Inverse Design of a Broad Bandwidth W-Band Meta-Surface Window for Gyro-TWT Based on Bidirectional Machine Learning

A bidirectional machine learning (ML) method is proposed in this work to design a broad bandwidth W-band meta-surface window (MSW) for a gyrotron traveling-wave tube (gyro-TWT). This ML method combines the autoencoder (AE) algorithm, the backpropagation (BP) neural network based on the particle swarm optimization (PSO) algorithm, and the deep neural network (DNN) algorithm, where the obtained training mean squared error (MSE) is as low as 0.00321 when deploying 750 sets of data. Moreover, a W-band gyro-TWT MSW with a bandwidth of 17 GHz and a reflectivity below −20 dB when operating at 95-GHz center frequency is designed using the presented ML method. The thickness of this MSW reaches 2.1 mm, providing a feasible application for high-power gyro-TWT. Finally, a sapphire MSW is fabricated to verify the simulation results, and the results agree well with each other.