A new model to optimize the knowledge exchange …scientiairanica.sharif.edu/article_4065_00e595332522b5a...Power and Lundmark [34] 836 H.R. Dezfoulian et al./Scientia Iranica, Transactions

Scientia Iranica E (2017) 24(2), 834{846

Sharif University of TechnologyScientia Iranica

Transactions E: Industrial Engineeringwww.scientiairanica.com

A new model to optimize the knowledge exchange inindustrial cluster: A case study of Semnan plasterproduction industrial cluster

H.R. Dezfoulian, A. Afrazeh� and B. Karimi

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

Received 24 November 2015; received in revised form 9 January 2016; accepted 27 February 2016

KEYWORDSKnowledge exchange;Optimization;Industrial cluster;New mathematicalmodel;Organizational socialrelationships.

Abstract. Industrial clusters bring member �rms the opportunities and advantages tosave resources and increase competitiveness through cooperation and joint activities. Oneof these opportunities is knowledge exchange, using shared resources. If cluster �rms wantto create knowledge directly or acquire it out of cluster, it is necessary to spend muchmoney and time. The aim is to maximize knowledge transfer between �rms of a clusterregarding the limitation of budget and time, using existing knowledge ow networks. Thisproblem is formulated with a new model of mixed integer programming and solved by theCPLEX solver for Semnan plaster production industrial clusters. The results of sensitivityanalysis show that knowledge transfer is much more in uenced by budget than by timeconstraints. The results help cluster managers to have a better understanding, regardingthe available resources and business conditions, to maximize the results obtained fromknowledge transfer process in industrial cluster members.© 2017 Sharif University of Technology. All rights reserved.

1. Introduction

Knowledge is considered to be a critical resource tocreate competitive advantage among organizations [1-4]. Management and implementation of these resourcesare important activities in each organization. Cre-ation, sharing, and using of knowledge e�ectively forinnovation and sustaining competitive advantage areessential [5]. Knowledge management includes theprocess of identi�cation, acquisition, development andcreation, transfer, keeping, evaluation, and applicationof knowledge and helps organizations in achieving theirobjectives [6]. Huang [7] and Hsu [8] introduced knowl-edge transfer as a vital issue for organizations, since itempowers organizations to develop skills and compe-

*. Corresponding author. Tel.: +98 21 64545364E-mail address: [email protected] (H.R. Dezfoulian);[email protected] (A. Afrazeh); [email protected] (B.Karimi)

tencies and competitive advantages [9,10]. Knowledgetransfer is also considered as a requisite of successfulknowledge management.

Knowledge transfer in intra-organizational andinter-organizational levels can be carried out [11-13].The inter-organizational knowledge transfer can bedone horizontally and vertically. The former takesplace among the �rms with similar grounds of pro-duction and activity so that they can belong to anindustrial cluster and the latter usually occurs amongthe organizations existing in a supply chain in di�erentlevels [14]. Since the end of the 20th century, industrialcluster has been considered as the in ection point ofindustrial development programs and technology inmost countries of the world. The industrial cluster is anetwork of �rms with related industries that are locatedin an area [15-17]. These �rms form relationships witheach other to gain economic savings and have jointactivities.

After the business atmosphere became more com-

H.R. Dezfoulian et al./Scientia Iranica, Transactions E: Industrial Engineering 24 (2017) 834{846 835

petitive, the competition from �rm to �rm changedinto competition from cluster to cluster and chainto chain [18] and it became inevitable for �rms tocollaborate with each other to do joint activities toobtain success. Therefore, �rms try to venture jointactivities in integrated groups such as industrial clus-ters. One collaborative activity is knowledge transferamong cluster �rms. Acquiring knowledge from outsideor creating it within organization can be costly andtime{consuming. The close relationship between the�rms helps to facilitate the knowledge transfer and gainthe knowledge from others more cheaply in a shorterlength of time [19,20]. In addition to reduction in timeand cost of knowledge acquiring, knowledge transferenhances the synergy between the cluster members, im-proving the ability of cluster innovation and promotingthe overall competitiveness of the cluster members [21].

There have been di�erent studies about knowl-edge transfer in industrial clusters, but the academicresearch is limited to design and e�ective use of knowl-edge ow networks [22]. There has not been any specialresearch on the design of e�ective networks of knowl-edge transfer in industrial clusters with optimizationapproaches. Design of e�ective network of knowledgetransfer among the members of an industrial clusterbased on the existing organizational relationships in thecluster leads to the facilitation of knowledge transferand promotion of productivity of knowledge transferin the industrial cluster. This, in turn, provides theground to guide the managers to do e�ective actions.Although there is limited research about design ofknowledge ow networks, the necessity of this subjectdemands more studies [22].

In this paper, our focus is on how it is possible todesign e�ective knowledge transfer network based onthe existing social networks between the members of acluster, so that the level of knowledge among the clustermembers is maximized regarding the constraint of bud-get and time. To gain the best result and improve pro-ductivity of knowledge transfer, the important typesof knowledge are identi�ed for cluster members andin transfer of knowledge between members, they areprioritized, because the value of knowledge is the mostimportant factor in the transfer of knowledge [23,24].Also, in knowledge transfer, degree of closeness mustbe paid attention to and �rms with closer relationshipsare better candidates for knowledge transfer [25] as theknowledge transfer is carried out with lower expenses.Response to this question shows what knowledge inwhat time is transferred between which �rms in anindustrial cluster to maximize the knowledge level ofcluster members regarding the constraint of budgetand time. We formulated the problem in a newmixed integer programming model, solved it for a realproblem, and presented some insights for managers toproceed with sensitivity analysis.

Therefore, important and in uential parametersof this problem were identi�ed and it became possibleto make best decisions to exchange knowledge in clusterregarding the limitation of resources.

The rest of the paper will be presented as follows.The second section presents the literature review.Then, in the third section, the methodology is intro-duced and the de�nition of the problem and its mathe-matical formulation are presented in the fourth section.The �fth section is devoted to describing the case studyand computational results. Finally, concluding remarksare summarized in the sixth section.

2. Literature review

Literature review of this research contains the previousresearch in two parts. The �rst part refers to theresearch on knowledge transfer and related issues inindustrial clusters and the second part relates to orga-nizational social relationship and its use in knowledgetransfer.

