratio of parts, reduces component mass, meets or exceeds structural load requirements, reduces material usage and manufacturing costs, and enhances the manufacturability and efficiency of the design.
Smart Images

Figure CN113196273B_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims the benefit under 35 U.S.C. 119(e) of U.S. Patent Application No. 62 / 758,404, filed on November 9, 2018, entitled “MACROSTRUCTURE TOPOLOGY GENERATION WITH DISPARATE PHYSICAL SIMULATION FOR COMPUTER AIDED DESIGN AND MANUFACTURING,” which is incorporated herein by reference in its 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 structures of those objects (e.g., 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 being modeled. A B-Rep model is a set of connected surface elements that specify the boundaries between the solid and non-solid portions of the modeled 3D object. In a B-Rep model (often referred to as a B-Rep), the geometry is stored in the computer using smooth and precise mathematical surfaces, in contrast to the discrete and approximate surfaces of a mesh model that may be difficult to work with in a CAD program.
[0005] In addition, CAD programs have been used in conjunction with additive manufacturing systems and techniques. Additive manufacturing (also known as solid free-form construction or 3D printing) refers to any manufacturing process in which a 3D object is built from a raw material (generally a powder, liquid, suspension, or molten solid) in a series of layers or cross-sections. Examples of additive manufacturing include fused filament construction (FFF) and selective laser sintering (SLS). In addition, subtractive manufacturing refers to any manufacturing process in which a 3D object is built from a blank material (generally a "blank" or "workpiece" larger than the 3D object) by cutting away portions of the blank material.
[0006] In addition, CAD software has been designed to perform 3D geometry automatic generation (generative design) for a part to be manufactured or one or more parts in a larger part system. This 3D geometry automatic generation is usually limited to a design space specified by a user of the CAD software, and the 3D geometry generation is usually subject to design goals and constraints, which can be defined by the user of the CAD software or another party and imported into the CAD software. Design goals (such as minimizing waste or the weight of the designed part) can be used to drive the geometry generation process towards a better design. Design constraints can include both structural integrity constraints for individual parts (i.e., the requirement that the part should not fail under the expected structural load during the use of the part) and physical constraints imposed by the larger system (i.e., the requirement that the part does not interfere with another part in the system during use). In addition, certain CAD software has included tools to promote 3D geometry enhancement using lattices and skins of various sizes, thicknesses, and densities, wherein the lattice is composed of beams or struts connected to each other at the nodes, and the skin is a shell structure covering or encapsulating the lattice. Such tools allow 3D parts to be redesigned to make them lighter while still maintaining desired performance characteristics (e.g., stiffness and flexibility). Such software tools have used various types of lattice topologies that can be used to generate manufacturable lattice structures. Summary of the invention
[0007] This specification describes techniques related to computer-aided design of physical structures using a generative design process, wherein a three-dimensional (3D) model of the physical structure can be generated to include lattices, hollow portions, holes, and combinations thereof, and wherein the physical structure including these designed features can then be manufactured using additive manufacturing, subtractive manufacturing, and / or other manufacturing systems and techniques.
[0008] In general, one or more aspects of the subject matter described in this specification may be embodied in one or more methods, the one or more methods comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, settings for physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; and iteratively modifying, by the computer-aided design program, both a three-dimensional topology of a generative model of the object and one or more external shapes of the three-dimensional topology using a generative design process, the generative design process representing the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space. The modification may include: changing a constitutive model for the physical simulation according to a current iteration of the three-dimensional topology and the one or more external shapes of the three-dimensional topology, wherein (i) the one or more solid regions are calculated for the constitutive model according to the one or more solid regions within the design space including at least one void, or (ii) the one or more void regions are calculated for the constitutive model according to the one or more void regions within the design space including at least one solid; performing the physical simulation of the object using the changed constitutive model to generate a physical evaluation relative to the at least one design goal and the at least one design constraint; calculating a shape change rate of the one or more external shapes of the three-dimensional topology in the current iteration according to the physical evaluation; and updating the generative model of the object through the generative design process using the shape change rate. In addition, the one or more methods may include: after the modification completes the change of both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, providing a three-dimensional model of the object according to the three-dimensional topology and the one or more external shapes of the three-dimensional topology by the computer-aided design program for manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
[0009] Physical simulations may include finite element analysis (FEA) simulations. It should be noted that various types of physical simulations may be used. Physical simulations performed by the systems and techniques described in this document may simulate one or more physical properties, and may use one or more types of simulations. For example, FEA (including linear static FEA), finite difference methods, and material point methods may be used. In addition, simulations of physical properties may include simulating buckling, natural frequencies, heat fluxes, electrical fluxes or electromagnetic fluxes, and material solidification properties, among other possibilities. Moreover, different types of generative models and generative design processes may be used.
[0010] In some implementations, the generative model is a level set representation of a three-dimensional topology and one or more external shapes of the three-dimensional topology, and the generative design process uses a level set topology optimization method to iteratively modify the three-dimensional topology and the one or more external shapes of the three-dimensional topology. The change may include: according to a homogenized lattice material representation that expresses the stiffness of the lattice as a function of the lattice topology and volume fraction of the lattice at a uniform thickness of the beams in the lattice composed of solid beams and void regions, changing the constitutive model. Moreover, the method may include: after the modification completes the change of both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts (i) the thickness of the beams in the lattice or (ii) the density of the lattice while keeping the three-dimensional topology constant according to the physical simulation, the at least one design goal and the at least one design constraint; and the providing may include: providing the three-dimensional model of the object according to the three-dimensional topology, the one or more external shapes of the three-dimensional topology and the adjusted lattice.
[0011] In some implementations, the changing includes: changing the constitutive model according to the one or more external shapes being offset inward to define a hollow structure including at least one void. The method may include: after the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts the thickness of the hollow structure while keeping the one or more external shapes of the three-dimensional topology constant according to the physical simulation, the at least one design goal and the at least one design constraint. Moreover, the providing may include: providing the three-dimensional model of the object according to the three-dimensional topology, the one or more external shapes of the three-dimensional topology and the adjusted thickness of the hollow structure.
[0012] In some implementations, the changing includes: changing the constitutive model according to (i) the one or more external shapes are offset inward to define a hollow structure including at least one void, and (ii) a homogenized lattice material representation that expresses the stiffness of the lattice as a function of the lattice topology and volume fraction of the lattice at a uniform thickness of the beams in the lattice composed of solid beams and void regions. The method may include: after the modification completes the change of both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts the thickness of the hollow structure while keeping the one or more external shapes of the three-dimensional topology constant according to the physical simulation, the at least one design goal and the at least one design constraint; and after the modification completes the change of both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts (i) the thickness of the beams in the lattice or (ii) the density of the lattice while keeping the one or more external shapes of the three-dimensional topology and the adjusted thickness of the hollow structure constant according to the physical simulation, the at least one design goal and the at least one design constraint. Furthermore, the providing may include providing the three-dimensional model of the object according to the three-dimensional topology, the one or more external shapes of the three-dimensional topology, the adjusted thickness of the hollow structure, and the adjusted lattice.
[0013] One or more aspects of the subject matter described in this specification may also be embodied in one or more methods, the one or more methods comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, a setting for a physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; and iteratively modifying, by the computer-aided design program, both a three-dimensional topology of the generative model of the object and one or more external shapes of the three-dimensional topology, the generative design process using a level set topology optimization method and a level set representation of the generative model in the design space. The modification comprises: changing a constitutive model used for the physical simulation according to a current iteration of the level set representation and expressing the stiffness of the lattice as a homogenized lattice material representation of the lattice topology and volume fraction of the lattice at a uniform thickness of the beams in the lattice composed of beams and void regions; performing the physical simulation of the object using the changed constitutive model to generate a physical evaluation relative to the at least one design goal and the at least one design constraint; calculating a shape change rate of the level set representation in the current iteration according to the physical evaluation; and updating the level set representation of the object using the shape change rate by the level set method.
[0014] The one or more methods also include: after the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts (i) the thickness of the beams in the lattice or (ii) the density of the lattice while keeping the three-dimensional topology constant based on the physical simulation, the at least one design goal and the at least one design constraint; and the computer-aided design program provides a three-dimensional model of the object based on the three-dimensional topology, the one or more external shapes of the three-dimensional topology and the adjusted lattice, for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
[0015] The modification may include adjusting the volume fraction of the homogenized lattice material representation according to a constitutive matrix of the homogenized lattice material representation of the lattice topology. Performing the physical simulation may include performing a finite element analysis simulation using a body-fitting mesh-based solver and changing the constitutive model by mapping geometric field data and finite element simulation data between voxel grid points of the level set representation of the generative model and elements of a solid mesh model used for the finite element analysis simulation. Changing the constitutive model for the physical simulation may include, according to the current iteration of the level set representation, (i) using void simulation elements outside the one or more external shapes of the three-dimensional topology, and (ii) using the homogenized lattice material representation for simulation elements inside the one or more external shapes of the three-dimensional topology.
[0016] The obtaining may include receiving input indicating an initial three-dimensional model defining at least a portion of the design space. The initial three-dimensional model may define a boundary of the design space, and changing the constitutive model used for the physical simulation may include, based on the current iteration of the level set representation, (i) using the homogenized lattice material representation for simulation elements outside the one or more external shapes of the three-dimensional topology but inside the boundary of the design space, and (ii) using solid simulation elements inside the one or more external shapes of the three-dimensional topology.
[0017] Calculating the shape change rate may include calculating an advection rate using an augmented Lagrangian method. The homogenized lattice material representation may express the structural behavior of a given lattice as an anisotropic solid material, wherein the anisotropic solid material is a continuous material having properties approximately equivalent to the given lattice. The one or more computer-controlled manufacturing systems may include an additive manufacturing machine, and the providing may include: generating a tool path description for the additive manufacturing machine from the three-dimensional model; and using the tool path description to manufacture the physical structure corresponding to the object by the additive manufacturing machine.
[0018] One or more aspects of the subject matter described in this specification may be embodied in one or more methods, the one or more methods comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, a setting for a physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; and iteratively modifying, by the computer-aided design program, both a three-dimensional topology of a generative model of the object and one or more external shapes of the three-dimensional topology, the generative design process representing the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space. The modification comprises: the three-dimensional topology and the one or more external shapes of the three-dimensional topology are shifted inward to define a current iteration of a hollow structure in the three-dimensional topology to change a constitutive model for the physical simulation; performing the physical simulation of the object using the changed constitutive model to generate a physical evaluation relative to the at least one design goal and the at least one design constraint; calculating a shape change rate of the one or more external shapes of the three-dimensional topology in the current iteration according to the physical evaluation; and updating the generative model of the object through the generative design process using the shape change rate. The one or more methods also include: after the modification completes the changes to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts the thickness of the hollow structure while keeping the one or more external shapes of the three-dimensional topology constant based on the physical simulation, the at least one design goal and the at least one design constraint; and the computer-aided design program provides a three-dimensional model of the object based on the three-dimensional topology, the one or more external shapes of the three-dimensional topology and the adjusted thickness of the hollow structure, for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
[0019] The offset inward to define the hollow structure may be a constant offset selected by a user of the computer-aided design program. Performing the physical simulation may include performing a finite element analysis simulation using a body-fitted mesh solver and changing the constitutive model by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used in the finite element analysis simulation. Calculating the shape change rate may include calculating the advection rate using an augmented Lagrangian method.
[0020] Adjusting the thickness of the hollow structure may include: starting from one or more internal shapes of the hollow structure, iteratively modifying the one or more internal shapes of the three-dimensional topology using the generative design process, including changing the constitutive model used for the physical simulation according to the current iteration of the three-dimensional topology and the one or more internal shapes, so that after completing the modification of the one or more internal shapes, the thickness of the hollow structure is variable across the three-dimensional topology. The generative model may include a level set representation of the one or more internal shapes and the one or more external shapes of the three-dimensional topology, and the generative design process may employ a level set topology optimization method to iteratively modify the one or more internal shapes and the one or more external shapes of the three-dimensional topology. The one or more methods may include: identifying (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during construction, and / or (iv) visibility of holes for aesthetic considerations in different regions of the three-dimensional topology of the hollow structure; and for a subset of the different regions compared to the remaining regions, determining the location of one or more holes in the three-dimensional topology of the hollow structure for forming or removing support material used during manufacturing based on the identified (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during construction, and / or (iv) visibility of holes for aesthetic considerations. The obtaining may include: receiving input indicating an initial three-dimensional model defining at least a portion of the design space. Moreover, the one or more computer-controlled manufacturing systems may include an additive manufacturing machine, and the providing may include: generating a tool path description for the additive manufacturing machine from the three-dimensional model; and using the tool path description to manufacture the physical structure corresponding to the object by the additive manufacturing machine.
[0021] One or more aspects of the subject matter described in this specification may be embodied in one or more methods, the one or more methods comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, settings for physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; and iteratively modifying, by the computer-aided design program, both a three-dimensional topology of a generative model of the object and one or more external shapes of the three-dimensional topology using a generative design process, the generative design process representing the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space. The modification includes: changing the constitutive model used for the physical simulation based on (i) the three-dimensional topology and the one or more external shapes of the three-dimensional topology are offset inward to define a hollow structure in the three-dimensional topology in a current iteration, and (ii) a homogenized lattice material representation that expresses the stiffness of the lattice as a function of the lattice topology and volume fraction of the lattice at a uniform thickness of the beams in the lattice composed of solid beams and void regions; performing the physical simulation of the object using the changed constitutive model to generate a physical evaluation relative to the at least one design objective and the at least one design constraint; calculating a shape change rate of the one or more external shapes of the three-dimensional topology in the current iteration based on the physical evaluation; and updating the generative model of the object through the generative design process using the shape change rate.
[0022] The one or more methods also include: after the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts the thickness of the hollow structure while keeping the one or more external shapes of the three-dimensional topology constant according to the physical simulation, the at least one design goal and the at least one design constraint; after the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts (i) the thickness of the beams in the lattice or (ii) the density of the lattice while keeping the one or more external shapes of the three-dimensional topology and the adjusted thickness of the hollow structure constant according to the physical simulation, the at least one design goal and the at least one design constraint; and the computer-aided design program provides a three-dimensional model of the object according to the three-dimensional topology, the one or more external shapes of the three-dimensional topology, the adjusted thickness of the hollow structure and the adjusted lattice, for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
[0023] The modification may include adjusting the volume fraction of the homogenized lattice material representation according to a constitutive matrix of the homogenized lattice material representation of the lattice topology. The physical simulation may include a finite element analysis (FEA) simulation. It should be noted that various types of physical simulations may be used. The physical simulation performed by the systems and techniques described in this document may simulate one or more physical properties, and one or more types of simulations may be used. For example, FEA (including linear static FEA), finite difference methods, and material point methods may be used. In addition, simulations of physical properties may include simulating buckling, natural frequencies, heat fluxes, electrical fluxes or electromagnetic fluxes, and material solidification properties, among other possibilities. Moreover, different types of generative models and generative design processes may be used.
[0024] Performing the physics simulation may include performing a finite element analysis simulation using a body-fitted mesh-based solver, and the constitutive model may be altered by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used in the finite element analysis simulation. Calculating the shape change rate may include calculating the advection rate using an augmented Lagrangian method.
[0025] Adjusting the thickness of the hollow structure may include: starting from one or more internal shapes of the hollow structure, iteratively modifying the one or more internal shapes of the three-dimensional topology using the generative design process, including changing the constitutive model used for the physical simulation based on (i) the current iteration of the three-dimensional topology and the one or more internal shapes and (ii) the homogenized lattice material representation, so that after completing the modification of the one or more internal shapes, the thickness of the hollow structure is variable across the three-dimensional topology. The generative model may include a level set representation of the one or more internal shapes and the one or more external shapes of the three-dimensional topology, and the generative design process may employ a level set topology optimization method to iteratively modify the one or more internal shapes and the one or more external shapes of the three-dimensional topology.
[0026] The one or more methods may include: identifying (i) stress levels, (ii) ease of forming or removing support materials, (iii) the ability to support or inhibit material formation during construction, (iv) visibility of holes for aesthetic considerations, or (v) a combination thereof in different regions of the three-dimensional topology of the hollow structure; and determining the location of one or more holes in the three-dimensional topology of the hollow structure for forming or removing support material used during manufacturing based on the identified (i) stress levels, (ii) ease of forming or removing support materials, (iii) the ability to support or inhibit material formation during construction, (iv) visibility of holes for aesthetic considerations, or (v) a combination thereof for a subset of the different regions compared to the remaining regions.
[0027] The offset inward to define the hollow structure can be a constant offset selected by a user of the computer-aided design program. The obtaining may include: receiving input indicating an initial three-dimensional model that defines at least a portion of the design space. The homogenized lattice material representation can express the structural behavior of a given lattice as an anisotropic solid material, which is a continuous material with properties approximately equivalent to the given lattice. Moreover, the one or more computer-controlled manufacturing systems may include an additive manufacturing machine, and the providing may include: generating a tool path description for the additive manufacturing machine from the three-dimensional model; and using the tool path description to manufacture the physical structure corresponding to the object by the additive manufacturing machine.
[0028] These and other methods described herein may be implemented using a non-transitory computer-readable medium encoding a computer-aided design program operable to cause one or more data processing devices to perform the methods. In some implementations, a system includes: a non-transitory storage medium having instructions of a computer-aided design program stored thereon; and one or more data processing devices configured to run the instructions of the computer-aided design program to perform the methods. In addition, such a system may include an additive manufacturing machine or other manufacturing machine, and the one or more data processing devices may be configured to run the instructions of the computer-aided design program to generate instructions for such a machine (e.g., tool path instructions for the additive manufacturing machine) from the three-dimensional model, and use the instructions to manufacture the physical structure corresponding to the object by the machine (e.g., the additive manufacturing machine uses the tool path instructions).
[0029] Specific embodiments of the subject matter described in this specification may be implemented to achieve one or more of the following advantages. The CAD program may provide a variety of different generative design synthesis methods from which to choose, including: level set-based topology optimization, which provides a basic level set method for topology optimization; lattice and skin optimization, which provides thickness optimization of lattices and skins; hybrid topology optimization, which provides topology optimization using lattice filling; inside-out hybrid topology optimization, in which there is lattice filling in the negative space between the topology optimized design and the original design space; hollow topology optimization, which provides a method for topology optimization using internal hollow areas; and / or hybrid-hollow topology optimization, which provides a method for topology optimization using lattice filling and internal hollow areas. Users may be enabled to mix and match different generative design synthesis methods and multiple input design variables to produce a generative design process that helps create a new generative design that meets user goals. Moreover, one or more generative design processes may use macrostructure topology generation in combination with different physical simulation operations that process modeled objects in a manner different from the macrostructure topology generation, which allows the macrostructure topology generation to be informed by more accurate structural characteristics of the type of generative design being created. This can result in an improved topology and shape of the generative design that achieves the physical structural requirements of the object to be manufactured from the generative design.
[0030] Introducing a lattice inside a part during the topology optimization process can increase the specific stiffness (stiffness / mass) of the part and improve the manufacturability of the design, such as when using additive manufacturing. By considering the impact of the lattice behavior during topology optimization, the performance-to-weight ratio of the part can be improved. In addition, components containing optimized lattice structures within a topology optimized body can be designed based on expected structural loads resulting in a lightweight design with high stiffness.
[0031] Additionally, hollow structural components can be generatively designed, which can result in a significant reduction in component mass while still meeting or exceeding specified structural load requirements. This can reduce material usage in the manufacture of such components, improve efficiency, ease transportation, and reduce the cost of manufacturing and transporting physical structures. Generating such hollow structures can produce better strength-to-weight ratios than similar solid components, especially under bending loads. Topological optimization of components can be performed on parts while maintaining the hollow interior within the volume of the component. It should be noted that by considering the hollow structural aspects of the component being designed during the topology optimization process, rather than performing solid topology optimization with high safety factors and / or volume targets so that the designed component can be manually hollowed out later using standard geometric operations, the mass can be reduced for a given design while maintaining or improving structural performance without the significant risk of significant changes in structural response after topology optimization.
[0032] In addition, the shape and topology of the outer surface of the object and the shape and topology of the inner surface of the object (the surface defining the hollow interior) can both be optimized. When the topology of the outer surface is optimized, the inner surface can be created when activating / deactivating elements after the shape changes, which means that only the skin-like element layer needs to be kept active. In addition, the outer surface does not need to be re-meshed at each iteration, which can reduce the required processing resources and increase the reliability of the method. Moreover, the described hollow topology optimization approach can enable the generative design process to expand its ability to synthesize shapes, which can enable more design problems to be solved using generative design.
