Abstract: Aiming at Failure Mode and Effects Analysis (FMEA) problems on product under uncertain circumstances,a novel integrated FMEA method which combined hybrid multi-attribute decision with correlative propagation analysis was proposed.To deal with fuzziness and uncertain evaluation information of experts from all stages' FMEA process effectively,fuzzy rough number in the form of interval was adapted to model in this method.Under the modeling framework,fuzzy rough analytic hierarchy process was used to construct risk factor importance pair-wise comparation matrix,which was then optimized and solved by automata algorithm to get importance weight vector of risk factors.Kullback-Leibler divergence was introduced to improve VlseKriterijumska Optimizacija-Komp-romisno Resenje (VIKOR) model,and the improved VIKOR model was used to compute risk priority number.By considering the correlative influence between failure modes,the correlative propagation complex networks of failures modes was constructed.Through computing complex network topology parameters such as correlative output degree,betweenness and clustering coefficient,the risk priority number was revised,so that risk level of failures modes was sorted afresh comprehensively.A case study of large scale gearbox special for metallurgical equipment was offered to illustrate the practicability and validation of the proposed method.

Key words: failure mode and effects analysis, fuzzy rough number, automata-analytical hierarchy process, vlsekriterijumska optimizacija-komp-romisno resenje, complex network

摘要： 为解决面向不确定环境下的产品失效模式与影响分析(FMEA)问题,提出基于混合多属性决策和关联传播分析相集成的FMEA新方法。该方法为有效处理FMEA过程中各阶段专家的模糊性、不确定性评价信息,运用区间形式的模糊粗糙数进行建模。在此建模框架下,首先以模糊粗糙层次分析法构建风险因子权重分析矩阵,并通过自动机算法进行寻优求解得到权重向量;其次引入Kullback-Leibler散度对多准则妥协解排序法模型进行改进并用于风险优先数的计算;最后考虑到失效模式间的交互影响,构建失效模式关联传播复杂网络,通过计算关联输出度、介数、集簇系数等网络拓扑参数对风险优先数进行修正,进而对失效模式的风险性进行全面重新排序。以冶金设备专用大型齿轮箱为例,验证了该方法的可行性与有效性。

关键词: 失效模式与影响分析, 模糊粗糙数, 自动机&mdash, 层次分析法, 多准则妥协解排序法, 复杂网络