2.1. Knowledge transfer in industrial clustersThere are a few studies about the transfer of knowl-edge in industrial clusters. Majority of studies aboutknowledge transfer in clusters are related to themethods and mechanisms of transfer. Sreckovic andWindsperger [26] examined the role of trust in usinga knowledge transfer mechanism. Richardson [27]focused on knowledge sharing through social inter-action in a policy-driven industrial cluster. Stackeet al. [28] analyzed knowledge transfer between clus-tered �rms and its relation to the competitivenessof tourism destinations in southern Brazil. Sreckovicand Windsperger [29] argued that \tacitness of thepartners' knowledge determines the information rich-ness of the knowledge transfer mechanisms in clus-ters". Lopez-Saez et al. [30] assessed the applicabilityof the SECI model (Nonaka and Takeuchi) to theprocesses of external knowledge acquisition for �rmslocated on knowledge-intensive clusters. Wilson andSpoehr [31] argued that informal knowledge transfersbetween skilled employees working in spatially boundedindustrial clusters might have an association with thelabour relationship between employers and employees.

The next group of studies examines the impactof various factors on knowledge transfer in industrialclusters. Ho�mann et al. [32] studied inter-�rm co-operation, industrial support institutions, workforcemobility, and social ties in the Brazilian furniturecluster. Xiong et al. [33] utilized the theory ofsystem dynamics to analyze the causal relationshipsof knowledge transfer in clusters. Four factors wereincorporated in this model: the supplier of knowledge,the recipient of knowledge, the knowledge gap, andthe transferring knowledge. Power and Lundmark [34]

836 H.R. Dezfoulian et al./Scientia Iranica, Transactions E: Industrial Engineering 24 (2017) 834{846

focused on the role of labor market and labor mobilityas sources of knowledge and ideas in industrial clusters.Dayasindhu [35] studied the e�ect of knowledge trans-fer and tacit factors, such as trust, experience, oppor-tunism, task complexity, and human asset speci�city,on international competitiveness of industrial clusters.

Another group of studies deals with knowledgenetwork in clusters. Guo and Guo [36] exploredthe evolution of knowledge network in manufacturingclusters considering the roles of participating actorsin evolution stages of knowledge network. Chen etal. [37] studied the impact of the regional socio-culturalbackground, depth of the shared knowledge, extent ofthe shared knowledge, and attributes of knowledge,partners, and network on e�ciency of knowledge net-works in industrial clusters. Giuliani [38] explored thestructural properties of knowledge networks in threeindustrial clusters by analyzing social networks.

Other papers have studied di�erent subjects.Zhou [39] introduced a four-stage spiral cycle modelfor knowledge di�usion in high-tech clusters. Yu [40]studied product knowledge ontology modeling in theindustrial cluster in order to solve the product infor-mation sharing problem. Ho�mann et al. [41] analyzedwhether �rm innovation was in uenced by the transferof knowledge among cluster �rms. Fang et al. [21] tookenterprise clusters as the research object in order toidentify various risks during the process of knowledgesharing by use of fuzzy mathematics and arti�cialneural network. Bocquet and Mothe [42] turned to thegovernment role in integration of knowledge in smallindustrial clusters.

The review of studies performed in this area showsthat these studies are of quality type and pay attentionto methods and factors in uencing knowledge transferin the cluster. The increasing competition in businessand resource limitation necessitate that we do notlimit ourselves to identify and survey the e�ect offactors in uential in knowledge transfer. Therefore,it is necessary to determine the change and propercombination of these factors along each other to get thebest result. The realization of this goal is possible byformulating the problem in mathematical optimizationmodel.

2.2. Organizational social relationships andtheir use in knowledge transfer

Previous studies in knowledge management show thatpower of organizational social relationships a�ects theproductivity of knowledge transfer remarkably [43-45].In these studies, there are two relationships identi�ed,which are used in knowledge transfer: strong and weakrelationships [46,47]. Hansen proposes a seven-levelscale to determine the power of relationships betweenorganizations [48]. Using a scale of greater degrees canbe useful in using the resources better. To determine

the power of relationships between �rms, the factorsin uencing them must be known.

The studies of factors in uencing the formation ofsocial ties are growing [22]. In these studies, di�erentfactors such as trust, resource exchange, communica-tion method, frequent communications, and physicalcloseness are known to be e�ective in formation and re-inforcement of social ties between organizations and thesta� [28,49]. In this paper, we focused on factors suchas �nancial relationships between �rms and their levels,labor force, raw materials and equipment exchange [28],physical closeness [42], and family relationships be-tween sta� [50] to determine the power of relationships.The presence of strong relationship between �rmsincreases the knowledge transfer e�ciency [51] andreduces the cost [25]. On the other hand, researcherswho have studied the social networks and organizationsbelieve that �rm success is dependent not only ontalents of the sta� and organizational capabilities, butalso on their interactions [52].

Of the research carried out, which turned to thee�ect of social networks on promotion of knowledgesharing in organizations, some can be mentioned.Cowan and Jonard [53], using simulation, studiedthe e�ect of di�erent types of network structures onknowledge distribution in organizations. Levine andPrietula [54], using agent-based simulation, studied thee�ect of di�erent types of ties between workers onknowledge sharing. The results showed that the perfor-mative ties in organizations reduced the average timeof �nishing tasks. Dong et al. [55] proposed a modelto integrate social network information, knowledge ofemployees, and availability of employees to benchmarkmanaging performance of knowledge-intensive serviceorganizations. They aimed to maximize the �nancialperformance of organization for a given social network.The sta� recognized the importance of using socialnetwork information and used it to facilitate knowledgesharing. In these studies, focus is on integratingfeatures of sta� and social network information ande�ective use of the available social network instead ofoptimum design and e�ective use of knowledge ownetwork.

More recently, research on designing organiza-tional social network has developed to improve knowl-edge management in organizations [22]. Leung andGlissmann [56] designed the organizational social net-works to improve skills of the sta�. They used cluster-ing approach to make connections among sta� basedon features and resources. In their clustering approach,there is no attention to the dynamic nature of sta�'sknowledge and the type of relationship among thesta� (in terms of strength and weakness). Zhuge [57]introduced a frame to combine the organizational socialrelationships and capabilities of knowledge sharing inorder to facilitate the design of a knowledge ow


network. He used trial and error in this frameworkand did not maximize the total knowledge level oforganization. Meanwhile, he did not pay attention topower of relationship amog the sta�, knowledge trans-fer time, and cost constraints for knowledge transferprogram. In another research, Dong et al. [22] studiedthe optimum design of knowledge ow networks amongthe sta� and sought to maximize the total knowledgelevel of organization in their mathematical model.

