Developing fuzzy expert system for supplier and...

Scientia Iranica A (2016) 23(3), 842{855

Sharif University of TechnologyScientia Iranica

Transactions A: Civil Engineeringwww.scientiairanica.com

Developing fuzzy expert system for supplier andsubcontractor evaluation in construction industry

E. Shahvand, M.H. Sebt� and M.T. Banki

Department of Construction Engineering and Management, Amirkabir University of Technology, Tehran, Iran.

Received 10 November 2014; received in revised form 4 August 2015; accepted 19 October 2015

KEYWORDSFuzzy expert system;Supply chainmanagement;Supplier;Subcontractor;Evaluation.

Abstract. Increasing competition in the global market to gain customer satisfactionand maximize pro�tability has been a challenge in the last decade for many constructioncorporations. In recent years, Supply Chain Management (SCM) has been introduced asone of the most e�ective approaches in the construction industry literature; further, the�rst step in SCM implementation is supplier evaluation and thus �nding and resolvingits weaknesses. In this regard, this research identi�ed the criteria for evaluation of twotypes of suppliers in the construction industry (�rst, material and equipment suppliersnamed suppliers, and second, service suppliers named subcontractors) and their fuzzymembership functions, distinctly, through literature review, questionnaire survey, andstatistical analysis of expert judgment. Since most of them are linguistic parameters, fuzzyapproach through Mamdani's inference mechanism has been utilized to develop a newmethodology of fuzzy expert system. Thus, the developed expert system would evaluateand select the subcontractors and suppliers, distinctly, through three main criteria which areQuality, Cost, and Work relations history-timely delivery. The developed system was testedin three major companies and the results revealed signi�cant performance improvement ofthese construction companies.© 2016 Sharif University of Technology. All rights reserved.

1. Introduction

Nowadays, strong competitive pressure forces manycompanies to make strategic changes. Constructioncompanies purify their organization, decrease theirorganizational vertical hierarchy, concentrate on spe-ci�c activities, and provide their requirements withoutsourcing. In the previous decades, companies havefound that cooperation with the factories and servicecompanies in the �elds of collaboration, planning,and execution is vital. Therefore, companies canshare technologies and planning information using thedevelopment of a vertical coalition of the organizations

*. Corresponding author. Tel.: +98 21 64543006;Fax: +98 21 64543037E-mail addresses: [email protected] (E. Shahvand);[email protected] (M.H. Sebt); [email protected] (M.T. Banki)

within the supply chain [1]. Rapid changes in envi-ronment convinced organizations to establish highly exible, agile supply chains and prepare a fast responseto evolutions. In this sequence, an e�ective supplierevaluation and selection is essential [2]. Gu�n etal. believe that at the present time, the purchase isconsidered as a strategic task instead of the technicalone. This can be assigned to the appreciable in uenceof suppliers on the success or failure of a �rm. Hence,the process of purchasing should meet the followingrequirements: 1) De�nition of criteria for supplier-selection; 2) Decision-making in supplier-selection; and3) Evaluation of supplier performance [3]. Banki et al.presented that increasing the number of bidders woulddecrease the prices in tendering and purchase activitiesof construction �rms [4]. Selecting suppliers andsubcontractors depends just on tendering price and thiswould be risky and result in project failure; therefore,

E. Shahvand et al./Scientia Iranica, Transactions A: Civil Engineering 23 (2016) 842{855 843

it is essential to evaluate based on multiple criteria toachieve satisfaction of owners. Due to the uncertainnature of construction projects and subjective judg-ment of the decision-makers, fuzzy decision models aredeveloped for construction contractor selection throughutilization of multiple criteria [5,6].

The evaluation of the performance of suppliers ina supply chain is a process to analyze and managetheir performance as well as to reduce costs and riskand to maintain continual improvement. The supplierevaluation system enables the organization to:

1. Control and evaluate the performance of suppliers;

2. Employ the same criteria and methodology for theevaluation;

3. Select suppliers in a systematic framework; in otherwords, to select robust suppliers for long-termcooperation and improve weak suppliers to achievean acceptable level of performance.

The performance evaluation systems should havethree main characteristics:

1. They should be systematic;

2. They have to be integrated with other systems inthe organization;

3. They should be e�cient and applicable.

In traditional models, performance evaluationsystem is only based on the e�ciency and produc-tion cost control. At the present time, due tothe unpredictable changes in the work environmentand managerial challenges, companies are involved ina competition which forces them to consider non-�nancial factors [7]. Chien-Ho et al. developed aSubcontractor Performance Evaluation Model (SPEM)by employing Evolutionary Fuzzy Neural InferenceModel (EFNIM) [8].

Construction companies have close relationshipswith two groups of suppliers, namely suppliers andsubcontractors. The �rst group supply materials,equipment, and machineries and the second groupsupply services. Therefore, evaluation of both suppliersseems essential. This study aims to develop a fuzzyexpert system for evaluation and selection of suppliersbased on the required criteria of the organization.First, the methodology is described; then, performancecriteria are speci�ed through literature review andexpert judgment; next, the architecture of the systemis represented; afterwards, fuzzy system membershipfunctions and rules are described; and �nally, theimplementation results are presented.

