Procedural generation of metamaterials
The procedural graph representation with a unified skeletal design space and conjugate surface construction method addresses the challenges of metamaterial design by enabling the exploration and generation of diverse structures, including TPMS, resulting in efficient and accessible metamaterial design and fabrication.
Patent Information
- Application Number
- PCT/US2024/021281
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-06-26
- Filing Date
- 2024-03-25
- Publication Date
- 2024-12-05
AI Technical Summary
Existing representations for metamaterial design are cumbersome and not compatible with each other, limiting the exploration of a wide range of structures, including straight trusses, curved beams, shellular structures, and solid bulks, as they require unique derivations and are difficult to edit, especially for triply periodic minimal surfaces (TPMS) which are often approximated due to complex mathematical requirements.
A procedural graph representation that uses a unified skeletal design space with simple elements like lines and surfaces, annotated with spatially varying thickness functions, and incorporates the conjugate surface construction method within an optimization loop to facilitate automatic design and editing of metamaterials, making it accessible to a wider audience.
This solution streamlines metamaterial design by enabling the exploration of various classes, including TPMS, with a compact, intuitive, and editable representation, allowing for the generation of novel structures with promising material properties, and facilitates the fabrication of metamaterials with diverse architectures.
Smart Images

Figure US2024021281_05122024_PF_FP_ABST
Abstract
Description
Attorney Docket No.70009-114WO1 PROCEDURAL GENERATION OF METAMATERIALS CROSS-REFERENCES TO RELATED APPLICATIONS
[0001] This application claims the benefit of U.S. Provisional Application No.63 / 469,189 filed on May 26, 2023, and U.S. Provisional Application No.63 / 523,136 filed on June 26, 2023. The entire contents of U.S. Provisional Application No.63 / 469,189 and U.S. Provisional Application No.63 / 523,136 are incorporated herein by reference. STATEMENT AS TOFEDERALLYSPONSOREDRESEARCH ORDEVELOPMENT
[0002] This invention was made with government support under 1122374 awarded by the National Science Foundation Graduate Research Fellowship Program. The government has certain rights in the invention. BACKGROUND OF THE INVENTION
[0003] This invention relates to methods and systems using a compact, intuitive representative computational graph for procedurally generating a wide spectrum of cellular metamaterials, which are small-scale, tileable structures that can be architected to exhibit many useful material properties.
[0004] Metamaterials are structures of long-standing interest, as they induce material properties that differ from those of their constituent base material(s). Metamaterials often exhibit behaviors that are not found in nature, such as tunable compliant, chiral, auxetic, and nonreciprocal behaviors and impressive strength-to-weight ratios. The behavior of a given metamaterial is primarily governed by its cellular architecture, which is the regular or random spatial arrangement of solid regions and voids used to fill a designated volume. The set of possible cellular architectures of metamaterials is uncountably large, even when only considering a small portion of the space – namely, the regular cellular solids that fill a unit cube (cell) and tile periodically in three-dimensional space. SUMMARY OF THE INVENTION
[0005] Effective metamaterial design can be challenging, because no existing representation is well-suited for a full suite of structures including straight trusses, curvedAttorney Docket No.70009-114WO1 beams, shellular structures, triply periodic minimal surfaces (TPMS) shellular structures, and solid bulks. Generic volume representations like voxel grids can express structures of any class, but the description for any individual structure is cumbersome. Moreover, editing is difficult because even simple changes (e.g., thickening a beam) trigger many independent voxel updates. Class-specific representations are common in practice, as they efficiently describe individual structures and permit simple edits. However, the underlying representations are as varied as the structures themselves: trusses and beams are often represented with graphs; solid bulks often stem from constructive solid geometry (CSG) operations, and thin-shell cellular (shellular) structures use surface meshes or implicit functions. These representations are not immediately compatible with one another, so it is difficult to explore over more than one class.
[0006] In some cases, it is difficult to explore variations even within a given class, as each structure requires a unique derivation. For example, many shellular metamaterials are based on TPMS, which are precisely defined as the integral of an Enneper-Weierstraß function. For ease of use, TPMS are commonly approximated by e.g., level sets of an implicit trigonometric function. However, both function types are structure-specific and difficult (if not impossible) to derive. This limits the set of structures accessible to engineers – and thus, available for exploration.
[0007] To alleviate these challenges, aspects described herein use a procedural graph representation that streamlines the process of metamaterial design for a wide range of common classes, including generic TPMS. The representation is specific to cellular metamaterial design, which capitalizes on characteristics like symmetries while ensuring that the representation is equally suitable for all target classes. Moreover, the representation is compact, intuitive, and easily editable, such that it is amenable to manual or automated design space exploration. Finally, it is expressive enough to represent not only known structures but also novel structures containing elements from one or more class(es).
[0008] One important aspect of the representation is the unified skeletal design space, which uses simple elements like lines and surfaces to capture the wide range of shapes found in metamaterials. Each skeletal element is constructed from a small set of simple, high-level specifications comprised of vertices and edges, a bounding volume, and the type of skeletal element (e.g., line, surface) to be instantiated. Each element is also annotated with a spatially varying thickness function that determines how it should eventually be thickened into a physically realizable volume (e.g., beam, shell, or solid bulk).Attorney Docket No.70009-114WO1
[0009] Although this design space naturally accommodates most of the target classes, the diverse TPMS shellulars present a considerable obstacle. To address this, the approaches described herein are grounded in the mathematical principles of TPMS, yet compatible with the design space described above and intuitive enough to be used by novices. A critical piece of the approach is based on the conjugate surface construction method (CSCM), which is "one of the most powerful techniques to construct minimal surfaces with a proposed shape in mind.” However, existing CSCM algorithms are largely restricted to mathematicians, as they require extensive human intervention and domain expertise. The CSCM is embedded in an optimization loop to realize the first fully automatic version of this pipeline, making it accessible to a wider audience.
[0010] In a general aspect, a method for forming a specification for a cellular metamaterial includes configuring a graph-based computation processor according to a graph-based computation specification to perform data processing steps, including steps to yield a specification of a unit cell for replication to form the cellular metamaterial. The data processing steps include forming a skeletal specification of a sub-cell of the unit cell from input data, the skeletal specification characterizing at least one of a line and a surface within the sub-cell, forming a skeletal specification of the unit cell from the skeletal specification of the sub-cell using one or more geometric transformations, the one or more transformations including a least one of a mirror transformation, a rotation transformation, a rotation transformation, and a scaling transformation, and performing the volumetric transformation of the skeletal specification of the unit cell to form a volumetric specification of the unit cell of the metamaterial for replication to form the specification of a volume of the metamaterial including a number of unit cells.
[0011] Aspects may include one or more of the following features.
[0012] The method may include accepting the graph-based computation specification and executing said process using a graph-based computation processor to yield the specification of the cellular metamaterial. The method may include using a graph-based computation processor to yield the specifications of a number of different types of cellular metamaterial, said types including a triply periodic minimal surface (TPMS) metamaterial, and a truss- based metamaterial. Forming a skeletal specification of a sub-cell may include determining locations of a number of vertices on a boundary of the sub-cell from the input data, the number of vertices forming a sequential chain, determining geometric constraints associated with at least some of the vertices from the input data, and forming the at least one of the lineAttorney Docket No.70009-114WO1 and the surface within the sub-cell from the locations of the vertices and the geometric constraints.
