Optimization method and system for frequency-band-divided high-power microwave source

Abstract

The invention provides an optimization method and system for a frequency-band-divided high-power microwave source, and the method comprises the steps: constructing a data set for describing the structural parameters of a high-power microwave (HPM) source device to be optimized based on the type of the HPM source device, and dividing the working frequency into a plurality of frequency bands; the method comprises the following steps: calculating relevance information of device structure parameters of a plurality of frequency bands by utilizing a Shapley plus interpretation (SHAP) value, encoding the device structure parameters into chromosomes, and generating an initialized population; a genetic operator improved based on an SHAP value is utilized to realize population evolution, elite individuals are found in an iteration mode, and optimization of an HPM source is realized. The method has the characteristics of capability of processing complex parameter relevance and efficient space exploration capability in technical performance, and has the advantages of good interpretability, visual presentation, adaptability, flexibility and the like in the application level. Therefore, the method becomes a powerful prototype optimization tool and is suitable for various prototype optimization design problems needing to consider parameter relevance and hierarchical structures.

Description

Technical Field

[0001] This invention belongs to the field of high-power microwave source optimization technology, and specifically relates to a method and system for optimizing a frequency-divided high-power microwave source. Background Technology

[0002] High-power microwave (HPM) sources operate based on the complex nonlinear interaction between electromagnetic fields and charged particles, generating high-power microwaves with operating frequencies ranging from 1 to 300 GHz and output power exceeding 100 MW. High-power microwave sources are the core components for generating high-power microwaves and are widely used in military, civilian, and scientific research fields.

[0003] Since the introduction of HPM (High-Performance Particle) source devices, researchers both domestically and internationally have primarily focused on physical mechanisms, proposing new research ideas through theoretical analysis, designing device prototypes, and optimizing devices by combining particle-in-cell (PIC) simulation methods with experimental verification. In recent years, with the continuous development of HPM technology, numerous new theories and device performance indicators have been proposed, aiming to help researchers design HPM sources with superior performance. However, as researchers' performance requirements for devices continue to increase, the design difficulty of HPM sources is also constantly increasing, posing new challenges to traditional HPM source design methods based solely on theoretical analysis and PIC numerical simulations.

[0004] Traditional device prototyping methods have significant shortcomings, specifically: 1. It relies heavily on the designer's subjective consciousness and personal preferences. This may lead to design results that are limited by personal preferences and experience, lacking objectivity and comprehensiveness; 2. It is severely limited by the technical and knowledge level of the designers. This may result in a lack of innovation and originality in the design, failing to fully realize the design's potential; 3. It typically requires a significant amount of time and resources. From concept to actual production, multiple iterations and adjustments are needed, resulting in a long design cycle and high design costs; 4. Lack of automation and optimization capabilities. Manual device optimization is typically limited to local search and local optimization. Designers may only be able to improve the design through limited trial and error and adjustments, unable to fully explore the design space and find the global optimum. HPM devices have complex structures, numerous parameters, and significant nonlinear characteristics, making it difficult to find the global optimum using only manual methods.

[0005] 5. Lack of adaptive optimization capabilities based on the characteristics of different frequency bands. The importance of the structural and performance parameters of devices varies greatly in each frequency band. In the optimization process of traditional algorithms, the optimization direction of the device is generally global and tends to change randomly, resulting in low optimization efficiency.

[0006] Genetic algorithms are a commonly used parameter optimization algorithm. This method is based on the mechanism of natural genetics and can be effectively applied to solving complex nonlinear problems. Due to its high efficiency and robustness, it has been widely used in solving optimization problems.

[0007] In recent years, researchers have attempted to combine evolutionary optimization algorithms (such as genetic algorithms and particle swarm optimization) with PIC simulation methods to assist in the optimal design of HPM sources, and have achieved some results. Traditional genetic algorithms mainly obtain superior individuals in the offspring through selection and mutation, but these individuals may be damaged during the mutation process.

