Reed design method and device, computer equipment and storage medium

Abstract

The invention relates to a reed design method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining a structure parameter condition and a constraint condition of a reed; performing batch simulation based on the structure parameter condition to obtain multiple groups of simulation data of the reed; determining associated parameters in the structure parameters corresponding to the structure parameter data according to the multiple groups of simulation data; according to the correlation parameters, the multiple sets of simulation data and a preset model, a performance prediction model is obtained through fitting; and determining at least one group of target parameter data of the reed based on an optimization algorithm and the performance prediction model according to the structure parameter condition and the constraint condition. And through the optimization algorithm and the performance prediction model, automatic design of the reed is carried out, so that the design time of the reed is shortened, and the design cost of the reed is further reduced.

Description

Technical Field

[0001] This application relates to the field of structural design technology, and in particular to a spring design method, apparatus, computer equipment, and storage medium. Background Technology

[0002] With the development of camera modules, the voice coil motor (VCM) has become a crucial component. A VCM is a drive device that operates based on the Lorentz force principle of a energized coil operating within a permanent magnetic field created by a magnet. The VCM incorporates a reed that supports the camera module's lens and provides precise torque balance and restoring force during lens movement. For example, by supporting the lens with the reed, high-precision linear movement along the optical axis (Z-axis) is allowed, along with micro-motion in the X, Y, and Z directions. High-precision linear movement along the Z-axis enables autofocus, while micro-motion in the X, Y, and Z directions enables optical image stabilization. Furthermore, when the drive current to the VCM is disconnected, the reed's elasticity allows the lens to return to its initial equilibrium position. Therefore, the reed's structural design is particularly important for the performance of the camera module.

[0003] The structural design of a reed has a decisive impact on its mechanical properties and reliability. Currently, reed design typically requires experienced engineers to design, simulate, verify, and optimize the reed. Therefore, the current reed design methods require a significant amount of research and development time and are excessively costly in terms of manpower. Summary of the Invention

[0004] Therefore, it is necessary to provide a reed design method, apparatus, computer equipment, and storage medium to address the aforementioned technical problems.

[0005] In a first aspect, this application provides a reed design method, the method comprising: obtaining structural parameter conditions and constraint conditions of the reed; performing batch simulations based on the structural parameter conditions to obtain multiple sets of simulation data of the reed; each set of simulation data comprising: structural parameter data of the reed and corresponding performance parameter data; determining related parameters among the structural parameters corresponding to the structural parameter data based on the multiple sets of simulation data; fitting a performance prediction model based on the related parameters, the multiple sets of simulation data and a preset model; and determining at least one set of target parameter data of the reed based on the structural parameter conditions and the constraint conditions, using an optimization algorithm and the performance prediction model.

[0006] In one embodiment, obtaining the structural parameter conditions of the reed includes: constructing an initial simulation model of the reed; determining the initial structural parameter data of each structural parameter in the reed based on the initial simulation model; obtaining the floating range corresponding to each structural parameter; and determining the structural parameter conditions according to the initial structural parameter data of each structural parameter and the floating range corresponding to each structural parameter.

[0007] In one embodiment, the step of performing batch simulation based on structural parameter conditions to obtain multiple sets of simulation data for the reed includes: sampling the structural parameter conditions using a preset sampling algorithm to obtain multiple sets of structural parameter data; performing simulation based on the multiple sets of structural parameter data and an initial simulation model to obtain performance parameter data corresponding to each set of structural parameter data; and obtaining multiple sets of simulation data based on each set of structural parameter data and the performance parameter data corresponding to each set of structural parameter data.

[0008] In one embodiment, the structural parameter data includes parameter data corresponding to multiple structural parameters; the performance parameter data includes parameter data corresponding to multiple performance parameters; determining the correlation parameter among the structural parameters based on multiple sets of simulation data includes: determining a parameter data set for each structural parameter and a parameter data set for each performance parameter based on multiple sets of simulation data; determining the correlation coefficient between each structural parameter and each performance parameter based on a correlation analysis algorithm, according to the parameter data set for each structural parameter and the parameter data set for each performance parameter; and using the structural parameter corresponding to the correlation coefficient that meets preset conditions as the correlation parameter.

[0009] In one embodiment, fitting a performance prediction model based on the correlation parameters, multiple sets of simulation data, and a preset model includes: determining multiple sets of fitting data based on the correlation parameters and multiple sets of simulation data; each set of fitting data includes: correlation parameter data of the correlation parameters and corresponding performance parameter data; constructing an initial prediction model based on the multiple sets of fitting data and multiple preset models; and iterating the initial prediction model based on a genetic algorithm and the multiple sets of fitting data to obtain a performance prediction model.

[0010] In one embodiment, the multiple preset models include: a Kriging fitting model, a Gaussian process regression fitting model, a radial basis function fitting model, an inverse distance weighted fitting model, and an energy minimization spline interpolation fitting model; the step of constructing an initial prediction model based on multiple sets of fitting data and multiple preset models includes: fitting a first function using a Kriging fitting model based on multiple sets of fitting data; fitting a second function using a Gaussian process regression fitting model based on multiple sets of fitting data; fitting a third function using a radial basis function fitting model based on multiple sets of fitting data; fitting a fourth function using an inverse distance weighted fitting model based on multiple sets of fitting data; fitting a fifth function using an energy minimization spline interpolation fitting model based on multiple sets of fitting data; and setting weights for the first, second, third, fourth, and fifth functions respectively to construct the initial prediction model.

[0011] In one embodiment, the step of iterating the initial prediction model based on the genetic algorithm and multiple sets of fitted data to obtain a performance prediction model includes: determining prediction performance data based on the correlation parameter data in the multiple sets of fitted data and the initial prediction model; and optimizing the weights of the first function, the second function, the third function, the fourth function, and the fifth function in the initial prediction model based on the prediction performance data, the performance parameter data in the multiple sets of fitted data, and the genetic algorithm to obtain the performance prediction model.

[0012] In one embodiment, determining at least one set of target parameter data for the reed based on the structural parameter conditions and the constraints, using an optimization algorithm and a performance prediction model, includes: generating an initial population based on the structural parameter conditions and an optimization algorithm; the initial population includes multiple sets of associated structural parameter data corresponding to the associated parameters; determining associated performance parameter data corresponding to each set of associated structural parameter data based on the initial population and the performance prediction model; and iterating the population based on the initial population, the associated performance parameter data corresponding to each set of associated structural parameter data, and the constraints, using the optimization algorithm and the performance prediction model, until a preset termination condition is met, thereby determining at least one set of target parameter data for the reed.

[0013] Secondly, this application also provides a reed design apparatus, the apparatus comprising: an acquisition module for acquiring structural parameter conditions and constraint conditions of the reed; a simulation module for performing batch simulations based on the structural parameter conditions to obtain multiple sets of simulation data of the reed; each set of simulation data includes: structural parameter data of the reed and corresponding performance parameter data; a parameter determination module for determining related parameters from the structural parameters corresponding to the structural parameter data based on the multiple sets of simulation data; a fitting module for fitting a performance prediction model based on the related parameters, the multiple sets of simulation data, and multiple preset models; and an optimization module for determining at least one set of target parameter data of the reed based on the structural parameter conditions and the constraint conditions, using an optimization algorithm and the performance prediction model.

[0014] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any of the reed design methods described in the first aspect.

[0015] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements any of the reed design methods described in the first aspect above.

[0016] The aforementioned reed design method, apparatus, computer equipment, and storage medium acquire the structural parameters and constraints of the reed. Based on these structural parameters, batch simulations are performed to obtain multiple sets of simulation data for the reed. Based on these simulation data, relevant parameters are determined from the structural parameters corresponding to the structural parameter data. Based on the relevant parameters, the multiple sets of simulation data, and various preset models, a performance prediction model is fitted. Based on the structural parameters and constraints, and using an optimization algorithm and the performance prediction model, at least one set of target parameter data for the reed is determined. Through the optimization algorithm and performance prediction model, the reed is designed automatically, thereby reducing the design time and further reducing the design cost. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a reed design method in one embodiment;

[0018] Figure 2 This is a flowchart illustrating a method for obtaining structural parameter conditions in one embodiment;

[0019] Figure 3 This is a flowchart illustrating an automated simulation method in one embodiment;

[0020] Figure 4This is a flowchart illustrating a method for determining association parameters in one embodiment;

[0021] Figure 5 This is a flowchart illustrating a method for fitting a performance prediction model in one embodiment;

[0022] Figure 6 This is a flowchart illustrating a method for determining target parameter data in one embodiment;

[0023] Figure 7 This is a flowchart illustrating an automatic reed optimization design method in one embodiment;

[0024] Figure 8 This is a structural block diagram of a reed design device in one embodiment;

[0025] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] With the development of camera modules, the voice coil motor (VCM) has become a crucial component. A VCM is a drive device that operates based on the Lorentz force principle of a energized coil operating within a permanent magnetic field created by a magnet. The VCM incorporates a reed that supports the camera module's lens and provides precise torque balance and restoring force during lens movement. For example, by supporting the lens with the reed, high-precision linear movement along the optical axis (Z-axis) is allowed, along with micro-motion in the X, Y, and Z directions. High-precision linear movement along the Z-axis enables autofocus, while micro-motion in the X, Y, and Z directions enables optical image stabilization. Furthermore, when the drive current to the VCM is disconnected, the reed's elasticity allows the lens to return to its initial equilibrium position. Therefore, the reed's structural design is particularly important for the performance of the camera module.

