GA-BP Neural Networks for Optimal Design of Multistage Depressed Collector for 340-GHz Traveling Wave Tubes

In this article, a three-stage depressed collector for a 340-GHz traveling wave tube (TWT) is designed, and thermal analysis is finished to test the thermal reliability. To quickly finish the design of the collectors with high efficiency, low reflux rate, and heat dissipation, the back-propagating (BP) neural network (NN) and a multiobjective optimization algorithm are combined to optimize the collector. First, BP NN is used to construct an approximate calculation model between the ten input parameters of the collector and the three target values: 1) recovery efficiency; 2) reflux rate; and 3) heat balance. Then, the nondominated sorting genetic algorithm III (NSGA-III) was used to find Pareto optimal solutions for the three targets. The proposed method’s optimal recovery efficiency is 88.17%, the reflux rate is 0.09%, and the temperature difference inside the collector is only <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$16~^{\circ }$ </tex-math></inline-formula>C, achieving thermal balance. It provides a reference for the fast and accurate design of a multistage depressed collector (MDC) of a 340-GHz TWT.