Vibration reduction optimization design method and device for optical platform
By optimizing the vibration isolator parameters of the optical platform using active learning based on the RVM model and multi-island genetic algorithm, the problems of low accuracy and high cost in calculating the vibration reduction performance of the optical platform are solved, and efficient vibration reduction performance evaluation and vibration isolator selection are achieved.
Patent Information
- Application Number
- CN202510944131.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-09
- Publication Date
- 2025-11-04
AI Technical Summary
Existing technologies suffer from low accuracy, high cost, and low efficiency in calculating the vibration reduction performance of optical platforms, making it difficult to accurately predict the jitter of optical devices and evaluate beam pointing accuracy.
An active learning method based on the RVM model is adopted to generate an initial sample space, optimize the design using input variables, and combine multi-island genetic algorithm and sensitivity analysis to optimize the vibration isolator parameters of the optical platform, thereby reducing computational cost and improving computational efficiency.
It enables high-precision and high-efficiency vibration reduction performance evaluation of optical platforms under external vibration environments, provides guidance for vibration isolator selection, reduces computational costs, and improves the vibration reduction performance of optical platforms.
Smart Images
Figure CN120893291A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of optical mechanical vibration reduction design, and particularly relates to an optical platform vibration reduction optimization design method and device. BACKGROUND
[0002] An optical platform is a mechanical device capable of integrating optical instrument installation, adapting to complex external load environment and guaranteeing stable transmission of an optical path, and is usually composed of an external box, an upper optical device and a bottom vibration isolator. The optical platform must have good vibration reduction performance to guarantee stable transmission of the optical path. However, due to a large number of uncertain design parameters in the structural design stage of the optical platform, it is difficult to accurately predict the jitter of the upper optical device, and it is also difficult to evaluate the beam pointing accuracy of the entire optical link, thereby causing the problem of low calculation accuracy.
[0003] The conventional parameter optimization often selects the Monte Carlo method, and the disadvantage is that a large number of samples are required, the calculation cost is high, and the production efficiency in the engineering field is largely limited. At the same time, the optical platform is a complex strong nonlinear mechanical system, and the mapping relationship between the input and the output is relatively complex. The difficulty in obtaining samples caused by low calculation efficiency is also a thorny problem. SUMMARY
[0004] The present application provides an optical platform vibration reduction optimization design method and device to solve the problems of low calculation accuracy, high cost and low efficiency in the prior art.
[0005] In a first aspect, the present application provides an optical platform vibration reduction optimization design method, comprising:
[0006] Determine the input variables based on the optical platform, wherein the optical platform comprises a box, a first number of vibration isolators and a second number of optical devices, the vibration isolators are located at the bottom of the box, the optical devices are located inside the box, and the input variables include the dynamic stiffness and viscous damping of each vibration isolator and the elastic modulus of the box;
[0007] Generate a third number of input variables according to the distribution type and distribution parameters of the input variables as an initial sample space;
[0008] Input the input variables in the initial sample space into the trained RVM model respectively to obtain the output response, wherein the output response includes the angular displacement of the centroid of each optical device in the pitch and azimuth directions;
[0009] Establish an optimization model by taking the interval of the vertical natural frequency as a constraint condition, taking the minimum of the maximum value of the output response as an optimization target and taking the input variables as design parameters, optimize the initial sample space and the output response thereof based on the optimization model, and obtain the optimal input variables and the output response thereof.
[0010] Optionally, the RVM model is trained based on active learning, comprising:
[0011] extracting a fourth number of input variables from the initial sample space, and submitting the fourth number of input variables to an optical platform dynamics model in batches for finite element simulation calculation to obtain a fourth number of output responses;
[0012] composing the fourth number of input variables and the fourth number of output responses into a fourth number of training samples as an initial training set, and each group of training samples comprises a group of input variables and output responses thereof;
[0013] initially training the RVM model by randomly selecting a fifth number of training samples from the initial training set;
[0014] continuously training the RVM model that has been initially trained by extracting new training samples from the remaining training samples in the initial training set as supplementary training samples, and stopping the training when a convergence condition is met, wherein the convergence condition is that the error between the output responses of the same group of input variables obtained by the RVM model and the output responses obtained by performing the finite element simulation calculation is not more than 2%.