[0028] The structural design of the reed has a decisive impact on its mechanical performance and reliability. Currently, the reed structure design of voice coil motors (VCMs) primarily relies on engineers' experience and manual trial-and-error methods. Specifically, the reed is first manually modeled; for example, engineers use 3D CAD software to manually draw the geometric model of the reed based on experience. Next, the reed model is manually simulated. Engineers import the reed model into finite element analysis software, manually set material properties, mesh generation, boundary conditions, and perform the solution. Finally, engineers interpret the results, analyzing whether they meet the design requirements. If not, engineers, based on their experience, manually modify one or more geometric parameters in the CAD model, resimulate, and analyze the results until the requirements are met. Therefore, reed design typically requires experienced engineers to design, simulate, verify, and optimize the reed.

[0029] Currently, the design of reeds in related technologies relies entirely on manual operation by engineers, based on a cycle of modeling, simulation, and parameter tuning. Each iteration takes several hours, resulting in a long design cycle and low R&D efficiency. Furthermore, this process heavily depends on the experience of senior engineers, demanding high skill levels and consuming significant human resources. Because reeds have numerous structural parameters, manual parameter tuning typically only allows for local optimization within a limited set of options, making it difficult to explore the entire design space. The final design may only be a usable solution, not the optimal one. Simultaneously, during the reed design process, engineers struggle to accurately identify which structural parameter has the greatest impact on design requirements, leading to unclear tuning directions and increased trial-and-error attempts.

[0030] In one embodiment, such as Figure 1 As shown, a reed design method is provided, including the following steps:

[0031] Step 101: Obtain the structural parameters and constraints of the reed.

[0032] The reed is a spring in the voice coil motor (VCM) used to support the lens of the camera module. This camera module can be used in mobile phones, vehicles, and surveillance equipment; this embodiment is not specifically limited, as long as the camera module includes this reed. The structural parameter conditions refer to the required conditions for all structural parameters corresponding to this reed design, i.e., the value range of all structural parameters. The constraint conditions are the required conditions that this reed design needs to meet, i.e., the restrictive requirements for all performance parameters of the reed. The structural parameter conditions and constraint conditions can be set according to actual usage requirements.

[0033] Structural parameter conditions can be a set of input conditions describing the geometry and dimensional range of a reed, used to limit the adjustable design variables and their value ranges during simulation and optimization. Specifically, the reed can be divided into multiple spring segments, including straight spring segments and curved spring segments. The structural parameters corresponding to straight spring segments are length and width, etc.; the structural parameter corresponding to curved spring segments is the radius of curvature (R-angle). The structural parameter conditions include the value range of the structural parameter corresponding to each spring segment, for example, the length and width value range for a straight spring segment, and the R-angle radius value range for a curved spring segment. Taking a reed consisting of two straight spring segments and three curved spring segments as an example, the structural parameter conditions include: the length and width value ranges of the first straight spring segment, the length and width value ranges of the second straight spring segment, the R-angle radius value ranges of the first curved spring segment, the R-angle radius value ranges of the second curved spring segment, and the R-angle radius value ranges of the third curved spring segment.

[0034] Constraints can be restrictive requirements on the performance parameters of the reed, used to determine whether the reed design meets practical application standards. Performance parameters can include the maximum stress of the reed under a given preset displacement, and the stiffness coefficients in the X, Y, and Z directions. Constraints include: restrictions on the maximum stress, stiffness coefficients in the X, Y, and Z directions of the reed under a given preset displacement.

[0035] Step 102: Perform batch simulations based on structural parameter conditions to obtain multiple sets of simulation data for the reed.

[0036] After obtaining the structural parameter conditions, multiple sets of structural parameter data are first generated based on these conditions. The structural parameters in each set of data are identical to those corresponding to the structural parameter conditions. For example, the structural parameter conditions include: the length and width ranges of the first linear spring segment, the length and width ranges of the second linear spring segment, the radius (R) of the first curved spring segment, the radius (R) of the second curved spring segment, and the radius (R) of the third curved spring segment. Therefore, each set of structural parameter data generated based on these conditions includes the values ​​of the length and width of the first linear spring segment, the length and width of the second linear spring segment, and the radius (R) of the first, second, and third curved spring segments. Then, batch simulations are performed based on the generated sets of structural parameter data to determine the performance parameter data corresponding to each set. The structural parameter data and the corresponding performance parameter data constitute a set of simulation data; that is, each set of simulation data includes the structural parameter data of the spring and the corresponding performance parameter data. Among them, the performance parameter data represents the numerical values ​​of the performance parameters of the reed formed by the set of structural parameter data. The performance parameters include the maximum stress of the reed under a given preset motion displacement condition, the stiffness coefficient in the X direction, the stiffness coefficient in the Y direction, and the stiffness coefficient in the Z direction.

[0037] Step 103: Based on multiple sets of simulation data, determine the associated parameters among the structural parameters corresponding to the structural parameter data.

[0038] Each set of simulation data includes structural parameter data and corresponding performance parameter data for the reed. Correlation analysis is performed on the structural parameter data and performance parameter data to determine the associated parameters among the structural parameters corresponding to the structural parameter data. Associated parameters represent structural parameters that have a significant impact on performance parameters and can be obtained through sensitivity analysis or statistical analysis of multiple sets of simulation data. For example, correlation analysis, sensitivity analysis, or feature selection algorithms are used to extract associated parameters from multiple sets of simulation data. Specifically, statistical analysis is performed on multiple sets of simulation data to identify structural parameters that contribute significantly to changes in performance parameters and these are used as associated parameters. Statistical analysis can use algorithms such as Pearson correlation coefficient and random forest, etc., and this embodiment does not impose specific limitations.

[0039] Step 104: Based on the correlation parameters, multiple sets of simulation data, and the preset model, a performance prediction model is fitted to obtain the model.

[0040] After determining the correlation parameters, multiple sets of fitting data are obtained by filtering from multiple sets of simulation data based on the correlation parameters. Specifically, each set of simulation data includes the structural parameter data of the reed and the corresponding performance parameter data. The correlation parameter data corresponding to the correlation parameters is filtered from the structural parameter data, and the correlation parameter data and the corresponding performance parameter data are used as a set of fitting data. Each set of simulation data is filtered based on the correlation parameters to obtain multiple sets of fitting data. Then, the multiple sets of fitting data are fitted with various preset models to obtain a performance prediction model. The preset model is a pre-selected type of mathematical function or machine learning architecture used to fit the relationship between the correlation parameters and the performance parameters. For example, the preset model can be at least one of the following: Kriging fitting model, Gaussian process regression fitting model, radial basis function fitting model, inverse distance weighted fitting model, and energy minimization spline interpolation fitting model. Specifically, if there is only one preset model, the preset model is fitted based on multiple sets of fitting data to obtain the performance prediction model; if there are multiple preset models, weights are assigned to each preset model, all preset models are superimposed, and the model is fitted based on multiple sets of fitting data to obtain the performance prediction model. The performance prediction model is used to represent the relationship between correlation parameters and performance parameters. In actual use, by inputting the specific values ​​corresponding to the correlation parameters, the performance parameters of the reed corresponding to that set of correlation parameters can be obtained.

[0041] Step 105: Based on the structural parameter conditions and constraints, and using the optimization algorithm and performance prediction model, determine at least one set of target parameter data for the reed.