[0008] Traditional genetic algorithms, limited by their genetic operators, can only fine-tune the device structure within a global range of parameter variations when applied to HPM source optimization design. They cannot optimize based on the specific frequency band characteristics of the structural parameters, resulting in slow convergence and low optimization efficiency, which significantly restricts the efficiency of HPM source device optimization design. Therefore, traditional genetic algorithms are unsuitable for the requirements of HPM source device prototype optimization design, and there is an urgent need for an algorithm that can optimize based on the current frequency band range of the device during the global optimization process. Summary of the Invention

[0009] The purpose of this invention is to provide an optimization method and system for a high-power microwave source divided into frequency bands, so as to solve the problem that the optimization cannot be assisted by the structural parameter characteristics of specific frequency bands, resulting in slow convergence speed and low optimization efficiency.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an optimization method for a frequency-divided high-power microwave source, comprising: Based on the type of high-power microwave (HPM) source device to be optimized, a dataset describing the characteristics of this type of HPM source device is constructed, and the operating frequency is divided into multiple frequency bands; The correlation information of device structure parameters in multiple frequency bands was calculated using Shapley Additive Explanations (SHAP) values, and the device structure parameters were encoded into chromosomes to generate an initial population. By utilizing genetic operators based on SHAP values, population evolution is achieved, and elite individuals are found through iteration to optimize the HPM source.

[0011] Furthermore, based on the type of high-power microwave (HPM) source device to be optimized, a dataset describing the characteristics of that type of HPM source device is constructed, including: Based on the characteristics of the HPM source device to be optimized, a dataset of device samples is collected. This dataset includes the device's structural parameters, performance parameters to be optimized, and corresponding operating frequency information.

[0012] Furthermore, the division of the operating frequency into multiple frequency bands includes: Based on the operating frequency range and performance characteristics of the devices, the operating frequency is divided into several frequency bands, and the device samples in each frequency band have similar structural parameter characteristics.

[0013] Furthermore, the correlation information of device structure parameters across multiple frequency bands calculated using Shaplectic Interpretation (SHAP) values ​​includes: For device samples within each frequency band after segmentation, the SHAP values ​​are used to calculate their parameter correlation results, including the SHAP main effect and interaction effect. The main effect is the independent impact of each parameter on device performance. The interaction effect is the impact when it works together with other parameters, in order to determine the relative importance and direction of influence of each parameter in different frequency bands.

[0014] Furthermore, the process of encoding the device structural parameters into chromosomes and generating an initial population includes: By utilizing the range and precision of parameter changes, parameter values ​​are determined using a uniform random algorithm, and individual population encoding is performed using a tree-like chromosome encoding method, thus achieving population initialization.

[0015] Furthermore, device performance parameters are predicted using PIC numerical simulation methods, and individual fitness is evaluated using a multi-objective fitness evaluation method based on a fuzzy simple added weighting system (FSAWS) (hereinafter referred to as the FSAWS method), including: A concrete numerical model of the device described by the individual population is performed, and the simulation is carried out using the particle simulation PIC numerical simulation software. Based on the numerical simulation results and the pre-set performance index rating criteria, the fitness of the individual is evaluated using the FSAWS method.

[0016] Furthermore, the use of genetic operators based on SHAP values ​​to achieve population evolution, finding elite individuals through iteration, and optimizing the HPM source includes: Using gene crossover and gene mutation operators suitable for tree-like chromosomes, a new population is generated to achieve population genetic evolution. During the iterative evolution process of the genetic algorithm, an improved genetic operator based on SHAP values ​​is introduced. According to the frequency band of the performance parameters of the device corresponding to the individual in the current population, the parameter correlation SHAP value results of that frequency band are found. According to the importance magnitude and direction of the parameters, corresponding gene crossover and gene mutation operators are designed. Elite individuals are found through iteration to achieve prototype optimization design of HPM source. Based on the updated new population, the above optimization process is repeated until the maximum number of iterations is reached or a device prototype structure that meets the design requirements is found.

[0017] Secondly, the present invention provides an optimization system for a frequency-divided high-power microwave source, comprising: The data acquisition module is used to construct a dataset describing the characteristics of the high-power microwave (HPM) source device type to be optimized, and to divide the operating frequency into multiple frequency bands. The initialization module is used to calculate the correlation information of device structure parameters in multiple frequency bands using Shaplectic Interpretation (SHAP) values, encode the device structure parameters into chromosomes, and generate an initial population. The optimization module is used to achieve population evolution by using genetic operators based on SHAP values, and to find elite individuals through iteration to optimize the HPM source.

[0018] Thirdly, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the optimization method for a frequency-division high-power microwave source.

