Process for designing a component

Abstract

Computer-implemented method (120) for designing a component, comprising the steps: - holding (111) at least one selected from a design, in particular a basic design, a design parameter, a simulation input (52) or a deviation from a design (53), in particular a previous design, further in particular from a basic design, in a trained Neural Circuit Policy (NCP) Network (200) comprising Liquid Time Constant (LTC) neurons; - modifying (53) at least one selected from the design or the at least one design parameter (51); - simulating (40) a deviation selected from at least one deviation (53) from the design, in particular the basic design, or a deviation (5) from a physical parameter.

Description

[0001] The invention relates to a method. The invention relates to a training method. The invention relates to a computing device. The invention relates to a computer program product. The invention relates to a memory device. The invention relates to an electronic signal.

[0002] The design of components using computer-implemented methods is known in the state of the art.

[0003] US20230237219A1 describes methods, systems, and apparatus, including media-encoded computer program products, for designing three-dimensional lattice structures, comprising, in one aspect, a method comprising: obtaining a mechanical problem definition including a 3D model of an object; generating a numerical simulation model for the 3D model of the object using one or more loading cases and one or more isotropic solid materials identified as a base material model for a design space; predicting the performance of different lattice settings in different orientations in the design space using a lattice structure behavior model in place of the base material model in the numerical simulation model;and presenting a set of grid proposals for the design space based on the predicted performance of the different grid settings in the different orientations; wherein the grid structure behavior model has been pre-calculated for the different grid settings that are generatable by the 3D modeling program.

[0004] US20210141985A1 describes that a component of the technical system, a component designation, and a characteristic designation for a characteristic relevant to the design of the components are retrieved and queried using a search engine. The documents found by the search engine are retrieved, and component information, e.g., product information for a specific component, is extracted from them. The extracted component information is fed to a machine learning routine that has been trained using a large number of predefined training component information and training characteristic parameter values ​​to reproduce predefined training characteristic parameter values ​​using predefined training component information. The output data of the machine learning routine is selected as characteristic parameter values ​​and inserted into a plan data set. The plan data set is then output for the design of the technical system.

[0005] The described processes, devices and systems do not allow for individual adaptation of components, are resource-intensive to implement and are complex.

[0006] The object of the invention is to be able to provide individualized components, while reducing the effort and saving resources.

[0007] The problem is solved in particular by a method having the features of claim 1. The problem is solved in particular by a training method having the features of claim 8. The problem is solved in particular by a computing device having the features of claim 9. The problem is solved in particular by a computer program product having the features of claim 10. The problem is solved in particular by a memory device having the features of claim 11. The problem is solved in particular by an electronic signal having the features of claim 12. Further features and details can be found in the subclaims, the description and the drawings.Features and details described in connection with the method according to the invention naturally also apply in connection with the training method according to the invention, the computer program product according to the invention, the electronic signal according to the invention, and the control device according to the invention. This also applies vice versa, so that with regard to the disclosure of the individual aspects of the invention, reference is always made to each other.

[0008] According to one aspect, the problem is solved in particular by a method having the features of claim 1.

[0009] According to this, a computer-implemented method for designing a component can be implemented. The method comprises, in particular, the step of maintaining a design or design parameter in a trained Neural Circuit Policy (NCP) network comprising Liquid Time Constant (LTC) neurons and modifying at least one design element or design parameter. Furthermore, a simulation of a deviation selected from at least one deviation from the design, in particular from the basic design, or a deviation from a physical parameter is performed.

[0010] Here and elsewhere, holding can be referred to in particular when the corresponding design and / or the design parameter is at least partially stored in a corresponding NCP, processed in it, evaluated by it or stored and / or processed in any other way as an electronic signal on a computing device and / or a storage device, in particular as described elsewhere herein.

[0011] The described method for designing a component using Liquid Time Constant (LTC) Neural Circuit Policy (NCP) Networks offers the possibility, compared to conventional techniques, that the model used for the design of the component can be smaller and thus more cost-effective (to train and operate) than conventional models.

[0012] Another advantage is that the attention maps of the neurons in the NCP network can be explainable, as the number of neurons can be very small. This allows for better understanding and tracking of the network's decisions.

