Complex Network Method Towards Evaluating Industrial Symbiosis

Transcription

1 169 A publcaton of CHEMICAL ENGINEERING TRANSACTIONS VOL. 61, 2017 Guest Edtors: Petar S Varbanov, Rongxn Su, Hon Loong Lam, Xa Lu, Jří J Klemeš Copyrght 2017, AIDIC Servz S.r.l. ISBN ; ISSN The Italan Assocaton of Chemcal Engneerng Onlne at DOI: /CET Complex Network Method Towards Evaluatng Industral Symboss Zheng Wang a, *, Yng Jang a, Yaoguo Huang a, Xaopng Ja b a College of Chemcal Engneerng,Qngdao Unversty of Scence and Technology, No. 53, Zhengzhou Road,Qngdao,Chna b College of Envronment and Safety Engneerng,Qngdao Unversty of Scence and Technology, No. 53, Zhengzhou Road, Qngdao, Chna jyemly@163.com The constructon of ndustral park has become one of the man ways of cleaner producton, energy savng and emsson reducton n ndustral development. Most ndustral symboss (IS) networks are very complex. A method of evaluatng IS networks s proposed based on complex network theory. Frstly, accordng to TOPSIS (technque for order preference by smlarty to an deal soluton), takng the mult-attrbute decson-makng method to dentfy key enterprses n IS networks. Secondly, t realzes quanttatve detecton by the communty partton of complex network. Fnally, we evaluate the results of detecton based on Grvan-Newman modularty and the coheson of communtes. The case analyss shows that the mult-attrbute decson-makng method can dentfy core enterprses effectvely. The quanttatve detecton for communty structure of IS networks s feasble wth the method of complex network communty detecton. The ntegraton degree of IS networks can be evaluated from the aspects of nodes and communty structures. 1. Introducton Symboss s materal contact between two or more dfferent members. Wth further research, the deas of symboss have been developed more extensve research and appled n many felds. IS networks are the producton organzaton pattern of symbotc relatonshp. It s the major form of cooperaton among enterprses, and has the dual nature of ndustral ecologcal system and network organzaton. It manly emphaszes chan dstrbuton of resources and energy among enterprses (Porkka et al., 2013). Not only does t mprove energy and resource effcency, but t reduces the destructon of ecologcal envronment. Makng full use of resource and energy can realze clean producton of IS system and mnmze the waste of resource and energy, such as wastewater treatment (Ochoa, 2016) and waste heat utlzaton (Yong et al., 2016). IS system has become the effectve envronmental management way of ndustral parks. The wn-wn pattern of ndustral producton and envronmental protecton has been acheved (Raafat et al., 2013). Amed at the evaluaton of symbotc system, many scholars put forward evaluaton methods. The asymmetrc dstrbuton coeffcent s proposed to analysze the stablty of symbotc boenergy park (Ng et al., 2014). Jung et al. (2013) use dscounted cash flow and mult-attrbute global nference of qualty methods to evaluate the performance of eco-ndustral park plot projects. Emergy analyss s proposed to conduct an assessment on Hefe economc and technologcal development area to quantfy the performance of ndustral symboss (Fan et al., 2016). In recent years, the method of evaluatng IS networks based on complex network theory has been developed gradually. Meanwhle, the theory system of symboss networks structural analyss could be constructed. It acheved characterstc analyss of IS networks. The structure and characterstc of IS networks are studed. A comprehensve methodologcal and analytcal framework s provded to understand the structural elements of IS networks (Domenech et al., 2011). For key enterprses of IS networks, key nodes are dentfed by analysng nternal structure of IS networks and enterprses relatonshp, on the bass of combned socal network analyss and complex network theory (Zhong et al., 2010). The network connectedness and related attrbutes of these hybrd ecologcal and IS systems are studed (Zhang et al., 2013). Accordng to calculaton of network topology parameters, the vulnerablty of ecologcal symboss networks s analysed quanttatvely, such as node betweenness and network effcency. Then mportant enterprses are dentfed by Please cte ths artcle as: Wang Z., Jang Y., Huang Y., Ja X., 2017, Complex network method towards evaluatng ndustral symboss, Chemcal Engneerng Transactons, 61, DOI: /CET