In the above studies, while ignored among orga-nizations, designing of social network among sta� hasbeen of interest. The closest research to ours is thatof Dong et al. [22] who turned to optimum designof knowledge ow network among the sta� using amathematical model. They sought to maximize theknowledge level of sta� in the objective function anddid not consider the time and cost constraints forknowledge transfer in the organization. There are someconstraints in real-life problems on time and cost, whichmost organizations face with. To have more attentionto social ties between organizations that facilitate andreduce the knowledge exchange cost, in addition tomaximization of knowledge level of cluster �rms in theobjective function, another section should be includedto increase the knowledge transfer between �rms thathave stronger relationships with each other. Attentionto �rm power to acquire and transfer knowledge andcapacity of training classes for simultaneous knowledgetransfer to various �rms is e�ective in making themodel real. We tried to consider these issues in ourproposed model to focus on real-life problems. Such amodel can help us to better understand the optimumdesign of knowledge ow network between �rms.

3. Methodology

This research is based upon the philosophy of posi-tivism while it uses quantitative method and followsdeductive approach. Evaluation of results, analysis ofdata and modeling, empirical data collection, gener-ation of models, and development of instruments aresome approaches of quantitative method. The strategythat is suitable for this study is survey, which is atechnique in which secondary data is collected fromdocuments of the �rms. In this study, quantitativeapproach is preferred as an appropriate tool and allthe results are presented in numbers and mixed integerprogramming model as well. Quantitative approachis a good �t for deductive approaches, in which atheory or hypothesis justi�es the variables, and thepurpose of quantitative research is imbuing the �ndingswith larger populations through an objective [58]. Thepurpose of this study is to carry out a case study.Population and units of analysis are �rms in Semnanplaster production industrial cluster and their panelsof experts.

4. De�nition and formulation of the problem

A number of �rms, which are located in a geographicarea with similar activities and products, have thepotential of creating an industrial cluster. If these�rms make strong relationships and create a networkof cooperation, then an industrial cluster is formed.Member �rms of an industrial cluster have the op-portunity of increasing their e�ciency by doing jointactivities like supplying, using distribution networks,utilizing communications technology, sta� training, do-ing research, and knowledge development. Knowledgetransfer among members is another opportunity, whichis the subject of this paper. It forms a knowledgenetwork, which brings member �rms the opportunity ofjointly transferring knowledge with lower costs. Clustermanagers can de�ne an incentive system to form suchknowledge networks. Their objective is maximizingmember �rms level of knowledge through increasingknowledge transfer in the shortest time and at thelowest cost.

Since the cluster member �rms are not quitesimilar in terms of size, purpose, type of market,product features, etc., the importance of di�erent typesof knowledge is not the same for them. Also, varioustypes of knowledge have not the same value for a�rm. Therefore, prioritization of knowledge types isnecessary when time and budget are limited. At thebeginning of the planning horizon, it is necessary todetermine the knowledge level of each �rm in all typesof knowledge. Three levels are de�ned for each type ofknowledge: beginner, medium, and expert. The �rmsthat have a higher knowledge level can transfer theirknowledge to other �rms in the process of knowledgetransfer. The time of the transfer of knowledge isdependent on complexity and tacitness of knowledge.

A mathematical model is developed to maximizeknowledge level of member �rms with limited budgetand time. This model is presented in the following. Inthe �rst part, the assumptions, notations, parameters,and variables of the model are described. Then, themixed integer programming model for this problem isintroduced.

AssumptionsThere are some assumptions related to the problem:

- The cost of the transfer of knowledge is dependenton the type of knowledge and closeness of �rms;

- Member �rms have limited capability of transfer-ring and acquiring knowledge simultaneously. Thiscapability is dependent on attributes of the �rm,such as the number of employees and their level ofexperience;

- Knowledge transferring �rm is at higher level ofknowledge than knowledge acquiring �rm;


- The �rm that is transferring a kind of knowledgeto others cannot simultaneously acquire the sameknowledge from other �rms;

- No �rm can acquire a higher level of the knowledgethat is being acquired until knowledge transferringperiod �nishes;

- The level of knowledge for a �rm increases when theknowledge transferring period �nishes;

- The total cost of knowledge transfers cannot exceedthe total budget allocated for the knowledge transferin the industrial clusters.

Notationsi; j; g Firms belonging to cluster 1; 2; � � � ;MK Types of knowledge 1; 2; � � � ;KT Time periods 1; 2; � � � ; TVariablesXtijk If the transfer of knowledge k from

�rm i to �rm j starts in the period t,the value is one, otherwise zero

Etjk If the �rm j in the period t isacquiring knowledge k, the value isone, otherwise it is zero

F tik If the �rm i in the period t istransferring knowledge k, the value isone, otherwise it is zero

Ltik Level of knowledge k in the �rm i atthe end of the period t (a positiveinteger variable)

ParametersDk The number of periods required to

transfer a level of knowledge kWik Importance of knowledge k for �rm iCijk Cost of transfer of knowledge k from

�rm i to �rm jC The total budget allocated for the

knowledge transfer in the industrialclusters

Ai The maximum number of knowledgetypes the �rm i can transfer toother �rms simultaneously (transfercapability of �rm i)

Bj The maximum number of knowledgetypes �rm j can acquire from other�rms simultaneously (acquiringcapability of �rm j)

� The maximum number of �rms thatcan acquire a certain type of knowledgefrom a �rm simultaneously (trainingclass capacity in cluster)

Rij Closeness of �rm i to �rm j

�; � Weight coe�cients to determine theimportance of each part of objectivefunction, sum of which must be one

4.1. Mathematical modelObjective function and 14 sets of constraints of themodel are the following.

4.1.1. Objective functionThe objective function consists of two parts. The�rst part is to maximize the total knowledge level ofindustrial cluster �rms regarding the knowledge values.The second part is to maximize the knowledge transferbetween �rms that have a closer relationship with eachother.

