Data optimization method, device, apparatus and storage medium

Abstract

Embodiments of the present application provide a data optimization method, device and equipment and a storage medium. The method comprises: obtaining an initial population data set, the initial population data set comprising at least one individual data, the individual data comprising at least one design parameter; performing reverse learning processing on the initial population data set to obtain a first population data set; performing differential evolution processing and crossover processing on the first population data set to obtain a second population data set; performing non-dominated sorting processing and clustering processing on the second population data set to obtain a target population data set, the target population data set being used for multi-objective optimization. Through multi-working-condition constraint processing and multi-modal solution set maintenance, the output target population data set can meet the design requirements under different working conditions, provide diversified candidate schemes for engineers, shorten the design cycle and improve the comprehensive performance of the suspension system.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a data optimization method, apparatus, device, and storage medium. Background Technology

[0002] With the rapid development of the automotive and construction machinery industries, vehicle vibration and noise performance has become a core indicator for measuring ride comfort and product quality. The powertrain (such as the engine and transmission) is connected to the vehicle body through the suspension system. The design of the suspension system needs to achieve decoupling in multiple degrees of freedom, while meeting the vibration suppression requirements under different operating conditions (such as idling, acceleration, and cruising).

[0003] The dynamic model of the suspension system exhibits high dimensionality, nonlinearity, and strong coupling, posing multiple challenges to optimization design. Among related technologies, the application of multi-objective optimization algorithms in suspension decoupling optimization mainly relies on the Non-dominated Sorting Genetic Algorithm II (NSGA-II). NSGA-II achieves Pareto front search through non-dominated sorting, crowding calculation, and a fast convergence mechanism, which can handle multi-objective conflicts and generate a relatively uniformly distributed solution set.

[0004] However, in suspension system optimization, NSGA-II suffers from problems such as low search efficiency, weak local search capability, insufficient diversity preservation, and poor adaptability to multiple operating conditions. Therefore, there is an urgent need for a dataset applicable to NSGA-II to address these issues. Summary of the Invention

[0005] This application provides a data optimization method, apparatus, device, and storage medium to solve the problems of low search efficiency, weak local search capability, insufficient diversity preservation, and poor adaptability to multiple operating conditions in NSGA-II.

[0006] In a first aspect, embodiments of this application provide a data optimization method, including:

[0007] Obtain an initial population dataset, which includes at least one individual data point, and the individual data point includes at least one design parameter.

[0008] The initial population dataset is processed by reverse learning to obtain the first population dataset.

[0009] Differential evolution and crossover processing are performed on the first population dataset to obtain the second population dataset;

[0010] The second population dataset is subjected to non-dominated sorting and clustering to obtain the target population dataset, which is used for multi-objective optimization.

[0011] In one possible implementation, the second population dataset undergoes non-dominated sorting and clustering to obtain the target population dataset, including:

[0012] Perform non-dominated sorting on at least one experimental individual data in the second population dataset to obtain at least one parent data and at least one child data;

[0013] Determine the crowding distance of at least one parent data;

[0014] The third population dataset is determined based on the crowding distance. The third population dataset includes at least the K data points with the largest crowding distance among the parent data.

[0015] Clustering is performed on the third population dataset to obtain the target population dataset.

[0016] In one possible implementation, the third population dataset is clustered to obtain the target population dataset, including:

[0017] Gaussian perturbation is applied to the parent data in the third population dataset to generate perturbation data corresponding to each parent data.

[0018] The perturbation data corresponding to each parent generation data is optimized by neighborhood search to obtain the optimized data corresponding to each parent generation data.

[0019] Based on the data of each parent generation and the corresponding optimized data of each parent generation, at least one candidate data is determined;

[0020] Clustering is performed on at least one candidate data to obtain the target population dataset, which includes at least one category and the target data corresponding to each category.

[0021] In one possible implementation, at least one candidate data is determined based on each parent generation data and the corresponding optimized data, including:

[0022] For any parent data, if the parent data dominates the optimized data corresponding to the parent data, then the candidate data is determined to be the parent data;

[0023] If the optimized data corresponding to the parent data dominates the parent data, then the candidate data is determined to be the optimized data corresponding to the parent data;

[0024] If the parent data and the corresponding optimized data do not dominate each other, then the data with a crowding distance greater than or equal to the preset distance are determined as candidate data.

[0025] In one possible implementation, the initial population dataset is subjected to back-learning processing to obtain a first population dataset, including:

[0026] Determine the reverse individual data corresponding to each individual data point in the initial population dataset;

[0027] Get constraints and preset functions;

[0028] Based on the constraints and preset functions, at least one individual data and at least one reverse individual data are filtered to obtain a first type of group dataset. The first type of group dataset includes at least one first individual data, and the at least one first individual data includes at least one individual data and the first N data from at least one reverse individual data.

[0029] In one possible implementation, the first population dataset undergoes differential evolution and crossover processing to obtain a second population dataset, including:

[0030] In the first type of population dataset, M first volume data are randomly selected, where M is 3;

[0031] Based on a preset scaling factor, differential evolution is performed on M first-body data to obtain at least one differential data.

[0032] Obtain the preset crossover probability;

[0033] Based on the preset crossover probability, the first population dataset, and at least one differential evolution data, a second population dataset is determined, which includes data from at least one experimental individual.

[0034] In one possible implementation, the method further includes:

[0035] For any target data point in the target population dataset, determine the hypervolume boost rate corresponding to that target data point;

[0036] If the hypervolume improvement rate corresponding to each target data is greater than or equal to the first preset value, and / or the number of processing times of the target population dataset is greater than or equal to the second preset value, then it is determined that the target population dataset can be used for multi-objective optimization.

[0037] If the hypervolume improvement rate corresponding to at least one target data is less than the first preset value, and / or the number of processing times for obtaining the target population dataset is less than the second preset value, then the target population dataset is determined to be unsuitable for multi-objective optimization. The target population dataset is then determined as the initial population dataset, and processing continues to obtain a new target dataset.