[0019] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the optimization method for a frequency-division high-power microwave source.

[0020] Compared with the prior art, the present invention has the following technical effects: This invention uses parameter correlation analysis to analyze the correlation between parameters, thereby representing the hierarchical structure and dependencies of parameters in depth. By introducing this physical parameter correlation into the optimization process, it assists in the optimization of devices under specific conditions. This invention rationally divides the operating frequency into frequency bands and uses the SHAP method to deeply analyze the correlation of device parameters in each frequency band. This allows the optimization process to no longer be limited to a general global scope, but to perform precise optimization based on the characteristics of different frequency bands. This effectively overcomes the limitations of traditional genetic algorithms in dealing with the complex characteristics of multi-frequency bands of HPM source devices, and improves the targeting and effectiveness of optimization. In the genetic algorithm optimization process, this invention, by accurately grasping the importance and direction of parameters in each frequency band, enables the next generation of individuals to perform targeted optimization adjustments based on key parameters. This targeted guidance avoids a large amount of invalid exploration and redundant computation during the search process of traditional genetic algorithms, significantly improving the convergence speed of the algorithm and finding a better device structure scheme in a shorter time, thus greatly improving the overall efficiency of HPM device optimization design.

[0021] This invention, based on SHAP value analysis, not only clarifies the main and interaction effects of each parameter on device performance, but also presents these complex parameter relationships in intuitive and visual charts. This enables designers to clearly understand the inherent logical connections between device structural parameters and performance, providing a strong basis for subsequent optimization decisions and enhancing the confidence and operability of the optimization process and results.

[0022] This invention uses a relativistic backward wave tube (RBWO) device as an example, but its core concept and architecture are universal and can be flexibly extended to the optimization design of other types of HPM source devices. Furthermore, the methods for dividing different frequency bands, the depth and dimensions of parameter correlation analysis, etc., can be flexibly adjusted according to the actual needs and design goals of specific devices, adapting to various complex HPM source device optimization scenarios.

[0023] This invention introduces the FSAWS multi-objective fitness evaluation method based on a fuzzy simple weighted system, which can be applied to multi-objective prototype optimization design scenarios of HPM source devices.

[0024] In summary, the global band division HPM source device optimization design method based on SHAP values ​​possesses the technical performance characteristics of handling complex parameter correlations and efficient spatial exploration capabilities. At the application level, it offers advantages such as good interpretability, visualization, adaptability, and flexibility. This makes this method a powerful prototyping optimization tool suitable for various prototyping optimization design problems that require consideration of parameter correlations and hierarchical structures. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the structural parameters of a relativistic backward wave tube type device.

[0026] Figure 2 This is a schematic diagram of chromosome coding.

[0027] Figure 3 This is an example of a parent chromosome in a crossover.

[0028] Figure 4 This is a cross example based on the main effects of SHAP values.

[0029] Figure 5 This is an example of a feature interaction matrix based on SHAP value interaction in the crossover operator.

[0030] Figure 6 This is a cross example of main effects and interactions based on SHAP values.

[0031] Figure 7 This is an example of a mutated parent chromosome.

[0032] Figure 8 This is an example of variation based on the main effects of SHAP values.

[0033] Figure 9 This is an example of a feature interaction matrix based on SHAP value interactions in mutation operators.

[0034] Figure 10 This is an example of variation based on the main effects and interactions of SHAP values.

[0035] Figure 11 This is a flowchart of a high-power microwave source prototype design method based on frequency band SHAP value optimization. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings: Example 1, please refer to Figure 11 This invention provides an optimization method for a frequency-divided high-power microwave source, comprising: Based on the type of high-power microwave (HPM) source device to be optimized, a dataset describing the characteristics of this type of HPM source device is constructed, and the operating frequency is divided into multiple frequency bands; The correlation information of device structure parameters in multiple frequency bands is calculated using Shaplectic Interpretation (SHAP) values, and the device structure parameters are encoded into chromosomes to generate an initial population. By utilizing genetic operators based on SHAP values, population evolution is achieved, and elite individuals are found through iteration to optimize the HPM source.