1.1. Supply chainSeveral de�nitions have been proposed in the literature.A system in which the raw suppliers, production

processes, distributors, and customers are regulated inone direction and information ow in the other, whichare related to each other and are called the supplychain [9]. In some cases, it includes pecuniary ow [10].Indeed, supply chains include all the activities relatedto material ow and transformation, from the raw ma-terial preparation phase to customer product deliveryphase. In correlation with material ow, there arealso two other types of ow: information ow and�nancial resources ow or pecuniary ow [1]. Supplychain management is strategic; it provides the system-atic trade coordination between companies and makesuse of any correlated trade process which leads toperformance improvement of each and every companyand their supply chain [11]. Nowadays, supply chainimprovement through partnership is widely known asan e�ective tool.

1.2. Supplier performance evaluation criteriaSupplier-selection decisions are very complicated be-cause of various factors that must be consideredin the decision-making process. Many researcherspointed out that the numbers and types of criteriatotally depend on the corporate policy, objectives, andstrategies [4,5,8,12,13]. Hence, in this research, theperformance criteria are de�ned by reviewing literatureand obtaining expert's knowledge. According to thestudies carried out by Neely et al.~[14,15], the favorablecriteria should:

1. Be based on the organizational strategy;2. Be simply understood;3. Provide precise feedback;4. Be based on the quantities which can be controlled

by users;5. Represent the business process - in other words,

both client and supplier contribute to de�ne them;6. Be related to the organizational goals;7. Be compatible;8. Be a part of the management loop;9. Be clearly de�ned;

10. Have distinct e�ects;11. Focus on the development;12. Be valid within the time;13. Provide fast feedback;14. Provide one de�nite goal.

The problem of determination of the supplier-selection criteria has been the focus of a signi�cantamount of studies since 1960s [2,3,5-8,16-13,17,18].Dickson presented 23 supplier-selection criteria, includ-ing [17,18]: (1) net price, (2) capability to provideall quality characteristics, (3) quality of maintenance

844 E. Shahvand et al./Scientia Iranica, Transactions A: Civil Engineering 23 (2016) 842{855

Table 1. Performance criteria as short term e�ect and capability criteria as long term e�ect on the achievement of thesupply chain.

Capability factors Performance factorsQuality systems for operation with the place/qualityphilosophy of the supplier

Ability to meet delivery promise/delivery leadtime/consistent delivery

Financial capability of the supplier Price

Technological capability/R&D capability Quality/reliability of the product

Reputation for integrity/believability andhonesty/vendor's image

Management sensitivity to requirementsof the buyer/attitude

Management sensitivity to requirementsof the buyer/attitude

After-sales support/the available technicalsupport

Performance awards/performance history Positive attitudes towards complaints

Bidding procedural compliance

Pro�tability of suppliers

Breadth of product line/ability of a supplier tosupply a number of items

Proximity of the supplier/geographic location

Management and organization

Contribution to productivity

Con ict resolution

Production facilities and capacity

Communication openness

Labor problems at the supplier's place

Business volume/amount of past business

services, (4) timely delivery, (5) geographical location,(6) �nancial stability, (7) production facilities, (8)commercial experience, (9) technological capability,(10) managing and organizing, (11) possible future pur-chases of each vendor, (12) employing communicationsystem, (13) operational controls, (14) position of thevendor in industry, (15) work relations history, (16) or-ganizational ethics, (17) job motivation, (18) complaintand warranty policies, (19) capabilities of productpackaging, (20) rate of in uence on the contact, (21)possibility to satisfy the educational demands, (22)improvement of the procedures, and 23) performancehistory.

Garvin explained the performance criteria in de-tail. He suggested the most dominant factors including:quality, cost, timely delivery, services, and exibil-ity [19].

Sarkar et al. divided the criteria of performance

as short term e�ect and those of capability as longterm e�ect on the achievement of the supply chain(Table 1) [16].

1.3. Fuzzy expert systemIn recent years, expert systems or knowledge-basedsystems have been the focus of a signi�cant amountof studies. These studies declare that expert systemprocesses knowledge while other software processesdata and information [20]. Expert systems are oneof the most practical �elds of the arti�cial intelligencethat implicate a collection of the understandings andregulations in a special �eld and ultimately simulatethe abilities of experts in speci�c problem domains [21].These are some powerful and exible tools which cansolve many problems that could not be solved bytraditional methods.

Fuzzy expert system is a knowledge-based system


which uses fuzzy logic instead of Boolean logic in itsknowledge-base and conclusions are made based on userinputs and fuzzy inference [22].

Further, some well-known methods such as partialcompleteness [23], optimized association rules [24], andCLIQUE [25] divided the qualitative attributes intomany crisp partitions. Since there are no interactionsbetween the partitions, this division does not seemfavorable for linguistic and qualitative parameters; forexample, if we tried to partition the range (70, 80)of the \COST" parameter for a supplier into twopartitions, then the separable point would not bedi�erent between 75.01 and 74.99. Hence, interactionwith any of the neighbor partitions can be promised.Moreover, we considered that the fuzzy associationrules described by the natural language were well suitedfor thinking about human subjects. They enhance the exibility of users in making decisions or designingthe fuzzy systems for agility evaluation. Thus, afuzzy partition method has been employed to �ndthe fuzzy association rules [26]. Marsh et al. utilizedthe combination of fuzzy logic and expert systemto evaluate contractors. Expert rules were used tocapture the expert's reasoning and fuzzy logic andmodel both subjective and objective factors [12]. Thecharacteristics of the developed fuzzy expert system aredescribed in greater details in the next section.

