Curved surface optimization processing method and device, electronic equipment, medium and product

Abstract

The embodiment of the invention provides a curved surface optimization processing method and device, electronic equipment, a medium and a product. The method comprises the following steps: in response to an adjustment instruction for a to-be-optimized curved surface, performing performance prediction based on an agent model according to the adjustment instruction to obtain a performance prediction result; wherein the performance prediction result is the physical performance of the to-be-optimized curved surface after optimization and adjustment based on the adjustment instruction; performing optimization adjustment on the to-be-optimized curved surface based on the adjustment instruction to obtain a target curved surface; and visually displaying the target curved surface and the performance prediction result. The method is used for achieving the effect of improving the efficiency of curved surface optimization treatment.

Description

Technical Field

[0001] This application relates to the field of simulation design technology, and in particular to a method, apparatus, electronic device, medium and product for optimizing curved surfaces. Background Technology

[0002] In the field of industrial design, it is often necessary to perform simulation design for the geometry of complex curved surfaces. The curved surfaces obtained from the simulation design can be used to analyze their physical properties in products used in actual industrial applications.

[0003] In some techniques, users adjust the parameters of a surface in simulation design software to obtain a surface model. This model is then imported into physical simulation software for high-precision simulation to obtain simulation results of the surface model's physical properties. However, this technique is time-consuming and inefficient in terms of surface optimization.

[0004] Therefore, there is an urgent need for a solution that can improve the efficiency of surface optimization processing. Summary of the Invention

[0005] The surface optimization processing method, apparatus, electronic device, medium, and product provided in the embodiments of this application are used to improve the efficiency of surface optimization processing.

[0006] In a first aspect, embodiments of this application provide a method for optimizing curved surfaces, comprising:

[0007] In response to the adjustment command for the surface to be optimized, a performance prediction is performed based on the surrogate model according to the adjustment command, and the performance prediction result is obtained; wherein, the performance prediction result is the physical performance of the surface to be optimized after optimization based on the adjustment command;

[0008] The surface to be optimized is optimized and adjusted based on the adjustment command to obtain the target surface;

[0009] The target surface and performance prediction results are visualized.

[0010] In one possible implementation, the adjustment instructions include: the matrix index of the control points on the surface to be optimized, and the displacement change vector of the control points;

[0011] Based on the adjustment instructions, performance prediction is performed using the proxy model to obtain the performance prediction results, including:

[0012] Based on the matrix index, the first region on the surface to be optimized is determined; where the first region is the surface region on the surface to be optimized that is affected by the adjustment command;

[0013] For the first region, the geometric feature vector of the first region is determined based on the displacement change vector;

[0014] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction results.

[0015] In one possible implementation, the surface to be optimized has the original physical properties;

[0016] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction results, including:

[0017] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the physical performance change; whereby the physical performance change represents the change in physical performance caused by the optimization adjustment of the surface to be optimized based on the displacement change vector.

[0018] The performance prediction results are determined based on the original physical properties and the changes in physical properties.

[0019] In one possible implementation, the first region includes at least one parameter point;

[0020] The surface to be optimized is optimized and adjusted based on adjustment commands to obtain the target surface, including:

[0021] For the parameter points in the first region, determine the vertex coordinates of the parameter points based on the displacement change vector;

[0022] Based on the vertex coordinates, determine at least one geometric face in the first region;

[0023] At least one geometric surface in the first region is rendered, and the target surface is constructed based on the rendering result and the second region; wherein, the second region is the surface region in the surface to be optimized other than the first region.

[0024] In one possible implementation, after optimizing the surface to be optimized based on adjustment instructions to obtain the target surface, the method further includes:

[0025] The simulation module is invoked to perform performance analysis on the target surface and obtain the actual physical properties of the target surface.

[0026] Training data is constructed based on the surface parameters of the target surface and its actual physical properties.

[0027] The proxy model is trained based on the training data to update its parameters, resulting in an updated proxy model.

[0028] In one possible implementation, the surrogate model is any one or a combination of the following models: Gaussian regression model, multinomial model, radial basis function network, neural network model, and Kriging model.

[0029] In one possible implementation, the performance prediction results are visualized, including at least one of the following:

[0030] Visualize the numerical results of performance predictions;

[0031] The trend of performance prediction results is visualized in the form of a curve; the trend is determined based on the original physical properties and performance prediction results of the surface to be optimized.

[0032] The gradient information of the performance prediction results is visualized in the form of arrows; where the gradient information is calculated by the surrogate model based on the displacement change vector of the control point indicated by the adjustment command and the performance prediction results; the direction of the arrow is the same as or opposite to the direction of the gradient information; the direction of the gradient information represents the displacement direction of the control point that increases the physical performance of the surface to be optimized the fastest.

[0033] In one possible implementation, the method further includes:

[0034] If the optimization objective of the surface to be optimized is to maximize physical performance, then the direction of the arrow is determined to be the same as the direction of the gradient information.

[0035] If the optimization objective of the surface to be optimized is to minimize physical performance, then the direction of the arrow is determined to be opposite to the direction of the gradient information.

[0036] Secondly, embodiments of this application provide a surface optimization processing apparatus, comprising:

[0037] The processing module is used to respond to the adjustment command for the surface to be optimized, and to perform performance prediction based on the proxy model according to the adjustment command to obtain the performance prediction result; wherein, the performance prediction result is the physical performance of the surface to be optimized after optimization based on the adjustment command;

[0038] The processing module is also used to optimize and adjust the surface to be optimized based on adjustment instructions to obtain the target surface;

[0039] The display module is used to visualize the target surface and performance prediction results.

[0040] In one possible implementation, the adjustment instructions include: the matrix index of the control points on the surface to be optimized, and the displacement change vector of the control points;

[0041] Based on the adjustment instructions, performance prediction is performed using the proxy model to obtain the performance prediction results. The processing module is used for:

[0042] Based on the matrix index, the first region on the surface to be optimized is determined; where the first region is the surface region on the surface to be optimized that is affected by the adjustment command;

[0043] For the first region, the geometric feature vector of the first region is determined based on the displacement change vector;

[0044] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction results.

[0045] In one possible implementation, the surface to be optimized has the original physical properties;

[0046] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction results. The processing module is used for:

[0047] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the physical performance change; whereby the physical performance change represents the change in physical performance caused by the optimization adjustment of the surface to be optimized based on the displacement change vector.

[0048] The performance prediction results are determined based on the original physical properties and the changes in physical properties.

