关键词: 自适应步长, 萤火虫算法, BP神经网络, 突防效能, 评估, 优化

Abstract: Firefly algorithm has problems of slow convergence and easy to get into local optimal. To solve the problems, a self-adaptive step firefly algorithm (SASFA) is presented, which replaces fixed search step by self-adaptive step associated with individual density. The algorithm combines self-adaptive step firefly algorithm and the BP neural network, and obtains the initial optimal weights and threshold value of BP neural network by using self-adaptive step firefly algorithm. The initial weights and threshold value are used as the initial parameters of BP neural network to improve the training precision and rate. Taking the penetration effectiveness evaluation of ballistic missile as an example, the penetration effectiveness index system is established, and the penetration effectiveness evaluation model of ballistic missile based on improved BP neural network is built. The example analysis and simulation experiments show that BP neural network optimized by approved firefly algorithm has high accuracy and strong convergence. It can be used in ballistic missile penetration effectiveness evaluation.

Key words: self-adaptive step, firefly algorithm(FA), BP neural network(BPNN), penetration effectiveness, evaluation, optimization