[0037] This invention divides the operating frequency into frequency bands and combines the SHAP value analysis results of each frequency band to enable the optimization process to fully consider the characteristics of device structural parameters in different frequency bands, achieving precise optimization and overcoming the problem that traditional genetic algorithms cannot take into account the characteristics of local frequency bands in global optimization. By using the parameter importance and influence direction determined by the SHAP value, a clear directional guide is provided for the evolution of the genetic algorithm, avoiding a large amount of blind search and random trial in traditional algorithms, and significantly improving optimization efficiency and convergence speed. SHAP value analysis can intuitively show the influence of various parameters on device performance, providing interpretability and visualization support for the optimization process. It helps designers to deeply understand the working principle and optimization mechanism of the device, and further improves the scientificity and effectiveness of optimization decisions. By using PIC numerical simulation to predict device performance parameters and employing the FSAWS method for fitness evaluation, the optimization process and results can be interpreted and understood. This helps designers assess the reliability and feasibility of the optimization results.

[0038] Example 2: This invention provides an optimization method for a frequency-divided high-power microwave source, specifically including: The purpose of this invention is to provide a global frequency band optimization design method for HPM sources. This method introduces SHAP value analysis to accurately grasp the correlation between device parameters in different frequency bands; it improves the crossover and mutation operators to enable precise optimization based on the characteristics of device parameters in different frequency bands; and it introduces a multi-criteria decision analysis method based on a fuzzy simple weighted system to achieve multi-objective fitness evaluation.

[0039] A global frequency band HPM source optimization design method based on SHAP values. The method includes: Based on the HPM source device type to be optimized, a dataset describing the characteristics of this type of HPM source device is constructed, and the operating frequency is divided into multiple frequency bands; The correlation information of device structural parameters in multiple frequency bands is calculated using SHAP values, and the structural parameters are encoded into chromosomes to generate an initial population. Based on the device structure described by individual populations, the PIC numerical simulation method is used to predict device performance parameters, and the FSAWS method is used to evaluate individual fitness. Population evolution is achieved using genetic operators based on SHAP values; elite individuals are found through iterative methods to achieve optimized design of HPM source.

[0040] Based on the type of HPM source device to be optimized, a dataset describing the structural parameters of that type of HPM source device is constructed, and the operating frequency is divided into multiple frequency bands. According to the characteristics of the HPM source device to be optimized, a dataset of relevant device samples is collected. This dataset covers information such as the device's structural parameters, the performance parameters to be optimized, and the corresponding operating frequency. Combining the device's operating frequency range and performance characteristics, the operating frequency is divided into several frequency bands to ensure that device samples within each frequency band have similar structural parameter characteristics. The process involves using SHAP values ​​to calculate the correlation information of device structural parameters across multiple frequency bands, encoding these parameters into chromosomes, and generating an initial population. First, for device samples within each frequency band, SHAP values ​​are used to calculate the parameter correlation results, including SHAP main effects and interaction effects. Specifically, for each parameter, its independent impact on device performance (main effect) and its impact when interacting with other parameters (interaction effect) are calculated, thereby determining the relative importance and direction of influence of each parameter in different frequency bands. The initial population is generated by using the parameter variation range and precision information, employing a uniform random algorithm to assign parameter values, and then encoding individual population members using a tree-like chromosome encoding method, thus achieving population initialization. The method uses PIC numerical simulation to predict device performance parameters and FSAWS method to evaluate individual fitness. It performs a concrete numerical model of the device described by the individual population, performs simulation using PIC numerical simulation software, and evaluates individual fitness using FSAWS method based on the numerical simulation results and pre-set performance index rating criteria. The process utilizes genetic operators improved based on SHAP values ​​to achieve population evolution. Elite individuals are identified iteratively to optimize the design of the HPM source. Gene crossover and mutation operators suitable for tree-like chromosomes are used to generate a new population for genetic evolution. Improved genetic operators based on SHAP values ​​are introduced during the iterative evolution of the genetic algorithm. The SHAP values ​​of parameter correlation for each individual in the current population are searched based on the frequency band of the device's performance parameters. Corresponding gene crossover and mutation operators are designed according to the importance and direction of the parameters. For example, for highly important parameters, their weight and magnitude in crossover and mutation operations are increased to guide new individuals towards directions that better improve device performance. The process of finding elite individuals iteratively enables the prototype optimization design of the HPM source. Based on the updated new population, the above optimization process is repeated until the maximum number of iterations is reached or a device prototype structure that meets the design requirements is found.