[0049] In one possible implementation, the first region includes at least one parameter point;

[0050] The surface to be optimized is optimized and adjusted based on adjustment commands to obtain the target surface. The processing module is used for:

[0051] For the parameter points in the first region, determine the vertex coordinates of the parameter points based on the displacement change vector;

[0052] Based on the vertex coordinates, determine at least one geometric face in the first region;

[0053] At least one geometric surface in the first region is rendered, and the target surface is constructed based on the rendering result and the second region; wherein, the second region is the surface region in the surface to be optimized other than the first region.

[0054] In one possible implementation, after the surface to be optimized is optimized based on the adjustment command to obtain the target surface, the processing module is further configured to:

[0055] The simulation module is invoked to perform performance analysis on the target surface and obtain the actual physical properties of the target surface.

[0056] Training data is constructed based on the surface parameters of the target surface and its actual physical properties.

[0057] The proxy model is trained based on the training data to update its parameters, resulting in an updated proxy model.

[0058] In one possible implementation, the surrogate model is any one or a combination of the following models: Gaussian regression model, multinomial model, radial basis function network, neural network model, and Kriging model.

[0059] In one possible implementation, the performance prediction results are visualized, and the visualization module is used to perform at least one of the following:

[0060] Visualize the numerical results of performance predictions;

[0061] The trend of performance prediction results is visualized in the form of a curve; the trend is determined based on the original physical properties and performance prediction results of the surface to be optimized.

[0062] The gradient information of the performance prediction results is visualized in the form of arrows; where the gradient information is calculated by the surrogate model based on the displacement change vector of the control point indicated by the adjustment command and the performance prediction results; the direction of the arrow is the same as or opposite to the direction of the gradient information; the direction of the gradient information represents the displacement direction of the control point that increases the physical performance of the surface to be optimized the fastest.

[0063] In one possible implementation, the display module is also used for:

[0064] If the optimization objective of the surface to be optimized is to maximize physical performance, then the direction of the arrow is determined to be the same as the direction of the gradient information.

[0065] If the optimization objective of the surface to be optimized is to minimize physical performance, then the direction of the arrow is determined to be opposite to the direction of the gradient information.

[0066] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0067] The memory stores the instructions that the computer executes;

[0068] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0069] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0070] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0071] The surface optimization processing method, apparatus, electronic device, medium, and product provided in this application embodiment allow users to adjust the surface to be optimized. Responding to adjustment commands, a performance prediction is performed on the adjusted surface based on a proxy model to obtain a performance prediction result. Furthermore, the surface to be optimized is adjusted according to the adjustment commands, and a target surface is rendered. The adjusted target surface and the corresponding performance prediction result can be visualized on a preset interface. This achieves the effect of improving the efficiency of surface optimization processing. Attached Figure Description

[0072] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0073] Figure 1 A flowchart illustrating the surface optimization method provided in this application. Figure 1 ;

[0074] Figure 2 A flowchart illustrating the surface optimization method provided in this application. Figure 2 ;

[0075] Figure 3 A flowchart illustrating the surface optimization method provided in this application. Figure 3 ;

[0076] Figure 4 A flowchart illustrating the surface optimization method provided in this application. Figure 4 ;

[0077] Figure 5 A schematic diagram of the structure of the surface optimization processing device provided in this application;

[0078] Figure 6 A schematic diagram of the structure of the electronic device provided in this application.

[0079] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0080] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0081] First, let me explain the terms used in this application:

[0082] The surface to be optimized refers to the original simulated surface, which may have multiple control points. Users can drag these control points with the mouse to adjust the shape of the surface. Specifically, the surface involved in this application can be a non-uniform rational B-spline surface (NURBS surface). Due to its powerful expressive capabilities and unified description of conic sections and free curves, NURBS surface has become the industry standard for describing complex surfaces in Computer-Aided Design (CAD).

[0083] Physical properties refer to the objective physical characteristics, such as mechanical, thermodynamic, or fluid dynamics, exhibited by a surface based on its geometric shape under actual engineering operating conditions. These properties can be used to evaluate the practicality of the surface in engineering applications.

[0084] In aerospace, automotive, shipbuilding, and high-end consumer product design, the geometry of complex surfaces directly impacts core product performance indicators such as aesthetics, aerodynamics, and structural strength. Therefore, simulation-based design is typically required for complex surface geometries. The surfaces derived from these simulations can then be used to analyze their physical properties in actual industrial applications.

[0085] For example, when optimizing the curves of a car body, car designers need to balance the drag coefficient with the aesthetics of the design; aircraft wing designs need to consider both aerodynamic efficiency and material stress distribution; and ship hull designs need to make trade-offs between hydrodynamic properties and construction costs.

[0086] In some embodiments, this is achieved based on a traditional simulation design process. Users modify the control points, weights, or node vectors of the NURBS surface in the simulation design software. The updated geometric model is then exported and submitted to physical simulation software such as Computational Fluid Dynamics (CFD) or Finite Element Analysis (FEA) for high-precision simulation calculations of its physical properties. This process typically takes several hours or even days.

[0087] Furthermore, based on the simulation results (such as drag coefficient and stress distribution), the design strategy of modifying the NURBS surface is judged manually to determine whether it is reasonable, and then the surface is readjusted in the simulation software.

[0088] As can be seen from the above embodiments, the above embodiments have technical problems such as long processing time and low efficiency of surface optimization processing.

[0089] The surface optimization method provided in this application involves the user adjusting the surface to be optimized. Responding to the adjustment command, a proxy model is used to predict the performance of the adjusted surface, yielding the performance prediction result. This avoids the need for high-precision simulation to predict the physical performance of the adjusted surface in traditional simulation design processes, thus shortening the computation time for physical performance calculations. The method involves adjusting the surface to be optimized according to the adjustment command and rendering the target surface. The adjusted target surface and the corresponding performance prediction result can be visualized on a pre-defined interface, avoiding the problem in traditional simulation design processes where the adjustment results and performance prediction results cannot be simultaneously perceived by the user. Based on these two aspects, the efficiency of surface optimization processing can be improved.

[0090] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0091] Figure 1 A flowchart illustrating the surface optimization method provided in this application. Figure 1 ,like Figure 1 As shown, the method includes:

[0092] Step 101. In response to the adjustment command for the surface to be optimized, perform performance prediction based on the surrogate model according to the adjustment command to obtain the performance prediction result.

[0093] Among them, the performance prediction result is the physical performance of the surface to be optimized after optimization based on the adjustment command.