[0033] Finally, hollow topology optimization can be combined with lattice-based topology optimization to obtain the benefits of both generative design approaches. The macrostructural topology generative process can be informed by the structural properties of the design object with both hollow regions and lattices. This can produce even better generative designs that are very lightweight and have very high stiffness through improved topology and shape of the generative design that meets the physical structural requirements of the object to be manufactured from the generative design. In addition to stiffness, improved topology and shape of the generative design can be achieved, such as efficient heat dissipation, higher / lower natural vibration modes, depending on the design goals used.
[0034] The details of one or more embodiments of the subject matter described in this specification are set forth in the drawings and description below. Other features, aspects, and advantages of the invention will be apparent from the description, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1A An example of a system that can be used to design and manufacture physical structures is shown.
[0036] Figure 1B An example of the process of designing and manufacturing a physical structure is shown.
[0037] Figure 2A A graphical representation showing an example of narrowband level sets.
[0038] Figure 2B A graphical representation showing an example of an octree data structure.
[0039] Figure 2C A graphical representation showing an example of a geometric mapping between an initial design configuration and a current design configuration.
[0040] Figure 2D Graphical representation showing an example of data mapping from a solid grid to a level set grid.
[0041] Figure 3AAn example of a process of generating one or more portions of a 3D model of an object to be manufactured is shown, wherein the process includes various available topology optimizations using one or more generative design processes.
[0042] Figure 3B A graphical representation of an example showing a snapshot from a topology optimization iteration history.
[0043] Figure 3C A graphical representation showing an example of a process for thickness optimization of a lattice and skin.
[0044] Figure 3D Graphical representation showing examples of different lattice topologies.
[0045] Figure 3E A graphical representation showing an example of a process for generating a lattice using unit cells.
[0046] Figure 4A An example of a process for optimizing the topology of a lattice population of one or more portions of a 3D model of an object being generatively designed for manufacturing is shown.
[0047] Figure 4B An example of a process illustrating a design workflow for hybrid topology optimization.
[0048] Figure 4C A graphical representation showing an example of a 3D view of the output from a solid-lattice hybrid topology optimization.
[0049] Figure 4D Show Figure 4C A graphical representation of an example of a slice view of the output from a solid-lattice hybrid topology optimization.
[0050] Figure 4E A graphical representation showing an example of a 3D view of the output from a lattice-void hybrid topology optimization.
[0051] Figure 4F Show Figure 4E A graphical representation of an example of a slice view of the output from a lattice-void hybrid topology optimization.
[0052] Figure 5A An example of a process for optimizing the topology of one or more hollow portions of a 3D model of an object being generatively designed for manufacturing is shown.
[0053] Figure 5B An example of a process showing a design workflow for hollow topology optimization.
[0054] Figure 5C A graphical representation showing an example of a hollow design containing an internal cavity.
[0055] Figure 5D Shown from Figure 5C Graphical representation of an example of a slice view of a hollow design.
[0056] Figure 5E A graphical representation showing an example of a hollow level set.
[0057] Fig. 5F Graphical representation showing the transition of the hollow level set from hollow to basic topology optimization (TO).
[0058] Fig. 6A An example of a process for optimizing the topology of a lattice filling of a structure surrounding one or more hollow portions of a 3D model of an object being generatively designed for manufacturing is shown.
[0059] Figure 6B An example of a process for a hybrid-hollow topology optimization workflow is shown.
[0060] Figure 6C A graphical representation showing an example of a hybrid-hollow design containing an internal cavity and a lattice.
[0061] Fig.6D Show Figure 6C Graphical representation of an example of a slice view of a hybrid-hollow design.
[0062] Fig. 7A An example of a generatively designed hollow part is shown, with a heat map illustrating load paths and exemplary hole locations in areas of lower stress.
[0063] Figure 7B A visualization showing the distance from a given hole to various points inside the geometry.
[0064] Figure 8 is a schematic diagram of a data processing system including a data processing device that can be programmed as a client or a server.
[0065] Like reference numbers and names in the various drawings indicate like elements. DETAILED DESCRIPTION
[0066] Figure 1AAn example of a system 100 that can be used to design and manufacture physical structures is shown. Computer 110 includes processor 112 and memory 114, and computer 110 can be connected to network 140, which can be a private network, a public network, a virtual private network, etc. Processor 112 can be one or more hardware processors that can each include multiple processor cores. Memory 114 can include both volatile and non-volatile memory, such as random access memory (RAM) and flash RAM. Computer 110 can include various types of computer storage media and devices (including memory 114) to store instructions for programs running on processor 112, the programs including computer-aided design (CAD) program 116, the CAD program 116 implements three-dimensional (3D) modeling functions and includes one or more generative design processes for topology optimization through physical simulation (e.g., using at least one level set method as described). The physical simulation performed by the systems and techniques described in this document can simulate one or more physical properties, and one or more types of simulations can be used. For example, finite element analysis (FEA) (including linear static FEA), finite difference method and material point method can be used. Additionally, simulations of physical properties may include computational fluid dynamics (CFD), acoustics / noise control, heat transfer, computational injection molding, electrical or electromagnetic flux, and / or material solidification simulations (material solidification simulations are useful for phase changes during molding). Furthermore, the CAD program 116 may potentially implement hole generation techniques to support manufacturing and / or manufacturing control functions.
[0067] As used herein, CAD refers to any suitable program for designing a physical structure that meets design requirements, regardless of whether the program is capable of docking with and / or controlling manufacturing equipment. Thus, the CAD program 116 may include a computer-aided engineering (CAE) program, a computer-aided manufacturing (CAM) program, and 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 providers that the computer 110 can access via a network 140), or 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., 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.
[0068] 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 1A 118 as separate devices, but 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 device 118 / output device 120). Moreover, the computer 110 may include or be part of a virtual reality (VR) or augmented reality (AR) system. For example, the input device 118 / output device 120 may include a VR / AR input glove 118a and / or a VR headset 120a. In any case, the user 160 interacts with the CAD program 116 to create and modify the 3D model, which may be stored in the 3D model document 130.
[0069] In the example shown, the initial 3D model 132 is a seed model for input into the generative design process. In this example, the user 160 has defined a mechanical problem on which the generative design process is to operate to generate a new 3D model from the starting 3D model 132. In this case, the defined problem is a Michel type arch problem, where the user 160 has specified the domain 134 and the load case 136. However, this is only one of many possible examples.
[0070] In some implementations, user 160 (or other person or program) may specify a design space for an object to be manufactured, settings (e.g., loads and materials) for a physical simulation of the object (e.g., FEA, CFD, acoustic / noise control, heat conduction, computational injection molding simulation, electrical or electromagnetic flux, material curing, etc.), at least one design goal for the object (e.g., minimize material usage), and at least one design constraint for the object (e.g., a volume constraint). In some implementations, inputs for use in the physical simulation and generative design process may include: one or more regions of a current 3D model in which a new 3D model is to be generated; load cases defining one or more loads to be carried by the physical structure being designed in one or more different directions; one or more materials (e.g., one or more isotropic solid materials identified as a baseline material model for the design space); one or more seed model types to be used as input to the generative design process; one or more generative design processes to be used; and / or one or more lattice topologies to be used in one or more regions of the design space. Inputs to the generative design and physical simulation process may include: non-design space; different types of components (e.g., rods, bearings, housings); one or more target manufacturing processes and associated parameters; obstacle geometry that should be avoided; retained geometry that should be included in the final design; and parameters related to various aspects, such as the resolution of the design, the type of synthesis, etc.
[0071] Moreover, the CAD program 116 provides user interface elements in the UI 122 to enable the user 160 to specify the various types of inputs mentioned above, and all (or various subsets) of these inputs can be used in the generative design and physical simulation processes described in this document. In addition, the UI 122 of the CAD program 116 can enable the user 160 to design a part using traditional 3D modeling functions (to build an accurate geometric description of the 3D design model), and then use the generative design and simulation processes in the design space specified within one or more portions of the 3D design model. Therefore, as will be appreciated, many possible types of physical structures can be designed using the systems and techniques described in this document, the UI 122 can be used to create a complete mechanical problem definition for a part to be manufactured, and the generative design and physical simulation processes can accelerate new product development by achieving improved performance without time-consuming physical testing.
[0072] In addition, as described herein, the CAD program 116 implements at least one generative design process that enables the CAD program 116 to automatically generate one or more portions of the 3D model (or the entirety of the 3D model) based on design goals and constraints, wherein the geometric design is iteratively optimized based on simulation feedback. It should be noted that as used herein, "optimization" (or "optimal") does not mean that the best design among all possible designs is achieved in all cases, but rather that the best (or near-optimal) design can be selected from a finite set of possible designs that can be generated within a specified time given available processing resources. Design constraints can be defined by the user 160 or by another party and imported into the CAD program 116. Design constraints can include both structural integrity constraints for individual parts (i.e., the requirement that the part should not fail under expected structural loads during the use of the part) and physical constraints imposed by the larger system (e.g., the requirement that the part is contained within a specified volume so as not to interfere with other parts in the system during use).
[0073] Various generative design processes can be used to optimize the shape and topology of at least a portion of the 3D model. Iterative optimization of the geometric design of the 3D model by the CAD program 116 involves topology optimization, which is a lightweight method in which the optimal distribution of materials is determined by minimizing an objective function subject to design constraints (e.g., structural compliance with volume as a constraint). Topology optimization can be solved using a variety of numerical methods, which can be roughly classified into two groups: (1) material or microstructural techniques, and (2) geometric or macrostructural techniques. Microstructural techniques are based on determining the optimal distribution of material densities and include solid isotropic material penalty (SIMP) methods and homogenization methods. In the SIMP method, intermediate material densities are penalized in favor of having ρ=0 or p=1, which represent voids or solids, respectively. In the homogenization method, intermediate material densities are treated as composite materials.
[0074] In contrast, macrostructural techniques treat the material as homogeneous and represent the three-dimensional topology of the modeled object being generated as one or more boundaries between one or more solid regions (where homogeneous material is present) and one or more void regions (where no material is present) within a design space (also referred to as a domain or a subspace of a domain for topology optimization). The shape of the one or more boundaries is optimized during the generative design process, while the topology changes in the domain due to shape optimization combined with adding / removing and shrinking / growing / merging void regions. Therefore, the type of final optimized topology that can be generated by a generative design process using macrostructural techniques can depend largely on the number and size of voids within the seed geometry of the process.
[0075] It should be noted that although Figure 1A132 (where this model 132 includes a complex solid region 132A with many holes 132B surrounding a void region), but it should be understood that for any given generative design process iteration, the generative design process described in this document may employ two or more seed geometries / models in order to improve the final results of topology and shape optimization. In addition, during the shape and topology optimization process, one or more voids may be introduced into the solid domain and / or one or more solids may be introduced into the void domain in order to improve the final results of topology and shape optimization. Therefore, the CAD program 116 may include various types of available seed geometries and intermediate process geometry introductions, along with user interface elements that allow users 160 to design their own seed geometries and intermediate process geometry introductions. Likewise, the user 160 may run two or more generative design process iterations (saving the results of each iteration) until a preferred generative design geometry is produced.
[0076] Once the user 160 is satisfied with the generatively designed 3D model, the 3D model may be stored as a 3D model document 130 and / or another representation of the generative model may be generated using the 3D model (e.g., a .STL file for additive manufacturing). This may be performed at the request of the user 160 or in view of a user request for another action, such as sending the 3D model 132 to an additive manufacturing (AM) machine 170 or other manufacturing machinery that may be directly connected to the computer 110 or connected via a network 140 as shown. This may involve post-processing performed on the local computer 110 or a cloud service to export the 3D model 132 to an electronic document to be manufactured from. It should be noted that an electronic document (which will be referred to simply as a document for brevity) 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 the document in question, or in multiple coordinated files.
[0077] In any case, the CAD program 116 can provide the document 135 (which has tool path instructions in an appropriately formatted format) to the AM machine 170 to create a complete structure 138, which includes an optimized topology and shape (in this example, an arch design generated for a Michel-type arch problem). 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 addition, the user 160 can save or transfer the 3D model for later use. For example, the CAD program 116 can store the document 130 including the generated 3D model.
[0078] In some implementations, a subtractive manufacturing (SM) machine 174 (e.g., a computer numerically controlled (CNC) milling machine, such as a multi-axis, multi-tool milling machine) may also be used in the manufacturing process. Such a SM machine 174 may be used to prepare an initial workpiece on which the AM machine 170 will operate. In some implementations, a partially complete structure 138 is generated by the AM machine 170 and / or using a casting method (e.g., investment casting (IC) using a ceramic shell or sand casting (SC) using a sand core), and then one or more portions of this partially complete structure 138 are removed (e.g., finished) by the CNC machine 174 to form a completed structure. Moreover, in some implementations, the CAD program 116 may provide a corresponding document 135 (which has tool path instructions in an appropriately formatted format, such as a CNC numerical control (NC) program) to the SM machine 174 for use in manufacturing parts using various cutting tools, etc.
[0079] In various implementations, the CAD program 116 of the system 100 can implement one or more generative design processes as described in this document. The generative design process seeks optimal geometry, topology, or both. For example, the generative design process seeks the best geometry among alternative designs by minimizing a performance-related objective function subject to constraints:
[0080]
[0081] Make g i (s,u(s))=0i=1,...,n g (2)
[0082] where s is a vector of design variables related to the geometry of the domain, and u is a vector of state variables (e.g., displacements) that depend on s. Additional constraints (e.g., equilibrium) are defined by the set g i For simplicity, equality constraints are used here. Mathematical programming methods for minimizing (1) can be gradient-based or non-gradient-based. Gradient-based methods (compared to non-gradient-based methods) generally use more information associated with design sensitivity, such as:
[0083]
[0084] This is the derivative of the performance-related objective function with respect to the design variables. In lattice-based methods, s represents the lattice thickness. In level set-based topology optimization methods, s represents the boundary of the solid region.
[0085] Figure 1BAn example of a process for designing and manufacturing a physical structure is shown. Design variables are obtained 180, for example, by a CAD program 116 for use in generating a generative 3D model. Different generative design processes can be formulated by using different combinations of design variables (which may include lattices, density fields, and level sets). In some implementations, the design variables may include (e.g., received through UI 122) various types of input, such as a selection between different generative design synthesis methods made available by a CAD program in system 100. In some implementations, available generative design synthesis methods include: (1) level set-based topology optimization, which provides a basic level set method for topology optimization; (2) lattice and skin optimization, which provides thickness optimization of lattices and skins; (3) hybrid topology optimization, which provides topology optimization using lattice filling; (4) inside-out hybrid topology optimization, in which lattice filling exists in the negative space between the topology optimized design and the original design space; (5) hollow topology optimization, which provides a method for topology optimization using internal hollow regions; and / or (6) hybrid-hollow topology optimization, which provides a method for topology optimization using lattice filling and internal hollow regions.
[0086] Additionally, design variables are possible, such as (1) a design space for generative design geometry generation, such as a boundary representation (B-Rep) 3D model designed or loaded into the CAD program 116, the design space being used as a subspace of the optimization domain of the described generative design process; and / or (2) a set of input entities that specify boundary conditions for generative design geometry generation, such as a B-Rep selected using the UI 122 to specify a subspace, the set of input entities remaining for use as connection points with other components in a larger 3D model or with a separate 3D model. Different combinations of design variables may be used in response to input from the user 160 (e.g., the CAD program 116). For example, the user 160 may select different generative design synthesis methods to be used in various different design spaces within a single 3D model.
[0087] Other design variables may include settings for physical simulations, such as the density of elements in a FEA model or homogenized lattice material representation of a selected lattice topology to be used with the optimized 3D topology of the part being generatively designed. Design variables may include various design goals and constraints, such as those described in this document. In addition, functionality to help users specify design variables may be provided, for example, by the CAD program 116. For example, a lattice recommender may provide a prediction of suitable lattice settings for a given problem using a single solid simulation. In some implementations, the lattice recommender described in PCT Publication No. WO 2017 / 186786 A1 filed on April 26, 2017 and U.S. Application No. 16 / 096,623 filed on October 25, 2018, both of which are named "METHOD AND SYSTEMFOR GENERATING LATTICE RECOMMENDATIONS IN COMPUTERAIDED DESIGN APPLICATIONS", is used, and both documents are hereby incorporated by reference.
[0088] With the generative design variables specified, one or more 3D models are generated 185, for example, by the CAD program 116 using one or more generative design processes (e.g., using one or more selected generative design synthesis methods). In some implementations, the one or more generative design processes may use the described level set method, where s from equations 1, 2, and 3 represents the boundary of a solid region implicitly represented using one or more level sets, which are signed distance values calculated on a Cartesian background grid. In a level set-based topology optimization method, the external shape of a structure is represented by a one-dimensional high-level level set function, and the changes in shape and configuration are replaced by changes in the value of the level set function in order to obtain an optimal structure. The level set function refers to a function that indicates whether each portion of the design domain in which the initial structure is set corresponds to a material domain (material phase) in which the structure is formed and occupied by a material, a void domain (void phase) in which voids are formed, or a boundary between the two domains, where a predetermined value between a value representing the material domain and a value representing the void domain represents a boundary between the material domain and the void domain.
[0089] Figure 2AA graphical representation of an example of a narrowband level set 200 is shown, where the level set value is constant outside the narrowband. The level set 200 is plotted on a Cartesian background grid relative to the x-axis, y-axis, and z-axis (202, 204, 206, respectively). Scale bar 208 shows different levels of the level set corresponding to signed distance values calculated on the Cartesian background grid, including the 0 isosurface corresponding to the level set representation of the current structural model. The level set representation enables a continuous and clear representation of the outline of the optimal structure, and during the process of structural optimization based on the level set method, one or more gaps can be introduced into the material domain based on the topological derivative of the objective function, thereby allowing topological changes (configuration changes), such as the introduction of holes in the material domain. In addition, in some implementations, one or more octree data structures are used to accurately resolve the geometry, such as Figure 2B An example of an octree data structure 220 is shown in the graphical representation of . The octree data structure 220 is plotted on a Cartesian background grid with respect to the x-axis, y-axis, and z-axis ( 222 , 224 , 226 , respectively).
[0090] As mentioned above, in some implementations, available generative design synthesis methods include level set-based topology optimization. Level set-based topology optimization involves optimizing the shape of a design domain using shape derivatives, which are derivatives of a constrained minimization problem with respect to the shape. Shape changes are applied to level sets, allowing topological changes during shape modifications. The result of this type of generative design process is a partitioning of the design space into solid regions and void regions, resulting in an optimized shape, which typically has topological changes. For this type of level set-based topology optimization, as well as variations of this type of level set-based topology optimization described in this document, one or more of the following approaches may be used.
[0091] Linear Elastic Topology Optimization
[0092] Consider a domain Ω
[0093]
[0094] u=0,at Γ D Previous (5)
[0095]
[0096] where ∈(u) is the linear strain tensor, D is the fourth-order constitutive tensor, u is the displacement vector, f is the external load vector, and is the Neumann boundary Γ N with a prescribed traction force on the outward normal vector n. For simplicity, homogeneous Dirichlet boundary conditions can be assumed on and . Then the constrained topology optimization problem can be:
[0097] Minimize J(Ω,u) (7)
[0098]
[0099] u=0,at Γ N Previous (9)
[0100] D∈(u)n=t,in Γ N Previous (10)
[0101] where compliance minimization can be used as the objective function:
[0102]
[0103] Figure 2C A graphical representation of an example of a geometric mapping 240 between an initial design configuration and a current design configuration is shown. The solution space in topology optimization can be defined by different perturbations of the geometry within the design space. In this context, a linear mapping 242 that maps a given domain 246Ω into a perturbed domain 248Ωt can be defined. By this mapping, a material point with coordinates x∈Ω can be mapped 244 onto:
[0104] x t =x+tδv, t≥0 (12)
[0105] where δv is a constant vector field and t is a scalar parameter (see Figure 2C ). It should be noted that solving the equations using a gradient-based mathematical programming approach involves using the directional derivative of the objective function in the direction of the velocity field, δv
[0106]
[0107] More than one approach can be used to obtain directional derivatives of the objective function for use in gradient-based optimization methods. Approaches suitable for use with gradient-based optimization methods include direct differentiation, semi-analytical derivatives, adjoint methods, and finite differences.
