DEMATEL方法介绍解读课件.ppt

1、Decision making trial and evaluation laboratoryDEMATEL2010暑期培训班 张文力1References 1. S.M. Seyed-Hosseini *, N. Safaei, M.J. Asgharpour 2. Tongji University 3. 吉林大学4. 长春理工大学2DEMATEL起源？ 时间：November 2005 作者：S.M. Seyed-Hosseini *, N. Safaei, M.J. AsgharpourIran 发表：RELIABILITY ENGINEERING &SYSTEM SAFETY 名称：

2、Reprioritization of failures in a system failure mode and effects analysis decision making trial and evaluation laboratory technique3 DEMATEL中文:决策试行与评价实验室。 an effective approach for analyzing relation between components of a system in respect to its type (direct/indirect) and severity 它是进行因素分析与识别因素分

3、析与识别的一种有效的方法。例1 一种运用图论与矩阵工具图论与矩阵工具进行系统因素分析的方法，是通过系统中各因素之间的逻辑关系与直接影响矩各因素之间的逻辑关系与直接影响矩阵阵，计算出每个因素对其他因素的影响程度以及被影响度，从而计算出每个因素的中心度与原因度。 这种方法是充分利用专家专家的经验和知识来处理复杂的社会问题，尤其对那些要素关系不确定要素关系不确定的系统更为有效。DEMATEL是什么？ 下一页4归根到底DEMATEL是对(1)失效模型failure mode and effects analysis(2)影响因素(3)决策标准排序！排序！DEMATEL诞生的原因，见文章1见例1C1C2

4、C3C4A1XXXXA2XXXXA3XXXX传统的多目标决策矩阵：看各个标准之间的关系排序之后可以做什么排序之后可以做什么?例如例如(1)应该更关注哪个失效模型应该更关注哪个失效模型?(2)哪个影响因素对公司的业绩影响最大哪个影响因素对公司的业绩影响最大?(3)怎样给决策标准赋予权重怎样给决策标准赋予权重?5DEMATEL 输入什么，得到什么?i例1：怎样培养一个好学生？ (1)专家说专家说四大因素:A自身努力，B家庭因素，C学校因素，D社会因素(2)专家说专家说各要素之间的关系传统方法是否考虑了各因素之间的关系?ADCB哪个因素更重要?54321观察：BA:家庭因素直接影响一个人的努力程度D

5、A: 社会因素直接影响一个人的努力程度D BA：社会因素通过影响家庭，最终影响个人努力程度不仅有直接关系还有间接关系6(3)我们通过(2)得到一个矩阵:直接影响直接影响矩阵MDEMATEL 输入什么，得到什么?iiADCB54321ABCDA0000B5000C4000D3210注意用词！(4)把(3)中得到的矩阵标准化，得到矩阵MABCDA0000B1000C0.8000D0.60.40.20怎样标准化？文章17DEMATEL 输入什么，得到什么?iii(5)通过一系列计算得到综合影响综合影响矩阵MABCDA0000B1000C0.8000D0.560.40.20每行元素分别求和每列元素分别

6、求和ABCDA00000.00 B10001.00 C0.80000.80 D0.560.40.201.16 2.360.40.20RCR+CR-CA02.362.36-2B10.41.40.6C0.80.210.6D1.1601.161.2最后根据最后根据R+C 或者或者R-C对对ABCD排序排序注意用词！R+C 或者或者R-C 的具体的具体含义？含义？8问题：(1)DEMATEL为什么诞生？(2)中间的计算过程？(3)排序的依据？(4)DEMATEL的优点？文章文章1逐一解答逐一解答9Reprioritization of failures in a system failure mode

7、 and effects analysis decision making trial and evaluation laboratory technique作者：S.M. Seyed-Hosseini *, N. Safaei, M.J. Asgharpour Industrial Engineering Department, Iran University of Science and Technology, Narmak, Tehran, 16844, IranAvailable online 10 November 2005发表：RELIABILITY ENGINEERING &SY

8、STEM SAFETY最早提出dematel失效模式与影响分析10综述：研究意义（为什么？） some of inherently shortcomings of conventional Risk Priority Number (RPN) method and like.失效模式与影响分析这个领域传统的RPN 方法存在一定缺陷。研究目的（做什么？） an effective methodology related to decision making field has been developed for reprioritization of failure modes in a sy

9、stem Failure Mode and Effects Analysis (FMEA) for corrective actions. 优化失效模式与影响分析中失效模式的顺序顺序用来纠正偏差.研究内容（怎么做？） Decision Making Trial and Evaluation Laboratory (DEMATEL) analyzing relation between components of a system in respect to its type (direct/indirect) and severity.分析各元素之间直接+间接的关系研究结果 The demon

10、strated results have shown that DEMATEL method can be an efficient, complementary and confident approach for reprioritization of failure modes in a FMEA.有效、完备、可信11背景知识：失效模式与影响分析背景知识：失效模式与影响分析-FMEA 起源：1960，航空工业。研究系统中每一个组成部分可能的失效模式可能的失效模式，确定各个失效模式的后果后果或对系统可靠性的影响，并按其影响的严重程度严重程度进行分类的分析方法。 用途：某个系统对可靠性和安全