[0042] An optimization algorithm is a computational method used to find the optimal solution that satisfies constraints within a given search space. Optimization algorithms can be global search strategies such as genetic algorithms and particle swarm optimization, as well as local search methods such as gradient descent. For example, an optimization algorithm can be a multi-objective optimization genetic algorithm. The optimization algorithm searches under structural parameter conditions through iterative operations such as population initialization, selection, crossover, mutation, and fitness evaluation. During the search process, it receives structural parameter conditions and constraints as the definition of the search space, calls a performance prediction model to evaluate the performance parameter data corresponding to each set of associated structural parameter data in the population, and drives population iteration. Population iteration is a repetitive computational process in genetic algorithms that gradually approaches the optimal solution set through intergenerational evolution, and can be used to achieve asymptotic convergence from an initial random solution to a high-quality candidate solution. Population iteration generates a new population through selection, recombination, and mutation of individuals in each generation, and evaluates its performance using a performance prediction model. The target parameter data is the combination of structural parameters that satisfies all constraints and has the optimal performance parameter index, output by the optimization algorithm.

[0043] Specifically, within the structural parameter conditions, an initial population is generated based on an optimization algorithm. Each population includes multiple sets of associated structural parameter data corresponding to the reed's associated parameters. For example, these multiple sets of associated structural parameter data may include 100 sets or 200 sets; this embodiment does not impose a specific limitation and can be set according to actual usage requirements. For example, the structural parameter conditions include all associated parameter conditions. An initial population is generated based on these conditions. For the initial population, each set of associated structural parameter data is input into a performance prediction model to obtain the associated performance parameter data corresponding to each set of associated structural parameter data. The associated performance parameter data corresponding to each set of associated structural parameter data in the population is fed back to the optimization algorithm, allowing the algorithm to evaluate individual performance using the associated performance parameter data corresponding to each set of associated structural parameter data in the population, perform genetic operations to generate offspring, and iterate continuously until a preset termination condition is reached to obtain the target parameter data. The target parameter can be a set of numerical values ​​for the associated parameter combinations of the reed that satisfy all design requirements.

[0044] This embodiment obtains the structural parameters and constraints of the reed, providing clear design boundaries and evaluation criteria for subsequent automated processes, avoiding invalid searches, and thus providing search boundaries for design variables for batch simulation, ensuring that the generated model is within the physically realizable range. Batch simulation based on structural parameter conditions yields multiple sets of simulation data for the reed, acquiring a large number of simulation samples at once, replacing manual modeling and simulation step-by-step, and improving data acquisition efficiency. By identifying associated parameters from the structural parameters corresponding to the structural parameter data, the dimensionality of the optimization problem is reduced, resources are concentrated on optimizing key parameters, and redundant calculations are reduced, thereby narrowing the dimensionality of the optimization problem and improving the efficiency of subsequent modeling and optimization. A performance prediction model is fitted based on associated parameters, multiple sets of simulation data, and various preset models, avoiding repeated use of high-cost simulation tools, replacing time-consuming finite element simulation with the performance prediction model, and accelerating the design evaluation process. Based on the structural parameter conditions and constraints, and using optimization algorithms and performance prediction models, at least one set of target parameter data for the reed is determined, achieving efficient global search, overcoming the limitations of manual experience, and automatically searching for the parameter combination that satisfies the constraints and has the best performance under the structural parameter conditions, serving as the final adoptable reed design parameters to guide actual modeling and manufacturing. The entire process forms a closed-loop automated mechanism from parameter input, data generation, model building to solution output, which shortens the design cycle of the reed and reduces the continuous reliance on highly skilled engineers, thereby effectively reducing the design time and R&D cost of the reed.

[0045] In one embodiment, such as Figure 2 As shown, a method for obtaining structural parameter conditions is provided, which specifically includes the following steps:

[0046] Step 201: Construct the initial simulation model of the reed.

[0047] The initial simulation model is a model constructed based on the physical properties of the reed, including its 3D structure, material type, and thickness. For example, the initial simulation model can be constructed by an engineer using modeling software; alternatively, it can be obtained from historical reed designs, and a reed model can be selected as the initial simulation model. This embodiment does not impose specific limitations; only the initial simulation model of the reed needs to be constructed. The material type refers to the material of the reed, such as titanium copper, beryllium copper, or stainless steel. The thickness represents the thickness of the reed, which can be set to, for example, between 0.03 mm and 0.05 mm. The above material type and thickness are merely illustrative examples; specific settings can be made according to actual usage requirements when constructing the initial simulation model. The modeling software is CAD software. For example, the modeling software can be SolidWorks, CATIA, Pro / E (Creo), AutoCAD, UG (Siemens NX), etc. This embodiment does not impose specific limitations on the modeling software; only CAD software that supports API calls is required.

[0048] Step 202: Based on the initial simulation model, determine the initial structural parameter data for each structural parameter in the reed.

[0049] After constructing the initial simulation model, the initial structural parameter data for each structural parameter of the corresponding reed is automatically measured in the modeling software. Taking a reed consisting of two straight spring segments and three curved spring segments as an example, the structural parameters include: the length and width of the first straight spring segment, the length and width of the second straight spring segment, the radius of curvature of the first curved spring segment, the radius of curvature of the second curved spring segment, and the radius of curvature of the third curved spring segment. The initial structural parameter data are the specific values ​​of the above structural parameters corresponding to the initial simulation model.

[0050] Step 203: Obtain the floating range corresponding to each structural parameter.

[0051] The floating range is a numerical variation interval set for the initial structural parameter data. It represents the maximum and minimum boundaries that a certain structural parameter is allowed to be adjusted. The floating range is usually set in the form of a percentage. For example, the floating range can be ±5%, indicating that the structural parameter is allowed to fluctuate within ±5% based on the initial structural parameter data. Taking a reed consisting of two straight spring segments and three curved spring segments as an example, the floating range includes: the floating range of the length of the first straight spring segment, the floating range of the width of the first straight spring segment, the floating range of the length of the second straight spring segment, the floating range of the width of the second straight spring segment, the floating range of the radius of the rounded corner of the first curved spring segment, the floating range of the radius of the rounded corner of the second curved spring segment, and the floating range of the radius of the rounded corner of the third curved spring segment. The floating range can be set according to the usage requirements of the reed. This embodiment does not specifically limit the specific value of the floating range.

[0052] Step 204: Determine the structural parameter conditions based on the initial structural parameter data for each structural parameter and the corresponding floating range for each structural parameter.

[0053] After obtaining the initial structural parameter data for each structural parameter and the corresponding floating range for each structural parameter, the structural parameter conditions for the corresponding structural parameter can be determined. For example, the length range of the first linear spring segment is determined based on its length and its floating range; the width range of the first linear spring segment is determined based on its length and its floating range; the length range of the second linear spring segment is determined based on its length and its floating range; the width range of the second linear spring segment is determined based on its length and its floating range; the radius of curvature of the first curved spring segment is determined based on its radius and its floating range; the radius of curvature of the second curved spring segment is determined based on its radius and its floating range; and the radius of curvature of the third curved spring segment is determined based on its radius and its floating range.

[0054] This embodiment transforms the manual setting process, which originally relied on engineers' subjective judgment, into an automated and structured process, reducing human intervention and trial-and-error costs. Simultaneously, the standardized output of structural parameters improves the efficiency of subsequent batch simulations and optimization, thereby shortening the overall design cycle and reducing the continuous reliance on highly skilled personnel.

[0055] In one embodiment, such as Figure 3 As shown, an automated simulation method is provided, which specifically includes the following steps:

[0056] Step 301: Sample the structural parameter conditions using a preset sampling algorithm to obtain multiple sets of structural parameter data.

[0057] The preset sampling algorithm is a mathematical method for systematically selecting sample points within a given parameter space, enabling efficient coverage of the design space with fewer sampling iterations. The preset sampling algorithm can be one or more of the following: Latin hypercube sampling, orthogonal experimental design, Sobol sequence sampling, optimized Latin hypercube sampling, or adaptive sampling based on a surrogate model. This embodiment does not impose specific limitations, only requiring that sampling be performed within the structural parameter conditions to obtain multiple sets of structural parameter data. These multiple sets of structural parameter data are discrete parameter sets generated from the structural parameter conditions by the preset sampling algorithm, each set representing an input design scheme to be simulated. Specifically, within the design space defined by the structural parameter conditions, the preset sampling algorithm automatically generates multiple representative combinations of structural parameters. For example, Latin hypercube sampling generates multiple evenly distributed sampling points, each set of sampling points corresponding to a set of structural parameter data. Taking a reed consisting of two straight spring segments and three curved spring segments as an example, the structural parameter conditions include: the length and width ranges of the first straight spring segment, the length and width ranges of the second straight spring segment, the radius of the first curved spring segment (R-angle), the radius of the second curved spring segment (R-angle), and the radius of the third curved spring segment (R-angle). Then, a set of structural parameter data includes: the length and width of the first straight spring segment, the length and width of the second straight spring segment, the radius of the first curved spring segment (R-angle), the radius of the second curved spring segment (R-angle), and the corresponding values ​​of the radius of the third curved spring segment (R-angle).

