Optimal working parameter searching method for power module of high-power millimeter wave gyrotron traveling wave tube

Abstract

The invention belongs to the technical field of high-power millimeter waves, and particularly provides an optimal working parameter searching method for a power module of a high-power millimeter wave gyrotron traveling wave tube. The method comprises the following steps: firstly, through combination and correlation analysis of test data, selecting working parameters which can be used for establishing a prediction model as input, and taking efficiency as output to form sample data; then, establishing a prediction model through a radial basis function neural network, and effectively training the prediction model; and finally, taking prediction output of the trained radial basis function neural network prediction model as a fitness function, performing global optimization by adopting a genetic algorithm, and obtaining predicted optimal efficiency and corresponding working parameters thereof. According to the method, the optimal efficiency and the corresponding working parameters can be predicted based on a large amount of test data provided by an automatic test system, the workload of parameter design of the gyrotron traveling wave tube is greatly reduced, the design and test cost is further greatly reduced, and the practical value is good.

Description

technical field [0001] The invention belongs to the technical field of high-power millimeter waves, and relates to an automatic test system for high-power millimeter-wave gyro-traveling-wave tubes; specifically, it provides a high-power millimeter-wave gyro-traveling-wave tube power module optimal operation based on a radial basis neural network prediction model Parameter search method. Background technique [0002] As a mainstream device in high-power millimeter-wave devices, the cyclotron traveling-wave tube has a wide range of applications in the fields of national defense, scientific research, and civil communications because of its high power and wide frequency bandwidth in the millimeter-wave band. The design of the gyrotron TWT mainly includes two parts: size design and current-voltage-velocity aspect ratio of the electron beam. After the structural design is completed, the working parameters of the electron beam usually require a large number of theoretical calculati...

AI Technical Summary

Problems solved by technology

[0003] At present, the prediction module is usually used in the automated test system to predict the optimal efficiency of the test data, and then obtain the working parameters corresponding to the optimal efficiency under the preset device structure; among them, the common prediction modules are constructed based on the BP neural network, but , because the BP neural network has the following deficiencies when making predictions: first, it is easy to fall into a local optimal solution; second, it is difficult to determine the hidden layer and the number of neurons in the hidden layer; third, whether a new BP neural network can be trained Achieving convergence is related to many factors (such as the size of the data set, the way the data set is selected, the network structure (input and output nodes, hidden layer nodes and activation functions), loss function and number of iterations, etc.), and finally leads to the prediction based on BP neural network The model is greatly affected by various factors, the prediction accuracy cannot be kept stable, and the generalization ability is poor

Images