[0015] Optionally, the extraction of the fourth number of input variables from the initial sample space is achieved by Latin hypercube sampling.
[0016] Optionally, the submission of the fourth number of input variables to the optical platform dynamics model in batches for finite element simulation calculation to obtain a fourth number of output responses comprises:
[0017] modifying the input variables in the CMD macro command stream file according to the fourth number of input variables by using a programming language, and submitting the modified CMD macro command stream file to the finite element software in batches for the finite element simulation calculation to obtain the fourth number of output responses, wherein the CMD macro command stream file is a text file containing all information of the optical platform dynamics model after being solved by the finite element software.
[0018] Optionally, the continuous extraction of new training samples from the remaining training samples in the initial training set as supplementary training samples is achieved by an adaptive U learning function.
[0019] Optionally, the optimization of the initial sample space and the output responses thereof based on the optimization model to obtain the optimal input variables and the output responses thereof is achieved by a multi-island genetic algorithm.
[0020] Optionally, it further comprises:
[0021] Performing sensitivity analysis based on the trained RVM model to obtain an importance ranking of the input variables.
[0022] In a second aspect, the present application provides an optical platform vibration reduction optimization design device, comprising an input variable determination module, an initial sample space creation module, an output response calculation module and an optimization module, wherein:
[0023] The input variable determination module is configured to determine input variables based on an optical platform, the optical platform comprising a box, a first number of vibration isolators and a second number of optical devices, the vibration isolators being located at the bottom of the box, the optical devices being located inside the box, the input variables comprising dynamic stiffness and viscous damping of each vibration isolator and elastic modulus of the box;
[0024] The initial sample creation module is configured to generate a third number of input variables as an initial sample space according to the distribution type and distribution parameters of the input variables;
[0025] The output response calculation module is configured to input the input variables in the initial sample space into a trained RVM model to obtain output responses thereof, the output responses comprising angular displacements of the mass centers of each optical device in the pitch and azimuth directions;
[0026] The optimization module is configured to establish an optimization model by taking a vertical natural frequency interval as a constraint condition, a minimum maximum value of the output responses as an optimization target and the input variables as design parameters, and to optimize the initial sample space and the output responses thereof based on the optimization model to obtain optimal input variables and output responses thereof.
[0027] In a third aspect, the present application provides a computing device, comprising:
[0028] a memory configured to store program instructions;
[0029] a processor configured to invoke the program instructions stored in the memory to execute any of the above methods according to the obtained program execution.
[0030] In a fourth aspect, the present application provides a computer-readable non-volatile storage medium comprising computer-readable instructions, which, when read and executed by a computer, cause the computer to execute any of the above methods.
[0031] The above scheme quantitatively evaluates the vibration reduction performance of the optical platform under external vibration environment with high precision and high efficiency, and realizes efficient and accurate optimization of the vibration reduction parameters of the optical platform and sensitivity analysis with low computational cost, which can provide guidance for the vibration reduction optimization design (i.e. vibration isolator selection) of the optical platform. BRIEF DESCRIPTION OF DRAWINGS
[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description only constitute some embodiments of the present application, and all other drawings obtained by those of ordinary skill in the art without creative effort based on these drawings belong to the protection scope of the present application.
[0033] Figure 1 A flowchart of an optical platform vibration reduction optimization design method provided by the embodiments of the present application is shown in the figure.
[0034] Figure 2 A flowchart of the RVM model based on active learning for training provided by the embodiments of the present application is shown in the figure.
[0035] Figure 3 A schematic diagram of the optical platform dynamics model provided by the embodiments of the present application is shown in the figure.
[0036] Figure 4 A histogram of the input variable sensitivity index provided by the embodiments of the present application is shown in the figure.
[0037] Figure 5 A structural diagram of an optical platform vibration reduction optimization design device provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION
[0038] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings. Obviously, the described embodiments only constitute some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the protection scope of the present application.
[0039] It should be noted that the terms "first", "second", etc. involved in the documents of the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.
[0040] Figure 1 The flow of the optical platform vibration reduction optimization design method provided by the embodiments of the present application is shown in detail, including:
[0041] S101, determining input variables based on an optical platform.
