System and method for generative design and additive manufacturing of non-newtonian smart materials for energy absorption

Abstract

The present invention relates to a system and method for generative design and additive manufacturing of non-Newtonian smart materials, and the system and method are optimized for energy absorption. The proposed system includes a generative design module that creates candidate structures based on rheological inputs, a finite element analysis module that simulates performance using strain rate sensitive models, and an additive manufacturing module that manufactures optimized structures. The designs can be represented in a voxel grid and converted to signed distance fields (SDF) for simulation and comparison. The optimization process can iteratively refine structural features, such as cell size, density, and void geometry, to meet impact performance and manufacturing constraints. As such, the integrated approach can automatically generate high-performance printable structures for protective applications.

Description

Technical Field

[0001] This invention generally relates to the fields of structural mechanics and smart materials, laminated fabrication, impact engineering, rheology, computer science, and data analysis; more specifically, this invention relates to a system and method for use in the design and laminated fabrication of energy-absorbing non-Newtonian smart materials. Background Technology

[0002] Energy-absorbing materials are crucial for a wide range of applications, protecting equipment, machinery, automotive components, aerospace structures, and all moving objects. Traditional energy-absorbing structure design methods typically rely on manipulating geometric features, such as honeycomb or lattice structures, to control the dissipation of kinetic energy during impact. However, these techniques often reach performance limits due to factors such as material density, manufacturability, and the need to adapt to varying impact conditions. Materials exhibiting non-Newtonian behavior (viscosity varying with shear rate) have significant potential for enhanced energy absorption. Their ability to harden at high strain rates compared to traditional Newtonian materials provides superior impact protection. However, designing structures using non-Newtonian materials presents significant challenges. This is due to the complex relationships between material properties, geometry, and dynamic loading conditions, making it difficult to predict and optimize energy absorption performance using traditional design methods.

[0003] Current computational tools typically rely on simplified material models that fail to capture the complex rheological behavior of non-Newtonian materials. This limitation hinders accurate predictions of energy absorption in real-world impact scenarios. Furthermore, existing design optimization techniques often struggle to handle the nonlinear characteristics introduced by non-Newtonian material models, making it difficult to effectively explore the design space and, consequently, find optimal solutions.

[0004] Additive manufacturing technology offers the potential to create complex geometries, enabling advanced energy-absorbing structures. However, effectively combining additive manufacturing with the design of non-Newtonian materials requires a sophisticated approach that considers both the unique behavior of the materials and the manufacturing constraints imposed by the additive manufacturing process.

[0005] Therefore, there is a need for a system and method to improve the use of non-Newtonian materials in the design and manufacture of energy-absorbing structures in order to overcome the limitations of current design and manufacturing technologies. Summary of the Invention

[0006] The present invention aims to meet the above-mentioned needs by providing an integrated approach that combines generative design, advanced finite element analysis (FEA) with non-Newtonian material models and laminated fabrication.

[0007] This invention provides a system and method for the generative design and stacking fabrication of non-Newtonian smart materials that optimize energy absorption. The proposed system and method overcome the limitations of existing technologies by integrating advanced computational design, simulation, and fabrication techniques. Furthermore, this invention employs a signed distance field (SDF) representation, providing inherent characteristics of the design and facilitating shape and topology comparisons independent of specific geometric parameters.

[0008] According to a first aspect of the invention, a system is provided for designing and manufacturing non-Newtonian energy-absorbing structures. The system receives user-defined input parameters and generates physical structures through an integrated design-to-manufacturing workflow. The system includes a user interface, a generative design module, a database, and a finite element analysis module. The user interface receives input parameters including material properties, load conditions, and geometric constraints, and converts the input parameters into structured data. The generative design module receives the structured data and generates candidate structures using a 3D-conditional diffusion model, wherein the 3D-conditional diffusion model can synthesize voxel-based geometries based on the physical and rheological properties of the selected non-Newtonian material. The dataset contains experimentally derived rheological data, including viscosity curves, viscoelastic moduli, and manufacturing parameters. During the generation of candidate structures, the rheological data is provided to the generative design module. The finite element analysis module simulates the mechanical properties of the candidate structures using a material model specific to non-Newtonian behavior and a voxel mesh resolution consistent with the laminated manufacturing process.

[0009] According to a second aspect of the invention, a method is provided for designing and manufacturing non-Newtonian energy-absorbing structures, and is designed to receive user-defined input parameters and generate physical structures through an integrated design-to-manufacturing workflow. The method includes the following steps: receiving input parameters, including material properties, load conditions, and geometric constraints, via a user interface and converting the input parameters into structured data; receiving the structured data via a generative design module; generating candidate structures using a 3D-conditional diffusion model via the generative design module, wherein the 3D-conditional diffusion model synthesizes voxel-based geometries based on the physical and rheological properties of the selected non-Newtonian material; storing a dataset in a database, wherein the dataset contains experimentally derived rheological data, including viscosity curves, viscoelastic moduli, and manufacturing parameters; providing the rheological data to the generative design module during the generation of the candidate structures; and simulating the mechanical properties of the candidate structures using a material model specific to non-Newtonian behavior and a voxel mesh resolution consistent with the laminated manufacturing process via a finite element analysis module.

[0010] The above configuration can achieve the following key advantages:

[0011] Enhanced energy absorption performance: The integration of advanced design, simulation and manufacturing technologies enables the structure to have excellent energy absorption characteristics.

