NEW RISK ANALYSIS METHOD to EVALUATE BCP of SUPPLY CHAIN DEPENDENT ENTERPRISE

Transcription

1 The 14 th Word Conference on Earthquake Engineering NEW RISK ANALYSIS ETHOD to EVALUATE BCP of SUPPLY CHAIN DEPENDENT ENTERPRISE Satoru Nishikawa 1, Sei ichiro Fukushima 2 and Harumi Yashiro 3 ABSTRACT : 1 Director, inistry of Land, Infrastructure, Transport and Tourism, Tokyo, Japan 2 Senior Researcher, Tokyo Eectric Power Services Co., Ltd., Tokyo, Japan 3 anager, Tokio arine & Nichido Risk Consuting Co., Ltd., Tokyo, Japan Emai: nishikawa-s2iv@mit.go.jp Business continuity is now being regarded as an important factor in management of enterprises. ost of the enterprises are not stand-aone; they depend on Suppy (SC), for their operations. In order to draw BCP (Business Continuity Pan) for suppy chain dependent enterprise, there is a need for a quantitative risk anaysis method to evauate such manufacturing system. The business interruption time is a crucia factor in terms of economic oss. A mode enterprise with severa factories connected through suppy chain was appied in anayzing the risk. Severa ocation types are considered. The suppy chain connection, parae, series and mixed are considered. The damage probabiity of factories and their business interruption time are cacuated based on the seismic risk of different ocations in Japan. The annua exceedance probabiity of business interruption time of these ocation and suppy chain types are cacuated and compared. Through combining these factors a reaistic risk anaysis method for evauating business interruption is proposed and is proven to be functiona. This method may be appied for strategic ocation panning of suppy dependent enterprises who wish to minimize their risk. KEYWORDS: business continuity, suppy chain, seismic risk, business interruption, economic oss, risk curve 1. INTRODUCTION ost of the modern enterprises are not stand-aone; they depend on SC, for their operations. In many of the competitive manufacturing industries, the management, for the sake of decreasing the cost of stockpiing and aso to respond quicky to changing market demands, is shifting to the Just in Time procurement and manufacturing system. This system is based on the assumption that the SC to their factory functions without interruption. Therefore, once the SC is interrupted, they face the immediate risk of hat of operations. On the other hand, business continuity is now being regarded as an important factor in risk management of enterprises. In the case of the Niigata-Chuetsu-Oki Earthquake on 16 Juy 2007 in Japan, an automobie engine piston ring manufacturer suffered serious damage. Since this manufacturer had a dominant share, most of the major automobie companies in Japan had to hat their assemby ines for one week, due to the fact that they did not have extra stock of piston rings in their hands. Since it was difficut to identify an aternate suppier for the sophisticated piston rings in a short period, the automobie companies sent their own engineers to this piston ring manufacturer to assist the quick restoration of the piston ring manufacturing ines. This is a typica exampe of a oca disaster affecting production in other ocations since they are SC connected. This incident demonstrated the need to consider the suppy chain as an important factor in drawing BCP in such enterprises. Simiar cases were seen in the October 2004 Niigata-Chuetsu Earthquake and in the 1995 Hanshin-Awaji Earthquake, where the direct damage to the factories hit by the earthquake spread to the enterprises ocated outside the affected area through the hat of the suppy chain of parts and materias.