[0094] For example, the surface to be optimized is a NURBS surface. The user can interact by dragging control points on the NURBS surface. In response to the user's dragging of the control points on the NURBS surface, an adjustment command is generated. This adjustment command instructs the user to adjust the shape of the surface to be optimized.

[0095] The performance of the surface after the adjustment command is executed based on the surrogate model is predicted, and the performance prediction result is obtained. This performance prediction result characterizes the physical properties of the surface after the adjustment command is executed. Optionally, the physical properties may include one or more of the following: mechanical properties, thermodynamic properties, hydrodynamic properties, acoustic properties, and optical properties.

[0096] For example, mechanical properties can include stress distribution (such as the maximum stress value of a NURBS surface under load in a mechanical part), stiffness (such as the resistance to deformation of a NURBS surface in an aerospace structural component), and fatigue strength (such as the durability of a NURBS surface under alternating loads in an automotive structural component).

[0097] For example, thermodynamic properties can include heat transfer efficiency and thermal stability. For instance, the heat dissipation effect and resistance to high-temperature deformation of NURBS surfaces in aero-engine blades.

[0098] For example, fluid dynamics performance can include aerodynamic drag and fluid lift. Examples include the air drag of NURBS surfaces in aircraft wings, or the fluid lift of NURBS surfaces in ship hulls.

[0099] For example, acoustic performance can refer to the sound wave reflection characteristics of the NURBS surface in a sound-generating device. Optical performance can refer to the light transmission uniformity of the NURBS surface in an optical lens.

[0100] It should be noted that the proxy model can be pre-trained. The trained proxy model can be deployed to the terminal device or embedded in the application on the terminal device. On the terminal device, the proxy model only performs the forward propagation inference process, which has relatively low computational cost and is relatively fast. Optionally, the training process of the proxy model can be executed on the server side. Furthermore, multiple Graphics Processing Units (GPUs) can be deployed on the server to train the initial model, thereby obtaining the proxy model.

[0101] Specifically, adjustment instructions can include the user's operational features of control points on the surface to be optimized. For example, operational features include: the coordinates of the control points, the weight of each control point on the surface to be optimized, and the vector used to move the control points. These operational features are then input into the surrogate model for performance prediction processing to obtain the performance prediction results.

[0102] Step 102. Optimize and adjust the surface to be optimized based on the adjustment command to obtain the target surface.

[0103] For example, taking a NURBS surface as the surface to be optimized, the surface to be optimized has original parameters. The original parameters of the surface to be optimized may include, but are not limited to: control point matrix, B-spline basis function degree, node vector, and weight of each control point.

[0104] Furthermore, the adjustment command can be a user dragging a control point on the NURBS surface. Based on the selected control point, the support region affected by that control point during this adjustment process is calculated. For example, if the degree of the NURBS surface is p in the U direction and q in the V direction, then the support region affected by that control point can be represented as follows: .

[0105] It is understandable that the parameter points in this support area will change shape in response to the adjustment command.

[0106] Furthermore, based on the new coordinates after dragging the control points, the basis function values ​​within the support region are recalculated. According to the parametric equations of the NURBS surface itself, the parameter values ​​of the parameter points within the support region are substituted to obtain the coordinates of the parameter points in the new surface, replacing the corresponding parameter points in the surface to be optimized. Then, based on the coordinates of the parameter points in the new surface, a geometric surface is generated as a sub-surface. This sub-surface within the support region is then joined with the non-support region portion of the surface to be optimized to obtain the target surface. This target surface is the updated surface obtained after the surface to be optimized has been optimized in response to the adjustment command.

[0107] Step 103. Visualize the target surface and performance prediction results.

[0108] For example, through a human-computer interactive visual interface, the target surface after adjustment and optimization can be visualized and a graphical representation of the target surface can be obtained.

[0109] Optionally, during the visualization of the target surface, sub-surfaces in the supporting area and some surfaces in the non-supporting area can be displayed using graphics of different colors.

[0110] For example, the performance prediction results can also be visualized through the above-mentioned visualization interface, thereby directly showing users how the adjustment command has affected the relevant physical properties of the NURBS surface.

[0111] Optionally, the numerical values ​​of the performance prediction results can be displayed as text boxes. The text boxes will display the numerical text of the performance prediction results.

[0112] Optionally, the trend of the performance prediction results can be displayed in a text box. The text box contains descriptive text describing the trend. For example, the descriptive text could be: "This adjustment to the surface reduced stress concentration by 20% and increased fatigue strength by 0.4 MPa."

[0113] It should be noted that the labels for steps 101 and 102 above do not indicate any limitation on the execution order; and Figure 1 The exemplary arrows in steps 101 and 102 shown are only one example, where steps 101 and 102 can be executed sequentially according to the label order. However, in another example, steps 101 and 102 can also be executed simultaneously. It can be understood that during the simultaneous execution of steps 101 and 102, in response to the adjustment instructions for the surface to be optimized, the following actions are performed: based on the adjustment instructions, performance prediction is performed using a surrogate model to obtain the performance prediction result; and the surface to be optimized is optimized based on the adjustment instructions to obtain the target surface.

[0114] The surface optimization method provided in this application involves the user adjusting the surface to be optimized. Responding to the adjustment command, a proxy model is used to predict the performance of the surface after adjustment, yielding a performance prediction result. This avoids the need for high-precision simulation to predict the physical performance of the adjusted surface in traditional simulation design processes, significantly reducing computational complexity and shortening the calculation time for physical performance. The method involves adjusting the surface to be optimized according to the adjustment command and rendering the target surface. The adjusted target surface and the corresponding performance prediction result can be visualized on a preset interface, avoiding the problem in traditional simulation design processes where the adjustment result and performance prediction result cannot be simultaneously perceived by the user. This allows designers to obtain millisecond-level performance feedback during interactive design. Based on these two aspects, smooth interaction is ensured, and the efficiency of surface optimization processing is improved.

[0115] As can be seen from the foregoing embodiments, the surface to be optimized can be a NURBS surface. NURBS surfaces exhibit locality characteristics. The locality characteristic of a NURBS surface means that modifying a single control point, weight, or local node vector only affects a limited local region of the surface and does not affect the global area.

[0116] To further improve the efficiency of surface optimization processing, for the surface performance prediction process, we can first identify the affected parts of the surface, and then use a surrogate model to predict the performance prediction results of the optimized and adjusted surface for the affected parts.

