Laser-driven multi-shell fusion target design method and device based on hybrid optimization

Abstract

The present invention relates to a laser-driven multi-shell fusion target design method and device based on hybrid optimization. The method comprises: obtaining an initial neutron yield; generating a first input parameter by normalizing a random variable; performing numerical simulation calculation according to the first input parameter to obtain a first neutron yield; comparing the first neutron yield with the initial neutron yield to perform cyclic convergence control until the random walk step length is less than or equal to the random walk step length accuracy, and obtaining multiple sets of parameters in the optimization process; constructing an initial input-output model using multiple sets of parameters as initial data; solving a second input parameter, and calculating the corresponding second neutron yield; iteratively optimizing the initial input-output model through the second input parameter and the second neutron yield to obtain an optimized input-output model; and estimating the optimal parameters based on the optimized input-output model. The present invention does not require manual adjustment of input parameters, accelerates optimization through cyclic convergence control, has high efficiency, and has low time cost.

Description

Technical Field

[0001] The invention relates to the technical field of inertial confinement fusion, and in particular to a method and device for designing a laser-driven multi-shell fusion target based on hybrid optimization. Background Art

[0002] Nuclear fusion energy is a clean and safe energy source and an important way to solve the energy problem. Among them, inertial confinement fusion (ICF) is an important way to achieve fusion energy. Its basic principle is to drive the fuel particles with high-power density laser or particle beam in a short time, so that the fuel is quickly compressed to ultra-high density and temperature, thereby triggering nuclear fusion reaction and releasing huge energy. However, the process of achieving inertial confinement fusion is extremely complex, involving the interaction of multiple disciplines such as radiation, fluid mechanics, and plasma physics.

[0003] In the optimization of inertial confinement fusion, in order to predict and control the parameter changes of physical processes, the one-dimensional radiation fluid dynamics program MULTI is usually used for numerical simulation. MULTI can quickly complete the calculation based on the initial parameters input, thereby providing support for experimental design and scheme improvement. However, to obtain ideal simulation results, more than a dozen or even more parameters need to be carefully adjusted. Since these parameters are highly sensitive during the optimization process, small changes may lead to large differences in the results. Such parameters include drive power, time series, temperature distribution during compression, etc., and their mutual coupling makes it difficult to grasp the laws. Therefore, exploring the laws of inertial confinement fusion reactions by manually adjusting and inputting parameters one by one is not only a huge workload, but also inefficient, which increases the difficulty of obtaining ideal results. Summary of the invention

[0004] Based on this, it is necessary to provide a laser-driven multi-shell fusion target design method and device based on hybrid optimization that can reduce the need for human intervention, improve the accuracy of parameter adjustment, and accelerate the design of inertial confinement fusion in response to the above-mentioned technical problems.

[0005] A method for designing a laser-driven multi-shell fusion target based on hybrid optimization, the method comprising:

[0006] Step 1, performing numerical simulation calculation according to initial random walk parameters and conditional parameters to obtain initial neutron yield;

[0007] Step 2, generating a random variable, and normalizing the random variable to generate a first input parameter;

[0008] Step 3, performing numerical simulation calculation according to the first input parameter to obtain a first neutron yield;

[0009] Step 4, compare the first neutron yield with the initial neutron yield. If the first neutron yield is greater than the initial neutron yield, update the initial neutron yield to the first neutron yield, and the number of cycles is reset to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles is increased by 1; if the number of cycles reaches the convergence control number, update the initial random walk step length; until the random walk step length is less than or equal to the random walk step length precision, the random walk program ends, and multiple sets of parameters in the optimization process are obtained;

[0010] Step 5, using the multiple sets of parameters as initial data to construct an initial input-output model;

[0011] Step 6, using the acquisition function to solve the second input parameter, and performing numerical simulation calculation according to the second input parameter to obtain a second neutron yield;

[0012] Step 7, iteratively optimizing the initial input-output model through several groups of the second input parameters and the second neutron yield to obtain an optimized input-output model;

[0013] Step 8: Estimating optimal parameters based on the optimized input-output model.