[0013] The method may include forming the at least one of the line and the surface within the sub-cell includes forming the surface including forming a non-planar minimal surface constrained by the chain of vertices. Forming the minimal surface may include constructing a conjugate surface of the nonplanar surface. The unit cell and sub-cell may have parallelepiped shapes. The unit cell and sub-cell may have cuboid shapes. The cellular metamaterial may be triply periodic in space. The method may include providing a user interface for accepting the graph-based computation specification from a user and displaying a result of one or more of the data processing steps.
[0014] The method may include using a machine-implemented search using the graph- based computation processor to generate candidate metamaterials, to computationally predict properties of said metamaterials, and to select from the generated candidate metamaterials according to predicted properties of said metamaterials. The search may include varying the input data for forming the skeletal specification of a sub-cell, while maintaining a same graph specification of the number of data processing steps. The search may include varying the graph specification of the number of processing steps. The method may include causing fabrication of the cellular metamaterial. Causing fabrication of the cellular metamaterial may include three-dimensional printing of said metamaterial.
[0015] In another general aspect, a computing system has a processor, and storage for processor instructions for causing the processor to perform all the steps described above.
[0016] In another general aspect, a non-transitory machine-readable medium has instructions stored thereon, said instructions when executed by a process cause the processor to perform all the steps described above.
[0017] In another general aspect, a method for forming a specification for a metamaterial using a number of transformations in a computational or procedural graph includes receiving a specification of a number of vertices located in a part of a fundamental bounding volume (sometimes referred to as a sub-cell of the unit cell) for a unit cell of the metamaterial, processing the specification of the vertices in a surface or line transformation to generate a fundamental skeleton (sometimes referred to as a skeletal specification of a sub-cell of the unit cell) for a unit cell of the metamaterial, processing the fundamental skeleton in one or more geometric transformations to generate a complete skeleton for a unit cell of theAttorney Docket No.70009-114WO1 metamaterial, and processing the complete skeleton in a volumetric transformation to generate a unit cell representation of the metamaterial.
[0018] Aspects may have one or more of the following features.
[0019] The method may also include receiving a specification of the computational graph, the specification specifying the number of transformations, and specifying a configuration for some or all of the transformations of the number of transformations. The configurability of the computational graph and the transformations allows a user to specify a wide spectrum of types of metamaterials including trusses, triply periodic minimal surfaces (e.g., Schwarz-P), solid bulks, and combinations thereof.
[0020] The fundamental skeleton may include a nonplanar surface. Physical material properties of the unit cell representation or of a metamaterial formed using replications of the unit cell representation may be determined by simulation. The geometric transformations may include one or more of a mirror transformation, a rotation transformation, and a scaling transformation. Processing the specification of the number of vertices may include forming one or more edge chains. The edge chains may have different smoothness properties.
[0021] The procedural graph may include and / or be capable of representing other types of skeletal elements (e.g., constant mean curvature surfaces) that are similarly parameterized or derived from vertices and smoothness-annotated chains. Other types of thickening operations may also be used in the procedural graph. The procedural graph may also be capable of additional types of simulation mesh discretizations (e.g., tetrahedralization) and / or property evaluation nodes. The building blocks of the procedural graph may be used to define a set of rules, a formal grammar, or some other approach that can be used as a generative model, to automatically produce procedural graphs describing a variety of valid structures (e.g., for material exploration).
[0022] The framework described herein may be able to optimize over the procedural graph representation to automatically identify a metamaterial structure that achieves some target material property (“inverse design”). Advantageously the framework doesn’t need to limit the search to a particular architecture a priori, as is required by most other methods.
[0023] Some aspects may be able to strategically arrange meso-scale metamaterial structures to fill in a larger macro-scale object with spatially varying properties. The desired spatially varying property distribution may be defined explicitly by the user, or it may be inferred automatically, to achieve a high-level functional goal (e.g., achieving a target deformation under a given load) specified by the user. Neighboring metamaterial structuresAttorney Docket No.70009-114WO1 may be optimized within the arrangement for connectivity, to ensure the integrity of the full object (e.g., so there are no missing or extremely weak connections between adjacent metamaterial cells). Transitions between structures of distinct architectures may be smooth (e.g., from beam-based metamaterial structures to shellular metamaterial structures) within a single macro-scale object. Aspects may leverage the compactness of the metamaterial representation to co-optimize the shape of a macro-scale object along with the structure and arrangement of the meso-scale metamaterials that comprise it.
[0024] Among other advantages, aspects described herein provide a unified framework for designing many different types of cellular structures for metamaterials, where different types of cellular structures were previously only obtained using separate, architecture- specific representations or through overly generic shape / volume representations such as voxel grids.
[0025] Aspects also advantageously provide a simple, intuitive, and easily editable framework for designing metamaterials that is usable by non-experts. Aspects also advantageously allow for the exploration and simulation of metamaterials to discover novel metamaterials (e.g., those that are particularly suitable to a particular application or fit a material specification). Aspects are advantageously able to leverage conjugate surface construction to easily make structures that previously required complex mathematics that wasn’t usable to make multiple different metamaterial structures.
[0026] Advantageously, aspects provide a compact, intuitive procedural graph representation for a wide spectrum of cellular metamaterials, which are small-scale, tileable structures that can be architected to exhibit many useful material properties. The “architecture” of each structure may contain elements such as beams, thin shells, or solid bulks. However, it is challenging to explore these diverse architectures using existing representations. Generic approaches like voxel grids are versatile, but representing and editing any individual structure is cumbersome. Architecture-specific representations address these issues but are generally incompatible with one another. A unified procedural graph is provided that succinctly represents the construction process for any structure using a simple skeleton annotated with spatially varying thickness. To express the highly constrained triply periodic minimal surfaces (TPMS) in this manner, the first fully automated adaptation of the conjugate surface construction method is presented, which allows non-experts to generate arbitrary TPMS from intuitive input. By extending this design philosophy to all architectures, engineers’ capacity for exploration is greatly expanded. The representation’s expressiveness, accuracy, compactness, and ease of use by conducting a user study and recreating a wideAttorney Docket No.70009-114WO1 range of established structures verified. Finally, the representation can generate large collections of novel structures with promising material properties.
[0027] Other features and advantages of the invention are apparent from the following description and from the claims. BRIEFDESCRIPTION OF THEDRAWINGS
[0028] FIG.1 is a flowchart for procedurally generating a metamaterial using a procedural graph.
[0029] FIGs.2a and 2b show edge chain construction.
[0030] FIG.3 is a flowchart for a surface / line construction module.
[0031] FIG.4 is an example of constructing a TPMS surface using a conjugate surface solver.
[0032] FIG.5 is an exemplary algorithm for a conjugate surface solver.
[0033] FIG.6 is a flowchart for a geometric transformation module.
[0034] FIG.7 is a flowchart for a solidify / object / volumetric transformation module.
[0035] FIG.8 is a number of classes of metamaterial unit cell structures.
[0036] FIG.9 is a flowchart for fabricating metamaterial structures based on unit-cell structures.
[0037] FIG.10 is an example of a fabricated metamaterial.