[0117] Figure 2 A flowchart illustrating the surface optimization method provided in this application. Figure 2 This embodiment is in Figure 1 Based on the examples, the process of obtaining performance prediction results by performing performance prediction on the proxy model in step 101 is described in detail.

[0118] The adjustment instructions include: the matrix index of the control points on the surface to be optimized, and the displacement vector of the control points. In other words, the adjustment instructions include the control points selected by the user on the surface to be optimized, and the movement of those control points when dragged. Specifically, the matrix index is used to determine which point on the surface to be optimized is the user-selected control point. Based on this matrix index, the range of the surface affected by the control point can be determined. Specifically, the displacement vector is used to represent the movement of the control point when dragged by the user.

[0119] For example, the adjustment command is the displacement vector of the control points caused by the user dragging the control points on the surface to be optimized.

[0120] Optionally, if the user uses a mouse for 2D interaction, the displacement vector is a two-dimensional vector; if the user uses a VR controller for 3D interaction, the displacement vector is a three-dimensional vector.

[0121] Based on this, such as Figure 2 As shown, the method includes:

[0122] Step 201. Determine the first region on the surface to be optimized based on the matrix index.

[0123] The first region is the surface region in the surface to be optimized that is affected by the adjustment command.

[0124] For example, the first region on the surface to be optimized is determined according to the matrix index in the adjustment instruction. As can be seen from the foregoing embodiments, the first region mentioned in this embodiment can be understood as the support region in the foregoing embodiments.

[0125] For example, the adjustment command instructs the user to drag control point A. The matrix index of control point A is (i, j). Then, the first region of the surface to be optimized affected by control point A can be represented as follows: The degree of the NURBS surface is p in the U direction and q in the V direction.

[0126] It can be understood that in the first area corresponding to control point A Only when the parameter points in the control point A move will their shapes change.

[0127] Step 202. For the first region, determine the geometric feature vector of the first region based on the displacement change vector.

[0128] For example, the geometric feature vector characterizes the geometric features of the subsurface formed by the first region after the control points change. Specifically, the geometric feature vector may include: the average curvature of the first region, and / or the change in the normal vector of the first surface.

[0129] Specifically, determining the average curvature of the first region based on the displacement change vector can be achieved through the following steps A1 to A4:

[0130] Step A1. Update the three-dimensional coordinates of the control points using the displacement change vector.

[0131] Step A2. For each parameter point in the first region, recalculate the tangent direction and curvature trend of the surface at that parameter point using the updated three-dimensional coordinates of the control points.

[0132] The tangent direction includes the tangent direction of the surface along the U direction and the tangent direction of the surface along the V direction. The bending tendency includes the degree to which the surface deviates from the tangent direction along the U direction and the degree to which the surface deviates from the tangent direction along the V direction.

[0133] Step A3. Based on the tangent direction and curvature trend of each parameter point in the first region, calculate the principal curvature in the U direction and the principal curvature in the V direction of each parameter point in the first region. Then, take the average value of the principal curvature in the U direction and the principal curvature in the V direction to obtain the average curvature of each parameter point in the first region.

[0134] Step A4. Sum the average curvature of each parameter point in the first region, and then divide by the number of parameter points in the first region to obtain the average curvature of the first region.

[0135] Specifically, determining the change in the normal vector of the first region based on the displacement change vector can be achieved through the following steps B1 to B5:

[0136] Step B1. For the first region of the surface to be optimized, calculate the original normal vectors of each parameter point in the first region before the control points are moved.

[0137] Step B2. Update the first region according to the displacement change vector so that the surface shape of the first region changes.

[0138] Step B3. For each parameter point in the first region of the surface after the shape change, recalculate the updated normal vector of each parameter point in the first region.

[0139] Step B4. For each parameter point, subtract the corresponding updated normal vector from the original normal vector to obtain the change in the normal vector of each parameter point in the first region.

[0140] Step B5. Sum the changes in the normal vectors of all parameter points in the first region, and then divide by the number of parameter points in the first region to obtain the changes in the normal vectors of the first region.

[0141] Step 203. Based on the displacement change vector and geometric feature vector, perform performance prediction using the surrogate model to obtain the performance prediction results.

[0142] For example, the displacement change vector and geometric feature vector are used to construct the input vector, which is then input into the surrogate model. The surrogate model is used for performance prediction processing to obtain the physical performance of the surface after the adjustment command is executed, i.e., the performance prediction result.

[0143] In the above embodiments, the region affected by the adjustment command in the surface to be optimized is first determined according to the adjustment command, and then the local geometric features of the region are determined. Based on the surrogate model, performance prediction is performed according to the local geometric features and the displacement change vector of the adjustment command. On the one hand, it eliminates the need to wait for the reconstruction of the entire surface before performing high-precision physical performance simulation, shortening the surface optimization processing time. On the other hand, by extracting only the geometric features of the affected region as input to the surrogate model, global feature redundancy is avoided, allowing the prediction model to focus more on key regions, thereby improving the accuracy of the performance prediction results.

[0144] Based on the above embodiments, the surface to be optimized has the original physical properties.

[0145] For example, when the surface to be optimized has not undergone any optimization adjustments, its original physical properties can be obtained using high-precision physical simulation software. Optionally, the original physical properties are denoted as... .

[0146] Furthermore, step 203 above may specifically include:

[0147] Step 2031. Based on the displacement change vector and geometric feature vector, perform performance prediction based on the surrogate model to obtain the physical performance change.

[0148] Among them, the change in physical properties characterizes the change in physical properties of the surface to be optimized after optimization and adjustment based on the displacement change vector.

[0149] For example, an input vector is constructed by combining the displacement change vector and the geometric feature vector, and then input into a surrogate model. The surrogate model is then used for performance prediction to obtain the change in physical performance.

[0150] It should be noted that the original physical performance refers to the overall physical performance of the surface to be optimized. While the physical performance change output by the surrogate model is based on the input vector of the first region, the surrogate model can determine, based on the relevant change characteristics of the first region, how such a change will affect the overall physical performance of the surface.

[0151] Optionally, the change in physical properties can be denoted as For example, a change in physical properties can be a change in the drag coefficient.

[0152] Step 2032. Determine the performance prediction results based on the original physical properties and the changes in physical properties.

[0153] For example, the original physical properties and the changes in physical properties are added together to obtain the performance prediction result. Optionally, the performance prediction result is denoted as... That's understandable. .