[0012] Automated design and manufacturing: The proposed system simplifies the entire process and reduces design time and costs.

[0013] Optimization for specific materials: Focus on non-Newtonian materials to ensure that their unique properties are fully utilized in the final design.

[0014] Functionality suitable for a variety of applications: The proposed system can be widely used in protective equipment, automotive and aerospace and other scenarios. Attached Figure Description

[0015] The embodiments of the present invention will now be described in more detail with reference to the accompanying drawings, wherein:

[0016] Figure 1 A schematic block diagram illustrating the architecture of a system according to an embodiment of the present invention is shown;

[0017] Figure 2 An internal flowchart of a diffusion model of 3D conditions implemented in a generative design module according to an embodiment of the present invention is shown.

[0018] Figure 3 A flowchart illustrating the mesh-to-3D image conversion process performed in a generative design module according to an embodiment of the present invention is provided.

[0019] Figure 4 A flowchart illustrating the calibration algorithm implemented in the optimization module according to an embodiment of the present invention is provided.

[0020] Figure 5 A flowchart illustrating the optimization algorithm executed in the optimization module according to an embodiment of the present invention is provided.

[0021] Figure 6 Example diagrams of optimized designs are provided, showing voxel-based structures representing generative outputs, with annotations for relative density and specific energy absorption.

[0022] Figure 7 The training loss curves and validation loss curves for the artificial intelligence model used in the generative design module are shown.

[0023] Figure 8 The graph shows the predictive performance of the AI ​​model used in the generative design module. Detailed Implementation

[0024] In the following description, a system and method are proposed, and aspects of their use in the design and fabrication of non-Newtonian smart materials for energy absorption are illustrated as preferred examples. Those skilled in the art will understand that various modifications, including additions and / or substitutions, can be made without departing from the scope and spirit of the invention. Specific details may be omitted to avoid obscuring the invention; however, this disclosure is intended to enable those skilled in the art to practice the teachings herein without extensive experimentation.

[0025] Figure 1 A schematic block diagram illustrating the architecture of a system 100 according to an embodiment of the present invention is shown. (Refer to...) Figure 1 System 100 is used for generative design and modular fabrication of non-Newtonian energy-absorbing protective structures. System 100 is applicable to a variety of scenarios, including but not limited to those involving structural mechanics, smart material systems, modular fabrication, impact engineering, rheology, computer science, and data analysis. System 100 can be deployed in scenarios involving the design and manufacture of products using non-Newtonian materials, such as protective equipment, energy-absorbing panels, or lightweight aerospace structures. System 100 can also be used in simulation-based workflows, such as those involving modeling material behavior, computationally designing and generating structures, and evaluating performance under dynamic loading conditions.

[0026] System 100 includes a user interface 102, a generative design module 110, a database 120, a finite element analysis (FEA) module 130, an optimization module 140, and a multilayer manufacturing module 150. In the proposed architecture, the generative design module 110, the dataset 120, the FEA module 130, and the optimization module 140 can construct a design framework based on artificial intelligence (AI).

[0027] User interface 102 provides an access platform for users to define design goals and material constraints in a simple and structured manner. User interface 102 allows users to input key parameters such as target energy absorption, geometric boundaries, load conditions, and specific rheological properties of the material to be used. User interface 102 also converts this human-readable input into a machine-readable structured data format so that downstream modules within the system can subsequently parse and process it. For example, this formatted data can serve as the base input for generative design module 110.

[0028] Generative design module 110 generates candidate structures based on formatted input parameters received from user interface 102. Input parameters may include energy absorption targets, cell sizes, geometric constraints, loading conditions, and rheological properties specific to the selected non-Newtonian material. As a core AI-driven component, generative design module 110 integrates 3D-conditional diffusion models and utilizes a machine learning-based generative engine.

[0029] 3D conditional diffusion models can iteratively construct voxel-based structural geometries using noise tensors and conditional vectors derived from input parameters. These geometries can respond to material behavior, such as viscosity and process-specific constraints related to shear rate.

[0030] The generative design module 110 also employs a structural representation method that allows the generated structures to be represented as tetrahedral meshes and further converted to signed distance fields (SDFs), thus providing a consistent and differentiable format for subsequent simulations, optimizations, and topology analyses. Specifically, the generated design can initially be represented as a tetrahedral mesh, and then converted to SDF. The conversion process involves extracting oriented closed surfaces for each structure and calculating the signed distance from each point on the Cartesian mesh to the nearest surface location.

[0031] In this regard, the descriptions of structural features mentioned, including unit cell size, density, pore size and shape, are defined with full consideration of the liquid phase properties of non-Newtonian materials and the relevant manufacturing constraints, such as injection speed and curing time.

[0032] In one embodiment, the generative design module 110 may include a model for processing user-defined input parameters, performing a 3D conditional diffusion process, and applying structural representation methods to generate consistent design outputs. In one embodiment, the model of the generative design module 110 may apply gradient-based algorithms or evolutionary algorithms.

[0033] The following representative examples are the contents of input parameters defined through user interface 102 and processed by generative design module 110. In one embodiment, the input parameters used in generative design module 110 include material-specific and geometry-specific constraints, such as strain rate sensitivity, target energy absorption, cell size, and overall structure density.

