Application of Generalized RBF Network Based on K-means Clustering in Solving Complex Mappings

Abstract

Initial perturbation is one of the important sources of the distribution of projectile dropping points. However, due to the complex structure of the virtual prototyping system of coupling system, there is no clear functional relationship between the parameters of the initial disturbance index and its influencing parameters. It is difficult to establish scientific mapping relationship. In this paper, for the complex mapping problem of multi-objective optimization of initial perturbation, based on K-means clustering generalized RBF network, the nonlinear mapping relationship between initial disturbance index parameters and its important influence parameters is established. Uniform design is used to establish a virtual shooting test scheme, and a generalized radial basis neural network is used to solve complex mappings between initial disturbance index parameters and important influencing parameters, thereby providing an important basis for initial disturbance optimization.