[0038] FIG.11 is an example of the design process for a complex metamaterial unit cell structure.
[0039] FIG.12 is a list of nodes for computational graphs.
[0040] FIG.13 is a metamaterial simulation node.Attorney Docket No.70009-114WO1 DETAILED DESCRIPTION 1 OVERVIEW
[0041] Referring to FIG.1, a user 100 uses a unified metamaterial design framework 101 to generate a metamaterial structure 180. Very generally, the framework 101 processes a guiding topology 103 for the metamaterial using a series of procedures such as surface or line construction, geometric transformation, and thickening procedures to generate a unit cell structure 160 that can be replicated to form a metamaterial structure 180. The metamaterial structure 180 can be used in a simulator 182 to simulate properties 184 of materials made using the metamaterial structure and / or the metamaterial structure can be used to physically fabricate a metamaterial 190.
[0042] Both the guiding topology 103 and the procedural graph 600 are specified by the user 100, where the procedural graph 600 processes the guiding topology to generate the metamaterial structure 180. The configurability of the guiding topology and the procedural graph enables the user to construct a wide spectrum of cellular metamaterials, including metamaterials with straight truss, curved beam, shellular, triply periodic minimal surface (TPMS) shellular, and solid bulk cellular structures. When coupled with the simulation mentioned above, the user can explore the metamaterial space by procedurally generating large collections of novel structures to discover structures with promising material properties.
[0043] In the exemplary procedural graph 600 of FIG.1, the guiding topology 103 is first provided to an edge chain construction module 105. The edge chain construction module 105 processes the guiding topology 103 to form one or more edge chains 106. As is described in greater detail below, each edge chain includes a set of edges connecting two or more vertices, where each edge is associated with a smoothness flag indicating whether the edge is required to be a straight line. Each edge chain is confined to a “fundamental” bounding volume (sometimes referred to as a sub-cell of the unit cell), which is a part of a unit cell volume that is ultimately occupied by the unit cell structure 160. A detailed explanation of examples of edge chains and the construction of edge chains can be found in Section 2, below.
[0044] The edge chain 106 is then processed in a surface or line construction module 130 (configured with a solver type parameter 131) to generate a fundamental skeleton 110 (sometimes referred to as a skeletal specification of a sub-cell of the unit cell) as a surface 112 or line constrained by the edge chain. The user can configure the surface construction module to specify the type and feature(s) of the generated surface (e.g., mixed minimalAttorney Docket No.70009-114WO1 surface, conjugate surface, direct surface, surface-curvature requirements, etc.). The fundamental skeleton 110 may also include other elements, such as lines or beams that are straight or curved, and / or planar surfaces. A detailed explanation of generating a fundamental skeleton from an edge chain can be found in Section 3, below.
[0045] One particularly interesting type of surface that can be generated by the surface construction module 130 is a triply periodic minimal surface (TPMS). A detailed explanation of how the surface construction module generates TPMSs from edge chains can be found in Section 3.4.3, below.
[0046] The fundamental skeleton 110 is then processed using one or more geometric transformation modules 140 (configured with transformation parameters 133) to generate a complete skeleton 150 (sometimes referred to as a skeletal specification of the unit cell). In general, the geometric transformation modules replicate the fundamental skeleton 110 throughout the unit cell through geometric transformations such as mirroring, translation, rotation, and / or scaling, as specified by the user 100. Additional details related to the use of geometric transformations to generate the complete skeleton can be found in Section 4, below.
[0047] In some examples, the complete skeleton 150 is represented as a collection of surfaces and / or lines with zero thickness and is processed in a solidification module 164, which generates a unit cell structure 160 by thickening the skeleton 150 according to a spatially varying thickness profile 135 specified by the user 100. Additional details related to generating the unit-cell structure using the thickening procedure can be found in Section 5, below.
[0048] In some examples, the unit-cell structure 160 is provided to a replication module 170 that generates a fully specified metamaterial structure 180 through a replication operation 170 of the unit-cell structure 160 while specifying the number of repetitions in each spatial dimension for the replication operation 170 (e.g., as specified by replication parameters 137). Then the fully specified metamaterial structure 180 can be provided to a fabrication device, such as a 3D printing machine, to fabricate the actual metamaterial 190. Optionally, the user 100 can also provide fabrication parameters 139 such as material types, printing speed, etc., to the fabrication device.
[0049] Optionally, once the metamaterial structure 180 is fully specified, the user can predict material properties 184 of the metamaterial 190 or the metamaterial structure 180, such as material matrix, directional elasticity, and phononic bandgap, by performingAttorney Docket No.70009-114WO1 computational simulations 182. The user 100 can specify simulation parameters, such as Young’s modulus, mass density, and Poisson’s ratio, for the simulations 182. Additional details and examples of simulation of material properties can be found in Section 8, below.
[0050] In general, when coupled with the simulation mentioned above, the user can explore the metamaterial space by procedurally generating large collections of novel structures to discover new structures with promising material properties, as is described in Section 9, below. 2 GUIDING TOPOLOGY
[0051] Referring to FIG.2, in some examples, the guiding topology 103 includes a number of vertices 103 and edge chain parameters 107 including a vertex order parameter 207a and a smoothness parameter 207b. The edge chain construction module 105 processes the vertices 103 using the parameters 107 to generate the one or more edge chains 106. In some examples, the edge chain construction module 105 includes a path creation module 211 and a smoothness annotation module 213.
[0052] In general, each of the vertices 103 represents a location of a point in a fundamental bounding volume 104 in a unit cell 102 (i.e., the volume that eventually contains the unit cell structure 160 of the metamaterial). The fundamental bounding volume 104 in FIG.2 is a cuboid that occupies a small part (e.g., 1 / 8th) of the unit cell 102, but other sizes and shapes (e.g., prism or tetrahedron) of fundamental bounding volumes are possible. Within the edge chain construction module 105, the vertices 103 are first provided to the path creation module 211, which is parameterized using the vertex order parameter 207a. The path creation module 211 processes the vertices 103 to create a path 106 traversing the vertices 103 in an order specified by the vertex order parameter 207a.
[0053] The path 106 is provided to the smoothness annotation module 213, which is configured using the smoothness parameters 207b. The smoothness annotation module 213 annotates each segment of the path 106 with either a “straight” flag or a “smooth” flag. Downstream components of the design framework 101 use the flags to create different types of fundamental skeleton designs, as is described in the next section. In general, a given edge chain does not have a mixture of edges flagged as smooth and edges flagged as straight—if a boundary contains both smooth and straight portions, multiple edge chains are required.Attorney Docket No.70009-114WO1
[0054] In the example of FIG.2, there is a single edge chain 106 with six edges connecting six vertices 104. As will become evident below, the edges are annotated with a “smooth” flag. 3 FUNDAMENTAL SKELETON
[0055] Referring to FIG.3, the edge chains 106 output by the edge are provided to the surface / line construction module 130, which is parameterized using a solver type parameter 131. The surface / construction module 130 processes the edge chains 106 using a solver 241 specified by the solver type parameter 131 to generate the fundamental skeleton 110 having skeletal elements that include a surface and / or line constrained by the edge chain. In some examples, the fundamental skeleton 110 includes and / or is associated with (1) a smoothness- annotated path over 3d vertices, (2) a corresponding skeletal type / solver, (3) a spatially varying thickness profile over the element’s domain, and for sliding solvers, (4) a set of bounding planes.