[0038] Secondly, embodiments of this application provide a data optimization apparatus, the apparatus comprising:

[0039] The acquisition module is used to acquire the initial population dataset, which includes at least one individual data point, and the individual data point includes at least one design parameter.

[0040] The processing module is used to perform reverse learning on the initial population dataset to obtain the first population dataset.

[0041] The processing module is also used to perform differential evolution processing and crossover processing on the first type of population dataset to obtain the second type of population dataset;

[0042] The processing module is also used to perform non-dominated sorting and clustering on the second population dataset to obtain the target population dataset, which is used for multi-objective optimization.

[0043] In one possible implementation, the processing module is specifically used for:

[0044] Perform non-dominated sorting on at least one experimental individual data in the second population dataset to obtain at least one parent data and at least one child data;

[0045] Determine the crowding distance of at least one parent data;

[0046] The third population dataset is determined based on the crowding distance. The third population dataset includes at least the K data points with the largest crowding distance among the parent data.

[0047] Clustering is performed on the third population dataset to obtain the target population dataset.

[0048] In one possible implementation, the processing module is specifically used for:

[0049] Gaussian perturbation is applied to the parent data in the third population dataset to generate perturbation data corresponding to each parent data.

[0050] The perturbation data corresponding to each parent generation data is optimized by neighborhood search to obtain the optimized data corresponding to each parent generation data.

[0051] Based on the data of each parent generation and the corresponding optimized data of each parent generation, at least one candidate data is determined;

[0052] Clustering is performed on at least one candidate data to obtain the target population dataset, which includes at least one category and the target data corresponding to each category.

[0053] In one possible implementation, the processing module is specifically used for:

[0054] For any parent data, if the parent data dominates the optimized data corresponding to the parent data, then the candidate data is determined to be the parent data;

[0055] If the optimized data corresponding to the parent data dominates the parent data, then the candidate data is determined to be the optimized data corresponding to the parent data;

[0056] If the parent data and the corresponding optimized data do not dominate each other, then the data with a crowding distance greater than or equal to the preset distance are determined as candidate data.

[0057] In one possible implementation, the processing module is specifically used for:

[0058] Determine the reverse individual data corresponding to each individual data point in the initial population dataset;

[0059] Get constraints and preset functions;

[0060] Based on the constraints and preset functions, at least one individual data and at least one reverse individual data are filtered to obtain a first type of group dataset. The first type of group dataset includes at least one first individual data, and the at least one first individual data includes at least one individual data and the first N data from at least one reverse individual data.

[0061] In one possible implementation, the processing module is specifically used for:

[0062] In the first type of population dataset, M first volume data are randomly selected, where M is 3;

[0063] Based on a preset scaling factor, differential evolution is performed on M first-body data to obtain at least one differential data.

[0064] Obtain the preset crossover probability;

[0065] Based on the preset crossover probability, the first population dataset, and at least one differential evolution data, a second population dataset is determined, which includes data from at least one experimental individual.

[0066] In one possible implementation, the processing module is further configured to:

[0067] For any target data point in the target population dataset, determine the hypervolume boost rate corresponding to that target data point;

[0068] If the hypervolume improvement rate corresponding to each target data is greater than or equal to the first preset value, and / or the number of processing times of the target population dataset is greater than or equal to the second preset value, then it is determined that the target population dataset can be used for multi-objective optimization.

[0069] If the hypervolume improvement rate corresponding to at least one target data is less than the first preset value, and / or the number of processing times for obtaining the target population dataset is less than the second preset value, then the target population dataset is determined to be unsuitable for multi-objective optimization. The target population dataset is then determined as the initial population dataset, and processing continues to obtain a new target dataset.

[0070] Thirdly, embodiments of this application provide a data optimization device, including:

[0071] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method described in any one of the first aspects.

[0072] Fourthly, embodiments of this application provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method described in any one of the first aspects.

[0073] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method described in any one of the first aspects.

[0074] The data optimization method, apparatus, device, and storage medium provided in this application embed differential evolution and reverse learning strategies into the NSGA-II framework, forming a collaborative mechanism for global search and local development. Differential evolution guides the generation of differential data based on individual data through differential vectors, improving search directionality. Reverse learning expands the search scope and enhances diversity by constructing reverse data. Through classification processing, it ensures that the output target population dataset contains multiple equivalent design schemes, meeting the engineering requirements for adaptability to multiple working conditions and design diversity. Attached Figure Description

[0075] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0076] Figure 1 A schematic diagram illustrating the application scenarios provided in the embodiments of this application;

[0077] Figure 2 A flowchart illustrating a data optimization method provided in an embodiment of this application;

[0078] Figure 3 A flowchart illustrating another data optimization method provided in an embodiment of this application;

[0079] Figure 4A schematic diagram of the data optimization device provided in the embodiments of this application;

[0080] Figure 5 A schematic diagram of the structure of the data optimization device provided in the embodiments of this application.

[0081] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0082] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0083] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0084] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0085] To facilitate understanding, the following will be combined with... Figure 1 The application scenarios applicable to the embodiments of this application will be described.

[0086] Figure 1 This is a schematic diagram illustrating an application scenario provided in an embodiment of this application. Please refer to [link / reference]. Figure 1This includes server 101. Server 101 is equipped with NSGA-II. The suspension system can be decoupled and optimized through NSGA-II in server 101 to simultaneously meet the vibration suppression requirements under different operating conditions (such as idling, acceleration, and cruising).

[0087] Among related technologies, NSGA-II achieves Pareto front search through non-dominated sorting, crowding calculation, and a fast convergence mechanism, which can handle multi-objective conflicts and generate a relatively uniformly distributed solution set. However, in suspended system optimization, NSGA-II suffers from problems such as low search efficiency, weak local search ability, insufficient diversity preservation, and poor adaptability to multiple operating conditions. Therefore, there is an urgent need for a dataset applicable to NSGA-II to address these issues.