max

0BB@� MXi=1

KXk=1

Wik:LTik+�TXt=1

MXi=1

MXj=1;i 6=j

KXk=1

Rij :Xtijk

1CCA :(1)

4.1.2. ConstraintsThe constraints of this model can be listed as follows:

� If the level of �rm i in knowledge k is higher than�rm j at the beginning of the period t, then Xt

ijkcan be one (M 0 is a large enough number):

(1 +M 0):Xtijk � Ltik � Ltjk +M 0

i; j = 1; � � � ;M; i 6= j; k = 1; � � � ;K;t < (T �Dk + 1): (2)

� If Xtijk at the beginning of period t is one, then �rm

j cannot acquire knowledge k from another �rms inthe next Dk � 1 periods:

MXi=1;i6=j

t+Dk�1Xq=t+1

Xqijk �

0@1�MX

i=1;i6=jXtijk

1Aj = 1; � � � ;M; k = 1; � � � ;K;t � (T �Dk + 1): (3)

� In the last Dk � 1 periods of planning horizon, thetransfer of knowledge k cannot be started, becausethere is not enough time to transfer it completely:

TXq=T�Dk+1

Xqijk � 0 i; j = 1; � � � ;M;

i 6= j; k = 1; � � � ;K: (4)

� Knowledge level of �rm j remains the same forevery type of knowledge in the initial Dk periodsof planning horizon:

Lt+1jk =Ltjk j=1; � � � ;M; k=1; � � � ;K;t < Dk: (5)


� The knowledge k of �rm j increases one level aftercompletion of transfer period (Dk period):

Ltjk = Lt�1jk +

MXi=1;i6=j

Xt�Dkijk j = 1; � � � ;M;

k = 1; � � � ;K; t > Dk: (6)

� The knowledge level k of �rm i in all periods cannotbe higher than the de�ned maximum level (expertlevel):

Ltik�Lmax i=1; � � � ;M; k=1; � � � ;K;t = 1; � � � ; T: (7)

� No more than � �rms can acquire knowledge k from�rm i in period t simultaneously:

MXj=1;j 6=i

t+DkXq=t

Xqijk � � i = 1; � � � ;M;

k = 1; � � � ;K; t = 1; � � � ; T: (8)

� Constraints 9 and 10 control the capability of a�rm in simultaneous knowledge transferring. Every�rm i can transfer maximum Ai types of knowledgeto other �rms simultaneously in each period:

KXk=1

F tik�Ai i=1; � � � ;M; t=1; � � � ; T;(9)0@ MX

j=1;j 6=i

tXq=t�Dk+1

Xqijk=M

01A � F tik

i = 1; � � � ;M; k = 1; � � � ;K;t = 1; � � � ; T: (10)

� Constraints 11 and 12 control the capability of a �rmin simultaneous knowledge acquiring. Every �rm jcan acquire maximum Bj types of knowledge fromother �rms simultaneously in each period:

KXk=1

Etjk�Bj j=1; � � � ;M; t=1; � � � ; T;(11)0@ MX

i=1;i 6=j

tXq=t�Dk+1

Xqijk=M

01A � Etjk

j = 1; � � � ;M; k = 1; � � � ;K;t = 1; � � � ; T: (12)

� Firm j cannot acquire higher levels of knowledge k

from other �rms as long as it receives it from �rm i:

MXi=1;i6=j

t+DkXq=t

Xqijk � 1 j = 1; � � � ;M;

k = 1; � � � ;K; t = 1; � � � ; T: (13)

� Firm i cannot acquire knowledge k from other �rmsif it transfers the knowledge to �rm j:

Xtijk +

MXg=1;g 6=i

t+Dk�1Xq=t

Xqgik � 1

i; j = 1; � � � ;M; i 6= j; k = 1; � � � ;K;t = 1; � � � ; T: (14)

� The total costs of knowledge transfer from transfer-ring �rm i to acquiring �rm j in planning horizoncannot exceed the allocated budget C:

TXt=1

MXi=1

MXj=1;i6=j

KXk=1

Cijk:Xtijk � C: (15)

5. Case study

In this section, the proposed model for a real prob-lem is applied and the results from the model andsensitivity analysis are presented while denoting thecluster characteristics. The studied industrial cluster isSemnan production plaster that has 35 industrial active�rms in the production of plaster and its products.The distribution of �rms is seen in Figure 1. Around1000 and 4000 persons work in this cluster, directlyand indirectly, respectively. Semnan, with an annualproduction of 6 million tonnes of plaster, produces80% of the production of plaster in Iran. Averagely,29 people are employed in each �rm. At this stage,according to the idea of cluster managers, it is decidedto incorporate 10 highly prioritized types of knowledgeof clusters in the knowledge transfer program.

Figure 1. Distribution of plaster producing �rms in thecluster.


Initial knowledge level of member �rms was iden-ti�ed for each type of knowledge at the beginning ofthe horizon as input parameters. This identi�cation ofinitial knowledge level was based on the criteria de�nedby cluster experts and tailored to the performance ofthe �rm in the �eld of knowledge.

More complex and tacit knowledge needs moretime to be transferred between two �rms. Accordingto the view of experts, transfer of identi�ed typesof knowledge in this cluster takes 1 to 3 months,depending on their complexity and tacitness. Eachmonth is considered a single period in this model.

The cost of knowledge transfer from one �rmto another is related inversely to the closeness ofthe �rms and directly to the complexity and tac-itness of knowledge. In other words, the closerthe relationship between the �rms, the less the costis; and the higher the complexity and tacitness ofknowledge, the higher the cost of knowledge transferwould be [59]. The closeness of �rms can be in-terpreted in terms of �nancial transactions betweenthe �rms, exchange of labor, raw materials andequipment, adjacency, and the relationship amongtheir employees. The closeness of cluster membersis expressed with numbers between 1 and 9, where

number 1 indicates the weakest relationship and num-ber 9 represents the strongest relationship between the�rms.

The capability of �rms in simultaneous trans-ferring or acquiring of knowledge is limited, and isdependent on the number of employees and their levelsof education and experience. This capability variesbetween 2 and 6 types of knowledge for the �rms, whichare incorporated in this study.