[0056] The different types of solvers (e.g., mixed minimal surface, conjugate surface, and direct surface, described below) used to process the edge chains 106 enable the design of several classes of metamaterial unit cell structures, including straight trusses, curved beams, shellular structures, triply-periodic minimum surface (TPMS) shellular structures, and solid bulks. 3.1 Lines and Beams
[0057] Straight and curved beams are well-represented by line skeletons, which follow a path given by an ordered list of vertices. Curves can be created from concise vertex lists via, for example, natural cubic spline interpolation. Thus, one needs only introduce the “smoothness” flag described above to determine whether individual segments should be straight or smoothed. A spatially varying thickness profile may be associated with the fundamental skeleton 110 to control a spatially varying thickness over a length of the line. 3.2 Shellular Structures
[0058] Shellular structures can be represented by surface-type fundamental skeletons (described in greater detail below) annotated with a spatially varying thickness profile. Surface-type skeletons are also amenable to an ordered-vertex parametrization, as surfaces are frequently generated over a target boundary loop. For example, the Plateau problemAttorney Docket No.70009-114WO1 spans a fixed boundary with a minimal surface, which locally minimizes surface area and has zero mean curvature everywhere. Minimal surfaces can also be given by the free boundary problem, in which the input boundary is not fixed: each boundary portion is restricted to lie in its given plane, but its specific shape is inferred automatically as it “slides” along the plane to improve the metric. For the general construction of a triply periodic minimal surface (TPMS), both boundary types are required. To capture them, a pair of sliding solvers are devised that take smoothness-annotated boundaries as input: smooth boundary portions are permitted to slide, while straight portions remain fixed. The mixed minimal sliding solver described above is used when at least one fixed edge is present. For fully sliding boundaries – which may otherwise degenerate – a conjugate solver is introduced (described in detail below). Lastly, direct solvers are introduced to generate (not necessarily minimal) surfaces over fully fixed boundaries; here, straight boundary portions remain straight while smooth portions are interpolated. All the solvers assume disc topology (genus 0), as higher-genus surfaces can generally be decomposed. Critically, this also holds for TPMS: despite having genus ≥ 3, TPMS can be decomposed via their symmetry lines (see below).
[0059] For volumetric primitives, when combined with simple thickening methods, lines, and surfaces yield many basic primitives. For example, by offsetting the thickness along the skeleton’s normal direction(s), one can transform a line into a cylinder or a square-bounded planar surface into a cuboid. Similarly, by sweeping a sphere of the desired radius across the line / surface skeleton, one can create cylinders / cuboids with rounded ends / edges. This yields a sphere if the underlying line has length zero. 3.3 Line Solver
[0060] Line skeleton elements of fundamental skeleton 110 are represented by a sequence of 3D points, which can form any open, non-branching path or simple closed loop (if the endpoints are identical). Along straight edge chains, neighboring vertices are connected by straight lines. Each smooth edge chain is interpolated by the line solver to form a natural cubic spline that passes through the input points with C2 continuity everywhere. Natural cubic splines are used because they are simple, intuitive, and familiar, thanks to their widespread use in other modeling tools. To address concerns that are specific to cellular metamaterial design, the standard spline solver used for the line solver is adjusted to permit (1) C2 continuity along closed loops and (2) curves that tile smoothly across a periodic boundary. Any ill-defined curves (with < 4 vertices) are addressed by computing a quadratic spline (3 vertices) or a straight line (2 vertices).Attorney Docket No.70009-114WO1 3.4 Surface Solvers
[0061] In some examples, the surface design space for fundamental skeleton 110 includes the mixed minimal, direct, and conjugate surface solvers. 3.4.1 Direct Surface
[0062] In some examples, the direct surface solver yields more general (non-minimal) surfaces over fully fixed boundaries. Segments of a straight edge chain remain straight, while smooth edge chains are first interpolated into a curve according to the line skeleton description above. Then, a standard thin-shell model is applied, which is defined over a surface that is subject to fixed boundary constraints: , where ^^^^^^^^^penalizes in-plane stretching, ^^^^^penalizes bending, and ^ is the energy weight. By default, it is set that ^ = 0.1 to prioritize ^^^^^^^^.
[0063] To measure the deformation energy, a rest shape must be defined for the surface. One way to define the rest shape is to use a hole-filling algorithm to generate an initial surface that spans the boundary loop. However, complex boundaries often yield poorly behaved or low-quality triangle meshes. Instead, a default rest shape is used for all boundary loops: namely, a circular surface patch that is much smaller than the input boundary. This forces a large stretching deformation to meet the target boundary; coupled with a de- emphasized bending energy (e.g., ^ = 0.0), this encourages a smooth surface that approximates a minimal surface. By adjusting ^, it is possible to deviate from this behavior to produce a wide range of surfaces as elements of the fundamental skeleton 110. 3.4.2 Conjugate Surface Construction 3.4.2.1 Overview
[0064] Referring to FIGs. 4 and 5, one particularly interesting type of surface that can be generated by the surface construction module 130 is a TPMS. A conjugate surface solver is introduced to provide a stable approach for TPMS generation, whose fundamental skeleton is the solution to a free boundary problem – i.e., the fundamental skeleton boundary includes only sliding edges that are permitted to change to reduce surface area. Under standard approaches such as mesh-based gradient flows, these problems are prone to degeneration. However, by leveraging an associate family of a minimal surface, the conjugate surfaceAttorney Docket No.70009-114WO1 solver provides a stable 3-step approach to these problems, as shown in FIG. 4. In a first set of steps 421, 423, a related Plateau problem is constructed. In a second step 425, the related Plateau problem is solved. Finally, in a third step 427, a conjugate transformation is applied to obtain the solution to the original problem. Although this approach is powerful, existing algorithms require considerable domain expertise. Therefore, an optimization loop is further introduced to automate the conjugate surface solver and make it accessible to novices. The critical intuition for the algorithmic treatment of the conjugate surface problem is provided below before discussing the conjugate surface solver in detail.
[0065] An associate family is a set of minimal surfaces that can be continuously transformed into one another the scalar . Some well-known families transform acatenoid into a helicoid, or a P surface into a Schwarz D. For special pairs of surfaces , the family can be parametrized as follows:where. As suggested by the complex formulation, the special surfaces (at^ ^ ^0) and (at = ) are said to be conjugate to one another. More generally, any two members and are conjugate to one another, and every minimal surface is part of such a pair.pairs have several pertinent properties: Proposition 1 (Conjugate Surface Properties): let be a minimal surface, and be its conjugate. Then, the following are true: (a) For any arbitrary point in the domain, the surface normals and areisometric, so the angles at corresponding points along the boundary are identical on both surfaces. (c) If some portion of is bounded by a straight line, then the corresponding portion of is bounded by a planar symmetry line, and vice versa.
[0066] The symmetry lines noted in Proposition 1(c) are critical for minimal surfaces (particularly TPMS), as they allow the surface to be extended while preserving smoothness:Attorney Docket No.70009-114WO1 Proposition 2 (Minimal Surface Extension): let be a minimal surface partially bounded by a symmetry line . Then: (a) If is a planar symmetry line in plane , then must be fully contained in , ^ must meet orthogonally and ^ can be extended by reflecting across . (b) If is a straight line, then ^ can be extended by rotating 180° about .