[0088] In this embodiment, differential evolution and reverse learning strategies are embedded into the NSGA-II framework to form a collaborative mechanism between global search and local development. Differential evolution guides the generation of differential data based on individual data through differential vectors, improving search directionality. Reverse learning expands the search scope and enhances diversity by constructing reverse data. Through classification processing, it is ensured that the output target population dataset contains multiple equivalent design schemes, meeting the engineering requirements for adaptability to multiple working conditions and design diversity.

[0089] The method described in this application will now be illustrated through specific embodiments. It should be noted that the following embodiments may exist independently or in combination with each other; identical or similar content will not be repeated in different embodiments.

[0090] Figure 2 This is a flowchart illustrating a data optimization method provided in an embodiment of this application. Please refer to [link / reference]. Figure 2 The method may include:

[0091] S201. Obtain the initial population dataset.

[0092] The execution entity in this application embodiment can be a data optimization device or a data optimization apparatus installed within a data optimization device. The data optimization apparatus can be implemented through software or a combination of software and hardware. The data optimization device can be a terminal device, a server, etc. The terminal device can be a mobile phone, tablet computer, etc.

[0093] The initial population dataset includes at least one individual data point, and each individual data point includes at least one design parameter.

[0094] For example, design parameters can include suspension stiffness parameters, damping parameters, and placement location.

[0095] At least one individual data point can be randomly selected from at least one design variable to form the initial population dataset. The number of randomly selected individual data points can be a default setting or a user-defined number; this application does not impose any restrictions.

[0096] For example, at least one design variable includes data A1 to data A200, and 100 individual data points can be randomly selected from data A1 to data A200 to form the initial population dataset. Specifically, the initial population dataset includes data A1 to data A100.

[0097] S202. Perform reverse learning on the initial population dataset to obtain the first population dataset.

[0098] The initial population dataset can be processed by reverse learning to obtain the first population dataset in the following way: determine the reverse individual data corresponding to each individual data in the initial population dataset; obtain the constraints and preset functions; according to the constraints and preset functions, filter at least one individual data and at least one reverse individual data to obtain the first population dataset. The first population dataset includes at least one first individual data, and at least one first individual data includes at least one individual data and the first N data in at least one reverse individual data.

[0099] The inverse individual data corresponding to the individual data can be the inverse solution corresponding to the individual data.

[0100] Constraints can be set based on specific usage scenarios and the parameters of the suspension system. For example, constraints can be stiffness constraints, frequency constraints, etc.

[0101] The preset function can be a function generated based on the optimization objective. For example, the preset function can be a function based on the decoupling rate of the main direction.

[0102] For any individual data point, after processing it according to constraints and a preset function, the first function value corresponding to that individual data point is obtained. For any reverse individual data point, after processing it according to constraints and a preset function, the second function value corresponding to that individual data point is obtained. Based on the first and second function values, the individual data points and the number of reverse individuals are sorted in ascending order to obtain the first sort. The first N data points in the first sort constitute the first individual data point.

[0103] N is an integer greater than or equal to 1. N can be a default setting or a user-defined value; this application does not impose any restrictions. N can be set in advance and stored in the preset storage space of the data optimization device.

[0104] For example, as illustrated above, the initial population dataset is determined to include data A1 to A100. The data optimization device determines the reverse individual data corresponding to each individual data in the initial population dataset, obtaining at least one reverse individual data including data A200 to A300. The data optimization device acquires constraints and preset functions, and processes data A1 to A100 and data A200 to A300 according to the constraints and preset functions to obtain at least one first function value and at least one second function value.

[0105] Based on the first and second function values, the individual data and the reverse number of individuals are sorted in ascending order to obtain the first sort. In this first sort, the data optimization device determines the first 100 individual data points as data A1 to data A100. Therefore, the data optimization device determines that the first population dataset includes data A1 to data A100.

[0106] By using reverse learning, the solution space is quickly covered during the initialization phase, avoiding local convergence problems caused by uneven distribution of the initial population dataset.

[0107] S203. Perform differential evolution and crossover processing on the first group dataset to obtain the second group dataset.

[0108] The second population dataset can be obtained by performing differential evolution and crossover processing on the first population dataset in the following manner: In the first population dataset, M first individual data points are randomly selected, where M is 3; Differential evolution processing is performed on the M first individual data points according to a preset scaling factor to obtain at least one differential data point; A preset crossover probability is obtained; The second population dataset is determined based on the preset crossover probability, the first population dataset, and at least one differentially evolved data point, and the second population dataset includes at least one experimental individual data point.

[0109] The preset scaling factor can be between 0 and 1. It can be a default setting or a user-defined setting; this application does not impose any restrictions. The preset scaling factor can be set in advance and stored in the preset storage space of the data optimization device.

[0110] The number of at least one difference data point is the same as the number of the first volume data point in the first population dataset.

[0111] The preset crossover probability can be a default setting or a user-defined one; this application does not impose any restrictions. The preset crossover probability can be set in advance and stored in the preset storage space of the data optimization device.

[0112] At least one experimental individual data can be generated from at least one difference data and at least one first individual data according to a preset crossover probability.

[0113] For example, as illustrated above, the first type of cluster dataset is determined to include data A1 to data A100. The data optimization device randomly selects three first-order data points from data A1 to data A100, including data A1, data A13, and data A30. The data optimization device obtains a preset scaling factor of 0.6 from a preset storage space and performs differential evolution processing on data A1, data A13, and data A30 according to the preset scaling factor to obtain differential data 1. The differential processing process is repeated to obtain 100 differential data points, including differential data 1 to differential data 100.

[0114] The data optimization device obtains a preset crossover probability of 0.8 in a preset storage space, and determines a second population dataset based on the preset crossover probability, data A1 to data A100 and at least one differential evolution data. The second population dataset includes experimental data A1 to experimental data A100.

[0115] Optionally, if the data of the test individuals exceeds the set range, the test individual data can be updated by boundary reflection or truncation to bring the test individual data within the set range.