Parameter � depends on facilities of the clusterand possibility of coordinating �rms to participate inthe program. The value of � in the Semnan plasterproduction industrial cluster was considered 4.

The values � and � are determined according tothe view of cluster managers. They suggest values �and � giving the importance of each objective function.In Table 1, values of parameters and the way oftheir measurement in the Semnan plaster productionindustrial cluster are presented.

The sum of levels of 10 highly prioritized knowl-edge types of member �rms in the cluster, whichare considered in the transfer process, is 714 at thebeginning of the planning horizon and would be 1050in the ideal conditions. It means that the maximumamount of knowledge that can be transferred is 336.

Table 1. Summary of the model parameters.

Parameters Range/value Average Unit Measurement and calculation basis

L1ik� 1-3 2.04 Level Performance of each �rm in the knowledge of

interest based on the de�ned criteriaDk 1-3 | Period (month) Rate of complexity and tacitness of knowledgeWik 0.03-0.19 0.1 | Opinion of the managers of each �rm

Cijk 11-49 | Currency Rate of complexity and tacitness of knowledgeand closeness degree of �rms to each other

C 7500 | Currency Financial power of cluster members and thebudget allocated to cluster managers

Rij 1-9 3.2033 |

Factors such as �nancial relationships between�rms and level of these relationships, exchangerate of labour force, raw materials andequipment, adjacency, and kin relationship amongsta� of the �rm

Ai 2-6 3.1428 |Factors such as number of workers in the sta�,their education and experience level, facilities,and the opportunity the �rm considers for education

Bj 2-6 3.2571 |Factors such as number of workers in the sta�,their education and experience level, facilities,and the opportunity the �rm considers for education

� 4 | No. Available software and hardware facilities in thecluster and the member �rms for education

�, � 0.5-0.5 | | Opinion of the cluster managers� The value Ltik for t = 1 is a parameter and for t = 2; � � � ; T is a variable.


5.1. ResultsThe model is solved with the exact solution methodand solver of CPLEX software of GAMS. The �rst partof the objective function is related to maximizationof the �rms' knowledge level. The results show thatin a 13-month horizon, knowledge level of �rms inthe cluster, in 10 types, with cost of 7160 monetaryunits increases to the expert level (reaching ideal levelof 1050 from the initial sum of 714). Limitation ofresources does not allow the cluster managers to attainthis level of knowledge transfer. Therefore, they canmake decision about the selection of a combination ofbudget and time based on the results of sensitivityanalysis and resources available for the best level ofknowledge transfer. The second part of the objectivefunction is related to enhancing the knowledge transferbetween �rms with stronger relationships. This reducesthe cost of transferring knowledge between membersof the cluster. Figure 2 represents the distribution ofdi�erent levels of relationship between member �rms.Relationship level 1 is the weakest relationship andrelationship level 9 is the strongest relationship. Whilemajority of relationships between member �rms areweak (less than 5), knowledge transfers are mainlydone through strong relationships (more than 5). Thisre ects that existing networks in cluster can be used toreduce the cost and facilitate the knowledge transfer.

Another important issue that should be noted isthe role of the �rms with expert and medium knowledgein the transfer of knowledge to other members of thecluster. The results indicated that 90.5% of the trans-ferred knowledge in the cluster was from the �rm thathad experts in knowledge and the the share of �rmswith a medium knowledge in transfer was only 9.5%. InTable 2, the share of transfer of knowledge between the�rms according to their knowledge level is presented.

The contributions of �rms to knowledge transferprogram are not the same. As seen in Figure 3, 13�rms with less than 5 cases of knowledge transferhave the least contribution and 3 �rms with more

Figure 2. The existing relationships between the clustermembers and relationship types used in knowledgetransfer process.

Table 2. The share of �rms in knowledge transferregarding their knowledge level.

From mediumto beginner

From expertto medium

From expertto beginner

9.5% 82% 8.5%

Figure 3. Firms frequency in terms of number ofknowledge transfers.

than 20 cases of knowledge transfer have the greatestcontribution to knowledge transfer. Close scrutinyof the input data of the model shows that mostly�rms with closer relationships with other membersand higher initial levels of knowledge have greatercontributions to knowledge transfer to others.

The results of this section show that the �rmswhich are expert in the knowledge of interest and havea close relationship with other members have the mostin uential role in the knowledge ow network of thecluster. Strength of network of �rms in a cluster hasprime importance in increasing the cluster potential forknowledge transfer.

In the next section, the results of examiningthe e�ects of model parameters on the �nal solutionare presented so that the managers can choose thebest combination of resources based on conditions andresources available.

5.1.1. Sensitivity analysis of budgetWithout budget constraint, 336 transfers of knowledgeoccur and the ideal level of cluster knowledge (i.e.,1050) at a cost of 7313 monetary units is achievable. Inthe case of budget constraint, the number of knowledgetransfers decreases. This relation between reduction ofbudget and decreasing number of knowledge transfersis shown in Figure 4, in which, percentages are de�nedwith respect to the ideal point, i.e. allocation of 7313monetary units and 336 knowledge transfers betweenmembers.

5.1.2. Sensitivity analysis for planning horizonAchieving 336 knowledge transfers in the cluster re-quires a planning horizon of at least 13 periods. Short-ening the length of planning horizon reduces the totalnumber of knowledge transfers in the cluster, which


Figure 4. Sensitivity analysis of budget.

Figure 5. Sensitivity analysis for planning horizon.

in turn leads to reduction in the total level of clusterknowledge. Figure 5 shows the e�ect of reduction inthe planning horizon on the total knowledge transfersin the cluster.

5.1.3. Analysis of the relationship between trainingclass capacity and knowledge transfer cost

The presence of more representatives of �rms in a classin order to acquire a type of knowledge is allowed ifthe facilities to increase the capacity of the classes arecreated. This requires spending part of the budget toincrease the capacity of training classes in the �rms.The increased capacity will help to reduce the cost andtime of knowledge exchange; however, the results ofsensitivity analysis in Figure 6 indicate that the changein capacity of trainnig classes from the present situation(� = 4) contributes up to 0.5% to the reduction in thecost of transfering knowledge, which can be completelyignored; in facet, it does not have any e�ect on thechange of the knowledge transfer time (horizon ofplanning) and is neglected.