[0154] In the example above, the change in the physical performance of the entire surface after optimization, predicted by the surrogate model, combined with the original physical performance of the surface to be optimized, allows for the calculation of the final performance prediction result. This avoids the surrogate model directly predicting the performance of the entire surface; instead, it predicts the shape changes of the affected surface regions and their impact on the overall physical performance of the surface. On one hand, this avoids directly predicting the overall surface performance, reducing computational load and processing time; on the other hand, it improves the accuracy of the performance prediction results by inferring the performance changes of the affected surface regions.

[0155] In the foregoing Figure 2 Based on the illustrated embodiment, it can be seen that by combining the user's adjustment instructions for the surface to be optimized, a first region can be determined. This first region is understood to be the surface region affected by the user's adjustment. Furthermore, because NURBS surfaces possess locality characteristics, the efficiency of surface optimization processing can be further improved by allowing for local updates during the adjustment of the surface to be optimized.

[0156] In one example, the first region includes at least one parameter point.

[0157] For example, since the surface to be optimized includes multiple control points, the adjustment command moves one of the control points, and the control point enclosed by the first region is at least one parameter point. These parameter points move as a result of the adjustment command moving the control point.

[0158] Figure 3 A flowchart illustrating the surface optimization method provided in this application. Figure 3 This embodiment details the process of optimizing and adjusting the surface to obtain the target surface. For example... Figure 3 As shown, the method includes:

[0159] Step 301. For the parameter points in the first region, determine the vertex coordinates of the parameter points based on the displacement change vector.

[0160] It should be noted that the vertex coordinates of the parameter point refer to its three-dimensional coordinates in actual three-dimensional space. Furthermore, it can be understood that the displacement vector in the adjustment command is applied to a control point, which also belongs to the parameter point in the first region.

[0161] Therefore, the updated vertex coordinates corresponding to the control points are first determined based on the displacement change vector.

[0162] For each parameter point in the first region, the intermediate coordinates of each parameter point are calculated based on the displacement change vector. Then, the intermediate coordinates are multiplied by the weights of each parameter point relative to the control points to obtain the vertex coordinates of each parameter point.

[0163] For example, taking a three-dimensional operation, the original coordinates of the parameter point are three-dimensional, and the displacement vector is also three-dimensional. The elements in the displacement vector are added to the corresponding dimensions in the original coordinates to obtain the vertex coordinates.

[0164] Step 302. Determine at least one geometric face in the first region based on the vertex coordinates.

[0165] For example, the geometric surface can be a polygonal surface. Taking a triangular surface as an example, the parameter points in the first region are traversed, and the triangles formed by three adjacent parameter points are connected in order according to the vertex coordinates of the parameter points to form a triangular surface.

[0166] Repeat this operation until all parameter points in the first region are reasonably divided into at least one continuous triangular face.

[0167] Step 303. Render at least one geometric surface in the first region, and construct the target surface based on the rendering result and the second region.

[0168] The second region refers to the surface region in the surface to be optimized, excluding the first region. In other words, the second region is the surface region in the surface to be optimized that was not affected by this adjustment command. Therefore, during the re-rendering and reconstruction of the target surface, the rendering results of the second region can be reused to reduce the amount of rendering computation.

[0169] Based on the previous example, multiple consecutive triangular faces in the first region are rendered to obtain the rendering result of the first region. Then, the rendering results of the first region and the second region are merged to obtain the target surface.

[0170] Optionally, the process of fusing the rendering results to obtain the target surface may also include: performing a connectivity check on the connection between the first and second regions. This connectivity check can be performed by checking whether the edges of the triangular faces in the first region and the triangular faces in the second region coincide at the connection point. If they coincide, the connectivity check is considered passed; otherwise, it is considered failed. If the connectivity check fails, the triangular faces in the first region can be re-divided to improve the smoothness and continuity of the connection.

[0171] Optionally, the above process of rendering the triangular facets of the first region to obtain the rendering result can be completed on the graphics card using the parallel computing capabilities of modern GPUs, achieving extreme speed.

[0172] Optionally, steps 101 and 102 mentioned in the foregoing embodiments can be executed synchronously. During the synchronous execution of steps 101 and 102, the first region on the surface to be optimized is determined according to the displacement change vector indicated by the adjustment instruction. Then, the above steps are divided into thread A and thread B, and thread A and thread B are executed in parallel. Thread A is used to execute steps 202 to 203 above; thread B is used to execute steps 301 to 303 above.

[0173] In the example above, when the user adjusts the surface to be optimized according to the adjustment command, the affected surface region, i.e., the first region, is accurately calculated based on the locality characteristics of NURBS surfaces. Then, the vertex coordinates are calculated only for the parameter points in this region, and the geometric surface formed by the parameter points is recalculated based on the vertex coordinates. The new geometric surface in the first region and the unaffected second region are rendered to obtain the target surface. This reduces the computational complexity from the entire surface to a local region. The complexity decreases from O(M×N) to O(m×n), where m is much smaller than M and n is much smaller than N, thus achieving real-time visual updates.

[0174] In some embodiments, once the surrogate model is built, it remains static during the optimization process applied to the surface to be optimized. The surrogate model is constructed using a large amount of sample data, including surface parameters and the surface's physical performance response. This process is computationally intensive and inefficient. Furthermore, the static nature of the surrogate model may lead to inaccurate performance predictions.

[0175] Based on any of the foregoing embodiments, a simulation module can also be invoked to perform high-precision physical performance simulation and to learn the proxy model online.

[0176] Figure 4 A flowchart illustrating the surface optimization method provided in this application. Figure 4 .like Figure 4 As shown, after optimizing the surface to be optimized based on adjustment commands to obtain the target surface, the method also includes:

[0177] Step 401. Call the simulation module to perform performance analysis on the target surface and obtain the actual physical properties of the target surface.

[0178] For example, during a design intermittent phase, or after the user confirms the adjustment instruction, the simulation module can be invoked to perform precise analysis and high-precision simulation verification of the target surface obtained from the current adjustment instruction.

[0179] The simulation module can be physical simulation software configured on the user terminal, such as computational fluid dynamics (CFD) software or finite element analysis (FEA) software mentioned in the previous example.

[0180] The simulation module will be called the day after tomorrow. The target surface will be input into the simulation module, allowing it to perform high-precision physical property simulations to obtain the actual physical properties. Optionally, the actual physical properties will be denoted as... .

[0181] Step 402. Construct training data based on the surface parameters of the target surface and its actual physical properties.

