基于混合编码策略的电弧炉冶炼钢种的碳含量预报模型

提出一种混合编码策略的遗传算法(GA)训练电炉钢碳含量神经网络预报模型。先采用二进制编码策略,再采用十进制编码策略继续优化预报模型的权阈值,这种混合编码策略综合了二进制编码搜索能力强和十进制编码变异量可任意小的优点。仿真结果表明,混合编码策略的遗传算法(GA)具有更快的收敛速度和更好的寻优性能。对100 t电弧炉冶炼0.85%～1.00%C的钢种,预报碳含量的精度为±0.04%时混合编码GA的命中率为96%,二进制编码GA的命中率为90%。

Predictive Model for Carbon Content in Steel Grade Melting by Arc Furnace Based on Hybrid Coding Method

A new genetic algorithm (GA) using hybrid coding method to train the neural network model for predication of carbon content in arc furnace steel is proposed, that is first to use binary coding method then to use decimal coding method continuously to optimize and train the weighted threshold of predictive model. The hybrid coding method combines the advantages of binary coding method with strong search ability and decimal coding method with arbitrarily small variance. Simulation results show that the genetic algorithm (GA) using hybrid coding method has faster convergence rate and better search performance, for instance as steel grade with 0. 85% ~ 1. 00% C is melted by 100 t arc furnace, for ±0. 04% precision of predictive carbon content in steel the hit ratio by using hybrid coding method GA is 96% , but the hit ratio by using binary coding method GA is 90% .