[0041] Example 3: This invention provides an optimization method for a frequency-divided high-power microwave source, comprising: (1) Perform structural parameterization on the HPM source device to be optimized. Taking a relativistic backward wave tube as an example, the device structure is parameterized, such as... Figure 1 As shown, r 0、 r 1. r 2. r 3. r 4. d 0、 z 0、 z 1. z 2. z All three are independent parameters, while the remaining parameters are combined parameters. n r The number of reflecting cavities. h ri1 , w ri1 , h ri2 and w ri2 For the first i The structural dimensions of each reflecting cavity (0≤ i ≤ n r ); n s The number of slow-wave structures. h sj1 , w sj1 , h sj2 and w sj2 For the first j The structural dimensions of each slow wave (0≤ j ≤ n s ); n e To extract the number of cavities, h ek1 , w ek1 , h ek2 and w ek2 For the first k The structural dimensions of each extraction cavity (0≤ k ≤ n e The parameter variation range and precision are set as shown in Table 1. Taking the parameters described by the reflective cavity structure as an example, then { n r , h ri1 , w ri1 , hri2 , w ri2} represents a set of combined parameters. Among these combined parameters... n r These are advanced parameters used to determine the number of reflecting cavities and the number of structural parameters for each reflecting cavity.

[0042] Table 1

[0043] (2) Dataset processing and frequency band allocation Based on the characteristics of the HPM source device to be optimized, a dataset of relevant device samples is collected, i.e. D = { X , Y}.in X = { x 1, x 2, …, x m}, x i It is the first of the devices i Each structural parameter, x i = { x i1 , x i2 , … , x in}, m It is the number of structural parameters. Y= { y 1, y 2, …, y p}, y q It is the device number q One performance parameter, y q = { y q1 , y q2 , … , y qn}, p It refers to the number of performance parameters. n Number of sample data. This dataset covers the structural parameters of the devices and the corresponding performance parameters (output power, operating frequency, etc.) calculated using PIC. Combining the operating frequency range and performance characteristics of the devices, the operating frequency is divided into several frequency bands to ensure that the device samples within each band have relatively consistent structural parameter characteristics. Assuming the frequency range is divided into K frequency bands, the range of the kth frequency band is [freq_start].k freq_end k (k = 1, 2, ..., K).

[0044] (3) SHAP value calculation For device samples within each frequency band after segmentation, the SHAP method is used to calculate their parameter correlation results, including SHAP main effects and interaction effects. For the [missing information - likely a specific frequency band]... i The parameter is calculated for the first parameter. q Independent effects on the performance of individual devices (main effects) And the effects when it interacts with other parameters (interaction effect). (i≠j), thus determining the relative importance and direction of influence of each parameter in different frequency bands. For the i-th i The parameters have a total SHAP value of: (1) In the formula: F —All collections in the dataset, S — Not included i a subset of f q ( S — Using subsets S The q Predicted values ​​of device performance parameters.

[0045] The main effect is calculated by excluding the interaction effects of other features, and the formula is: (2) In the formula: ——No. i The parameter corresponds to the first... q The total SHAP value of each performance parameter ——No. i The parameter corresponds to the first... q The main effects of each performance parameter ——No. i The parameter and the first j The parameter corresponds to the first... q The interaction effect of performance parameters.

[0046] The interaction effect describes the additional contribution when two parameters act together. It is calculated using the Shaley interaction index: (3) (4) In the formula: —Introduced at the same time i and j The incremental effect. The interaction effect is symmetrically distributed, that is... .

[0047] By calculating the interaction effect value, the strength of the interaction between two features can be quantified.

[0048] (4) Structural parameter encoding and population initialization Assuming the population contains N Individual, denoted as x l (0≤) l ≤ N Using the range and precision of parameter variations, a uniform random algorithm is used to determine parameter values, and individual population encoding is performed based on the floating-point chromosome encoding method. Figure 2 As shown, population initialization is implemented. x l The chromosomes are denoted as: (5) in, , (6) , (7) , (8) The encoded individuals formed the initial population.