[0108] Adjoint method
[0109] Evaluating the shape derivative (Equation 13) may require the directional derivative of the state variable u in the direction of the velocity vector δv. This can be seen by using the chain rule:
[0110]
[0111] However, in some implementations, the following adjoint method may be used, which involves forming the Lagrangian L(Ω, u, λ), which depends on the domain shape Ω, the displacement field u, and the Lagrangian parameter λ
[0112]
[0113] The stability condition of the Lagrangian (i.e., δL(Ω, u, λ) = 0) leads to a complete set of equations for shape optimization. For example, the adjoint problem of compliance minimization (Eq. 11) can be given by considering the variation of the Lagrangian with respect to the displacement u. After introducing the cost function (Eq. 11) and revising the domain terms using the divergence theorem:
[0114]
[0115] The corresponding boundary value problem (called the adjoint problem) can be transformed into:
[0116]
[0117] λ=0,at Γ D Medium (18)
[0118]
[0119] This can lead to the determination that λ = -u is a solution to the adjoint problem. This means that for the compliance minimization problem (Eq. 11), the adjoint problem (Eqs. 17-19) does not need to be solved explicitly. Such problems are called self-adjoint problems, where the solution to the direct problem also yields the adjoint solution. However, this is not the common case, and depending on the nature of the direct problem and the objective function, a different adjoint problem may have to be solved. The advantages of using the Lagrangian as is include the identification of:
[0120]
[0121] This equation allows the shape derivative (Eq. 13) to be expressed as a boundary integral of the form:
[0122]
[0123] Without loss of generality, some boundary variations can be assumed to be irrelevant to the actual shape optimization. In solid mechanics, boundary variations can generally have the following form:
[0124] δv=0,at Γ D Previous (22)
[0125] δv=0,at Γ N On, among them
[0126] δv≠0, in Γ N Above, where σn=0.
[0127] This means that during shape optimization, only the boundary Γ NThe parts without traction are free to move. In this context, with structural compliance (Eq. 11) as the cost function, the Lagrangian (Eq. 15) is The above changes can become:
[0128]
[0129] Without limiting δv as described in Eq. 22, the change in the Lagrangian may contain several more terms. During iterative optimization of the shape, the shape derivative (Eq. 23) may be used as gradient information. To achieve the maximum reduction in the objective function, the boundary perturbations may be chosen as follows:
[0130] δv=-(2u·fD∈(u):∈(u)). (twenty four)
[0131] Along the direction of the normal vector Apply this boundary perturbation, where v is the shape change rate and is given by the following equation:
[0132]
[0133] Volume Control
[0134] Topology optimization using only the compliance minimization objective (Eq. 11) can produce an optimal topology that covers the entire design space. Therefore, some form of volume constraint is usually required. Moreover, in some implementations, controlling the volume change during topology optimization may be important for several reasons: 1) imposing volume constraints; 2) providing user control over the progress of topology optimization, e.g., more volume changes during initial iterations and less volume changes during later iterations; and 3) ensuring that arbitrary constraints without shape derivatives are satisfied.
[0135] It should be noted that constraining the existence of shape derivatives may require modifying the shape change rate in Equation 25. A modified objective function may be considered where the volume is penalized by a penalty parameter μ in the following equation:
[0136]
[0137] The corresponding shape derivative (Eq. 25) is then given by:
[0138]
[0139] where μ is constant along the boundary. The rate term in the shape derivative (Eq. 25) can now have an additional term as follows:
[0140] v=-(2u·fD∈(u):∈(u)+μ) (28)
[0141] Augmented Lagrangian method
[0142] In some implementations, an augmented Lagrangian method is used. Certain approaches (see volume control above) may have limitations, such as difficulty in achieving a specified volume target. In essence, the final volume of the design may depend on the value of μ specified in Equation 26. In such cases, achieving the specified design constraint target can be achieved by using an augmented Lagrangian method. Consider the following Lagrangian for compliance minimization, where the final volume target is V f (Ω):
[0143]
[0144] The shape derivative is then given by the following equation:
[0145]
[0146] The penalty parameters λ and μ may be updated in an increasing sequence so that they converge to the optimal Lagrange multiplier. In some implementations, one or more heuristic methods may be used to update the penalty parameters.
[0147] Body-fitted solver
[0148] In some implementations, one or more body-fitting grid-based solvers are used. Using such body-fitting grid-based solvers with level set methods involves mapping the data from the solid grid to a Cartesian grid (note that the inverse mapping is trivial due to the structured nature of the Cartesian grid). This involves Figure 2D The two mappings shown, Figure 2D A graphical representation of an example of a data mapping 260 from a solid grid 262 to a level set grid 264 is shown.
[0149] The data (e.g., strain energy, von Mises stress) in the solid mesh elements may first be mapped to the solid mesh nodes. This mapping may be accomplished by data averaging. For example, by averaging the solid mesh 262 at the solid node n j The solid mesh element e at i The data is averaged. In addition, a linear shape function can be used to map the data in the solid grid nodes to the voxel grid points. j The data at is linearly interpolated to the level set grid point g in the level set grid 264 i Having this mapping allows the use of level set methods with complex FEA models solved using a body-fitted solver.
[0150] As mentioned, the above approaches can be combined with the following FIG. 3A to FIG. 6DThe various types of level set-based topology optimization used together are described further below. Figure 1B , all of these level set methods of generative design, as well as other generative design processes described herein, may be implemented (e.g., in a CAD program 116) to provide both: (1) substantial user control over the generative design process and post-processing of the generative design output to form a final acceptable 3D model of an object; and (2) control functionality for providing the generatively designed 3D model for use in fabricating a physical structure corresponding to the object. Thus, the results of the generative design process may be presented to the user, for example, in a UI 122 on a display device 120, along with an option 190 to accept or reject the design.
[0151] If the design is rejected, Figure 1B The process may return to obtaining 180 new design variables, for example from the CAD program 116, for use in generating a new generated 3D model. Once the design is not rejected 190, Figure 1B The process may provide 195 the 3D model, 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 195 may involve sending or saving the 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. In some implementations, providing 195 involves: using the 3D model, for example by the CAD program 116, to generate 195A a tool path description for the computer-controlled manufacturing system; and using the tool path description generated for the additive manufacturing machine, for example by the CAD program 116, to manufacture 195B at least a portion of the physical structure corresponding to the object using the computer-controlled manufacturing system.
[0152] It should be noted that the 3D model provided 195 may be a post-processed version of the 3D model or generative design output generated 185 by the generative design synthesis method. For example, in some implementations, the 3D mesh model generated by the generative design synthesis method may be converted into a watertight B-Rep 3D model before providing 195. In some implementations, the generative design output may be post-processed using the systems and techniques described in U.S. Patent Application No. 62 / 758,053, filed on November 9, 2018, and entitled “CONVERSION OF GENERATIVE DESIGN GEOMETRY TOEDITABLE AND WATERTIGHT BOUNDARY REPRESENTATION IN COMPUTER AIDED DESIGN”. Furthermore, in some implementations, the post-processed generative design output can be edited using the systems and techniques described in U.S. Application No. 16 / 186,136, filed on November 9, 2018, and entitled “FACILITATED EDITING OF GENERATIVE DESIGN GEOMETRY IN COMPUTER AIDED DESIGN USER INTERFACE”. FIG. 3A to FIG. 7B The generative design processes described may also be implemented using the post-processing, editing, and / or providing 195 systems and techniques described above. Finally, while described in the context of a CAD program that provides multiple options for generative design, each of the generative design processes described in this document may be implemented as a stand-alone generative design process in a CAD program. Thus, not all of the following may be combined with FIG. 3A to FIG. 7B The generative design processes described need to be implemented together in any given implementation.
[0153] Figure 3A A process for generating one or more portions of a 3D model of an object to be manufactured (e.g., as described by Figure 1A 116), wherein the process includes various available topology optimizations performed using one or more generative design processes. A design space for the selected generative design process, settings for physical simulation, and other inputs are identified 300. This may involve, for example, receiving user input via UI 122 on display device 120, importing information from another program or third-party source, and / or one or more of these inputs may be predefined in a given implementation.
[0154] The settings for the physical simulation may include one or more physical properties to be simulated and one or more types of simulations to be performed (as discussed above), as well as potentially alternative modeling or other approximation methods. In some implementations, the type of physical simulation is predefined for all uses of the program from which the generative design process has been initiated or in view of a specific context in the program. In addition, the settings for the physical simulation include at least one set of load conditions and / or other physical environment information associated with the type of physical simulation to be performed.
[0155] The design space can be an initial 3D model or one or more parts of the initial 3D model to be used as a starting geometry. In some cases, the design space can be determined as a bounding box of all initial geometries, and the bounding box is used to specify loads or other physical environment information associated with the type of physical simulation to be performed. In some cases, the design space can be boundless. In some implementations, the part of the design space to be used as the starting geometry can be automatically set by a genetic algorithm or other process. For example, bubble-shaped holes (e.g., hole 132B) can be placed in the domain, and a genetic algorithm can be used to change the bubble size and spacing. Moreover, other inputs may depend on the type of physical simulation to be performed and / or the type of generative design process to be used. For example, when a lattice is to be used in a generative design process, other inputs may include lattice topology, volume fraction, unit size, and thickness.
[0156] When the generative process to be used employs a basic level set method for topology optimization, a level set is initialized 302A for the design space. It should be noted that the level set method is an example of a macrostructure generative modeling technique, in which the generative model represents the object being designed as one or more boundaries between one or more solid regions (with homogeneous material therein) and one or more void regions (without material therein) within the design space. In this case, after initialization, a physical simulation of the modeled object is performed 302B, in which each solid region is considered to be a homogeneous solid material and each void region is considered to have no material therein.
[0157] The level set is updated 302C using the results from the simulation and checked 302D for convergence. The simulation, update and check process is then iterated until convergence. A detailed example of this type of macrostructure topology optimization process is now described. To simplify the presentation, the compliance minimization problem is used together with the penalty volume (Equation 26). Moreover, for all of the following detailed examples, for ease of presentation, it is assumed that FEA is used for physical simulation, but other physical simulation types mentioned above may also be used.
[0158] The starting shape may simply be the design space or the intersection of the design space with a suitable seed geometry. An initial level set ψ0 may be created by converting the starting shape to a signed distance field (SDF). The conversion may be performed using an implementation of the OpenVDB toolkit. Other approaches are also possible. Next, the shape may be iteratively optimized until the objective function has converged. The FEA model used for the simulation includes solid elements throughout the solid region of the generative design space being optimized. In each iteration, the constitutive model D of the elements may be updated by changing the elastic modulus according to the relative position of each element e in the FEA model relative to the current level set. For example, elements outside the level set (called void elements) may be given a very low stiffness D v , and the constitutive model of the internal elements is set to the constitutive model of the original solid material D s This can be done by checking the average level of convergence of the element nodes:
[0159]
[0160] where n j denotes the coordinates of the element nodes. Once the FEA model is synchronized with the current geometry represented by the level set, the boundary value problem (Equations 4-6) can be solved to compute the advection rate. The shape change can be applied by advecting the level set ψ using, for example, the Hamilton Jacobi equation:
[0161]
[0162] where v is the shape derivative (Eq. 28). Note that one or more heuristic shape updating methods may be used, and one or more methods of moving the 0th contour of the level set may be used given the direction of movement specified by the shape derivative. A linear mapping between FEA nodes and level set grid points (see body-fitting solver above) may allow FEA results such as D∈(u):∈(u) to be transferred to the level set where the advection rate v is calculated.
[0163] Once the objective function has converged, the surface of the final level set can be extracted using a contour method that extracts the 0th contour of ψ. Figure 3B A graphical representation of an example of a snapshot 320 from a topology optimization iteration history is shown. For each of the snapshots 320a, 320b, 320c, 320d, Figure 3B The level set 322 and the corresponding FEA model 324 are shown in FIG. An example of the algorithm is as follows:
[0164] Algorithm 1-Level Set-Based Topology Optimization
[0165] Input: (Ω, D v , D s , μ)
[0166] Output: (Ω s )
[0167] / / Initialize the level set ψ according to the design space
[0168] 1: ψ0 = f SDF (Ω)
[0169] / / Iterate until the convergence tolerance c is met
[0170] 2: i = 0
[0171] 3: When i=0 or |J i -J i-1 |>c
[0172] / / Set the constitutive model of the FEA element to void D x or entity D s
[0173] 4:
[0174] / / Formulate and solve FEA problems
[0175] 5.
[0176] 6.uK -1 f
[0177] / / Calculate the advection rate
[0178] 7:
[0179] / / Solve the Hamiltonian Jacobian and obtain the new level set ψ i+1
[0180] 8:
[0181] 9: ψ i+1 ←ψ i
[0182] / / Calculate the target
[0183] 10:
[0184] / / Increment the iteration
[0185] 11: i←i+1
[0186] 12: End the loop
[0187] / / Get the final level set as the solid area
[0188] 13:
[0189] It should be noted that generalizing this algorithm for arbitrary objectives and constraints may require modifications, such as the following. In line 6, any adjoint problems required to compute shape derivatives for each objective and constraint should be solved (see adjoint method above). In line 7, the different shape derivatives should be combined using an augmented Lagrangian method (see augmented Lagrangian square method above) to produce a single advection rate. For further details on implementing these modifications to the generative design process in general, see the Augmented Lagrangian Algorithm for Constrained Shape Optimization below.
[0190] return Figure 3A , when the generative process to be used is lattice and skin optimization, a lattice and skin model 304A is initialized for a design space, where the design space is divided into a lattice surrounded by a skin. It should be noted that a lattice is a set of solid material beams arranged in a 3D configuration (topology), and a skin is a thin shell of solid material wrapped around the lattice material. After initialization, a physical simulation of the modeled object is performed 304B, and the lattice model portion of the model is updated 304C using the results from the simulation using a heuristic method that relates stress ratios to the thickness of the beams in the lattice. This simulation 304B and update 304C process is iteratively performed until convergence 304D. A detailed example of this lattice and skin optimization is now described.
[0191] Figure 3C A graphical representation of an example of a process 340 for thickness optimization of a lattice and skin is shown. The lattice and skin optimization can be a two-step process, where a lattice and skin of uniform thickness are created in a first step, followed by an optimization of the lattice and skin thickness driven by full stress design principles. For example, an initial solid domain Ω 342 can be divided 344 (automatically by a computer or in response to user input) into a skin (Ω h ) region 348 and the lattice (Ω l ) region 350, and then subjected to thickness optimization 346, such as skin thickness optimization 348A and / or lattice thickness optimization 350A. It should be noted that in some implementations, one or more suitable ways of creating and optimizing lattices may be used.
[0192] Lattice and Skin Creation
[0193] Figure 3DA graphical representation of examples of different lattice topologies 360 is shown. For example, different lattice topologies 360 are shown as unit cells (e.g., cubes of size 1×1×1). Different lattice topologies 360 include X topology 362, grid topology 364, tetra topology 366, octet topology 368, snowflake topology 370, and vin tile topology 372. Other lattice topologies are also possible. Lattice design creation can be performed by partitioning the original design space to obtain volume subsets for lattice creation. Next, the available size is u = {u x ,u y ,u z The lattice domain can then be tiled with the lattice filled cells. l The boundaries of any cubes intersecting the lattice ensure that the lattice does not protrude outside the lattice area.
[0194] There may be a library of pre-computed lattice cells for creating different lattice microstructures T inside the cell. The output lattice may depend on the following parameters: 1) lattice topology T: representing different lattice microstructures; 2) unit size vector u = {u x ,u y ,u z}: limit the size of the unit cell; 3) the thickness of the lattice beam t∈{t min ≤t≤t max}: thickness of each lattice beam; and 4) offset vector f = {f x , f y , f z}: Use this vector to tile the unit cell at an offset from the origin of the lattice domain.
[0195] Figure 3E A graphical representation of an example of a process 380 for generating a lattice using unit cells is shown. The process 380 includes transitioning from a design space 382 and unit cells 384 to a design space 386 populated with unit cells, and then to a final lattice 388 with optimized beam thicknesses.
[0196] The input for skin creation may include one or more user-selected subsets of the design domain surface, hereafter referred to as skin regions. The goal of skinning can include converting a skinned region into a volume surface with a given thickness. In each skinned region, the thickness of each vertex v can be calculated by averaging the polygon normal vectors. i ∈Γ h Each polygon can then be cloned and the corner vertices can be moved along the vertex normal vector according to the specified skin thickness.
[0197] Lattice and Skinning Optimization
[0198] In some implementations, a linear elastic size optimization approach is adopted. The lattice and the skin can be composed of n b 、n h beam and shell elements, where
[0199]
[0200]
[0201] Among them B i , S i represent beam and shell elements of variable thickness, respectively. Variable thickness can be achieved by having different thickness values at each vertex. The goal of this generative design process can be to optimize each vertex v i ∈Ω l ∪Ω h The thickness at t i ,For example
[0202] Minimize J(Ω l ,Ω s ,u) (35)
[0203] Subject to Ku=f (36)
[0204] t min ≤t i ≤t max (37)
[0205] In full stress design, the objective function can be a weighted sum of volume and stress
[0206]
[0207] The constrained thickness optimization problem of Eqs. 35-37 can be optimized using a heuristic approach to the lattice optimization process (eg, Algorithm 2 below).
[0208] Geometry initialization involves the creation of the lattice and skin (see Lattice and Skin Creation above). The lattice can be initialized using the lattice topology T, the unit size u, the offset vector f, and the minimum bound t on the lattice thickness variable. min to be created inside the lattice design space (which can be a subset of the full domain). The skin can be created by applying a subset of polygons Γ on the domain surface Γ to the h Extrude to create a thickness equal to the minimum limit t minThe skin can be created with a number of layers. Before starting the optimization loop, the thickness variables can be initialized to be at minimum bounds, as reflected during lattice and skin creation. Note that each lattice can have two thickness variables corresponding to the thickness at each end of each beam. Similarly, each polygon on the skin can have the same number of thickness variables as the number of vertices in the polygon. Thickness variables can be shared at lattice joints and at skin polygon junctions.
[0209] The FEA model used for simulation can use beam and shell elements to represent the lattice and skin, respectively. The FEA model can be composed of beam, shell, and solid elements, while the geometric model can use beam and shell thickness values to create the lattice and skin, where the solid area remains unchanged. During each iteration of the optimization process, the stiffness of all beam and shell elements can be updated according to the current value of the corresponding thickness design variable. Once the FEA model is solved, the stress ratio at each lattice beam end and shell polygon vertex can be calculated. This stress ratio information can be used to update the thickness for the next iteration, for example, using a heuristic rule or another approach.
[0210] Algorithm 2 - Lattice and Skin Thickness Optimization:
[0211] Input: (T,u,f,t min , t max , Γ h ,Ω,σ t )
[0212] Output:
[0213] / / Create lattice and skin using minimum thickness
[0214] 1.
[0215] 2.
[0216] / / Initialize the thickness of all elements in the lattice and skin to the minimum thickness
[0217] 3.
[0218] / / Iterate until the convergence tolerance c is met
[0219] 4. i = 0
[0220] 5. When i = 0 or |J i -J i-1 |>c
[0221] / / Calculate element stiffness using the thickness of the vertices in the element
[0222] 6.
[0223] / / Formulate and solve FEA problems
[0224] 7.
[0225] 8.u=K -1 f
[0226] / / Calculate the stress ratio at each vertex
[0227] 9.
[0228] / / Calculate thickness increment based on stress ratio
[0229] 10. l=# load cases
[0230] / / Update the thickness without violating bounds
[0231] 11.
[0232] / / Calculate the target
[0233] 12.
[0234] / / Increment the iteration
[0235] 13.i←i+1
[0236] 14. End the loop
[0237] Note that f in row 10 above is a function of the stress ratios from different load cases. For example, the full stress design (FSD) method devised in Patnaik et al., "Optimality of a Fully Stressed Design", NASA / TM-1998-207411, may be used. Note that other methods suitable for converting stress ratios (or other simulation data) into thickness adjustments may be used.
[0238] return Figure 3A, when the generative process to be used employs a lattice hybrid level set method for topology optimization, a level set is initialized 306A for the design space. It should be noted that because the level set method is an example of a macrostructure generative modeling technique, the generative model (e.g., level set) represents the object being designed as one or more boundaries between one or more solid regions (having homogeneous material therein) and one or more void regions (having no material therein) within the design space. However, in the case of hybrid topology optimization (as described in this document), a physical simulation of the modeled object is performed 306B, wherein at least a portion of the solid region is treated as having at least one void therein (e.g., the macrostructure generative modeling technique treats it as a solid and the physical simulation treats it as something that partially contains the void in the form of a lattice), or at least a portion of the void region is treated as containing at least one solid therein (e.g., the macrostructure generative modeling technique treats it as a void and the physical simulation treats it as something that partially contains solid material in the form of a lattice).