[0034] The strain rate sensitivity of the selected non-Newtonian material can range from 1 to 50, determined by the ratio of the complex modulus at high strain rates to low strain rates (e.g., 3 Hz and 0.1 Hz, respectively). The target energy absorption can be at least 1 J (joules), depending on the application requirements. The cell size can range from 1 to 30 mm to allow for finely tuned structural control, while also selecting geometric constraints, such as 100 to 2,000 kg / m². 3 The density range and minimum pore size of 0.01 mm are used to balance energy absorption efficiency and manufacturability.

[0035] Database 120 serves as a material property database, storing datasets containing rheological data derived from experiments with various non-Newtonian smart materials. The experimentally derived rheological data includes at least shear rate-dependent viscosity curves, viscoelastic moduli, strain rate sensitivity curves, and manufacturing parameters such as extrusion pressure, printing speed, and curing time. In addition to acting as a centralized repository, database 120 also provides direct data access to the generative design module 110.

[0036] During the generation of candidate structures, the generative design module 110 can retrieve relevant material properties from the database 120 to guide the formation of a geometry consistent with the physical behavior and manufacturing constraints of the selected material.

[0037] In addition, database 120 is also used to receive feedback from FEA module 130. If the simulation results show deviations in predicted performance or material response, these results can be used to update or improve the stored dataset. This feedback mechanism allows the dataset in database 120 to remain adaptive over time and contributes to the continuous improvement of the design and simulation process.

[0038] In one embodiment, the dataset stored in database 120 includes information related to the rheological properties of non-Newtonian materials, structural design records, and dynamic response and performance data associated with previously evaluated configurations.

[0039] In one embodiment, the dataset stored in database 120 may also include rheological data of non-Newtonian smart materials, including but not limited to polymer formulations of polyurethane, polydimethylsiloxane (PDMS), styrene-ethylene-butylene-styrene (SEBS), silicone, blends of natural and synthetic rubbers, and shear-thickening elastomers.

[0040] FEA Module 130 is used to simulate energy absorption, deformation modes, and structural elasticity under dynamic loading scenarios, such as drop impacts or compression. FEA Module 130 incorporates a nonlinear rheological model specific to non-Newtonian behavior and uses the same voxel mesh resolution as 3D printing. This consistency between simulation and manufacturing improves prediction accuracy and helps screen design candidates that fail to meet desired performance thresholds under actual manufacturing constraints.

[0041] The input to the FEA module 130 can be directly derived from the structural output generated by the generative design module 110. After simulation, if performance deviations, instabilities, or material description errors are observed, the FEA module 130 can pass feedback to the database 120 to update the stored dataset, thereby prompting the refinement or expansion of the stored rheological curves. This workflow allows the FEA module 130 to serve not only as a verification stage for the generated structure by the generative design module 110, but also as an adaptive quality gate, thereby driving the continuous accuracy and relevance of the material data in the dataset stored in the database 120.

[0042] In one embodiment, the FEA module 130 includes a rheological model tailored for non-Newtonian materials, and a simulation program for evaluating dynamic response and performance under specific load conditions.

[0043] The optimization module 140 optimizes the structural geometry based on simulation feedback received from the FEA module 130. The optimization module 140 uses optimization algorithms to iteratively adjust key design parameters, such as cell shape, spatial connectivity, and density gradient. The goal of the optimization module 140 is to enhance energy absorption performance while meeting base fabrication constraints, including feature resolution, injection speed, and curing time. The optimized configuration can then be further provided as feedback to the generative design module 110.

[0044] In one embodiment, the generative design module 110 also includes a comparator for evaluating the differences between the original design and the optimized version returned from the optimization module 140. If the comparator identifies deviations in geometry, material distribution, or other structural parameters that exceed a predetermined threshold, a notification is triggered and can be sent to the user via the user interface 102. This functionality allows the user to identify significant changes that may affect the original design intent and decide whether to further adjust the input parameters before proceeding.

[0045] The additive manufacturing module 150 is used to fabricate the final optimized structure using a liquid phase deposition process compatible with high-viscosity, shear-sensitive non-Newtonian materials. The additive manufacturing module 150 receives optimized design data presented in voxel-level structural information from the optimization module 140. The received data can be converted by the additive manufacturing module 150 into machine-executable manufacturing instructions, defining the precise paths and control sequences required for physical implementation. The manufacturing process can be carried out under operating conditions including parameters such as injection speed, nozzle diameter, printing speed, and curing temperature.

[0046] In one embodiment, the build-up manufacturing module 150 may employ liquid phase deposition technology, suitable for processing high-viscosity and highly shear-sensitive non-Newtonian materials; the build-up manufacturing module 150 includes a material inlet for receiving a printable formulation, a controllable extrusion nozzle for dispensing material at precise pressure and speed, and a movable build platform supporting layer-by-layer deposition. The nozzle's operating settings, including extrusion rate, temperature, and motion path, can be dynamically adjusted based on design inputs.

[0047] In simple terms, the generative design module 110 receives input parameters from the user interface 102, including material properties, loading conditions, and geometric constraints, and generates initial structural candidates using a 3D conditional diffusion model. During this process, the generative design module 110 can retrieve relevant rheological and processing data from the database 120 to guide the generation of designs consistent with the physical behavior of the selected non-Newtonian material. Furthermore, the structural features involved in the process, such as cell size, density, pore size, and shape, can be determined based on the material's rheological properties and manufacturing constraints, such as injection speed and curing time.