[0067] As hinted above, symmetry lines can also be used to decompose a complex surface into a fundamental patch bounded by symmetries. Then, by Proposition 1(c), has a conjugate bounded by the opposite symmetries. Moreover, it is possible to construct one surface from the other via a “conjugation” process. Thus, the problems of solving for and are equivalent and, often, at least one of them is well-behaved.
[0068] Since the term "conjugate" is overloaded – as the relationship between and as well as the procedure that maps between them – one can call the primary surface and the dual. Now, if the primary boundary contains any straight lines, the dual would still require planar (sliding) symmetry lines. The conjugate surface solver is not especially helpful in this case, and one should apply a mixed minimal solver to be introduced later. However, if is a free boundary problem (i.e., has only planar symmetry lines), is a simple polygonal contour. The conjugate surface solver is beneficial here, as any such admits a Plateau solution that can be conjugated to a non-degenerate .
[0069] The latter steps (Plateau problem solving and conjugation) are well-established. However, the construction process for ^is briefly described here, as this is the major limitation of existing conjugate surface algorithms.
[0070] By Proposition 1(a), the surface normal at each vertex must match that of the corresponding vertex ^. Moreover, since is orthogonal to both bounding planes incident on (Proposition 2(a)), the normal at aligns with the intersection line on which sits. By Proposition 1(b), the angle between adjacent edges incident on (as measured within the plane spanned by the pair of lines) must equal the dihedral angle spanned by the planes incident on . Now, the only remaining unknown is the length of each edge in . Boundaries with 4 vertices are determined up to scaling due to the loop closure condition, but for vertices, there are generally ( − 4) edge lengths that can be set arbitrarily.
[0071] Previous works require users to manually compute these edge lengths through an iterative process that requires a deep understanding of the relationship between and . This is especially true when should align with a given fundamental bounding volume, asAttorney Docket No.70009-114WO1 the edges’ relative lengths determine this alignment – thus, they cannot be set arbitrarily. Improper edge lengths may also preclude a valid solution entirely: for the period problem, in which multiple curves reside on the same bounding plane, most length assignments fail. Existing methods rely on a manual intermediate value argument that cannot readily expose the permissible values. The conjugate surface solver approach of the current invention identifies a valid solution while obscuring the solver details and reducing the required user expertise. 3.4.2.2 Edge Length Optimization
[0072] To address the limitation of existing conjugate surface methods, a fully automated solver, shown as Algorithm 1 in FIG. 5, is devised for free-boundary problems given by a set of boundary planes and an ordered edge loop over these planes. An optimization loop is proposed that finds edge lengths ^for the dual contour , such that the final surface conforms to .
[0073] To prepare for the optimization, the normals and angles at each vertex of ^are first inferred, as described above. In addition, the convex polyhedron ^is computed which is enclosed by the half spaces given by . It is also inferred that : = { , …, }, where is the number of planes in and is the set of vertices from that belong to plane . These vertex sets can be used with respect to any primary or dual surface since the pointwise correspondences for any point in by construction are known. Finally, a rotation is computed that will align the candidate primary surfaces to for comparison. To do this, a sample primary mesh is first constructed. A plane is fit to each co-planar vertex set along the boundary of , which results in a set of boundary planes , as shown in Algorithm 1 (FIG. 5). The is found that best aligns the normals of and . only needs to be computed once because all future candidate primary surfaces (and thus, their bounding planes) will share the same orientation; only the scale will differ.
[0074] To judge each candidate contour with lengths = { }, a two-term objective function is devised , wheresolver on a contour with lengths , penalizes differences between the boundary planes of and the target planes , and penalizes the misalignment of boundary vertices in that are supposed to beAttorney Docket No.70009-114WO1
[0075] First, the bounding planes for are found by fitting a plane to each co-planar vertex set . Since has already been applied to , the normal is set to be the normal of plane in . The plane’s origin is given by , where is the position of the vertex. Then, can be defined as.
[0076] polyhedron is computed that is enclosed by the half-spaces of . The iterative closest point (ICP) algorithm is used to align this polyhedron with the reference, . Only translation is permitted, as the rotation has already been accounted for. After alignment, is given by the mean squared distances between the two polyhedra. Since only translation is considered, no unconverging ICP results are encountered during the optimization.
[0077] The method above is designed for arbitrary convex bounding polyhedra. However, when the polyhedron is an Axis-Aligned Bounding Box (AABB) – i.e., six axis-aligned planes forming a cuboid space – one can leverage a stronger objective. Namely, for any pair of points , on opposite, parallel planes of the AABB with normal , the expected distance between and along ^is known. Thus, both and can be encoded if all such pairs , are ensured to be distance direction,i.e., , where isis the unit axis- direction, and are the sets of boundary vertices on each opposing plane with normal , and is the target distance between the two planes in direction .
[0078] On each evaluation of the energy function (Algorithm 2, FIG. 5), there is a set of suggested lengths input over the contour that may or may not form a closed loop. The closedness is verified and / or the lengths are adjusted as necessary by solving the following mini-minimization problem during SolveContour: , ,Attorney Docket No.70009-114WO1 where is the edge length that will be used as the base guess for edge on this iteration and is the contour edge direction given by vA and vN. The linear constraint guarantees that forms a closed loop. Since this is a quadratic energy with linear constraints, is obtained by directly solving a small linear system.
[0079] To seed the optimization and create , an initial guess is found for the contour lengths. By construction, all input vertices lie on the edges of the convex polyhedron , and contour edges lie on the faces of . Each contour edge is categorized based on its endpoints, which can be on (1) two neighboring polyhedron edges , ; (2) two non- neighboring polyhedron edges , on the same face; or (3) the same polyhedron edge . Each case behaves in a relatively consistent manner – as evidenced by, e.g., Proposition 2, which implies that final contour edges must intersect polyhedron edges perpendicularly. The expected length of each green curve predicts the corresponding dual length because lengths are preserved due to isomorphism (Proposition 1(b)). Thus, based on the expected shape of each final contour, a simple heuristic is developed for the initial length of in each case. In case (1), is the length of an arc between and , assuming the arc is part of the circle whose center is the intersection point of and , and whose radius is the average distance from each endpoint of to the circle center. In case (2), is the distance between the endpoints of . In case (3), the distance between theis set to be the diameter of a circle whose center is the midpoint of ; then, is half of the circle’s perimeter.
[0080] Beginning from , the energy given in Algorithm 2 of FIG. 5 is optimized using the gradient-free method.it is common to have large differences in absolute edge lengths, the optimization converges faster if one optimizes the ratio to each edge’s base length, rather than the absolute length itself. By defining the optimization over this ratio space, generic lower and upper bounds for each edge length can also be easily defined, which may be, e.g., 0.1 and 10 respectively.