11、性要求很高。预防错误发生预防错误发生+及时纠错及时纠错(1)what might go wrong?什么地方可能出问题(2)What might cause it to wrong?什么原因引起的(3)And what effect would it have?会造成怎样的影响12传统的FMEA分析方法RPN risk priority number风险优先数风险优先数 Traditionally, the prioritization of failures for corrective actions is performed by eveloping a risk priority nu

12、mber (Ford Motor Company, 1988). 传统对错误排序的方法 RPN Number=S*O*D依据这个值的大小来排序S： severity of its failure effect 严重性O：probability of the failure-mode occurrence 可能性 D：probability of the failure being detected 被发现的可能性下一页下一页 例子例子13RPN打分：SOD14RPN的缺点(1) the RPN elements are not equally weighted with respect to

13、one another in terms of risk.(2) the RPN scale itself has some non-intuitive statistical properties.DEMATEL诞生！诞生！ 例如：A: RPN 9X3X2=54 B:RPN 4X5X6=120A一旦发生就会产生严重后果，应该优先处理，但是RPN值很小1. 平均值是166不是5002. 中点不在5003. 有1000种可能的RPN值，实际上只有120种15DEMATEL methodology i16DEMATEL methodology ii各行求和，找各行求和，找最大，此处最大，此处101

14、7DEMATEL methodology iii有证明不详讲，计算过程下一页含义含义；(1)B到到E有一条直接有一条直接或者间接通路或者间接通路(2) B对对E施加影响，影响施加影响，影响因子因子0.098 失效模式与失效模式与影响影响(3)B与与E的重要程度，的重要程度，B比比E重要重要0.098 决策标准决策标准18M计算过程excel10000-0.41-0.2-0.40001-0.3000-0.21-0.2-0.10001100000.4110.30.490.10.0101.060.320.060.0200.211.060.210.10001000000.400.20.400000.3

15、0000.200.20.100000.000 0.000 0.000 0.000 0.000 0.410 0.000 0.298 0.489 0.098 0.006 0.000 0.064 0.319 0.064 0.021 0.000 0.213 0.064 0.213 0.100 0.000 0.000 0.000 0.000 M=E-M=(E-M)-1=M*(E-M)-1=19DEMATEL methodology ivR+C和R-C的含义，下一页20DEMATEL methodology vR ,C,R+C,R-C的含义：R+C： degree of relation between

16、each alternative with others and alternatives having more values of R+C have more relationship with another and those having little values of R+C have litter relationship with others.与其他元素的关关系系，值越大，关系越密切。R-C：indicates severity of influence for each alternative. having more values of R-C have higher

17、influence to another and are assumed to have higher priority and those having little values of R-C receiving more influence from another are assumed to have lower priority.值大对其他元素施加很大影响，自己很少受其他元素影响。优先级由此判定.R-C0:dispatcher R-CCEDA.22DEMATEL 优点：(1)considering indirect relationships in relation analyze

18、(2)capability of clustering alternatives in large systems with many failure modes(3)RPN or similar approaches can be as an input data for DEMATEL method.与其它方法结合(4)allocating as possible as a unique rank number to each alternative 例子 下一页23优点优点(3)区区分度更好分度更好CF24应该CF27重要, RPN体现不出来，原因下一页24CF24 has a seve

19、rity of influence 360 on MF3 and a severity of influence 360 on MF5 too. However, according to row 32, CF27 has only a severity of influence 392 on MF4. It is notable that values 360 and 392 are relative large value in RPN implementation in Table 7.25Obviously, the performance of proposed method is

20、depends on input data, i.e. weights matrix. Therefore, other methods such as RPR can be considered for this method. Also DEMATEL can be considered in Fuzzy or probability environments that suggested for further researches.DEMATEL 未来的发展26 Assessing the architectural design services by using DEMATEL a

21、pproach 作者：Shouming Chen School of Economics and Management Tongji University发表：2009 International Conference on computational Intelligence and Natural Computing提供一个case study 加深印象27 Architectural Design Companies what factors should be used to assess the service? a more precise weigh vector ?是什么在影响

22、建筑设计公司的服务质量？28(1) Delphi method 确定6个主要因素 The committee chose six main factors for the evaluation of design service, which are: quality (F 1 ), delivery time (F 2 ), technological capability (F 3 ), reputation (F 4 ), flexibility (F 5 ), price (F 6 ). 29(2) Delphi method 确定6个因素间的相互关系用来标准化30(3)标准化direct-relation matrix31(4)根据公式得到Total relation matrix，计算D+R,D-R32(5)本文的特别之处(1)用用D+R来排序来排序(2)顺带求出了顺带求出了6个影响个影响因素的权重因素的权重结论在这里结论在这里33Thank you for your attention.34