[0013] Another possibility of the method is that NCP networks can be capable of making decisions without a reward function. This allows for better understanding and control of the network's decisions.

[0014] Finally, LTC calibration methods can be more general and robust than conventional methods, meaning they can be better able to meet different requirements and environmental conditions. This makes the process for designing components with LTC-NCP networks particularly interesting for the development of complex systems and applications where conventional techniques may not be sufficient.

[0015] A computer-implemented method for designing a component may be configured to include a step of maintaining a design or design parameter in a trained Neural Circuit Policy (NCP) network based on Liquid Time Constant (LTC) neurons. This means that the method may use an NCP network specifically trained for designing components, in particular by means of a training method as described elsewhere herein, and in which the information about the design or design parameter may be mapped via LTC neurons, such as described elsewhere herein.

[0016] Modifying at least one design, i.e., a design element, or design parameter, involves changes to the design's properties in order to adapt it to the desired requirements. This can be done, for example, by changing the geometry, material, or security features.

[0017] Simulating a deviation means in particular that the method can simulate a deviation from the design or a physical parameter, in particular by means of CAD, CAE, FEM and / or SW simulation, in order to be able to examine the effects of these changes on the component or to be able to design a component and thereby determine an optimized shape and / or design in order to be able to transfer this to a manufacturing process, for example as a digital twin, in particular in the form of an electronic signal, as described elsewhere herein, in order to be able to manufacture the component. These deviations can be selected from at least one deviation from the design, in particular from the basic design, or a deviation from a physical parameter.

[0018] CAD (Computer-Aided Design) refers specifically to the use of computers and software tools to support the design and construction of products or components. It encompasses a wide range of tools and techniques for creating, editing, and improving designs, including appropriate simulation capabilities and capabilities.

[0019] CAE (Computer-Aided Engineering) refers specifically to the use of computers and software tools to support the engineering process, especially in the development of products or components. CAE encompasses a variety of methods and techniques that can be used to conduct simulations to investigate the behavior of products or components, particularly in the area of ​​simulated stress tests.

[0020] FEM (Finite Element Method) refers specifically to a numerical method for solving partial differential equations, which can be used in mechanics, particularly in structural mechanics and mechanical engineering. The objects under consideration can be divided into small components (elements), whose structure and behavior can be calculated analytically or numerically. The results of the individual elements can then be combined to determine the behavior of the entire object, such as a component.

[0021] Simulation software refers specifically to programs and tools that can be used to simulate the behavior of components, systems in which these components may be integrated, or related relationships. This can support product development, mechanism investigation, or installation planning, helping to anticipate potential problems and improve the efficiency of planning, installation, and performance of the corresponding systems.

[0022] Holding a design means that the method can use an existing design to run through the NCP network, for example, to determine the corresponding effects of changes to the design or to modify a design. Alternatively or additionally, holding a design can be used to train the NCP network in a training procedure as described elsewhere herein, particularly independently of any inference of the procedure.

[0023] Holding a design parameter refers in particular to the fact that the method can use an input, determined, or simulated value of a design parameter of an existing design to run through the NCP network, for example, to determine the corresponding effects of changes to the design or to change a design. In particular, corresponding predictions about design parameters and their values ​​can also be made, which in turn can be mapped to a simulation. Alternatively or additionally, a design can be held in order to train the NCP network in a training method, as described elsewhere herein, in particular independently of any inference of the method.

[0024] A neural circuit policy network is, in particular, an artificial neural network that can be used to make decisions regarding the design of components. Liquid time constant neurons are, in particular, a type of neuron in a neural circuit policy network that is used to process information about the design or design parameters.