[0239] The level set is updated 306C using the results from the simulation, and converged 306D for convergence. The simulation, updating, and checking process is then iterated until convergence. A detailed example of this type of macrostructural hybrid topology optimization process is now described, in which topology optimization and lattice can be used together to produce a final design containing solids or voids as well as lattices. However, rather than manually creating such a design by replacing solid or void regions (of the generated topology) with lattices (which can result in a suboptimal design), the effects of lattice behavior can be considered during topology optimization, which can result in an improved structural design, such as a stronger or more durable part, less material usage, or a combination thereof, such as an improved strength to weight ratio.
[0240] In order to take into account the lattice behavior during topology optimization, a homogenized lattice material representation can be created for the lattice. The homogenized lattice material representation expresses the stiffness of the lattice composed of beams and void regions as a function of the lattice topology and the volume fraction of the lattice (and possibly other lattice settings) at a uniform thickness of the beams in the lattice. For example, each of the available lattice topologies (e.g., lattice topology 360) can have a representative volume element (RVE) generated for it. One or more methods for creating a representative volume element (RVE) for a lattice can be used, including the methods described in PCT Publication No. WO 2017 / 186786 Al, filed on April 26, 2017, and U.S. Application No. 16 / 096,623, filed on October 25, 2018, both entitled "METHOD AND SYSTEMFOR GENERATING LATTICE RECOMMENDATIONS IN COMPUTER AIDED DESIGN APPLICATIONS".
[0241] The RVE concept assumes that a specific shape / pattern of the microstructure is repeated in the lattice. Therefore, instead of modeling the entire lattice structure, the repetitive part of the shape / pattern (RVE) can be modeled using appropriate periodic boundary conditions. Once the behavior of the RVE is calculated, the repetitive part can be homogenized and represented as an anisotropic material. Therefore, the lattice microstructure can be represented as a continuous material with equivalent properties. These equivalent properties can be obtained through appropriate micromechanical modeling processes, such as using a homogenization method to create representative volume elements for different lattice topologies. The lattice widely used in homogenization is the octet lattice and its analytically calculated RVE. A database of numerically calculated lattice RVE properties can be constructed for use in hybrid topology optimization. Therefore, hybrid topology optimization can combine lattice construction with level set-based topology optimization using homogenized lattice RVE.
[0242] Homogenization
[0243] Various lattice topology types (see e.g. Figure 3D ) can be used for lattice creation. Each topology type can have unique structural behavior and can produce different optimal designs when used to optimize structural components. In addition, the structural behavior can vary with the matrix material, lattice orientation, unit size and thickness. Therefore, in various implementations, there can be different levels of integration of topology optimization with lattice structures.
[0244] For integration into topology optimization methods, the lattice stiffness can be expressed as a function of the material density (volume fraction) and the lattice topology type. The characteristic length c that depends on the lattice topology type can be used. T Let the lattice volume fraction x l With the unit size u(u=u x =u u =u z ) is related to the thickness t as follows:
[0245]
[0246] The behavior of the homogenized lattice can be expressed using a linear elastic constitutive model as:
[0247]
[0248] where σ and E contain the components of the Cauchy stress tensor and the corresponding strain tensor respectively given by:
[0249] σ T ={σ xx , σ yy , σ zz , σ xy , σxz , σ yz} (41)
[0250] ∈ T ={∈ xx ,∈ yy ,∈ zz , γ xy , γ xz , γ yz (42)
[0251] It should be noted that for an isotropic linear elastic material, the constitutive matrix is given by:
[0252]
[0253]
[0254] Where E and v represent Young's modulus and Poisson's ratio, respectively.
[0255] In addition, the stress and strain in Eq. 40 are not the stress and strain in the lattice beam, but the stress and strain of the homogenized RVE. Therefore, the stress and strain information in the lattice beam should be converted into the stress and strain information of the homogenized RVE. First, the stress can be applied on the boundary of the RVE as a surface traction. For example, the uniform surface traction t = u 2 To achieve the boundary σ 11 = 1. On the other hand, the strain can be obtained by averaging over the boundary.
[0256] Additionally, since the RVE represents a potentially infinitely repeating shape / pattern, periodic boundary conditions should be applied. Finally, due to the lattice layout and orientation, the RVE exhibits anisotropic material properties. In this application, the lattice can be considered to be an anisotropic material with no symmetry planes with a constitutive matrix given by:
[0257]
[0258] Referring to Equation 40, Each column in the matrix contains the strain component obtained by applying a unit stress component on the boundary of the RVE of the desired lattice. Essentially, The entries in the i-th column of the matrix can be given by the components of the vector ∈ obtained by setting the i-th stress component to 1. For example, this can be obtained by applying σ 22 =1Get The second column of the matrix is as follows:
[0259]
[0260] ∈=[d 12 d 22d 32 d 42 d 52 d 62 ] T (47)
[0261] Variation with lattice volume fraction
[0262] Since the lattice volume fraction can be changed during topology optimization, the constitutive matrix is limited A functional relationship with the volume fraction can be advantageous. ij and the lattice volume fraction x l The relationship between can be expressed using a fourth-order polynomial:
[0263] d ij =ξ(x l ) T b ij (48)
[0264] in is the vector of the monomial basis, and b ij ={a ij b ij c ij d ij e ij} is the vector of unknown coefficients.
[0265] The unknown polynomial coefficients can be obtained by l ∈D (e.g., where D = {.1, .15, .2, ..., .75}) Numerical calculations are performed to estimate. As expected, the constitutive properties can be a monotonic function of volume fraction. Therefore, the fourth-order polynomial approximation in Eq. 48 can be a reasonable choice. It should be noted that x is not calculated due to its dependence on Poisson's ratio. l Alternatively, you can use the same l =1 corresponds to a completely solid material Let's consider this. For further details on the variation of matrix entries with lattice volume fraction, see PCT Publication No. WO 2017 / 186786 Al, filed on April 26, 2017, and U.S. Application No. 16 / 096,623, filed on October 25, 2018, both entitled “METHODAND SYSTEM FOR GENERATING LATTICE RECOMMENDATIONS IN COMPUTER AIDED DESIGN APPLICATIONS.”
[0266] Predicting RVE for arbitrary lattice properties
[0267] Generating a complete lattice RVE database involves l ∈D Computation and Storage Matrix d ij entries. Then you can Conventional regression (e.g., least squares method) is used during matrix calculation in Equation 48. Note that in the numerical Some references can be used during matrix calculations E 0 and v 0 Use this database to calculate Know The constitutive matrix of RVE The process can be as follows.
[0268] First, n discrete data points around the requested x1 value in D may be selected. Since there are five unknown coefficients, it is advantageous to use at least 10 data points in order to prevent overfitting.
[0269] {x1, x2, …, x n}∈D (49)
[0270] x1<x2<…<x n (50)
[0271]
[0272] At each data point, the basis vector in Equation 48 can be evaluated. By stacking the row vectors of the basis, the following n x 4 moments can be generated:
[0273]
[0274] Similarly, d can be collected at all data points ij to construct the response vector y ij ={d ij (x1), d ij (x2),…,d ij (x n ),} T .
[0275] Second, if the selected data points include x l =1, then equations 43-44 can be used at x l = 1 calculation Third, the pre-computed discrete d ij value to take the actual elastic modulus into account:
[0276]
[0277] Fourth, linear regression may be used to determine the polynomial coefficients in Equation 48.
[0278] b ij =[X T X] -1 {X T y ij} (54)
[0279] Fifth, once the unknown coefficients are calculated from Equation 54, Equation 48 can be used to calculate the The components of the constitutive matrix of the lower lattice RVE are as follows:
[0280] d ij (x l )=ξ(x l ) T b ij (55)
[0281] Design Workflow
[0282] Figure 4B As shown, for example, by Figure 1A 1 . An example of a process 440 of a design workflow for hybrid topology optimization performed by a CAD program 116 of FIG. 1 . In process 440, the topology can be optimized while keeping the lattice thickness constant. The topology is optimized 442 for a uniform thickness lattice. Due to the two possible options 448 and 450, two types of designs can be obtained from process 440. At 446, a check can be made as to whether there is a lattice inside the optimized topology design (vs. outside the optimized topology design). The first option involves dividing the design space into solid and lattice regions, in which case a solid lattice topology optimization 450 can be performed using Algorithm 3, such as Figure 4C and Figure 4D shown.
[0283] Figure 4C A graphical representation of an example of a 3D view 460 of output from a solid-lattice hybrid topology optimization is shown. Figure 4D Show Figure 4C 4 is a graphical representation of an example of a slice view 470 of the output from a solid-lattice hybrid topology optimization. As shown, the lattice 462 is located outside the topology optimized surface around the solid region 464. This is an example of inside-out hybrid topology optimization, where the lattice fill exists in the negative space between the topology optimized design and the original design space.
[0284] The second option involves partitioning the design space into lattice and void regions, in which case a lattice void topology optimization 448 may be performed using Algorithm 4, such as Figure 4E and Figure 4F shown. Figure 4E A graphical representation of an example of a 3D view 480 of output from a lattice-void hybrid topology optimization is shown. Figure 4F Show Figure 4E 4 is a graphical representation of an example of a slice view 490 of the output from a lattice-void hybrid topology optimization. As shown, the lattice 482 is inside the topology optimized surface 484. This is an example of a hybrid topology optimization that provides a topology optimization with lattice filling.
[0285] It should be noted that some implementations include either the first option or the second option, while some implementations include both options and allow the user to select the type of hybrid lattice topology optimization to be performed for any given design space based on aesthetics, structural integrity, or other reasons. In addition, some implementations also include a selection of how to optimize 444 the variable thickness of beams in the lattice after optimizing 442 the topology of the structure while the lattice thickness remains constant. In some implementations, the type of lattice thickness optimization performed depends on the type of topology optimization previously performed (e.g., different types of thickness optimizations may be performed given potential inaccuracies in the homogenization process (e.g., for coarse resolution lattices).
[0286] For example, a check may be made at 452 for the resolution of the lattice used during topology optimization 442, and if a fine resolution lattice was used, option 454 may be used (e.g., lattice density optimization using Algorithm 5); otherwise, option 456 may be used (e.g., lattice thickness optimization using Algorithm 2). It should be noted that the resolution of the lattice that constitutes a "fine resolution lattice" may be determined experimentally in a given implementation, and / or a heuristic rule may be used (e.g., lattice unit cell volume / total volume < 0.02 means the lattice is fine).
[0287] A detailed example of how to perform hybrid topology optimization using the level set method is now described. To simplify the algorithm, a compliance minimization problem with a penalty volume (Eq. 26) can be used.
[0288] Topology optimization stage
[0289] An exemplary solid lattice design process is listed in Algorithm 3, where the starting shape and level set are exactly the same as those of Algorithm 1. Unlike the lattice and skin optimization (Algorithm 2), the starting lattice thickness (and corresponding volume fraction) can be set to a high value close to the maximum allowed value. The corresponding lattice constitutive model can be obtained using the method outlined above (see Homogenization). The FEA model used for simulation includes solid elements throughout the solid region of the generative design space being optimized. Next, the shape can be iteratively optimized until the objective function has converged. In each iteration, the constitutive model D of each element e in the FEA model can be updated as in the basic topology optimization method, but the elements outside the level set can be given a lattice constitutive model D l , and the constitutive model of the elements inside the level set can be set to the constitutive model of the original solid material D s The remaining steps are exactly the same as those of Algorithm 1 until the final level set surface defining the final solid region is extracted.
[0290] Algorithm 3-Solid Lattice Hybrid Topology Optimization
[0291] Input: (T, ut min , t max ,Ω,D s , μ)
[0292] Output: (Ω1, Ω s )
[0293] / / Initialize the level set ψ according to the design space
[0294] 1.ψ0=f SDF (Ω)
[0295] / / Set the initial lattice thickness
[0296] 2.t0=t min , +.75(t max -t min )
[0297] / / Calculate the initial lattice volume fraction and obtain the lattice constitutive model D l
[0298] 3.
[0299] 4.D l ←f RVE (T,u,x l )
[0300] / / Iterate until the convergence tolerance c is met
[0301] 5. i = 0
[0302] 6. When i = 0 or |J i -J i-1 |>c
[0303] 7.
[0304] / / Formulate and solve FEA / problems
[0305] 8.
[0306] 9.u=K -1 f
[0307] / / Calculate the advection rate
[0308] 10.
[0309] / / Solve the Hamiltonian Jacobian and obtain the new level set ψ i+1
[0310] 11.
[0311] 12.ψ i+1 ←ψ i
[0312] / / Calculate the target
[0313] 13.
[0314] / / Increment the iteration
[0315] 14.i←i+1
[0316] 15. End the loop
[0317] / / Get the final level set as the solid area
[0318] 16.
[0319] / / Get the final lattice area
[0320] 17.Ω t ←Ω\Ω l
[0321] The lattice area can be determined as the difference between the original design space and the solid area. In some implementations, the OpenVDB method can be used to extract the final level set. It should be noted that the grid, B-Rep or level set space can be used to extract the final level set in a form that can be used in the next stage.
[0322] The lattice void design process (Algorithm 4) differs from the solid lattice process (Algorithm 3) in two steps. For example, in step 7, the element update rule can be changed to set the lattice constitutive model D of the elements inside the level set. t In addition, elements outside the level set (void elements) can be given a very low stiffness D v .
[0323] Algorithm 4 - Lattice-Void Hybrid Topology Optimization
[0324] enter:
[0325] Output: (Ωx)
[0326] / / Initialize the level set ψ according to the design space
[0327] 1.ψ0=f SDF (Ω)
[0328] / / Set the initial lattice thickness to 75% of the maximum thickness
[0329] 2.
[0330] / / Calculate the initial lattice volume fraction and obtain the lattice constitutive model D l
[0331] 3.
[0332] 4.D l ←f RVE (T,u,x l )
[0333] / / Iterate until the convergence tolerance c is met
[0334] 5. i = 0
[0335] 6. When i = 0 or |J i -J I |>c
[0336] / / Set the constitutive model of FEA element to lattice D l or gap D v
[0337] 7.
[0338] / / Formulate and solve FEA problems
[0339] 8.
[0340] 9.u=K -1 f
[0341] / / Calculate the advection rate
[0342] 10.
[0343] / / Solve the Hamiltonian Jacobian and obtain the new level set ψ i+1
[0344] 11.
[0345] 12.ψ i+1 ←ψ i
[0346] / / Calculate the advection rate
[0347] 13.
[0348] 14.i←i+1
[0349] 15. End the loop
[0350] / / Get the final level set as the lattice region
[0351] 16.
[0352] Finally, in step 16, the 0th contour surface of the final level set may define a lattice region. In some implementations, the OpenVDB method may be used to extract the final level set. It should be noted that a grid, B-Rep, or level set space may be used to extract the final level set in a form that can be used in the next stage.
[0353] Thickness optimization stage
[0354] Once an optimized topology of the lattice region is obtained using Algorithm 3 or 4, the lattice thickness optimization can be performed using the Lattice and Skin Thickness Optimization process (see Lattice and Skin Optimization above). This may require a change in the FEA model where solid elements with lattice RVEs are replaced with beam elements before the thickness is optimized. Alternatively, solid elements with equivalent lattice properties (RVE) can be used for thickness optimization, as outlined in Algorithm 5. It should be noted that this approach may be more relevant when a fine resolution lattice is used. Such a lattice may result in a large number of beam elements in the FEA model when the thickness optimization process is used (see Lattice and Skin Optimization above).
[0355] In Algorithm 5, the thickness design variable can be represented by the volume fraction at each solid element in the lattice region. At the beginning of the algorithm, the thickness value is initialized to the minimum thickness t min. The FEA model used for simulation may represent the lattice as solid elements with equivalent lattice properties (RVE). During each iteration of the optimization process, the RVE for each real element may be recalculated based on the current value of the corresponding thickness design variable. This may be done by first converting the thickness to a volume fraction using Equation 39. The RVE may then be recalculated (see Homogenization above). Once the FEA model is solved, the stress ratios at each solid element may be calculated. The same full stress design approach as Algorithm 2 may be used to update the thickness using the stress ratios for the next iteration.
[0356] At the end of the optimization, a real lattice geometry can be created in the Lattice region, where the thickness values at each beam are mapped from the solid mesh used for the simulation. For this purpose, the same linear mapping between FEA nodes and level set grid points can be used (see Body-fitted Solver above).
[0357] Algorithm 5- Lattice Density Optimization
[0358] Input: (T, u, t min , t max ,Ω l , σ t )
[0359] Output:
[0360] / / Initialize the lattice thickness at all elements in the lattice region to the minimum thickness
[0361] 1.
[0362] / / Iterate until the convergence tolerance c is met
[0363] 2. i = 0
[0364] 3. When i = 0 or |J i -J i-1 |>c
[0365] / / Calculate volume fraction and update lattice constitutive model
[0366] 4.
[0367] 5.D l (e)←f RVE (T,u,x l (e))
[0368] / / Formulate and solve FEA problems
[0369] 6. K = f Ω B T D l (e)B
[0370] 7.u=K -1 f
[0371] / / Calculate the stress ratio at each vertex
[0372] 8.
[0373] / / Calculate thickness increment based on stress ratio
[0374] / / (e.g., α=.5, β=0.1, γ=0.5)
[0375] 9.
[0376] / / Update the thickness without violating bounds
[0377] 10.
[0378] / / Calculate the target
[0379] 11.
[0380] / / Increment the iteration
[0381] 12.i←i+1
[0382] 13. End the loop
[0383] / / Create the lattice after mapping the thickness at the solid elements to the level set grid
[0384] 14.Ωl(T,u,f,t)∈Ω
[0385] It should be noted that although the detailed examples above use level set representations and level set methods, other macrostructure type topology optimization systems and techniques may be used.
[0386] Topology generation using lattices for CAD applications
[0387] Figure 4A As shown (for example, by Figure 1A 1 , an example of a process of optimizing the topology of a lattice packing for one or more portions of a 3D model of an object being generatively designed for manufacturing (performed by a CAD program 116 of a computer program). Design criteria, boundary conditions, and / or other design variables are obtained 400, for example, by the CAD program 116. This may involve receiving user input, for example, through the UI 122 on the display device 120, importing information from another program or third party source, and / or one or more of these inputs may be predefined in a given implementation.
[0388] Boundary conditions can be mechanical problem definitions and input entities that are retained to serve as connection points with other components in a larger 3D model or a separate 3D model. In some implementations, obtaining 400 includes, for example, receiving input 400 through a UI 122 on a display device 120, the input 400 indicating an initial 3D model that defines at least a portion of a design space for topological optimization of an object. In some implementations, for example, a user is enabled to select multiple parts from an assembly shown in the UI through a UI 122 on a display device 120, and the user is enabled to indicate load points and areas away from the design. Then, the software (e.g., a CAD program 116) can use the inverse of these as a design space. Other inputs, such as a selected lattice topology, may also be received 400.
[0389] Design criteria may include one or more design objectives and one or more design constraints. Design objectives may include those that minimize one or more cost functions of the object to be manufactured and / or maximize one or more utility functions thereof. In some implementations, the design objectives include one or more of the following: (1) minimizing material use (e.g., waste); (2) minimizing weight; (3) minimizing manufacturing time; (4) maximizing tool life; (5) maximizing the ability to remove support material; (6) targeting natural frequencies (e.g., achieving higher / lower natural vibration modes for vibration reduction); and (7) positioning and packaging of parts within the structure (e.g., electronic components for circuit actuation (including reserving space for wiring between components in the assembly, limiting wiring length) and / or components for structural health monitoring, and component considerations to ensure that each component can be added / removed independently or in a limited number of steps). Design constraints may include those that address manufacturing requirements and / or address usage requirements. In some implementations, the design constraints include one or more of the following: (1) manufacturing constraints that restrict geometries to those that are manufacturable (e.g., minimum self-support angles for AM processes in which separate support material is not used, subject to build orientation when appropriate; or mechanical access constraints for AM or SM processes, such as 3-axis, 5-axis, and / or 2.5-axis CNC constraints); (2) thermomechanical constraints; (3) material fraction or variation constraints; (4) structural integrity; (5) stiffness; (6) compliance; (7) stress criteria (e.g., von Mises or Tsai-Wu stresses for manufacturing or use); (8) physical constraints (not interfering with larger part systems); and (9) failure constraints (e.g., brittle elastic fractional resistance). In some implementations, the design criteria include minimizing compliance with volume, stress, displacement, buckling safety factor, and / or natural frequency constraints, but the method can be generalized to other design objectives and constraints, including minimizing volume, deviation from a target (compliant mechanism) (e.g., least squares deviation), and / or design-related loads (pressure loads).