[0081] After constructing a suitable primary mesh , the previously computed rotation is applied to ensure that the returned surface aligns with . Although the overall agreement is quite high, numerical issues frequently cause the vertices of the conjugated surface boundary to be slightly offset from their intended planes. To resolve this issue, all boundary vertices are “snapped” to their intended boundary planes. This is done via the direct surface solver (described above). For this, is treated as the rest shape and the target position is computed for each boundary vertex by projecting its current position to the target plane. The small discrepancies only induce a slight deformation of , so the accuracy of the results is unaffected.Attorney Docket No.70009-114WO1 3.4.3 Mixed Minimal Surface
[0082] In some examples, the mixed minimal surface solver is a sliding solver for boundaries that are at least partially fixed. As discussed above, the shape and location of all straight boundary segments are preserved, such that the boundary of the computed surface coincides precisely with all fixed input. Smooth (or “sliding”) segments have much more freedom, as they are treated like free boundaries that are permitted to deform within the given boundary plane. To achieve a minimal surface subject to these constraints, a mean-curvature flow algorithm is performed that imposes Dirichlet boundary conditions along any fixed boundaries and sliding constraints along sliding boundaries. 4 COMPLETE SKELETON
[0083] Referring to FIG.6, the fundamental skeleton 110 is then processed using geometric transformation modules 140 (configured according to the transformation parameters 133) to generate the complete skeleton 150. In some examples, the geometric transformation modules replicate the fundamental skeleton 110 throughout unit cell 102 through geometric transformations such as mirroring, translation, rotation, and / or scaling, as specified by user 100. One exemplary sequence of geometric transformations is shown in FIG.6, where three “mirror” transformations 340, 342, and 344 are applied to the fundamental skeleton 110. The first mirror transformation 340 adds a rotated replica of the fundamental skeleton 110, forming ¼ of the complete skeleton. The second mirror transformation 342 adds a rotated replica of the ¼ completed complete skeleton, forming ½ of the complete skeleton. The third mirror transformation 344 adds a rotated replica of the ½ completed complete skeleton, forming the complete skeleton 150. 5 UNIT-CELL STRUCTURE
[0084] Referring to FIG.7, the complete skeleton 150 is represented as a collection of surfaces and / or lines with zero thickness and is processed in a solidify (or object) module 164 (configured according to the thickness profile 135) to generate the unit cell structure 160. In some examples, the solidify module 164 generates the unit cell structure 160 by thickening (sometimes referred to as a volumetric transformation) the skeleton 150 according to a spatially varying thickness profile specified by the user 100 (as mentioned above). In some examples, to control how each skeletal element is instantiated as a volumetric object, a spatially varying thickness function is introduced over the element’s domain. This function isAttorney Docket No.70009-114WO1 given by sparse input: the user specifies the thickness at a few sample points, and then these values are interpolated over the full domain. The values are computed and stored with respect to a simple parametrization of the domain, rather than actual vertex positions along the skeletal element. For lines, a 1D parametrization for [0, 1] is used. The full thickness profile is linearly interpolated from the provided sample points. For surfaces, which have disc-topology, thickness is defined with respect to a UV-domain computed via the as-rigid- as-possible parameterization. Then, the thickness is interpolated by treating each sample point as a handle and computing bounded biharmonic weights over the domain. 5.1 Unit Cell Types
[0085] In some examples, the approaches described herein are suitable to design static 3D cellular metamaterials with regular (rather than random) unit cells, are tilable in 3- dimensions, and are confined to a unit cube. Referring to FIG. 8, examples of such metamaterials include trusses 831, curved beams 833, solid bulks 839, shellulars 835, and TPMS shellulars 837. 5.1.1 Trusses and Curved Beams.
[0086] Truss-based structures have long been used for metamaterials, as they are easy to describe yet widely varied. The space of truss topologies alone is large enough to demand its own taxonomy, even before accounting for continuous parameters like vertex positions or thickness profiles over beams and their junctions. Previous approaches hand-designed several curved-beam structures exhibiting chiral, auxetic, rigid, and compliant behaviors. For diverse Poisson’s ratio and Young’s modulus values. Other approaches described 1,205 distinct truss-based topologies using a tetrahedral decomposition of a cube and a graph over 15 possible nodes. Others created 262 topologies by deforming and superimposing 7 fundamental lattice units. Yet others used topology optimization to compute candidate structures, over which they fit graph templates to create parametrized truss families with extremal properties. The approach described herein is related to these approaches, and trivially captures truss- or beam-based topologies that reside in a unit cube. The approach described herein expands a powerful graph-based representation by porting it to cellular architectures that are traditionally less amenable to exploration.Attorney Docket No.70009-114WO1 5.1.2 Solid Bulks
[0087] Solid bulks appear in many forms, including non-periodic spinodoid topologies, foams, and open-cell porous structures. Aspects described herein relate to regular topologies, such as topology-optimized, cubic-symmetric structures. Others have proposed another set of cubic-symmetric structures based on the level sets of 36 crystallographic space groups. Many CSG-like structures have also been hand-designed for specific properties like phononic bandgap. The approaches described herein cover the vast majority of these regular structures. 5.1.3 Shellulars
[0088] Shellular units are often based on 2-manifold surfaces and non-manifold surfaces like cubic lattices and honeycomb cells. Others have designed functionally graded porous structures by spatially varying these base surfaces. Approaches described herein capture a wide range of (non-)manifold surfaces. 5.1.4 TPMS Shellulars
[0089] TPMS are minimal surfaces that tile seamlessly along 3 mutually orthogonal directions. Non-self-intersecting TPMS are of particular interest, as they partition the surrounding volume into two or more distinct labyrinthine channels. Shellulars based on TPMS are smooth and uniquely suitable for additive manufacturing, making them ideal for tasks such as bone scaffolding, static mixing, thermal energy management, and strength- preserving lightweighting. TPMS are often created by Enneper-Weierstraß functions, specialized triangle meshes, or level sets of implicit functions. The latter have enabled several TPMS explorer tools and methods for hybrid TPMS creation. Others have constructed composite TPMS structures by extracting isosurfaces from weighted sums of the TPMS’ trigonometric approximations. Yet others have created hybrid TPMS structures by computing Boolean operations on existing meshes. Lastly, others propose hierarchical TPMS structures, in which TPMS cells at one length scale are used to construct larger TPMS. Although these approaches show promise, they are limited by their reliance on structure specific TPMS representations, which are difficult to construct and combine with other classes. The approach described herein addresses both limitations.
[0090] Others have constructed surface- and truss-based approximations of TPMS from physical principles, by using 3D graphic statics to explore the dual graphs of geometric form and static forces. They also design hybrid TPMS by specifying the desired labyrinth(s) – i.e. channels of negative space on either side of the surface – and then constructing a surface thatAttorney Docket No.70009-114WO1 partitions the volume accordingly. This approach may be powerful, but even domain experts struggle to imagine the relationship between a structure and its labyrinth. Graph representations described herein operate on information about the desired surface itself, rather than its voids. One important element of the TPMS approach described herein is the CSCM. This method leverages a surface’s associate family to construct complex TPMS indirectly. The first automated, easy-to-use pipeline for this method is introduced herein. 6 METAMATERIAL STRUCTURE
[0091] Referring to FIG.9, in some examples, the unit-cell structure 160 is provided to a replication module 170, which generates a fully specified metamaterial structure 180 through a replication operation 170 of the unit-cell structure 160. The replication module 170 replicates the unit-cell structure 160 according to replication parameters 137 including a number of repetitions in each spatial dimension. Then the fully specified metamaterial structure 180 can be provided to a fabrication device, such as a 3D printing machine, to fabricate the physical metamaterial 190 according to fabrication parameters 139 (e.g., specifications including material and / or fabrication parameters such as material types, printing speed, etc.). Referring to FIG.10, one example of a fabricated metamaterial 1090 illustrates the result of replicating a unit cell structure to form a metamaterial. 7 EXEMPLARY PROCEDURAL GRAPH
[0092] Very generally, the process for designing metamaterials described herein includes creating a procedural graph that unifies the skeletal design space with other pertinent operations. The procedural graph includes nodes that perform operations such as vertex creation, line / surface inference, mirroring, or skeleton thickening. Each node also has properties that control its behavior. For example, an edge chain has a smoothness flag, and each surface node has a solver type and a thickness profile. By chaining and sequentially evaluating these nodes, a variety of structures can be formed.
