Multi-objective optimization design method based on machine learning and related equipment
By establishing a 3D model, adjusting the hyperparameter combination, and iterative optimization, the problem of low efficiency in machine learning hyperparameter tuning was solved, and more efficient and accurate multi-objective optimization design was achieved.
Patent Information
- Application Number
- CN202510910191.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-02
- Publication Date
- 2025-11-25
AI Technical Summary
Machine learning hyperparameter tuning is inefficient, relying on domain experience and trial and error, resulting in high training costs and low accuracy for machine learning models.
By establishing multiple 3D models to be optimized, obtaining sample datasets using parametric modeling methods, setting different combinations of hyperparameters for machine learning surrogate models, calculating the error between predicted values and actual simulation values, adjusting the hyperparameter combinations, and using multi-objective optimization algorithms to find the optimal combination, the 3D models are iteratively optimized.
It improves the efficiency of hyperparameter tuning, reduces training costs, and enhances the accuracy of machine learning models and the precision of multi-objective optimization design.
Smart Images
Figure CN121009769A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of machine learning technology, and in particular to a multi-objective optimization design method and related equipment based on machine learning. Background Technology
[0002] Data-driven multi-objective optimization is a method that combines data science with multi-objective optimization techniques. It aims to efficiently solve complex optimization problems involving multiple conflicting objectives by utilizing historical, experimental, or real-time data. By replacing computationally expensive objective functions or constraints with machine learning models and combining them with optimization algorithms, multi-objective optimization design can be achieved.
[0003] Hyperparameter tuning in machine learning is a crucial step in building high-performance models. Machine learning hyperparameters refer to parameters that need to be manually set before model training, such as learning rate, tree depth, batch size, etc. However, manual parameter tuning often relies on domain experience and trial and error, which is relatively inefficient. Summary of the Invention
[0004] The main objective of this application is to propose a multi-objective optimization design method and related equipment based on machine learning, so as to improve the efficiency of hyperparameter tuning and the accuracy of multi-objective optimization design.
[0005] To achieve the above objectives, one aspect of this application proposes a multi-objective optimization design method based on machine learning, the method comprising the following steps:
[0006] Establish three-dimensional models with multiple objectives to be optimized;
[0007] The sample dataset is obtained based on the three-dimensional model using a parametric modeling method.
[0008] Set different combinations of hyperparameters for machine learning agent models;
[0009] The machine learning agent model is fed with uniform and fixed input parameters to obtain the predicted values output by the machine learning agent model under different combinations of the hyperparameters.
[0010] Determine the error value between the predicted value and the actual simulation value under the unified and fixed input parameters;
[0011] The hyperparameters of the corresponding combination are adjusted according to the error value to obtain the adjusted hyperparameter combination;
[0012] The adjusted hyperparameter combination is optimized to obtain the optimal hyperparameter combination;
[0013] The 3D model is iteratively optimized based on the machine learning proxy model with the optimal hyperparameter combination and the sample dataset to obtain the optimized 3D model.
[0014] In some embodiments, establishing multiple target 3D models to be optimized includes the following steps:
[0015] A CAD model is designed based on the shape and size of the workpiece to be modeled, and the CAD model is imported into CAE software for settings to obtain the three-dimensional model with multiple objectives to be optimized; wherein, each objective to be optimized is a set of objective response values under a set working condition.
[0016] In some embodiments, obtaining the sample dataset based on the 3D model using the parametric modeling method includes the following steps:
[0017] The sample dataset is obtained from the three-dimensional model using experimental methods or the finite element method.
[0018] In some embodiments, setting different combinations of hyperparameters for the machine learning agent model includes the following steps:
[0019] Different combinations of hyperparameters are set for all the hyperparameters of the machine learning agent model; wherein the setting parameters of each combination include the range of change of the hyperparameters and whether they are enabled or disabled.
[0020] In some embodiments, adjusting the hyperparameters of the corresponding combination based on the error value to obtain the adjusted hyperparameter combination includes the following steps:
[0021] Analyze the contribution of the hyperparameters to the error value under each combination;
[0022] The hyperparameters under the corresponding combination are adjusted according to the contribution level to obtain the adjusted hyperparameter combination.