[0116] Differential evolution guides the generation of differential data, leveraging information from superior solutions to enhance search directionality. The synergistic effect of scaling factors and crossover probabilities ensures a balance between exploration and exploitation, thereby accelerating convergence and avoiding the inefficiencies of random search.

[0117] S204. Perform non-dominated sorting and clustering on the second population dataset to obtain the target population dataset.

[0118] The target population dataset is used for multi-objective optimization.

[0119] The target population dataset can be obtained by performing non-dominated sorting and clustering on the second population dataset as follows: Perform non-dominated sorting on the data of at least one experimental individual in the second population dataset to obtain at least one parent data and at least one child data; determine the crowding distance of at least one parent data; determine the third population dataset based on the crowding distance, the third population dataset including the K data with the largest crowding distance among at least one parent data; perform clustering on the third population dataset to obtain the target population dataset.

[0120] Non-dominated non-processing can divide the data of at least one experimental individual into different levels (frontiers), thus providing a basis for multi-objective optimization algorithms to select excellent individuals.

[0121] The parent data consists of experimental individuals in the second population dataset whose corresponding rank is greater than or equal to a preset rank. The offspring data consists of experimental individuals in the second population dataset whose corresponding rank is less than a preset rank.

[0122] Crowding distance measures the distribution density of a solution relative to its surrounding solutions in the target space. Given equal quality, solutions with less "crowded" surroundings are preferred, thus promoting a more uniform distribution of the solution set on the Pareto front. Therefore, when determining the third population dataset, at least the K data points with the largest crowding distance from at least one parent are selected.

[0123] K is an integer greater than or equal to 1. K can be a default value or a user-defined value; this application does not impose any restrictions. K can be set in advance and stored in the preset storage space of the data optimization device.

[0124] After obtaining the third population dataset, Gaussian perturbation can be applied to each parent data in the third population dataset. In this way, the randomness of Gaussian perturbation can ensure that the parent data retains its exploratory ability when used for local search in multi-objective optimization algorithms.

[0125] For example, as illustrated above, the second population dataset is determined to include experimental data A1 to A100. The data optimization device performs a non-dominated sorting process on experimental data A1 to A100, obtaining at least one parent dataset including experimental data A1 to A60, and at least one child dataset including experimental data A61 to A100. The data optimization device determines the crowding distance between experimental data A1 to A60, and based on this distance, identifies the 50 datasets with the largest crowding distances as including experimental data A11 to A60. Therefore, the data optimization device determines that the third population dataset includes experimental data A11 to A60. The data optimization device performs clustering on experimental data A11 to A60, obtaining the target population dataset as dataset A.

[0126] The target population dataset can serve as input data for multi-objective optimization algorithms. Because the target data in the dataset exhibit diversity, stability, and uniform distribution, it can meet design requirements under different operating conditions, providing engineers with diverse candidate solutions, shortening the design cycle, and improving the overall performance of the suspension system. For example, the multi-objective optimization algorithm could be NSGA-II.

[0127] The data optimization method provided in this application expands the search range through a reverse learning strategy, avoids data centralization, and ensures the generation of multimodal data. The differential evolution operator utilizes superior solution information to reduce inefficient iterations of random search and accelerate algorithm convergence. A local perturbation mechanism refines the parent data, improving the smoothness of the Pareto front and the stability of the solution. Through multi-condition constraint processing and multimodal solution set maintenance, the output target population dataset can meet the design requirements under different conditions, providing engineers with diverse candidate solutions, shortening the design cycle, and improving the overall performance of the suspension system.

[0128] Based on any of the above embodiments, the following, in conjunction with Figure 3 The detailed process of data optimization methods is explained.

[0129] Figure 3 This is a flowchart illustrating another data optimization method provided in an embodiment of this application. Please refer to... Figure 3 The method includes:

[0130] S301. Obtain the initial population dataset.

[0131] A coupled dynamic model of the powertrain-mount-body system can be established in advance, taking into account excitations under different operating conditions, and the optimization objective can be determined.

[0132] Based on the modeling method of the engine powertrain mounting system, the electric motor powertrain mounting system is simplified into an undamped multi-degree-of-freedom vibration system consisting of a rigid body, springs, and a rigid ground. The general vibration differential equation and the free vibration differential equation of this system can be specifically represented by the following formula 1:

[0133]

[0134] in, Mass matrix of powertrain mounting system; The damping matrix of the powertrain mounting system; This represents the stiffness matrix of the powertrain mounting system. Let be the displacement column vector of the powertrain's center of mass.

[0135] By obtaining the eigenvalues ​​and eigenvectors from Equation 1, the natural modal frequencies and configuration of the powertrain mounts can be obtained, as shown in Equation 2 below:

[0136]

[0137]

[0138] in, The natural angular frequency of the system; This represents the system's mode shape. See above for explanations of other parameters.

[0139] Optimization objectives may include .in, To maximize the decoupling rate (reduce the coupling between degrees of freedom); To minimize vibration transmissibility (to improve comfort); This refers to the structural strength and stiffness satisfaction under constraints.

[0140] Individual data can serve as design parameters. The dynamic characteristics of the powertrain mounting system are closely related to the parameters of both the powertrain and the mounting system. Since parameters such as the powertrain's mass and moment of inertia are difficult to change, when performing decoupling optimization of the mounting system, it is usually considered to optimize the mounting parameters, such as the mounting position, mounting angle, and stiffness in all directions. The model selects the triaxial stiffness of the mounting point as the design parameter.

[0141] S302. Perform reverse learning processing on the initial population dataset to obtain the first population dataset.

[0142] For the Individual data The first one can be determined using the following formula 3. The reverse individual data corresponding to each individual data point:

[0143]

[0144] in, The total number of design parameters in individual data; For the first The upper boundary of individual data; For the first The lower boundary of individual data; For the first The first design parameter in each individual data set; For the first The second design parameter in the individual data; For the first In the data of each individual, the first Design parameters.

[0145] The upper and lower boundaries of individual data can be determined based on the range set for individual data.