[0049] (5) PIC numerical simulation By utilizing the specific values ​​of device structural parameters within the population, a concrete numerical model of the device is constructed. The performance parameters of interest are clearly defined, such as operating frequency, output power, frequency purity, energy conversion efficiency, and particle throughput. Numerical simulations of each device structure are performed using PIC numerical simulation software to predict the performance parameters of all devices described by the population. Individual fitness is calculated using the individual fitness evaluation method described below, and population evolution is achieved through an iterative approach to obtain elite individuals.

[0050] (6) Individual fitness assessment The target weight W is determined based on the priority of the device performance to be optimized. The weight setting is divided into 5 levels with reference to the FSAWS method, as shown in Table 3.

[0051] Table 3

[0052] Evaluations are conducted based on the type of device performance parameters. Assumptions This is a description function for the performance parameters of range-type devices. This is the evaluation function for interval-type performance parameters. The interval-type performance parameters are evaluated using a graded approach, with nine levels based on the FSAWS method, as shown in Table 4.

[0053] Table 4

[0054] Trend-based performance parameters can be divided into two categories: positive trend indicators and negative trend indicators. Among them, positive trend indicators are better when their relative values ​​are relatively large, while negative trend indicators are better when their relative values ​​are relatively small.

[0055] Assumption For describing the performance parameters of trend-based devices, for The evaluation function, the set to be evaluated contains a total of p Group results.

[0056] If the trend indicator is a positive trend indicator, then a relatively large value is preferred: , , (9) If the trend target is a negative trend indicator, a relatively small value is preferred: , , (10) Based on the above objective evaluation method, the fitness of an individual can be determined as follows: (11) in, W Weights for performance parameters, F s For rating interval-type performance parameters, F o Rating for trend-based performance parameters. For individual fitness.

[0057] (7) Population evolution Population evolution is achieved using genetic operators based on improved SHAP values. The SHAP values ​​of parameter correlation within the frequency band corresponding to the performance parameters of the devices associated with individuals in the current population are retrieved. Corresponding gene crossover and mutation operators are designed according to the magnitude and direction of parameter importance.

[0058] 1) Gene selection The roulette wheel algorithm is used to select individuals to serve as parents for reproduction. Using the selected parents, the main effects and interaction strengths of their SHAP values ​​are used to improve gene crossover and gene mutation operators, generating a new population while retaining the elite individuals from the original population.

[0059] 2) Gene crossover operator Assume the parent chromosomes that need to be crossed are respectively P 1 and P 2, such as Figure 3 As shown, gene loci on chromosomes i The first corresponding device i The encoding of each structural parameter. Then: First, using the operating frequency of the device described by the parent chromosome as the key value, we find the characteristic importance between the structural parameters and performance parameters within the operating frequency range. Normalization is performed using equation (12) to obtain the normalized gene crossover probability W, which takes into account the importance of features. (1) c =[W 1 c W 2 c W 3 c W 4 c W 5 c W 6 c W 7 c W 8 c ].

[0060] (12) In the formula, k c This represents the preset fixed gene crossover probability. W q Indicates the first q The evaluation weight of each performance parameter, Indicates the first i The probability of crossover between genes.

[0061] Normalization is performed using equation (13) to obtain the normalized gene crossover probability W, which takes into account the feature interactions between structural parameters. (2) c =[[W 11 c W 12 c , …, W 18 c ], [W 21 c W 22 c , …, W 28 c ],…, [W81 c W 82 c ,…, W 88 c ]).

[0062] (13) In the formula, W q Indicates the first q The evaluation weight of each performance parameter, Indicates the first i When the first gene crosses over, the... j The probability of crossover between genes.

[0063] For each gene locus i ,use As the crossover probability, each gene will undergo crossover operation according to this probability, such as... Figure 4 As shown. When p i <W i c At that time, P 1 and P 2nd i The crossover operation is performed at the position of each gene; when p i ≥W i c If the crossover occurs, the crossover operation will not be performed. Green indicates that crossover has occurred, and black indicates that crossover has not occurred.