[0093] Referring to FIG.11, one example of an exemplary procedural graph 1160 for generating a relatively complex unit cell structure is shown. The procedural graph 1160 generates a final unit cell structure 1161 by a Boolean combination of three component unit cell structures: a first component unit cell structure 1162a, a second component unit cell structure 1162b, and a third component unit cell structure 1162b. Each of the component unit cell structures 1162 begins as a guiding topology 1163a, 1163b, 1163c, which isAttorney Docket No.70009-114WO1 processed in an edge chain node 1164a, 1164b, 1164c to form edge chains. The edge chains are then processed in surface / line construction nodes 1165a, 1165b, 1165c to form fundamental skeletons.
[0094] The fundamental skeletons for the second and third component unit cell structures 1162b, 1162c are processed in geometric transformation nodes 1167b, 1167c to form complete skeletons. Complete skeletons for all three component unit cell structures are then thickened in object nodes 1166a, 1166b, 1166c to generate volumetric objects. The volumetric objects are then combined in Boolean combination steps 1168 to form the final unit cell structure 1161.
[0095] Referring to FIG.12, a listing of nodes and their associated inputs and properties is shown. Some of those nodes are described in more detail below. 7.1 Vertex and Edge Chain Nodes
[0096] A vertex node places a vertex at the 3D location given by its “position” property. Vertex nodes do not accept any input node connections. The next simplest node is the edge chain, which accepts a set of vertex nodes as input, and then instantiates a path over these vertices in the traversal order that is stored as a property of the edge chain node. As discussed above, each edge chain also has a “smoothness” flag which is used by the subsequent line or surface node(s) to facilitate the creation of various types of surfaces / lines. 7.2 Line Nodes
[0097] Line nodes accept a set of edge chains that are continuously traversable – i.e., the end of one edge chain is the start of another, such that they form a simple, non-branching path (open or closed). As discussed above, the shape of a line is determined by the smoothness of each constituent edge chain. A line node can accept any combination of smooth and fixed edge chains, as long as they are continuously traversable. For more complicated paths, multiple line nodes must be defined. The spatially varying thickness function for each line node is defined as described above. 7.3 Surface Nodes
[0098] Surface nodes accept a set of edge chains that form a simple closed edge loop, over which a surface is instantiated. Users first select a surface type from among mixed minimal, conjugate, and direct. This selection determines the algorithms used to interpret theAttorney Docket No.70009-114WO1 boundary and solve for the final surface, as detailed above. It also dictates the requirements and properties of the surface node.
[0099] For direct surfaces, the user needs only provide the energy weight and a non- degenerate, simple, closed boundary loop over vertices located anywhere in the fundamental bounding volume. For conjugate and mixed-minimal surfaces, each sliding segment must lie on a fundamental bounding volume face. Mixed minimal surfaces permit vertices anywhere on a fundamental bounding volume face. Conjugate boundaries are more restricted, to satisfy the properties described above. Specifically, all vertices must lie on fundamental bounding volume edges and form property-respecting configurations.
[0100] There are also dual surfaces and associated family nodes for capturing and exploring intermediate surfaces like the Gyroid. 7.4 Mirror, Transform, and Group Nodes
[0101] To build full translational units from a structure’s fundamental skeleton, standard geometric transformations are provided such as the mirror node. Translation, rotation, and scaling transformations are supported via the transform node, as well as the ability to combine a set of skeletal elements into one unit via the group node. Each of these nodes takes one or more “skeletons” as input, which may be a line, a surface, or the output of one of the transformation nodes above. The user can select whether the transformation is applied to a copy of the input skeleton (“doCopy”=true) or to the input skeleton itself (“doCopy”=false). 7.5 Object, CSG Boolean, and Voxel Nodes
[0102] To generate volumetric objects as the unit-cell structure, the solidify / object node thickens each skeletal element based on its interpolated thickness profile. First, a regular grid over a unit cell is instantiated, which will store an indicator function that tracks whether each grid point is inside or outside of the object. The grid resolution is a property of the object node, along with the desired extrusion method. As suggested above, two extrusion approaches are supported: spherical and normal. For spherical extrusion, the skeleton is densely sampled, and – for each sample point – a sphere is splatted onto the indicator function grid, using a sphere diameter equal to the sample point’s prescribed thickness. For normal extrusion, the offsets are applied along the normal direction at each sample point. For surfaces, each vertex is offset by one-half of the desired thickness along each orientation (±) of its normal. For lines, a scaled circular cross-section is swept along the skeleton to create a tube. After extrusion, the indicator function is computed for each resulting geometry andAttorney Docket No.70009-114WO1 union them together. To preserve tileability, identical function values are required for each pair of corresponding grid points on opposite boundary faces. Finally, marching cubes are performed on the indicator function to extract the object mesh.
[0103] A constructive solid geometry (CSG) boolean node is also provided that supports union, intersection and difference operations on a set of volumetric objects. For efficiency, Boolean operations on the underlying indicator functions are computed for each object; then, tileability is enforced and marching cubes are performed as before.
[0104] Lastly, the indicator functions are used to implement a voxel node, which creates voxelized meshes suitable for property prediction. 8 MATERIAL PROPERTY SIMULATION
[0105] Referring to FIG.13, optionally, once the metamaterial structure 180 is fully specified, the user can predict material properties 184 of the metamaterial 190 with the metamaterial structure 180, such as material matrix, directional elasticity, and phononic bandgap, by performing computational simulations 182. User 100 can specify simulation parameters, such as Young’s modulus, mass density, and Poisson’s ratio, for the simulations 182.