2 The 14 th Word Conference on Earthquake Engineering These cases ca for the need to deveop an earthquake risk evauation method for business continuity panning of SC dependent enterprise. For such evauation, mutipe ocations of the nodes of the SC system and the individua risks of the buidings on these ocations need to be combined and considered. Not ony the direct materia damage by the earthquake to the nodes, but moreover the (BIT) of the enterprise is a crucia factor for the surviva of the enterprise in the competitive market. The BIT is a crucia factor in terms of economic oss. If an enterprise hats its suppy of products over a consumer acceptabe timeframe, the consumer wi be quicky taken over by riva enterprises and if the business resumption time exceeds a certain imit, the enterprise wi no onger be abe to return to the previous market. Therefore a method to quantify the BIT of the SC dependent system and to evauate options to decrease the BIT is needed 1). This study, proposes a quantitative risk evauation method of BIT of SC dependent enterprise, and appies to mode enterprises with severa factories connected through SC. Three ocation types are considered, factories concentrated in Tokyo, factories ocated in the Kanto pain and factories dispersed in Eastern Japan. The suppy chain connection, parae, series and mixed are considered. The damage probabiity of factories and their BIT are cacuated based on the seismic risk of different ocations. The annua exceedance probabiity (AEP) of BIT of these ocation and suppy chain types are cacuated and compared. Through combining these factors a reaistic risk anaysis method for evauating business interruption is proposed and is proven to be functiona. This method may be appied for strategic ocation panning of suppy dependent enterprises who wish to minimize their risk. Aso aternative options to decrease the BIT are proposed. 2. BUSINESS INTERRUPTION TIE AS THE CRITICAL FACTOR IN BCP There are two main objectives for enterprises to draw BCP 2) (Fig.2.1). First, to avoid the tota hat of operations, even in case of disasters and maintain the minima eve of operations for business continuity. Second, to resume the operations to pre-disaster eve within an acceptabe timeframe from the viewpoint of corporate management in the competitive market. Leve of Operations Recovery Curve Now without BCP Desirabe Risk Recovery Curve Curve after with BCP Resume operations DISASTER within acceptabe time Before After 100% Recovery Target inimum Timeframe Target Shorten the Gap of Business aintain the minima Gap: the Target vs. Status Quo Interruption Time between the Status Leve of Operations Quo and the BCP Appied Case Figure taken from Cabinet Office BCP Guideine ver.1 Fig.2.1 The Overa Concept of Business Continuity Panning Let us suppose that a factory was hit by an earthquake. For the user of the product of the factory, his biggest interest woud be when the factory woud resume suppy to him. If he sees that he cannot expect the suppy within the period he can wait, he wi search for an aternative product or manufacturer. If the user decides that he cannot wait anymore, the factory wi ose the user. Even if the factory resumes production afterwards, since the previous user is aready taken by other manufacturer, the enterprise who runs the factory wi have to pay marketing efforts to regain the user or to find new customers, which wi require additiona marketing costs. The onger the BIT, the marketing cost necessary wi augment. If the BIT of the factory is further proonged, the enterprise may no onger

3 The 14 th Word Conference on Earthquake Engineering be abe to return to the market. Thus BIT is an important factor for business continuity of an enterprise. 2. STRUCTURE OF THE RISK EVALUATION ETHOD OF BIT There has been previous proposa and study on a method to evauate earthquake damage risk of an enterprise who possesses a portfoio of mutipe buiding assets in various ocations 3). This method, by generating numerous possibe scenario earthquakes (hereinafter earthquake events), estimates possibe damage to the individua buidings and by summation of these damages cacuates the damage risk of the portfoio. By expanding this methodoogy, 1) BIT of individua nodes of a SC is cacuated according to each earthquake event and 2) by combining the resuts of BIT of individua nodes according to the SC patterns, thus generating event trees, the BIT of the SC dependent enterprise wi be obtained. 2.1 of Individua Node BIT of individua node j by earthquake event i sha be represented as t j (i) and is defined by Eqn.2.1 as foows. 4 t = [ p, t, ] (2.1) j k = 0 k is the variabe representing eves of damage as foows, k=0: no damage, k=1: sight damage, k=2: moderate damage, k=3: severe damage, k=4: coapse. t j, k represents BIT of node j by event i according to eve of damage k. p k j, represents conditiona probabiity of t, k j, is drawn from the fragiity curve of individua node j. p k j k j. j k 2.2 of Suppy BIT of SC by earthquake event i sha be represented as t SC (i) and is defined by Eqn.2.2 as foows. t SC N = f [ p = 1 ] g[ t is the variabe representing individua end-branches of event trees. N is the tota number of end-branches. t (i) represents the vector composed from t j, k for end-branch case. p (i) represents the vector of conditiona probabiity corresponding to t (i). f [ ] represents the conditiona probabiity of occurrence of end-branch case and is defined by Eqn.2.3 as a function of p (i). ] (2.2) f [ ] = p1, k1 p2, k 2 L pn, kn p (2.3) n represent the number of nodes composing the SC. g [ ] represents the BIT of SC for each end-branch and is a function of t (i). This function is determined by the connection pattern of the nodes of the SC. For a series SC the maximum, for a parae SC the minimum, is chosen. Fig.2.2 shows the outine of evauation of BIT for each earthquake event. Fig.2.3 shows the procedure of the proposed BIT risk evauation method. In order to consider the uncertainties, a onte-caro simuation with

