摘要:
提出了一种基于字符串编码克隆选择的建模算法, 其核心是通过字符串编码的克隆选择策略在初等函数空间上进化、选择初等函数和函数间的运算关系, 构成预测模型, 并优化模型参数, 实现预测模型形式的自动生成和参数的自适应确定. 字符串编码设计使随机生成函数表达式成为可能; 克隆选择策略实现了字符串编码抗体的进化训练; 对基于字符串编码的克隆选择操作算子设计, 增加了种群的多样性, 使得不同抗体协同进化, 有效避免了进化``早熟''现象. 用该算法建立某武器系统可靠度预测模型, 实验结果表明: 该模型具有较高的精度, 为可靠性预测、故障预报与维修奠定了基础.
Abstract:
Modeling algorithm based on string-coding clone selection is put forward. In the algorithm, evolution and choice of function structure, arithmetic relation and parameters are completed in elementary function space with string-coding clone selection, therefore, automatic production of forecasting model structures and adaptive determination of forecasting model parameters are realized. Design of string-coding makes it possible to randomly produce an expression in elementary function space; clone selection strategy achieves evolutionary training of string-coding antibodies. Moreover, by design of clone operator to string-coding antibodies, different antibodies evolve coordinately, diversity of antibodies restrains ``precocity'' in evolution. Reliability forecasting model of the weapon system is built uisng the algorithm. Results of experiments show that the forecasting model is of high accuracy, and lays a foundation of reliability forecasting, fault forecasting and maintenance for the control system.