[0014] A laser-driven multi-shell fusion target design device based on hybrid optimization, the device comprising:

[0015] An initialization module is used to perform numerical simulation calculations based on initial random walk parameters and conditional parameters to obtain an initial neutron yield;

[0016] A random variable generation module, used to generate a random variable, normalize the random variable, and generate a first input parameter;

[0017] A first neutron yield calculation module, used for performing numerical simulation calculation according to the first input parameter to obtain a first neutron yield;

[0018] A loop convergence control module is used to compare the first neutron yield with the initial neutron yield. If the first neutron yield is greater than the initial neutron yield, the initial neutron yield is updated to the first neutron yield, and the number of cycles is reset to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles is increased by 1; if the number of cycles reaches the convergence control number, the initial random walk step length is updated; until the random walk step length is less than or equal to the random walk step length precision, the random walk program ends, and multiple groups of parameters in the optimization process are obtained;

[0019] An initial model building module, used to build an initial input-output model using the multiple sets of parameters as initial data;

[0020] A second neutron yield calculation module, used to solve a second input parameter using an acquisition function, and perform numerical simulation calculation according to the second input parameter to obtain a second neutron yield;

[0021] an iterative parameter generation module, configured to iteratively optimize the initial input-output model through a plurality of groups of the second input parameters and the second neutron yield to obtain an optimized input-output model;

[0022] The parameter generation module is used to estimate the optimal parameters based on the optimized input-output model.

[0023] The above-mentioned laser-driven multi-shell fusion target design method and device based on hybrid optimization perform numerical simulation calculations according to initial random walk parameters and conditional parameters to obtain an initial neutron yield; generate random variables, and normalize the random variables to generate first input parameters; perform numerical simulation calculations according to the first input parameters to obtain a first neutron yield; compare the first neutron yield with the initial neutron yield, if the first neutron yield is greater than the initial neutron yield, update the initial neutron yield to the first neutron yield, and reset the number of cycles to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles increases by 1 .... When the number reaches the convergence control number, the initial random walk step size is updated; until the random walk step size is less than or equal to the random walk step size precision, the random walk program ends, and multiple groups of parameters in the optimization process are obtained; the multiple groups of parameters are used as initial data to construct an initial input-output model; the acquisition function is used to solve the second input parameter, and a numerical simulation calculation is performed according to the second input parameter to obtain a second neutron yield; the initial input-output model is iteratively optimized through several groups of second input parameters and the second neutron yield to obtain an optimized input-output model; optimal parameter estimation is performed based on the optimized input-output model.

[0024] The present invention does not need to manually adjust the input parameters, and automatically generates the initial data required for building the model through the random walk algorithm. In the random walk algorithm, the first neutron yield is compared with the initial neutron yield for cyclic convergence control, so that the random walk algorithm is not easy to fall into the local extreme value range, the optimization speed is accelerated, and the optimization efficiency is improved; the initial input-output model is constructed by using multiple groups of parameters as initial data, and the second input parameter is generated by the acquisition function, and the second neutron yield corresponding to the second input parameter is calculated. The initial input-output model is iteratively optimized and trained by several groups of second input parameters and the second neutron yield, which can improve the prediction accuracy and the reliability of the model. And based on the optimized input-output model, it can quickly generate appropriate output according to the requirements of the input parameters, improve the prediction ability of the neutron yield, and has the advantages of high efficiency and low time cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0026] Figure 1 A schematic flow chart of a method for designing a laser-driven multi-shell fusion target based on hybrid optimization in one embodiment;

[0027] Figure 2 A structural block diagram of a laser-driven multi-shell fusion target design device based on mixing optimization in one embodiment.

[0028] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] It should be noted that in the present invention, descriptions such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0031] It can be understood that the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0032] The following will describe the implementation of the present invention in detail with reference to the accompanying drawings in the embodiment diagram of the present invention.

[0033] Example 1

[0034] like Figure 1As shown, a method for designing a laser-driven multi-shell fusion target based on hybrid optimization is provided, comprising the following steps:

[0035] Step 1: Perform numerical simulation calculations based on initial random walk parameters and conditional parameters to obtain initial neutron yield.

[0036] Step 2: Generate a random variable, normalize the random variable, and generate a first input parameter.

[0037] Step 3: Perform numerical simulation calculation according to the first input parameter to obtain a first neutron yield.