4 The 14 th Word Conference on Earthquake Engineering ground motion strength and buiding strength as stochastic variabes is appied. BIT for buiding 1 Eqn. (2.1) Expected BIT of SC Eqn. (2.2) BIT for buiding j Eqn. (2.1) Start Generation of Earthquake Events Cacuation of Conditiona Faiure Probabiities of Nodes Identification of BIT of Nodes by Damage Leves buiding 1 buiding j Evauation of BIT of Individua Nodes Evauation of BIT of Suppy A Events N Y event i Seismic zone Evauation of Risk Curve End Fig.2.2 Outine of Evauation Process of BIT of SC Fig.2.3 Fowchart of Risk Anaysis of Suppy 3. APPLICATION TO ODEL SUPPLY CHAIN 3.1 Setting of ode Suppy Five nodes(factories) ocated in three ocation types, A: factories concentrated in Tokyo (Fig.3.1), B: factories ocated in the Kanto pain (Fig3.2) and C: factories dispersed in eastern Japan (Fig.3.3) were set as exampes. Based on the previous findings in Ref.4) and Ref.5), the seismic strength of each nodes and BIT by damage eve were set as shown in Tabe 3.1. Fig.3.4 shows the three connection patterns of the five nodes, series, parae and mixed, as mode SCs. 3.2 Setting of Seismic Activity ode Seismic activity zone mode was set according to Ref.6). Fig.3.5 shows the ocation of seismic sources and the seismic parameters are shown in Tabe 3.2. With reference to Annaka & Yashiro 7) attenuation reation of ground motion was cacuated foowing the Annaka ode as shown in Eqn.3.1. og a = h 2.203og d (3.1) 2 2 d = Δ h exp(0.699 ) a stands for maximum ground acceeration, Δ for distance from the epicenter, h for depth of seismic source, for magnitude of the earthquake event. Logarithmic standard deviation representing the variabiity of the range attenuation was set at natura ogarithm 0.5.

5 The 14 th Word Conference on Earthquake Engineering Tabe 3.1 Fragiity Curve and BIT by Leve of Destruction A-4 A-5 A-1 A-2 Damage Leve Characteristic Vaues of Fragiity edian (cm/s/s) Log-norma Std. Deviation BIT (day) Sight oderate Severe Coapse A-3 #-1 #-2 #-3 #-4 # N Series System 139.5E 140.0E Fig.3.1 Location A: Tokyo Concentration #-1 #-2 #-3 #-4 # N B-5 Parae System #-2 #-1 #-4 #-5 B-4 B-1 B-2 #-3 ixed System Fig.3.4 Suppy Pattern odes 35.5N B-3 Tabe 3.2 Parameters of Seismic s 139.5E 140.0E 140.5E Fig.3.2 Location B: Kanto Pain C N C N C-4 C N C E 138.0E 139.0E 140.0E 141.0E Fig.3.3 Location C: Eastern Japan Return Period (years) Return Period (years) Return Period (years) th th th th th th th th kt kt kt kt kt kt kt kt kt kt kt kt cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb cb kk kk kk kk kk kk e e j j j s s s n n A Vaue in G-R Equation A Vaue in G-R Equation A Vaue in G-R Equation c c c c c c c c 注 :b-vaue in G-R equation for sources c08, c09, c10, c11, c12, c14, c15 and c18 is set at 0.9