2. Research methodology

The �rst step in the evaluation of supplier performanceis to select the evaluation approach and then theevaluation criteria which should be identi�ed [27]. Theproposed method and criteria should have the followingfeatures:

1. They should have a logical basis;2. They should have adequate acceptance in the work

environment;3. The required information and tools should be

handy;4. They should provide the possibility for review and

continuous improvement.

In the developed expert system, the authorscovered all the above steps; thus, after modeling thefuzzy expert system for evaluation of the suppliers,the criteria were speci�ed by investigating a vast areaof literature after holding several justi�cation sessionswith many experts to persuade them to cooperate inthis research. Some meeting sessions were organizedby a group of 14 construction industry managers (con-struction experts) and 6 university professors (fuzzylogic experts); the average age and average experienceof the construction experts were almost 47 and 19 years,respectively, and for the fuzzy logic experts 44 and 17

years, respectively. By gaining bene�t of one set ofquestionnaires and two sets of statistical analyses, thecriteria were �nalized and the membership functionswere speci�ed by the experts. Next, the fuzzy logicalgorithms were developed by coding the rules usingMamdani's inference engine; after testing and con�rm-ing the results by the experts, the expert system wasdeveloped and the preparation phase was completed.

After �nal preparation of the fuzzy expert system,the validation and veri�cation steps were performed inthree major construction companies. These steps arementioned in Section 4.

2.1. Supplier and subcontractor evaluationcriteria

It is clear that e�ective and e�cient application of eval-uation criteria in the organizations requires acceptanceby the users. On the other hand, applying di�erentcriteria is in uential in the results. Furthermore,linguistic and fuzzy variables may have an apprecia-ble in uence on the decision-making process [7]. Inthis research, after collecting the criteria mentionedin the literature for supplier evaluation, the authorsimplemented two phases of criteria evaluation. In thesephases, the chosen criteria were classi�ed and �lteredby the experts and the resulted criteria and theirmembership functions were analyzed through one set ofquestionnaires and two statistical analyses. At last, two�nal groups of criteria with their membership functionswere speci�ed, distinctly, for evaluation of suppliersand subcontractors. The two evaluation phases arementioned in the following sections.

2.1.1. Evaluation phaseIn this phase, the whole collected criteria (over 82criteria) were evaluated by 20 experts. There were twomeeting sessions for achieving this goal. In the �rstsession, the research goals, plan, and collected criteriawere presented to the experts; the authors wantedthe experts to survey the criteria and prepare theirpreferred list with priorities for the next session. Inthe second session, the lists of experts were gatheredand the rating was scaled and averaged; then, thecompatibility of high-score criteria with constructionindustry was speci�ed; thus, after improving compati-bility between criteria and the construction industry,19 criteria were chosen as essential criteria for theconstruction industry to post-evaluate in the nextphase. These chosen criteria were classi�ed in fourgroups, including quality, price, timely delivery, and exibility.

2.1.2. Post-evaluation phaseIn this phase, 18 criteria were classi�ed in two groupsfor evaluation and selection of suppliers and subcon-tractors, distinctly. Then, the evaluation algorithmof the developed fuzzy expert system was presented.


Table 2. Selected criteria for suppliers.

Level 1(overall

evaluation)Level 2 Level 3 Level 4

Supp

lier

grad

e(m

ater

ial&

equi

pmen

t) 1) Quality

1- Process quality1. System evaluation score2. Adaptation with standards3. Ability to improve quality

2- Products quality

4. Products quality(return from customer)5. After-sale service quality6. Quality of initial materials

2) Cost

7. Superiority of purchase costscompared to other suppliers8. Superiority of transportationcosts compared to other suppliers9. Flexibility in payment conditionand discount

|

3) Work relationshistory and timely delivery

10. Company capabilities and proper exibility in the timely delivery11. History of successful cooperation12. Capacity

|

After consideration of inference algorithm and inves-tigation within criteria, 12 criteria were selected asthe �nal essential criteria for each group of suppli-ers and subcontractors (including 5 common crite-ria), distinctly, which will be presented in the nextsection. For improvement in evaluation results, theexperts suggested to merge exibility criterion intoall other three criteria and further combine timelydelivery with work relations history. Thus, the �nal12 criteria were classi�ed in three groups of quality,cost, and timely delivery-work relations history, inwhich there were some exibility-related sub-criteria.The criteria for both groups are illustrated in Ta-bles 2 and 3.

After the agreement of experts with these 19criteria, one questionnaire was circulated among themand the numbers in the range of values for eachlinguistic term in regard to the trapezoidal membershipfunctions needed for the fuzzy inference engine werespeci�ed for each �nal criterion by the experts; thevalues and membership functions were �nalized afterstatistical analysis for utilization in the expert system,which will represented in the next sections.

