基于循环神经网络的行波管大信号输出特性预测

Non-linear Output Characteristic Prediction of TWTs Based on Recurrent Neural Network Model

深度学习是机器学习领域的一个研究方向,已应用于多种人工智能技术的研究。本文基于行波管大信号理论建立了循环神经网络训练模型,将深度学习用于行波管非线性特性的预测,并对行波管的输入参数进行了核函数变换,增强了模型对非线性特征的传递能力。经过训练得到一个适用于8~18 GHz螺旋线行波管的模型,该模型可以预测行波管的输出功率。

Deep learning is a subset of machine learning,and it has been applied to many artificial intelligence technologies.A training model of recurrent neural network based on large signal theory of traveling wave tubes(TWTs)is established,with which deep learning is used to predict the nonlinear characteristics of TWTs,and kernel function transformations for the input parameters of TWT is constructed to enhance the nonlinear features transfer ability of the model.After training,a model for 8~18 GHz helix TWTs is obtained,which can predict the output power of TWTs.