[0058] Step 302: Perform simulation based on multiple sets of structural parameter data and the initial simulation model to obtain the performance parameter data corresponding to each set of structural parameter data.

[0059] Combining finite element analysis (FEM) software and modeling software, the FEM software automatically calls the initial simulation model in the modeling software to simulate each set of generated structural parameter data, thereby obtaining the performance parameter data corresponding to each set of structural parameter data. The FEM software can be ANSYS Mechanical, Abaqus and its Python scripting interface, COMSOL Multiphysics, Nastran, etc., but this embodiment is not specifically limited, as long as it can perform simulations. The FEM software and modeling software are connected through bridging tools, such as Abaqus Python Scripting and Pyansys, etc., but this embodiment is not specifically limited. Specifically, each set of structural parameter data is imported into the initial simulation model, the structural parameters of the initial simulation model are updated, and the FEM software performs finite element simulation to solve for the performance parameter data corresponding to that set of structural parameter data. The performance parameter data represents the numerical values ​​of the performance parameters of the spring formed by that set of structural parameter data, including the maximum stress of the spring under a given preset displacement condition, the stiffness coefficient in the X direction, the stiffness coefficient in the Y direction, and the stiffness coefficient in the Z direction. Specifically, during the simulation, a forced displacement of 0.46 mm is applied to the reed in the Z direction. The performance parameters represent the maximum stress, stiffness coefficient in the X direction, stiffness coefficient in the Y direction, and stiffness coefficient in the Z direction under this condition. The stiffness coefficient is a fundamental physical quantity describing the elastic deformation of the reed under external force. Understandably, to improve simulation efficiency, scripts can be used to batch call the modeling software interface to update model parameters, and parametric modeling software and finite element analysis software can be used to automatically perform the updates and solutions.

[0060] Step 303: Based on each set of structural parameter data and the corresponding performance parameter data, multiple sets of simulation data are obtained.

[0061] After obtaining each set of structural parameter data and its corresponding performance parameter data, multiple sets of simulation data are paired. Specifically, each set of input structural parameter data is paired with the output performance parameter data to form a structured data unit, and all simulation data are summarized to obtain multiple sets of simulation data. These multiple sets of simulation data can be stored in tabular form, with each set including the input structural parameter data and the output performance parameter data for that simulation.

[0062] This embodiment transforms the original serial, manual modeling and simulation process into a parallel, standardized data production process, shortening the simulation cycle and improving the efficiency of design exploration. This effectively reduces the design time and R&D cost of the reed and lays a data foundation for discovering better solutions globally.

[0063] In one embodiment, such as Figure 4As shown, a method for determining correlation parameters is provided, which specifically includes the following steps:

[0064] Step 401: Based on multiple sets of simulation data, determine the parameter data set for each structural parameter and the parameter data set for each performance parameter.

[0065] Each set of simulation data includes structural parameters and corresponding performance parameters. The structural parameter data includes the parameter data corresponding to multiple structural parameters; the performance parameter data includes the parameter data corresponding to multiple performance parameters, and the parameter data is the specific value of the corresponding parameter.

[0066] Taking a reed consisting of two linear spring segments and three curved spring segments as an example, a set of structural parameter data includes: the length and width of the first linear spring segment, the length and width of the second linear spring segment, the radius of curvature (R) of the first curved spring segment, the radius of curvature (R) of the second curved spring segment, and the corresponding values ​​of the radius of curvature (R) of the third curved spring segment. Performance parameter data includes: the maximum stress of the reed under a given preset displacement condition, and the values ​​of the stiffness coefficient in the X direction, Y direction, and Z direction.

[0067] The parameter data set for each structural parameter is an ordered array or vector that organizes the values ​​of a specific structural parameter across all simulation data, used for correlation calculations. For example, the parameter data set for the width of all first linear spring segments in multiple sets of simulation data is formed; the parameter data set for the width of all second linear spring segments in multiple sets of simulation data is formed; the parameter data set for the radius of curvature spring segments in multiple sets of simulation data is formed; the parameter data set for the radius of curvature spring segments in multiple sets of simulation data is formed; the parameter data set for the radius of curvature spring segments in multiple sets of simulation data is formed; and the parameter data set for the radius of curvature spring segments in multiple sets of simulation data is formed.

[0068] The parameter data set for each performance parameter is an ordered array or vector that organizes the values ​​of a specific performance parameter across all simulation data, used for correlation calculations. For example, it retrieves the values ​​of all maximum stresses from multiple sets of simulation data to form a parameter data set for maximum stress; it retrieves the values ​​of all stiffness coefficients in the X direction from multiple sets of simulation data to form a parameter data set for the X-direction stiffness coefficient; it retrieves the values ​​of all stiffness coefficients in the Y direction from multiple sets of simulation data to form a parameter data set for the Y-direction stiffness coefficient; and it retrieves the values ​​of all stiffness coefficients in the Z direction from multiple sets of simulation data to form a parameter data set for the Z-direction stiffness coefficient.

[0069] Step 402: Based on the correlation analysis algorithm, determine the correlation coefficient between each structural parameter and each performance parameter according to the parameter data set of each structural parameter and the parameter data set of each performance parameter.

[0070] Association analysis algorithms are mathematical methods for quantifying the statistical dependency between two sets of parameter data. Examples include Pearson correlation coefficient analysis and analysis of variance (ANOVA). Pearson correlation coefficients range from -1 to 1, with -1 corresponding to negative correlation, 0 to no correlation, and 1 to positive correlation. By calculating the ratio of the covariance to the standard deviation between structural and performance parameters, the strength of the linear association between them is quantified. Specifically, the correlation coefficient is calculated for each structural parameter and each performance parameter, and the structural parameters are ranked according to their absolute values ​​to identify those parameters that significantly influence the performance parameters.

[0071] For example, considering the width of the first linear spring segment in the structural parameters, calculate the correlation coefficient between the width of the first linear spring segment and the maximum stress, the correlation coefficient between the width of the first linear spring segment and the stiffness coefficient in the X direction, the correlation coefficient between the width of the first linear spring segment and the stiffness coefficient in the Y direction, and the correlation coefficient between the width of the first linear spring segment and the stiffness coefficient in the Z direction. Understandably, it is necessary to calculate the correlation coefficient between each structural parameter and each performance parameter.

[0072] Taking the calculation of the correlation coefficient between the width of the first linear spring segment and the maximum stress through Pearson correlation coefficient analysis as an example, the calculation of the correlation coefficient is illustrated below:

[0073]

[0074] Where r represents the correlation coefficient between the width of the first linear spring segment and the maximum stress; n represents the number of data points in the parameter data set; X i The i-th data point in the parameter data set representing the width of the first linear spring segment; The average value of the parameter data set representing the width of the first linear spring segment; Y i The i-th data point in the parameter data set representing the maximum stress; This represents the average value of the parameter data set indicating maximum stress. Understandably, for each structural parameter and each performance parameter, the corresponding correlation coefficient needs to be calculated using the above formula.

[0075] Step 403: Use the structural parameters corresponding to the correlation coefficients that meet the preset conditions as the correlation parameters.

[0076] The preset condition can be that the absolute value of the correlation coefficient is greater than a preset coefficient threshold. The structural parameters corresponding to the correlation coefficients that meet the preset condition are used as association parameters. Specifically, after calculating the correlation coefficient between each structural parameter and each performance parameter, the absolute value of all correlation coefficients is calculated, and the structural parameters corresponding to the correlation coefficients whose absolute values ​​are greater than the preset coefficient threshold are used as association parameters. The preset coefficient threshold can be set according to actual usage requirements; for example, a preset coefficient threshold of 0.5. For example, the absolute value of the correlation coefficient between the width of the first linear spring segment and the maximum stress is calculated. If this absolute value is greater than the preset coefficient threshold, then the width of the first linear spring segment is used as the association parameter.

[0077] This embodiment effectively reduces the dimensionality of high-dimensional structural parameters by filtering related parameters, thereby reducing computational redundancy caused by irrelevant structural parameters. Furthermore, it eliminates reliance on subjective experience in parameter selection, enhancing the scientific rigor of subsequent optimization algorithms.

[0078] In one embodiment, such as Figure 5 As shown, a method for fitting a performance prediction model is provided, which specifically includes the following steps:

[0079] Step 501: Determine multiple sets of fitting data based on the correlation parameters and multiple sets of simulation data.