[0023] In some embodiments, adjusting the hyperparameters for each combination based on the contribution to obtain the adjusted hyperparameter combination includes the following steps:
[0024] The hyperparameters of combinations whose contribution is greater than the first error contribution threshold are deleted.
[0025] The hyperparameters for combinations whose contribution is less than the second error contribution threshold are retained.
[0026] Wherein, the first error contribution threshold is greater than the second error contribution threshold.
[0027] In some embodiments, optimizing the adjusted hyperparameter combination to obtain the optimal hyperparameter combination includes the following steps:
[0028] The adjusted hyperparameter combination is optimized using a multi-objective optimization algorithm to obtain the optimal hyperparameter combination.
[0029] To achieve the above objectives, another aspect of this application proposes a multi-objective optimization design apparatus based on machine learning, the apparatus comprising:
[0030] The model building unit is used to build three-dimensional models of multiple targets to be optimized.
[0031] The sample acquisition unit is used to obtain a sample dataset based on the three-dimensional model using a parametric modeling method.
[0032] The hyperparameter setting unit is used to set different combinations of hyperparameters for the machine learning agent model;
[0033] An output prediction unit is used to input uniform and fixed input parameters into the machine learning agent model and obtain the predicted values output by the machine learning agent model under different combinations of the hyperparameters.
[0034] An error determination unit is used to determine the error value between the predicted value and the actual simulation value under the unified and fixed input parameters;
[0035] The hyperparameter adjustment unit is used to adjust the hyperparameters of the corresponding combination according to the error value to obtain the adjusted hyperparameter combination;
[0036] The hyperparameter optimization unit is used to optimize the adjusted hyperparameter combination to obtain the optimal hyperparameter combination.
[0037] The model optimization unit is used to iteratively optimize the 3D model based on the machine learning proxy model set with the optimal hyperparameter combination and the sample dataset to obtain the optimized 3D model.
[0038] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.
[0039] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.
[0040] The embodiments of this application include at least the following beneficial effects:
[0041] This application provides a multi-objective optimization design method and related equipment based on machine learning. The method involves: establishing multiple 3D models with different objectives to be optimized; obtaining a sample dataset from the 3D models using a parametric modeling method; setting different combinations of hyperparameters for a machine learning surrogate model; inputting uniform and fixed input parameters into the machine learning surrogate model to obtain predicted values output by the model under different combinations of hyperparameters; determining the error between the predicted values and the actual simulation values under the uniform and fixed input parameters; adjusting the hyperparameters of the corresponding combinations based on the error values to obtain adjusted hyperparameter combinations; optimizing the adjusted hyperparameter combinations to obtain the optimal hyperparameter combination; and iteratively optimizing the 3D model using the machine learning surrogate model with the optimal hyperparameter combination and the sample dataset to obtain the optimized 3D model. This application establishes multiple 3D models with different objectives to be optimized, selects the most suitable hyperparameter combination for each model, and achieves multi-objective optimization for that model, further improving the efficiency of hyperparameter optimization and multi-objective optimization design. This application provides a more accurate and targeted multi-objective design optimization scheme. Compared with traditional optimization methods, it can significantly improve the efficiency of hyperparameter tuning for machine learning models, reduce the cost of hyperparameter tuning, thereby reducing training costs and improving model accuracy. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 A flowchart illustrating a multi-objective optimization design method based on machine learning, provided for an embodiment of this application;
[0044] Figure 2 An example flowchart of a machine learning-based multi-objective optimization design method provided in this application embodiment;
[0045] Figure 3 This is an example diagram of a wheel hub 3D CAD model before track intersection optimization provided in an embodiment of this application;
[0046] Figure 4 This is a schematic diagram of the structure of the deep learning model provided in the embodiments of this application;
[0047] Figure 5 A graph illustrating the contribution of deep learning hyperparameters to error values, provided in an embodiment of this application.