[0048] The FEA module 130 uses a non-Newtonian material model to simulate the mechanical properties of the generated design under dynamic loads. The simulation can be performed using the same voxel mesh resolution as in the stacking fabrication process. Based on the simulation results, the optimization module 140 iteratively refines the design to improve energy absorption and ensure compatibility with manufacturing requirements. The final optimized design is then transferred to the stacking fabrication module 150, which uses a liquid phase deposition process consistent with the material viscosity and shear sensitivity to fabricate the target structure.

[0049] In the above workflow, this invention also provides a generative design and stacked manufacturing method for non-Newtonian smart materials. The proposed method includes the following key steps: (1) defining input parameters for the generative design module; (2) using the generative design module to generate candidate structural designs; (3) converting the generated designs into a signed distance field (SDF) representation; (4) using the FEA module and a material model for non-Newtonian behavior to simulate the mechanical properties of the candidate designs; (5) using an optimization module to optimize the designs based on the simulation results; and (6) using a stacked manufacturing module to manufacture the optimized structure. The provided integrated sequence realizes a data-driven, automated design and manufacturing process for complex energy-absorbing structures. Compared with traditional methods, the proposed method provides enhanced energy absorption capacity, improves adaptability to material-specific constraints, shortens the design cycle time, and enables the production of high-fidelity, performance-optimized structures for specific applications.

[0050] The functions and internal components of the generative design module 110 and the optimization module 140 will be further described below.

[0051] Figure 2 An internal flowchart of a 3D conditional diffusion model implemented in generative design module 110 according to an embodiment of the present invention is shown. The 3D conditional diffusion model can generate a 3D design by performing at least five steps.

[0052] Step S210 involves receiving 3D input parameters including a noise tensor and mechanical property conditions. In one embodiment, the diffusion model with 3D conditions receives input parameters including a noise tensor and one or more mechanical property conditions. The mechanical property input parameters may include parameters related to the selected non-Newtonian material, such as strain rate sensitivity, target modulus range, or rheological threshold. This property information can be retrieved from database 120, which provides material-specific information to regulate the design generation process. In this way, the model can initiate structural synthesis within the limits of material feasibility and application requirements.

[0053] Step S220 involves forward diffusion, a process that introduces noise into the structural representation. During the forward diffusion phase, the 3D conditional diffusion model progressively adds noise to the initial 3D structural representation, gradually transforming the structured data into a completely random state. This process can be performed internally within the generative design module 110 without requiring input from external data. The purpose of this phase is to prepare input parameters for the subsequent reverse denoising process while preserving conditional information.

[0054] Step S230 involves pure noise. At the end of the forward diffusion process, the structural form is represented as pure noise. This latent representation can serve as the starting point for the reverse diffusion stage. The pure noise state does not contain explicit structural information but is still implicitly constrained by the material and performance parameters defined in the input stage.

[0055] Step S240 involves performing backdiffusion for denoising using a 3D U-Net model. In the backdiffusion phase, a trained 3D U-Net model is applied to predict the geometry of the underlying structure from a noisy representation in a stepwise manner. The prediction process can be driven by a deep learning model, which can be pre-trained to learn and reconstruct mechanically and materially feasible complex spatial patterns. The 3D U-Net architecture extends the standard U-Net to the volumetric data domain, which is particularly suitable for modeling three-dimensional structures with high spatial resolution. Its encoder-decoder design with skip connections allows the model to capture both global context and local geometric details. 3D U-Net can act as an AI-based generative engine, enabling the system to synthesize structural candidates that meet conditional input parameters, such as rheological parameters and design constraints. The predictive power of the 3D U-Net model allows it to generate complex geometries that conform to mechanical specifications, which are practically difficult to design using rule-based or purely parametric methods. Although the denoising stage does not actively query external data during inference, the 3DU-Net model can be pre-trained based on data from a dataset stored in Database 120, which includes previously validated structures and their associated performance records.

[0056] Step S250 involves outputting the generated 3D design. After denoising, the diffusion model with 3D conditions outputs a 3D design that reflects geometric constraints and material-aware considerations. The final structure is generated through the combined effect of the model's predictive capabilities and mechanical property inputs retrieved from database 120.

[0057] Figure 3 A flowchart illustrating the mesh-to-3D image conversion process performed in the generative design module 110 according to an embodiment of the present invention is shown. Figure 3 The process involves six steps, including extracting the oriented closed surfaces of each structure and calculating the SDF of points on the Cartesian mesh based on the mesh representation, thereby generating a 3D image based on the SDF. Furthermore, the shape and topology of the structure can be accurately represented using algorithms.

[0058] Step S310 begins with the tetrahedral mesh representation of the candidate design. The mesh is composed of... Figure 2 The diffusion model of the 3D conditions described in the paper is generated, and the geometry of the structure is encoded by volume elements.

[0059] Step S320 involves extracting oriented closed surfaces from the tetrahedral mesh. In this step, the outer surface boundaries of the mesh are identified and extracted while preserving the surface normal orientation, which is an important step in accurately calculating the directed distance.

[0060] Step S330 involves calculating the directed distance from the extracted geometry to the surface. In this step, an algorithm capable of efficiently and accurately evaluating the number of wraps and the distance field can be used to calculate the directed distance from each point in space to the nearest point on the surface.

[0061] Step S340 involves creating a Cartesian mesh containing the extracted surface boundary volume. The mesh can be discretized into uniformly distributed nodes, with each node serving as a sampling location for storing geometric distance data.