[0025] In one embodiment, the development of a component can be carried out using a specific class of interpretable feedback (neural circuit policies, NCPs) and a specific class of continuous-time (liquid time-constant, LTC) neural networks. In particular, the design of a mechatronic component can be carried out in which a user, especially an engineer, interacts with a CAD / CAE model by changing certain mechanical parameters such as diameter, length, and material. The CAD software can then calculate the simulation of a physical parameter, such as torsional load, and displays the results, for example, on a screen. Using this dataset, the NCPs can be further developed and the LTC network trained. Neural circuit policy networks are, in particular, a special type of recursive learning neural network that is based on the structure of the C. elegans nervous system.The behavior of the artificial neural network can be described by a policy that can map a network state and an observation to a new network state and an action. An observation can be understood as any of the inputs to the process described elsewhere. In particular, an output of the neural network can be any of the outputs described elsewhere. In particular, as explained above, the neural network introduces a design and / or a design modified by a user as an observation and maps this to other physical parameters, such as the changing torsional load. The neural network itself can be further developed in the process.Because NCP networks consist of only 5 to 50, especially 10 to 29, and especially 15 to 19 neurons, they allow for the investigation of the neurons' effects and even the prediction of how the neurons themselves might behave. This makes the NCP network faster than convoluted neural networks (CNNs), for example, and the training effort is significantly lower. This improves the traceability of the results and saves resources in the application and implementation of the method.

[0026] Modifying at least one design element or design parameter specifically involves changes to the design's properties to adapt it to the desired requirements. This also allows for rapid testing of possible configurations without having to first manufacture correspondingly designed components. This allows for faster analysis.

[0027] Simulating a deviation refers specifically to the process simulating a deviation from the design or a physical parameter in order to investigate the effects of these changes on the component. This allows changes to be simulated, for example, using a computing device as described elsewhere herein, which can then be fed into the neural network. This can be done alternatively or in addition to the changes that can be introduced into the process by a user.

[0028] Deviation from a design refers in particular to a deviation from the properties of an original design, but can also refer to deviations from a design of a previous (recursive) cycle.

[0029] Deviation from a physical parameter refers in particular to the fact that the method examines the effects of a change in a physical parameter that may be relevant to the component. This can either originate from a simulation or have already been determined. This can be done alternatively or in addition to manual input. It can also be provided that a change to a design outputs a corresponding change in a physical parameter, i.e., outputs its value change, in order to understand the effects of the design change.

[0030] According to one aspect, the method steps can be performed recursively, in particular until a predetermined value of the physical parameter is reached. Alternatively or additionally, the method steps can be performed recursively until a limit value, in particular a minimum and / or a maximum, of the physical parameter is reached.

[0031] Recursively performing the process steps means that the steps can be repeated until a specific goal is achieved. In relation to the process described above, this can mean repeating the simulations and modifications until a specified limit for the physical parameter is reached or until a minimum of the parameter is reached.

[0032] Achieving a given physical parameter means that the value of the parameter can reach a certain specified value. In relation to the process described above, this can mean achieving a certain material thickness, torsional stress (in Nm), or load.

[0033] Reaching a limit value means that, in particular, a specific value has been specified that must be reached. This can be a minimum or a maximum; in other embodiments, this can also be a value determined by other conditions.

[0034] Achieving a minimum of a physical parameter means reducing the parameter's value to the lowest possible level. In relation to the process described above, this may mean minimizing the stress on a component to maximize its service life or to save costs, particularly because that component can be replaced less frequently.

[0035] Achieving a maximum of the physical parameter means that the value of the parameter can be increased to the largest possible value. In relation to the method described above, this may mean maximizing the load on a component to maximize the service life (of the overall structure, for example) (for example, because transferring the load to a specific component can protect other components) or to save costs, particularly because other components can be replaced less frequently since the redirection of a load can be implemented.

[0036] Recursively performing the process steps means repeating the steps until the goal is achieved. In relation to the process described above, this can mean repeating the simulation and modification of the design until an optimal result is achieved or until certain requirements are met.

[0037] According to one aspect, at least one deviation from the retained design, in particular from the basic design, can be introduced by a user.

[0038] This means that at least one deviation from the held (i.e. also the processed or stored) design, in particular from the basic design, can be introduced by a user into the process for designing a component using Liquid Time Constant (LTC) Neural Circuit Policy (NCP) Networks.

[0039] The term "basic design" refers specifically to the original design of the component, which can serve as a starting point for the design. The basic design contains all the important parameters and properties of the component and forms the basis for further modifications and improvements. The component can be available as a digital twin, which can be stored in a storage medium, a storage device, or a computing device to replicate the physical parameters as faithfully as possible, thus enabling digital experiments to be conducted.