[0048] Figure 6 A flowchart illustrating the adjustment of deep learning hyperparameters provided in this application embodiment;
[0049] Figure 7 The optimized deep learning model provided in this application embodiment combines the Pareto front plot obtained through multi-objective optimization;
[0050] Figure 8 Example diagram of the optimized 3D CAD model of the wheel hub provided in the embodiments of this application;
[0051] Figure 9 A schematic diagram of a machine learning-based multi-objective optimization design device provided in this application embodiment;
[0052] Figure 10 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0053] 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 of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.
[0054] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”
[0055] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.
[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.
[0057] This application provides a machine learning-based multi-objective optimization design method and related equipment, relating to the field of machine learning technology. The machine learning-based multi-objective optimization design method and related equipment provided in this application can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited to these; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application implementing a machine learning-based multi-objective optimization design method, but is not limited to the above forms.
[0058] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.
[0059] Reference Figure 1 This application provides a multi-objective optimization design method based on machine learning. This method may include, but is not limited to, steps S100 to S170, as follows:
[0060] S100: Establish a 3D model with multiple objectives to be optimized;
[0061] S110: Obtain the sample dataset based on the three-dimensional model using the parametric modeling method;
[0062] S120: Set different combinations of hyperparameters for the machine learning agent model;
[0063] S130: Input uniform and fixed input parameters into the machine learning agent model to obtain the predicted values output by the machine learning agent model under different combinations of the hyperparameters;
[0064] S140: Determine the error value between the predicted value and the actual simulation value under the unified and fixed input parameters;
[0065] S150: Adjust the hyperparameters of the corresponding combination according to the error value to obtain the adjusted hyperparameter combination;
[0066] S160: Optimize the adjusted hyperparameter combination to obtain the optimal hyperparameter combination;
[0067] S170: The 3D model is iteratively optimized based on the machine learning proxy model under the optimal hyperparameter combination setting and the sample dataset to obtain the optimized 3D model.
[0068] Optionally, establishing multiple target 3D models to be optimized includes the following steps:
[0069] A CAD model is designed based on the shape and size of the workpiece to be modeled, and the CAD model is imported into CAE software for settings to obtain the three-dimensional model with multiple objectives to be optimized; wherein, each objective to be optimized is a set of objective response values under a set working condition.
[0070] Optionally, obtaining the sample dataset based on the 3D model using the parametric modeling method includes the following steps:
[0071] The sample dataset is obtained from the three-dimensional model using experimental methods or the finite element method.
[0072] Optionally, setting different combinations of hyperparameters for the machine learning agent model includes the following steps:
[0073] Different combinations of hyperparameters are set for all the hyperparameters of the machine learning agent model; wherein the setting parameters of each combination include the range of change of the hyperparameters and whether they are enabled or disabled.
[0074] Optionally, adjusting the hyperparameters of the corresponding combination based on the error value to obtain the adjusted hyperparameter combination includes the following steps:
[0075] Analyze the contribution of the hyperparameters to the error value under each combination;
[0076] The hyperparameters under the corresponding combination are adjusted according to the contribution level to obtain the adjusted hyperparameter combination.
[0077] Optionally, adjusting the hyperparameters for each combination based on the contribution to obtain the adjusted hyperparameter combination includes the following steps:
[0078] The hyperparameters of combinations whose contribution is greater than the first error contribution threshold are deleted.
[0079] The hyperparameters for combinations whose contribution is less than the second error contribution threshold are retained.
[0080] Wherein, the first error contribution threshold is greater than the second error contribution threshold.
[0081] Optionally, optimizing the adjusted hyperparameter combination to obtain the optimal hyperparameter combination includes the following steps:
[0082] The adjusted hyperparameter combination is optimized using a multi-objective optimization algorithm to obtain the optimal hyperparameter combination.
[0083] The following section will provide a detailed introduction and explanation of the solutions in the embodiments of this application, using specific application examples.