[0062] Step S350 involves assigning directed distance values ​​to the mesh nodes. For each node in the Cartesian mesh, a signed scalar value is assigned based on its relative position to the extraction surface. A positive value indicates a point outside the surface, while a negative value indicates a point inside the structure.

[0063] Step S360 involves generating a final 3D image based on the signed distance field. The resulting 3D image serves as a unified geometric representation, enabling downstream modules (e.g., FEA module 130 and optimization module 140) to evaluate, compare, and manipulate the current structural design independently of the original mesh topology.

[0064] In one embodiment, the generative design module 110 is used to generate at least two candidate structures for each iteration, thereby enabling parallel evaluation and comparative optimization across multiple design paths.

[0065] In one embodiment, the FEA module 130 can work in conjunction with the optimization algorithm to form a closed-loop feedback loop for iterative design refinement. This involves generating the initial structure and transforming it into a closed-loop feedback loop. Figure 3 Following the signed distance field, FEA module 130 can use a material model specific to the selected non-Newtonian formula to evaluate the structural response under specified loading conditions. The simulation program transmits the resulting performance data to optimization module 140, which then adjusts the design accordingly. Through this feedback process, system 100 can iteratively modify and re-evaluate the structure to improve energy absorption and manufacturability. The optimization logic can dynamically update structural parameters, such as cell size, density, pore size, and shape, based on rheological and mechanical properties retrieved from database 120.

[0066] Figure 4 A flowchart illustrating the calibration algorithm implemented in the optimization module 140 according to an embodiment of the present invention is shown. Figure 4 The proposed procedure involves measuring the rheological properties of non-Newtonian materials, such as shear rate-dependent viscosity and viscoelastic behavior. The goal is to ensure the accuracy of the simulation and to maintain consistency with actual manufacturing conditions.

[0067] Step S410 involves receiving experimental data and initial simulation parameters. The experimental data may include the behavioral response of a non-Newtonian material measured under controlled loading conditions, such as stress-strain curves, energy absorption values, or viscoelastic behavior. The simulation parameters may include initial estimates of the rheological constants, boundary conditions, and material coefficients used in FEA module 130.

[0068] Step S420 involves running a simulation using the received parameters. The optimization module 140 and the FEA module 130 can collaborate to perform the simulation and generate a predicted structural response for subsequent comparison with experimental data.

[0069] Step S430 involves comparing the simulation results with experimental data. The system evaluates the differences between the predicted behavior and the observed behavior, such as deviations in energy absorption, deformation profile, or failure mode.

[0070] Step S440 involves adjusting simulation parameters based on the identified deviations. Parameter tuning may include modifying the viscosity profile, updating the constitutive model, or optimizing the mesh density and solver tolerance to better reflect physical reality.

[0071] Step S450 involves outputting the calibrated simulation parameters. The output simulation parameters can be stored for future use in simulations.

[0072] Figure 5 A flowchart illustrating the optimization algorithm executed in optimization module 140 according to an embodiment of the present invention is shown. Figure 4 The calibration procedure described herein allows optimization module 140 to refine the structural design generated by generative design module 110 based on performance targets and simulation results obtained in FEA module 130. By applying specific algorithms, such as gradient-based optimization algorithms, optimization module 140 can iteratively adjust design parameters to improve energy absorption performance while minimizing design time and computational costs.

[0073] Step S510 involves receiving the input design and the corresponding performance targets. The input design can be previously determined by the FEA module 130. Figure 4 In this process, candidate structures are evaluated using calibrated simulation parameters. Performance targets may include energy absorption thresholds, deformation limits, or manufacturing constraints.

[0074] Step S520 involves evaluating the performance of the current design. The evaluation process can be based on simulation data obtained from the FEA module 130 or predicted internal performance metrics. The optimization module 140 can assess how close the current design is to the desired functional objectives, such as in terms of energy absorption efficiency or stress distribution.

[0075] Step S530 involves modifying the design based on the evaluation results. In one embodiment, this step includes applying gradient-based or heuristic optimization techniques to adjust design variables, such as cell geometry, material distribution, or topological features. The updated design aims to improve performance while maintaining manufacturability.

[0076] Step S540 involves checking convergence. Optimization module 140 can determine whether the modified design meets the predetermined performance target within acceptable tolerances. If the target has not been met, the process returns to step S520 for further refinement.

[0077] Step S550 involves outputting the optimized design. Once convergence is complete, the final design can be transferred to downstream modules for manufacturing. After completing step S550, the final design is evaluated against manufacturing constraints to verify its feasibility. Upon confirmation, the optimization module 140 compiles the design and its related process parameters into a structured data package, which is then transferred to the multilayer manufacturing module 150 to initiate the manufacturing process and produce a physical product.

[0078] In this invention, the generative design module 110 employs an artificial intelligence-driven approach. Specifically, it implements a diffusion model with 3D conditions to synthesize design candidate solutions that respond to material properties and manufacturing constraints. Based on artificial intelligence mechanisms, the system can explore complex design spaces inaccessible to traditional rule-based or parametric methods, particularly in handling the nonlinear, strain-rate-dependent material behavior of non-Newtonian smart materials. By integrating the AI-generated functionality with the FEA module 130 and the optimization module 140, the system 100 establishes a data-driven feedback loop where the generated designs are rigorously evaluated and iteratively improved. Simulation results can accurately predict energy absorption and deformation details, while optimization algorithms can adjust geometric features to enhance structural performance and manufacturability.

