A parallel multi-objective optimization method for traveling wave tube components

Abstract

The present invention relates to the field of traveling wave tube design and optimization, and specifically to a parallel multi-objective optimization method for traveling wave tube components. The present invention optimizes and sets the optimization decision variables and optimization objectives for traveling wave tube components, introduces a general-purpose parallel computing processor to adaptively perform parallel calculations on various traveling wave tube component projects and obtain corresponding simulation calculation results, and finally performs non-dominated sorting and updates the optimal solutions and population evolution. Based on the present invention, the traveling wave tube components can be designed and optimized more flexibly, with a higher degree of freedom in optimal design, and further iteratively update the design optimization and development technology of traveling wave tubes; effectively solves the problem of low efficiency in the existing multi-objective optimization method for traveling wave tube components, which requires separate serial optimization execution for each calculation project.

Description

Technical Field

[0001] The present invention relates to the field of traveling wave tube design and optimization, and specifically to a parallel multi-objective optimization method for traveling wave tube components. Background Art

[0002] Traveling wave tubes are widely used in military applications, communication systems and other fields due to their characteristics such as wide operating frequency band, high output power and high efficiency. A traveling wave tube consists of an electron gun, a high-frequency circuit, a collector, a magnetic system, and an input / output window structure. During the design process of a traveling wave tube, multi-objective optimization problems will be encountered for each component of the traveling wave tube.

[0003] For example, for the electron gun that generates an electron beam, it is necessary to optimize the cathode radius and the distance between the cathode and the anode to obtain a larger total cathode emission current and a smaller beam waist radius, so as to generate more electron beam energy; the high-frequency circuit that provides dispersion characteristics needs to optimize the structural parameters to obtain a more flat dispersion characteristic within the operating frequency band, and a better coupling impedance is required to improve the electron efficiency of the beam-wave interaction; after the beam-wave interaction is completed, the collector needs to optimize the structure and the voltages of each stage to meet higher recovery efficiency and collector efficiency. The optimization decision variables and optimization objectives of different components in a traveling wave tube are different, and there is still a lack of a unified optimization architecture that can perform multi-objective optimization for all components of the traveling wave tube.

[0004] Moreover, in the previous multi-objective optimization methods for traveling wave tube components, each calculation project needs to be executed separately, and the next calculation project can only be executed after one calculation project is completed. This calculation method is a serial optimization calculation method. This serial optimization method has a slow calculation speed and will take more time. Summary of the Invention

[0005] In view of the above problems or deficiencies, in order to solve the problem of slow serial optimization calculation speed, and at the same time form a unified architecture that can perform multi-objective optimization for all components of the traveling wave tube and improve the flexibility of the optimization design of traveling wave tube components, the present invention proposes a parallel multi-objective optimization method for traveling wave tube components.

[0006] A parallel multi-objective optimization method for traveling wave tube components is as follows:

[0007] Step1. Set the optimization objectives of the traveling wave tube components;

[0008] According to the design requirements of the traveling wave tube components, set the optimization decision variables of the traveling wave tube components in the optimization setting interface, including whether the traveling wave tube optimization decision variables are optimized, the lower limit value and the upper limit value of the optimization decision variables; set the optimization objectives of the traveling wave tube components, including the optimization objective value and the optimization direction. The optimization direction includes two directions: greater than and less than.

[0009] Specifically, the optimization decision variables of the electron gun include the cathode position, anode position, and electron gun voltage of the electron gun; the optimization objectives of the electron gun include the total cathode emission current, the cathode back-bombardment current, the intercepted current, the beam waist radius, and the beam waist position.

[0010] Specifically, the optimization decision variables of the high-frequency structure include the inner radius and pitch of the helix; the optimization objectives of the high-frequency structure include the normalized phase velocity, the coupling impedance, and the attenuation constant.

[0011] Specifically, the optimization decision variables of the collector include the collector entrance energy distribution, the collector structure parameters, and the electrode voltages of each electrode of the collector; the optimization objectives of the collector include the collector efficiency and the reflux rate.

[0012] Specifically, the optimization decision variables of the magnetic system include the structural parameters of the magnet steel, the structural parameters of the pole shoe, and the material; the optimization objectives of the magnetic system include the magnetic field period, the peak on-axis magnetic field, and the magnetic system performance parameters.