[0084] To address the aforementioned technical problems, this embodiment provides a multi-objective optimization design method based on machine learning, comprising the following steps:
[0085] S1. Establish a 3D model that needs to be optimized, and obtain a sample dataset using methods such as experiments and finite element analysis.
[0086] S2. Set machine learning hyperparameters to obtain the predicted output values of the machine learning surrogate model under a specific input and the error values of the simulation / experiment results under different hyperparameter combinations.
[0087] S3. Analyze the contribution of each hyperparameter to the error magnitude, modify or delete the hyperparameters to obtain new combinations of machine learning hyperparameters;
[0088] S4. Based on the new hyperparameter combination, use multi-objective optimization methods to optimize the hyperparameters and obtain the machine learning model under the optimal hyperparameter combination. Use this model to iteratively optimize the problem to obtain a more accurate optimization result.
[0089] In one implementation, the sample dataset is generated by using the finite element method, combined with a parametric model for rapid modeling, and then solving the model.
[0090] In one implementation, before training the machine learning agent model, the sample dataset needs to be preprocessed, including data cleaning, data normalization, handling missing values, and handling outliers.
[0091] In one implementation, the machine learning model can be one of the following methods: neural network, decision tree / random forest, support vector machine, XGBoost, K-means clustering, etc.
[0092] In one implementation, the input values for the machine learning model when calculating errors should be a uniform and fixed set of input parameters.
[0093] In one implementation, the hyperparameters of the machine learning model need to include all hyperparameters that affect the model's fitting performance. The specific range of values for each hyperparameter is determined based on the size of the collected sample dataset to avoid underfitting and overfitting.
[0094] In one implementation, the multi-objective optimization algorithm adopted should be one of the following: a three-generation non-dominated sorting multi-objective genetic optimization algorithm, a multi-objective optimization algorithm based on particle swarm optimization algorithm, or a multi-objective optimization algorithm based on evolutionary strategy.
[0095] In one implementation, the hyperparameter sampling method for the machine learning agent model should be one of random sampling, uniform sampling, full factorial design sampling, orthogonal experimental design sampling, or optimal Latin hypercube sampling.
[0096] In one implementation, the error value of the machine learning agent model should be the arithmetic mean of the error values of multiple optimization objectives.
[0097] In one implementation, the contribution analysis method adopted should be one of the following: Pareto contribution method, variance decomposition method, cooperative game method, or hypervolume contribution method.
[0098] In one implementation, after analysis using a specific contribution analysis method, the contribution relationship between hyperparameters and errors can be obtained. Based on this contribution relationship, hyperparameters that have a positive effect on increasing errors can be turned off or adjusted, while hyperparameters that have a positive effect on decreasing errors can be adjusted.
[0099] In one implementation, after obtaining the adjusted hyperparameter combination, the same multi-objective optimization algorithm is used to perform multi-objective optimization within the range of hyperparameter values. In this optimization, R... 2 The optimization targets are the highest and the lowest MSE values.
[0100] In one implementation, after obtaining the parameter combination, a multi-objective optimization algorithm is used to iteratively optimize the problem, thereby obtaining a more accurate multi-objective optimization result.
[0101] Reference Figure 2 This embodiment can also be implemented through the following methods:
[0102] S1. Establish a 3D model that needs to be optimized, and obtain a sample dataset using methods such as experiments and finite element analysis.
[0103] A 3D CAD model of a wheel hub for rail transit is created based on specific dimensional parameters, such as... Figure 3 As shown, the geometric model is preprocessed using finite element software, including geometric cleaning and mesh generation. Specific working conditions are set according to static, vibration, and acoustic requirements. Using the parametric finite element model and a secondary development batch processing file, the response values under different working conditions are automatically calculated, solved, and read, and collected into an Excel spreadsheet as an optimization sample dataset.
[0104] S2. Set machine learning hyperparameters to obtain the predicted output values of different machine learning surrogate models under specific inputs and the error values of simulation / experiment results;
[0105] The machine learning model used in this embodiment is a deep learning model. The hyperparameters involved in a deep learning model include: the number of hidden layers, the number of neurons in a single hidden layer, the types of activation functions, the batch size, the initial learning rate, and regularization-related parameters such as Dropout probability. A schematic diagram of the deep learning model is shown below. Figure 4 As shown.