[0182] For example, taking a NURBS surface as the surface to be optimized, the target surface after optimization based on the adjustment instructions is also a NURBS surface. The surface parameters of the target surface may include, but are not limited to: control point matrix, B-spline basis function degree, node vector, and weights of each control point.

[0183] Training data is constructed using surface parameters and the actual physical properties obtained from simulation by the simulation module. This training data is used for online learning of the surrogate model. Specifically, the constructed training data consists of surface parameters and their corresponding actual physical properties.

[0184] Step 403. Train the proxy model based on the training data to update the parameters of the proxy model and obtain the updated proxy model.

[0185] For example, training data is input into a surrogate model, which is then fine-tuned and corrected online to reduce prediction errors. Specifically, surface parameters from the training data are input into the surrogate model to obtain predicted values ​​for the surfaces represented by these parameters. The error between the predicted values ​​and the actual physical properties in the training data is calculated, and the parameters of the surrogate model are fine-tuned based on this error to obtain an updated surrogate model.

[0186] The updated surrogate model can be used to generate correction suggestions for adjustment instructions. It can be understood that the updated surrogate model can more accurately predict the physical properties of the surface; if an adjustment instruction leads to worse physical properties of the surface, it can generate correction suggestions for the adjustment instruction.

[0187] The updated proxy model can also run a global optimization algorithm, such as gradient descent, to find the optimal solution in the current design space and then present the optimal solution to the user.

[0188] In the above embodiments, the surrogate model can be trained using the actual physical performance obtained from high-precision simulation, enabling online learning and fine-tuning of the surrogate model. This allows the updated surrogate model to make more accurate performance predictions. In the long-term design process, the surrogate model can gradually improve its predictive ability for complex performance indicators, reducing reliance on high-cost simulations. Furthermore, the updated surrogate model allows for re-performance prediction, yielding new prediction results. Based on these new predictions, corrective suggestions can be generated for the user's adjustment instructions, making adjustments to the surface to be optimized more accurate.

[0189] In addition, the above process can form a dual-loop collaborative optimization mechanism, with the inner loop responding to the user's adjustment instructions in real time through the agent model, and the outer loop updating the agent model through high-precision verification in the background.

[0190] Based on any of the foregoing embodiments, the proxy model can be a single lightweight model or a combination of multiple lightweight models.

[0191] In one example, the surrogate model is any one or a combination of the following models: Gaussian regression model, multinomial model, radial basis function network, neural network model, and Kriging model.

[0192] Among them, Gaussian Process Regression (GPR) is a probabilistic surrogate model based on Gaussian process theory. It describes the correlation between surface-related characteristics and surface physical performance through a kernel function, and can output performance prediction results while providing confidence intervals.

[0193] Among them, the polynomial model refers to using first-order, second-order, or higher-order polynomial functions to fit the relationship between the relevant characteristics of a surface and the physical properties of the surface. In essence, it captures the linear or simple nonlinear trends between variables.

[0194] Radial Basis Function Network (RBFN) is a feedforward network consisting of an input layer, a hidden layer, and an output layer. It uses local basis functions to fit the relationship between the surface's correlation characteristics and the surface's physical properties.

[0195] Among them, the neural network model refers to a multi-layer network that simulates the structure of neurons in the human brain. Through the input layer, hidden layer, and output layer, it learns the mapping relationship between complex surface-related characteristics and surface physical properties.

[0196] Among them, the Kriging model refers to a surrogate model based on spatial interpolation theory. It uses a variogram function to describe the approximate physical properties of adjacent surfaces corresponding to their related features, thereby achieving high-precision interpolation prediction.

[0197] The following section provides some possible model combination methods.

[0198] One possible combination is to use a surrogate model that includes a polynomial model and a kriging model. Specifically, a polynomial model is first used to fit the global trend of the physical properties of the NURBS surface, and then a kriging model is used to correct for local biases.

[0199] One possible combination is to use a surrogate model that includes a radial basis function network (RBFN) and a neural network model. Specifically, the RBFN model is used to capture local abrupt changes in the physical properties of the surface, while the neural network is used to fit the overall complex nonlinear relationship.

[0200] One possible combination is a surrogate model that includes a Gaussian regression model and a radial basis function network (RBFN). Specifically, the GPR model is used to provide predicted values ​​and confidence intervals for physical properties, while the RBFN model is used to improve local fitting accuracy.

[0201] The above optional combinations are merely examples. In practical applications, the above models can be combined into two or more models to obtain a proxy model, depending on actual business needs. Alternatively, one of the above models can be selected as the proxy model.

[0202] In the above examples, lightweight surrogate models are used instead of high-precision physical simulation software, reducing the time required for performance prediction and improving the efficiency of surface optimization. Furthermore, a multi-model combination strategy is employed to enhance the comprehensiveness of the surrogate model. By aggregating the advantages of different models, the deviation in performance prediction results caused by the bias of a single model is reduced, thereby enhancing the reliability of performance prediction.

[0203] In some embodiments, simulation design software can update the shape of a surface. However, such software only provides visual feedback on geometry and cannot visualize changes in physical properties. With the development of Virtual Reality (VR), Augmented Reality (AR), and digital prototyping technologies, the industry's demand for real-time, interactive design and optimization tools is increasingly urgent. Users expect to see real-time visual feedback on surface shape and key performance indicators (such as airflow and temperature field) while dragging surface control points, thereby achieving a true "what you see is what you get" surface optimization process.

[0204] Therefore, in one possible implementation, the performance prediction results are visualized in multiple forms.

[0205] In one example, the numerical values ​​of the performance prediction results are visualized.

[0206] For example, a text box is generated on the human-computer interaction visualization interface. The text box displays the numerical value of the performance prediction result in the form of numerical text. Optionally, in addition to the numerical text used to display the numerical value of the performance prediction result, the text box may also include string text used to display the type of physical performance of the performance prediction result.

[0207] For example, the text box includes the string text "The drag coefficient of the curved surface is:" and the numerical text "0.28".

[0208] In one example, the trend of performance prediction results is visualized as a curve. This trend is determined based on the original physical properties of the surface to be optimized and the performance prediction results.

[0209] For example, the trend of performance prediction results is determined by the original physical properties of the surface to be optimized and the performance prediction results corresponding to multiple adjustment commands. For instance, if the user makes a total of 20 adjustments to the surface to be optimized, the proxy model will calculate the corresponding performance prediction result in response to each adjustment command. Based on these 20 performance prediction results and the original physical properties, a curve showing the change in physical performance can be generated and displayed as a graph on the human-computer interaction visualization interface.