[0013] Specifically, the optimization decision variables of the input / output window include the structural design parameters of the input / output window; the optimization objectives of the input / output window include the S-parameters and the standing wave ratio.

[0014] Step2. Set the optimization algorithm;

[0015] Select a multi-objective optimization algorithm, including the NSGA-II or MOPSO optimization algorithm, and set the corresponding optimization algorithm parameters according to the selected different optimization algorithms.

[0016] Specifically, the parameters of the NSGA-II multi-objective optimization algorithm include the maximum number of iterations N Iter , the size of the population N Pop , the crossover rate d cross , the mutation rate d mutate and the variable precision N ps .

[0017] The parameters of the MOPSO multi-objective optimization algorithm include the maximum number of iterations N Iter , the size of the population N Pop , the inertia weight factor d w , the individual learning factor d c1 , the mutation rate d c2 and the variable precision N ps .

[0018] Specifically, the mentioned variable precision N ps is used to constrain the number of significant digits of the optimization decision variables to be N ps digits after the decimal point, that is, when N ps = 2, it means that the number of significant digits of the optimization decision variables is 2 digits after the decimal point.

[0019] Step3. Generate the initial population of the optimization algorithm;

[0020] Randomly generate N optimization algorithm individuals within the parameter range of the optimization decision variables of the traveling wave tube component according to the optimization decision variables and optimization objectives of the traveling wave tube component set in Step1 and the optimization algorithm set in Step2, generate the initial population Pop(I pop ), and record I Iter = 0. Iter

[0021] Specifically, use I Iter to represent the current iteration number of the optimization algorithm, that is, I Iter = 0 indicates that the current population is the initialized population, and I Iter = 1 indicates that the current population Pop(I Iter ) is the population of the first generation.

[0022] Step4. Based on the current population Pop(I Iter ), send an instruction of "start parallel computing" to the parallel computing processor.

[0023] Step5. The parallel computing processor starts to execute;

[0024] The parallel computing processor receives the instruction of "start parallel computing" in Step4, identifies the calculation engineering type of the traveling wave tube component, and the parallel computing processor generates the set corresponding number of processes N Pop according to the population size N Thread in Step2, that is, N Thread = N pop and correspondingly updates the optimization decision variables of the traveling wave tube component, and then starts the corresponding solver for parallel computing. When the solver finishes the calculation, obtain all the parallel computing results of the current population Pop(I Iter ), and the parallel computing processor sends an instruction of "parallel computing completed" to Step6.

[0025] Specifically, the calculation engineering types of the traveling wave tube components that the parallel processor can handle include electron gun, high-frequency circuit, collector, magnetic system, and input / output window, and the corresponding solvers are EOS, HFCS, EOS, MFS, and WS respectively. Among them, the calculation engineering of both the electron gun and the collector is simulated and calculated by the EOS solver.

[0026] Step6. Perform non-dominated sorting and update the optimal solution;

[0027] Receive the instruction of "parallel computing completed" sent by the parallel processor in Step5, perform fast non-dominated sorting and crowding degree calculation on the current population Pop(I Iter ), update to obtain the optimal solution, and determine the Pareto front.

[0028] Step 7. Perform population evolution and generate a new population;

[0029] Perform elitist strategy, crossover, and mutation on the current population Pop(I Iter ) to form a new generation population Pop(I Iter +1), and record I Iter =I Iter +1.

[0030] Step 8. Perform termination condition judgment;

[0031] Judge whether the current iteration number I Iter is greater than the maximum iteration number N Iter . If I Iter is greater than N Iter , then end the optimization calculation of the traveling wave tube components and execute Step 9; otherwise, return to execute Step 4 to send the "start parallel calculation" instruction to the parallel computing processor until I Iter is greater than N Iter .

[0032] Step 9. Result analysis and display;

[0033] According to the current optimization results, store the optimization decision variables and optimization objectives of each generation in a data file, and perform display processing on the saved data.