[0080] Each set of fitted data includes: correlation parameter data and corresponding performance parameter data. Multiple sets of fitted data are subsets extracted from multiple sets of simulation data, retaining only input-output pairs related to the correlation parameters. Specifically, for a set of simulation data, the simulation data includes structural parameter data and corresponding performance parameter data. Since the correlation parameters are determined within the structural parameters, the structural parameter data includes the correlation parameter data. The correlation parameter data corresponding to the correlation parameters is selected from the structural parameter data, and then integrated with the corresponding performance parameter data to obtain a set of fitted data. Through this method, correlation parameter data corresponding to the correlation parameters is extracted from each set of simulation data to obtain multiple sets of fitted data.

[0081] Step 502: Construct an initial prediction model based on multiple sets of fitted data and various preset models.

[0082] Multiple preset models are available, including: Keligin fitting model, Gaussian process regression fitting model, radial basis function fitting model, inverse distance weighted fitting model, and energy minimization spline interpolation fitting model.

[0083] Based on multiple sets of fitted data, the first function is obtained by fitting a Kriging model. The Kriging model is a geostatistical interpolation model based on the spatial autocorrelation assumption. It describes the spatial dependence between parameters through a variogram and is suitable for performance response modeling with strong spatial structure characteristics. For example, the Kriging method. The Kriging model takes multiple sets of fitted data as input, estimates the semi-variogram, and solves the Kriging equations to generate a weighted linear interpolation function, i.e., the first function. The fitting of the Kriging method is implemented using existing conventional methods, which will not be elaborated here.

[0084] Based on multiple sets of fitted data, a second function is obtained by fitting a Gaussian process regression model. The Gaussian process regression model is a Bayesian nonparametric regression method that models the performance response as a Gaussian stochastic process. Its output includes not only the predicted mean but also a quantification of prediction uncertainty, allowing it to maintain robust predictive ability even with small sample sizes and supporting active learning strategies. For example, Gaussian process regression (GPR). The performance response mapping function, i.e., the second function, generated by the Gaussian process regression model based on multiple sets of fitted data. The fitting of Gaussian process regression (GPR) is implemented using existing conventional methods, which will not be elaborated here.

[0085] Based on multiple sets of fitted data, a third function is obtained by fitting a model using radial basis functions (RBF). The RBF fitting model is an interpolation model composed of a linear combination of basis functions with the distance from the input point to the center point as the independent variable. It possesses global approximation capability and good smoothness, and can be used to achieve high-precision interpolation in the key parameter space. It also has a strong characterization ability for local nonlinear responses such as deformation sensitivity. For example, the RBF. The performance response mapping function generated by the RBF fitting model, based on multiple sets of fitted data, is the third function. The fitting of the RBF is achieved through existing conventional methods, which will not be elaborated here.

[0086] Based on multiple sets of fitted data, the fourth function is obtained through an inverse distance weighted fitting model. The inverse distance weighted fitting model is a spatial interpolation method based on neighborhood weighted averaging. The predicted value is obtained by summing the neighboring samples according to the inverse power of their distance. It can be used for fast, coarse predictions with extremely low computational cost, suitable for scenarios with high real-time requirements and relatively uniform data distribution. For example, the inverse distance weighted method (IDW). The inverse distance weighted fitting model generates a performance response mapping function, i.e., the fourth function, based on multiple sets of fitted data. The fitting of the inverse distance weighted method (IDW) is implemented through existing conventional methods, which will not be elaborated here.

[0087] Based on multiple sets of fitted data, the fifth function is obtained by fitting an energy-minimum spline interpolation model. The energy-minimum spline interpolation model is a mathematical model that constructs a smooth interpolation function by minimizing a certain energy functional, such as the curvature integral. This ensures that the response curve is physically reasonable and oscillating, and can be used to guarantee the physical interpretability and continuity of performance prediction results. For example, Energy Minimum Spline Interpolation (EMS). The performance response mapping function generated by the energy-minimum spline interpolation model, based on multiple sets of fitted data, is the fifth function. The fitting of Energy Minimum Spline Interpolation (EMS) is implemented using existing conventional methods, which will not be elaborated here.

[0088] After obtaining the first, second, third, fourth, and fifth functions, weights are assigned to each function to construct an initial prediction model. The initial prediction model is as follows:

[0089] Y=f(X)=w k f k (X)+w s f s (X)+w r f r (X)+w i f i (X)+w e f e (X)

[0090] Where Y=f(X) represents the initial prediction model; Y represents the performance parameter data; X represents the correlation parameter data; w k f represents the weight of the first function; k (X) represents the first function, which is the function obtained by the Kriging method; w s f represents the weight of the second function; s (X) represents the second function, which is the function obtained from Gaussian process regression; w r f represents the weight of the third function; r (X) represents the third function, which is the function obtained from the radial basis functions; w i f represents the weight of the fourth function; i (X) represents the fourth function, which is the function obtained by the inverse distance weighting method; w e f represents the weight of the fifth function; e (X) represents the fifth function, which is the function obtained by minimizing energy through spline interpolation.

[0091] Step 503: Based on the genetic algorithm and multiple sets of fitting data, the initial prediction model is iterated to obtain the performance prediction model.

[0092] After obtaining the initial prediction model, the prediction performance data is determined based on the correlation parameter data from multiple sets of fitted data and the initial prediction model itself. Taking a set of fitted data as an example, this set includes correlation parameter data and corresponding performance parameter data. The correlation parameter data is input into the initial prediction model, and the initial prediction model outputs the prediction performance data. By inputting the correlation parameter data from each set of fitted data into the initial prediction model, the prediction performance data corresponding to each correlation parameter data can be obtained.

[0093] Based on the predicted performance data, performance parameter data from multiple sets of fitted data, and the genetic algorithm, the weights of the first, second, third, fourth, and fifth functions in the initial prediction model are optimized to obtain the performance prediction model. The genetic algorithm is a global optimization algorithm that simulates the mechanism of biological evolution. It searches for the optimal solution in the solution space through selection, crossover, and mutation operations, automatically optimizing the weights in the initial prediction model to improve the overall predictive ability of the model. With the goal of minimizing the difference between the predicted performance data and the performance parameter data, the weights of the five functions are set as the optimization variables of the genetic algorithm. The population is initialized in the interval [0, 1], and iterative evolution is performed through selection, crossover, and mutation operations until the fitness converges, outputting the optimal weight combination and the corresponding performance prediction model. The termination conditions of the genetic algorithm include at least one of the following: minimizing the difference between the predicted performance data and the performance parameter data, reaching a preset number of iterations, and the optimal solution showing no significant improvement over multiple generations. The weight values ​​of the five functions generated by the genetic algorithm are substituted into the initial prediction model to obtain the performance prediction model.

[0094] This embodiment transforms the traditional trial-and-error simulation that relies on human experience into a data-driven automated modeling process, which significantly reduces the number of finite element simulations and shortens the model training and verification cycle. This enables rapid performance prediction and parameter optimization in subsequent designs, effectively reducing reliance on highly skilled engineers and computational resource consumption, thereby shortening the reed design cycle and reducing design costs.

[0095] In one embodiment, such as Figure 6 As shown, a method for determining target parameter data is provided, which specifically includes the following steps:

[0096] Step 601: Based on the structural parameter conditions, generate the initial population using an optimization algorithm.

[0097] The initial population includes multiple sets of associated structural parameter data corresponding to the associated parameters of the reed. These multiple sets of associated structural parameter data constitute the first-generation candidate solution set generated during the iteration of the optimization algorithm. The multiple sets of associated structural parameter data are generated based on the value range of the associated parameters in the structural parameter conditions, using uniform random sampling or Latin hypercube sampling methods. Each set of associated structural parameter data includes the numerical values ​​corresponding to all associated parameters. The optimization algorithm can be a combination of GA, PSO, and gradient descent, or other mature optimization algorithms such as Simulated Annealing, Ant Colony Optimization, Bayesian Optimization, etc., or any combination of the above algorithms.

[0098] Step 602: Based on the initial population and performance prediction model, determine the associated performance parameter data corresponding to each group of associated structural parameter data.

[0099] After obtaining the initial population, the associated structural parameter data of each group in the initial population are input into the performance prediction model to obtain the associated performance parameter data corresponding to each group of associated structural parameter data. The associated performance parameter data includes: the numerical values ​​corresponding to the performance parameters, which include the maximum stress of the reed under a given preset motion displacement condition, the stiffness coefficient in the X direction, the stiffness coefficient in the Y direction, and the stiffness coefficient in the Z direction.