5.1.4. Analysis of the relationship between the powerof �rms in simultaneous knowledge exchangeand planning horizon

Capacity of �rms for simultaneous knowledge exchngeis an important factor in this model. This factor is

Figure 6. Relationship between training class capacityand knowledge transfer cost.

Figure 7. Relationship between strengthening the powerof �rms for simultaneous knowledge exchange andplanning horizon.

entered into the model with parameters Ai and Bj .Increase in these parameters does not a�ect knowledgetransfer costs, but it may provide the possibility ofreaching the ideal transfer number (336) in a shorterlength of time (Figure 7). For example, 10% increase inpower of knowledge exchnge in the �rms makes transferof 336 cases of knowledge become possible in 9 periods(4 periods fewer).

5.1.5. Analysis of the relationship between closenessdegree of the cluster members and knowledgetransfer cost

The quality of relationship between the �rms is anotherin uencing factor on the costs of knowledge transfer inindustrial clusters. Strengthening these relationshipsbrings the cluster managers the opportunity of reducingthe costs. For example, in Figure 8, 25% increasein strength of the relationships between the memberscauses 13% reduction in the costs of knowledge trans-fers.

5.2. DiscussionThe status of cluster members' level of knowledge in themost important 10 types of knowledge at the beginning


Figure 8. The relationship between strengtheningcloseness degree of the cluster members and knowledgetransfer cost.

Table 3. Knowledge level of �rms in the commencementof planning horizon.

Beginner level Medium level Expert level

17.5% 61% 21.5%

of the horizon is presented in Table 3. Results ofTable 3 show that expert-level �rms are the mostin uential members for knowledge transfers and are incharge of 90% of the total knowledge transfers in thecluster. It is worth mentioning that an important partof the expert �rms are those who were at medium levelat �rst and reached the expert level through knowledgetransfer process.

The strength of relationships among membersof the cluster is an important factor that facilitatesknowledge transfer process and reduces the costs.The results show that the knowledge transfer in theplanning horizon occurs only between �rms with strongrelationships. This suggests utilizing the existingnetworks of the �rms for the knowledge transfers.

Table 4 summarizes the impact of model parame-ters on cost and time of knowledge transfers. Changingparameters � and � by determining the importance ofparts of each objective function and the increase in thecapacity of the training classes, i.e. parameter �, donot have a signi�cant impact on cost and duration ofthe knowledge transfer in the cluster. Parameter Rijis an indicator of closeness of �rms, and parametersAi and Bj stand for the �rm's power of simultaneous

knowledge transfer that can reduce the cost and time,respectively. The results show that cluster managerscan focus on strengthening the networks and increasingthe capability of �rms in absorbing and transferringknowledge to facilitate knowledge exchange with lowertime and costs.

The sensitivity of the model to knowledge transferbudget is much more than time horizon, according toFigures 4 and 5. This fact helps cluster managers intheir decision-making and shows that budget cuts canstrongly reduce the chance of implementing a successfulknowledge transfer program.

6. Conclusion

The most important advantage of the formation ofindustrial clusters is joint activities of cluster members,which saves resources and facilitates synergy. Animportant instance is knowledge transfer using sharedresources. With the identi�cation of the in uencingfactors, we can design a cost-e�ective knowledge ownetwork, which maximizes knowledge transfer in agiven time span. Accordingly, a mathematical modelwas developed in this paper for optimal design ofknowledge ow network in industrial clusters. Then,the proposed model was applied in a real-world casestudy.

The results of solving the model revealed severalimportant facts about the knowledge transfer betweenthe member �rms. First, expert members were keyplayers in knowledge ow network and made a mo-mentum, which accelerated knowledge transfers in thecluster. Therefore, the importance of these membersshould not be neglected in knowledge transfer program.Second, the process of knowledge transfer in the clusterdepended heavily on knowledge transfer funding andbudget cuts had a great impact on reducing the numberof knowledge transfers among members. Third, in-creasing the capability of �rms in knowledge absorptionand transfer provided the opportunity of reducingtime horizon of knowledge transfer program in indus-trial clusters. Fourth, strengthening the relationshipbetween the �rms signi�cantly reduced the costs ofknowledge transfer in the cluster. These facts helpcluster managers to understand the most importantfactors in designing and implementing the knowledge

Table 4. E�ect of model parameters on cost and time of knowledge transfer.

Parameter Knowledgetransfer cost

Knowledgetransfer time

E�ciency

Changes in � and � | | Ignorable

Increase of � | | Ignorable

Increase of Rij Reduction | To 13%

Increase of Ai and Bj | Reduction To 46%


transfer program in industrial clusters. They also candesign knowledge ow network to maximize knowledgetransfers in the cluster, with minimum cost and time.

In further research, the dynamic nature of rela-tionships among cluster members should be consideredas the interactions between cluster �rms can causethem to become closer over time. The horizontalintegration in supply chain is another issue to bestudied in the development of the suggested model.

References

1. Jasimuddin, S.M. \Exploring knowledge transfermechanisms: The case of a UK-based group withina high-tech global corporation", International Journalof Information Management, 27, pp. 294-300 (2007).

2. Narteh, B. \Knowledge transfer in developed-developing country inter�rm collaborations: a concep-tual framework", Journal of Knowledge Management,12(1), pp. 78-91 (2008).

3. Hammami, H., Amara, N. and Landry, R. \Organiza-tional climate and its in uence on brokers` knowledgetransfer activities: A structural equation modeling",International Journal of Information Management,33, pp. 105-118 (2013).

4. Sheng, M.L., Chang, S.Y., Teo, T. and Lin, Y.F.\Knowledge barriers, knowledge transfer, and inno-vation competitive advantage in healthcare settings",Management Decision, 51(3), pp. 461-478 (2013).

5. Chamberlin, T. and Doutriaux, J. \Sourcing knowl-edge and innovation in a low-technology industry",Industry and Innovation, 17(5), pp. 487-510 (2010).

6. Afrazeh, A. \Knowledge management (concepts, mod-els, measurement and implementation)", AmirkabirUniversity of Technology, Tehran, Iran (2009) (InPersian).

7. Huang, Ch. \Knowledge sharing and group cohesive-ness on performance: an empirical study of technologyR&D teams in Taiwan", Technovation, 29, pp. 786-797(2009).

