基于差分-分布估计算法的铝热连轧多目标规程优化

Aluminum hot strip mill rolling schedule optimization based on the algorithm of DE-EDA

铝热连轧生产中,合理的规程制定,能够改善产品质量,提高产量。利用最小二乘支持向量机对轧制力模型建模,并以降低能耗和预防打滑为目标,对河南某铝热连轧精轧机组进行规程优化。针对多目标分布估计算法(MOEDA)存在收敛速度慢和精度低的问题,采用改进的差分进化算法(DE)与之结合。改进了差分进化算法的差向量和最优粒子选取方法,并设计了合理的算法切换机制,该组合算法的收敛性和分布性相对原算法有明显改善。在河南某铝热连轧精轧机组的规程优化中,该方法能够获得收敛性和分布性较好的近似Pareto前沿,求解精度和算法可靠性优于传统方法。

Reasonable procedures can improve the quality of the product in the aluminum strip production.In this paper,using the least squares support vector machine（SVM）to predict the rolling force,aiming at reducing energy consumption and prevent slippage,the procedures of finishing mill group are optimized.To solve the problems of slow convergence speed and convergence precision for the Multi-Objective Estimation of Distribution Algorithm（MOEDA）,the improved differential evolution algorithm（DE）is combined with EDA.The difference vector and the method of selecting the best individual are improved,and a more reasonable switch method is designed.The convergence and distribution for the new algorithm is better than those of the old algorithms.To optimize some procedures of finishing mill group in Henan province,the new algorithm can obtain the approximate Pareto frontier with good convergence and distribution,the precision of solution and the reliability are superior to the traditional methods.