6 The 14 th Word Conference on Earthquake Engineering 38.0N e4 s of Large Earthquakes e5 cb01 th11 th13 th14 th12 th16 th07 j5 38.0N c N 35.0N cb18 cb25 cb10 cb19 cb05 cb24 cb13 cb26 cb20 cb12 cb29 cb16 cb27 cb14 cb07 cb28 cb30 cb21 kk05 cb31 cb17 cb23 cb22 kk02 cb15 kk01 cb35 kk03 cb34 kk06 n1 kk04 n2 cb02 cb03 cb04 cb11 cb09 kt02 kt03 kt04 kt05 kt09 th17 kt01 kt11 cb33 kt10 kt07 cb32 kt12 s3 s1 kt08 kt06 s2 th15 j6 j7 37.0N 35.0N c15 c14 c12 c10 c11 c09 c08 s ofedium&sma EQs 137.0E 138.0E 139.0E 140.0E 141.0E 4. BIT RISK CURVE OF ODEL SUPPLY CHAIN Fig. 3.5 Location of Seismic s 137.0E 138.0E 139.0E 140.0E 141.0E Risk Curve of BIT for the Series ode SC is shown in Fig.4.1. Risk Curve of BIT for the Parae ode SC is shown in Fig.4.2. Risk Curve of BIT for the ixed ode SC is shown in Fig.4.3. In these figures, the horizonta axis shows the 90 percentie exceedance of, the vertica axis shows the AEP. The bod ine shows the risk curve of the SC, the thin ines show the risk curve of the different nodes. In Series SC, if one or more of the five nodes suffers damage, the SC hats. Therefore, as shown in Fig.4.1, the risk curve of the SC is arger than any of the individua risk curve of the different nodes in a ocations A, B and C. When we compare the risk curves of the three ocation patterns, C<A<B. Risk curve of ocation in Kanto pain B is arger than concentration in Tokyo A, since the possibiity of one of the nodes in Kanto pain being hit by an earthquake is reativey high, due to the fact that Kanto pain has many seismic sources distributed. Risk curve of distribution in eastern Japan C is smaer than others since 4 out of the 5 nodes are ocated outside the Kanto pain. In Parae SC, the SC hats ony when a five nodes suffer damage at the same earthquake event. Therefore, as shown in Fig.4.2, the risk curve of the SC is smaer than any of the individua risk curve of the different nodes in a ocations A, B and C. When we compare the risk curves of the three ocation patterns, C<B<A. The more the nodes are geographicay dispersed, the parae SC has greater redundancy and the hat risk of SC becomes smaer. In ocation pattern C the expected BIT of SC is zero, since the possibiity of a five nodes dispersed in eastern Japan being hit by a same earthquake event is zero. In ixed SC, if one of the nodes out of #1, #4 and #5 suffer damage, the SC hats. If both nodes #2 and #3 suffer damage at the same earthquake event, the SC hats. When we compare the risk curves of the three ocation patterns, C<B<A. Since nodes #2 and #3 are aternates, the more the two are geographicay dispersed; the hat risk of SC becomes smaer. When we compare ocation A and B, the possibiity of nodes B-2 and B-3 suffering damage at the same earthquake event is smaer than the possibiity of nodes A-2 and A-3 suffering damage at the same earthquake event. Therefore, athough ocation B has nodes in Kanto pain where many seismic sources are distributed, the risk curve of SC is smaer than that of ocation A. This is different from the Series SC and this indicates that provision of aternate node contributes to decreasing the risk of a series SC. When we compare the risk curve of the Series SC and ixed SC for ocation B, in Series SC the risk curve of SC is greater than the risk curve of node B-3, by providing node B-2 as aternative and making ixed SC, the risk curve of ixed SC is much smaer than the risk curve of node B-3. Node B-3 is Yokohama which has the highest possibiity, among a the nodes in this exampe, of being hit by earthquakes and may become the botteneck in the Series SC. This indicates that in order to decrease the risk of a series SC, providing aternative node to the most vunerabe node and ocating it away from the origina node woud be a good soution.

7 The 14 th Word Conference on Earthquake Engineering Node 1-1A-1 Node 1-2A-2 Node 1-3A-3 Node 1-4A-4 Node 1-5A-5 Location A:Tokyo Concentration Node 1-1A-1 Node 1-2A-2 Node 1-3A-3 Node 1-4A-4 Node 1-5A-5 Location A:Tokyo Concentration Node 1-1A-1 Node 1-2A-2 Node 1-3A-3 Node 1-4A-4 Node 1-5A-5 Fig. 4.1 BIT Risk Curve of Series SC Fig. 4.2 BIT Risk Curve of Series SC Node 2-1B-1 Node 2-2B-2 Node 2-3B-3 Node 2-4B-4 Node 2-5B-5 Location B:Kanto Pain Node 2-1B-1 Node 2-2B-2 Node 2-3B-3 Node 2-4B-4 Node 2-5B-5 Location B:Kanto Pain Node 2-1B-1 Node 2-2B-2 Node 2-3B-3 Node 2-4B-4 Node 2-5B-5 Location C: Eastern Japan Node 3-1C-1 Node 3-2C-2 Node 3-3C-3 Node 3-4C-4 Node 3-5C-5 Location A:Tokyo Concentration Location B:Kanto Pain Location C: Eastern Japan Fig. 4.3 BIT Risk Curve of ixed SC Node 3-1C-1 Node 3-2C-2 Node 3-3C-3 Node 3-4C-4 Node 3-5C-5 Location C: Eastern Japan Node 3-1C-1 Node 3-2C-2 Node 3-3C-3 Node 3-4C-4 Node 3-5C-5 5. ANALYSIS OF THE RELATION BETWEEN THE LOCATION OF NODES AND BIT RISK CURVE In addition to the ocation of the nodes (incuding the possibiity of mutipe nodes hit by the same earthquake) and the suppy chain patterns, the earthquake environment of the ocation of the individua nodes is another factor of the BIT risk curve. In order to excude the third factor and examine the infuence of the former two factors to the BIT risk curve, comparison of the sum of the BIT of individua nodes and the BIT of SC is shown in Fig.5.1. In Series SC, as the nodes are geographicay dispersed, the BIT of SC becomes coser to the sum, and in Location C the two curves overap. In Location C, the nodes in SC are infuenced by individua earthquakes, and the SC hats if one of the nodes stops. In Location A, since there are cases which an earthquake affects mutipe nodes, the BIT of SC is smaer than the sum. In Parae SC, in a three ocation types, BIT of SC is far smaer than the sum, and as the nodes are geographicay dispersed, BIT of SC nears to zero. The ixed SC risk curve, in the domain where the AEP is arger than, neary overaps the Series SC, whereas, in the domain where AEP is smaer than, shows intermediary nature of the Series SC and Parae SC. Fig.5.2 shows the ratio of BIT of SC against the sum of the BIT of individua nodes.

