Parameter Identification of Projectiles Based on an Improved Butterfly Algorithm Optimizing a Mixed Kernel Extreme Learning Machine

Accurate projectile dynamic modelling requires identifying aerodynamic parameters. The traditional methods for identifying aerodynamic parameters of missiles suffer from significant modeling errors. Therefore, this study proposes an improved butterfly-shaped optimization hybrid extreme learning machine algorithm. It combines the butterfly algorithm with a hybrid extreme learning machine, Cauchy mutation, and adaptive weight. The search ability of the Butterfly algorithm is enhanced by introducing the Cauchy distribution function and adaptive weighting factors. In addition, to balance the weights of searches and to optimize the regularization coefficients and kernel function parameters, the dynamic switching probability p is introduced. The identification accuracy of four different algorithms was compared under noise-free conditions. The feasibility of the improved butterfly-optimized hybrid extreme learning machine was verified. When there is noise, the strength of the algorithm is confirmed by comparing the effect of different noise levels on how well it can identify things. The simulation results show that the improved butterfly optimization hybrid extreme learning machine algorithm has higher accuracy and better robustness in identifying projectile aerodynamic parameters. The simulation results show that the improved butterfly optimization hybrid extreme learning machine algorithm has higher accuracy and better robustness in identifying projectile aerodynamic parameters.