[0079] In the generative design module 110, artificial intelligence is particularly well-suited for feedback-driven and iterative design workflows. Unlike static design methods, AI-based generative models can dynamically adapt to performance evaluations and constraint updates received from other modules in the system. When integrated with the FEA module 130 and the optimization module 140, the AI ​​model can integrate simulation results and optimization feedback to dynamically improve the design output based on real-time performance data. This adaptive capability supports efficient convergence to high-performance structures, especially in applications involving non-Newtonian materials where the relationship between design and performance is highly nonlinear and sensitive to manufacturing conditions. The iterative nature of the design loop benefits from the AI ​​model's ability to generate diverse but constraint-compliant structural candidates in each iteration, thereby enhancing the system's ability to explore, learn, and optimize in complex design spaces.

[0080] Figure 6 Example diagrams of optimized designs are provided; the voxel-based structural representations shown are generative outputs with annotations for relative density and specific energy absorption. The provided examples reflect the structural configurations generated and evaluated during the design process presented in this invention.

[0081] Figure 7 Training and validation loss curves for the artificial intelligence model used in generative design module 100 are shown. The provided curves show that the loss gradually decreases as the training steps proceed, indicating that the model, after effective training, can generalize across structured samples.

[0082] Figure 8 The graphs show the predictive performance of the artificial intelligence model used in the generative design module 100. The left graph compares the predicted value of relative density with the reference value, and the right graph compares the predicted value of specific energy absorption (SEA) with the reference value at 70% strain. The corresponding mean and median relative errors indicate that the model can reasonably estimate the structural performance indicators based on the geometric input.

[0083] Here are some examples for reference:

[0084] Example 1

[0085] Generative Design and Simulation of Non-Newtonian Energy Absorbing Structures

[0086] Non-Newtonian energy-absorbing structures can be designed using the generative design module. The generative design module can generate multiple candidate designs, each represented as a voxel mesh.

[0087] Input parameters include a density of at least 0.05, a target energy absorption capacity of 5J, and a geometric constraint that limits the cell size to 4mm.

[0088] The design can be converted to SDF for analysis. Then, using the FEA module, each design is simulated under dynamic load conditions of a 2.5kg drop scenario and an impact energy of 5J. The optimization module iteratively optimizes the top few candidate designs, thus improving energy absorption capacity by at least 5%.

[0089] The optimized design features a cell density of at least 100 kg / m³ and an average pore size of at least 0.1 mm, with an energy absorption capacity of at least 1 J, which is at least 5% higher than the performance of existing technologies.

[0090] Example 2

[0091] Calibration and manufacturing process of non-Newtonian structures

[0092] It can measure the rheological properties of non-Newtonian materials to calibrate the system. The target material exhibits a shear rate-dependent viscosity, ranging from a shear rate of 1 s⁻¹. -1 10 at the time 4 Pa·s to a shear rate of 100 s -1 10 at the time -3 The viscoelastic properties were also evaluated, and the results showed that the storage modulus was 0.5-1.5 MPa and the loss modulus was 0.1-0.2 MPa.

[0093] These properties can be input into the FEA module to simulate the performance of the optimized design. The design can be fabricated using a liquid phase deposition build-up process with an injection speed of 1000-3000 mm / s, a curing time of 120-240 seconds, and a temperature of 70-90 degrees Celsius (e.g., 80 degrees Celsius). The final structure achieves an energy absorption efficiency of at least 5%, demonstrating that the proposed system can translate designs into high-performance materials.

[0094] Furthermore, a comparative experiment was conducted to evaluate the energy absorption performance of the materials of this invention and existing technologies. The non-Newtonian material structure optimized using a generative design module was tested under dynamic loading conditions. The non-Newtonian material structure could withstand an impact of 2.5 kg, with an impact energy of 5 J. Compared to a traditional honeycomb structure with a similar density (kg / m³), the optimized structure exhibited better energy absorption performance. In addition, the structure designed using this invention also demonstrated higher stability.

[0095] Example 3

[0096] Application of protective equipment

[0097] A prototype of an impact-resistant glove was also designed using the proposed system. Target performance indicators included an energy absorption threshold of 5 J and a maximum permissible transmitted force of 4 kN. The proposed system allows for optimization of the non-Newtonian material structure to meet these criteria. Furthermore, the manufactured gloves were subjected to a 2.5 kg drop test, simulating an impact force of at least 1 N.

[0098] Example 4

[0099] Aerospace structural applications

[0100] Lightweight aerospace panels can be designed using generative design and additive manufacturing modules. The designed panels can employ a non-Newtonian energy-absorbing lattice structure. Compared to traditional aluminum honeycomb panels, the optimized panels are significantly lighter while maintaining the same impact resistance under standardized testing conditions. This result demonstrates that the proposed system can manufacture practical aerospace components that combine lightweight and high performance, making it industrially viable.

[0101] In one embodiment, the resulting optimized structure comprises a size ranging from 1 to 30 mm and a bulk density ranging from 100 to 2000 kg / m³. 3 The range of elements is consistent with the geometric constraints applied during generative design.

[0102] The above description of the present invention is for illustrative and descriptive purposes only. It is not intended to be an exhaustive disclosure, nor is it intended to limit the invention to the precise forms disclosed. Those skilled in the art will understand that various modifications and variations can be made.

[0103] These embodiments were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling those skilled in the art to understand the various embodiments of the invention and the various variations suitable for particular purposes.