[0106] When initializing a deep learning model, all hyperparameters should be enabled; the number of samples is 300, so the number of hidden layers should be between 2 and 5, the number of neurons in a single hidden layer should be between 16 and 64, the activation function should be one of Sigmoid, tanh, or ReLU, the batch size should be between 16 and 64, the initial learning rate should be between 1x10⁻⁶ and 1x10⁻¹, and the Dropout probability should be between 0.2 and 0.5.
[0107] It should be noted that in this embodiment, after determining the above hyperparameter sampling intervals, the optimal Latin hypercube method is used for sampling, wherein the number of hidden layers, the number of neurons in a single hidden layer, and the batch size should be integers.
[0108] S3. Analyze the contribution of each hyperparameter to the error magnitude, adjust the hyperparameters, and obtain the adjusted machine learning hyperparameter combination.
[0109] After completing step S2, for each set of hyperparameter combinations generated, the machine learning model under this hyperparameter combination is used to make predictions using the previously determined set of input parameter combinations. The simulated and predicted values of the output response under this hyperparameter combination are then subjected to error analysis to obtain the error value of this set of machine learning surrogate models. The output error value under each working condition is calculated to obtain the average error under multiple working conditions.
[0110] Based on the obtained average error value, the Pareto contribution analysis method is used to analyze the contribution of all hyperparameters, and the relationship between them and the average error is obtained, such as... Figure 5 As shown in the Pareto contribution plot, a positive correlation indicates that as the target dependent variable increases, the target dependent variable also increases; a negative correlation indicates that as the target dependent variable decreases, the target dependent variable also decreases. When the number of hidden layers is 4-5, it shows a positive correlation with the error; when the number of hidden layers is 2-3, it shows a negative correlation with the error. The number of neurons per layer shows a positive correlation with the error when 48-64, and a negative correlation when 16-48. The ReLU activation function, when enabled, shows a negative correlation with the error in this problem. The Tanh and Sigmoid activation functions, when enabled, show a positive correlation with the error in this problem. The batch size, when 40-64 layers, shows a positive correlation with the error; when the number of layers is 16-40, it shows a negative correlation with the error. The learning rate, when 1x10^(-1)-1x10^(-2), shows a positive correlation with the error; when the rate is 1x10^(-2)-1x10^(-6), it shows a negative correlation with the error. The Dropout probability, when 0.35-0.5, shows a positive correlation with the error; when 0.2-0.35, it shows a negative correlation with the error.
[0111] It can be seen that among the parameters positively correlated with error, batch size and Drouout probability are the top two contributors in the intervals of 40-64 and 0.35-0.5, respectively. Therefore, the values of these hyperparameters in these intervals are discarded, and only the values in the intervals of 16-40 and 0.2-0.35 are considered. Similarly, the intervals of other hyperparameters showing positive correlation are discarded, and only the intervals showing negative correlation are retained. Among the negatively correlated hyperparameters, batch size and number of hidden layers are the top two contributors, so they are retained and their values are kept unchanged. The contribution analysis is then performed on the remaining negatively correlated hyperparameters. After each analysis, the top two hyperparameters by contribution percentage are retained until the analysis ends. If the correlation is weak in the end, i.e., the percentage is less than 0.1 after analysis, the hyperparameter is deleted. The hyperparameter adjustment process is as follows: Figure 6 As shown.
[0112] Based on the above contribution analysis, the contribution relationships between each hyperparameter and the error were obtained. The intervals for the number of hidden layers, the number of neurons, the batch size, the learning rate, and the Drouout probability were set to show a negative correlation. ReLU was chosen as the activation function, and Tanh and Sigmoid were no longer used. Based on these selected hyperparameters, a three-generation non-dominated sorting multi-objective genetic optimization algorithm was used, with the R-squared value of the machine learning model... 2 The hyperparameter was optimized with the goal of maximizing its value and MSE value. The resulting hyperparameter combinations are shown in Table 1.