2.2. System architectureAppropriate system architecture would lead to facilita-tion of the development process. System architecturespeci�es system components and their interaction witheach other [28]. As it is shown in Figure 1, theproposed system architecture includes three main and

Figure 1. System architecture which include three mainand two subsidiary modules.

two subsidiary modules. The main modules are asfollows:

� Fuzzy inference engine: a module which analyzesand interprets the rules, membership functions, anddata in the database and logically concludes aresult. There are several fuzzy inference engines; inthis research, Mamdani's inference engine, which issuitable for capturing expert knowledge, has beenutilized. It allows us to describe the expertisein a more intuitive, more human-like manner. Inthe following paragraphs, the fuzzy logic operationsand Mamdani's inference engine mechanism arerepresented:

Fuzzy logic operations:Fuzzy union operation or fuzzy OR:

�A+B(x) = max (�A(x); �B(x)) :

Fuzzy intersection operation or fuzzy AND:

�A:B(x) = min (�A(x); �B(x)) :


Table 3. Selected criteria for subcontractors.

Level 1(overall

evaluation)Level 2 Level 3 Level 4

Supp

lier

grad

e(s

ervi

ce)

1) Quality1- Process quality

1. Knowledge level andskills of the personnel2. Certi�cates and standardsof activity3. Ability to improve quality

2- Service quality4. Service quality5. Relevant background6. Quality of initial materials

2) Cost

7. The superiority of the service costscompared to other suppliers8. Capabilities in minimizing waste of stu�and optimizing machinery consumption9. Flexibility in payment condition

|

3) Work relationshistory and timely delivery

10. Subcontractor capabilities and proper exibility in the timely delivery11. History of successful cooperation12. Ethics

|

Figure 2. Mamdani inference mechanism.

Mamdani [30] uses the following mechanism (Fig-ure 2):- Rule 1: If input1 is A11 and input2 is A12 THEN

output is C1;- Rule 2: If input1 is A21 and input2 is A22 THEN

output is C2;- Fact: Input1 is x0 and input2 is y0 consequence:

output is C.The fuzzy implication is modeled by Mam-

dani [30] as:

A AND B ! C = (A \B) \ C:

The �ring levels of the rules, denoted by �I, i = 1; 2,are computed by:

�1 = A11(x0) \A12(y0) = min fA11(x0); A12(y0)g ;�2 = A21(x0) \A22(y0) = min fA21(x0); A22(y0)g :

The individual rule outputs are computed by:

C 01(z) = �1 \ C1(z) = min f�1; C1(z)g ;C 02(z) = �2 \ C2(z) = min f�2; C2(z)g :

Then, the overall system output is computed by:

C(z) = C©1 (z) [ C©

2 (z) = maxnC©

1 (z); C©2 (z)

o:

� User interface: Users of the proposed fuzzy expertsystem are decision-makers who input the real valueof each linguistic variable into the system through auser interface. However, user interface displays thegrade and scores of producers and also o�ers therequired guidelines for making a contract.

� Fuzzy rule base: This module includes two parts,namely data and rules, which provide a mecha-nism for sort fuzzy rules. These fuzzy rules areconditional statements which are represented asfollow [29].

If x is Xi and y is Yi, THEN o is Oi., inwhich x and y are linguistic input variables; Xi and


Yi are their linguistic quantities which have beendetermined for a fuzzy function. Similarly, o andOi are respectively the fuzzy output variables andquantities [30].

In this fuzzy system, real values are employedfor scoring suppliers. As a result, the system hastwo subsidiary modules. Firstly and foremost isa fuzzi�er in the input that converts real variableto fuzzy set. In this research, we used trape-zoidal membership functions which are based onthe advices of experts and their re ections on theirknowledge. Secondly is a defuzzi�er in the outputthat converts the fuzzy set to a real value [31].For defuzzi�ng, we used Center Of Gravity (COG)method which �nds the geometrical center of theoutput variable.

Z� =�X

z:�c(z)�=�X

�c(z)�:

3. The proposed fuzzy expert system

The fuzzy expert system employs linguistic charac-teristics for the evaluation of factors such as quality,cost, exibility, and work relations history. Then,

these factors are merged by fuzzy inference and afterconducting an overall evaluation, they are assignedto one of the grades A, B, C, D, or E according tothe obtained scores. By clicking on each block, thesystem asks a series of questions about that block fromthe user. The consulting process is very user-friendly.Users can select a crisp within the range of 0-10 for eachcharacteristic. Afterwards, the fuzzy�er converts thecrisp variables into a fuzzy set which can be employedby fuzzy rule-base system and Mamdanis' inferenceengine for the evaluation of three criteria to determinethe supplier grade. The fuzzy�er performs the fuzzi-�cation process using trapezoidal fuzzy membershipfunctions (with low, medium, and high trapezoidalfunctions) available in the knowledge-based system. Itshould be noted that the above fuzzy functions havebeen determined separately for each characteristic, asmentioned in Section 2.1.2.

As shown in Table 4, the inference process is acombination of three stages (four levels) in this fuzzyexpert system. After receiving input through level 4,the �rst stage of fuzzy inference begins and transfersthe results to the higher stage within two sections centX and cent Y . Fuzzy inference engine uses cent X

Table 4. Fuzzy inference stages for the evaluation of suppliers.

