多决策专家系统.pdf

《多决策专家系统.pdf》由会员分享，可在线阅读，更多相关《多决策专家系统.pdf（6页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。

1、UNCORRECTED PROOF12A decision support system for coating selection based on fuzzy logic3and multi-criteria decision making4Georgios Athanasopoulosa,*,Carles Riba Romevab,15aDepartment of Project Engineering,Technical University of Catalonia,Spain6bCentre de Disseny dEquips Industrials(CDEI),Technica

2、l University of Catalonia,Spain79a r t i c l ei n f o1011Keywords:12Decision support13Coating selection14Qualitative and quantitative properties15Fuzzy sets16Multi-criteria171 8a b s t r a c t19Wear and corrosion are the most important factors that the surface of the engineering parts must con-20fro

3、nt.The need for protection and improvement of the mechanical characteristics of the surface of engi-21neering parts can be to some extent satisfied by coatings.Coatings are considered as an excellent solution22when resistance to corrosion,oxidation or low friction is demanded,but due the complexity

4、of selecting23the appropriate one,engineers often avoid them.The need for simultaneous consideration of qualitative24and quantitative properties,render the use of classic material selection theories inadequate.An expert25system for coating selection is presented in this paper,which can handle both q

5、ualitative and quantita-26tive variables.The mathematical model used combines the multi-criteria decision making theories27(MCDM)together with the fuzzy sets theory.The Max-Min set”method is applied to calculate the order-28ing value of the alternatives while the TOPSIS method is used to rank them.A

6、 numerical example is pro-29vided to illustrate the method.Finally,the process presented can be easily computerized,to create the30relative software.31?2009 Published by Elsevier Ltd.3233341.Introduction35Mechanical parts in engineering applications usually have to36endure more than a single load,to

7、 serve a single purpose or to have37only one characteristic.Frequently a material has to successfully38withstand a combination of mechanical and chemical loads and39serve different purposes.The various characteristics required can40be divided into two categories,the substrate requirements and41the s

8、urface requirements.42Wear and corrosion are the most important factors that the sur-43face of the engineering parts must confront.Davis(2001)Q1states44that economical reports for the US market,refer almost 300 bil-45lion$annual losses due to corrosion of metals,while scientists46support that 40%of

9、the cost could be avoided by proper corrosion47prevention methods.The need for protection and improvement of48the mechanical characteristics of the surface of engineering parts49can be to some extent satisfied by coatings.50A coating is considered to be a thin layer of a material which is51deposited

10、 over the surface of another material in order to prevent52or decrease the rhythm of the chemical and physical interaction53between the substrate and the environment.Resuming in a few54lines the attributes of coatings,it can be said that they can provide55resistance against corrosion,oxidation,wear

11、and also provide ther-56mal insulation.Moreover,they can improve the mechanical prop-57erties of the surface and change the electronic ones.Even more,58improve the appearance and the optical properties of the surface59and protect the substrate from microbes(ASM Handbook,1994).60By using coatings,the

12、 engineer can choose a simple material as61a base to fabricate a mechanical part and then coat it with a differ-62ent and better material to meet the service criteria.The profits of63that are tremendous and actually the invention of high technology64coatings opened new paths in fabrication design.Gr

13、ainer and Blunt65(1998)present a thorough analysis of the properties and the appli-66cations of the commercial coatings and propose further uses.By67using a simpler material for the substrate,the engineer can dimin-68ish the cost of purchase.Moreover,during forming the tool wear69and the use of lubr

14、icants are diminished as well as the energy con-70sumption.Furthermore,high dimensional accuracy can be easier71achieved.Concerning the use of the part,the wear and corrosion72are diminished,thus the service life increased.These are some of73the characteristics of coatings and how they can be exploi

15、ted.74Not every coating can provide protection against everything.75Each type of coating has its own properties,which can change radi-76cally by changing the process details,like the temperature or depo-77sition time.On the other hand,a material will have different78chemical and mechanical propertie

16、s,depending on the coating0957-4174/$-see front matter?2009 Published by Elsevier Ltd.doi:10.1016/j.eswa.2009.01.016*Corresponding author.Address:Av.Diagonal,647 Edificio H,Planta 10,08028Barcelona,Spain.Tel.:+34 652675912.E-mail addresses:georgios.athanasopoulosupc.edu(G.Athanasopoulos),car-les.rib