8 The 14 th Word Conference on Earthquake Engineering Sum of BIT of of BITs Nodes Parae S.C.(2) SC ixed S.C.(3) SC Sum of BIT of of BITs Nodes Parae S.C.(2) SC ixed S.C.(3) SC Location A:Tokyo Concentration Location B:Kanto Pain Location C: Eastern Japan Fig. 5.1 Comparison of BIT of SC and the Sum of BIT of Individua Nodes Parae S.C.(2) EC ixed S.C.(3) SC Series S.C.(1) SC S.C.(2) Parae S.C.(3) EC Sum of BIT of of BITs Nodes Parae S.C.(2) SC ixed S.C.(3) SC Parae S.C.(2) EC ixed S.C.(3) SC BIT of SC/Sum of BIT of Nodes BIT (S.C.) / Sum of BITs BIT of SC/Sum of BIT of Nodes BIT (S.C.) / Sum of BITs BIT of SC/Sum of BIT of Nodes BIT (S.C.) / Sum of BITs Location A:Tokyo Concentration Location B:Kanto Pain Location C: Eastern Japan Fig.5.2 The Ratio of BIT of SC against the Sum of the BIT of Individua Nodes 6. CONCLUSION The proposed method enabes the BIT risk curve anaysis of SCs composed of mutipe nodes in different geographica ocations. BIT of different types of SCs, Series, Parae and ixed, can be quantitativey anayzed and be compared according to the various ocations of the nodes. Therefore the enterprises dependant on SCs for their operations can quantitativey compare the earthquakes risks from one ocation pattern to another. They can aso identify the most critica node in the Series SC, and evauate the reduction of the risk by providing additiona aternative node to that node, and furthermore evauate the risk reduction according to the ocation of this aternative node. Three ocation patterns were appied; however this method may be appied to other ocations provided that the seismic source data is avaiabe. Aso the SC patterns can be expanded to various combinations. REFERENCES 1) Nishikawa, S., Fukushima, S. and Yashiro H. (2007). "Proposa of a new ethod for Risk Evauation of Suppy- Dependent Business Continuation Pan, Proceedings of the Annua Conference of the Institute of Socia Safety Science, Vo.21, pp ) The Centra Disaster anagement Counci Specia Committee (2005). Business Continuity Panning Guideine First Version, Cabinet Office, Government of Japan. 3) Fukushima, S. and Yashiro H. (2002). Seismic Risk Anaysis on Portfoio of Buidings, Journa of Architecture, Panning and Environmenta Engineering, No.552, pp ) Sakamoto, S. (2006). Evauation of Recovery Curve for Business Continuity Panning, Report of Taisei Technoogy Center, No.39, pp ) Takahashi, I., Hayashi, T., Enomoto, T., iyamura,. and Fukushima, S. (2007). "A Consideration of Earthquake Damage ode atrix for Business Continuity Pan, Proceedings of the Annua Conference of the Institute of Socia Safety Science, Vo.21, pp ) Architectura Institute of Japan (2004). Recommendations for Loads on Buidings, Architectura Institute of Japan 7) Annaka, T. and Yashiro, H. (1998). A seismic source mode with tempora dependencies of arge earthquake occurrence for probabiistic seismic hazard anaysis in Japan, Risk Anaysis, WIT PRESS, pp ,