[0100] Step 603: Based on the initial population, the associated performance parameter data corresponding to each set of associated structural parameter data, and the constraints, the population is iterated using an optimization algorithm and a performance prediction model until the preset termination condition is met, and at least one set of target parameter data for the reed is determined.

[0101] After obtaining the association performance parameter data corresponding to each group of association structure parameter data in the initial population, the association performance parameter data is fed back to the optimization algorithm. The optimization algorithm calculates the individual fitness based on the performance prediction results, and performs operations such as selection, crossover, and mutation to generate a new population under the premise of satisfying constraints. The performance is then repeatedly evaluated through the performance prediction model to find the global optimal solution, which is the target parameter data.

[0102] The constraints include: the maximum stress limit of the reed under a given preset displacement condition, and the limits for the stiffness coefficient in the X, Y, and Z directions. Specifically, the maximum stress limit under a given preset displacement condition includes: the maximum stress of the reed under this condition must be less than the yield strength of the reed material, and the maximum stress must be minimized. The stiffness coefficient limit in the X direction includes: the stiffness coefficient in the X direction must be greater than a first stiffness coefficient threshold, and the stiffness coefficient in the X direction must be at its maximum. The first stiffness coefficient threshold can be set according to actual usage requirements, for example, 5000 N / m. The stiffness coefficient limit in the Y direction includes: the stiffness coefficient in the Y direction must be greater than a second stiffness coefficient threshold, and the stiffness coefficient in the Y direction must be at its maximum. The second stiffness coefficient threshold can be set according to actual usage requirements, for example, 5000 N / m. The constraints on the stiffness coefficient in the Z direction include: the stiffness coefficient in the Z direction must be within a preset threshold range, and the difference between the stiffness coefficient in the Z direction and the third stiffness coefficient threshold must be minimized. The preset threshold range and the third stiffness coefficient threshold can be set according to actual usage requirements; for example, the preset threshold range could be 65 ± 1 N / m, and the third stiffness coefficient threshold could be 65 N / m. Since the constraints all include maximum / minimum conditions for each performance parameter, the optimization algorithm ultimately outputs the optimal solution, i.e., the target parameter data. The target parameter data includes a set of values ​​for all associated parameters.

[0103] The preset termination condition is used to determine whether the population iteration is complete, ensuring that the algorithm stops and outputs the current optimal result within a reasonable time. For example, the termination condition could be that the change in the correlation performance parameter data between two adjacent generations is less than 1e to the power of -10, where e is a natural constant. Alternatively, it could be that iteration stops after reaching a preset number of iterations. This embodiment does not impose specific limitations on the termination condition; it can be set according to actual usage requirements.

[0104] This embodiment introduces an initial population mechanism into the optimization algorithm, generating multiple sets of key structural parameter data covering the design space based on structural parameter conditions. A performance prediction model is then used to quickly estimate the key performance parameters corresponding to each set of key structural parameter data, achieving efficient solution evaluation. Furthermore, by combining constraints, high-quality solutions are continuously screened and evolved during the population iteration process, gradually approaching the optimal design region. A preset termination condition ensures the optimization process automatically ends, avoiding resource waste. This data-driven approach replaces manual trial and error based on experience, reducing reliance on high-cost simulations, improving search efficiency and automation, effectively shortening the design cycle, and reducing R&D manpower and time costs.

[0105] In one specific embodiment, such as Figure 7As shown, an automatic reed optimization design method is provided, the core of which is to construct a closed-loop system consisting of automated modeling, automated simulation, data processing, and optimization algorithms. Specifically, it includes the following steps:

[0106] Step 1: Define design goals and constraints.

[0107] First, define the structural parameters and constraints for the reed design. Design parameter variables, determining the variable structural parameters in the reed's geometric model. The reed can be divided into multiple spring segments, including straight spring segments and curved spring segments. The structural parameters corresponding to straight spring segments are length and width, etc.; the structural parameter corresponding to curved spring segments is the radius of curvature (R-angle). Set the value range for each structural parameter. Set the material type of the reed to titanium copper, beryllium copper, stainless steel, etc. Set the thickness of the reed to between 0.03 mm and 0.05 mm, preferably fixed at 0.04 mm. Design constraints: Constraint 1: Under a given displacement in the Z direction, the maximum stress of the reed is less than the material's yield strength. For example, a given displacement of 0.46 mm and a material yield strength of 1380 MPa are constants determined based on the material type. Constraint 2: The stiffness coefficient Kx in the X direction meets a specific value or range, for example, Kx > 5000 N / m. Constraint 3: The stiffness coefficient (Ky) in the Y direction meets a specific value or range, for example, Ky > 5000 N / m. Constraint 4: The stiffness coefficient Kz in the Z direction meets a specific value or range, for example, Kz = 65±1 N / m.

[0108] Step 2, Automated parametric modeling.

[0109] The model is automatically generated using the API of the CAD software. API stands for Application Programming Interface. CAD software includes SolidWorks, CATIA, Pro / E (Creo), AutoCAD, and UG (Siemens NX). A script is written that accepts a set of structural or related parameter data from step 1 or step 7 and automatically drives the CAD software to generate the corresponding 3D model. The model is exported in a standard format for subsequent simulation.

[0110] Step 3, Experimental design based on Latin hypercube (LHS).

[0111] In the initial optimization phase, an efficient sampling method was employed to quickly explore the entire design space. For the range of all structural parameters defined in step 1, Latin Hypercube Sampling (LHS) was used to generate N sets of evenly distributed and well-covered initial sampling points, effectively generating multiple sets of structural parameter data. LHS ensures uniform sampling across each dimension of the structural parameter, offering higher efficiency compared to random sampling.

[0112] Step 4: Automated batch simulation.

[0113] The simulation process is automated by connecting modeling and simulation software. Finite element analysis (FEM) software is used, and the Pyansys library is employed for automated operation. The Pyansys library is a Python library used for scripting control of the FEM software. The FEM software can be ANSYS Mechanical, Abaqus and its Python script interface, COMSOL Multiphysics, Nastran, etc. Specifically, a master Python script is first written, which loops through the structural parameter data generated by LHS sampling in step 3. Within the loop, the script from step 2 is called to generate a CAD model of the structural parameter data. The Pyansys library is then used to start the FEM software and automate the simulation. The specific simulation flow is as follows: first, the CAD model is input, the material properties and thickness of the springs in the model are defined, the mesh is automatically generated, boundary conditions are applied, and finally, the static structural analysis is solved. The boundary conditions specifically include: simulating the fixed end of the etching process, such as the outer ring fixation; and applying a forced displacement in the Z direction, such as 0.46 mm, at the carrier connection.

[0114] Step 5: Automatic extraction and data processing of simulation results.

[0115] After the finite element analysis software completes the automated simulation, performance parameter data is extracted from the ANSYS simulation results. These performance parameters include the overall stress of the CAD model and the stiffness coefficients in the X, Y, and Z directions at the point of forced displacement. Specifically, the ANSYS library launches the finite element analysis software and automatically executes the simulation. After the simulation solution is completed, the result file is automatically queried to extract the specific values ​​of the following performance parameters: the overall stress of the model and the stiffness coefficients in the X, Y, and Z directions at the point of applied displacement. The extracted performance parameter data, along with the corresponding structural parameter data, are stored in a database or table to form a dataset of input structural parameter data and output performance parameter data.

[0116] Step 6: Sensitivity analysis and feature selection.

[0117] After executing steps 3-5 multiple times, based on the dataset of input structural parameter data and output performance parameter data, the correlation parameters with the strongest relationship to the performance parameters are selected from the structural parameters. That is, the dataset is analyzed to determine which structural parameters have the most significant impact on the performance parameters, and structural parameters with insignificant effects are removed, extracting the correlation parameters. An accurate mathematical model is established to express the relationship between the correlation parameters and the performance parameters. Methods such as analysis of variance (ANOVA) and Pearson correlation coefficient can be used to determine which structural parameters have the most significant impact on the performance parameters. Specifically, statistical analysis is performed on the dataset collected in step 5 using methods such as analysis of variance (ANOVA) and Pearson correlation coefficient. For example, the structural parameters include: the width W1 of the first straight spring segment, the width W2 of the second straight spring segment, the width R1 of the first curve spring segment, and the width R2 of the second curve spring segment. The width W1 of the first linear spring segment has the greatest impact on the overall stress of the model, while the width W2 of the second linear spring segment has the greatest impact on the stiffness coefficient in the Z direction. The width R1 of the first curved spring segment affects both the overall stress and the stiffness coefficient in the Z direction, while the width R2 of the second curved spring segment has no impact on either the overall stress or the stiffness coefficient in any single direction. Based on this, the structural parameters that affect performance parameters—the widths W1, W2, and R1 of the first linear spring segment, are considered as correlation parameters. Determining these correlation parameters is beneficial for establishing an accurate mathematical model between them and the performance parameters.