[0390] In some implementations, for example, Figure 1AThe CAD program 116 obtains 400 a design space for an object to be manufactured, a setup for a physical simulation of the object, at least one design goal for the object, and at least one design constraint for the object. The physical simulation performed by the systems and techniques described in this document can simulate one or more physical properties, and one or more types of simulations can be used. For example, FEA (including linear static FEA), finite difference methods, and material point methods can be used. In addition, simulations of physical properties can include simulating buckling, natural frequencies, heat flux, electrical flux or electromagnetic flux, and material solidification properties, among other possibilities. It should be noted that the setup for the physical simulation will be different for different physical simulations.
[0391] For example, the CAD program 116 iteratively modifies 405 both the three-dimensional topology of the generative model of the object and one or more external shapes of the three-dimensional topology using a generative design process that employs a level set topology optimization method and a level set representation of the generative model in the design space. For example, this may include using a gradient-based algorithm to optimize the 3D topology and shape of the object model. In some implementations, the iterative modification 405 involves using the above combined Figure 3A (mixed lattice) and FIG. 4B to FIG. 4F Describes the systems and techniques.
[0392] In a hybrid lattice approach for topology optimization, a homogenized lattice material representation is used, where the homogenized lattice material expresses the stiffness of a lattice consisting of beams and void regions as a function of the lattice topology and the volume fraction of the lattice, for example using RVEs corresponding to different lattice topologies, as described above. It should be noted that the volume fraction is a function of the unit cell size and beam thickness, and that alternative lattice arrangements can be used, as mentioned above. Various approaches to generating RVEs can be used. In general, a homogenized lattice material representation can express the structural behavior of a given lattice as an anisotropic solid material, which is a continuous material with properties approximately equivalent to a given lattice. Various homogenization methods can be used, such as numerical homogenization and analytical homogenization, and / or experimental data can be used to define the homogenized lattice material representation.
[0393] In some implementations, iteratively modifying 405 includes adjusting 405A a volume fraction of a homogenized lattice material representation according to a constitutive matrix of the homogenized lattice material representation of the lattice topology. In some implementations, the beam thickness is varied while keeping the unit size constant. In some implementations, the unit size is varied, but this may result in a lattice with mixed unit sizes, which may present challenges during manufacturing. In some implementations, a uniform lattice is used during the initial topology optimization.
[0394] In any case, the current iteration of the model being iteratively modified is changed 405B for physical simulation. This may involve: changing 405B the constitutive model for physical simulation according to the current iteration of the level set representation and the homogenized lattice material representation at a uniform thickness of the beams in the lattice. In some implementations, changing 405B the constitutive model for physical simulation includes: such as described above in conjunction with Algorithm 4, according to the current iteration of the level set representation, (i) using void simulation elements outside one or more external shapes of the three-dimensional topology, and (ii) using a homogenized lattice material representation for simulation elements inside one or more external shapes of the three-dimensional topology. For example, the void elements may have a constitutive model in which the Young's modulus is reduced to about 0.001 of the Young's modulus of a conventional solid material (see Equations 43 and 44), and the lattice elements may use a constitutive model of the lattice, as explained above under the section headings Homogenization, Variation with Lattice Volume Fraction, and Predicting RVE for Arbitrary Lattice Properties.
[0395] In some implementations, changing 405B the constitutive model for physical simulation includes: according to the current iteration of the level set representation, (i) using a homogenized lattice material representation for simulation elements outside one or more external shapes of the three-dimensional topology but inside the boundary of the design space, and (ii) using solid simulation elements inside one or more external shapes of the three-dimensional topology. It should be noted that in FEA, each element has a stiffness matrix that contributes to the overall stiffness matrix. After inverting this stiffness matrix, the solution (stress, displacement, etc.) can be obtained. When formulating the element stiffness matrix, the solid element stiffness matrix can use a linear elastic material model, while the lattice element can use a lattice RVE constitutive model. In addition, the boundary of the design space can be defined by an initial three-dimensional model (e.g., as specified by a user).
[0396] A physical simulation of the object is performed 405C using the altered constitutive model to generate a physical evaluation relative to at least one design objective and at least one design constraint. As mentioned above, various types of physical simulations can be used, and therefore various types of physical evaluations can be generated. In some implementations, performing 405C the physical simulation involves performing a finite element analysis simulation using a body-fitting mesh-based solver, and the constitutive model is altered 405B by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used for the finite element analysis simulation. In some implementations, the solid mesh-based simulation can be non-body-fitting, for example, solved using a voxel solver.
[0397] The shape change rate of the level set representation in the current iteration is calculated 405D based on the physical evaluation. It should be noted that this is where the different objectives and constraints come into play. In some implementations, calculating 405D the shape change rate involves using the augmented Lagrangian method described herein to calculate the advection rate, such as when there are multiple shape derivatives. Then, before the next iteration, the shape change rate is used to update 405E the level set representation of the object, for example, by a level set method. For example, updating 405E can use the advection rate obtained by the augmented Lagrangian method (ALM).
[0398] After the modification 405 completes the change to both the three-dimensional topology and one or more external shapes of the three-dimensional topology, the thickness of the beams in the lattice or the density of the lattice is adjusted 410, for example, by the CAD program 116, based on the physical simulation, at least one design goal, and at least one design constraint, while keeping the three-dimensional topology constant. For example, see the thickness optimization in step 10 of Algorithm 2 and Algorithm 5, and also see step 4 of Algorithm 5, where the thickness is converted to an equivalent density. It should be noted that the lattice thickness optimization can be performed using the density.
[0399] After adjusting the lattice for the optimized 3D topology, a three-dimensional model of the object may be provided 415 based on the three-dimensional topology, one or more external shapes of the three-dimensional topology, and the adjusted lattice for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems. Figure 1B Provide 195 as described. Thus, providing 415 may include generating instructions (e.g., tool path instructions) for an additive manufacturing machine, a subtractive manufacturing machine, or other machine, system, or device. It should be noted that the described process need not be used with any particular manufacturing system or technology.
[0400] return Figure 3A , when the generative process to be used employs a hollow level set method for topology optimization, a level set is initialized 308A for the design space. It should be noted that the level set method is an example of a macrostructure generative modeling technique, where the generative model represents the object being designed as one or more boundaries between one or more solid regions (having homogeneous material therein) and one or more void regions (having no material therein) within the design space. However, in the case of hollow topology optimization, as described in this document, a physical simulation of the modeled object is performed 308B, where at least a portion of the solid regions are treated as having voids therein (i.e., the macrostructure generative modeling technique treats it as a solid, and the physical simulation treats it as something containing a hollow region).
[0401] The level set is updated 308C using the results from the simulation and checked 308D for convergence. The simulation, updating, and checking process is then iterated until convergence. The hollow topology optimization method can be motivated by the observation that common non-solid structural elements such as I-beams, sandwich panels, and tubes can achieve mass reduction by placing material away from the neutral axis.
[0402] Compared to solid structural elements, material near the neutral axis can be removed and / or redistributed farther away to become lighter and / or stiffer. Non-solid structural elements can often achieve better strength-to-weight ratios than their solid counterparts, especially when curved. But rather than using the naive approach of running a basic topology optimization with a high safety factor and / or volumetric goal, and then manually converting to a hollow structure using standard geometry operations, the hollow topology optimization process can consider the effects of the hollow interior during topology optimization, which can result in improved structural designs, such as stronger or more durable parts, less material usage, or a combination thereof, such as improved strength-to-weight ratios.
[0403] Compared to keeping the skin thickness of the hollow structure constant or using only the full stress design algorithm (e.g., Algorithm 2) to optimize the skin thickness to produce a variable skin thickness, this can provide benefits because the topology optimization stage of the described hollow topology optimization process can complement or benefit from the hollow process and thickness optimization. Compared to adding cavities to the generated geometry after performing solid topology optimization based on the overall solid geometry (which can significantly change the structural response during topology optimization and / or lose any optimal properties enjoyed by the solid geometry), this can improve the structural response of the completed design. In addition, it should be noted that the described hollow topology optimization does not require the topology changes to be limited to the shell surface. In addition to the external topology of the solid material, the internal topology of the hollow can also be optimized.
[0404] A detailed example of this type of macrostructure topology optimization process is now described, in which topology optimization and hollow structure physics simulation can be used together to produce a final design that contains solid material around one or more hollow cavities. Level set-based topology optimization methods can be modified to produce hollow designs, such as Figure 5C and Figure 5D shown. Figure 5C A graphical representation of an example of a hollow design 560 containing an internal cavity is shown. Figure 5D A graphical representation of an example of a slice view 562 of a hollow design 560 is shown, thereby exposing an internal cavity 564. Hollow topology optimization may be performed using a level set method as described below. To simplify the algorithm, a compliance minimization problem with a penalty volume may be used (Equation 26).
[0405] Design Workflow
[0406] Figure 5B As shown, for example, by Figure 1A 16. The process 540 is an example of a design workflow for hollow topology optimization performed by the CAD program 116 of FIG. 16. The process 540 is similar to the process 440 (see hybrid topology optimization above) in that the process 540 has two main steps. The design workflow can be used to maintain the hollow skin thickness t in the first step. h Topology optimization is performed while being constant. At step 542, a uniform thickness skin is created. For example, at step 546, hollow topology optimization may occur (e.g., using Algorithm 6). At step 544, thickness optimization occurs. At 548, a check is made as to whether a thick skin is to be used. If a thick skin is to be used, then at 550, the basic topology optimization algorithm 1 (with slight modifications) or the skin thickness optimization algorithm 7 may be used. Otherwise, at 552, the lattice and skin thickness optimization algorithm 2 may be used. The selection at 548 may depend on whether shell elements are to be used in the lattice and skin thickness optimization (Algorithm 2) (which is not convenient for thick skin (due to the large hollow thickness t h For thick shells, Algorithm 7 or a slightly modified Algorithm 1 may be used. It should be noted that the thickness of the shell that constitutes a "thick shell" may be determined based on the aspect ratio of the shell elements, such that when the aspect ratio (maximum surface length / thickness) exceeds a predetermined value (e.g., as determined experimentally and / or by a heuristic rule in a given implementation), the shell is thin, otherwise the shell is thick.
[0407] Figure 5E A graphical representation of an example of a hollow level set 570 is shown. i|0 The 0 contour line 572 is offset by an amount t in the negative direction of the normal vector n 574. h 576 to create an internal level set. The 0th contour line 578 of the internal level set is indicated as ψ c|0 . Fig. 5F A graphical representation showing the transition of the hollow level set 570 from hollow to basic topology optimization (TO). Positions 582 and 584 are not hollow (I c =0). However, position 586 is hollow (I c =1). It should be noted that Figure 5E and Fig. 5F Helps explain one of the terms in the derivative of the hollow shape as described.
[0408] Therefore, a hollow design can be achieved by offsetting the 0th contour of the design level set in the negative normal vector direction to create a hollow skin. h is called the hollow thickness and can be kept constant during the topology optimization phase. This hollow thickness t hThis may be specified by the user at the start of the hollow topology optimization, for example, via UI 122. The offset process may produce two level sets. The outer level set defines the outer surface of the design, and the inner level set defines the surface of the interior cavity. The hollow skin thickness optimization stage performed using Algorithms 2 or 7 may require the creation of a skin. Therefore, the final design from the uniform thickness hollow topology optimization (Algorithm 6) may need to be divided into solid regions and skin regions. The first step is to use the hollow indicator function I from the last iteration of the topology optimization stage as follows c Limit the skin surface area:
[0409] Γ h : = {x∈Γ|I c (x)=1} (56)
[0410] Next, the same thickness t used during the hollow topology optimization can be used h To create the skin Γ h (t h , Γ h ). For this purpose, the skin creation process can be used (see Lattice and Skin Creation above). The solid area can now be given as the difference between the solid area at the end of the hollow topology optimization and the newly created skin.
[0411] Topology optimization stage
[0412] The design process used during the topology optimization phase is outlined in Algorithm 6, where the starting shape and level set are exactly the same as those of Algorithm 1. The FEA model used for simulation includes solid elements in the non-hollow portion of the solid region of the generative design space being optimized. In each iteration, a hollow design can be created (see the design workflow above). Next, the constitutive model D of each element e in the FEA model can be updated as in the basic topology optimization method, but when updating the constitutive model of each element, the difference between the external and internal level sets (i.e., ψ ) (e.g., calculated using a Boolean difference operation) is used. d =Ψ i \ψ c ).
[0413] The shape derivative in step 10 uses some additional terms to take into account the hollow regions. The exact expression for the advection rate is given by:
[0414]
[0415] where v(x) is the advection rate from the basic topology optimization (Eq. 28), and K1, K2 represent the Gaussian curvature and mean curvature at the sampling point x, respectively. Operator I c= {0, 1} allows for a smooth transition between non-hollow topology optimization (using the advection rate from the base topology optimization) and hollow topology optimization. For examples of indicating hollow and non-hollow regions, see Fig. 5F Further details on shape derivative calculations that may be used are described below with reference to "Shape Derivatives for Thick Shell Representations".
[0416] It should be noted that in Algorithm 6, the topology optimization only advects the external level set ψ in step 11. i , the internal level set ψ c This can usually be created when updating and solving the FEA model during optimization iterations. This ensures that only the skin-like solid element layer contributes to the stiffness of the design. The remaining steps can be the same as those of Algorithm 1 until the final solid region is extracted.
[0417] Algorithm 6-Hollow Topology Optimization
[0418] Input: (Ω, D v , D s , μ, t h )
[0419] Output: (Ω s )
[0420] / / Initialize the level set ψ according to the design space
[0421] 1.ψ0=f SDF (Ω)
[0422] / / Iterate until the convergence tolerance c is reached
[0423] 2. i = 0
[0424] 3. When i = 0 or |J i -J i-1 |>c
[0425] / / Offset the design level set to create a hollow level set
[0426] 4.ψ c : = {ψ c (x) = ψ i (x)-t h}
[0427] 5.ψ d =ψ i \ψ c
[0428] / / Set the constitutive model of the FEA element to void D v or entity D s
[0429] 6.
[0430] / / Formulate and solve FEA problems
[0431] 7.
[0432] 8.u=K -1 f
[0433] / / Calculate hollow shape derivatives and advection rates
[0434] 9.
[0435] 10.
[0436] / / Solve the Hamiltonian Jacobian and obtain the new level set ψ i+1
[0437] 11.
[0438] 12.ψ i+1 ←ψ i
[0439] / / Calculate the target
[0440] 13.
[0441] / / Increment the iteration
[0442] 14.i←i+1
[0443] 15. End the loop
[0444] / / Get the final hollow level set
[0445] 16.ψ c : = {ψ c (x) = ψ i (x)-t h}
[0446] / / Get the final entity area
[0447] 17.
[0448] The extraction of the final solid region can be defined by the 0th contour of the Boolean difference between the external level set and the internal level set. In some implementations, the OpenVDB method can be used to extract the final level set. It should be noted that the final level set can be extracted in a form that can be used in the next stage using a grid, B-Rep or level set space.
[0449] Skin thickness optimization stage
[0450] The second stage 544 of the hollow topology optimization may include optimizing the thickness of the hollow region using Algorithms 2, 1, or 7. It should be noted that Algorithms 1 and 7 may be more relevant when thick skins are used because both algorithms may represent the skin regions using solid elements in the FEA model. When using Algorithm 2, the skin may be represented as shell elements in the FEA model. The numerical accuracy of Algorithm 2 may be reduced when thick shells are represented.
[0451] When using the basic level set algorithm 1, the internal level set ψ representing the cavity from the final iteration can be calculated while keeping the external shape constant. c Perform topology optimization. Note that (given that the interior level set has a negative sign inside the cavity and a positive sign in the domain) the FEA element update rule in step 4 may need to be changed as follows:
[0452]
[0453] The other steps in Algorithm 1 can remain unchanged. This process can efficiently optimize the shape of the cavity defining the hollow design, which can produce more than one cavity in the generated design.
[0454] Using the previously described lattice and skin thickness optimization process (see Lattice and Skin Optimization above) may require changes to the FEA model where solid elements are replaced with shell elements before optimizing thickness. In some implementations, thickness optimization is performed using solid elements as outlined in Algorithm 7. Algorithm 7 is the skin equivalent of the lattice density optimization (Algorithm 5). The design variables may include thickness values for each polygon vertex in the skin region, which are initialized at the beginning of the algorithm to the minimum thickness t min During each iteration of the optimization process, a level set can be created from the current skin domain, and the level set can be used to update the constitutive model of the solid elements in the FEA model. Once the FEA model is solved, the stress values at each skin polygon can be obtained using the same mapping created to transfer data from the FEA model to the level set grid (see the body-fitted solver above). The remaining steps can be the same as for the lattice and skin optimization (Algorithm 2), except that the skin domain needs to be updated at the end of each iteration.
[0455] Algorithm 7-Skin Thickness Optimization
[0456] Input: D v , D s , t min , t max ,Ω h , σ t
[0457] Output:
[0458] / / Initialize the skin thickness of all elements in the skin area to the minimum thickness
[0459] 1.
[0460] / / Iterate until the convergence tolerance c is met
[0461] 2. i = 0
[0462] 3. When i = 0 or |J i -J i-1 |>c
[0463] / / Create a level set based on the skin
[0464] 4.ψ h =f SDF (Ω h )
[0465] / / Set the constitutive model of the FEA element to void D v or entity D s
[0466] 5.
[0467] / / Formulate and solve FEA problems
[0468] 6. K = f Ω K e
[0469] 7.u=K -1 f
[0470] / / Calculate the stress ratio at each vertex
[0471] 8.
[0472] / / Calculate thickness increment based on stress ratio
[0473] / / Use, for example, α=0.5, β=0.1, β=0.33
[0474] 9.
[0475] / / Update the thickness without violating the boundaries
[0476] 10.
[0477] / / Calculate the target
[0478] 11.
[0479] / / Update the skin using the current thickness value
[0480] 12.Ω h ←Ω h (t,Γ h )
[0481] / / Increment the iteration
[0482] 13.i←i+1
[0483] 14. End the loop
[0484] / / Create the lattice after mapping the thickness at the solid elements to the level set grid
[0485] 15.Ω l (T,u,f,t min )∈Ω
[0486] It should be noted that although the detailed examples above use level set representations and level set methods, other macrostructure type topology optimization systems and techniques may be used.
[0487] Hollow topology generation for CAD applications
[0488] Figure 5A Shown (for example, as shown by Figure 1A 1 , an example of a process of optimizing the topology of one or more hollow portions of a 3D model of an object being generatively designed for manufacturing (performed by a CAD program 116 of a computer program). Design criteria, boundary conditions, and / or other design variables are obtained 500, for example, by the CAD program 116. This may involve receiving user input, for example, via the UI 122 on the display device 120, importing information from another program or third-party source, and / or one or more of these inputs may be predefined in a given implementation.
[0489] Boundary conditions can be mechanical problem definitions and input entities that are retained to serve as connection points with other components or separate 3D models in a larger 3D model. In some implementations, obtaining 500 includes, for example, receiving input 500 through a UI 122 on a display device 120, the input 500 indicating an initial 3D model of at least a portion of a design space defining topology optimization for an object. In some implementations, for example, a user is enabled to select multiple parts from the components shown in the UI through a UI 122 on a display device 120, and the user is enabled to indicate load points and areas away from the design. Then, the software (e.g., a CAD program 116) can use the inverse of these as a design space. Other inputs may also be received 500, such as a hollow structure offset value input (e.g., a constant offset selected by a user of the CAD program 116 to define a hollow structure during topology optimization).