[0113] It should be noted that the dataset used in the training process is the sample dataset obtained in S1. Its input parameters are a total of ten input parameters, including the shape parameter of the wheel hub weight reduction hole and the parameter of the spoke thickness. The output parameters are the stress value of the optimized area of the wheel hub under three static working conditions, the natural frequency value of the free mode under dynamic working conditions, the equivalent acoustic radiation response ERP under acoustic working conditions, and the mass of the wheel hub itself.
[0114]
[0115] Table 1
[0116] Optimized deep learning model R 2 The MSE values are shown in Table 2:
[0117] Evaluation indicators <![CDATA[R 2 Value RMSE value Curved stress 0.9853 0.0268 linear stress 0.9814 0.0262 Turnout stress 0.9976 0.01 quality 0.9987 0.0104 Mode1 0.9930 0.0224 Mode2 0.9935 0.0212 Mode3 0.9952 0.0183 Mode4 0.9963 0.0164 ERP 0.9812 0.0275
[0118] Table 2
[0119] The prediction results before and after hyperparameter tuning were compared to determine the error, using the same set of parameters as input: [0.05,4,2.48,0,0.06,3.43,3.84,1.71,-3.58,25.06]. Before optimization, the average error of the response values was 6.70%, with the largest error occurring in the prediction of ERP (Emergency Participation Rate) at 16.10%. After optimization, the average error of the response values was 3.17%, with the largest error occurring in the prediction of turnout stress at 8.79%. The maximum error after optimization was significantly improved compared to before optimization, and the average error for each prediction condition was reduced by nearly half, resulting in a more accurate machine learning surrogate model.
[0120] The prediction results are shown in Table 3, which compares the prediction results before and after hyperparameter optimization. In the table, the stress units for turnouts, straight lines, and curves are MPa, the mass units are kg, the units for first-order mode, second-order mode, third-order mode, and fourth-order mode are Hz, the ERP unit is dB, and the unit for average error is %.
[0121]
[0122]
[0123] Table 3
[0124] S4. Based on the new hyperparameter combination, use multi-objective optimization methods to optimize the hyperparameters and obtain the machine learning model under the optimal hyperparameter combination. Use this model to iteratively optimize the problem to obtain a more accurate optimization result.
[0125] The initial model's response values for each operating condition are as follows: turnout stress 70.21 MPa, linear stress 22.71 MPa, curvilinear stress 79.05 MPa, mass 321 kg, first-order mode 295.05 Hz, second-order mode 340.55 Hz, third-order mode 340.60 Hz, fourth-order mode 533.21 Hz, and ERP acoustic radiation response 84.65 dB. Optimization of both lightweight design and acoustic performance is required for this model.
[0126] Based on the optimized hyperparameter combination obtained from S3, an optimized machine learning model is established to perform multi-objective optimization and solution for the above-mentioned wheel hub optimization problem. The multi-objective optimization problem is described as follows: Optimization objective: Minimize the mass and ERP equivalent acoustic radiation response. Design variables: Wheel hub weight reduction hole shape parameters + web thickness parameters. Optimization constraints: 1. The stress in the spoke region under the three static conditions does not exceed 150 MPa; 2. The first four modes of the wheel hub in free mode are not lower than 200 Hz, 290 Hz, 290 Hz, and 450 Hz respectively compared to before optimization; 3. The maximum decibel value of the radial acoustic radiation power curve does not exceed 85 dB.
[0127] The three-generation non-dominated sorting multi-objective genetic optimization algorithm is configured as follows: the population size is 200, the initial population sampling method is Latin hypercube sampling, simulated binary crossover is used with a crossover probability of 0.9 and a distribution index of 5, multinomial mutation is used with a mutation probability of 0.2 and a distribution index of 10; at the same time, in order to avoid population degradation and improve population diversity, duplicate individuals are eliminated, and the number of iterations is set to 1000.