Level 1(overall

evaluation)Stage 3 Level 2 Stage 2 Level 3 Stage 1 Level 4

Supp

lier

grad

e

Fuzz

yin

fere

nce

1) QualityFuzzy

inference1- Process quality2- Products quality

Fuzzyinference

1. System evaluationscore2. Adaptationwith standards3. Ability toimprove quality4. Products quality(return from customer)5. After-sale servicequality6. Quality of initialmaterials

2) CostFuzzy

inference

7. Superiority of purchase costscompared to other suppliers8. Superiority of transportationcosts compared to other suppliers9. Flexibility in payment conditionand discount

- -

3) Work relationshistory andtimely delivery

Fuzzyinference

10. Company capabilities and proper exibility in the timely delivery11. History of cooperation12. Capacity

- -


Figure 3. Fuzzy expert system for evaluation and selection of suppliers.

and cent Y to sort results of the previous level (withinlow, medium, or high) and transfer these sorted resultsto the upper level for the next fuzzy inference stage.Finally, through exchange of two �nal levels, the resultsof level 2 (within low, medium, or high) are sent to theinference engine and the �nal result will be determinedat level 1, within very low, low, medium, high, and veryhigh (Table 4).

Since software development has a high level ofimportance in the design of expert systems, an attemptis made in the present study to develop user-friendlysoftware.

At the end, after inputting the information onall criteria, the �nal grade of the supplier as well asthe required guidelines for making a contract with thesupplier and the obtained scores by the company aredisplayed in the block. In addition, the condition ofthree main criteria is also displayed (Figure 3). Thefollowings are some of the advantages of the proposedexpert system:

1. The knowledge-base is reliable and easy to employ;

2. This system is cost-e�ective;

3. It can be utilized in di�erent situations at the sametime;

4. Reportage is easy by it;

5. It provides the possibility of analysis and inter-pretation of the results during the process and,accordingly, educates the experts.

3.1. Membership functionsA Membership Function (MF) is de�ned as a curve inwhich each input point is mapped by a membershipvalue (or degree of membership) between 0 and 1; MFsplay crucial role in the overall performance of fuzzysystems. The nature of problems determines shape ofMFs; in this regard, triangular or trapezoidal shapesare simple to implement and fast for computation. Inthis research, the de�ned fuzzy MFs aim to utilize the

knowledge of the experts. Therefore, trapezoidal MFsare chosen based on the judgment of experts; theybelieve that trapezoidal shape could better representtheir objectives by the selected criteria and simplifyand facilitate the judgment process.

A trapezoidal membership function is formulated,as follows, with four parameters a, b, c, and d:

�F (x : a; b; c; d) =

8>>>>>>>>><>>>>>>>>>:

0; if x < a

(x� a)=(b� a); if a � x � b1; if b < x < c

(d� x)=(d� c); if c � x � d0; if d < x

A questionnaire survey was made to specify a, b, c,and d parameters for each criterion of membershipfunction. After averaging and statistical analysis,the �nal results were classi�ed and similar resultswere speci�ed. Therefore, based on similarities, forfacilitating the judgment process, experts decided toarrange trapezoidal input functions in four classes(Table 5); each criterion related function is shownin Table 6. Figures 4-8 illustrate the membershipfunctions related to the variable; classes A, B, C, D;and output variables, respectively.

Figure 4. Input trapezoidal fuzzy MF (class A).


Table 5. Four classes of trapezoidal membership functions.

Linguisticvariables

Fuzzy trapezoidalmembership

function(class A)


function(class B)


function(class C)


function(class D)

High (6.5, 8, 10, 10) (7.5, 8.5, 10, 10) (8.5, 9.5, 10, 10) (9, 9.8, 10, 10)Medium (3, 4.5, 6.5, 8) (4.5, 5.5, 7.5, 8.5) (6.5, 7.5, 8.5, 9.5) (7, 8, 9, 9.5)

Low (0, 0, 3, 4.5) (0, 0, 4.5, 5.5) (0, 0, 6.5, 7.5) (0, 0, 7, 8)

Table 6. Classi�cation of the criteria for suppliers within four classes of trapezoidal membership functions.

Class CriteriaA History of cooperation, and ability to improve quality.

B

Capacity, system evaluation score, adaptation withstandards, quality of initial materials, superiority ofpurchase costs compared to other suppliers, andsuperiority of transportation costs compared to other suppliers.

CCompany capabilities and proper exibility in thetimely delivery, exibility in payment condition anddiscount, and after-sale service quality.

D Products' quality (return from customer).

Figure 5. Input trapezoidal fuzzy MF (class B).

Figure 6. Input trapezoidal fuzzy MF (class C).

3.2. Description of fuzzy system rulesAs stated before, the developed fuzzy expert systemutilizes rules-data base and Mamdanis' inference enginefor data inference. Since several parameters have beenused for the evaluation of suppliers in this research,

Figure 7. Input trapezoidal fuzzy MF (class D).

Figure 8. Output trapezoidal fuzzy membership function.

a great number of rules are ready to be �red in thesystem. Further, all the mentioned rules are presentedsymmetrically.

In this system, by arranging and subdividing eachgeneral parameter in the duplex and triplex groups,


Table 7. 27 rules of inference in a triplex group.