8. Hsu, C. \Knowledge sharing practices as a facilitat-ing factor for improving organizational performancethough human capital: Preliminary test", ExportSystems with Applications, 35, pp. 1316-1326 (2008).

9. Ji, L-M., Hung, J., Chen, S-W. and Jou, C. \Fosteringthe determinants of knowledge sharing, virtual com-munities", Computers in Human Behavior, 25(4), pp.929-939 (2009).

10. Wang, S. and Noe, R.A. \Knowledge sharing: A reviewand directions for future research", Human ResourceManagement Review, 20, pp. 115-131 (2010).

11. Fang, S.C., Yang, C.W. and Hsu, W.Y. \Inter-organizational knowledge transfer: the perspective ofknowledge governance", Journal of Knowledge Man-agement, 17(6), pp. 943-957 (2013).

12. Hutzschenreuter, T. and Horstkotte, J. \Knowledgetransfer to partners: a �rm level perspective", Journalof Knowledge Management, 14(3), pp. 428-448 (2010).

13. Wijk, R.V., Jansen, J.J.P. and Lyles, M.A. \Inter-and intra-organizational knowledge transfer: A meta-analytic review and assessment of its antecedents andconsequences", Journal of Management Studies, 45,pp. 830-853 (2008).

14. Kotabe, M., Martin, X. and Domoto, H. \Gaining fromvertical partnerships: knowledge transfer, relationshipduration, and supplier performance improvement inthe US and Japanese automotive industries", StrategicManagement Journal, 24, pp. 293-316 (2003).

15. Malmberg, A. and Maskell, P. \The elusive conceptof localization economies: towards a knowledge basedtheory of spatial clustering", Environment and plan-ning, 34(3), pp. 429-449 (2002).

16. Porter, M.E. \Clusters and the new economics ofcompetition", Harvard Business Review, 76(6), pp. 77-90 (1998).

17. Porter, M.E. \Location, competition, and economicdevelopment: local clusters in a global economy",Economic Development Quarterly, 14(1), pp. 15-34(2000).

18. Hult, G.T.M., Ketchen, D.J. and Arrfelt, M. \Strategicsupply chain management: Improving performancethrough a culture of competitiveness and knowledgedevelopment", Strategic Management Journal, 28, pp.1035-1052 (2007).

19. Broekel, T. and Binder, M. \The regional dimension ofknowledge transfers-A behavioral approach", Industryand Innovation, 14(2), pp. 151-175 (2007).

20. Kesidou, E. and Snijders, C. \External knowledgeand innovation performance in clusters: Empiricalevidence from the uruguay software cluster", Industryand Innovation, 19(5), pp. 437-457 (2012).

21. Fang, Y., Liang, Q. and Jia, Z. \Knowledge sharingrisk warning of industry cluster: An engineering per-spective", Systems Engineering Procedia, 2, pp. 412-421 (2011).

22. Dong, S., Johar, M. and Kumar, R. \Understandingkey issues in designing and using knowledge ownetworks: An optimization-based managerial bench-marking approach", Decision Support Systems, 53, pp.646-659 (2012-2).

23. Fatahi, K. and Afrazeh, A. \An integrated model forevaluating knowledge value in organizations", Interna-tional Journal, WSEAS Transaction on InformationScience and Applications, 2(2), pp. 138-143 (2005).

24. Liu, C. and Chen, S. \Determinants of knowledge shar-ing of e-learners", International Journal of Innovationand Learning, 2(4), pp. 434-445 (2005).

25. Fritsch, M. and Kau�eld-Monz, M. \The impact ofnetwork structure on knowledge transfer: an applica-tion of social network analysis in the context of regional


innovation networks", The Annals of Regional Science,44(1), pp. 21-38 (2010).

26. Sreckovic, M. and Windsperger, J., The Impact ofTrust on the Choice of Knowledge Transfer Mecha-nisms in Clusters, Network Governance Contributionsto Management Science, Springer-Verlag Berlin Hei-delberg, pp. 73-85 (2013).

27. Richardson, C. \Knowledge-sharing through social in-teraction in a policy-driven industrial cluster", Journalof Entrepreneurship and Public Policy, 2(2), pp. 160-177 (2013).

28. Stacke, A.R.N.P., Ho�mann, V.E. and Costa, H.A.\Knowledge transfer among clustered �rms: a studyof Brazil", Anatolia - An International Journal ofTourism and Hospitality Research, 23(1), pp. 90-106(2012).

29. Sreckovic, M. and Windsperger, J. \Organization ofknowledge transfer in clusters: a knowledge-basedview", In New Developments in the Theory of Net-works, Franchising, Alliances and Cooperatives, M.Tuunanen, J. Windsperger, G. Cliquet, G. Hendrikse,Eds., Springer, Berlin, pp. 318-334 (2011).

30. Lopez-Saez, P., Navas-Lopez, J.E., Martin-de-Castro,G. and Cruz-Gonzalez, J. \External knowledge ac-quisition processes in knowledge-intensive clusters",Journal of Knowledge Management, 14(5), pp. 690-707 (2010).

31. Wilson, L. and Spoehr, J. \Labour relations and thetransfer of knowledge in industrial clusters: why doskilled workers share knowledge with colleagues inother �rmsin Industrial Clusters: Why do skilled work-ers share knowledge with colleagues ?", GeographicalResearch, 48(1), pp. 42-51 (2010).

32. Ho�mann, V.E., Lopes, G.S.C. and Medeiros, J.J.\Knowledge transfer among the small businesses of aBrazilian cluster", Journal of Business Research, 67,pp. 856-864 (2014).

33. Xiong, J., Duan, Z. and Wang, Y. \Modeling and sim-ulation of the inter-organizational knowledge transferimpact factors in industrial clusters", The 19th Inter-national Conference on Industrial Engineering and En-gineering Management, Springer-Verlag, Berlin, Hei-delberg, pp. 161-171 (2013).

34. Power, D. and Lundmark, M. \Working throughknowledge pools: Labor market dynamics, the trans-ference of knowledge and ideas, and industrial clus-ters", Urban Studies, 41(5/6), pp. 1025-1044 (2004).