[0210] In one example, the gradient information of the performance prediction result is visualized as arrows. The gradient information is calculated by the surrogate model based on the displacement change vector of the control point indicated by the adjustment command and the performance prediction result; the direction of the arrow is the same as or opposite to the direction of the gradient information; the direction of the gradient information represents the displacement direction of the control point that causes the fastest increase in the physical performance of the surface to be optimized.

[0211] For example, gradient information is obtained by calculating the partial derivatives of the displacement change vector with respect to the performance prediction results. Specifically, the partial derivatives of the performance prediction results with respect to each dimension of the displacement change vector are calculated to obtain the partial derivatives for each dimension, thus forming the gradient information.

[0212] Correspondingly, the gradient information has a direction, which represents the displacement direction that increases the physical properties of the surface to be optimized the fastest.

[0213] Specifically, in the process of visualizing the gradient information of performance prediction results in the form of arrows, the method may also include:

[0214] If the optimization objective of the surface to be optimized is to maximize physical performance, then the direction of the arrow is determined to be the same as the direction of the gradient information; if the optimization objective of the surface to be optimized is to minimize physical performance, then the direction of the arrow is determined to be opposite to the direction of the gradient information.

[0215] In practical engineering applications, the optimization objectives may differ depending on the physical property. For example, for the maximum stress of a curved surface, the physical property needs to be minimized; for the stiffness of a curved surface, the physical property needs to be maximized.

[0216] Therefore, if the optimization objective of the surface to be optimized is to maximize physical performance, then the direction of the arrow is determined to be the same as the direction of the gradient information. It can be understood that the direction of the arrow at this point indicates the displacement direction of the control point corresponding to maximizing physical performance. This is used to instruct the user to adjust the surface in this direction to optimize the surface and thus improve its physical performance.

[0217] Furthermore, if the optimization objective of the surface to be optimized is to minimize physical performance, then the direction of the arrow is determined to be opposite to the direction of the gradient information. This can be understood as the arrow direction indicating the displacement direction of the control point corresponding to minimizing physical performance. It instructs the user to adjust the surface in this direction to optimize the surface and thus improve physical performance.

[0218] Furthermore, the arrow with the determined direction is displayed on the visual interface. Optionally, the position of the arrow's tail can be the position of the control point indicated by the adjustment command.

[0219] In the above embodiments, the target surface and its corresponding performance prediction results can be visualized. Furthermore, different visualization methods can be used to display the performance prediction results. This enables the visualization of physical performance during surface optimization, improving the user experience and making the surface optimization process more intelligent.

[0220] The surface optimization method provided in this application involves the user adjusting the surface to be optimized. Responding to the adjustment command, a proxy model is used to predict the performance of the surface after adjustment, yielding a performance prediction result. This avoids the need for high-precision simulation to predict the physical performance of the adjusted surface in traditional simulation design processes, significantly reducing computational complexity and shortening the calculation time for physical performance. The method involves adjusting the surface to be optimized according to the adjustment command and rendering the target surface. The adjusted target surface and the corresponding performance prediction result can be visualized on a preset interface, avoiding the problem in traditional simulation design processes where the adjustment result and performance prediction result cannot be simultaneously perceived by the user. This allows designers to obtain millisecond-level performance feedback during interactive design. Based on these two aspects, smooth interaction is ensured, and the efficiency of surface optimization processing is improved.

[0221] There's no need to wait for a complete reconstruction of the entire surface before performing high-precision physical performance simulations, thus shortening the surface optimization processing time. By extracting only the geometric features of the affected region as input to the surrogate model, global feature redundancy is avoided, allowing the prediction model to focus more on key areas and thereby improving the accuracy of performance prediction results.

[0222] Instead of directly predicting the overall performance of the surface, this approach reduces computational load and processing time. Instead, it infers the performance changes in the affected surface regions, improving the accuracy of performance prediction results.

[0223] By training the surrogate model online, the updated model can make more accurate performance predictions. In the long-term design process, the surrogate model can gradually improve its predictive ability for complex performance metrics, reducing reliance on costly simulations.

[0224] The target surface and its corresponding performance prediction results are visualized, enabling a visual representation of the physical performance during the surface optimization process, thus improving the user experience and making the surface optimization process more intelligent.

[0225] Figure 5 A schematic diagram of the structure of the surface optimization processing device provided in this application is shown below. Figure 5 As shown, the surface optimization processing device 50 provided in this embodiment includes:

[0226] The processing module 501 is used to respond to the adjustment command for the surface to be optimized, and to perform performance prediction based on the proxy model according to the adjustment command to obtain the performance prediction result; wherein, the performance prediction result is the physical performance of the surface to be optimized after optimization based on the adjustment command.

[0227] Processing module 501 is also used to optimize and adjust the surface to be optimized based on adjustment instructions to obtain the target surface;

[0228] The display module 502 is used to visualize the target surface and performance prediction results.

[0229] In one possible implementation, the adjustment instructions include: the matrix index of the control points on the surface to be optimized, and the displacement change vector of the control points;

[0230] Based on the adjustment instructions, performance prediction is performed using the proxy model to obtain the performance prediction results. Processing module 501 is used for:

[0231] Based on the matrix index, the first region on the surface to be optimized is determined; where the first region is the surface region on the surface to be optimized that is affected by the adjustment command;

[0232] For the first region, the geometric feature vector of the first region is determined based on the displacement change vector;

[0233] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction results.

[0234] In one possible implementation, the surface to be optimized has the original physical properties;

[0235] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction results. Processing module 501 is used for:

[0236] Based on the displacement change vector and geometric feature vector, performance prediction is performed using a surrogate model to obtain the physical performance change; whereby the physical performance change represents the change in physical performance caused by the optimization adjustment of the surface to be optimized based on the displacement change vector.

[0237] The performance prediction results are determined based on the original physical properties and the changes in physical properties.

[0238] In one possible implementation, the first region includes at least one parameter point;

[0239] The surface to be optimized is optimized and adjusted based on adjustment commands to obtain the target surface. Processing module 501 is used for:

[0240] For the parameter points in the first region, determine the vertex coordinates of the parameter points based on the displacement change vector;

[0241] Based on the vertex coordinates, determine at least one geometric face in the first region;

[0242] At least one geometric surface in the first region is rendered, and the target surface is constructed based on the rendering result and the second region; wherein, the second region is the surface region in the surface to be optimized other than the first region.