[0042] Specifically, the optical platform comprises a box, a first number of vibration isolators and a second number of optical devices, the vibration isolators are located at the bottom of the box, the optical devices are located inside the box, and the input variables (i.e. parameters to be optimized) include the dynamic stiffness and viscous damping of each vibration isolator and the elastic modulus of the box.
[0043] In a possible implementation, the optical platform comprises a box, four vibration isolators at the bottom and six optical devices inside, the dynamic stiffness of the four vibration isolators are represented as k1, k2, k3 and k4 respectively, the viscous damping are represented as c1, c2, c3 and c4 respectively, and the elastic modulus of the box is represented as E.
[0044] S102, generating a third number of input variables as an initial sample space according to the distribution type and distribution parameters of the input variables.
[0045] In the above possible implementation, the distribution type and distribution parameters of the input variables are shown in Table 1, wherein the mean and standard deviation are the distribution parameters. 10000 groups of input variables are generated as the initial sample space according to the distribution type and distribution parameters of the input variables shown in Table 1.
[0046] Table 1
[0047] Serial number Input variable name (unit) Distribution type Mean Standard deviation 1 1st vibration isolator dynamic stiffness k1 (N / mm) Normal distribution 2400 120 2 Viscous damping c1 (kg / s) of 1st vibration isolator Normal distribution 360 18 3 2nd vibration isolator dynamic stiffness k2 (N / mm) Normal distribution 2400 120 4 Viscous damping c2 (kg / s) of 2nd vibration isolator Normal distribution 360 18 5 3rd vibration isolator dynamic stiffness k3 (N / mm) Normal distribution 2400 120 6 Viscous damping c3 (kg / s) of 3rd vibration isolator Normal distribution 360 18 7 4th vibration isolator dynamic stiffness k4 (N / mm) Normal distribution 2400 120 8 Viscous damping c4 (kg / s) of 4th vibration isolator Normal distribution 360 18 9 Elastic modulus E (MPa) Normal distribution 110 5.5
[0048] S103, inputting the input variables in the initial sample space into the trained RVM model respectively to obtain the output responses thereof.
[0049] Specifically, the output responses include the angular displacement of the centroid of each optical device in the pitch and azimuth directions.
[0050] In an example, the RVM model is trained based on active learning, as shown in Figure 2 , including:
[0051] S201, extracting a fourth number of input variables from the initial sample space, and submitting the fourth number of input variables to the optical platform dynamics model in batches for finite element simulation calculation to obtain a fourth number of output responses.
[0052] In an example, the extraction of the fourth number of input variables from the initial sample space is achieved by Latin hypercube sampling.
[0053] In an example, the submission of the fourth number of input variables to the optical platform dynamics model in batches for finite element simulation calculation to obtain the fourth number of output responses includes:
[0054] The input variables in the CMD macro command stream file are modified according to the fourth number of input variables using a programming language. The modified CMD macro command stream files are then submitted in batches to the finite element software for finite element simulation calculations to obtain the fourth number of output responses. The CMD macro command stream file is a text file containing all information of the optical platform dynamic model exported after being solved by the finite element software.
[0055] In the above possible implementations, Latin hypercube sampling is used to extract 1000 sets of input variables from the initial sample space. Python is then used to modify the input variables in the CMD macro command stream file according to each of the 1000 sets of input variables. The modified CMD macro command stream files are then submitted in batches for finite element simulation calculations to obtain 1000 sets of output responses. Finite element simulation calculations based on the optical platform dynamics model include:
[0056] Establish a dynamic model of the optical platform: such as Figure 3 As shown, it consists of a box, four vibration isolators at the bottom, and six optical devices inside. In the finite element software, the box is meshed. The optical device model is replaced by a concentrated mass point (the mass point position coincides with the centroid of the optical device). The vibration isolator model is replaced by a bushing element and is bound to the ground.