[0038] Step 4, compare the first neutron yield with the initial neutron yield. If the first neutron yield is greater than the initial neutron yield, update the initial neutron yield to the first neutron yield and reset the number of cycles to zero. If the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged and the number of cycles increases by 1. If the number of cycles reaches the convergence control number, update the initial random walk step length. When the random walk step length is less than or equal to the random walk step length precision, the random walk program ends and multiple sets of parameters in the optimization process are obtained.

[0039] It can be understood that in the process from step 2 to step 4, the parameter space is mainly explored by random walk, so as to quickly generate diverse input parameters, and then effectively find combinations that may increase neutron yield, reduce human intervention and adjustment of parameters, and provide diverse data for subsequent model construction.

[0040] During the random walk, unlike the traditional method of adjusting the step size after comparing the yield size, the present invention generates a set of input parameters and neutron yields at one time, and then optimizes the cycle process by the convergence control number, so that the entire cycle process will not fall into the local extreme value range, and the entire cycle process is more adaptable and flexible.

[0041] Step 5: Use multiple sets of parameters as initial data to build an initial input-output model.

[0042] Step 6: Use the acquisition function to solve the second input parameter, and perform numerical simulation calculation based on the second input parameter to obtain the second neutron yield.

[0043] Step 7, iteratively optimizing the initial input-output model through several groups of second input parameters and second neutron yields to obtain an optimized input-output model.

[0044] It can be understood that constructing the initial input-output model based on multiple sets of parameters obtained based on the random walk method ensures that the foundation of the model is based on real data, thereby improving the reliability and predictive ability of the model.

[0045] Then, in the process from step 5 to step 7, the second input parameter is solved by the acquisition function, and the corresponding second neutron yield is calculated; the initial input-output model is optimized by the second input parameter and the corresponding second neutron yield, and the model is continuously improved to make the trained final model more accurate.

[0046] Step 8: Estimating optimal parameters based on the optimized input-output model.

[0047] The above-mentioned laser-driven multi-shell fusion target design method based on hybrid optimization performs numerical simulation calculation according to the initial random walk parameters and conditional parameters to obtain the initial neutron yield; generates random variables, normalizes the random variables, and generates first input parameters; performs numerical simulation calculation according to the first input parameters to obtain the first neutron yield; compares the first neutron yield with the initial neutron yield, and if the first neutron yield is greater than the initial neutron yield, updates the initial neutron yield to the first neutron yield, and the number of cycles is reset to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles is increased by 1; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles is increased by 1; When the number of loops reaches the convergence control number, the initial random walk step size is updated; until the random walk step size is less than or equal to the random walk step size precision, the random walk program ends and multiple sets of parameters in the optimization process are obtained; using multiple sets of parameters as initial data, an initial input-output model is constructed; the acquisition function is used to solve the second input parameter, and a numerical simulation calculation is performed based on the second input parameter to obtain the second neutron yield; the initial input-output model is iteratively optimized through several sets of second input parameters and the second neutron yield to obtain an optimized input-output model; optimal parameter estimation is performed based on the optimized input-output model.

[0048] The present invention does not need to manually adjust the input parameters, and can quickly perform automatic optimization after the input parameters are input, which is highly efficient; the initial data required for model construction is automatically generated by the random walk algorithm, and the initial data is usually distributed around the extreme value, which is more efficient. In addition, in the random walk algorithm, the first neutron yield is compared with the initial neutron yield for cyclic convergence control, so that the random walk algorithm is not easy to fall into the local extreme value range, accelerate the optimization speed, and improve the optimization efficiency. The initial input-output model is constructed by using multiple groups of parameters as initial data, and the second input parameter is generated by the acquisition function, and the second neutron yield corresponding to the second input parameter is calculated. The initial input-output model is iteratively optimized and trained by several groups of second input parameters and the second neutron yield, which can improve the prediction accuracy and the reliability of the model. And based on the optimized input-output model, it is possible to quickly generate appropriate outputs according to the requirements of the input parameters, improve the prediction ability of the neutron yield, and has the advantages of high efficiency and low time cost.