[0064] Next, considering the interaction effects between structural parameters, if a crossover occurs at a gene locus, it is necessary to further select another pair of genes that interact with the structural parameters for crossover operation, according to the weighted probability distribution of the feature interaction matrix between structural parameters. Figure 4 Taking the first gene as an example, p 1 < W 1 c Then, gene pairs at that locus undergo crossover, according to Figure 5 The feature interaction matrix W shown (2) c The weighted probability distribution is obtained. The part marked in green indicates that the gene at that position has interacted with the gene undergoing crossover, while the rest indicates that genes at other positions have not interacted with that gene. Therefore, a crossover operation is also performed on the gene at the position marked in green, such as... Figure 6 As shown, the green part represents the gene loci where crossover originally occurred, and the red part represents the gene loci that performed the crossover operation based on the interaction.

[0065] chromosomeP 1 and P 2. Two heterogeneous device schemes are characterized respectively. Their gene loci and device structural parameters form a strict spatial mapping relationship. In the crossover operation stage, a gene exchange mechanism under spatial position constraints is adopted, that is, only homologous gene loci are allowed to exchange information.

[0066] 3) Gene mutation operator Assume the parent chromosome that needs to be mutated is P 1, such as Figure 7 As shown. Then: Firstly, according to P Using the operating frequency corresponding to the structural parameter 1 as the key value, we can identify the characteristic importance between structural parameters and performance parameters within the operating frequency range. Normalization is performed using equation (14) to obtain the normalized gene mutation probability W that takes into account the importance of features. (1) m =[W 1 m W 2 m W 3 m W 4 m W 5 m W 6 m W 7 m W 8 m ].

[0067] (14) In the formula, k m This represents a preset, fixed probability of gene mutation. W q Indicates the first q The evaluation weight of each performance parameter, Indicates the first i The probability of a gene mutating.

[0068] Normalization was performed using equation (15) to obtain the normalized gene mutation probability W, which takes into account the interaction between structural parameters. (2) m =[[W 11 m W 12 m , …, W 18 m ], [W 21 m W22 m , …, W 28 m ],…, [W 81 m W 82 m ,…, W 88 m ]).

[0069] (15) In the formula, W q Indicates the first q The evaluation weight of each performance parameter, Indicates the first i When the first gene mutates, the second... j The probability that a gene may also mutate.

[0070] For each gene locus, the following is adopted: As the mutation probability, each gene will undergo mutation operation according to this mutation probability, such as... Figure 8 As shown. If and only if p i <W i m When the gene is within its current range of variation, a random mutation is performed to generate a new allele; otherwise, the original genotype is retained. Green indicates a gene mutation, and black indicates no mutation.

[0071] Next, consider the interaction effects between structural parameters. If a mutation is performed at one gene locus, it is necessary to further select another gene that interacts with the structural parameter and perform a mutation based on the weighted probability distribution of the feature interaction matrix between structural parameters. Assume that... x If a mutation occurs at 7 gene loci, then according to Figure 9 The feature interaction matrix W shown (2) m The weighted probability selects another gene locus, specifically the position in the 7th row and 1st column (highlighted in green). This indicates that because gene locus 7 has mutated, the same mutation operation will also be performed on gene locus 1. Figure 10 As shown, the green part represents the gene loci that originally underwent mutation, and the red part represents the gene loci that performed mutation operations based on interactions.

[0072] 4) Generate new populations Repeat the above process to build a new population using the newly generated offspring, and use these new individuals as candidate solutions for the next generation for further optimization and search.

[0073] (8) Repeat steps 4 to 7 until the maximum number of iterations is reached or a solution that meets the design requirements is found.

[0074] (9) Output the optimal solution The individual with the highest fitness in the final population is selected as the optimal solution, which is the prototype optimization design result of the high-power microwave source.

[0075] In another embodiment of the present invention, an optimization system for a frequency-divided high-power microwave source is provided, which can be used to implement the above-described optimization method for a frequency-divided high-power microwave source. Specifically, the system includes: The data acquisition module is used to construct a dataset describing the structural parameters of the high-power microwave (HPM) source device type to be optimized, and to divide the operating frequency into multiple frequency bands. The initialization module is used to calculate the correlation information of device structure parameters in multiple frequency bands using Shaplectic Interpretation (SHAP) values, encode the device structure parameters into chromosomes, and generate an initial population. The optimization module is used to achieve population evolution by using genetic operators based on SHAP values, and to find elite individuals through iteration to optimize the HPM source.

[0076] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0077] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of an optimization method for a frequency-divided high-power microwave source.

[0078] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the optimization method for a frequency-division high-power microwave source in the above embodiments.