Rule'snumber

Ability toimprove quality

Adaptationwith standards

Systemevaluation score

Processquality

1 High High Fine High2 Hight High Medium High3 High High Weak High4 High Medium Fine High5 High Medium Medium High6 High Medium Weak Medium7 High Low Fine High8 High Low Medium Medium9 High Low Weak Weak10 Medium High Fine High11 Medium High Medium High12 Medium High Weak Medium13 Medium Medium Fine High14 Medium Medium Medium Medium15 Medium Medium Weak Weak16 Medium Low Fine Medium17 Medium Low Medium Weak18 Medium Low Weak Weak19 Low High Fine High20 Low High Medium Medium21 Low High Weak Weak22 Low Medium Fine Medium23 Low Medium Medium Weak24 Low Medium Weak Weak25 Low Low Fine Weak26 Low Low Medium Weak27 Low Low Weak Weak

through combining them in the next levels and usingfuzzy inference at each stage, the �nal evaluation resultof each supplier will be determined.

Thus, at each triplex group, the system receivesthree inputs (in the form of low or medium or highstates). Accordingly, there are 27 rules as well as 3inputs for result inference, which are presented for onetriplex group shown in Table 7.

In a similar way, the other triplex criteria groupshave 27 registered rules in the rule base. As it is citedin Tables 2 and 3, the system involves a duplex criteriagroup including 9 rules. These groups for suppliers arerepresented in Table 8.

Inference of the last stage resulted in suppliergrade. The di�erence between the inference of thethree criteria of the group in the last stage and that ofother groups lies in the division of the results into �vecategories, i.e. very low, low, moderate, high, and veryhigh (E, D, C, B, and A). In other words, after changingthe name of parameter in Table 4, output of rows 1, 2,

4, and 10 became very high and that of rows 18, 24,26, and 27 turned into very low. It should be notedthat for the sake of better determination of suppliergrade in the last stage, the score of each reasoning wasrepresented in a round number within 1:100 which wascomputed from de�uzi�cation.

4. Implementation of fuzzy expert system

The presented expert system has been tested and usedin three construction companies for three years. Theobtained results from a fuzzy expert system were com-pared with those from the AHP expert system (whichwas designed by the authors, before) and validation ofthe results was approved by the experts. As a sample,the authors represent the application of fuzzy expertsystem for the selection and evaluation of vendorsand subcontractors in three projects of one of theconstruction companies. The projects are introducedin Table 9.


Table 8. Triplex and duplex criteria groups for suppliers.

Inferenceresult

Three criteria groups andtwo criteria groups

Supplier grade1. Quality2. Cost3. Work relations history- timely delivery

Work relations historyand timely delivery

1. Company capabilities and proper exibility inthe timely delivery2. History of cooperation3. Capacity

Cost

1. Superiority of purchase costs compared toother suppliers2. Superiority of transportation costs compared toother suppliers3. Flexibility in payment condition and discount

Quality of products1. Quality of products (return from customer)2. After-sale service quality3. Quality of initial materials

Process quality1. System evaluation score2. Adaptation with standards3. Ability to improve quality

Quality 1. Process quality2. Quality of products

Table 9. Three projects in which fuzzy expert system was utilized.

Project Utilization Location(province-country)

Ofogh tower Residential Isfahan-Iran

Aqueduct Porch water conduit Isfahan-Iran

Mehr project Residential in 14 blocks Isfahan-Iran

In these companies, an overall review was per-formed on the past processes of subcontractor selectionand purchases. Besides, the current and previousstrategies of the company in regard to evaluation andselection of suppliers were determined by utilizingthe available documents and performing a number ofinterviews. There were three clear strategies in theevaluation process of the company:

1. The supplier who has o�ered the lower price ispreferable;

2. It is advisable to avoid dealing with a single sup-plier, which increases the risk parameter in projectaccomplishment;

3. It is recommended not to work with too manysuppliers, because the supply chain managementmay get out of control.

The developed fuzzy expert system covers allthe above three strategies. In other words, it dealswith cost strategy through selecting the cost as aprimary parameter. In addition, it considers twoother strategies by presenting some guidelines andapproaches in relation to how to make a contract withan evaluated supplier. Utilization of the developedfuzzy expert system in each construction companyresulted in collection of a classi�ed archive of powerfulsuppliers in di�erent �elds. The above-mentioned


Table 10. Evaluation of steel framework subcontractors.

Producer Grade Score Recommendation

Supplier 1 A 85 Deal with a long term contract and allocate the major part of capacitySupplier 2 B 81 Deal with a short term contract and allocate the major part of capacitySupplier 3 B 68 Deal with a short term contract and allocate the major part of capacity

Table 11. Evaluation of steel framework subcontractors.

Frameworkcontractor

Grade Score Recommendation

Contractor 1 B 79 Deal with a short term contract in a major projectContractor 2 B 77 Deal with a short term contract in a major projectContractor 3 B 69 Deal with a short term contract in a major projectContractor 4 C 56 Deal with a short term contract in a minor projectContractor 5 C 53 Deal with a short term contract in a minor projectContractor 6 C 42 Deal with a short term contract in a minor projectContractor 7 D 31 No deal

archive includes a wide range of supplier informationsuch as technical speci�cations, capabilities, servicesor materials that they can provide, cost and productquality (service quality), and other speci�ed criteria.As two examples, results of evaluation and selection ofsome steel framework producers and steel frameworksubcontractors (without mentioning their names) arepresented in Tables 10 and 11.