[0049] In step 1, the initial random walk parameters include the initial random walk step size , Random Walk Step Accuracy and convergence control times ; The condition parameters include the radiation temperature waveform and the initial parameters of the pellet structure .

[0050] During the initialization process, the random walk step length is set to the initial random walk step length, that is, , and set the random walk step size accuracy , convergence control times Then input the radiation temperature waveform and the initial parameters of the pellet structure , the initial neutron yield is obtained by simulation calculation through numerical simulation program .in, Indicates each parameter.

[0051] Number of loops ; Then enter the random variable generation in step 2.

[0052] In step 2, the expression for generating random variables is:

[0053] ;

[0054] In the formula, represents a random variable; Represents the radiation temperature waveform and the initial parameters of the pellet structure; Represents a random number between [0, 1].

[0055] Normalize the random variable and generate the first input parameter. The expression is:

[0056] ;

[0057] In the formula, Represents the first input parameter; represents the random walk step length.

[0058] In step 3, the first input parameter Write into the numerical simulation program to perform simulation calculations and obtain the first neutron yield .

[0059] In step 4, it is a cyclic convergence control process, which specifically includes: if the first neutron yield is greater than the initial neutron yield, updating the initial parameter to the first input parameter, and updating the initial neutron yield to the first neutron yield, and returning to step 2 to regenerate the random variable after the number of cycles is reset to zero.

[0060] If the first neutron yield is less than or equal to the initial neutron yield, and the number of cycles is less than the convergence control number, the initial neutron yield remains unchanged, the number of cycles is increased by 1, and the process returns to step 2 to regenerate the random variable.

[0061] If the first neutron yield is less than or equal to the initial neutron yield, and the number of cycles is equal to the convergence control number, and the initial random walk step size is greater than the random walk step size accuracy, the initial random walk step size is updated to the preset random walk step size, and return to step 2 to regenerate random variables.

[0062] If the first neutron yield is less than or equal to the initial neutron yield, the number of cycles is equal to the convergence control number, and the random walk step is less than or equal to the random walk step precision, the random walk program ends and multiple sets of parameters in the optimization process are obtained.

[0063] Specifically, if , then let , and then let the number of cycles , regenerate random variables through step 2.

[0064] like ,and , then no update is performed, and the number of cycles is Then, regenerate the random variables through step 2.

[0065] like ,and ,when When , the initial random walk step length is updated to the preset random walk step length. In this embodiment, half of the initial random walk step length is used as the preset random walk step length for updating, that is, ; Then regenerate the random variables through step 2.

[0066] like ,and ,when When , the random walk program ends, and multiple sets of parameters in the optimization process are obtained. . Multiple groups of parameters It includes several sets of data consisting of input parameters and neutron yields, which are used as the initial basis for subsequent model construction.

[0067] In step 5, the number of iterations is now set , and then with multiple sets of parameters As the initial data, the initial estimation is performed through the Gaussian process to construct the initial input-output model. That is to say, in step 5, the initial input-output model can be constructed by the Bayesian method, or a model with input-output characteristics can be constructed by other methods. The method used is a general training function, which will not be described in detail here.

[0068] In step 6, the acquisition function is used to solve the new input parameter, which is recorded as the second input parameter, and a numerical simulation program is used to perform simulation calculations to obtain a corresponding new neutron yield, which is recorded as the second neutron yield.

[0069] It can be understood that the acquisition function is selected according to the requirements. In this embodiment, the radial basis function is preferably used to solve the input parameters to better adapt to the method of this application.

[0070] Specifically, through this step, several groups of second input parameters and second neutron yields are generated as data for subsequent iterative updates.

[0071] In step 7, the second input parameters and the second neutron yield obtained in step 6 are input into the initial input-output model for estimation again, and then the initial input-output model is iteratively updated according to the estimation results. When the number of iterations reaches When , the optimized input-output model is obtained.

[0072] Finally, in step 8, the optimized input-output model can be used to directly output the corresponding optimal parameters for the input parameters, thereby optimizing the target structure.

[0073] It can be understood that the inputs in the optimized input-output model are usually laser or radiation parameters, target parameters, etc. The optimal parameters obtained by the optimized input-output model include but are not limited to neutron yield, energy yield parameters, etc., which are output as optimal parameters according to requirements.