[0034] In summary, the present invention proposes a parallel multi-objective optimization method for traveling wave tube components. By optimizing the settings, the optimization decision variables and optimization objectives of the traveling wave tube components are selected and set. Then, by introducing a general parallel computing processor, the engineering of multiple traveling wave tube components is adaptively parallel computed to obtain the corresponding simulation calculation results. Finally, non-dominated sorting and updating of the optimal solution, and population evolution are performed. Based on the present invention, the traveling wave tube components can be designed and optimized more flexibly, with a higher degree of freedom in optimal design, and the design optimization and development technology of the traveling wave tube are further iteratively updated; effectively solving the problem of low efficiency in the existing multi-objective optimization method for traveling wave tube components that requires separate serial optimization execution for each calculation project. Brief Description of the Drawings

[0035] Figure 1 is the flow schematic block diagram of the present invention;

[0036] Figure 2 is the schematic block diagram of the parallel computing processor in the present invention;

[0037] Figure 3 is the schematic diagram of the parallel computing project of the electron gun embodiment in the present invention;

[0038] Figure 4 is the result of the optimized design scheme of the electron gun embodiment of the present invention;

[0039] Figure 5 is the comparison block diagram of the parallel execution mechanism of the present invention and the serial execution mechanism of the prior art;

[0040] Figure 6 is the result diagram of the optimized design of the electron gun implemented by the parallel execution mechanism of the embodiment;

[0041] Figure 7 is the result diagram of the optimized design of the electron gun implemented by the serial mechanism of the prior art. Detailed implementation manners

[0042] For the convenience of understanding the technical content of the present invention, taking the parallel multi-objective optimization of the electron gun as an example, the content of the present invention will be elaborated in detail with reference to the accompanying drawings.

[0043] A parallel multi-objective optimization method for traveling wave tube components, as Figure 1 shown, includes the following steps:

[0044] Step1. Set the optimization objectives of the traveling wave tube components;

[0045] The selected optimization decision variable is the cathode curvature radius Rc, the current value is 9.3 mm, and the set optimization range is [8.37, 10.23]; the lateral distance Z of the focusing electrode g The current value is 1.71 mm, and the set optimization range is [1.539, 1.881]; the longitudinal distance R g1 The current value is 6.1, and the set optimization range is [5.49, 6.71]; the axial distance Z of the anode a The current value is 9.51 mm, and the set optimization range is [8.559, 10.461].

[0046] The set optimization objectives include setting the total cathode emission current I total , setting the optimization direction to be greater than, and setting the optimization target value to 200 mA; setting the injection waist radius R b as the optimization objective, setting the optimization direction to be less than, and setting the optimization target value to 0.4 mm, that is, the optimization objective function is min{I total > 200 mA, R b < 0.4 mm}.

[0047] Step2. Set the optimization algorithm;

[0048] Select the NSGA-II multi-objective optimization algorithm, and the optimization algorithm parameters are set as the maximum number of iterations N Iter = 50, the population size N Pop = 16, the crossover rate dcross = 0.6, mutation rate d mutate = 0.2 and variable precision N ps = 2.

[0049] Step3. Generate the initial population of the optimization algorithm;

[0050] According to the traveling wave tube component optimization decision variables and optimization objectives set in Step1 and the optimization algorithm set in Step2, click Start Optimization on the optimization settings interface to randomly generate N Pop = 16 individuals within the parameter range of the optimization decision variables, generating the initial population Pop(I Iter ), and record I Iter = I Iter + 1.

[0051] Step4. Based on the current population Pop(I Iter ), send the instruction "Start parallel computing" to the parallel computing processor.

[0052] Step5. Perform non-dominated sorting and update the optimal solution;

[0053] As Figure 2 shown is the execution process of the parallel computing processor. The parallel computing processor receives the instruction "Start parallel computing" from Step4, identifies that the traveling wave tube component calculation engineering type is the electron gun calculation engineering. The parallel computing processor generates the corresponding number of processes N Thread , that is, N Thread = N pop and accordingly updates the structural parameters of the traveling wave tube component, and then starts the corresponding solver EOS for parallel computing. When the solver calculation is completed, obtain all the parallel computing results of the current population Pop(I Iter ), and the parallel computing processor sends the instruction "Parallel computing completed" to Step6.

[0054] As Figure 3 shown are the N Thread = 16 electron gun calculation engineering generated by the parallel computing processor.

