Genetic Stochastic Algorithm Application in Beam Dynamics Optimization Problem

Abstract

AbstractThe article discusses the application of the genetic global search algorithm to the problem of beam dynamics optimization. The algorithm uses normal distribution to form new generations and provides covariance matrix adaptation during random search. The method is easy to use because it does not require calculation of the covariance matrix. The algorithm application is illustrated in the problem of global extremum search for the functional that characterizes beam dynamics quality in a linear accelerator. The extremal problem under study has a large number of variables; the objective function is multi-extreme. Therefore, the use of the stochastic method is the preferred way to achieve the goal. The algorithm quickly converges and can be successfully used in solving multidimensional optimization problems, including its combination with directed methods. The optimization results are presented and discussed.