[0040] A "user" is specifically a person who uses the process for designing a component with Liquid Time Constant (LTC) Neural Circuit Policy (NCP) networks. A mechatronics engineer or a product engineer, for example, can be a user.

[0041] "Introducing a change" specifically refers to the process by which a user introduces a deviation from the established design, especially the base design. This can be done by changing parameters or properties of the component to adapt it to specific requirements or needs. This can relate, for example, to a length, a thickness, an angle within the component, or the orientation of certain groups or features.

[0042] Introducing a change can help improve the component design and increase component performance. It allows users to develop tailored solutions for specific use cases through the flexibility of the process for designing components with Liquid Time Constant (LTC) Neural Circuit Policy (NCP) networks.

[0043] In one aspect, the physical parameter may be one of a torsional stress or a tensile stress.

[0044] One aspect of the process for designing a component with Liquid Time Constant (LTC) Neural Circuit Policy (NCP) Networks is that the physical parameter can be one of a stress, in particular a torsional stress and / or a strain stress, a tension or a force.

[0045] Stress refers specifically to a parameter that can load the component. This can occur through a stress (in the material, for example) or through a force acting on the material. A physical parameter can represent intrinsic material properties, such as stress, i.e., the forces and / or stresses within the material itself. Alternatively or additionally, extrinsic properties can also be represented, such as the forces and / or stresses acting on the material from outside.

[0046] Torsional stress refers specifically to the load a component can experience due to rotational movement. This can lead to deformations and distortions of the component, which can impact its strength and durability. The influence of the design on the resulting torsional stress can be simulated using this method.

[0047] Tensile stress refers specifically to a load that a component can experience due to axial movement. This type of loading can lead to deformations and distortions in the component, which can impact its strength and durability. The influence of the design on the resulting tensile stress can be simulated using this method.

[0048] The described method for designing a component can also process other physical parameters, such as temperatures, forces, pressure, vibrations, and others. Exposure to elevated or abnormal temperatures can affect materials' strength and durability. Mechanical forces can deform, distort, or destroy the component. Compressive loads can distort, twist, or destroy the component. Vibrations can lead to deformations and warping of the component, which can affect its strength and durability. The influence of the design on changes in one or more of the described parameters can be simulated using the method.

[0049] The process of designing a component using Liquid Time Constant (LTC) Neural Circuit Policy (NCP) Networks can help to take these physical parameters into account and adapt the component design accordingly to improve its performance and durability.

[0050] In one aspect, the at least one design parameter may be at least one of a geometry parameter, a material parameter, or a safety parameter.

[0051] The term "design parameter" specifically refers to a property or factor that can be considered in the design and construction of a component. Design parameters can include both physical and technical properties of the component.

[0052] The term "geometry parameter" refers specifically to a geometric property of the component, such as its length, thickness, and / or angle. The length of a component can have a significant impact on its functionality and performance. The thickness of a component can affect both its durability and flexibility. The angle between different parts of the component can influence its functionality and performance. Any change in such a geometric parameter in an installed situation can impact the behavior of the component, which can be reflected in the physical parameters. This allows the component to be optimized using the method before a manufacturing process is carried out.

[0053] A “material parameter” refers in particular to a property of the material from which the component is made. Examples of material parameters can be a Young’s modulus, also called elastic modulus, and / or a strain coefficient. The Young’s modulus is in particular a measure of how much a material can stretch or shorten under load. The Young’s modulus (also called Young’s modulus) is a mechanical quantity that can describe the relationship between the mechanical stress and the resulting strain range in a material. In particular, it indicates how a material can react under stress by specifying how much the material can stretch when loaded under a certain stress. The Young’s modulus is in particular a physical parameter for describing the mechanical properties of materials and can be used in component simulation.The expansion coefficients describe in particular how a material can stretch or shorten in different directions.

[0054] The term "safety parameters" refers specifically to factors that contribute to the safety of a component during use. Examples of safety parameters can include load-bearing capacity and / or fracture protection. The load-bearing capacity of a component describes, in particular, how much stress it can withstand without sustaining damage, such as bending or deformation. Fracture protection can be a measure of how reliably the component is protected against fracture.

[0055] The process of designing a component using Liquid Time Constant (LTC) Neural Circuit Policy (NCP) Networks can help to consider these design parameters and adapt the component design accordingly to improve its performance, durability, and safety.