[0118] Step 7, Gene aggregation optimization modeling.

[0119] Following step 6, gene aggregation optimization modeling is performed to construct a performance prediction model. A mathematical model of the association parameters and performance parameters, i.e., the performance prediction model, is constructed. This performance prediction model is used to directly obtain the target association parameter data that satisfies the target performance parameter data. Specifically, gene aggregation optimization is used to construct the performance prediction model. Multiple preset surrogate models are selected and combined with weights. For example, the Kriging method is selected. k (Krigin), Gaussian process regression f s (GPR), radial basis function f r (RBF), Inverse Distance Weighted Method f i (IDW), Energy Minimization Spline Interpolation f e (EMS), etc., to construct a performance prediction model f=w k f k +w s f s +w r f r +w i f i +w e fe The weights w are obtained through a genetic algorithm for optimization. k w s w r w i and w e Specifically, the weights in the above model are optimized using a genetic algorithm (GA) to find the optimal weights. The genetic algorithm can search a wide range of values, avoiding getting trapped in local optima. By combining the relationships between the correlation parameters and performance parameters, the following performance prediction model is constructed:

[0120] Y=f(X)=w k f k (X)+w s f s (X)+w r f r (X)+w i f i (X)+w e f e (X)

[0121] Where Y=f(X) represents the initial prediction model; Y represents the performance parameter data; X represents the correlation parameter data; w k f represents the weight of the first function; k (X) represents the first function, which is the function obtained by the Kriging method; w s f represents the weight of the second function; s (X) represents the second function, which is the function obtained from Gaussian process regression; w r f represents the weight of the third function; r (X) represents the third function, which is the function obtained from the radial basis functions; w i f represents the weight of the fourth function; i (X) represents the fourth function, which is the function obtained by the inverse distance weighting method; w e f represents the weight of the fifth function; e (X) represents the fifth function, which is the function obtained by minimizing energy through spline interpolation. The termination conditions of a genetic algorithm include the population iteration continuing until one of the following conditions is met: 1. The difference between the predicted value of the performance prediction model and the corresponding true value in the dataset is minimized. 2. The number of iterations reaches an upper limit, such as 1000. 3. The optimal solution shows no significant improvement over multiple generations.

[0122] Step 8: Output the optimal design scheme.

[0123] After the performance prediction model is established, the relationship between the correlation parameters and the structural parameters is clear, that is, Y=f(X). By inputting the target performance parameter data, the target correlation parameter data can be obtained in reverse, that is, X=f^(-1)(Y).

[0124] For example, constraints were defined in step 1. A performance prediction model was built in step 7. The current need is to find a set of target-related parameter data whose corresponding target performance parameters satisfy the constraints. This problem can be transformed into a multi-objective optimization problem. Through multi-objective optimization, the overall stress of the model corresponding to the target-related parameter data is minimized and is less than the material's yield strength; the stiffness coefficient in the corresponding X direction is maximized and is greater than a specified value; the stiffness coefficient in the corresponding Y direction is maximized and is greater than a specified value; and the stiffness coefficient in the corresponding Z direction is minimized and within a specific range. Specifically, this can be represented as follows:

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133] in, Y represents the target-related parameter data, σ represents the performance parameter data, and σ represents the overall stress of the model. This represents the overall stress of the model corresponding to the target-related parameter data, where Kx represents the stiffness coefficient in the X direction. Ky represents the stiffness coefficient in the X direction corresponding to the target associated parameter data, and Ky represents the stiffness coefficient in the Y direction. Kz represents the stiffness coefficient in the Y direction corresponding to the target associated parameter data, and Kz represents the stiffness coefficient in the Z direction. This represents the stiffness coefficient in the Z direction corresponding to the target's associated parameters. The specified value represents the stress of the entire model. This indicates the specified value of the stiffness coefficient in the X direction. This indicates the specified value of the stiffness coefficient in the Y direction. This represents the specified value of the stiffness coefficient in the Z direction, where ε is a constant.

[0134] Multi-objective problems can be transformed into single-objective problems. Multiplying the maximization problem by -1 transforms it into a minimization problem. Therefore, a single-objective optimization minimization problem can be obtained through linear weighting, as follows:

[0135]

[0136]

[0137]

[0138]

[0139]

[0140] in, The weights represent the overall stress of the model. The weight representing the stiffness coefficient in the X direction. The weighting of the stiffness coefficient in the Y direction. This represents the weight of the stiffness coefficient in the Z direction. Each associated parameter has its own value range, and the weight of each performance parameter is set according to actual usage requirements. Depending on the actual project scenario, if stress is the primary concern, the weight of the force can be set larger. Then, through continuous iteration using a genetic algorithm combined with a performance prediction model, the target parameter data that meets the target performance parameters can be determined. The termination conditions of the genetic algorithm include: the population iteration continues until one of the following conditions is met: 1. The difference in performance parameters between two adjacent generations is less than 1e to the power of -10. 2. The number of iterations reaches the upper limit, for example, 300 times.

[0141] This embodiment achieves fully automated closed-loop design of reeds, from parametric modeling to finite element simulation and iterative optimization algorithm. Latin hypercube is used for initial DOE design, replacing the manual design trial-and-error approach in related technologies. This efficiently explores the high-dimensional design space with fewer sample points, providing high-quality initial data for subsequent optimization. A hybrid strategy combining global and local search is employed. Global search ensures breadth of optimization, escaping local optima traps, while gradient descent ensures depth and efficiency, quickly converging to the precise optimal solution. Deep integration with CAD and finite element analysis software via scripts such as Pyansys automates cross-platform data and control flows, eliminating bottlenecks and error sources associated with manual data import / export. The design process is transformed from a traditional manual design verification model to a goal-driven automatic generation model. Engineers are no longer simulation operators but rather definers of goals and constraints.

[0142] This embodiment enables unattended automated iteration, completing thousands of design-simulation cycles within 24 hours. Compared to related technologies where manual iteration takes hours and finding a usable solution takes weeks, this embodiment shortens the R&D cycle from weeks to days or even hours, significantly reducing the development timeline. It frees engineers from tedious and repetitive manual modeling and simulation work, allowing them to focus on defining optimization goals and analyzing optimal results, greatly improving human resource utilization efficiency and significantly reducing R&D costs. The automated optimization algorithm can explore complex design spaces that are difficult for humans to access, finding the global optimum solution. This ensures that the designed spring, while meeting stiffness coefficient requirements, optimizes stress to a lower level, thereby significantly improving product fatigue life and reliability. Step 6 quantifies which structural parameter has the greatest impact on performance parameters and uses it as a correlation parameter. This knowledge is quantifiable and reusable, helping to deepen the understanding of spring design and guiding the rapid design of subsequent products, thus achieving the accumulation and reuse of design knowledge. Fully automated data flow avoids problems such as data copying errors and model import errors that may occur in manual operations, ensuring the stability and reliability of the design process.

[0143] The above embodiments can be achieved by integrating existing commercial optimization platforms, such as modeFRONTIER, HyperStudy, and ANSYS DesignXplorer, and developing automated plugins specific to the voice coil motor (VCM) reeds.

[0144] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0145] Based on the same inventive concept, this application also provides a reed design apparatus for implementing the reed design method described above. The solution provided by this apparatus is similar to the solution described in the above method; therefore, the specific limitations in one or more reed design apparatus embodiments provided below can be found in the limitations of the reed design method described above, and will not be repeated here.

[0146] In one embodiment, such as Figure 8 As shown, a reed design device is provided, including: an acquisition module 100, a simulation module 200, a parameter determination module 300, a fitting module 400, and an optimization module 500, wherein:

[0147] The acquisition module 100 is used to acquire the structural parameters and constraints of the reed.

[0148] The simulation module 200 is used to perform batch simulations based on structural parameter conditions to obtain multiple sets of simulation data for the reed; each set of simulation data includes: structural parameter data of the reed and corresponding performance parameter data.

[0149] The parameter determination module 300 is used to determine the associated parameters from the structural parameters corresponding to the structural parameter data based on multiple sets of simulation data.

[0150] The fitting module 400 is used to fit a performance prediction model based on the correlation parameters, multiple sets of simulation data, and multiple preset models.

[0151] The optimization module 500 is used to determine at least one set of target parameter data for the reed based on the structural parameter conditions and the constraint conditions, and on the optimization algorithm and performance prediction model.

