Adaptive design of the magnetron injection gun using constrained multiobjective optimization

Abstract

The design of the magnetron injection gun (MIG) is an inverse problem with implicit objective functions and complex constraints. The computer-aided design using an optimization algorithm is a promising method to improve the performance of the MIG and reduce design complexity and time consumption. However, the dependence on the initial structure makes it difficult to realize the real sense of using algorithms to design. To address this issue, this paper proposes an adaptive design method for the MIG. The design problem is equivalent to a constrained multiobjective optimization problem. A pseudo-random structure generation algorithm is developed based on the geometric constraints and the first-order MIG design trade-off equations. The fully parameterized MIG structure is randomly generated by the decision variables composed of the scale coefficients ranging from (0,1) and the other structural parameters. This transformation can increase the effectiveness of the randomly generated MIG structures and make it easier to set the value range of decision variables. The proposed adaptive design method is further verified by successfully designing the MIG for an 800 GHz gyrotron via the co-simulation of MATLAB and CST.