[0490] Design criteria may include one or more design objectives and one or more design constraints. Design objectives may include those that minimize one or more cost functions of the object to be manufactured and / or maximize one or more utility functions thereof. In some implementations, the design objectives include one or more of the following: (1) minimizing material use (e.g., waste); (2) minimizing weight; (3) minimizing manufacturing time; (4) maximizing tool life; (5) maximizing the ability to remove support material; (6) targeting natural frequencies (e.g., achieving higher / lower natural vibration modes for vibration reduction); and (7) positioning and packaging of parts within the structure (e.g., electronic components for circuit actuation (including reserving space for wiring between components in the assembly, limiting wiring length) and / or components for structural health monitoring, and component considerations to ensure that each component can be added / removed independently or in a limited number of steps). Design constraints may include those that address manufacturing requirements and / or address usage requirements. In some implementations, the design constraints include one or more of the following: (1) manufacturing constraints that restrict geometries to those that are manufacturable (e.g., minimum self-support angles for AM processes in which separate support material is not used, subject to build orientation when appropriate; or mechanical access constraints for AM or SM processes, such as 3-axis, 5-axis, and / or 2.5-axis CNC constraints); (2) thermomechanical constraints; (3) material fraction or variation constraints; (4) structural integrity; (5) stiffness; (6) compliance; (7) stress criteria (e.g., von Mises or Tsai-Wu stresses for manufacturing or use); (8) physical constraints (not interfering with larger part systems); and (9) failure constraints (e.g., brittle elastic fractional resistance). In some implementations, the design criteria include minimizing compliance with volume, stress, displacement, buckling safety factor, and / or natural frequency constraints, but the method can be generalized to other design objectives and constraints, including minimizing volume, deviation from a target (compliant mechanism) (e.g., least squares deviation), and / or design-related loads (pressure loads).
[0491] In some implementations, for example, Figure 1AThe CAD program 116 obtains 500 a design space for an object to be manufactured, a setup for a physical simulation of the object, at least one design goal for the object, and at least one design constraint for the object. The physical simulation performed by the systems and techniques described in this document can simulate one or more physical properties, and one or more types of simulations can be used. For example, FEA (including linear static FEA), finite difference methods, and material point methods can be used. In addition, simulations of physical properties can include simulating buckling, natural frequencies, heat flux, electrical flux or electromagnetic flux, and material solidification properties, among other possibilities. It should be noted that the setup for the physical simulation will be different for different physical simulations.
[0492] For example, the CAD program 116 uses a generative design process to iteratively modify 505 both the three-dimensional topology of the generative model of the object and one or more external shapes of the three-dimensional topology, and the generative design process represents the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions in the design space. For example, this may include using a gradient-based algorithm to optimize the 3D topology and shape of the object model, or using other algorithms such as evolutionary algorithms and random algorithms. In addition, the generative model can be a level set representation of one or more external shapes of the three-dimensional topology, and a level set topology optimization method can be used. In some implementations, the iterative modification 505 involves using the above combined Figure 3A (Hollow) and FIG. 5B to FIG. 5F Describes the systems and techniques.
[0493] The current iteration of the model being iteratively modified in 505A is changed for physical simulation. This may involve changing 505A the constitutive model for physical simulation according to the current iteration of the three-dimensional topology and one or more external shapes of the three-dimensional topology to limit the hollow structure in the three-dimensional topology. For example, 505A Boolean difference operations can be performed. In some implementations, the offset of the hollow structure is a constant offset (as mentioned above) selected by the user of the computer-aided design program, thereby setting the initial thickness of the hollow structure. In some implementations, the initial thickness is retrieved from a database of known good values for a given manufacturing process. Typically, the manufacturing process determines the minimum wall thickness.
[0494] Performing 505B a physical simulation of the object using the altered constitutive model to generate a physical evaluation relative to at least one design objective and at least one design constraint. As mentioned above, various types of physical simulations can be used, and therefore various types of physical evaluations can be generated. In some implementations, performing 505B the physical simulation involves performing a finite element analysis simulation using a body-fitting mesh-based solver, and altering 505A the constitutive model by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used for the finite element analysis simulation. In some implementations, the solid mesh-based simulation can be non-body-fitting, for example, solved using a voxel solver.
[0495] Calculate 505C the shape change rate of one or more external shapes of the three-dimensional topology in the current iteration based on the physical evaluation. For example, this may involve calculating 505C hollow shape derivatives and advection rates based on lines 9 and 10 in Algorithm 6. It should be noted that this is where different objectives and constraints come into play. In some implementations, calculating 505C shape change rates involves using the augmented Lagrangian method described herein to calculate advection rates, such as when there are multiple shape derivatives.
[0496] Then, before the next iteration, the generative model of the object is updated 505D using the shape change rate. For example, updating 505D may use the advection rate obtained by the augmented Lagrangian method (ALM). In some implementations, a level set method is used, such as shown above in lines 11 and 12 in Algorithm 6. In some implementations, non-level set surface manipulations may be used, such as B-Rep control point movement or node movement in a T-spline or mesh.
[0497] After modifying 505 to change both the three-dimensional topology and one or more external shapes of the three-dimensional topology, the thickness of the hollow structure is adjusted 510, for example, by the CAD program 116, based on physical simulation, at least one design goal, and at least one design constraint while keeping the one or more external shapes of the three-dimensional topology constant. For example, see Algorithm 2, Algorithm 1, and Algorithm 7, and Figure 5B .
[0498] In some implementations, adjusting 510 the thickness of the hollow structure includes: starting from one or more internal shapes of the hollow structure, iteratively modifying the one or more internal shapes of the three-dimensional topology using a generative design process, including changing a constitutive model for physical simulation according to a current iteration of the three-dimensional topology and the one or more internal shapes, so that after completing the modification of the one or more internal shapes, the thickness of the hollow structure varies across the three-dimensional topology. In some implementations, a level set method is used to optimize the external shape and / or internal shape of the hollow structure.
[0499] Thus, the generative model may be a level set representation of one or more internal shapes and one or more external shapes of a three-dimensional topology, and the generative design process may employ a level set topology optimization method to iteratively modify the one or more internal shapes and one or more external shapes of the three-dimensional topology. In some implementations, other macrostructure representations may be used for generative design, wherein the generative design process represents the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space. For example, other options for representing the contours of hollow regions include other surface modeling techniques. For example, the outer surface may be a mesh, a T-spline, or a B-Rep surface with an auxiliary inner surface created with a fixed offset, and the outer surface may be shape optimized using a classical shape optimization method.
[0500] Additionally, in some implementations, (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during fabrication, and / or (iv) visibility of holes for aesthetic considerations may be identified 515 in different regions of the three-dimensional topology of the hollow structure. Then, for a subset of the different regions compared to the remaining regions, the location of one or more holes in the three-dimensional topology of the hollow structure for forming or removing support material used during fabrication may be determined 520 based on the identified (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during fabrication, and / or (iv) visibility of holes for aesthetic considerations.
[0501] After adjusting the thickness of the hollow structure (and potentially after determining one or more positions of the one or more holes), a three-dimensional model of the object can be provided 525 based on the three-dimensional topology, one or more external shapes of the three-dimensional topology, and the adjusted thickness of the hollow structure for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems. This can involve as described above in conjunction with Figure 1B The described providing 195. Thus, providing 525 may include generating instructions (e.g., tool path instructions) for an additive manufacturing machine, a subtractive manufacturing machine, or other machine, system, or device. It should be noted that the described process need not be used with any particular manufacturing system or technology.
[0502] return Figure 3A, when the generative process to be used employs a hybrid-hollow level set method for topology optimization, a level set is initialized 310A for the design space. It should be noted that the level set method is an example of a macrostructure generative modeling technique, in which the generative model represents the object being designed as one or more boundaries between one or more solid regions (having homogeneous material therein) and one or more void regions (having no material therein) within the design space. However, in the case of a hybrid-hollow topology optimization, as described in this document, a physical simulation of the modeled object is performed 310B, in which at least a portion of the solid region is treated as having both partial void regions and complete pores therein (i.e., the macrostructure generative modeling technique treats it as a solid, and the physical simulation treats it as something that partially contains a lattice structure around the hollow region).
[0503] The level set is updated 310C using the results from the simulation and checked 310D for convergence. The simulation, updating and checking process is then iterated until convergence. A detailed example of this type of macrostructure topology optimization process is now described, in which lattice and topology optimization and hollow structure physics simulation are combined to produce a final design that contains a lattice around one or more hollow cavities. The level set-based topology optimization method can be modified to use both hollow topology optimization (in which material is present only near the design surface, and the thickness and topology of the material skin around the internal hollow area are not optimized) and hybrid topology optimization (in which a homogenized lattice representation (such as RVE) is used during topology optimization to create a design with a lattice). By combining the hollow and hybrid topology optimization methods, a 3D model of an object (e.g., a stand-alone structure or a part or component of a larger structure) can be produced that contains both hollow regions and a lattice that is located between one or more hollow cavities at the outer limits of the object.
[0504] Figure 6C A graphical representation of an example of a hybrid-hollow design 660 including an internal cavity and a lattice is shown. Fig.6D A graphical representation of an example of a slice view 662 of a hybrid-hollow design 660 is shown, thereby exposing an internal cavity 664 and an internal lattice 666. A hybrid-hollow topology optimization may be performed using a level set method as described below. To simplify the algorithm, a compliance minimization problem with a penalty volume (Equation 26) may be used.
[0505] Design Workflow
[0506] Figure 6B As shown, for example, by Figure 1A 1. An example of a process 640 for a hybrid-hollow topology optimization workflow executed by the CAD program 116 of FIG. 1. Process 640 is a three-step process in which, in step 642, while maintaining the lattice and the hollow skin thickness th The topology is optimized while being constant. For example, a lattice void topology optimization (TO) 644 may be used, such as in Algorithm 8. During step 646, the hollow skin thickness is optimized, such as using skin thickness optimization 648, such as using Algorithm 4 or Algorithm 7, followed by a lattice thickness optimization step 650, such as using lattice density optimization 652, such as using Algorithm 5.
[0507] Topology optimization stage
[0508] An example of the level set-based hybrid-hollow design process is shown in Algorithm 8, where the starting shape and level set are exactly the same as those in Algorithm 1. The thickness of the hollow skin can be kept constant at t h , and the starting lattice thickness (and corresponding volume fraction) can be set to a high value close to the maximum allowed value. The corresponding lattice constitutive model can be obtained using a homogenization method (see Homogenization above). The FEA model used for simulation contains solid elements in the non-hollow part of the solid region of the generative design space being optimized, where the solid elements actually correspond to the homogenized representation of the lattice (e.g., lattice RVE elements).
[0509] Next, the shape can be optimized iteratively until the objective function has converged. In each iteration, Figure 5B Next, the constitutive model D of each element e in the FEA model may be updated as in the basic topology optimization method, but when updating the constitutive model of each element, the difference between the external and internal level sets (i.e., ψ d =ψ i \ψ c The elements inside this level set (lattice elements) are given the lattice constitutive model D l , while the constitutive model of the external elements can use a very low stiffness D v The shape derivatives used in step 13 may be exactly the same as those used in the hollow topology optimization and advection steps for advection of only the outer level set. The remaining steps may be exactly the same as those of Algorithm 1 until the final level set surface is extracted, which is now defined by the 0th contour of the Boolean difference between the outer level set and the inner level set. This surface may represent a lattice domain. In some implementations, the OpenVDB method may be used to extract the final level set. It should be noted that a grid, B-Rep, or level set space may be used to extract the final level set in a form that can be used in the next stage.
[0510] Algorithm 8-Hybrid-Hollow Topology Optimization
[0511] Input: (T, u, tmin , t max ,Ω,D v , μ, t h )
[0512] Output: (Ω l )
[0513] / / Initialize the level set ψ according to the design space
[0514] 1.ψ0=∫ SDF (Ω)
[0515] / / Set the initial lattice thickness
[0516] 2.t0=t min , +0.75(t max -t min )
[0517] / / Calculate the initial lattice volume fraction and obtain the lattice constitutive model D l
[0518] 3.
[0519] 4.D l ←∫ RVE (T,u,x l )
[0520] / / Iterate until the convergence tolerance c is met
[0521] 5. i = 0
[0522] 6. When i = 0 or |J i -J i-1 |>c
[0523] / / Offset the design level set to create a hollow level set
[0524] 7.ψ c : = {ψ c (x) = ψ i (×)-t h}
[0525] 8.ψ d =ψ i \ψ c
[0526] / / Set the constitutive model of FEA element to lattice D l or gap D v
[0527] 9. / / Formulate and solve FEA problems
[0528] 10.
[0529] 11.u=K -1 f
[0530] / / / / Calculate hollow shape derivatives and advection rates
[0531] 12.
[0532] 13.
[0533] / / Solve the Hamiltonian Jacobian and obtain the new level set ψ i+1
[0534] 14.
[0535] 15.ψ i+1 ←ψ i
[0536] / / Calculate the target
[0537] 16.
[0538] / / Increment the iteration
[0539] 17.i←i+1
[0540] 18. End the loop
[0541] / / Get the final hollow level set
[0542] 19.ψ c : = {ψ(x) = ψ i (x) = t h}
[0543] / / Get the final lattice area
[0544] 20.
[0545] Once the optimized topology of the hollow lattice region is obtained using Algorithm 8, the hollow skin thickness optimization algorithm (Algorithm 7) or a modified version of the lattice-void hybrid topology optimization algorithm (Algorithm 4) can be used for hollow skin thickness optimization.
[0546] When the latter is used, the internal level set ψ representing the cavity from the final iteration can be rewritten while keeping the external shape constant. c Perform topology optimization. Note that given that the interior level set has a negative sign inside the cavity and a positive sign in the domain, the FEA element update rule in step 7 may need to be changed as follows:
[0547]
[0548] The other steps in Algorithm 4 remain unchanged. This process can effectively optimize the shape of the cavity, resulting in a hollow skin with variable thickness. Using Algorithm 7 may also require the following modifications to the FEA element update rule in step 5, while the other steps remain unchanged:
[0549]
[0550] The lattice thickness optimization stage may use Algorithm 5 without modification. It should be noted that the FEA model includes solid elements that may represent the design space. The presence of the lattice may be accounted for by using a homogenized lattice representation such as a lattice RVE (see Homogenization above. Furthermore, it should be noted that although the detailed examples above use a level set representation and level set method, other macrostructure type topology optimization systems and techniques may be used.
[0551] Hollow topology generation using lattices for CAD applications
[0552] Fig. 6A Shown (for example, as shown by Figure 1A 1 , an example of a process of optimizing the topology of a lattice filling for a structure around one or more hollow portions of a 3D model of an object being generatively designed for manufacturing (performed by a CAD program 116 of a computer program). Design criteria, boundary conditions, and / or other design variables are obtained 600, for example, by the CAD program 116. This may involve receiving user input, for example, through the UI 122 on the display device 120, importing information from another program or third-party source, and / or one or more of these inputs may be predefined in a given implementation.
[0553] Boundary conditions can be mechanical problem definitions and input entities that are retained to serve as connection points with other components or separate 3D models in a larger 3D model. In some implementations, obtaining 600 includes, for example, receiving input 600 through a UI 122 on a display device 120, the input 600 indicating an initial 3D model of at least a portion of a design space defining topology optimization for an object. In some implementations, for example, a user is enabled to select multiple parts from the components shown in the UI through a UI 122 on a display device 120, and the user is enabled to indicate load points and areas away from the design. Then, the software (e.g., a CAD program 116) can use the inverse of these as a design space. Other inputs may also be received 600, such as a selected lattice topology and / or a hollow structure offset value input (e.g., a constant offset selected by a user of the CAD program 116 to define a hollow structure during topology optimization).
[0554] Design criteria may include one or more design objectives and one or more design constraints. Design objectives may include those that minimize one or more cost functions of the object to be manufactured and / or maximize one or more utility functions thereof. In some implementations, the design objectives include one or more of the following: (1) minimizing material use (e.g., waste); (2) minimizing weight; (3) minimizing manufacturing time; (4) maximizing tool life; (5) maximizing the ability to remove support material; (6) targeting natural frequencies (e.g., achieving higher / lower natural vibration modes for vibration reduction); and (7) positioning and packaging of parts within the structure (e.g., electronic components for circuit actuation (including reserving space for wiring between components in the assembly, limiting wiring length) and / or components for structural health monitoring, and component considerations to ensure that each component can be added / removed independently or in a limited number of steps). Design constraints may include those that address manufacturing requirements and / or address usage requirements. In some implementations, the design constraints include one or more of the following: (1) manufacturing constraints that restrict geometries to those that are manufacturable (e.g., minimum self-support angles for AM processes in which separate support material is not used, subject to build orientation when appropriate; or mechanical access constraints for AM or SM processes, such as 3-axis, 5-axis, and / or 2.5-axis CNC constraints); (2) thermomechanical constraints; (3) material fraction or variation constraints; (4) structural integrity; (5) stiffness; (6) compliance; (7) stress criteria (e.g., von Mises or Tsai-Wu stresses for manufacturing or use); (8) physical constraints (not interfering with larger part systems); and (9) failure constraints (e.g., brittle elastic fractional resistance). In some implementations, the design criteria include minimizing compliance with volume, stress, displacement, buckling safety factor, and / or natural frequency constraints, but the method can be generalized to other design objectives and constraints, including minimizing volume, deviation from a target (compliant mechanism) (e.g., least squares deviation), and / or design-related loads (pressure loads).
[0555] In some implementations, for example, Figure 1AThe CAD program 116 obtains 600 a design space for an object to be manufactured, a setup for a physical simulation of the object, at least one design goal for the object, and at least one design constraint for the object. The physical simulation performed by the systems and techniques described in this document can simulate one or more physical properties, and one or more types of simulations can be used. For example, FEA (including linear static FEA), finite difference methods, and material point methods can be used. In addition, simulations of physical properties can include simulating buckling, natural frequencies, heat flux, electrical flux or electromagnetic flux, and material solidification properties, among other possibilities. It should be noted that the setup for the physical simulation will be different for different physical simulations.
[0556] For example, the CAD program 116 uses a generative design process to iteratively modify 605 both the three-dimensional topology of the generative model of the object and one or more external shapes of the three-dimensional topology, and the generative design process represents the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions in the design space. For example, this may include using a gradient-based algorithm to optimize the 3D topology and shape of the object model, or using other algorithms such as evolutionary algorithms and random algorithms. In addition, the generative model can be a level set representation of one or more external shapes of the three-dimensional topology, and a level set topology optimization method can be used. In some implementations, the iterative modification 605 involves using the above combined Figure 3A (Mixed-Hollow) and FIG. 6B to FIG. 6D Describes the systems and techniques.
[0557] In a hybrid-hollow approach for topology optimization, a homogenized lattice material representation is used, where the homogenized lattice material expresses the stiffness of a lattice consisting of beams and void regions as a function of the lattice topology and the volume fraction of the lattice, for example using RVEs corresponding to different lattice topologies, as described above. It should be noted that the volume fraction is a function of the unit cell size and beam thickness, and that alternative lattice arrangements can be used, as mentioned above. Various approaches to generating RVEs can be used. In general, a homogenized lattice material representation can express the structural behavior of a given lattice as an anisotropic solid material, which is a continuous material with properties approximately equivalent to a given lattice. Various homogenization methods can be used, such as numerical homogenization and analytical homogenization, and / or experimental data can be used to define the homogenized lattice material representation.
[0558] In some implementations, iteratively modifying 605 includes adjusting 605A a volume fraction of a homogenized lattice material representation according to a constitutive matrix of the homogenized lattice material representation of the lattice topology. In some implementations, the beam thickness is varied while keeping the unit size constant. In some implementations, the unit size is varied, but this may result in a lattice with mixed unit sizes, which may present challenges during manufacturing. In some implementations, a uniform lattice is used during the initial topology optimization.
[0559] In any case, the current iteration of the model being iteratively modified is changed 605B for physical simulation. This may involve changing 605B the constitutive model used for physical simulation according to the current iteration of the three-dimensional topology and one or more external shapes of the three-dimensional topology offset inward to define the hollow structure in the three-dimensional topology, the homogenized lattice material representation at the uniform thickness of the beams in the lattice. For example, a Boolean difference operation may be performed 605B to define the hollow structure, and the homogenized lattice material representation is used within the 3D region of the Boolean difference.
[0560] In some implementations, the offset inward to define the hollow structure is a constant offset selected by the user of the computer-aided design program (as mentioned above), thereby setting the initial thickness of the hollow structure. In some implementations, the initial thickness is retrieved from a database of known good values for a given manufacturing process. Typically, the manufacturing process determines the minimum wall thickness. In any case, the homogenized lattice material is used to represent what would otherwise be solid material between one or more external shapes of the three-dimensional topology and their inward offsets. Thus, structures with internal hollows and lattices surrounding those internal hollows, such as Fig.6D shown.