2 170 smulaton analyss of node falure (Xang et al., 2016). Takng Kalundborg IS water network as an example, mportant nodes are found based on network metrcs and complete dsrupton. Chopra et al.(2014) desgn strateges to develop reslent and sustanable of IS networks. Amed at detecton problem of IS networks, Huang et al. (2014) propose a structure detecton method based on the algorthm of communty detecton. Consderng the mportance of core enterprses n IS networks, an dentfcaton method by evaluatng node mportance based on TOPSIS s proposed n ths work. In addton, through adoptng the communty structure theory, we could acheve the communty structure detecton of complex IS networks and realze the quanttatve descrpton of IS networks. It s effectve to analyse the structure characterstc quanttvely and the connecton degree of each communty structure of IS networks. 2. Core enterprses dentfcaton of complex IS networks 2.1 Selecton of node mportance ndex Four quanttatve ndces are selected to descrbe the node mportance. Table 1: The formula and defnton of ndex Node mportance ndces Formula Defnton DC(degree centralty) DC=k DC s equal to the node degree. FBC(flow betweenness gst It s the proporton of paths passed node, among all FBC() centralty) st g nonredundant paths. st n CC(closeness centralty) It s the recprocal of average the shortest dstance CC n / dj j 1 between node and others. n EC(egenvector centralty) 1 EC Its value s egenvector of the greatest egenvalue of aj x j j 1 adjacency matrx. 2.2 Determnaton of ndex weght Utlzng the analytc herarchy process (AHP) to calculate the weght of each ndex. (1) Establsh the comparson matrx We take the three-demarcaton method (0,1,2) to analyse the mportance of four ndexes. DC manly reflects local nformaton of network. Compared wth others, ts mportance s the lowest. Due to the value of DC only relates to the number of adjacent nodes, and t gnores nodes whch connected ndrectly. FBC represents the ablty of node to control the entre network. CC reflects the extent of the nodes n network center. They all consder from overall perspectve of network. Therefore, we gve them same evaluaton. Not only does EC consder the node mportance from overall perspectve of network, but t reflects the combned nfluence of each node. So, EC s the most mportant ndex. The comparson matrx E s establshed and shown n Table 2. 2 ndex s more mportant than ndex j E e 1 ndex s same mportant wth ndex j, b e. 4 j j j 1 0 ndex j s more mportant than ndex Table 2: Comparson matrx E DC FBC CC EC b DC FBC CC EC (2) Construct the judge matrx by range method c DC CC FBC EC M W W DC 1 1/ 3 1/ 3 1/ 9 1/ 81 1/ C ( c ) / j FBC CC / EC Where B=max(b1,,b4)-mn(b1,,b4), ( b bj )/ B j b c c, cb=9, 4 j j 1 (1) (2) M c, W 4 M, 4 W W / ( W ). Fnally, the 1 consstency s checked. The weghtng of four ndexes s W