[0079] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a 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, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0081] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. An optimization method for a frequency-band-divided high-power microwave source, characterized in that, include: Based on the type of high-power microwave HPM source device to be optimized, a dataset describing the structural parameters of this type of HPM source device is constructed, and the operating frequency is divided into multiple frequency bands; The correlation information of device structure parameters in multiple frequency bands was obtained by using Shaplectic interpretation of SHAP values, and the device structure parameters were encoded into chromosomes to generate an initial population. By utilizing genetic operators based on SHAP values, population evolution is achieved, and elite individuals are found through iteration to optimize the HPM source.

2. The optimization method for a frequency-band-divided high-power microwave source according to claim 1, characterized in that, Based on the type of high-power microwave HPM source device to be optimized, a dataset describing the characteristics of this type of HPM source device is constructed, including: Based on the characteristics of the HPM source device to be optimized, a dataset of device samples is collected. This dataset includes the device's structural parameters, performance parameters to be optimized, and corresponding operating frequency information.

3. The optimization method for a frequency-band-divided high-power microwave source according to claim 2, characterized in that, The division of the operating frequency into multiple frequency bands includes: Based on the operating frequency range and performance characteristics of the devices, the operating frequency is divided into several frequency bands, and the device samples in each frequency band have similar structural parameter characteristics.

4. The optimization method for a frequency-band-divided high-power microwave source according to claim 1, characterized in that, The correlation information of device structure parameters in multiple frequency bands calculated using the Shaplectic interpretation SHAP value includes: For device samples within each frequency band after segmentation, the SHAP values ​​are used to calculate their parameter correlation results, including the SHAP main effect and interaction effect. The main effect is the independent impact of each parameter on device performance. The interaction effect is the impact when it works together with other parameters, in order to determine the relative importance and direction of influence of each parameter in different frequency bands.

5. The optimization method for a frequency-band-divided high-power microwave source according to claim 4, characterized in that, The process of encoding device structural parameters into chromosomes and generating an initial population includes: By utilizing the range and precision of parameter changes, parameter values ​​are determined using a uniform random algorithm, and individual population encoding is performed using a tree-like chromosome encoding method, thus achieving population initialization.

6. The optimization method for a frequency-band-divided high-power microwave source according to claim 5, characterized in that, The performance parameters of devices are predicted using the PIC numerical simulation method, and the fitness of individuals is evaluated using a multi-objective fitness evaluation method based on fuzzy simple weighted system (FSAWS). The devices described by the individuals in the population are modeled numerically, and the simulation is performed using PIC numerical simulation software. Based on the numerical simulation results and the pre-set performance index rating criteria, the fitness of individuals is evaluated using the FSAWS method.

7. The optimization method for a frequency-band-divided high-power microwave source according to claim 1, characterized in that, The method of using genetic operators based on SHAP values ​​to achieve population evolution, finding elite individuals through iteration, and optimizing the HPM source includes: Using gene crossover and gene mutation operators suitable for tree-shaped chromosomes, a new population is generated to achieve population genetic evolution. In the iterative evolution process of the genetic algorithm, an improved genetic operator based on SHAP value is introduced. According to the frequency band of the performance parameters of the device corresponding to the individual in the current population, the parameter correlation SHAP value results of the frequency band are found. According to the importance magnitude and direction of the parameters, corresponding gene crossover and gene mutation operators are designed. Elite individuals are found through iteration to realize the prototype optimization design of HPM source. Based on the updated new population, the above optimization process is repeated until the maximum number of iterations is reached or a device prototype structure that meets the design requirements is found.

8. An optimization system for a frequency-band high-power microwave source, characterized in that, include: The data acquisition module is used to construct a dataset describing the characteristics of the high-power microwave HPM source device type to be optimized, and to divide the operating frequency into multiple frequency bands. The initialization module is used to calculate the correlation information of device structure parameters in multiple frequency bands using Shaplectic interpretation of SHAP values, encode the device structure parameters into chromosomes, and generate an initial population. The optimization module is used to achieve population evolution by using genetic operators based on SHAP values, and to find elite individuals through iteration to optimize the HPM source.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the optimization method for a frequency-division high-power microwave source 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 the processor, it implements the steps of the optimization method for a frequency-divided high-power microwave source as described in any one of claims 1 to 7.

Images