[0057] Apply external loads to the optical platform: Analyze the vibration load spectrum of the optical platform and apply the corresponding vibration load spectrum in the form of acceleration power spectral density (PSD) at the center position of the bottom of the four vibration isolators;
[0058] Set the dynamic simulation parameters: set the frequency response analysis range to 1Hz~100Hz, the frequency resolution to 0.1Hz, use the modal superposition method to solve the random vibration conditions of the optical platform, set the structural damping ratio to 0.01, and use the t-mm-s unit system for the dynamic model of the optical platform.
[0059] Extracting dynamic simulation results: After solving in the finite element software, the angular displacements (denoted as Ry and Rz) of the centroids of the six optical devices in the model in the pitch and azimuth directions are extracted based on the global coordinate system, and the CMD macro command stream file of the model is exported.
[0060] S202, the fourth number of input variables and the fourth number of output responses are combined to form the fourth number of training samples, which are used as the initial training set.
[0061] Specifically, each training sample includes a set of input variables and their output responses.
[0062] S203, arbitrarily select the fifth number of training samples from the initial training set to perform initial training on the RVM model.
[0063] In the possible implementation manner, the fifth quantity of training samples is 100 groups of training samples.
[0064] S204, continuously extract new training samples from the remaining training samples of the initial training set as supplementary training samples to continue training the RVM model after the initial training, and stop training when the convergence condition is met.
[0065] Specifically, the convergence condition is that the error between the output response obtained by the same group of input variables via the RVM model and the output response obtained by performing finite element simulation calculation does not exceed 2%.
[0066] Satisfying the convergence condition means that the accuracy of the RVM model meets the standard. In the possible implementation manner, only 100 groups of training samples are needed as the initial training set and 45 groups of supplementary training samples to train the RVM model with good calculation accuracy. At the same time, compared with the process of using finite element simulation all the way, this method greatly reduces the calculation cost.
[0067] In an example, continuously extracting new training samples from the remaining training samples of the initial training set as supplementary training samples is implemented by an adaptive U learning function.
[0068] By selecting more training samples to continue training the RVM model through the adaptive U learning function, the prediction accuracy of the RVM model is higher.
[0069] S104, an optimization model is established by taking the interval of the vertical natural frequency as a constraint condition, the minimum maximum value of the output response as an optimization target, and the input variable as a design parameter, and the initial sample space and its output response are optimized based on the optimization model to obtain the optimal input variable and its output response.
[0070] Since the angular displacement has a direction, that is, there is a positive and negative sign, the maximum value of the angular displacement is the maximum absolute value of the angular displacement.
[0071] Although the beam pointing accuracy is a broad concept, the maximum value of the angular displacement is only a specific physical quantity for evaluating the beam pointing accuracy, but the maximum value of the angular displacement can reflect the beam pointing accuracy from the side.
[0072] The vertical natural frequency is an intermediate quantity generated in the finite element simulation calculation process, which cannot be too high or too low. Taking it as a constraint condition can ensure that the optimized result meets the engineering practice.
[0073] In the possible implementation manner, the optimization model established is as follows:
[0074]
[0075] wherein, Ry i and Rzi f represents the angular displacement of the centroid of the i-th optical device in the pitch and azimuth directions, respectively. v For the vertical natural frequency, x j For the j-th design parameter, μ j and σ j denoted as the mean and standard deviation of the normal distribution of the j-th design parameter, respectively.
[0076] In one example, the initial sample space and its output response are optimized based on an optimization model to obtain the optimal input variables and their output response, which is achieved through a multi-island genetic algorithm.
[0077] The multi-island genetic algorithm is an off-the-shelf tool for quickly selecting a maximum or minimum value from a large amount of data.
[0078] In the above possible implementations, a multi-island genetic algorithm was used to perform 1015 iterations to screen 10,000 sets of input variables and their output responses, obtaining the optimal input variables and their output responses. Table 2 shows a comparison of the initial values and optimized values of the input variables, and Table 3 shows a comparison of the initial values and optimized values of the output responses.