35. Dayasindhu, N. \Embeddedness, knowledge transfer,industry clusters and global competitiveness: a casestudy of the Indian software industry", Technovation,22, pp. 551-560 (2002).

36. Guo, J. and Guo, B. \The evolution of knowledgenetwork in manufacturing cluster: A case in China",Proceedings of the IEEE IEEM, pp. 890-894 (2008).

37. Chen, J., Chen, D. and Li, Z. \The analysis ofknowledge network e�ciency in industrial clusters",

International Seminar on Future Information Technol-ogy and Management Engineering, IEEE, pp. 257-260(2008).

38. Giuliani, E. \The selective nature of knowledge net-works in clusters: evidence from the wine industry",Journal of Economic Geography, 7, pp. 139-168 (2007).

39. Zhou, S. \Study of knowledge di�usion FSAI modelfor high-tech SMES clusters", The 19th InternationalConference on Industrial Engineering and EngineeringManagement, Springer-Verlag, Berlin, Heidelberg, pp.783-794 (2013).

40. Yu, J. \Study on product knowledge ontology for in-dustrial cluster", International Symposium on Knowl-edge Acquisition and Modeling, IEEE, pp. 108-112(2008).

41. Ho�mann, V.E., Bandeira-de-Mello, R. and Molina-Morales, F.X. \Innovation and knowledge transfer inclustered interorganizational networks in Brazil", LatinAmerican Business Review, 12, pp. 143-163 (2011).

42. Bocquet, R. and Mothe, C. \Knowledge governancewithin clusters: the case of small �rms", KnowledgeManagement Research & Practice, 8, pp. 229-239(2010).

43. Afrazeh, A. and Zarinozv, I. \A conceptual frameworkfor presenting and promoting human resource knowl-edge sharing in the supply chain as a social networkpromoting human resource knowledge sharing in thesupply Chain as a social", International Journal ofServices, Economics and Management (Special Issueon: Modeling and Analysis of Supply Chain Networks),2(3/4), pp. 227-245 (2010).

44. Borgatti, S.P. and Cross, R. \A relational view ofinformation seeking and learning in social networks",Management Science, 49(4), pp. 432-445 (2003).

45. Cross, R., Parker, A., Prusak, L. and Borgatti, S.P.\Knowing what we know: supporting knowledge cre-ation and sharing in social networks", OrganizationalDynamics, 30(2), pp. 100-120 (2001).

46. Guy, I., Jacovi, M., Shahar, E., Meshulam, N., Soroka,V. and Farrell, S. \Harvesting with SONAR - Thevalue of aggregating social network information", InOnline Social Networks, Proceeding of the Twenty-Sixth Annual SIGCHI Conference on Human Factorsin Computing Systems, M. Burnett, M. Costabile, T.Catarci, B. Ruyter, D. Tan, M. Czerwinski, and A.Lund, Eds., ACM Press, New York, pp. 1017-1026(2008).

47. Levine, S.S. and Kurzban, R. \Explaining clustering insocial networks: towards an evolutionary theory of cas-cading bene�ts", Managerial and Decision Economics,27(2-3), pp. 173-187 (2006).

48. Hansen, M., Using the Wisdom of Others: Searchingfor and Transferring Knowledge, Presentation at theLondon Business School (1996).

49. Kotlarsky, J. and Oshri, I. \Social ties, knowledge shar-ing and successful collaboration in globally distributed


system development projects", European Journal ofInformation Systems, 14(1), pp. 37-48 (2005).

50. Arregle, J.L., Hitt, M.A., Sirmon, D.G. and Very,P. \The development of organizational social capital:Attributes of family �rms", Journal of ManagementStudies, 44(1), pp. 73-95 (2007).

51. Hansen, M.T. \The search-transfer problem: the roleof weak ties in sharing knowledge across organizationsubunits", Administrative Science Quarterly, 44(1),pp. 82-111 (1999).

52. Hidalgo, C.A. \The value in the links: networksand the evolution of organizations", (2010). Available:http://www.chidalgo.org/Papers/Hidalgo SageChap-ter .pdf.

53. Cowan, R. and Jonard, N. \Network structure andthe di�usion of knowledge", Journal of EconomicDynamics and Control, 28, pp. 1557-1575 (2004).

54. Levine, S.S. and Prietula, M. \Towards a contin-gency theory of knowledge exchange in organizations",Academy of Management Best Paper Proceedings, At-lanta, GA (2006).

55. Dong, S., Johar, M. and Kumar, R. \A benchmarkingmodel for management of knowledge-intensive servicedelivery networks", Journal of Management Informa-tion Systems, 28(3), pp. 127-160 (2012-1).

56. Leung, Y.T. and Glissmann, S.M. \A clusteringapproach to the design of knowledge-intensive serviceproviders", (2011). Available: http://domino.research.ibm.com/library/cyberdig.nsf/papers/58C6B1D509E8DCA68525780000603960.

57. Zhuge, H. \Knowledge ow network planning andsimulation", Decision Support Systems, 42, pp. 571-592 (2006).

58. Borrego, M., Douglas, E.P. and Amelink, C.T. \Quan-titative, qualitative, and mixed research methods inengineering education", Journal of Engineering Edu-cation, 98(1), pp. 53-66 (2009).

59. Leppala, S. \An epistemological perspective on knowl-edge transfers: From tacitness to capability and reli-ability", Industry and Innovation, 19(8), pp. 631-647(2012).

Biographies

Hamid Reza Dezfoulian is a PhD student in theDepartment of Industrial Engineering at AmirkabirUniversity of Technology and Lecturer in the De-partment of Industrial Engineering at Bu-Ali SinaUniversity. His main research interest lies in the �eld ofknowledge management and industrial plant planning.

Abbas Afrazeh is currently Associate Professor ofIndustrial Engineering at Amirkabir University ofTechnology, Tehran, Iran. His main research interestlies in the �eld of knowledge management and humanresource management.

Behrooz Karimi is currently Professor of IndustrialEngineering at Amirkabir University of Technology,Tehran, Iran. His main research interest lies in the�eld of supply chain management and productionplanning.

Documents

A new model to optimize the knowledge exchange …scientiairanica.sharif.edu/article_4065_00e595332522b5a...Power and Lundmark [34] 836 H.R. Dezfoulian et al./Scientia Iranica, Transactions