[0561] Performing 605C a physical simulation of the object using the altered constitutive model to generate a physical evaluation relative to at least one design objective and at least one design constraint. As mentioned above, various types of physical simulations can be used, and therefore various types of physical evaluations can be generated. In some implementations, performing 605C the physical simulation involves performing a finite element analysis simulation using a body-fitting mesh-based solver, and altering 605B the constitutive model by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used for the finite element analysis simulation. In some implementations, the solid mesh-based simulation can be non-body-fitting, for example, solved using a voxel solver.
[0562] The shape change rate of one or more external shapes of the three-dimensional topology in the current iteration is calculated 605D according to the physical evaluation. For example, this may involve calculating 605D hollow shape derivatives and advection rates according to lines 12 and 13 in Algorithm 8. It should be noted that this is where different objectives and constraints come into play. In some implementations, calculating 605D the shape change rate involves using the augmented Lagrangian method described herein to calculate the advection rate, such as when there are multiple shape derivatives.
[0563] Then, before the next iteration, the generative model of the object is updated 605E using the shape change rate. For example, updating 605E may use the advection rate obtained by the augmented Lagrangian method (ALM). In some implementations, a level set method is used, such as shown above in lines 14 and 15 in Algorithm 8. In some implementations, non-level set surface manipulations may be used, such as T-splines or B-Rep control point moves or node moves in a mesh.
[0564] After modifying 605 both the three-dimensional topology and the one or more external shapes of the three-dimensional topology are changed, the thickness of the hollow structure is adjusted 610, for example, by the CAD program 116, based on physical simulation, at least one design goal, and at least one design constraint while keeping the one or more external shapes of the three-dimensional topology constant. For example, see Algorithm 4 and Algorithm 7, and Figure 6B .
[0565] In some implementations, adjusting 610 the thickness of the hollow structure includes: starting from one or more internal shapes of the hollow structure, iteratively modifying the one or more internal shapes of the three-dimensional topology using a generative design process, including changing a constitutive model used for physical simulation according to (i) the current iteration of the three-dimensional topology and the one or more internal shapes and (ii) the homogenized lattice material representation, so that after completing the modification of the one or more internal shapes, the thickness of the hollow structure is variable across the three-dimensional topology. In some implementations, a level set method is used to optimize the external shape and / or the internal shape of the hollow structure.
[0566] Thus, the generative model may be a level set representation of one or more internal shapes and one or more external shapes of a three-dimensional topology, and the generative design process may employ a level set topology optimization method to iteratively modify the one or more internal shapes and one or more external shapes of the three-dimensional topology. In some implementations, other macrostructure representations may be used for generative design, wherein the generative design process represents the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space. For example, other options for representing the contours of hollow regions include other surface modeling techniques. For example, the outer surface may be a mesh, a T-spline, or a B-Rep surface with an auxiliary inner surface created with a fixed offset, and the outer surface may be shape optimized using a classical shape optimization method.
[0567] After modification 605 completes the change to both the three-dimensional topology and one or more external shapes of the three-dimensional topology, and after adjustment 610, while keeping both the one or more external shapes of the three-dimensional topology and the thickness of the adjusted hollow structure constant, the thickness of the beams in the lattice or the density of the lattice is adjusted 615, for example, by the CAD program 116 according to the physical simulation, at least one design goal, and at least one design constraint. For example, see Algorithm 2 and step 10 of Algorithm 5 for thickness optimization, and also see step 4 of Algorithm 5, where the thickness is converted to an equivalent density. It should be noted that the lattice thickness optimization can be performed using density.
[0568] Additionally, in some implementations, (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during fabrication, and / or (iv) visibility of holes for aesthetic considerations may be identified 620 in different regions of the three-dimensional topology of the hollow structure. Then, for a subset of the different regions compared to the remaining regions, the location of one or more holes in the three-dimensional topology of the hollow structure for forming or removing support material used during fabrication may be determined 625 based on the identified (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during fabrication, and / or (iv) visibility of holes for aesthetic considerations.
[0569] After adjusting the thickness and lattice of the hollow structure (and potentially after determining one or more positions of one or more holes), a three-dimensional model of the object can be provided 630 based on the three-dimensional topology, one or more external shapes of the three-dimensional topology, the adjusted thickness of the hollow structure, and the adjusted lattice for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems. This may involve as described above in conjunction with Figure 1BProvide 195 as described. Thus, providing 630 may include generating instructions (e.g., tool path instructions) for an additive manufacturing machine, a subtractive manufacturing machine, or other machine, system, or device. It should be noted that the described process need not be used with any particular manufacturing system or technology.
[0570] Further details of topology optimization systems and techniques for implementing macrostructure types for hollow topology optimization and / or hybrid-hollow topology optimization are now described with respect to shape derivatives of thick shell representations.
[0571] geometry
[0572] Let ∑ be a smooth, watertight oriented surface in Euclidean space, with a normal vector field N∑ pointing in the direction of "free space" outside ∑. Solid objects can be created from ∑ by "thickening" it in the direction of the negative normal vector. That is: we can pick a small To limit
[0573] Ω h : = {x-sN ∑ (x): for all x∈[0,h]} (61)
[0574] To put it bluntly, h consists of all points sandwiched between Σ and an offset of Σ in the direction of the negative normal vector by a distance h, the offset being expressed as:
[0575] ∑ h : = {x-hN ∑ (x): for all x∈∑} (62)
[0576] Ω h The border It consists of two non-intersecting surfaces: ∑ itself and the offset surface ∑ h .
[0577] It should be noted that the so-called The "outward" unit normal vector field (which is given by Ω h The inner part of the solid material is defined as pointing to the outer part of the solid material) equal to N on ∑ ∑ , but points to ∑ h N ∑ In the opposite direction. h The exact formula for -N ∑ (proj ∑ (y)), where proj ∑ is a map that takes a point y to its closest point on Σ, so for example, if we know that y = x - hN ∑ (x) and h is small enough, then proj ∑ (y) = x.
[0578] Geometry Deformation
[0579] ∑ allows deformation. The deformation can be expressed as This deformation can be achieved in a number of ways: for example, by expressing ∑ as the zero level set of a function in the Euclidean space context and using ⊥ The method described is equivalent to first order in deformation. What is important is the infinitesimal variation of ∑ itself, which is precisely a function of the surface normal velocity. Any deformation of ∑ will have a notion of the "amplitude" of the deformation, which is a positive scalar ε. For example, if the deformation is produced by advection, then the amplitude of the deformation corresponds to the advection time. The deformed surface can be described by ∑ ε express.
[0580] Once ∑ is deformed, then ∑ h This is simply to maintain ∑ ε and∑ h The offset distance h between the deformed versions of the object “drags the object” in this way. Therefore, the deformation Ω of the thickened object h is completely determined by the deformation of ∑. It should be noted that it can be shown that ∑ h At point y∈∑ h The infinitesimal variation at is just -θ ⊥ (proj Σ (y)), where proj Σ is a mapping that takes a point y to its closest point on ∑.
[0581] Optimization by Steepest Descent
[0582] The steepest descent method can be used to find a certain objective function Φ: This is based on the shape Taylor formula, which says that ∑ is relative to some θ as above. ⊥ : The resulting deformation has a deformation of magnitude ε that satisfies:
[0583] Φ(∑ ε )≈Φ(∑)+εDΦ ∑ (θ ⊥ ) (63)
[0584] For sufficiently small ε, where DΦ ∑ (Φ ⊥ ) symbolizes the shape derivative of Φ in ∑.
[0585] Use ⊥ The formula for the shape derivative can be chosen for a specific θ ⊥, so that the shape derivative term in the shape Taylor formula is negative. Therefore, if ∑ is relative to this selected θ ⊥ If ∑ is changed, the objective function will decrease, at least for sufficiently small ε. After updating ∑ by performing the change (e.g., by advection of the level set function representing ∑), the updated shape represents an improvement relative to the objective function. For further improvement, the procedure can be iteratively repeated until convergence occurs.
[0586] Shape derivative of a class of objective functions
[0587] Consider the objective function of the following form. Let Φ0: is some "volumetric" shape function (e.g., a shape function that can be evaluated over a volumetric domain), such as a convex linear combination of the average structural compliance with respect to a set of load cases and the total mass. This allows the "surface" shape function to be defined by:
[0588] Φ(∑):=Φ0(Ω h ) (64)
[0589] It can be assumed that we know how to calculate Φ0 at any domain Ω and relative to the boundary normal velocity function V ⊥ : Any change in Ω generated with respect to the outward unit normal vector field of Ω ∈ That is, Φ0 can be standard, meaning that the shape derivative satisfies the Hadamard-Zolésio structure theorem, and provides the formula:
[0590] Φ0(Ω ε )≈Φ0(Ω)+εDΦ 0,Ω (θ ⊥ )
[0591] in
[0592] For sufficiently small ε, the function G of Ω is Ω It is known that the shape gradient of Φ0 at shape Ω. Ω The exact form of depends on Φ0, and the derivative can be calculated (for example, with respect to the average structural compliance and volume) by To figure out.
[0593] The shape derivative of Φ can be expressed in terms of the shape derivative of Φ0. This can be done by using the shape derivative that is almost completely applicable to Ω h The above formula of the shape derivative of Φ0 and taking into account the composition That is, ∑ can have a boundary normal velocity V ⊥ : =θ ⊥ , and ∑ hMay have a boundary normal velocity V ⊥ : = -θ ⊥ proj ∑ Then, Φ is given by the normal velocity function θ ⊥ The changes in ∑ ε The first-order change is:
[0594]
[0595] Therefore, the expected shape derivative formula is:
[0596]
[0597] It should be noted that y can be simply used as ∑ h to emphasize that the two integrals are over different spaces and cannot be combined a priori in any simple way.
[0598] Extract descent direction
[0599] Recall that the exact formula for the shape derivative is useful in that it should allow for the selection of θ ⊥ , so that the shape derivative becomes negative, thus reducing the objective function to first order according to the shape Taylor formula. However, it is not clear how to do this for the shape derivative of Φ calculated above. This is due to the presence of two competing terms (e.g., the integral over ∑ and the integral over ∑ h ), and it is unclear how these competing terms cancel out for a given θ ⊥ fact.
[0600] There are two ways to proceed. The first is to apply the Hilbert space method of extracting the descent direction. The second is to apply the change of variables to ∑ h This allows us to express it as an integral over ∑. This is simple: recall that for any point y∈∑ h Both can be in the form y=x-hN ∑ (x) is written; for our purposes now, this should be considered as the number from ∑ to ∑ h This mapping is called n:∑→∑ h , where n(x): = x-hN ∑ (x). Therefore, by changing the variable formula of the surface integral
[0601]
[0602] where Jac is the Jacobian of n and is based on the fact that for all x∈∑, proj ∑ (x-hN ∑ (x))=x.
[0603] It turns out that the Jacobian of n can be determined. With some help from differential geometry, it can be shown that Jac(x) = 1 + hH Σ (x)+h 2 K Σ (x), where H Σ is the mean curvature of ∑, and K Σ is the Gaussian curvature of ∑, provided that
[0604]
[0605] DΦ ∑ (θ ⊥ ) provides a good formula for the shape gradient of Φ at ∑, which is expressed as:
[0606] Therefore, as required by our steepest descent procedure to update ∑ toward the optimal direction of Φ, θ ⊥ =-G ∑ The selection of DΦ ∑ (θ ⊥ ) is negative.
[0607] Hole generation in hollow structures
[0608] In addition to the details described above for hollow topology optimization and / or hybrid-hollow topology optimization, shape derivatives can be used to assist in post-processing of the 3D model of the generative design of the object, such as to address manufacturing considerations. For many manufacturing processes, it remains challenging to build parts with hollow internal structures. Support materials are often used during manufacturing to enable the raw materials to be formed into the desired shape. For example, powder bed fusion (PBF) metal additive manufacturing (AM) can use metal particles or support structures, investment casting (IC) can use ceramic shells, and sand casting (SC) can use internal sand cores. Support materials often need to be provided during manufacturing and then removed afterwards because the support materials will occupy the interior of the part during manufacturing.
[0609] To enable practical formation and removal of support material, holes can be automatically generated to penetrate the outer skin of the hollow geometry, where the holes are placed in lower stress areas of the design. Fig. 7A An example of a generatively designed hollow part 700 is shown with a heat map showing load paths and exemplary hole locations in lower stress areas 705. The examples of hole placement locations 705 indicate how holes may be located in lower stress areas of the design, as indicated by the holes being located in blue areas of the heat map, which correspond to lower stresses from the load paths.
[0610] Considerations for hole placement may include, but are not limited to: (1) the amount and location of von Mises (VM) stresses; (2) the ease of forming or removing support material; (3) the ability to support or inhibit material formation during construction; and (4) the visibility of the hole for aesthetic considerations. For a given hollow geometry, topological shape derivatives can be used to find the optimal location to place the hole. Shape optimization can then be used to optimize the shape of this hole. The exemplary goal above is to minimize compliance. However, additional goals (such as a measure of the ability to remove support material) can be used as goals, or multiple goals can be used.
[0611] For example, the ability to extract material can be measured as follows. First, assume that each internal cavity is filled with powder. Take any point on the surface (a potential hole location). Now calculate the shortest distance from a given particle to a hole location. For example, Figure 7B A visualization is shown of the distance from a given hole 720 to various points 725 inside the hollow geometry 730. By calculating this distance for all particles and summing it, the sum is an exemplary measure of the ease of removing the support material. This distance can be calculated for all points on the surface along with the surface gradient, and this goal and gradient can be combined with the goal and gradient of minimizing compliance.
[0612] Thus, holes may be automatically generated in the optimized hollow shell for manufacturability reasons (e.g., additive manufacturing powder removal, supporting sand casting cores, etc.). These holes may be automatically or semi-automatically placed as described above, and have parameters such as hole size, number of holes, etc. In some implementations, the hole placement process described above is performed by the CAD program 116, and the user controls the hole parameters and final approval of the hole locations, for example, through the UI 112.
[0613] Augmented Lagrangian algorithm for constrained shape optimization
[0614] introduce
[0615] Algorithms can be developed to solve constrained shape optimization problems, where a shape function (e.g., volume or compliance) is to be minimized under a mixture of equality and inequality constraints. These constraints can use shape functions (e.g., target volume, aggregate stress measure, or aggregate displacement measure). Different algorithms can be used to handle non-aggregate (also known as point-by-point) constraints, such as the norm of the stress at each point of the shape. In general, the optimization problem can have the following form:
[0616]
[0617] sG E (Ω,u Ω )=0 (74)
[0618] G I(Ω,u Ω )=0 (75)
[0619] For u in Ω that satisfies the linear elastic PDE Ω , where Adm is a set of admissible shapes, F is a scalar-valued shape differentiable shape function, and G E and G I is a vector-valued shape differentiable shape function that may depend on the shape explicitly or implicitly through its elastic response under load.
[0620] Classical Augmented Lagrangian Algorithm
[0621] Equality constraints
[0622] In the classical setting of optimizing a scalar function of a vector variable, the augmented Lagrangian method can be most simply applied to equality constrained optimization problems of the following form:
[0623]
[0624] stg i (x) = 0 for i = 1, ..., k (76)
[0625] The augmented Lagrangian method is an enhancement of the so-called penalty method. The penalty method considers penalizing the objective function and the sequence c: = ck tends to infinity. Then, we can define For any i, it is never g i (x k )=0. However, since the increasing sequence of c values imposes more and more penalties on constraints, if x k Converges to x * Then for all i, g i (x * )=0, it can be expected that x k Finally, the constraints are satisfied. In addition, it can be expected that x * is the solution to Equation 76.
[0626] The problem with the penalty method is that due to poor conditions, the increasing sequence of c values will make it increasingly difficult to numerically use L c (x) is minimized. The augmented Lagrangian method corrects this problem. This algorithm minimizes the “augmented Lagrangian”
[0627]
[0628] And also maintain the sequence of Lagrange multipliers μ k and the sequence c of increasing penalty parameters k Again, it is possible to define These sequences are c k The update is done in a way that is stable about some large but finite value (thus avoiding the ill-conditioning inherent in general penalty methods). This is done under the KKT condition x k Converges to the solution x of Equation 76 * (and is therefore optimal and feasible) and μ k Converges to x * The associated Lagrange multiplier μ * Since the augmented Lagrangian method iterates over both the primal variable x and the “dual” variable μ, this algorithm is an instance of a primal-dual optimization method.
[0629] The general framework of the augmented Lagrangian algorithm may require methods for increasing penalty parameters and methods for tightening tolerances.
[0630] 1. Select the final tolerance T 最终 .
[0631] 2. Set k = 0. Start with initial tolerance c0, μ0. Constrain less than T 最终 Starting with an initial tolerance T0.
[0632] 3. When it does not converge to the tolerance T 最终 Internal time:
[0633] a. To the problem Apply an unconstrained optimization algorithm. When convergence reaches tolerance T k When the value is within , stop the algorithm. k+1 The output is the x-value at this convergence level.
[0634] b. Check constraint satisfaction
[0635] (i) If the constraint is not satisfied within the tolerance T k , then let c k+1 :=Increase(c k ) and μ k+1 :=μ k .
[0636] (ii) If the constraint satisfies within the tolerance T k Within, through Update the Lagrange multipliers.
[0637] c. Set tolerance T k+1 :=Tighten(T k ).
[0638] d. Increase k.
[0639] It should be noted that the above algorithm is characterized by updating the Lagrangian multiplier. This can be understood as follows. The unconstrained minimum of the enhanced Lagrangian at iteration k satisfies or
[0640]
[0641] Of course, the equality constraint optimization solution x of Eq. 76 * and Lagrange multipliers satisfy
[0642]
[0643] g i (x * )=0, for i=1,…,k (80)
[0644] In the augmented Lagrangian algorithm, since the expectation is to ensure that the iteration x k Converges to x * , so it is expected and g i (x k )→0 is consistent with this purpose. Therefore, the Lagrange multiplier update can be interpreted as A fixed-point iteration scheme of the form is used to achieve this goal.
[0645] Inequality constraints
[0646] The good extension of the classical equality-constrained augmented Lagrangian algorithm to the classical inequality-constrained case depends on two facts. For fact 1, the problem
[0647]
[0648] stg i (x)≤0, for i=1,...,k (81)
[0649] Equivalent to the problem
[0650]
[0651] For i=1,...,k (82)
[0652] The new z variable in Eq. 82 is called the slack variable and satisfies at the feasible optimum where x * is the feasible optimal value of Equation 81.
[0653] For fact 2, the enhanced Lagrangian of equation 82 is
[0654]
[0655] And the unconstrained optimization generated in the augmented Lagrangian algorithm can be decomposed and partially solved as follows:
[0656]
[0657]
[0658] It has been introduced This is because elementary calculus techniques can be used to The minimization is performed explicitly in Function to word the result.
[0659] As a result, the inequality constraint equation 81 can be solved by applying the augmented Lagrangian algorithm (see equality constraints above) to the modified Lagrangian function.
[0660]
[0661] Its gradient (after some manipulation) is
[0662]
[0663] Note that for completeness, the solution to the minimization in the z-variable appears in Equation 84. Substituting this solution yields Equation 85. For simplicity, the minimization in the z-variable is rephrased as
[0664]
[0665] Where a, b, s are fixed parameters, which can be solved as follows.
[0666] The first step is to substitute y: =z 2 Use with Substitute the minimization problem above. This new problem is now quite simple, because the function Φ(y): = a(s+y)+b(s+y) 2 is a simple parabola, and the problem Find the minimum of this parabola when constrained to the region y ≥ 0. Therefore, the global constrained minimum lies at the global unconstrained minimum of Φ, or at the y = 0 boundary of the constrained region, whichever is smaller. This leads to y * =-a / 2b-s or y * =0; select Φ from * :=Φ(y * ) is the equation for which the result is minimal. Therefore, after algebraic rearrangement,
[0667] Φ * =min{Φ(0),Φ(-a / 2b-s)}
[0668] =min{as+bs 2 , -a 2 / 4b}
[0669] =a max{s,-a / 2b}+b[max{s,-a / 2b}] 2 (89)
[0670] When s = g i (x) and a=μ i And when b=c / 2, equation 85 is obtained.
[0671] Applications to constrained shape optimization problems
[0672] For simplicity, the augmented Lagrangian algorithm can be applied to solve equations 73-75 with one scalar equality constraint and one scalar inequality constraint. Then, Ω∈Adm can be interpreted to mean that Ω satisfies an admissibility constraint of the following form. The surface of each Ω∈Adm contains predefined ports; each Ω∈Adm contains predefined keep-in regions; and each Ω∈Adm avoids predefined keep-out regions.