[0056] According to one aspect, the method may comprise the step of outputting at least one mathematical mapping to a simulation.

[0057] The method may include the step of outputting at least one mathematical mapping to a simulation to visualize and verify the desired design. A mathematical mapping is, in particular, a set of formulas or equations that describe the behavior of a component in a specific context. These equations can be solved using data from the simulation to generate a graphical representation of the mapping on a screen.

[0058] An image on a screen can be a two-dimensional representation that shows various parameters of the component, such as length, width, and thickness. The representation can also be three-dimensional and show the shape of the component in various directions. A simulation of the loading of the component by physical parameters—such as weight or air resistance, especially alternatively or additionally—can also be displayed to verify the component's behavior after design changes (or in different situations).

[0059] Alternatively or additionally, in particular, a mapping into a function space can be carried out, which makes it possible to check whether the component adapts to a specification, a minimum or another consideration with regard to the at least one physical parameter as a result of the design change.

[0060] According to one aspect, the method may comprise at least one step selected from increasing, decreasing, or keeping the same values ​​of the at least one design parameter based on a mathematical mapping of the simulation input and / or the design deviation to design parameter changes.

[0061] The method may comprise at least one step selected from increasing, decreasing, or keeping the same values ​​of the at least one design parameter based on a mathematical mapping of the simulation input and / or the design deviation to design parameter changes.

[0062] Increasing a value of a design parameter means, in particular, having the value of the design parameter increased or recommending an increase as an output in order to achieve the desired design or a design with desired properties and / or behaviors.

[0063] In particular, reducing the value of a design parameter means reducing the value of the design parameter to achieve the desired design. For example, you could reduce the thickness of a wall to save weight.

[0064] In particular, keeping a value of a design parameter constant means setting the value of the design parameter to a specific value in order to achieve the desired design.

[0065] According to an independent aspect, a training method for an NCP network, in particular comprising LTC, can be designed, in particular for a computer-implemented method, as described elsewhere herein. The method can comprise the step of inputting at least one selected from a design parameter, a simulation input, a simulation output, a design parameter, in particular torsional stress or tensile stress, a deviation of a design parameter between a target design layout and a current design layout, or changes to the design, in particular by a user.

[0066] The training method can be described by the features, properties, and advantages of the method, the control device, the storage device, the electronic signal, and the computer program product. This also applies across the category boundaries of method, device, and system. Thus, the method, the control device, the storage device, the electronic signal, and the computer program product can also be described by the features, properties, and advantages of the training method. For the sake of readability and compactness, a repetition of all these features, properties, and advantages is omitted.

[0067] According to an independent aspect, a computing device may be designed and configured to carry out a method, in particular as described elsewhere herein.

[0068] The computing device can be described by the features, properties, and advantages of the training method, the method, the storage device, the electronic signal, and the computer program product. This also applies across the category boundaries of method, device, and system. Thus, the training method, the method, the storage device, the electronic signal, and the computer program product can also be described by the features, properties, and advantages of the computing device. For the sake of readability and compactness, a repetition of all these features, properties, and advantages is omitted.

[0069] According to an independent aspect, a computer program product may be designed and configured to carry out a method as described elsewhere, in particular when executed on a computing device as described elsewhere herein.

[0070] The computer program product can be described by the features, properties, and advantages of the training method, the control device, the storage device, the electronic signal, and the method. This also applies across the category boundaries of method, device, and system. Thus, the training method, the control device, the storage device, the electronic signal, and the method can also be described by the features, properties, and advantages of the computer program product. For the sake of compactness and readability, a repetition of all these features, properties, and advantages is omitted here.

[0071] The computer program product can be designed and configured to be executed on a machine. When the computer program product is executed on the machine, a method can be carried out as described elsewhere. The computer program product is, in particular, machine-readable code and / or an electrical signal that is / are configured to be read by a machine in order to transmit work instructions to a machine, such as to carry out a method of the type described elsewhere. A computer program product can, in particular, be designed as machine-readable code, in particular as an algorithm that can carry out corresponding compensation. Alternatively or additionally, the machine-readable code can also carry out instructions to produce at least one component from a digital twin using additive manufacturing.For this purpose, control commands can also be transmitted to corresponding devices in order to carry out production using additive manufacturing.