17、aupc.edu(C.R.Romeva).1Address:c/Llorens Artigues,46,Edificio U,Planta 0,08028 Barcelona,Spain.Tel.:+34 93 401 08 34;fax:+34 93 401 19 89.Expert Systems with Applications xxx(2009)xxxxxxContents lists available at ScienceDirectExpert Systems with Applicationsjournal homepage: 3430No.of Pages 6,Model

18、5G6 February 2009 Disk UsedARTICLE IN PRESSPlease cite this article in press as:Athanasopoulos,G.,&Romeva,C.R.A decision support system for coating selection based on fuzzy logic.Expert Systems with Applications(2009),doi:10.1016/j.eswa.2009.01.016UNCORRECTED PROOF79deposition method.Combining the a

19、vailable deposition techniques80with different materials,one can get finally the desired coating81properties.82The selection of the appropriate coating is a complicated pro-83cess.Service conditions(load type,chemically aggressive environ-84ment,temperature,etc.),compatibility of the materials,objec

20、ts85shape and deposition parameters are some of the variables that86need to be considered.The final problem results to be multi-87dimensional and classic material selection theories cannot cover88it in full length(Wang&Chang,1995).89The selection of the right coating type and its deposition param-90

21、eters is a process that undertakes a lot of time and experience.On91the other hand,the benefits that can be gained by the use of coat-92ings compensate the dedicated time.Nevertheless,the complexity93and the difficulty of selecting the suitable coating,call for the cre-94ation of a customized select

22、ion strategy.95The purpose of this paper is to present an expert system for96material selection adapted for coatings.The mathematical model97used,combinesthemulti-criteriadecisionmakingtheories98(MCDM)together with the fuzzy sets theory.A modified TOPSIS99method is applied(Triantaphyllou,2000),which

23、 includes a new100definition for the ideal solution”,while the fuzzy data are rated101with the Max-Min set”theory(Chen,1985).A numerical example102is provided to illustrate the method.Finally,the process presented103can be easily computerized,to create the relative software.1042.Material selection a

24、nd decision making105Selecting the best material for a part involves more than select-106ing a material that has the properties that provide the necessary107service performance;it is also intimately connected with the pro-108cessing of the material into a finished part.Thus,a poorly chosen109materia

25、l can add to manufacturing cost and unnecessarily increase110the cost of the part.Also,the properties of the part may be changed111by processing,and that may affect the service performance of the112part(Dieter,2000).113Naturally,material selection has been a case of study for many114researchers and

26、a great effort has been made to create efficient115selection strategies.Ashby(2004)created the Material property116charts”concerning various mechanical and thermal properties.117Diagrams that represent ratios between material properties are118illustrated on the charts.The optimal material for a part

27、icular engi-119neering application can be then selected,based on the desired ra-120tio.Another strategy developed by Ashby(2004),deals with the121exploitation of previous experience stored in huge databases.With122the use of analytical questionnaires this knowledge is drafted and123applied in the pa

28、rticulate application.Dieter(2000)divides the124problem into two categories:material selection for a new product125or for a re-designed product.Then he defines the different direc-126tions that the analysis may follow and states that material selec-127tion must be combined with the particulate desig

29、n of the128product.In every case,material selection is based on the compar-129ison of a small number of quantitative material properties,speci-130fied through the design requirements.131However,some times there is a plethora of design requirements132that impose the implication of qualitative propert

30、ies,like corrosion133resistance,apart from the quantitative properties,like hardness.134The performance of the materials in qualitative properties cannot135be expressed by crisp values but by linguistic terms and it is diffi-136cult to be handled.In order to confront these situations,Jee and137Kang(

31、2000)attempted to aid the material selection with decision138making theories.They developed a selection strategy,aided with139decision making theories,that considers quantitative as well as140qualitative variables.In this method a 110 scale is used to rate141the performance of the materials concerni

32、ng the qualitative vari-142ables.The data from both quantitative and the qualitative proper-143ties are then introduced into the final algorithm and the ranking is144made.However,the processing of the qualitative variables and the145way they are transformed into crisp values is rather simple.Several

33、146other customized strategies exist addressing single issues,which147however cannot comprise an integrated method for coating selec-148tion(Edwards,2005;Edwards&Deng,2007;Giudice,2005).149Dobrzanski and Madejski(2006),presented an expert system150prototype for selection of coatings for metals.The p