[0673] The classical augmented Lagrangian algorithm can be imitated for both equality and inequality constraints (see equality constraints above), but it can be adapted for shapes that satisfy admissibility constraints. The augmented Lagrangian is defined as
[0674]
[0675] in For simplicity, F and G are not shown. E , G I μ for linear elastic PDE Ω To consider only shape functions (which are surface or volume integrals of shape-dependent quantities), it is known from general principles that the shape gradient of a function L of this type evaluated at a given shape Ω is A scalar-valued function of , denoted dL(Ω): The shape gradient of the enhanced Lagrangian under shape Ω is
[0676] dL c (Ω,μ E , μ I )=dF(Ω)+(μ E +cG E (Ω))dG E (Ω)+max{0,μ I +cG I (Ω)}dG I (Ω) (91)
[0677] Where dF(Ω), dG E (Ω), dG I (Ω) are the shape gradients of the objective and constraint functions under shape Ω, respectively.
[0678] It is also known from general principles that by c The velocity function constructed by the projection of the shape gradient of Ω advects the level set function representing Ω to update the shape Ω to reduce L while maintaining the admissibility constraint. c The value of L c The shape gradients of dL are appropriately zeroed and the Adalsteinsson-Sethian rate expansion algorithm is used to convert dL c The value of (Ω) is Expand to This is used for the narrow band of c, μ E , μ I Solve the problem min for any fixed value of Ω∈Adm L c (Ω,μ E , μ I ) is the basis for an unconstrained gradient-based shape optimization algorithm.
[0679] The augmented Lagrangian algorithm for solving equations 73-75 can now be given as follows. It requires a method for increasing the penalty parameter and a method for tightening the tolerance.
[0680] 1. Select the final tolerance T 最终
[0681] 2. Set k = 0. With initial c0, Start with a constraint less than T 最终 Starting with an initial tolerance T0.
[0682] 3. Initialize the shape Ω0.
[0683] 4. When it does not converge to the tolerance T 最终 Internal time:
[0684] a. Apply the unconstrained gradient-based shape optimization algorithm to the problem Until convergence occurs within tolerance T k Output the shape Ω at this convergence level k+1
[0685] b. Check that constraints are satisfied.
[0686] (i) If the constraint is not satisfied within the tolerance T k Within, through Know Let c k+1 :=Increase(ck) multiplier.
[0687] (ii) If the constraint satisfies within the tolerance T k Within, through Know Update Lagrange multipliers
[0688] c. Set tolerance T k+1 :=Tighten(T k ).
[0689] d. Increase k.
[0690] Figure 8 8 is a schematic diagram of a data processing system including a data processing device 800, which can be programmed as a client or a server. The data processing device 800 is connected to one or more computers 890 via a network 880. Figure 8 Only one computer is shown as the data processing device 800, but multiple computers can be used. The data processing device 800 includes various software modules, which can be distributed between the application layer and the operating system. These may include executable and / or interpretable software programs or libraries, including tools and services of one or more 3D modeling programs 804 that implement the systems and technologies described above. Therefore, the 3D modeling program 804 can be a CAD program 804, and one or more generative design processes can be implemented (for example, using a level set-based method for generative design) to perform topology optimization and physical simulation operations (finite element analysis (FEA) or other operations). In addition, the program 804 can potentially implement manufacturing control operations (for example, generating and / or applying tool path instructions to implement the manufacture of the design object). The number of software modules used may vary depending on the implementation. In addition, the software modules may be distributed on one or more data processing devices connected by one or more computer networks or other suitable communication networks.
[0691] The data processing device 800 also includes a hardware or firmware device, which includes one or more processors 812, one or more additional devices 814, a computer-readable medium 816, a communication interface 818, and one or more user interface devices 820. Each processor 812 is capable of processing instructions for execution within the data processing device 800. In some implementations, the processor 812 is a single-threaded or multi-threaded processor. Each processor 812 is capable of processing instructions stored on a computer-readable medium 816 or a storage device (such as an additional device 814). The data processing device 800 uses a communication interface 818 (e.g., via a network 880) to communicate with one or more computers 890. Examples of user interface devices 820 include displays, cameras, speakers, microphones, tactile feedback devices, keyboards, mice, and VR and / or AR equipment. The data processing device 800 may store instructions for implementing operations associated with the above-described programs, for example, on a computer-readable medium 816 or one or more additional devices 814 (e.g., one or more of a hard disk device, an optical disk device, a tape device, and a solid-state storage device).
[0692] The embodiments and functional operations of the subject matter described in this specification may be implemented in a digital electronic circuit system, or in computer software, firmware or hardware (including the structures disclosed in this specification and their structural equivalents), or in a combination of one or more of them. The embodiments of the subject matter described in this specification may be implemented using one or more modules of computer program instructions encoded on a non-transitory computer-readable medium for execution by a data processing device or controlling the operation of a data processing device. The computer-readable medium may be a manufactured product (such as a hard drive in a computer system or a CD sold through a retail channel), or an embedded system. The computer-readable medium may be obtained separately, and then encoded with one or more modules of computer program instructions, for example, after one or more modules of computer program instructions are delivered through a wired or wireless network. The computer-readable medium may be a machine-readable storage device, a machine-readable storage substrate, a memory device, or a combination of one or more of them.
[0693] 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 creates an execution environment for the computer program in question, 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 network services, distributed computing, and grid computing infrastructures.
[0694] A computer program (also referred to as a program, software, software application, script, or code) can be written in any suitable form of a programming language (including compiled or interpreted languages, declarative or procedural languages), and it can be deployed in any suitable form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in 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 stores other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, 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 a single computer or on multiple computers located in one location or distributed across multiple locations and interconnected by a communication network.
[0695] 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, and the apparatus may be implemented as, a special purpose logic circuit system (e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit)).
[0696] For example, processors suitable for executing computer programs include both general-purpose microprocessors and special-purpose microprocessors, as well as 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 performing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices (e.g., disks, magneto-optical disks, or optical disks) for storing data, or be operably coupled to receive data from one or more mass storage devices, or transfer data to one or more mass storage devices, or perform both. However, a computer does not necessarily have such a device. Moreover, 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)). Suitable devices for storing computer program instructions and data include all forms of nonvolatile memory, media and storage devices, including, by way of 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 circuitry.
[0697] 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 (e.g., an LCD (liquid crystal display) display device, an OLED (organic light emitting diode) display device, or another monitor) for displaying information to the user, and a keyboard and a pointing device (e.g., a mouse or a tracking ball) by which the user can provide input to the computer. Other kinds of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback; and input from the user may be received in any suitable form, including sound, voice, or tactile input.
[0698] A computing system may include a client and a server. The client and the server are generally far 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 running on a corresponding computer and having 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 a back-end component (e.g., as a data server), or including a middleware component (e.g., an application server), or including a front-end component (e.g., a client computer with a graphical user interface or a browser user interface by which a user can interact with the implementation of the subject matter described in this specification), or any combination of one or more of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by digital data communication (e.g., a communication network) of any form or medium. Examples of communication networks include local area networks ("LAN") and wide area networks ("WAN"), internetworks (e.g., the Internet), and peer-to-peer networks (e.g., self-organizing peer-to-peer networks).
[0699] Although this specification contains many implementation details, these should not be interpreted as limitations on the scope of the claimed protection or the scope that can be claimed, but should be interpreted as descriptions of features specific to a particular embodiment of the disclosed subject matter. Certain features described in this specification in the context of a separate embodiment may also be implemented in a single embodiment in combination. Conversely, the individual features described in the context of a single embodiment may also be implemented in multiple embodiments individually or in any suitable sub-combination. Moreover, although the features are described above as acting in certain combinations and even initially required, one or more features from the required combination may be excluded from the combination in some cases, and the required combination may be directed to a sub-combination or a change in a sub-combination.
[0700] Similarly, although the operations are depicted in a particular order in the drawings, this should not be understood as requiring that such operations be performed in the particular order shown or in a sequential order or that all illustrated operations be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Moreover, the separation of the various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, but rather should be understood that the described program components and systems can be generally integrated together in a single software product or packaged into multiple software products.
[0701] 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 computer-aided design of a physical structure, comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, a setting for a physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; iteratively modifying, by the computer-aided design program, both a three-dimensional topology of a generative model of the object and one or more external shapes of the three-dimensional topology using a generative design process employing a level set topology optimization method and a level set representation of the generative model in the design space, wherein the modifying comprises: changing a constitutive model used for the physics simulation based on a current iteration of the level set representation and a homogenized lattice material representation expressing the stiffness of the lattice as a function of lattice topology and volume fraction of the lattice at a uniform thickness of the beams in the lattice composed of beams and void regions; performing said physical simulation of said object using the altered constitutive model to produce a physical evaluation relative to said at least one design objective and said at least one design constraint; calculating a shape change rate of the level set representation in the current iteration according to the physics evaluation; and updating the level set representation of the object using the shape change rate via the level set method; After the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, adjusting, by the computer-aided design program, (i) the thickness of the beams in the lattice or (ii) the density of the lattice while keeping the three-dimensional topology constant based on the physical simulation, the at least one design goal, and the at least one design constraint; and The computer-aided design program provides a three-dimensional model of the object based on the three-dimensional topology, the one or more external shapes of the three-dimensional topology and the adjusted lattice, so as to be used for manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
2. The method of claim 1, wherein the modifying comprises: The volume fraction of the homogenized lattice material representation is adjusted according to a constitutive matrix of the homogenized lattice material representation of the lattice topology.
3. The method of claim 1 , wherein performing the physical simulation comprises: A finite element analysis simulation is performed using a body-fitted mesh based solver, and the constitutive model is altered by mapping geometric field data and finite element simulation data between voxel grid points of the level set representation of the generative model and elements of a solid mesh model used in the finite element analysis simulation.
4. The method of claim 1 , wherein changing the constitutive model used for the physical simulation comprises: According to the current iteration of the level set representation, (i) void simulation elements are used outside the one or more external shapes of the three-dimensional topology, and (ii) the homogenized lattice material representation is used for simulation elements inside the one or more external shapes of the three-dimensional topology.
5. The method of claim 1, wherein the obtaining comprises: Input is received indicative of an initial three-dimensional model defining at least a portion of the design space.
6. The method of claim 5, wherein the initial three-dimensional model defines a boundary of the design space, and changing the constitutive model used for the physical simulation comprises: According to the current iteration of the level set representation, (i) the homogenized lattice material representation is used for simulation elements outside the one or more external shapes of the three-dimensional topology but inside the boundary of the design space, and (ii) solid simulation elements are used inside the one or more external shapes of the three-dimensional topology.
7. The method of claim 1, wherein calculating the shape change rate comprises calculating an advection rate using an augmented Lagrangian method.
8. The method of claim 1, wherein the homogenized lattice material represents expressing the structural behavior of a given lattice as an anisotropic solid material, the anisotropic solid material being a continuous material having properties approximately equivalent to the given lattice.
9. The method of claim 1, wherein the one or more computer-controlled manufacturing systems include an additive manufacturing machine, and the providing includes: generating a tool path instruction for the additive manufacturing machine from the three-dimensional model; as well as The physical structure corresponding to the object is manufactured by the additive manufacturing machine using the tool path instructions.
10. A system for computer-aided design of a physical structure, comprising: a non-transitory storage medium having instructions of a computer-aided design program stored thereon; as well as One or more data processing devices configured to execute the instructions of the computer-aided design program to perform the method of any one of claims 1-8.
11. The system of claim 10, comprising an additive manufacturing machine, wherein the one or more data processing devices are configured to execute the instructions of the computer-aided design program to generate a tool path description for the additive manufacturing machine from the three-dimensional model, and to use the tool path description to manufacture the physical structure corresponding to the object by the additive manufacturing machine.
12. A method for computer-aided design of a physical structure, comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, a setting for a physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; iteratively modifying, by the computer-aided design program, both a three-dimensional topology of a generative model of the object and one or more external shapes of the three-dimensional topology using a generative design process that represents the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space, wherein the modifying comprises: changing a constitutive model for the physical simulation in accordance with a current iteration of the three-dimensional topology and the one or more external shapes of the three-dimensional topology being offset inward to define a hollow structure in the three-dimensional topology; performing said physical simulation of said object using the altered constitutive model to produce a physical evaluation relative to said at least one design objective and said at least one design constraint; calculating a rate of shape change of the one or more external shapes of the three-dimensional topology in the current iteration based on the physical evaluation; and updating the generative model of the object via the generative design process using the shape change rate; After the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts the thickness of the hollow structure while keeping the one or more external shapes of the three-dimensional topology constant according to the physical simulation, the at least one design goal and the at least one design constraint; and The computer-aided design program provides a three-dimensional model of the object based on the three-dimensional topology, the one or more external shapes of the three-dimensional topology and the adjusted thickness of the hollow structure, so as to manufacture a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
13. The method of claim 12, wherein the offset inwardly to define the hollow structure is a constant offset selected by a user of the computer-aided design program.
14. The method of claim 12, wherein performing the physical simulation comprises: A finite element analysis simulation is performed using a body-fitted mesh based solver, and the constitutive model is altered by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used in the finite element analysis simulation.
15. The method of claim 12, wherein calculating the shape change rate comprises calculating an advection rate using an augmented Lagrangian method.
16. The method of claim 12, wherein adjusting the thickness of the hollow structure comprises: Starting from one or more internal shapes of the hollow structure, the one or more internal shapes of the three-dimensional topology are iteratively modified using the generative design process, including changing the constitutive model used for the physical simulation based on the three-dimensional topology and the current iteration of the one or more internal shapes, so that after completing the modification of the one or more internal shapes, the thickness of the hollow structure varies across the three-dimensional topology.
17. The method of claim 16, wherein the generative model comprises a level set representation of the one or more internal shapes and the one or more external shapes of the three-dimensional topology, and the generative design process employs a level set topology optimization method to iteratively modify the one or more internal shapes and the one or more external shapes of the three-dimensional topology.
18. The method of claim 16, comprising: identifying in different regions of the three-dimensional topology of the hollow structure (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during construction, and / or (iv) visibility of pores for aesthetic considerations; as well as For a subset of the different regions compared to the remaining regions, the location of one or more holes in the three-dimensional topology of the hollow structure for forming or removing support material used during manufacturing is determined based on the identified (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during construction, and / or (iv) visibility of the holes for aesthetic considerations.
19. The method of claim 12, wherein the obtaining comprises: Input is received indicative of an initial three-dimensional model defining at least a portion of the design space.
20. The method of claim 12, wherein the one or more computer-controlled manufacturing systems include an additive manufacturing machine, and the providing includes: generating a tool path instruction for the additive manufacturing machine from the three-dimensional model; as well as The physical structure corresponding to the object is manufactured by the additive manufacturing machine using the tool path instructions.
21. A system for computer-aided design of a physical structure, comprising: a non-transitory storage medium having instructions of a computer-aided design program stored thereon; as well as One or more data processing devices configured to execute the instructions of the computer-aided design program to perform the method of any one of claims 12-19.
22. The system of claim 21 , comprising an additive manufacturing machine, wherein the one or more data processing devices are configured to execute the instructions of the computer-aided design program to generate a tool path description for the additive manufacturing machine from the three-dimensional model, and to manufacture the physical structure corresponding to the object by the additive manufacturing machine using the tool path description.
23. A method for computer-aided design of a physical structure, comprising: obtaining, by a computer-aided design program, a design space of an object to be manufactured, a setting for a physical simulation of the object, at least one design goal of the object, and at least one design constraint of the object; iteratively modifying, by the computer-aided design program, both a three-dimensional topology of a generative model of the object and one or more external shapes of the three-dimensional topology using a generative design process that represents the three-dimensional topology of the generative model as one or more boundaries between one or more solid regions and one or more void regions within the design space, wherein the modifying comprises: changing a constitutive model used for the physical simulation based on (i) a current iteration of the three-dimensional topology and the one or more external shapes of the three-dimensional topology offset inwardly to define a hollow structure in the three-dimensional topology, and (ii) a homogenized lattice material representation expressing the stiffness of the lattice as a function of the lattice topology and volume fraction of the lattice at a uniform thickness of the beams in the lattice composed of solid beams and void regions; performing said physical simulation of said object using the altered constitutive model to produce a physical evaluation relative to said at least one design objective and said at least one design constraint; calculating a rate of shape change of the one or more external shapes of the three-dimensional topology in the current iteration based on the physical evaluation; and updating the generative model of the object via the generative design process using the shape change rate; After the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, the computer-aided design program adjusts the thickness of the hollow structure while keeping the one or more external shapes of the three-dimensional topology constant according to the physical simulation, the at least one design goal, and the at least one design constraint; After the modification completes the change to both the three-dimensional topology and the one or more external shapes of the three-dimensional topology, adjusting by the computer-aided design program (i) the thickness of the beams in the lattice or (ii) the density of the lattice while keeping both the one or more external shapes of the three-dimensional topology and the adjusted thickness of the hollow structure constant according to the physical simulation, the at least one design goal and the at least one design constraint; and The computer-aided design program provides a three-dimensional model of the object based on the three-dimensional topology, the one or more external shapes of the three-dimensional topology, the adjusted thickness of the hollow structure and the adjusted lattice, so as to be used to manufacture a physical structure corresponding to the object using one or more computer-controlled manufacturing systems.
24. The method of claim 23, wherein the modifying comprises: The volume fraction of the homogenized lattice material representation is adjusted according to a constitutive matrix of the homogenized lattice material representation of the lattice topology.
25. The method of claim 23, wherein performing the physical simulation comprises: A finite element analysis simulation is performed using a body-fitted mesh based solver, and the constitutive model is altered by mapping geometric field data and finite element simulation data between voxel grid points of a level set representation of the generative model and elements of a solid mesh model used in the finite element analysis simulation.
26. The method of claim 23, wherein calculating the shape change rate comprises calculating an advection rate using an augmented Lagrangian method.
27. The method of claim 23, wherein adjusting the thickness of the hollow structure comprises: Starting from one or more internal shapes of the hollow structure, the one or more internal shapes of the three-dimensional topology are iteratively modified using the generative design process, including changing the constitutive model used for the physical simulation based on (i) the three-dimensional topology and the current iteration of the one or more internal shapes and (ii) the homogenized lattice material representation, so that after completing the modification of the one or more internal shapes, the thickness of the hollow structure varies across the three-dimensional topology.
28. The method of claim 27, wherein the generative model comprises a level set representation of the one or more internal shapes and the one or more external shapes of the three-dimensional topology, and the generative design process employs a level set topology optimization method to iteratively modify the one or more internal shapes and the one or more external shapes of the three-dimensional topology.
29. The method of claim 27, comprising: identifying in different regions of the three-dimensional topology of the hollow structure (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during construction, and / or (iv) visibility of pores for aesthetic considerations; as well as For a subset of the different regions compared to the remaining regions, the location of one or more holes in the three-dimensional topology of the hollow structure for forming or removing support material used during manufacturing is determined based on the identified (i) stress levels, (ii) ease of forming or removing support material, (iii) ability to support or inhibit material formation during construction, and / or (iv) visibility of the holes for aesthetic considerations.
30. The method of claim 23, wherein the offset inward to define the hollow structure is a constant offset selected by a user of the computer-aided design program.
31. The method of claim 23, wherein said obtaining comprises: Input is received indicative of an initial three-dimensional model defining at least a portion of the design space.
32. The method of claim 23, wherein the homogenized lattice material represents expressing the structural behavior of a given lattice as an anisotropic solid material, the anisotropic solid material being a continuous material having properties approximately equivalent to the given lattice.
33. The method of claim 23, wherein the one or more computer-controlled manufacturing systems include an additive manufacturing machine, and the providing includes: generating a tool path instruction for the additive manufacturing machine from the three-dimensional model; as well as The physical structure corresponding to the object is manufactured by the additive manufacturing machine using the tool path instructions.
34. A system for computer-aided design of a physical structure, comprising: a non-transitory storage medium having instructions of a computer-aided design program stored thereon; as well as One or more data processing devices configured to execute the instructions of the computer-aided design program to perform the method of any one of claims 23-32.
35. The system of claim 34, comprising an additive manufacturing machine, wherein the one or more data processing devices are configured to execute the instructions of the computer-aided design program to generate a tool path description for the additive manufacturing machine from the three-dimensional model, and to manufacture the physical structure corresponding to the object by the additive manufacturing machine using the tool path description.