[0106] In one example, two property prediction nodes are implemented: material matrix and phononic bandgap. The material matrix node computes the stiffness tensor of the volumetric structure given by the computational graph. A homogenization approach is implemented to compute the equivalent stiffness tensor matrix, assuming periodic boundary conditions and a linear elastic material over the voxel mesh. The phononic bandgap node predicts a structure’s ability to prevent the propagation of waves in certain frequency ranges for applications such as frequency filters, beam splitters, waveguides, and sound / vibration protection devices. The blocked ranges are known as bandgaps, and they are often the result of structural frequency-filtering mechanisms such as Bragg scattering and local resonances. The bandgap is predicted using an approach that generates a set of dispersion curves showing the structure’s eigenmodes over varying wave vectors. These curves are processed in search of horizontal bands through which no curves pass. Each such area is a bandgap, as it indicates that the given frequency range has no viable transmission path through the structure.Attorney Docket No.70009-114WO1 9 MATERIAL EXPLORATION
[0107] In some examples, the approaches described herein can be used to explore the universe of possible metamaterials in a computerized search. For example, a computer can generate many (e.g., thousands or millions) of candidate metamaterials (e.g., by varying the guiding topology and / or the computational graph). The prediction and simulation techniques described above can be used to computationally predict the properties of the candidate metamaterials. Candidates with properties matching those properties desired by the user can be selected for further consideration and testing. In this way, the once arduous task of identifying novel and useful metamaterial structures can be automated. 10 IMPLEMENTATIONS
[0108] The approaches described above can be implemented, for example, using a programmable computing system executing suitable software instructions or it can be implemented in suitable hardware such as a field-programmable gate array (FPGA) or in some hybrid form. For example, in a programmed approach, the software may include procedures in one or more computer programs that execute on one or more programmed or programmable computing system (which may be of various architectures such as distributed, client / server, or grid) each including at least one processor, at least one data storage system (including volatile and / or non-volatile memory and / or storage elements), at least one user interface (for receiving input using at least one input device or port, and for providing output using at least one output device or port). The software may include one or more modules of a larger program. The modules of the program can be implemented as data structures or other organized data conforming to a data model stored in a data repository.
[0109] The software may be stored in non-transitory form, such as being embodied in a volatile or non-volatile storage medium, or any other non-transitory medium, using a physical property of the medium (e.g., surface pits and lands, magnetic domains, or electrical charge) for a period of time (e.g., the time between refresh periods of a dynamic memory device such as a dynamic RAM). In preparation for loading the instructions, the software may be provided on a tangible, non-transitory medium, such as a CD-ROM or other computer-readable medium (e.g., readable by a general or special purpose computing system or device), or may be delivered (e.g., encoded in a propagated signal) over a communication medium of a network to a tangible, non-transitory medium of a computing system where it is executed. Some or all of the processing may be performed on a special purpose computer, orAttorney Docket No.70009-114WO1 using special-purpose hardware, such as coprocessors or field-programmable gate arrays (FPGAs) or dedicated, application-specific integrated circuits (ASICs). The processing may be implemented in a distributed manner in which different parts of the computation specified by the software are performed by different computing elements. Each such computer program is preferably stored on or downloaded to a computer-readable storage medium (e.g., solid state memory or media, or magnetic or optical media) of a storage device accessible by a general or special purpose programmable computer, for configuring and operating the computer when the storage device medium is read by the computer to perform the processing described herein. The system may also be considered to be implemented as a tangible, non- transitory medium, configured with a computer program, where the medium so configured causes a computer to operate in a specific and predefined manner to perform one or more of the processing steps described herein.
[0110] A number of embodiments of the invention have been described. Nevertheless, it is to be understood that the foregoing description is intended to illustrate and not to limit the scope of the invention, which is defined by the scope of the following claims. Accordingly, other embodiments are also within the scope of the following claims. For example, various modifications may be made without departing from the scope of the invention. Additionally, some of the steps described above may be order independent, and thus can be performed in an order different from that described.
Claims
Attorney Docket No.70009-114WO1 WHAT IS CLAIMED IS:
1. A method for forming a specification for a cellular metamaterial comprising: configuring a graph-based computation processor according to a graph-based computation specification to perform a plurality of data processing steps, including steps to yield a specification of a unit cell for replication to form the cellular metamaterial, wherein the data processing steps include: forming a skeletal specification of a sub-cell of the unit cell from input data, the skeletal specification characterizing at least one of a line and a surface within the sub-cell; forming a skeletal specification of the unit cell from the skeletal specification of the sub-cell using one or more geometric transformations, the one or more transformations including a least one of a mirror transformation, a rotation transformation, and a scaling transformation; and performing the volumetric transformation of the skeletal specification of the unit cell to form a volumetric specification of the unit cell of the metamaterial for replication to form the specification of a volume of the metamaterial including a plurality of unit cells.
2. The method of claim 1, further comprising accepting the graph-based computation specification, and executing said data processing steps using the graph-based computation processor to yield the specification of the cellular metamaterial.
3. The method of any of claims 1 and 2, further comprising using the graph-based computation processor to yield the specifications of a plurality of different types of cellular metamaterials, said types including a triply periodic minimal surface (TPMS) metamaterial, and a truss-based metamaterial.
4. The method of claim 1, wherein forming a skeletal specification of a sub-cell comprises: determining locations of a plurality of vertices on a boundary of the sub-cell from the input data, the plurality of vertices forming a sequential chain;Attorney Docket No.70009-114WO1 determining geometric constraints associated with at least some of the vertices from the input data; and forming the at least one of the line and the surface within the sub-cell from the locations of the vertices and the geometric constraints.
5. The method of any of the preceding claims, wherein forming the at least one of the line and the surface within the sub-cell comprises forming the surface, said forming comprising: forming a non-planar minimal surface constrained by the chain of vertices.
6. The method of claim 5, wherein forming the nonplanar minimal surface comprises constructing a conjugate surface of the nonplanar minimal surface.
7. The method of any of the preceding claims, wherein the unit cell and sub-cell have parallelepiped shapes.
8. The method of claim 7, wherein the unit cell and sub-cell have cuboid shapes.
9. The method of any of claims 7 and 8, wherein the cellular metamaterial is triply periodic in space.
10. The method of any of the preceding claims, further comprising providing a user interface for accepting the graph-based computation specification from a user, and displaying a result of one or more of the data processing steps.
11. The method of any of claims 1 through 9, further comprising using a machine- implemented search using the graph-based computation processor to generate candidate metamaterials, to computationally predict properties of said candidate metamaterials, and to select from said candidate metamaterials according to predicted properties of said candidate metamaterials.Attorney Docket No.70009-114WO1 12. The method of claim 11, wherein the search comprises varying the input data for forming the skeletal specification of a sub-cell, while maintaining a same graph specification of the plurality of data processing steps.
13. The method of claim 11 or 12, wherein the search comprises varying the graph specification of the plurality of processing steps.
14. The method of any of the preceding claims, further comprising causing fabrication of the cellular metamaterial.
15. The method of claim 14, wherein causing fabrication of the cellular metamaterial comprises three-dimensional printing of said metamaterial.
16. A computing system comprising a processor, and storage for processor instructions for causing the processor to perform all the steps of any one of claims 1 through 14.
17. A non-transitory machine-readable medium having instructions stored thereon, said instructions when executed by a processor cause the processor to perform all the steps of any one of claims 1 through 14.
Citation Information
Patent Citations
Polymer conjugates targeting mediators of gastrointestinal disorders with reduced exposure
US62634691P0
Retractable Propeller Driven Pedal System and Method of Use
US62635231P0
Methods and apparati for implementing programmable pipeline for three-dimensional printing including multi-material applications
US20140324204A1
Cited By
TPMS-based pressure-resistant underwater low-frequency sound insulation superstructure and design method thereof
CN120408902A
Performance-driven metamaterial connectivity enhancement method based on sliding block
CN122333827A
Shell-based material
US20230265903A1