[0079] Table 2
[0080] Serial number Input variable (unit) Initial value Optimized value 1 1st vibration isolator dynamic stiffness k1 (N / mm) 2400 2654 2 Viscous damping c1 (kg / s) of 1st vibration isolator 360 381 3 2nd vibration isolator dynamic stiffness k2 (N / mm) 2400 2511 4 Viscous damping c2 (kg / s) of 2nd vibration isolator 360 268 5 3rd vibration isolator dynamic stiffness k3 (N / mm) 2400 2613 6 Viscous damping c3 (kg / s) of 3rd vibration isolator 360 368 7 4th vibration isolator dynamic stiffness k4 (N / mm) 2400 2498 8 Viscous damping c4 (kg / s) of 4th vibration isolator 360 365 9 Elastic modulus E (MPa) 110 116
[0081] Table 3
[0082] Serial number Output response (unit) Initial value Optimized value 1 Pitch angle displacement maximum max(|Ry i |) (μrad) 57.14 34.85 2 azimuth angle displacement maximum max(|Rz i |)(μrad) 10.21 6.61
[0083] The comparison shows that the maximum pitch angle displacement of the optical device's centroid was reduced by 39% and the maximum azimuth angle displacement was reduced by 35% after optimization, indicating that the optimization effect met expectations.
[0084] In one example, an optical platform vibration reduction optimization design method provided by an embodiment of the present invention further includes:
[0085] Sensitivity analysis was performed based on the trained RVM model to obtain the importance ranking of the input variables.
[0086] Sensitivity analysis is an existing tool that requires a large number of input variables and output responses. A trained RVM model can quickly calculate these data. Sensitivity analysis reveals the contribution of input variables to changes in the output response; higher sensitivity values indicate greater importance, allowing for priority optimization of more important input variables.
[0087] Among the above possible implementations, Figure 4A histogram of the input variable (i.e., the design variable in the figure) sensitivity index is shown, and the importance ranking of the input variables is k1>k3>c2>c4>c1>c3>E>k2>k4.
[0088] The scheme quantitatively evaluates the vibration reduction performance of the optical platform in the external vibration environment with high precision and high efficiency, and realizes efficient and accurate optimization of the vibration reduction parameters of the optical platform and sensitivity analysis with low calculation cost, thereby providing guidance for the vibration reduction optimization design (i.e., the selection of the vibration isolator) of the optical platform.
[0089] Based on the same technical concept, Figure 5 The structure of the optical platform vibration reduction optimization design device provided by the embodiment of the application is shown, which comprises an input variable determination module, an initial sample space creation module, an output response calculation module and an optimization module, wherein:
[0090] The input variable determination module is used to determine the input variable based on the optical platform, the optical platform comprises a box, a first number of vibration isolators and a second number of optical devices, the vibration isolators are located at the bottom of the box, and the optical devices are located inside the box, and the input variable comprises the dynamic stiffness and viscous damping of each vibration isolator and the elastic modulus of the box;
[0091] The initial sample creation module is used to generate a third number of input variables as an initial sample space according to the distribution type and distribution parameters of the input variables;
[0092] The output response calculation module is used to input the input variables in the initial sample space into the trained RVM model respectively to obtain the output response thereof, and the output response comprises the angular displacement of the centroid of each optical device in the pitch and azimuth directions;
[0093] The optimization module is used to establish an optimization model by taking the interval of the vertical natural frequency as a constraint condition, the minimum maximum value of the output response as an optimization target and the input variable as a design parameter, and optimize the initial sample space and the output response thereof based on the optimization model to obtain the optimal input variable and the output response thereof.
[0094] Based on the same technical concept, the embodiment of the application provides a computing device, comprising:
[0095] A memory is used to store program instructions;
[0096] A processor is used to call the program instructions stored in the memory, and execute the above method according to the obtained program execution.
[0097] Based on the same technical concept, the embodiment of the application provides a computer readable nonvolatile storage medium, comprising computer readable instructions, when the computer reads and executes the computer readable instructions, the computer executes the above method.
[0098] The present application is described in reference to the flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor 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 processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for carrying out the function specified by the block or blocks.
[0099] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for carrying out the function specified by the block or blocks.
[0100] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart and / or block diagram block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for carrying out the function specified by the block or blocks.
[0101] While preferred embodiments of the application have been described, modifications and variations can be apparent to those skilled in the art once aware of the general underlying concepts. Accordingly, the appended claims intend to encompass within their scope all such modifications and variations as fall within the true spirit and scope of the present application.
[0102] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.