[0128] The obtained Pareto front is as follows Figure 7 As shown, select Figure 7 The Pareto solution in the lower left box is taken as the optimal solution in this iteration. Its input parameters are substituted into the finite element model to obtain the optimized optimal solution parameters as shown in Table 4. The results of the optimized model are verified as shown in Table 5.
[0129]
[0130]
[0131] Table 4
[0132]
[0133] Table 5
[0134] Optimized wheel hubs, such as Figure 8 As shown in the figure, the largest error in the optimization results occurred in the predicted value of the curve stress, with an error of 4.64%. The average error of the nine predicted working conditions was 1.46%, indicating good accuracy. All the overall prediction results met the constraints. The mass was reduced from 321 kg to 305.38 kg, a reduction of 15.62 kg, and the acoustic response value was reduced from the initial 84.65 dB to 82.93 dB, a reduction of 2.17 dB, achieving both lightweighting of the structure and optimization of its acoustic performance.
[0135] The aforementioned machine learning-based multi-objective optimization design utilizes an initial model structure and a secondary-developed finite element method to generate a sample dataset. An optimized Latin hypercube sampling method is employed to sample the hyperparameters of the selected machine learning model. The machine learning model is trained for each hyperparameter combination based on the sample dataset, and the response values are predicted using the same parameter inputs to calculate the average error. A contribution analysis is performed on the relationship between hyperparameters and errors, and the hyperparameters are adjusted based on the analysis results. The adjusted hyperparameter combinations are then optimized using a multi-objective optimization algorithm to obtain an optimized machine learning model. Finally, the optimized machine learning model is combined with the multi-objective optimization algorithm to find the optimal solution for the problem, resulting in a more accurate and efficient machine learning multi-objective optimization design. This method is applicable to various scenarios requiring high-precision and high-efficiency machine learning multi-objective optimization design. Furthermore, the results achieve both lightweight design and performance optimization while meeting structural engineering performance requirements.
[0136] Reference Figure 9 This application also provides a machine learning-based multi-objective optimization design apparatus, which can implement the above-described machine learning-based multi-objective optimization design method. The apparatus includes:
[0137] The model building unit is used to build three-dimensional models of multiple targets to be optimized.
[0138] The sample acquisition unit is used to obtain a sample dataset based on the three-dimensional model using a parametric modeling method.
[0139] The hyperparameter setting unit is used to set different combinations of hyperparameters for the machine learning agent model;
[0140] An output prediction unit is used to input uniform and fixed input parameters into the machine learning agent model and obtain the predicted values output by the machine learning agent model under different combinations of the hyperparameters.
[0141] An error determination unit is used to determine the error value between the predicted value and the actual simulation value under the unified and fixed input parameters;
[0142] The hyperparameter adjustment unit is used to adjust the hyperparameters of the corresponding combination according to the error value to obtain the adjusted hyperparameter combination;
[0143] The hyperparameter optimization unit is used to optimize the adjusted hyperparameter combination to obtain the optimal hyperparameter combination.
[0144] The model optimization unit is used to iteratively optimize the 3D model based on the machine learning proxy model set with the optimal hyperparameter combination and the sample dataset to obtain the optimized 3D model.
[0145] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0146] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method of this application. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.
[0147] It is understood that the content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the methods of this application, and the beneficial effects achieved are the same as those achieved by the methods of this application.
[0148] Please see Figure 10 , Figure 10 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:
[0149] The processor 1001 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.
[0150] The memory 1002 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 1002 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1002 and is called and executed by the processor 1001.
[0151] Input / output interface 1003 is used to implement information input and output;
[0152] The communication interface 1004 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).
[0153] Bus 1005 transmits information between various components of the device (e.g., processor 1001, memory 1002, input / output interface 1003, and communication interface 1004);
[0154] The processor 1001, memory 1002, input / output interface 1003 and communication interface 1004 are connected to each other within the device via bus 1005.
[0155] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of this application.
[0156] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.
[0157] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0158] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.
[0159] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.
[0160] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0161] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.
[0162] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0163] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.
[0164] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0165] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0166] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0167] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0168] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.