[0072] According to an independent aspect, a storage device may be designed and configured to hold a computer program product as described elsewhere herein. The storage device may be designed and configured to be read by a computing device as described elsewhere herein, at least to the extent necessary to execute the computer program product to perform a method as described elsewhere herein.

[0073] A storage device can comprise a computer program product, in particular as described herein. Alternatively or additionally, the computing device, in particular as described herein, can comprise a computer program product, in particular as described herein. The computer program product is in particular designed and configured to be read from the storage device by the computing device in order to carry out one of the described methods. Thus, the storage device and / or the computing device can be described accordingly by the features, properties, and advantages as have been described and presented for one of the methods, the storage device, the electronic signal, and / or for the computer program product. This also applies vice versa. For reasons of compactness and readability, a repetition of all these features, properties, and advantages is omitted here.

[0074] In an independent aspect, an electronic signal may comprise at least one selected from a mathematical mapping to a simulation, a determined design parameter change, or a design change.

[0075] The electronic signal can be described by the features, properties, and advantages of the training method, the computing device, the storage device, the computer program product, and the method. This also applies across the category boundaries of method, device, and system. Thus, the training method, the computing device, the storage device, the computer program product, and the method can also be described by the features, properties, and advantages of the electronic signal. For the sake of compactness and readability, a repetition of all these features, properties, and advantages is omitted here.

[0076] Further advantages, features, and details of the invention will become apparent from the following description, which describes embodiments of the invention in detail with reference to the drawings. The features mentioned in the claims and in the description may be essential to the invention individually or in any combination. They show schematically: Fig. 1 shows an embodiment of a simulation; Fig. 2 shows an embodiment of a training method; and Fig. 3 a representation of an embodiment of a method.

[0077] Fig. 1 shows a representation of an exemplary, schematic embodiment of a simulation 40. In particular, a computer program product 10 is read from a storage device 20 by a computing device 30 in order to be able to carry out the simulation. In particular, a CAD module is executed which contains at least one design of a component, wherein geometric parameters such as diameter 12 or length 14 can be associated with the design. These can be mathematically represented in a design parameter map 2. Alternatively or additionally, shapes 16 of the components and / or a twist torque 18 can be mathematically represented in a further design parameter map 4. From the second design parameter map 4, a mathematical function 8 can be determined by means of an E-modulus 6 in order to, together with the first

[0078] Design parameter figure 2 simulates a simulated torsional stress 9. This can be compared with a specified target torsional stress 7 to determine a deviation 5 of a parameter, here, for example, a torsional stress deviation. This deviation can be transmitted as an electronic signal 55, in particular, further processed.

[0079] Fig. 2 shows an illustration of an embodiment of a training method 110. The training method 110 can be designed and configured to train an NCP 200 having LTC. The trained NCP 200 can be used, in particular, for a method 120 as described elsewhere herein. The training method 110 comprises, in particular, the step of inputting or holding 111 at least one selected from a current design parameter 51, a simulation input 52, and a design parameter 53. The NCP 200 here has, by way of example, six nodes arranged in three layers. The NCP 200 can be trained to predict parameter changes 113 and changes 114 in geometry parameters. The NCP 200 is stored, in particular, in the form of a computer program product 10 on a storage device 20 and can be read out by a computing device 30.The trained NCP 200 can be stored again on the storage device 20. The training method 110 can be executed fully, end-to-end, by training on the previously listed data set to detect the simulation inputs and the design deviations.

[0080] To reflect design parameter changes, in particular by increasing, decreasing or keeping the design parameter values ​​the same.

[0081] Fig. 3 shows an illustration of an embodiment of a method 120 for designing a component. The method 120 includes, in particular, the step of holding 111 or inputting at least one deviation from a design 53, in particular a previous design, further in particular from a base design, selected from a simulation input 52, into a trained Neural Circuit Policy (NCP) network (200) comprising Liquid Time Constant (LTC) neurons. A current parameter value 51 can also be input.

[0082] The method may include the step of modifying 53 at least one selected from the design or the at least one design parameter 51. This may be done, in particular, via an input mask of a computer device, for example, by a user, in order to influence, in particular, the geometry parameters.