[0055] Step6. Perform non-dominated sorting and update the optimal solution;

[0056] Receive the instruction "Parallel computing completed" sent by the parallel processor in Step5, perform fast non-dominated sorting and crowding degree calculation on the current population Pop(I Iter ), update to obtain the optimal solution, and determine the Pareto front.

[0057] Step7. Perform population evolution and generate a new population;

[0058] Perform the elitist strategy, crossover, and mutation on the current population Pop(I Iter ) to form a new generation population Pop(I Iter +1), and record I Iter = I Iter + 1.

[0059] Step8. Judgment of termination condition;

[0060] Judge whether the current optimization iteration number I Iter is greater than the maximum iteration number N Iter . If I Iter is greater than N Iter , then terminate the optimization calculation of the traveling wave tube components and execute Step9; otherwise, return to execute Step4 to send the "start parallel calculation" instruction to the parallel computing processor until I Iter is greater than N Iter .

[0061] Step9. Result analysis and display;

[0062] According to the obtained optimization results, store the optimization decision variables and optimization objectives of each generation in a data file, and perform display processing on the saved data. As Figure 4 shown in the result and display of the electron gun optimization design scheme of this embodiment, the design parameters and corresponding performance indicators set for the traveling wave tube components can be seen in the optimization results. It can be seen that the total cathode emission current I total is greater than the set optimization target value of 200 mA, and the waist radius R b is less than the set optimization target value of 0.4 mm.

[0063] Figure 5 is a comparison block diagram of the start calculation sequence of serial optimization calculation and parallel optimization calculation between the prior art and the present invention. Taking the serial multi-objective optimization of the electron gun as an example, it is necessary to execute each calculation project of the electron gun separately, and the next electron gun calculation project can be executed only after one electron gun project calculation is completed. This calculation method is the serial optimization calculation method. The parallel optimization calculation can start multiple electron gun projects for calculation simultaneously, which can improve the calculation speed of the multi-objective optimization of the electron gun.

[0064] In Figure 6 the description box in the optimization interface, the optimization start time and optimization end time of the electron gun optimization design scheme of this embodiment can be viewed. The parallel optimization takes 0 hours 2 minutes 58 seconds, that is, 178 seconds. If the serial optimization calculation is adopted, it is necessary to perform serial calculations on 16 electron gun calculation projects for one iteration. And when the number of optimization iterations increases, more electron gun calculation projects need to be serially optimized, resulting in more time consumption. Figure 7The computing time consumed by the serial optimization of the optimized design solution for the electron gun example is 0 hours, 12 minutes, and 48 seconds, which is 768 seconds. From the optimization time consumed by the two figures, it can be seen that the parallel optimization calculation can significantly improve the multi-objective optimization calculation speed of the traveling wave tube components.

[0065] As can be seen from the above embodiments, the present invention can design and optimize the traveling wave tube components more flexibly, with a higher degree of freedom in optimized design, and further iteratively update the design optimization and development technology of the traveling wave tube; effectively solve the problem of low efficiency in the existing multi-objective optimization method for traveling wave tube components, which requires separate serial optimization execution for each calculation project.