Claims
1. A multi-objective optimization design method based on machine learning, characterized in that, The method includes the following steps: Establish three-dimensional models with multiple objectives to be optimized; The sample dataset is obtained based on the three-dimensional model using a parametric modeling method. Set different combinations of hyperparameters for machine learning agent models; The machine learning agent model is fed with uniform and fixed input parameters to obtain the predicted values output by the machine learning agent model under different combinations of the hyperparameters. Determine the error value between the predicted value and the actual simulation value under the unified and fixed input parameters; The hyperparameters of the corresponding combination are adjusted according to the error value to obtain the adjusted hyperparameter combination; The adjusted hyperparameter combination is optimized to obtain the optimal hyperparameter combination; The 3D model is iteratively optimized based on the machine learning proxy model with the optimal hyperparameter combination and the sample dataset to obtain the optimized 3D model.
2. The multi-objective optimization design method based on machine learning according to claim 1, characterized in that, The process of establishing multiple 3D models to be optimized includes the following steps: A CAD model is designed based on the shape and size of the workpiece to be modeled, and the CAD model is imported into CAE software for settings to obtain the three-dimensional model with multiple objectives to be optimized; wherein, each objective to be optimized is a set of objective response values under a set working condition.
3. The multi-objective optimization design method based on machine learning according to claim 1, characterized in that, The method of obtaining the sample dataset based on the 3D model using a parametric modeling approach includes the following steps: The sample dataset is obtained from the three-dimensional model using experimental methods or the finite element method.
4. The multi-objective optimization design method based on machine learning according to claim 1, characterized in that, Setting different combinations of hyperparameters for the machine learning agent model includes the following steps: Different combinations of hyperparameters are set for all the hyperparameters of the machine learning agent model; wherein the setting parameters of each combination include the range of change of the hyperparameters and whether they are enabled or disabled.
5. The multi-objective optimization design method based on machine learning according to claim 1, characterized in that, The step of adjusting the hyperparameters of the corresponding combination based on the error value to obtain the adjusted hyperparameter combination includes the following steps: Analyze the contribution of the hyperparameters to the error value under each combination; The hyperparameters under the corresponding combination are adjusted according to the contribution level to obtain the adjusted hyperparameter combination.
6. The multi-objective optimization design method based on machine learning according to claim 5, characterized in that, The step of adjusting the hyperparameters for each combination based on the contribution to obtain the adjusted hyperparameter combination includes the following steps: The hyperparameters of combinations whose contribution is greater than the first error contribution threshold are deleted. The hyperparameters for combinations whose contribution is less than the second error contribution threshold are retained. Wherein, the first error contribution threshold is greater than the second error contribution threshold.
7. A multi-objective optimization design method based on machine learning according to any one of claims 1 to 6, characterized in that, The process of optimizing the adjusted hyperparameter combination to obtain the optimal hyperparameter combination includes the following steps: The adjusted hyperparameter combination is optimized using a multi-objective optimization algorithm to obtain the optimal hyperparameter combination.
8. A multi-objective optimization design device based on machine learning, characterized in that, The device includes: The model building unit is used to build three-dimensional models of multiple targets to be optimized. The sample acquisition unit is used to obtain a sample dataset based on the three-dimensional model using a parametric modeling method. The hyperparameter setting unit is used to set different combinations of hyperparameters for the machine learning agent model; An output prediction unit is used to input uniform and fixed input parameters into the machine learning agent model and obtain the predicted values output by the machine learning agent model under different combinations of the hyperparameters. An error determination unit is used to determine the error value between the predicted value and the actual simulation value under the unified and fixed input parameters; The hyperparameter adjustment unit is used to adjust the hyperparameters of the corresponding combination according to the error value to obtain the adjusted hyperparameter combination; The hyperparameter optimization unit is used to optimize the adjusted hyperparameter combination to obtain the optimal hyperparameter combination. The model optimization unit is used to iteratively optimize the 3D model based on the machine learning proxy model set with the optimal hyperparameter combination and the sample dataset to obtain the optimized 3D model.
9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.