3 2.3 Comprehensve evaluaton of node mportance based on mult-attrbute decson method The mult-attrbute decson method based on TOPSIS can evaluate node mportance n network. Node mportance s evaluated by comparson of the dstance between every choce and the postve or negatve deal soluton (Yu et al., 2013). (1) There are n nodes to be evaluated, x={x1,x2,,xn}. The number of evaluaton ndexes s m, s={s1,s2,s3,,sm}. x(sj) represents the jth ndex of node. P=[x(sj)]nxm s the decson matrx. The dmenson of each ndex s unfed n order to compare convenently. By standardzng P, we get the matrx R=(rj)nxm, where rj= x(sj)/x(sj)max. (2) The weght of the jth ndex s wj(j=1,2,,m), wj=1. Then caculate the weghted normalzed decson matrx Y=(yj)nxm =(rwj)nxm. The postve deal soluton A + and the negatve deal soluton A are determned as. max max max mn mn mn A {max( y1, y 2,, ym)} { y1, y2,, ym } A {mn( y1, y 2,, ym)} { y1, y2,, y m } (3) L, L (3) Based on Eucldean dstance, the dstance from evaluaton value of each node to the postve deal soluton A + and the negatve deal soluton A are calculated. Eq(5) s the comprehensve evaluaton ndex. m max 2 ( j j ) j 1 D y y D Z D D m D ( yj y j ) j 1, 3. Communty structure detecton of complex IS networks mn Methods of the communty detecton The research about communty structure of complex network has a close relatonshp between graph partton n computer scence and herarchcal clusterng n socology. The graph partton manly ncludes Kernghan- Ln algorthm and spectral bsecton method. Herarchcal clusterng manly ncludes GN (Grvan-Newman) algorthm. Combned wth the method of complex network communty detecton, we choose spectral bsecton method to acheve communty structure detecton of IS networks, whch has obvous communty structure. 3.2 Steps of the communty detecton In vew of general characterstcs of IS networks structure, Capocc spectral bsecton method based on the normal matrx s adopted to fnsh communty structure detecton (Capocc et al., 2005). Steps are as follows: (1) Constructng normal matrx N=K -1 A, where A s adjacency matrx, whose element aj s equal to 1 f ponts to j and 0 otherwse. K s the dagonal matrx wth elements k= aj. (2) Calculatng the egenvalue and egenvector of the matrx N, there s always an egenvalue equal to 1. (3) The frst non-trval egenvalues (closes to 1) to correspondng egenvectors drawng s roughly ladder-lke dstrbuton. The number of ladder seres matches wth the number of communty. Consderng structure characterstcs of communty n IS networks, we evaluate communty detecton method based on concepts of coheson coeffcent Coh(C) and modularty Q. Coh(C) and Q are lnear ftted by the form of y=a+bx. Accordng to the value of correlaton coeffcent r, the result of communty detecton can be evaluated. If r 0.75, there are better performance of ndependence among communtes, closely connected nteror nodes n communty and clear communty structure. On the contrary, the detecton result s not good. n lc dc 2 Q ( ) 1 L 2L E( C) Coh( C) E( C) A( C, Cj) j Where lc s the number of lnks nsde module C. L s the total number of lnks n network. dc s the total degree of nodes n module C. Where E(C) s the total number of lnks of communty C. n s the number of nodes n communty C, and A( C, Cj) s the total number of lnks between communty C and Cj. 4. Case study j Austran Styra eco-ndustry park s a typcal IS network, whch has 36 enterprses, such as cement mll, power plant and ceramcs factory. There are frequent materal exchange and energy flow among enterprses. Its materal flow dagram s shown n Fgure 1. Each enterprse s regarded as a node, then materal and energy exchange among nodes as the edge of network. Styra IS network has characterstcs of complex network based on measurement of complexty. Thus, we can use the node mportance evaluaton method to dentfy 171 (4) (5) (6) (7)

