求解板坯倒垛和落位问题的分布估计算法

Estimation of distribution algorithm for solving the slab stack shuffling and relocation problem

针对钢铁企业板坯库的板坯倒垛问题,与以往的研究不同,考虑了在倒垛过程中压在目标板坯上的障碍板坯可以不倒回原垛位的作业模式.提出采用板坯在板坯库中的移动次数来衡量板坯库的作业负荷,建立了以移动次数为优化目标的整数规划模型.在对模型的求解中设计了基于概率模型进化的分布估计算法(estimation of distribution algorithm,EDA)来确定轧制单元对应的板坯,采用基于辐射邻域的启发式算法为障碍板坯寻找最佳的落位位置.通过不同规模的数据实验,探讨了参数对算法性能的影响,确定了有效的参数组合.并与当前常用的改进遗传算法、单亲遗传算法进行对比,验证了所提算法的有效性和鲁棒性.

The slab stack shuffling （SSS） problem is studied in the slab yard of iron and steel industry. Different from the previous studies, in the process of slab stack shuffling, the obstacles slabs may not be moved back to the original stack. Thus, the movement times are put forward to measure the working load of slab warehouse and the integer programming model is established to minimize the number of the slab movement. An estimation of distribution algorithm （EDA） based on the probability model is proposed to determine the corresponding slabs for rolling units and a heuristic algorithm based on radiation neighborhood to find the best position for the obstacles slabs relocation. The influence of parameters on the performance of the algorithm is discussed through the experiments of different scales. Simulative experiments illustrate the effectiveness of the proposed method compared with modified genetic algorithm and partheno-genetic algorithm.