基于R2-MOEA/D算法的顶板支护决策模型

Decision model of roof support based on R2-MOEA/D algorithm

  • 摘要: 现有的顶板支护决策方法或片面分析安全因素,或对指标客观赋予权重,未能有效分配权重系数,不能满足高维多目标顶板案例决策的需求。针对该问题, 对顶板来压指标进行分析,提出了一种基于R2指标的差分高维多目标进化(R2-MOEA/D)算法的顶板支护决策模型。首先针对来压状态定义指标属性,建立顶板指标知识库,利用层次分析法和熵值法对知识库中的条件指标进行计算,得到指标的主观权重和客观权重;然后在确定主观、客观权重的基础上引入权重矩阵,构建基于R2-MOEA/D算法的顶板多目标决策模型;最后基于 R2-MOEA/D算法将多目标问题分解成多个子问题,利用切比雪夫函数作为R2指标排序标准进行个体选择,得到收敛性和多样性较好的Pareto最优解,即相似度最高的条件指标顶板案例,其对应的结果属性为事故案例的决策提供了支护方案。实验结果表明:R2-MOEA/D算法与NSGA2算法、NSGA3算法、RVEA算法相比,在数据集的收敛性和分布性上整体效果最优,改善了高维空间中的搜索能力。通过山西霍州三交河煤矿2-6011巷道和10-4151巷道对基于R2-MOEA/D算法的顶板支护决策模型进行可行性评定,结果表明:由R2-MOEA/D算法检索出的解决方案符合该矿的实际支护情况。

     

    Abstract: The existing roof support decision methods either analyze safety factors one-sidedly or assign weights to index objectively, fail to assign weight coefficients effectively, and cannot meet the demand of high-dimensional multi-objective roof case decision. In order to solve this problem, a decision model of roof support based on R2 index differential high-dimensional multi-objective evolution (R2-MOEA/D) algorithm is proposed by analyzing the roof weighting index. Firstly, the model defines the index attributes for the roof weighting state, establishes the roof index knowledge base, and calculates the conditional index in the knowledge base using analytic hierarchy process and entropy method, and obtains the subjective and objective weights of the index. Secondly, the model introduces the weight matrix on the basis of determining the subjective and objective weights to construct the roof multi-objective decision model which based on R2-MOEA/D algorithm. Finally, based on the R2-MOEA/D algorithm, the multi-objective problem is decomposed into multiple sub-problems, and the Chebyshev function is used as the R2 index ranking criterion for individual selection to obtain the Pareto optimal solution with better convergence and diversity, i.e., the conditional index roof case with the highest similarity. And its corresponding result attributes provide the support scheme for the decision of the accident case. The experimental results show that the R2-MOEA/D algorithm has the best overall effect in terms of the convergence and distribution of the data set compared with the NSGA2 algorithm, NSGA3 algorithm and RVEA algorithm, and improves the search capability in high-dimensional space. The feasibility evaluation of the decision model of roof support based on R2-MOEA/D algorithm is carried out through the roadway 2-6011 and roadway 10-4151 in Sanjiaohe Coal Mine, Huozhou, Shanxi. The results show that the solution retrieved by R2-MOEA/D algorithm is in line with the actual support situation of the coal mine.

     

/

返回文章
返回