4 172 core enterprses. On the other hand, t can be detected communty based on communty partton algorthm of complex network. To mprove accuracy of result and calculate convenently, Styra IS network s regarded as an undrected and unweghted network. The network of Styra IS park s shown n Fgure 2. Waste paper pulp Pulp Pulp mll Paper mll 3 Board and box lnng mll Fber pulp Ceramcs factory 2 Waste ol plant 1 Waste ol plant 2 Cement mll 2 Crushed aggregates collecton slt Sludge Wastewater treatment plant Fber pulp Fly ash Mnng company Votsberg Power plant 1 Bottom ash Lvng garbage Buldng rush 1 Power plant 2 Fltter Cement mll 4 Waste ol plant 3 Cement mll 6 Buldng rush 2 Cement mll 3 Heatng Graz Waste paper Cement mll 5 Slt Paper mll 2 Paper mll 4 Paper mll 5 Plastc packagng Plastc plant Paper mll 1 Grndng metal Grndng metal Steel mll Smeltng furnace sand Cement mll 1 Grstmll Textle mll 1 Textle mll 2 Waste textles Broken sawdust Asphalt Ceramcs factory 1 Packng-case Flax waste Produce Waste collecton Chemcal plant Fgure 1: Materal flow dagram of Austran Styra eco-ndustry park Fgure 2: The network of Austran Styra eco-ndustry park 4.1 Identfcaton of core enterprses DC, FBC, CC, EC and Z of each node of Styra IS network are shown n Table 3. Accordng to the numercal magntude of Z, node 1, node 2, node 10, node 9 and node 16 are key nodes. In other words, core enterprses contan cement mll 2, power plant 2, cement mll 1, steel mll and paper mll 2. So, we need to focus on the operaton of these enterprses. Compared wth the dentfcaton methods by Zhong et al. (2010), ths method can reduce the complexty of computaton and enhance the search effcency. In addton, the result s more relable due to usng the comprehensve ndcator. 4.2 Communty detecton Through the analyss and emprcal study of dfferent communty partton algorthm, the mproved spectrum splt method s more sutable for communty detecton of IS networks. At frst, wth the help of Matlab7.0, we get the frst non-trval egenvalue (selectng the maxmum) from normal matrx N and the frst non-trval egenvectors correspondng wth each node. They are shown n Table 4. Meanwhle, the frst non-trval egenvectors are sorted by sze and plotted by Orgn 8.0. Thus, we see that the dots present sx ladder-lke dstrbutons n Fgure 3a. Styra IS network s dvded nto sx communtes, ncludng CA, CB, CC, CD, CE and CF. The detecton of nodes n each communty s as follows. CA:3,4,20,21,22,24,33,34,35,36; CB:5,12,13,23; CC:6,15,31,32; CD:16; CE:1,17,18,19; CF:2,7,8,9,10,11,14,25; 26,27,28,29,30. Accordng to the dentfcaton of core enterprse as mentoned above, we fnd node 1 (cement mll 2) s located n CE. It s the key node of CE and lnk three communtes ncludng CD, CE and CF. Node 2 (power plant 2), node 10 (cement mll 1) and node 9 (steel mll) are located n CF. Among them, power plant 2 supples some

5 enterprses for power. CD only contans node 16 (paper mll 2). Node 16 not only lnks CC, CD and CE, but also plays an mportant role n keepng network connectvty. Table 3: The value of centrcty ndex and Z. Node DC FBC CC EC Z Node DC FBC CC EC Z Table 4: The frst non-trval egenvalue and egenvector of the matrx N Egenvalue Egenvector , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , The value of the frst non-trval egenvector The seral number of nodes (a) The value of Q The value of Coh(C ) (b) Fgure 3: a) s the dstrbuton of egenvalues and b) s the fttng curve of Q and Coh(C) 4.3 Result evaluaton of the communty detecton The modularty Q and coheson coeffcent Coh(C) of each communty of Styra IS network are calculated. Table 5 lsts the value of Q and Coh(C). The fttng curve of Q and Coh(C) s shown n Fgure 3b by usng Orgn 8.0. As show n Table 5, the correlaton coeffcent r s equal to It suggests that the evaluated method of communty partton s feasble. The result of Q and Coh(C) curve fttng reflects that the degree of ndependence among dfferent communtes. The correlaton coeffcent r=0.94>0.75, t ndcates each communty has a stronger ndependence n Styra IS network. Thus, Austran Styra eco-ndustry park has a hgher degree of enterprse cluster. The supply and demand can acheve a balance n ths IS network. Addtonally, all communtes ether connected each other or beng ndependent. It can mprove agglomeraton effect between ndustral clusters, acheve crcular economy, and mprove compettveness of enterprses.