[0104] definition

[0105] In this specification, unless the context otherwise requires, the word "comprising," or variations thereof, such as "including" or "comprising," shall be understood to include the stated integers or groups of integers, but not exclude any other integers or groups of integers. It should also be noted that in this disclosure, particularly in the claims and / or paragraphs, terms such as "comprising," "including," and "comprising" may have the meanings assigned to them by patent law or patent practice; for example, they may allow elements not expressly listed, but exclude elements present in the prior art or elements that affect the essential or novel features of the invention.

[0106] Furthermore, throughout the specification and claims, unless the context otherwise requires, the word "comprising" or variations thereof, such as "including" or "comprising," shall be understood to imply the inclusion of a group of one, a plurality of, or more than one object, without excluding any other possible integer ordinal numbers.

[0107] In this specification, references to "an embodiment," "an example embodiment," "exemplary embodiment," etc., indicate that the described embodiment may include specific features, structures, or characteristics, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, such expressions do not necessarily refer to the same embodiment. Moreover, when a specific feature, structure, or characteristic is described in connection with an embodiment, it should be understood that the knowledge of those skilled in the art can influence that feature, structure, or characteristic in connection with other embodiments, whether or not those embodiments are explicitly described.

[0108] The terms “approximately,” “substantially,” “essentially,” and “approximately” as used herein are used to describe and explain minute variations. When used in conjunction with an event or situation, the term can refer to a situation where the event or situation occurred precisely or approximately. The term “approximately” as used herein with respect to a given value or range typically refers to a range within ±10%, ±5%, ±1%, or ±0.5% of that given value or range. This range can be expressed herein as a distance from one endpoint to another or between two endpoints. Unless otherwise stated, all ranges disclosed in this disclosure include endpoints. When referring to the same numerical value or characteristic, the term can refer to a value within ±10%, ±5%, ±1%, or ±0.5% of that average.

[0109] In the preparation method described herein, the steps can be performed in any order without departing from the principles of the invention, unless the timing or order of operations is explicitly specified. Statements in the claims regarding performing a certain step first, followed by several other steps, should be understood as meaning that the first step is performed before any other step, but the other steps can be performed in any suitable order unless the order is further specified in other steps. For example, "step A, step B, step C, step D, and step E" recorded in the claim elements should be understood as meaning that step A is performed first, step E is performed last, and steps B, C, and D can be performed in any order between steps A and E, and this order is still within the literal scope of the method in the claim. A given step or subset of steps can also be performed repeatedly. Furthermore, unless the claims explicitly state that each step is performed individually, specified steps can be performed simultaneously.

[0110] The term "signed distance fields (SDF)" is used to evaluate the structural integrity and performance of non-Newtonian materials under different loading conditions.

[0111] The definition of a "Cartesian mesh" is to calculate the signed distance of each mesh point relative to the nearest surface point within the boundary volume containing the candidate design structure.

[0112] Other definitions of the selected terms used herein can be found in the detailed description of the invention and apply throughout. Unless otherwise defined, all other technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0113] References: The publicly available information in the following references is incorporated herein by reference.

[0114] Barill, G., Dickson, N. G., Schmidt, R., Levin, DIW & Jacobson, A. (2018). “Fast winding numbers for soups and clouds”. Published in: ACM Transactions on Graphics, 37(4).

[0115] The functional units and modules of the systems and methods according to the embodiments disclosed herein can be implemented using computing devices, computer processors, or electronic circuits, including but not limited to application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, and other programmable logic devices configured or programmed according to the teachings of this disclosure. Those skilled in the art of software or electronics can easily write computer instructions or software code that execute in computing devices, computer processors, or programmable logic devices based on the teachings of this disclosure.

[0116] All or part of the methods according to the embodiments can be performed in one or more computing devices, including server computers, personal computers, laptops, and mobile computing devices (such as smartphones and tablets).

[0117] Embodiments may include computer storage media, transient and non-transient memory devices storing computer instructions or software code that can be used to program or configure computing devices, computer processors, or electronic circuits to perform any of the processes of the present invention. Storage media, transient and non-transient memory devices may include, but are not limited to, floppy disks, optical disks, Blu-ray discs, DVDs, CD-ROMs, magneto-optical disks, ROMs, RAMs, flash memory devices, or any type of medium or device suitable for storing instructions, code, and / or data.

[0118] Each functional unit and module according to various embodiments may also be implemented in a distributed computing environment and / or cloud computing environment, wherein all or part of the machine instructions are executed in a distributed manner by one or more processing devices interconnected by a communication network, such as an intranet, a wide area network (WAN), a local area network (LAN), the Internet, and other forms of data transmission media.

[0119] The foregoing description of the present invention is for illustrative purposes only. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations will be apparent to those skilled in the art.

[0120] These embodiments were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling those skilled in the art to understand the various embodiments of the invention and the various modifications suitable for a particular purpose.

Claims

1. A system for designing and manufacturing non-Newtonian energy-absorbing structures, configured to receive user-defined input parameters and generate a physical structure through an integrated design-to-manufacturing workflow, characterized in that, The system includes: A user interface for receiving input parameters, including material properties, load conditions, and geometric constraints, and converting the input parameters into structured data; A generative design module is used to receive the structured data and use a 3D conditional diffusion model to synthesize voxel-based geometries based on the physical and rheological properties of non-Newtonian materials, thereby generating candidate structures. A database is used to store datasets containing experimentally derived rheological data, including viscosity profiles, viscoelastic moduli, and manufacturing parameters. The database is also used to provide the stored datasets to all generative design modules during the generation of the candidate structures. The finite element analysis module is used to simulate the mechanical properties of the candidate structures using material models specific to non-Newtonian behavior and voxel mesh resolution consistent with the laminated manufacturing process.