[0083] The method may comprise simulating 40 a deviation selected from at least one deviation 53 from the design, in particular the basic design, or a deviation 5 from a physical parameter.

[0084] As in the Fig. 3, the method steps are carried out recursively until a predetermined value 7 of the physical parameter 54 is reached. Alternatively or additionally, the method steps can be carried out recursively until a minimum of the physical parameter 54 is reached.

[0085] The simulation input 52 can in particular be submitted as a copy to a simulation in order to create the corresponding mathematical design parameter figures 2, 4, as in relation to the Fig. 1. The simulation 40 can be carried out in a corresponding manner in order to return a simulation output 15 as a parameter value change 5 to the NCP 200 as input 111. In this case, the NCP 200 can perform a mathematical Fig. into the simulation 60, whereby the introduced values, for example for geometry parameters, physical parameters or other corresponding values, are used to recursively reduce the optimization, in particular the minimization of the physical parameter 5.

[0086] The NPC 200 can be trained easily and quickly, and the method 120 enables a user to see the effects of changes to the component, particularly “on the fly,” and to modify them if necessary in order to obtain a shape and / or geometry-optimized component for a given requirement and / or safety.

[0087] The above explanation of the embodiments describes the present invention exclusively by way of examples. Of course, individual features of the embodiments can be freely combined with one another, provided they are technically feasible, without departing from the scope of the present invention. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] US 20230237219A1

[0003] US 20210141985A1

[0004]

Claims

[1] Computer-implemented method (120) for designing a component, comprising the steps: - a holding (111) at least one selected from a design, in particular a basic design, a design parameter, a simulation input (52) or a deviation from a design (53), in particular a previous design, further in particular from a basic design, in a trained Neural Circuit Policy (NCP) Network (200) comprising Liquid Time Constant (LTC) neurons; - one modification (53) of at least one selected from the design or the at least one design parameter (51); - simulating (40) a deviation selected from at least one deviation (53) from the design, in particular the basic design, or a deviation (5) from a physical parameter. [2] Method (120) according to claim 1, characterized by, that the process steps are carried out recursively until a predetermined value (7) of the physical parameter (54) is reached and / or a limit value, in particular a minimum and / or a maximum, of the physical parameter (54) is reached. [3] Method (120) according to one of claims 1 or 2, characterized by , that at least one deviation from the design (53), in particular from the basic design, is introduced by a user. [4] Method (120) according to any of the preceding claims, characterized by , that the physical parameter (54) is one of a stress, in particular a torsional stress and / or a tensile stress, tension or force. [5] Method (120) according to any of the preceding claims, characterized by , that the at least one design parameter (51) is at least one of a geometry parameter, a material parameter or a safety parameter. [6] Method (120) according to any of the preceding claims, characterized by the step of outputting (112) a mathematical mapping (60) to a simulation (40). [7] Method (120) according to any of the preceding claims, characterized by the step of at least one selected from increasing, decreasing or keeping constant values ​​of at least one design parameter (54), based on a mathematical mapping (60) of the simulation input (52) and / or the design deviation (53) to design parameter changes (5). [8] Training method (110) trained and set up to train an NCP (200) to train an LTC, in particular for a method (120) according to any of the preceding claims, comprising the step: -an input (111) of at least one selected from a design parameter, a simulation input (52), a simulation output (15), a design parameter (53), in particular torsional stress or strain stress, a deviation of a design parameter between a target design layout and an instantaneous design layout, or changes (53) to the design, in particular by a user. [9] Computing device (30) trained and set up to carry out a method (110, 120) according to one of the preceding claims. [10] Computer program product (10), designed and configured to perform a method (110, 120) according to any one of claims 1 to 8, in particular when executed on a computing device (30) according to claim 9. [11] Storage device (20) comprising a computer program product (10) according to claim 10, in particular configured and designed to be read out by a computing device (30) according to claim 9 at least to the extent necessary to execute the computer program product (10) in order to execute a method (110, 120) according to any one of claims 1 to 8. [12] Electronic signal (55) comprising at least one selected from a mathematical mapping (60) to a simulation (40), a determined design parameter change or a design change (53).

Images