[0146] After determining the reverse data, at least one individual data and at least one reverse individual data can be filtered according to the constraints and the preset function to obtain the first type of cluster dataset.

[0147] Constraints can include stiffness constraints and frequency constraints. Stiffness constraints can specifically constrain the principal stiffness of each suspension element. ( Representing the One suspension. (These represent the three directions of the suspension spindle, respectively), with the suspension stiffness as a constraint condition, and the initial value of the stiffness constraint range. .

[0148] Frequency constraints can be defined as requiring the natural frequency of the powertrain mounting system to be no less than 13Hz and the frequency intervals to differ by more than 2Hz.

[0149] The preset function can be a function corresponding to the energy decoupling rate. The preset function can be determined using the following formula 4:

[0150]

[0151]

[0152] in, In the expression, it is called the first The first mode in the second mode Directional energy decoupling; For the first The first mode shape Displacement; For the first The first mode shape Displacement; For mass array The Line 1 Column elements; The function corresponding to the decoupling rate in the main direction (preset function).

[0153] Reverse learning expands the search scope by constructing reverse data, ensuring a wider distribution of the initial population in the solution space. The generation of reverse data is based on upper and lower bounds of the design variables. By merging forward and reverse solutions and selecting high-quality individuals, population diversity is significantly improved. This strategy quickly covers the search space in the early stages of the algorithm, avoiding the local convergence problem caused by random initialization in traditional NSGA-II. Ultimately, the reverse learning strategy enhances the algorithm's global exploration capability, ensuring that the output Pareto solution set contains multiple approximately equivalent design schemes, meeting the multi-solution requirements of the suspension system under different operating conditions. This provides engineers with richer candidate options, significantly improving design flexibility and engineering practicality.

[0154] S303. Perform differential evolution and crossover processing on the first group dataset to obtain the second group dataset.

[0155] The differential evolution process can be performed on the first group dataset using the following formula 5 to obtain differential data:

[0156]

[0157] in, The data is differential; The first three data points are randomly selected. For the preset scaling factor, .

[0158] In determining the difference data Then, based on the preset crossover probability ,from and Data for individual participants in the experiment is generated based on a preset crossover probability. From the first The difference data and the first Generate the first body data from the first body data. Data from individual trials.

[0159] In this embodiment, differential evolution processing can generate a more directional offspring population, thereby improving global search efficiency.

[0160] Optionally, a dynamic scaling factor (F) and crossover probability (CR) adjustment mechanism can be introduced into the differential evolution operator to adaptively adjust the parameter values ​​according to the current iteration stage. For example, in the early stages of the algorithm, F and CR are set to higher values ​​to enhance global exploration capabilities. In the later stages of iteration, F and CR are gradually reduced to improve local exploration accuracy.

[0161] Specifically, the stage of the algorithm (exploration / development phase) is determined by the number of iterations or the convergence status, and the range of values ​​for F and CR is dynamically adjusted. In the early stages of the algorithm, high F and CR values ​​encourage the difference vector to generate mutated solutions with larger step sizes, covering a wider search space. In the later stages, low F and CR values ​​reduce randomness and focus on a refined search of the neighborhood of elite solutions.

[0162] Dynamic parameter adjustment enables the algorithm to match the optimal search strategy at different stages, avoiding the problems of underexploration or overexploration caused by traditional fixed parameters. Through staged parameter optimization, the algorithm can both quickly cover the solution space and efficiently develop high-quality solution neighborhoods, significantly improving the smoothness and distribution balance of the Pareto front.

[0163] S304. Perform non-dominated sorting on at least one experimental individual data in the second population dataset to obtain at least one parent data and at least one child data.

[0164] Non-dominated sorting was performed on at least one experimental individual data in the second population dataset to obtain parent data with better diversity and stability.

[0165] S305. Determine the crowding distance of at least one parent data.

[0166] For the Parent data, Take values ​​1, 2, ..., N-2 in sequence. Determine the... Parent data in the current optimization objective The previous and next parent data are used. The normalized difference between the previous and next parent data on the objective function value is calculated, and this difference is added to the crowding degree of the individual.

[0167] The number can be determined using the following formula 6. Crowding distance of parent data:

[0168]

[0169] in, For the first Crowding distance of parent data; For the first Parent data in the current optimization objective The previous parent data; For the first Parent data in the current optimization objective The next parent data; For at least one parent data in the current target The maximum value on; For at least one parent data in the current target The minimum value on.

[0170] Used for normalization, thereby eliminating the influence of different target dimensions.

[0171] S306. Determine the third group dataset based on the crowding distance.

[0172] The third type of dataset includes at least one set of K data points with the largest crowding distance from the parent data.

[0173] S307. Perform clustering on the third population dataset to obtain the target population dataset.

[0174] The target population dataset can be obtained by clustering the third population dataset as follows: Gaussian perturbation is applied to the parent data in the third population dataset to generate perturbation data corresponding to each parent data; neighborhood search is used to optimize the perturbation data corresponding to each parent data to obtain optimized data corresponding to each parent data; based on each parent data and the optimized data corresponding to each parent data, at least one candidate data is determined; clustering is performed on at least one candidate data to obtain the target population dataset, which includes at least one category and target data corresponding to each category.

[0175] The following formula 7 can be used to calculate the... The parent data is processed with Gaussian perturbation to obtain the first... Perturbation data corresponding to each parent data:

[0176]

[0177] in, For the first The perturbation data corresponding to each parent generation data; For the first Parent data; It is a standard Gaussian noise source.

[0178] The local perturbation mechanism avoids the algorithm from getting trapped in local optima through random Gaussian perturbations, while combining it with directional neighborhood search to gradually approach the optimal solution. The randomness of the Gaussian perturbation ensures that the algorithm maintains its exploratory ability in local searches, while the step size of the neighborhood search is adaptively adjusted iteratively (e.g., gradually decreasing with the number of iterations), focusing on the refined development of elite solution neighborhoods in later stages. This strategy, through the synergistic effect of "random exploration + directional development," significantly improves the convergence and smoothness of the solution set, avoiding the problem of stagnation in local regions in traditional NSGA-II. Ultimately, this technique optimizes the quality of the Pareto front, making the solution set more evenly distributed in the target space, providing engineers with more stable candidate design schemes, especially demonstrating stronger robustness in the multi-condition adaptive optimization of suspension systems.