5. Conclusion

In regard to the point that increases in the e�ciencyand service quality of suppliers and subcontractorsstrongly a�ect the performance of clients, revenue,and satisfaction, civil companies have increasingly em-ployed SCM in recent years. Therefore, the evaluationand selection of the right supplier and subcontractorare crucial for construction projects. There are so manycomplaints in the literature on traditional supplier-selection based on the lowest tender price. Therefore,the �rst step in supplier evaluation and selection wouldbe specifying the evaluation criteria. In this research,the criteria were speci�ed through 3 phases. First, theliterature was reviewed and the related supplier evalu-ation and selection criteria were gathered, comprehen-sively; next, the criteria database was transferred tothe experts for the pre-evaluation phase; and �nally,two sets of 12 criteria were selected, independently, forconstruction suppliers and subcontractors.

In this research, after specifying the criteria, theauthors prepared a multi criteria fuzzy framework byutilizing the judgment of experts for supplier andsubcontractor selection. As the proposed performancecriteria were linguistic, the fuzzy numbers, membershipfunctions, and fuzzy calculations were utilized to attain

the results in a realistic manner. The membershipfunctions which weighted the judgment process werespeci�ed through the judgment of experts. Therefore,the system user or client could utilize this facilitatedjudgment process through assigning his/her own scoreto each criterion for each supplier and attaining thesupplier preferences and the end results.

The developed model was tested in three ma-jor construction companies for two years and theimprovement results were satisfactory. Nevertheless,some challenges and improvement points exist for theidenti�ed system, due to the unique attributes of eachconstruction project and in addition, the di�erentinterests and preferences of clients and stakeholders;the unique way of weighting the criteria could beinappropriate for all the circumstances. Therefore, thefuture research could work on a dynamic approachto involve the clients and main stakeholders in theweighing of criteria in each project, simultaneously.On the other hand, although Mamdani method iswidely accepted for capturing expert knowledge, theMamdani-type fuzzy inference entails a substantialcomputational burden, so, the future research could beconcentrating on other inference mechanisms.

Acknowledgment

The authors deeply appreciate Dr. M.H. Fazel Zarandiand Engineer M. Hajimiri for their assistance to de-velop the fuzzy expert system.

References

1. Razmi, J. and Sifori, M. \Evaluation of the structuraldi�erences between lean and agile production in the


supply chain", In First International Conference on theSupply Chain Management and Information System,Tehran (2007).

2. Luo, X., Wu, C., Rosenberg, D. and Barnes, D. \Sup-plier selection in agile supply chains: An information-processing model and an illustration", Journal ofPurchasing & Supply Management, 15(4), pp. 249-262(2009).

3. Bhutta, K.S. and Huq, F. \Supplier selection problem:A comparison of the total cost of ownership andanalytic hierarchy process approaches", Supply ChainManagement: An International Journal, 7(3), pp. 126-135 (2002).

4. Banki, M.T., Smaeeli, B. and Ravanshadnia, M. \Theassessment of bidding strategy of Iranian construction�rm", International Journal of Management Scienceand Engineering Management, 4, pp. 153-160 (2008).

5. Singh, D. and Tiong, R.L.K. \A fuzzy decision frame-work for contractor selection", Journal of ConstructionEngineering and Management, 131, pp. 62-70 (2005).

6. Morote, A.N. and Vila, F.R. \A fuzzy multi-criteriadecision-making model for construction contractor pre-quali�cation", Automation in Construction, 25, pp. 8-19 (2012).

7. Salmanzadeh, S. and Javanrooh, S. \Evaluation theperformance of suppliers in the supply chain of autoparts", In First National Conference on the Logisticand Supply Chain Management and Information Sys-tem, Tehran (2004).

8. Chien, H.K., Min, Y.C. and Tsung, K.W. \Evaluatingsub-contractors performance using EFNIM", Automa-tion in Construction, 16, pp. 525-530 (2007).

9. Stevens, J. \Integrating the supply chain", Interna-tional Journal of Physical Distribution and MaterialsManagement, pp. 3-8 (1989).

10. Naim, M.M. \The book that changed the world",Manufacturing Engineer, pp. 13-16 (1997).

11. Mentzer, J.T., De Witt, W., Keebler, J.S., Min, S.,Nix, N.W., Smith, C.D. and Zacharia, Z.G. \De�ningsupply chain management", Journal of Business Lo-gistics, 22(2) pp. 1-25 (2001).

12. Marsh, C. and Fayek, A.R. \Surety Assist: Fuzzyexpert system to assist surety underwriters in evaluat-ing construction contractors for bonding", Journal ofConstruction Engineering and Management, 136(11),pp. 1219-1226 (2010).

13. Bevilacqua, M., Ciarapica, F.E. and Giacchetta, G. \Afuzzy-QFD approach to supplier-selection", Journalof Purchasing & Supply Management, 12, pp. 14-27(2006).

14. Neely, A., Richards, H., Mills, J., Platts, K. andBourne, M. \Designing performance measures: Astructured approach", International Journal of Oper-ations & Production Management, 17(11), pp. 1131-1152 (1997).

15. Neely, A., Mills, J., Platts, K. and Richards, H.G.\Performance measurement system design: Developingand testing a process-based approach", InternationalJournal of Operations & Production Management,20(10), pp. 1119-1145 (2000).