6 174 Table 5: The fttng result of Q and Coh(C) of Austran Styra eco-ndustry park IS network Methods of communty detecton Result of partton Coh(C) Q Ftted value r CA Austran CB Styra Capocc spectral bsecton method CC eco-ndustry CD 0 0 park CE CF Concluson Case analyss shows that we can fnd core enterprses based on the node mportance evaluaton method of complex network. Compared wth other methods, we take nto account multple ndces. Thus, ths method s more comprehensve, and result of dentfcaton s more accurate. Core enterprses play an mportant role n the entre IS networks. In order to ensure the effcency of system, we must pay more attenton on core enterprses. Furthermore, IS networks can be dvded nto communtes wth the communty structure theory of complex network, so as to realze quanttatve detecton of IS networks and quanttatve analyss of ts structure features. Then through lnear fttng of communty coheson coeffcent and modularty, we can evaluate quanttatvely the parttonng result accordng to fttng coeffcent. From the case of Austran Styra eco-ndustry park, we can conclude that the evaluaton method presented n ths study can evaluate quanttatvely the core enterprses dentfcaton and communtes dvson n ndustral symboss network. References Capocc A., Servedo V.D.P., Caldarell G., Colaor F., Detectng Communtes n Large Networks, Physca A Statstcal Mechancs & Its Applcatons, 352, Chopra S.S., Khanna V., Understandng Reslence n Industral Symboss Networks: Insghts from Network Analyss, Journal of Envronmental Management, 141, Domenech T., Daves M., Structure and Morphology of Industral Symboss Networks: The Case of Kalundborg, Proceda - Socal and Behavoral Scences, 10, Fan Y., Qao Q., Fang L., Yao Y., Emergy Analyss on Industral Symboss of an Industral Park a Case Study of Hefe Economc and Technologcal Development Area, Journal of Cleaner Producton, 141, Huang Y., Wang Z., Ja X., Wang F., The Communty Structure Dvson of Complex Industral Symboss Network, Computers & Appled Chemstry, 5, Jung S., Dodbba G., Song H.C., Fujta T., A Novel Approach for Evaluatng the Performance of Eco- Industral Park Plot Projects, Journal of Cleaner Producton, 39, Ng R.T.L., Wan Y.K., Ng D.K.S., Tan R.R., Stablty Analyss of Symbotc Boenergy Parks, Chemcal Engneerng Transactons, 39, Ochoa P., Wastewater Stablsaton Ponds System: Global Senstvty Analyss on Network Desgn, Chemcal Engneerng Transactons, 50, Porkka P.L., Mäknen, E.P., Hannu, V., Safety Culture Research n a Fnnsh Large-Scale Industral Park, Chemcal Engneerng Transactons, 31, Raafat T., Trokanas N., Cecelja F., Bm X., An Ontologcal Approach Towards Enablng Processng Technologes Partcpaton n Industral Symboss, Computers & Chemcal Engneerng, 59, Xang P., L B., Hu M., L J., Dong L., Study on Vulnerablty of Symboss Network n Eco-Industral Parks, Journal of engneerng studes, 8, Yong J.Y., Nemet A., Varbanov P.S., Klemeš J.J., Čuček L., Kravanja Z., Mantell V., Heat Exchanger Network Modfcaton for Waste Heat Utlsaton under Varyng Feed Condtons, Chemcal Engneerng Transactons, 43, Yu H., Lu Z., L Y.J., Key Nodes n Complex Networks Identfed by Mult-Attrbute Decson-Makng Method, Acta Physca Snca, 62, Zhang H, Chen Z, Yang N J, Socal Network Analyss and Network Connectedness Analyss for Industral Symbotc Systems: Model Development and Case Study, Fronters of Earth Scence, 7, Zhong G., Cao J., Cao L., Wang S., Quanttatve Analyss of Industral Symboss Network Structure, Academc annual conference of Chnese socety of envronment scence, Vol 2, Shangha, Chna, 9.