Claims
1. An optical bench vibration reduction optimization design method, characterized in that, The method comprises the following steps: determining input variables based on an optical platform, the optical platform comprising a box, a first number of vibration isolators and a second number of optical devices, the vibration isolators being located at the bottom of the box, the optical devices being located inside the box, the input variables comprising dynamic stiffness and viscous damping of each vibration isolator and elastic modulus of the box; generating a third number of input variables as an initial sample space according to distribution type and distribution parameters of the input variables; inputting the input variables in the initial sample space into a trained RVM model respectively to obtain output responses, the output responses comprising angular displacement of a centroid of each optical device in the pitch and azimuth directions; establishing an optimization model based on a constraint condition of a vertical natural frequency, an optimization objective of minimum maximum value of the output responses and design parameters of the input variables, and optimizing the initial sample space and the output responses based on the optimization model to obtain optimal input variables and output responses.
2. The method of claim 1, wherein, The RVM model is trained based on active learning, comprising the following steps: extracting a fourth number of input variables from the initial sample space, and submitting the fourth number of input variables to a finite element simulation calculation of an optical platform dynamics model in batches to obtain a fourth number of output responses; composing the fourth number of input variables and the fourth number of output responses into a fourth number of training samples as an initial training set, each training sample comprising a group of input variables and output responses thereof; initially training the RVM model by randomly selecting a fifth number of training samples from the initial training set; continuously extracting new training samples from the remaining training samples in the initial training set as supplementary training samples to continue training the RVM model which has been initially trained, and stopping training when a convergence condition is met, the convergence condition being that an error between output responses of the same group of input variables obtained by the RVM model and obtained by the finite element simulation calculation is not more than 2%.
3. The method of claim 2, wherein, The extraction of the fourth number of input variables from the initial sample space is achieved by Latin hypercube sampling.
4. The method of claim 2, wherein, The submission of the fourth number of input variables to the finite element simulation calculation of the optical platform dynamics model comprises the following steps: modifying input variables in a CMD macro command stream file according to the fourth number of input variables by using a programming language, and submitting the modified CMD macro command stream file to a finite element software for the finite element simulation calculation to obtain the fourth number of output responses, the CMD macro command stream file being a text file containing all information of the optical platform dynamics model after being solved by the finite element software.
5. The method of claim 2, wherein, The continuous extraction of new training samples from the remaining training samples in the initial training set as supplementary training samples is achieved by an adaptive U learning function.
6. The method of claim 1, wherein, The optimization of the initial sample space and the output responses based on the optimization model to obtain optimal input variables and output responses is achieved by a multi-island genetic algorithm.
7. The method of claim 1 wherein, The method further comprises the following steps: Based on the trained RVM model, sensitivity analysis is performed to obtain the importance ranking of the input variables.
8. An optical bench vibration reduction optimization design apparatus, characterized by comprising: The method comprises an input variable determination module, an initial sample space creation module, an output response calculation module, and an optimization module, wherein: The input variable determination module is configured to determine input variables based on an optical platform, the optical platform comprising a box, a first number of vibration isolators, and a second number of optical devices, the vibration isolators being located at the bottom of the box, and the optical devices being located inside the box, the input variables comprising dynamic stiffness and viscous damping of each vibration isolator and elastic modulus of the box; The initial sample creation module is configured to generate a third number of input variables as an initial sample space according to the distribution type and distribution parameters of the input variables; The output response calculation module is configured to input the input variables in the initial sample space into the trained RVM model to obtain output responses, the output responses comprising angular displacements of the centroids of each optical device in the pitch and azimuth directions; The optimization module is configured to establish an optimization model by taking a vertical natural frequency interval as a constraint condition, a minimum maximum value of the output responses as an optimization objective, and the input variables as design parameters, and to optimize the initial sample space and the output responses based on the optimization model to obtain optimal input variables and output responses.
9. A computing device, comprising: The method comprises: a memory configured to store program instructions; a processor configured to invoke the program instructions stored in the memory to execute the method according to any one of claims 1-7.
10. A computer-readable non-transitory storage medium, characterized in that, computer-readable instructions, when read and executed by a computer, cause the computer to execute the method according to any one of claims 1-7.