[0074] In one embodiment, the method further comprises: changing the radiation temperature waveform and the initial parameters of the target pellet structure , and repeat steps 1 to 8 to optimize the input and output models again.

[0075] In one embodiment, the radiation fluid dynamics program MULTI is used to perform numerical simulation calculations.

[0076] In one embodiment, a laser-driven multi-shell fusion target design method based on hybrid optimization is used to optimize a single-layer spherical target or a multi-layer spherical target structure.

[0077] Although this embodiment Figure 1 The steps in the process are shown in sequence as indicated by the arrows, but the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of the steps, and the steps can be executed in other orders. Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0078] Example 2

[0079] Based on the laser-driven multi-shell fusion target design method based on hybrid optimization in Example 1, this embodiment discloses a laser-driven multi-shell fusion target design device based on hybrid optimization, such as Figure 2 As shown, the laser driven multi-shell fusion target design device based on hybrid optimization includes: an initialization module 401, a random variable generation module 402, a first neutron yield calculation module 403, a cycle convergence control module 404, an initial model construction module 405, a second neutron yield calculation module 406, an iterative parameter generation module 407 and a parameter generation module 408, wherein:

[0080] The initialization module 401 is used to perform numerical simulation calculations according to the initial random walk parameters and conditional parameters to obtain the initial neutron yield.

[0081] The random variable generation module 402 is used to generate random variables, normalize the random variables, and generate first input parameters.

[0082] The first neutron yield calculation module 403 is used to perform numerical simulation calculation according to the first input parameter to obtain the first neutron yield.

[0083] The loop convergence control module 404 is used to compare the first neutron yield with the initial neutron yield. If the first neutron yield is greater than the initial neutron yield, the initial neutron yield is updated to the first neutron yield, and the number of loops is reset to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of loops is increased by 1; if the number of loops reaches the convergence control number, the initial random walk step length is updated; until the random walk step length is less than or equal to the random walk step length precision, the random walk program ends, and multiple sets of parameters in the optimization process are obtained;

[0084] The initial model building module 405 is used to build an initial input-output model using multiple sets of parameters as initial data.

[0085] The second neutron yield calculation module 406 is used to solve the second input parameter using the acquisition function, and perform numerical simulation calculation according to the second input parameter to obtain the second neutron yield.

[0086] The iterative parameter generation module 407 is used to iteratively optimize the initial input-output model through several groups of second input parameters and second neutron yields to obtain an optimized input-output model.

[0087] The parameter generation module 408 is used to estimate the optimal parameters based on the optimized input-output model.

[0088] In this embodiment, the specific working process and working principle of the initialization module 401, the random variable generation module 402, the first neutron yield calculation module 403, the loop convergence control module 404, the initial model construction module 405, the second neutron yield calculation module 406, the iteration parameter generation module 407 and the parameter generation module 408 are the same as those in the method of embodiment 1, so they are not repeated in this embodiment. Each unit module can be implemented in whole or in part by software, hardware and a combination thereof, and each unit module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above unit modules.

[0089] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0090] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] The above-mentioned embodiments only express several implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A method for designing a laser-driven multi-shell fusion target based on hybrid optimization, characterized in that: The method comprises: Step 1, performing numerical simulation calculation according to the initial random walk parameters and conditional parameters to obtain the initial neutron yield; wherein the initial random walk parameters include the initial random walk step length , Random Walk Step Accuracy and convergence control times The condition parameters include the radiation temperature waveform and the initial parameters of the target pellet structure. ; Step 2, generating a random variable, and normalizing the random variable to generate a first input parameter; Step 3, performing numerical simulation calculation according to the first input parameter to obtain a first neutron yield; Step 4, compare the first neutron yield with the initial neutron yield. If the first neutron yield is greater than the initial neutron yield, update the initial neutron yield to the first neutron yield, and the number of cycles is reset to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles is increased by 1; if the number of cycles reaches the convergence control number, update the initial random walk step length; until the random walk step length is less than or equal to the random walk step length precision, the random walk program ends, and multiple sets of parameters in the optimization process are obtained; Step 5, using the multiple sets of parameters as initial data to construct an initial input-output model; Step 6, using the acquisition function to solve the second input parameter, and performing numerical simulation calculation according to the second input parameter to obtain a second neutron yield; Step 7, iteratively optimizing the initial input-output model through several groups of the second input parameters and the second neutron yield to obtain an optimized input-output model; Step 8: Estimating optimal parameters based on the optimized input-output model.

2. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to claim 1, characterized in that: In step 2, the expression for generating random variables is: ； In the formula, represents a random variable; Represents the radiation temperature waveform and the initial parameters of the pellet structure; Represents a random number between [0, 1].

3. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to claim 2, characterized in that: In step 2, the random variable is normalized to generate the first input parameter, which is expressed as: ； In the formula, Represents the first input parameter; represents the random walk step length.

4. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to any one of claims 1 to 3, characterized in that: In step 4, include: If the first neutron yield is greater than the initial neutron yield, the initial parameter is updated to the first input parameter, and the initial neutron yield is updated to the first neutron yield, and the number of cycles is reset to zero before returning to step 2 to regenerate random variables; If the first neutron yield is less than or equal to the initial neutron yield, and the number of cycles is less than the convergence control number, the initial neutron yield remains unchanged, the number of cycles is increased by 1, and the process returns to step 2 to regenerate random variables; If the first neutron yield is less than or equal to the initial neutron yield, and the number of cycles is equal to the number of convergence control times, and the initial random walk step length is greater than the random walk step length precision, the initial random walk step length is updated to a preset random walk step length, and the process returns to step 2 to regenerate random variables; If the first neutron yield is less than or equal to the initial neutron yield, the number of cycles is equal to the convergence control number, and the random walk step is less than or equal to the random walk step precision, the random walk program ends and multiple sets of parameters in the optimization process are obtained.

5. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to any one of claims 1 to 3, characterized in that: Step 5 includes: The multiple groups of parameters are used as initial data, initial estimation is performed through a Gaussian process, and an initial input-output model is constructed.

6. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to any one of claims 1 to 3, characterized in that: Also includes: By changing the radiation temperature waveform and the initial parameters of the pellet structure , and repeat steps 1 to 8 to optimize the input-output model again.

7. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to any one of claims 1 to 3, characterized in that: The radiation fluid dynamics program MULTI is used for numerical simulation calculations.

8. The method for designing a laser-driven multi-shell fusion target based on hybrid optimization according to any one of claims 1 to 3, characterized in that: A laser-driven multi-shell fusion target design method based on hybrid optimization is used to optimize the structure of single-layer spherical targets or multi-layer spherical targets.

9. A laser-driven multi-shell fusion target design device based on hybrid optimization, characterized in that: The device comprises: An initialization module is used to perform numerical simulation calculations based on initial random walk parameters and conditional parameters to obtain an initial neutron yield; wherein the initial random walk parameters include an initial random walk step length , Random Walk Step Accuracy and convergence control times The condition parameters include the radiation temperature waveform and the initial parameters of the target pellet structure. ; A random variable generation module, used to generate a random variable, normalize the random variable, and generate a first input parameter; A first neutron yield calculation module, used for performing numerical simulation calculation according to the first input parameter to obtain a first neutron yield; A loop convergence control module is used to compare the first neutron yield with the initial neutron yield. If the first neutron yield is greater than the initial neutron yield, the initial neutron yield is updated to the first neutron yield, and the number of cycles is reset to zero; if the first neutron yield is less than or equal to the initial neutron yield, the initial neutron yield remains unchanged, and the number of cycles is increased by 1; if the number of cycles reaches the convergence control number, the initial random walk step length is updated; until the random walk step length is less than or equal to the random walk step length precision, the random walk program ends, and multiple groups of parameters in the optimization process are obtained; An initial model building module, used to build an initial input-output model using the multiple sets of parameters as initial data; A second neutron yield calculation module, used to solve a second input parameter using an acquisition function, and perform numerical simulation calculation according to the second input parameter to obtain a second neutron yield; an iterative parameter generation module, configured to iteratively optimize the initial input-output model through a plurality of groups of the second input parameters and the second neutron yield to obtain an optimized input-output model; The parameter generation module is used to estimate the optimal parameters based on the optimized input-output model.

Images