34、roposed analy-151sis for the definition of the design requirements is adequate,yet152the mathematical model that is used for rating the performances153and ranking the alternatives is relatively simplified.154There are two major issues that need to be confronted:first,the155extensive number of variab

35、les that need to be considered and sec-156ond the need for a combination of qualitative and quantitative157parameters in a systematic approach to coating selection.The158above reasons impose the use of a more sophisticate mathematical159model.160With the implication of qualitative properties in mate

36、rial selec-161tion,the situation becomes unclear and the introduction of fuzzy162logic becomes necessary(Zadeh,2008).Fuzzy sets and fuzzy logic163where invented during the 1960 by LotfiZadeh.Zadeh realized that164many data in modern science and engineering cannot be expressed165by crisp values,becau

37、se they enclose an amount of uncertainty.Za-166deh introduced a theory whose objects(fuzzy sets)are sets with167boundaries that are not precise.The membership in a fuzzy set is168not a matter of affirmation or denial,but rather a matter of degree169(Klir&Yuan,1995).170Generally,when a number of crit

38、eria or requirements should be171simultaneously considered to select a material,a MCDM method172can be applied,to designate the right choice.A widely used MCDM173method is the Technique for Order Preference by Similarity to Ideal174Solution(TOPSIS)(Shanian&Savadogo,2006;Shih,2007).The ba-175sic conc

39、ept of this method is that the selected alternative should176have the shortest distance from the ideal solution and the farthest177distance from the negative ideal solution,in some geometrical178sense.The Euclidean distance approach was proposed to evaluate179the relative closeness of the alternativ

40、es to the ideal solution.Thus,180the preference order can be derived by a series of comparisons of181these relative distances.The advantage of this method is its sim-182plicity and ability to yield an indisputable preference order(Yur-183dakula&Tansel,2008).1843.Method presentation185In the expert s

41、ystem presented in this paper,an attempt is186made to combine the multi-criteria decision making theories with187the theory of fuzzy sets.The data from the qualitative properties188are treated with the fuzzy logic theory and introduced into a TOP-189SIS like algorithm,where they are combined with th

42、e data from the190quantitative properties.To prescribe the relative priority among191the material properties,a weight factor is given to each of them.192Afterwards,the data are processed by the algorithm and the final193ranking is made.The proposed method for coating selection in-194cludes the follo

43、wing stages(Fig.1).1953.1.Specification of the design requirements196The material selection process begins with the specification of197the design requirements.Through the design requirements the198performance of the material in each property(variable)is defined,199thus the selection criteria are set

44、(Ashby,2004).The assembling of200the desired performances comprises the ideal solution”.However,201the qualitative properties included in the selection criteria,cannot202be processed in the same way like the quantitative ones;there is a2G.Athanasopoulos,C.R.Romeva/Expert Systems with Applications xx

45、x(2009)xxxxxxESWA 3430No.of Pages 6,Model 5G6 February 2009 Disk UsedARTICLE IN PRESSPlease cite this article in press as:Athanasopoulos,G.,&Romeva,C.R.A decision support system for coating selection based on fuzzy logic.Expert Systems with Applications(2009),doi:10.1016/j.eswa.2009.01.016UNCORRECTE

46、D PROOF203need for a clear distinction between them in the selection process204(Table 1).2053.2.Process of qualitative parameters206The performance of every material as it concerns the qualitative207properties cannot be expressed as a crisp number,but a linguistic208term,such as good,bad,medium and

47、excellent.Obviously the209boundaries of the above classes are not precisely defined.Thus,210the theory of fuzzy logic can be applied,to deal with these vari-211ables(McNeill&Thro,1994).212According to fuzzy logic,these linguistic terms can be repre-213sented by a membership function,which can be tri

48、angular,trape-214zoidal,etc.(Fig.2).In these graphs the x axis represents the215gradation of each property and the y axis the membership function,216that is the degree that an element(material)with the particulate217grade belongs to a specific class(set).Each set in the graph(trian-218gle,trapezoid,

49、etc.)describes a different class of performance.219Therefore,for a group of candidate materials,the designer can cre-220ate a graph with their performances,concerning a property(Klir&221Yuan,1995).222Once the performance of each material,concerning a qualitative223variable,has been connected to the

50、corresponding membership224function,a fuzzy ranking method must be applied,to defuzzify225them and transform them into manageable data.226For this step,the theory of the Max-Min set”,created by Chen227(1985)is applied.According to Chens theory,if a fuzzy number Li228has a membership function defined