[0152] The acquisition module 100 is also used to construct an initial simulation model of the reed; based on the initial simulation model, determine the initial structural parameter data of each structural parameter in the reed; acquire the floating range corresponding to each structural parameter; and determine the structural parameter conditions according to the initial structural parameter data of each structural parameter and the floating range corresponding to each structural parameter.

[0153] The simulation module 200 is further configured to sample the structural parameter conditions using a preset sampling algorithm to obtain multiple sets of structural parameter data; perform simulation based on the multiple sets of structural parameter data and an initial simulation model to obtain performance parameter data corresponding to each set of structural parameter data; and obtain multiple sets of simulation data based on each set of structural parameter data and the performance parameter data corresponding to each set of structural parameter data.

[0154] The parameter determination module 300 is also used to determine the parameter data set for each structural parameter and the parameter data set for each performance parameter based on multiple sets of simulation data; based on the correlation analysis algorithm, it determines the correlation coefficient between each structural parameter and each performance parameter based on the parameter data set for each structural parameter and the parameter data set for each performance parameter; and uses the structural parameter corresponding to the correlation coefficient that meets the preset conditions as the correlation parameter.

[0155] The fitting module 400 is further configured to determine multiple sets of fitting data based on the correlation parameters and multiple sets of simulation data; each set of fitting data includes: correlation parameter data of the correlation parameters and corresponding performance parameter data; construct an initial prediction model based on multiple sets of fitting data and multiple preset models; and iterate the initial prediction model based on the genetic algorithm and multiple sets of fitting data to obtain a performance prediction model.

[0156] The fitting module 400 is also used to fit a first function based on multiple sets of fitting data using a Kriging fitting model; to fit a second function based on multiple sets of fitting data using a Gaussian process regression fitting model; to fit a third function based on multiple sets of fitting data using a radial basis function fitting model; to fit a fourth function based on multiple sets of fitting data using an inverse distance weighted fitting model; and to fit a fifth function based on multiple sets of fitting data using an energy minimization spline interpolation fitting model; and to set weights for the first, second, third, fourth, and fifth functions respectively to construct an initial prediction model.

[0157] The fitting module 400 is further configured to determine prediction performance data based on the correlation parameter data in the multiple sets of fitting data and the initial prediction model; and to optimize the weights of the first function, the second function, the third function, the fourth function and the fifth function in the initial prediction model based on the prediction performance data, the performance parameter data in the multiple sets of fitting data and the genetic algorithm, so as to obtain the performance prediction model.

[0158] The optimization module 500 is further configured to generate an initial population based on the structural parameter conditions and an optimization algorithm; the initial population includes multiple sets of associated structural parameter data corresponding to the associated parameters; based on the initial population and the performance prediction model, determine the associated performance parameter data corresponding to each set of associated structural parameter data; based on the initial population, the associated performance parameter data corresponding to each set of associated structural parameter data, and the constraints, perform population iteration based on the optimization algorithm and the performance prediction model until a preset termination condition is met, and determine at least one set of target parameter data for the reed.

[0159] Each module in the aforementioned reed design device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0160] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 9As shown, the computer device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a reed design method.

[0161] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0162] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the reed design methods described in the above embodiments.

[0163] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements any of the reed design methods described in the above embodiments.

[0164] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0165] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0166] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method of designing a reed, characterized by, The method comprises: obtaining a structure parameter condition and a constraint condition of a reed; performing batch simulation based on the structure parameter condition to obtain a plurality of groups of simulation data of the reed; each group of simulation data comprises structure parameter data of the reed and corresponding performance parameter data; determining an associated parameter in the structure parameter data corresponding to the structure parameter according to a plurality of groups of simulation data; fitting a performance prediction model according to the associated parameter, a plurality of groups of simulation data and a preset model; determining at least one group of target parameter data of the reed based on an optimization algorithm and the performance prediction model according to the structure parameter condition and the constraint condition.

2. The method of claim 1, wherein, obtaining a structure parameter condition of a reed comprises: constructing an initial simulation model of the reed; determining initial structure parameter data of each structure parameter in the reed based on the initial simulation model; obtaining a floating range corresponding to each structure parameter; determining a structure parameter condition according to the initial structure parameter data of each structure parameter and the floating range corresponding to each structure parameter.

3. The method of claim 1, wherein, The batch simulation based on the structure parameter condition to obtain a plurality of groups of simulation data of the reed comprises: sampling the structure parameter condition by a preset sampling algorithm to obtain a plurality of groups of structure parameter data; performing simulation based on a plurality of groups of structure parameter data and an initial simulation model to obtain performance parameter data corresponding to each group of structure parameter data; obtaining a plurality of groups of simulation data according to each group of structure parameter data and the performance parameter data corresponding to each group of structure parameter data.

4. The method of claim 1, wherein, The structure parameter data comprises parameter data corresponding to a plurality of structure parameters; the performance parameter data comprises parameter data corresponding to a plurality of performance parameters; The determination of an associated parameter in the structure parameter based on a plurality of groups of simulation data comprises: determining a parameter data group of each structure parameter and a parameter data group of each performance parameter according to a plurality of groups of simulation data; determining a correlation coefficient between each structure parameter and each performance parameter based on an association analysis algorithm according to the parameter data group of each structure parameter and the parameter data group of each performance parameter; determining the structure parameter corresponding to the correlation coefficient satisfying a preset condition as the associated parameter.

5. The method of claim 1, wherein, The fitting of a performance prediction model according to the associated parameter, a plurality of groups of simulation data and a preset model comprises: determining a plurality of groups of fitting data according to the associated parameter and a plurality of groups of simulation data; each group of fitting data comprises associated parameter data of the associated parameter and corresponding performance parameter data; constructing an initial prediction model according to a plurality of groups of fitting data and a plurality of preset models; iterating the initial prediction model based on a genetic algorithm and a plurality of groups of fitting data to obtain a performance prediction model.

6. The method of claim 5, wherein, The plurality of preset models comprises a Kriging fitting model, a Gaussian process regression fitting model, a radial basis function fitting model, an inverse distance weighted fitting model and an energy minimization spline interpolation fitting model; the construction of an initial prediction model according to a plurality of groups of fitting data and a plurality of preset models comprises: fitting a first function by a Kriging fitting model based on a plurality of groups of fitting data; Fitting a second function based on multiple sets of fitting data by a Gaussian process regression fitting model; Fitting a third function based on multiple sets of fitting data by a radial basis function fitting model; Fitting a fourth function based on multiple sets of fitting data by an inverse distance weighted fitting model; Fitting a fifth function based on multiple sets of fitting data by an energy minimization spline interpolation fitting model; Respectively setting weights of the first function, the second function, the third function, the fourth function and the fifth function, and constructing an initial prediction model.

7. The method of claim 6, wherein, The performance prediction model is obtained by iterating the initial prediction model based on the genetic algorithm and the multiple sets of fitting data, including: Determining prediction performance data according to the correlation parameter data in the multiple sets of fitting data and the initial prediction model; Optimizing the weights of the first function, the second function, the third function, the fourth function and the fifth function in the initial prediction model based on the genetic algorithm and the prediction performance data and the performance parameter data in the multiple sets of fitting data, to obtain the performance prediction model.

8. The method of claim 1, wherein, The at least one set of target parameter data of the reed is determined based on the optimization algorithm and the performance prediction model according to the structure parameter condition and the constraint condition, including: Generating an initial population based on the optimization algorithm according to the structure parameter condition; the initial population includes multiple sets of correlation structure parameter data corresponding to the correlation parameters; Determining the correlation performance parameter data corresponding to each set of correlation structure parameter data based on the initial population and the performance prediction model; Performing population iteration based on the optimization algorithm and the performance prediction model according to the initial population, the correlation performance parameter data corresponding to each set of correlation structure parameter data and the constraint condition, until a preset termination condition is met, to determine the at least one set of target parameter data of the reed.

9. A reed designing apparatus characterized by comprising: The device includes: An acquisition module for acquiring a structure parameter condition and a constraint condition of a reed; A simulation module for performing batch simulation based on the structure parameter condition to obtain multiple sets of simulation data of the reed; each set of simulation data includes structure parameter data and corresponding performance parameter data of the reed; A parameter determination module for determining correlation parameters in the structure parameters corresponding to the structure parameter data according to multiple sets of simulation data; A fitting module for fitting a performance prediction model according to the correlation parameters, multiple sets of simulation data and multiple preset models; An optimization module for determining at least one set of target parameter data of the reed based on an optimization algorithm and a performance prediction model according to the structure parameter condition and the constraint condition. 10.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-9. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8.

11. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 8.

Images