[0179] Optionally, a hybrid perturbation strategy can be introduced into the local perturbation mechanism, combining Gaussian perturbation with neighborhood search (such as differential mutation or hill climbing search). Gaussian perturbation (random perturbation) and neighborhood search (directional perturbation) are applied simultaneously to the parent data, and the solution with better performance after perturbation is selected and retained.

[0180] The hybrid perturbation strategy balances randomness and directionality, avoiding the limitations of a single perturbation strategy and significantly optimizing the neighborhood search capability of elite solutions. Through a selective retention mechanism, the perturbed solutions are ensured to be more concentrated in the target space, improving the convergence and stability of the Pareto front.

[0181] At least one candidate data can be determined based on each parent data and the corresponding optimized data in the following way: For any parent data, if the parent data dominates the optimized data corresponding to the parent data, then the candidate data is determined to be the parent data; if the optimized data corresponding to the parent data dominates the parent data, then the candidate data is determined to be the optimized data corresponding to the parent data; if the parent data and the optimized data corresponding to the parent data do not dominate each other, then the data with a congestion distance greater than or equal to a preset distance is determined to be the candidate data.

[0182] If the parent data is at least as good as the corresponding optimized data across all optimization objectives, then the parent data dominates the corresponding optimized data. If the corresponding optimized data is at least as good as the parent data across all optimization objectives, then the corresponding optimized data dominates the parent data. If the parent data is nearly identical to the corresponding optimized data across all optimization objectives, then the parent data and the corresponding optimized data do not dominate each other.

[0183] The crowding distance of the optimized data corresponding to the parent data can be determined according to Formula 6 above.

[0184] If the parent data and the corresponding optimized data do not dominate each other, determining the candidate data based on the crowding distance can avoid excessive crowding of the frontier.

[0185] The preset distance can be a default setting or a user-defined setting; this application does not impose any restrictions. The preset distance can be set in advance and stored in the preset storage space of the data optimization device.

[0186] For example, suppose the third population dataset includes parent data 1 to parent data 50. The data optimization device performs Gaussian perturbation on parent data 1 to parent data 50 using Formula 7 above, obtaining perturbed data 1 to perturbed data 50. The data optimization device then optimizes the perturbed data corresponding to each parent data through neighborhood search, obtaining optimized data corresponding to each parent data, including optimized data 1 to optimized data 50.

[0187] Given parent data 1, the corresponding optimized data is determined to be optimized data 1. If the data optimization device determines that parent data 1 is at least as good as optimized data 1 across all optimization objectives, then parent data 1 dominates optimized data 1. Therefore, the data optimization device determines the candidate data as parent data 1.

[0188] Given parent data 2, the corresponding optimized data is determined to be optimized data 2. If the data optimization device determines that optimized data 2 is at least as good as parent data 2 across all optimization objectives, then optimized data 2 dominates parent data 2. Therefore, the data optimization device determines the candidate data as optimized data 2.

[0189] For parent data 3, the corresponding optimized data is determined to be optimized data 3. The data optimization device determines that parent data 3 is nearly identical to optimized data 3 in all optimization objectives, thus confirming that parent data 3 and optimized data 3 are not mutually exclusive. Therefore, the data optimization device determines the congestion distance between parent data 3 and optimized data 3. The data optimization device determines that the congestion distance of parent data 3 is greater than a preset distance, while the congestion distance of optimized data 3 is less than a preset distance. The data optimization device then determines parent data 3 as a candidate data. Through the above method, the data optimization device identifies at least one candidate data.

[0190] Clustering algorithms can be used to cluster at least one candidate dataset to obtain the target population dataset. Clustering algorithms can include K-means clustering, density-based spatial clustering of applications with noise (DBSCAN), and others.

[0191] Optionally, crowding or minimum distance thresholds can be combined to maintain a uniform distribution of the target population dataset in the target space.

[0192] A balanced distribution ensures that the final Pareto optimal solution set output by the algorithm not only satisfies convergence (approaching the true Pareto front) but also possesses good distribution (uniformly distributed along the Pareto front). Specifically:

[0193] (1) Prevent the aggregation of unsets and avoid information redundancy.

[0194] (2) Ensure comprehensive coverage and clearly demonstrate trade-offs. The target population dataset cannot fully depict the competitive and attrition-based trade-offs among various objectives. The diversity of target data in the target population dataset is very poor, and it is impossible to see the complete spectrum from "extreme pursuit of objective A" to "extreme pursuit of objective B". The uniformly distributed solution set clearly demonstrates the shape of the entire Pareto front.

[0195] (3) The uniformly distributed target population dataset provides a series of representative and differentiated candidate schemes. It is easy to locate the interval of interest according to preferences and make fine selections from the uniformly distributed target data around the interval, without having to get bogged down in a large number of similar target data.

[0196] A dynamic clustering strategy can be introduced to dynamically adjust the number of clusters (K value) based on the distribution density of the target population dataset. For example, when the target population dataset is sparsely distributed, the K value can be increased to retain more representative solutions; when the target population dataset is concentrated, the K value can be decreased to avoid excessive fragmentation.

[0197] Specifically, the number of clusters is dynamically determined by calculating the density distribution of the target population dataset in the target space (e.g., density estimation based on K-nearest neighbors). In high-density regions, the K value is reduced, and similar solutions are merged to avoid redundancy. In low-density regions, the K value is increased to retain more representative solutions and maintain diversity.

[0198] The dynamic K-value strategy adapts to changes in the solution set distribution, avoiding over-clustering or omission of sparse regions caused by a fixed K-value, and significantly improving the coverage of the multimodal solution set. The dynamically adjusted clustering strategy ensures that the output Pareto solution set has sufficient design options under different operating conditions, meeting the needs of engineers for diverse solutions.