[0243] In one possible implementation, after the surface to be optimized is optimized based on the adjustment command to obtain the target surface, the processing module 501 is further configured to:

[0244] The simulation module is invoked to perform performance analysis on the target surface and obtain the actual physical properties of the target surface.

[0245] Training data is constructed based on the surface parameters of the target surface and its actual physical properties.

[0246] The proxy model is trained based on the training data to update its parameters, resulting in an updated proxy model.

[0247] In one possible implementation, the surrogate model is any one or a combination of the following models: Gaussian regression model, multinomial model, radial basis function network, neural network model, and Kriging model.

[0248] In one possible implementation, the performance prediction results are visualized, and the visualization module 502 is configured to perform at least one of the following:

[0249] Visualize the numerical results of performance predictions;

[0250] The trend of performance prediction results is visualized in the form of a curve; the trend is determined based on the original physical properties and performance prediction results of the surface to be optimized.

[0251] The gradient information of the performance prediction results is visualized in the form of arrows; where the gradient information is calculated by the surrogate model based on the displacement change vector of the control point indicated by the adjustment command and the performance prediction results; the direction of the arrow is the same as or opposite to the direction of the gradient information; the direction of the gradient information represents the displacement direction of the control point that increases the physical performance of the surface to be optimized the fastest.

[0252] In one possible implementation, the display module 502 is further configured to:

[0253] If the optimization objective of the surface to be optimized is to maximize physical performance, then the direction of the arrow is determined to be the same as the direction of the gradient information.

[0254] If the optimization objective of the surface to be optimized is to minimize physical performance, then the direction of the arrow is determined to be opposite to the direction of the gradient information.

[0255] The surface optimization processing device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0256] Figure 6 A schematic diagram of the structure of the electronic device provided in this application. Figure 6 As shown, the electronic device 60 provided in this embodiment includes at least one processor 601 and a memory 602. Optionally, the electronic device 60 further includes a communication component 603. The processor 601, memory 602, and communication component 603 are connected via a bus 604.

[0257] In a specific implementation, at least one processor 601 executes computer execution instructions stored in memory 602, causing at least one processor 601 to perform the above-described method.

[0258] The specific implementation process of processor 601 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0259] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0260] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0261] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0262] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0263] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0264] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0265] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0266] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0267] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0268] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0269] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0270] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0271] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for optimizing curved surfaces, characterized in that, include: In response to an adjustment command for the surface to be optimized, a performance prediction is performed based on a proxy model according to the adjustment command to obtain a performance prediction result; wherein, the performance prediction result is the physical performance of the surface to be optimized after optimization based on the adjustment command. The surface to be optimized is optimized and adjusted based on the adjustment instructions to obtain the target surface; The target surface and the performance prediction results are visualized.

2. The method according to claim 1, characterized in that, The adjustment instructions include: the matrix index of the control points on the surface to be optimized, and the displacement change vector of the control points; Based on the adjustment instructions, performance prediction is performed using the proxy model to obtain performance prediction results, including: Based on the matrix index, a first region on the surface to be optimized is determined; wherein, the first region is the surface region on the surface to be optimized that is affected by the adjustment command; For the first region, the geometric feature vector of the first region is determined based on the displacement change vector; Based on the displacement change vector and the geometric feature vector, performance prediction is performed using a surrogate model to obtain the performance prediction result.

3. The method according to claim 2, characterized in that, The surface to be optimized has its original physical properties; Based on the displacement change vector and the geometric feature vector, performance prediction is performed using a surrogate model to obtain performance prediction results, including: Based on the displacement change vector and the geometric feature vector, performance prediction is performed using a surrogate model to obtain the physical performance change amount; wherein, the physical performance change amount characterizes the change in physical performance caused by the optimization adjustment of the surface to be optimized based on the displacement change vector; The performance prediction result is determined based on the original physical properties and the change in physical properties.

4. The method according to claim 2, characterized in that, The first region includes at least one parameter point; The surface to be optimized is optimized and adjusted based on the adjustment command to obtain the target surface, including: For the parameter points in the first region, the vertex coordinates of the parameter points are determined according to the displacement change vector; Based on the vertex coordinates, at least one geometric face in the first region is determined; At least one geometric surface in the first region is rendered, and the target surface is constructed based on the rendering result and the second region; wherein, the second region is the surface region other than the first region in the surface to be optimized.

5. The method according to claim 1, characterized in that, After optimizing the surface to be optimized based on the adjustment instructions to obtain the target surface, the method further includes: The simulation module is invoked to perform performance analysis on the target surface, thereby obtaining the actual physical properties of the target surface. Training data is constructed based on the surface parameters of the target surface and the actual physical properties. The proxy model is trained based on the training data to update its parameters, resulting in an updated proxy model.

6. The method according to any one of claims 1-5, characterized in that, The surrogate model is any one or a combination of the following models: Gaussian regression model, multinomial model, radial basis function network, neural network model, and Kriging model.

7. The method according to any one of claims 1-5, characterized in that, The performance prediction results are visualized, including at least one of the following: The numerical values ​​of the performance prediction results are then visualized. The trend of the performance prediction results is visualized in the form of a curve; wherein the trend is determined based on the original physical properties of the surface to be optimized and the performance prediction results. The gradient information of the performance prediction result is visualized in the form of arrows; wherein, the gradient information is calculated by the surrogate model based on the displacement change vector of the control point indicated by the adjustment command and the performance prediction result; the direction of the arrow is the same as or opposite to the direction of the gradient information; the direction of the gradient information represents the displacement direction of the control point that increases the physical performance of the surface to be optimized the fastest.

8. The method according to claim 7, characterized in that, The method further includes: If the optimization objective of the surface to be optimized is to maximize physical performance, then the direction of the arrow is determined to be the same as the direction of the gradient information; If the optimization objective of the surface to be optimized is to minimize physical performance, then the direction of the arrow is determined to be opposite to the direction of the gradient information.

9. A surface optimization processing device, characterized in that, include: The processing module is configured to respond to an adjustment command for the surface to be optimized, and perform performance prediction based on a proxy model according to the adjustment command to obtain a performance prediction result; wherein the performance prediction result is the physical performance of the surface to be optimized after optimization based on the adjustment command. The processing module is also used to optimize and adjust the surface to be optimized based on the adjustment instruction to obtain the target surface; The display module is used to visualize the target surface and the performance prediction results.

10. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-8.

12. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1-8.

Images