16. Sarkar, A. and Mohapatra, P.K.J. \Evaluation ofsupplier capability and performance: A method forsupply base reduction", Journal of Purchasing &Supply Management, 12(3), pp. 148-163 (2006).

17. Liu, J., Ding, F. and Lall, V. \Using data envelopmentanalysis to compare suppliers for supplier-selectionand performance improvement", Supply Chain Man-agement: An International Journal, 5(3), pp. 143-150(2000).

18. Vokurka, R.J., Choobineh, J. and Vadi, L. \A proto-type expert system for the evaluation and selection ofpotential suppliers", International Journal of Opera-tions & Production Management, 16(12), pp. 106-127(1996).

19. Garvin, D. \Management, strategic planning", Cali-fornia Management Review, pp. 85-106 (1993).

20. Darlington, K., The Essence of the Expert System,prentice-hall, England (2000).

21. Shariati, A. and Fatemi Ghomi, M. \Design of anexpert system and combination with analytic hierarchyprocess to evaluate and select the parts producer in thesupply chain", In Fourth International Conference onIndustrial Engineering, Tehran (2005).

22. Kandel, A.E. and Raton, B., Fuzzy Expert Systems,CRC Press (1992).

23. Srikant, R. and Agrawal, R. \Mining sequentialpatterns: Generalizations and performance improve-ments", in EDBT '96 Proceedings of the 5th Interna-tional Conference on Extending Database Technology:Advances in Database Technology, pp. 3-17 (1996).

24. Fukuda, T., Morimoto, Y., Morishita, S. andTokuyama, T. \Data mining using two-dimensionaloptimized association rules: Scheme, algorithms, andvisualization", in ACM SIGMOD International Con-ference on Management of Data, New York (1996).

25. Agrawal, R., Gehrke, J., Gunopul, D. and Raghavan,P. \Automatic subspace clustering of high dimensionaldata for data mining applications", in ACM SIGMODInternational Conference on Management of Data,New York, pp. 94-105 (1998).

26. Jain, V., Benyoucef, L. and Deshmukh, S.G. \A newapproach for evaluating agility in supply chains usingfuzzy association rules mining", Engineering Applica-tions of Arti�cial Intelligence, pp. 367-385 (2008).

27. Metaxiotis, K.S., Psarras, J.E. and Askounis, D.T.\GENESYS: an expert system for production schedul-ing", Industrial Management and Data Systems, pp.309-317 (2002).

28. Nunamaker, J., Chen, M. and Purdin, D. \Systemsdevelopment in information systems research", Journalof Management Information Systems, 7, pp. 89-106(1990).


29. Matthews, C. \A formal speci�cation for a fuzzy expertsystem", Information and software Technology, pp.419-429 (2003).

30. Klir, G.J. and Floger, T.A., Fuzzy Set, Uncetainty andInformation, Prentice-hall (1998).

31. Mohamadian, A. and Ghukasian, C.L. \Fuzzy expertsystem for the portfolio selection electronic businessmodels", in Second International Conference on Man-agement, Tehran (2004).

Biographies

Ehsan Shahvand obtained his BSc degree in CivilEngineering from Sharif University of Technology in2007 and his MSc degree in Construction Engineeringand Management from Amirkabir University of Tech-nology, with honor, in 2009. He is currently pursu-ing his studies as a PhD Candidate in ConstructionEngineering and Management at Amirkabir Universityof Technology, Tehran, Iran. His research interestsinclude strategic management, supply chain manage-ment, housing market analysis, economical analysisand �nancial management.

Mohammad Hasan Sebt obtained his BSc de-gree from the Jackson Mississippi State University inIndustrial Engineering and three MSc degrees fromJackson Mississippi State University and Universityof Kansas, Lawrence, in Computer Science, Busi-ness Administration, and Construction Engineeringand Management till 1990. Then, he obtained his

PhD degree from University of Kansas, Lawrence, inConstruction Engineering and Management in 1995.He has been a faculty member at The University ofKansas, USA, and Amirkabir University of Technology,Iran. He is Dean of the Management Department andConstruction Engineering and Management Group atAmirkabir University of Technology. He has authoredand coauthored many technical papers and books.His research interests include strategic management,project management, value engineering, and supplychain management.

Mohammad Taghi Banki obtained his BSc and MScdegrees from University of Tehran in Civil Engineeringin 1968 and then, he obtained another MSc degree inCivil Engineering (Structural Engineering) from NorthCarolina State University, Raleigh, in 1973. Then, heobtained his PhD degree in Civil Engineering (Con-struction Management) from University of Missouri,Department of Civil Engineering, Columbia, USA, in1980. He has been a faculty member at The Uni-versity of Calgary, Canada, and Amirkabir Universityof Technology, Iran. He was Minister of Planningand Budget Organization in 1982-1986 and Ministerof Energy (responsible for water and electivity) in1986-1988 in Islamic Republic of Iran and he hasother vital executive experiences. He has authoredand translated many books. His research interestsinclude project management, construction methodsand machineries, operational research, and economicanalysis.

Developing fuzzy expert system for supplier and...

Documents

Transcript of Developing fuzzy expert system for supplier and...