2. The system according to claim 1, wherein the finite element analysis module is used to transmit the performance evaluation results to the generative design module, and the generative design module is used to receive the evaluation results as feedback to guide the subsequent design iterations of the candidate structure.

3. The system according to claim 2, further comprising: An optimization module is used to receive simulation results from the finite element analysis module and, based on energy absorption performance and manufacturing constraints, iteratively refine the structural geometry of the candidate structure to generate an optimized structure. as well as A stacking fabrication module is used to fabricate the optimized structure using a liquid phase deposition process, wherein the liquid phase deposition process is configured to accommodate the shear sensitivity and high viscosity characteristics of the selected non-Newtonian material.

4. The system according to claim 3, wherein the optimization module is used to transmit the improved structure as feedback to the generative design module, and the generative design module is used to update subsequent design outputs based on the improved structure, thereby forming a feedback loop between the optimization module and the generative design module.

5. The system according to claim 1, wherein the 3D conditional diffusion model performs a forward diffusion process to progressively convert the structured data into a noisy representation, and then uses a trained 3D U-Net to perform a reverse diffusion process to reconstruct the structural geometry that meets the conditional input.

6. In the system according to claim 1, the generative design module extracts oriented closed surfaces and calculates the directed distances between points on a Cartesian grid, converting the generated structure into a representation of a directed distance field (SDF).

7. The system according to claim 1, wherein the non-Newtonian material comprises at least one of polyurethane, polydimethylsiloxane (PDMS), styrene-ethylene-butene-styrene (SEBS), silicone, natural rubber, synthetic rubber, or shear-thickening elastomer.

8. The system of claim 1, wherein the voxel mesh used in the finite element analysis module is matched with the resolution used in the layer fabrication module to maintain fidelity between the simulation program and the fabrication program.

9. A method for designing and manufacturing non-Newtonian energy-absorbing structures, configured to receive user-defined input parameters and generate a physical structure through an integrated design-to-manufacturing workflow, characterized in that... The method includes: The system receives input parameters, including material properties, load conditions, and geometric constraints, through a user interface and converts these input parameters into structured data. The structured data is received through the generative design module; The generative design module generates candidate structures using a 3D conditional diffusion model, which synthesizes voxel-based geometries based on the physical and rheological properties of the selected non-Newtonian material. The dataset is stored in a database, and the dataset contains experimentally obtained rheological data, including viscosity curves, viscoelastic modulus, and manufacturing parameters. During the generation of the candidate structures, the stored dataset is provided to the generative design module through the database; and The mechanical properties of the candidate structure are simulated using the finite element analysis module, employing a material model specific to non-Newtonian behavior and a voxel mesh resolution consistent with the laminated manufacturing process.

10. The method of claim 9, further comprising: The performance evaluation results are transmitted to the generative design module through the finite element analysis module. as well as The generative design module receives the evaluation results and uses them as feedback to guide subsequent design iterations of the candidate structures.

11. The method of claim 10, further comprising: The simulation results from the finite element analysis module are received, and the optimization module iteratively refines the structural geometry of the candidate structure based on energy absorption performance and manufacturing constraints, thereby generating an optimized structure. as well as The stacking fabrication module is used to fabricate the optimized structure using a liquid phase deposition process, wherein the liquid phase deposition process is configured to accommodate the shear sensitivity and high viscosity characteristics of the selected non-Newtonian material.

12. The method of claim 11, further comprising: The improved structure is transmitted as feedback to the generative design module through the optimization module. as well as The generative design module updates subsequent design outputs based on the improved structure, thereby forming a feedback loop between the optimization module and the generative design module.

13. The method of claim 11, further comprising: By operating the layering manufacturing module, the optimized structure is manufactured at an injection speed of 1000 to 3000 mm / s, a curing time of 120 to 240 seconds, and a temperature of 70 to 90 degrees Celsius, thereby generating a solid product.

14. The method of claim 13, wherein the physical product exhibits an energy absorption efficiency of at least 5% under dynamic load conditions.

15. The method of claim 13, further comprising: The physical product was tested under drop impact conditions involving the release of 2.5 kg of mass, an impact energy of at least 5 joules, and controlled deformation of the physical product.

16. The method of claim 13, wherein the physical product comprises units ranging in size from 1 to 30 mm and in density from 100 to 2000 kg / m³.

17. The method of claim 9, wherein the 3D conditional diffusion model performs a forward diffusion process to progressively convert the structured data into a noisy representation, and then performs a reverse diffusion process using a trained 3D U-Net to reconstruct the structural geometry that meets the conditional input.

18. The method of claim 9, further comprising: The generative design module extracts oriented closed surfaces and calculates the directed distances between points on a Cartesian grid, converting the generated structure into a directed distance field (SDF) representation.

19. The method of claim 9, wherein the voxel mesh used in the finite element analysis module is matched to the resolution used in the layer fabrication module to maintain fidelity between the simulation program and the fabrication program.

20. The method according to claim 9, wherein the non-Newtonian material comprises at least one of polyurethane, polydimethylsiloxane (PDMS), styrene-ethylene-butene-styrene (SEBS), silicone, natural rubber, synthetic rubber, or shear-thickening elastomer.

Images