风险可能数——一种基于最大信息熵理论的风险度量和风险排序新方法  

Risk Possibility Number—A New Model for Risk Evaluation and Prioritization Based on Maximum Entropy Theory

作　　者：王贵宝[1,2] 黄洪钟[1] 张小玲[1] 

机构地区：[1]电子科技大学机械电子工程学院,四川成都610054 [2]中国人民解放军驻三七二厂军事代表室,江西景德镇333002

出　　处：《航空学报》2009年第9期1683-1690,共8页Acta Aeronautica et Astronautica Sinica

基　　金：国家自然科学基金(50775026);教育部高等学校博士学科点专项科研基金(20060614016);四川省科技支撑计划(07GG012-002)

摘　　要：分析了风险优先数(RPN)方法中严重性S、发生率O和探测度D这3个评价指标的离散性与得分主观性不足。基于信息熵定理和最大信息熵推论,指出了RPN方法在评价风险不确定程度时,存在函数方差的依赖性和未能均衡各种风险要素的不合理性。针对多维信息的不确定性问题,分析了如何合理选取风险评价指标和权重分配。为了实现对风险程度的一致性测量,基于公理化的假设条件,给出了一种区别于RPN的风险可能数(RPoN)的定义和计算方法,并证明了RPoN方法能够一致性地用于风险的不确定程度评价。实验数据结果表明:RPoN方法用于风险的定量计算以及进行风险排序的效果优于RPN方法。This article analyzes the deficiencies of discreteness and subjectivity of the current method of risk priority number （RPN） in dealing with the three key information factors of severity, occurrence and detection of failure. Based on information entropy theory and maximum entropy inference, it demonstrates that when being used to evaluate the degree of uncertainty of risks, the substantial irrationalities of the RPN method are its dependency on the mean and the variance of the RPN function so that it cannot integrate the risk factors in proper proportion. Meanwhile, considering the uncertainty of multi-dimensional information, the problem of how to select the indexes for risk evaluation and allocate their weights for forecasting risk events is analyzed. For the realization of measuring the degree of risks consistently, a new definition and calculation by the name of the risk possibility number（RPoN）is proposed, which is different from the RPN. The formulation of the RPoN is based on and deduced from the continuity, monotonicity, and additivity assumptions of axioms of information entropy theory. The property of consistency of the RPoN approach is also proved when it is used to evaluate the degree of risks in the process of statistical decision-making. Experimental results suggest that the RPoN approach is better than the RPN method on quantitative calculation and prioritization of risks.

关 键 词：风险优先数 信息理论 多元控制系统 熵 测量 风险可能数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]