[0199] Optionally, after obtaining the target population dataset, it can also be determined whether the target population dataset can be used for multi-objective optimization.

[0200] The following method can be used to determine whether a target population dataset can be used for multi-objective optimization: For any target data in the target population dataset, determine the hypervolume improvement rate corresponding to the target data; if the hypervolume improvement rate corresponding to each target data is greater than or equal to a first preset value, and / or the number of processing steps for the target population dataset is greater than or equal to a second preset value, then the target population dataset can be used for multi-objective optimization; if there is at least one target data whose hypervolume improvement rate is less than the first preset value, and / or the number of processing steps for the target population dataset is less than the second preset value, then the target population dataset cannot be used for multi-objective optimization, and the target population dataset is determined as the initial population dataset, and processing continues to obtain a new target dataset.

[0201] The supervolume lift rate corresponding to the target data can be determined using the following formula 8:

[0202]

[0203] in, For the first The population of the first generation relative to the second generation Pareto front improvement of the generation population (percentage); For the first The hypervolume value of the Pareto front of the generation population; For the first The hypervolume value of the Pareto front of the generation population.

[0204] The first and second preset values ​​can be set by default or customized by the user; this application does not impose any restrictions. The first and second preset values ​​can be set in advance and stored in the preset storage space of the data optimization device.

[0205] If the hypervolume improvement rate corresponding to at least one target data is less than the first preset value, and / or the number of times the target population dataset is processed is less than the second preset value, then the target population dataset is determined as the initial population dataset, and a new target dataset is obtained by iteratively repeating the method provided in the embodiments of this application until the currently generated target dataset can be used for multi-objective optimization.

[0206] Figure 4 This is a schematic diagram of the data optimization device provided in an embodiment of this application. Please refer to... Figure 4 The data optimization device 10 may include:

[0207] Module 11 is used to acquire an initial population dataset, which includes at least one individual data point and the individual data point includes at least one design parameter.

[0208] Processing module 12 is used to perform reverse learning processing on the initial population dataset to obtain the first population dataset;

[0209] Processing module 12 is also used to perform differential evolution processing and crossover processing on the first group dataset to obtain the second group dataset;

[0210] Processing module 12 is also used to perform non-dominated sorting and clustering processing on the second population dataset to obtain the target population dataset, which is used for multi-objective optimization.

[0211] In one possible implementation, the processing module 12 is specifically used for:

[0212] Perform non-dominated sorting on at least one experimental individual data in the second population dataset to obtain at least one parent data and at least one child data;

[0213] Determine the crowding distance of at least one parent data;

[0214] The third population dataset is determined based on the crowding distance. The third population dataset includes at least the K data points with the largest crowding distance among the parent data.

[0215] Clustering is performed on the third population dataset to obtain the target population dataset.

[0216] In one possible implementation, the processing module 12 is specifically used for:

[0217] Gaussian perturbation is applied to the parent data in the third population dataset to generate perturbation data corresponding to each parent data.

[0218] The perturbation data corresponding to each parent generation data is optimized by neighborhood search to obtain the optimized data corresponding to each parent generation data.

[0219] Based on the data of each parent generation and the corresponding optimized data of each parent generation, at least one candidate data is determined;

[0220] Clustering is performed on at least one candidate data to obtain the target population dataset, which includes at least one category and the target data corresponding to each category.

[0221] In one possible implementation, the processing module 12 is specifically used for:

[0222] For any parent data, if the parent data dominates the optimized data corresponding to the parent data, then the candidate data is determined to be the parent data;

[0223] If the optimized data corresponding to the parent data dominates the parent data, then the candidate data is determined to be the optimized data corresponding to the parent data;

[0224] If the parent data and the corresponding optimized data do not dominate each other, then the data with a crowding distance greater than or equal to the preset distance are determined as candidate data.

[0225] In one possible implementation, the processing module 12 is specifically used for:

[0226] Determine the reverse individual data corresponding to each individual data point in the initial population dataset;

[0227] Get constraints and preset functions;

[0228] Based on the constraints and preset functions, at least one individual data and at least one reverse individual data are filtered to obtain a first type of group dataset. The first type of group dataset includes at least one first individual data, and the at least one first individual data includes at least one individual data and the first N data from at least one reverse individual data.

[0229] In one possible implementation, the processing module 12 is specifically used for:

[0230] In the first type of population dataset, M first volume data are randomly selected, where M is 3;

[0231] Based on a preset scaling factor, differential evolution is performed on M first-body data to obtain at least one differential data.

[0232] Obtain the preset crossover probability;

[0233] Based on the preset crossover probability, the first population dataset, and at least one differential evolution data, a second population dataset is determined, which includes data from at least one experimental individual.

[0234] In one possible implementation, the processing module 12 is further configured to:

[0235] For any target data point in the target population dataset, determine the hypervolume boost rate corresponding to that target data point;

[0236] If the hypervolume improvement rate corresponding to each target data is greater than or equal to the first preset value, and / or the number of processing times of the target population dataset is greater than or equal to the second preset value, then it is determined that the target population dataset can be used for multi-objective optimization.

[0237] If the hypervolume improvement rate corresponding to at least one target data is less than the first preset value, and / or the number of processing times for obtaining the target population dataset is less than the second preset value, then the target population dataset is determined to be unsuitable for multi-objective optimization. The target population dataset is then determined as the initial population dataset, and processing continues to obtain a new target dataset.

[0238] Figure 5 A schematic diagram of the data optimization device provided in an embodiment of this application. Please refer to... Figure 5 The data optimization device 20 may include a memory 21 and a processor 22. Exemplarily, the memory 21 and the processor 22 are interconnected via a bus 23.

[0239] Memory 21 is used to store program instructions;

[0240] The processor 22 is used to execute the program instructions stored in the memory, so that the data optimization device 20 performs the method shown in the above method embodiment.