Claims

1. A parallel multi-objective optimization method for traveling wave tube components, characterized in that The specific steps are as follows: Step1. Set the optimization objectives of the traveling wave tube components; According to the design requirements of the traveling wave tube components, set the optimization decision variables of the traveling wave tube components in the optimization setting interface, including whether the traveling wave tube optimization decision variables are optimized, the lower limit value and the upper limit value of the optimization decision variables; set the optimization objectives of the traveling wave tube components, including the optimization objective value and the optimization direction; Step2. Set the optimization algorithm; Select a multi-objective optimization algorithm, including the NSGA-II or MOPSO optimization algorithm, and set the corresponding optimization algorithm parameters according to the selected different optimization algorithms; Step3. Generate the initial population of the optimization algorithm; Optimize the decision variables and optimization objectives of the traveling wave tube components according to the settings in Step1, and randomly generate N within the parameter range of the optimization decision variables of the traveling wave tube components according to the optimization algorithm set in Step2 pop individuals of the optimization algorithm to generate the initial population Pop(I Iter ), record I Iter = 0; Specifically, use I Iter to represent the current iteration number of the optimization algorithm, that is, I Iter = 0 indicates that the current population is the initialized population, I Iter = 1 indicates that the current population Pop(I Iter ) is the population of the first generation; Step4. Based on the current population Pop(I Iter ), send an instruction to start parallel computing to the parallel computing processor; Step5. The parallel computing processor starts to execute; The parallel computing processor receives the instruction to start parallel computing in Step 4, identifies the calculation engineering type of the traveling wave tube component, and the parallel computing processor generates the corresponding number of processes N according to the population size N in Step 2 Pop and updates the optimization decision variables of the traveling wave tube component accordingly, and then starts the corresponding solver for parallel computing; Thread That is, N Thread = N pop and updates the optimization decision variables of the traveling wave tube component accordingly, and then starts the corresponding solver for parallel computing; When the solver finishes the calculation, obtain all the parallel calculation results of the current population Pop(I Iter ), and the parallel computing processor sends an instruction of completion of parallel calculation to Step6; Step6. Perform non-dominated sorting and update the optimal solution; Receive the instruction of the completion of parallel computing sent by the parallel processor in Step 5, and perform fast non-dominated sorting and crowding degree calculation on the current population Pop(I Iter ) to update and obtain the optimal solution and determine the Pareto front; Step7. Perform population evolution and generate a new population; Execute the elite strategy, crossover, and mutation on the current population Pop(I Iter ), to form a new generation population Pop(I Iter + 1), and record I Iter = I Iter + 1; Step8. Perform termination condition judgment; Determine the current iteration count I Iter whether it is greater than the maximum iteration count N Iter If I Iter is greater than N Iter then end the optimization calculation of the traveling wave tube component and execute Step9; otherwise, return to execute Step4 to send a start parallel calculation instruction to the parallel calculation processor until I Iter is greater than N Iter ; Step9. Result analysis and display; According to the currently obtained optimization results, store the optimization decision variables and optimization objectives of each generation in a data file, and perform display processing on the saved data.

2. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that: The optimization directions described in Step1 include two directions: greater than and less than.

3. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that: The optimization decision variables of the electron gun in Step1 include the cathode position, anode position, and electron gun voltage of the electron gun; the optimization objectives of the electron gun include the total cathode emission current, the back-bombarding cathode current, the intercepted current, the beam waist radius, and the beam waist position.

4. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that: The optimization decision variables of the high-frequency structure in Step1 include the inner radius of the helix and the pitch; the optimization objectives of the high-frequency structure include, but are not limited to, the normalized phase velocity, the coupling impedance, and the attenuation constant.

5. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that: The optimization decision variables of the collector in Step1 include the collector entrance energy distribution, the collector structure parameters, and the electrode voltages of each electrode of the collector; the optimization objectives of the collector include the collector efficiency and the reflux rate.

6. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that: The optimization decision variables of the magnetic system in Step1 include the structure parameters of the magnet, the structure parameters of the pole shoe, and the material; the optimization objectives of the magnetic system include the magnetic field period and the peak magnetic field on the axis and the magnetic system performance parameters.

7. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that: The optimization decision variables of the input / output window in Step1 include the structural design parameters of the input / output window; the optimization objectives of the input / output window include the S parameter and the standing wave ratio.

8. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that In the said Step 1: The parameters of the NSGA-II multi-objective optimization algorithm include the maximum number of iterations N Iter , the size of the population N Pop , the crossover rate d cross , the mutation rate d mutate and the variable precision N ps ; The parameters of the MOPSO multi-objective optimization algorithm include the maximum number of iterations N Iter , the size of the population N Pop , the inertia weight factor d w , the individual learning factor d c1 , the mutation rate d c2 and the variable precision N ps ; Among them, the variable precision N ps is used to constrain the number of significant digits of the optimization decision variable to N ps digits after the decimal point, that is, when N ps = 2, it means that the number of significant digits of the optimization decision variable is 2 digits after the decimal point.

9. The parallel multi-objective optimization method for traveling wave tube components according to claim 1, characterized in that In Step5: The types of traveling wave tube component calculation projects that can be processed by the parallel processor include the electron gun, high-frequency circuit, collector, magnetic system, and input / output window, and the corresponding solvers are EOS, HFCS, EOS, MFS, and WS respectively; among them, the calculation projects of the electron gun and the collector are both simulated and calculated by the EOS solver.

Images