[0241] The data optimization device provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0242] This application provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the above-described method when executed by a processor.

[0243] This application embodiment may also provide a computer program product, including a computer program that, when executed by a processor, can implement the above-described method.

[0244] All or part of the steps in the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above-described method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), random access memory (RAM), flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disc, and any combination thereof.

[0245] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0246] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0247] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0248] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

[0249] In this application, the term "comprising" and its variations can refer to non-limiting inclusion; the term "or" and its variations can refer to "and / or". The terms "first", "second", etc., in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

Claims

1. A data optimization method, characterized in that, include: Obtain an initial population dataset, wherein the initial population dataset includes at least one individual data point, and the individual data point includes at least one design parameter; The design parameters include suspension parameters; The initial population dataset is subjected to reverse learning processing to obtain the first population dataset; Differential evolution and crossover processes are performed on the first population dataset to obtain the second population dataset; The second population dataset is subjected to non-dominated sorting to determine the third population dataset; the third population dataset includes at least one parent generation. Gaussian perturbation is applied to each parent generation data in the third population dataset to generate perturbation data corresponding to each parent generation data; The perturbation data corresponding to each parent generation data is optimized by neighborhood search to obtain the optimized data corresponding to each parent generation data. Based on the data of each parent generation and the corresponding optimized data of each parent generation, at least one candidate data is determined; Clustering is performed on the at least one candidate data to obtain a target population dataset, which includes at least one category and target data corresponding to each category; the target population dataset is used for multi-objective optimization of the suspension system; The suspension system is an undamped multi-degree-of-freedom vibration system composed of a rigid body, springs, and a rigid ground. The general vibration differential equation and the free vibration differential equation of the undamped multi-degree-of-freedom vibration system are as follows: in, Mass matrix of powertrain mounting system; The damping matrix of the powertrain mounting system; This represents the stiffness matrix of the powertrain mounting system. Let be the displacement column vector of the powertrain's center of mass.

2. The method according to claim 1, characterized in that, The second group dataset is subjected to non-dominated sorting to determine the third group dataset, including: Perform non-dominated sorting on at least one experimental individual data in the second population dataset to obtain at least one parent data and at least one child data; Determine the crowding distance of the at least one parent data; The third group dataset is determined based on the crowding distance.

3. The method according to claim 2, characterized in that, Based on the data from each parent generation and the corresponding optimized data, at least one candidate data point is determined, including: For any parent data, if the parent data dominates the optimized data corresponding to the parent data, then the candidate data is determined to be the parent data; If the optimized data corresponding to the parent data dominates the parent data, then the candidate data is determined to be the optimized data corresponding to the parent data; If the parent data and the corresponding optimized data do not dominate each other, then the data whose congestion distance is greater than or equal to the preset distance are determined as candidate data.

4. The method according to claim 1, characterized in that, The initial population dataset is subjected to reverse learning processing to obtain the first population dataset, which includes: Determine the reverse individual data corresponding to each individual data in the initial population dataset; Get constraints and preset functions; Based on the constraints and the preset function, at least one individual data and at least one reverse individual data are filtered to obtain a first type of population dataset. The first type of population dataset includes at least one first individual data, and the at least one first individual data includes the at least one individual data and the first N data from the at least one reverse individual data.

5. The method according to claim 1, characterized in that, Differential evolution and crossover processes are performed on the first population dataset to obtain the second population dataset, which includes: In the first population dataset, M first-body data points are randomly selected, where M is 3; According to a preset scaling factor, differential evolution processing is performed on the M first volume data to obtain at least one differential data. Obtain the preset crossover probability; The second population dataset is determined based on the preset crossover probability, the first population dataset, and the at least one differential evolution data. The second population dataset includes data from at least one experimental individual.

6. The method according to any one of claims 1-5, characterized in that, The method further includes: For any target data in the target population dataset, determine the hypervolume improvement rate corresponding to the target data; the hypervolume improvement rate is the Pareto front improvement of the t-th generation population relative to the (t-1)-th generation population. If the supervolume improvement rate corresponding to each target data is greater than or equal to the first preset value, and / or the number of processing times of obtaining the target population dataset is greater than or equal to the second preset value, then it is determined that the target population dataset can be used for multi-objective optimization. If the hypervolume improvement rate corresponding to at least one target data is less than a first preset value, and / or the number of processing steps to obtain the target population dataset is less than a second preset value, then the target population dataset is determined to be unusable for multi-objective optimization, the target population dataset is determined as the initial population dataset, and processing continues to obtain a new target dataset.

7. A data optimization device, characterized in that, The device includes: An acquisition module is used to acquire an initial population dataset, the initial population dataset including at least one individual data, the individual data including at least one design parameter; the design parameter includes a suspension parameter. The processing module is used to perform reverse learning processing on the initial population dataset to obtain the first population dataset; The processing module is further configured to perform differential evolution processing and crossover processing on the first population dataset to obtain a second population dataset. The processing module is further configured to: perform non-dominated sorting on the second population dataset to determine a third population dataset; the third population dataset includes at least one parent generation; perform Gaussian perturbation on each parent generation in the third population dataset to generate perturbation data corresponding to each parent generation; optimize the perturbation data corresponding to each parent generation through neighborhood search to obtain optimized data corresponding to each parent generation; determine at least one candidate data based on each parent generation and the optimized data corresponding to each parent generation; perform clustering on the at least one candidate data to obtain a target population dataset, the target population dataset including at least one category and target data corresponding to each category; the target population dataset is used for multi-objective optimization of the suspension system. The suspension system is an undamped multi-degree-of-freedom vibration system composed of a rigid body, springs, and a rigid ground. The general vibration differential equation and the free vibration differential equation of the undamped multi-degree-of-freedom vibration system are as follows: in, Mass matrix of powertrain mounting system; The damping matrix of the powertrain mounting system; This represents the stiffness matrix of the powertrain mounting system. Let be the displacement column vector of the powertrain's center of mass.

8. A data optimization device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, in, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 6.

Images