CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY Y. T. Chen Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong yongtong.chen@connect.polyu.hk F. T. S. Chan* Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong f.chan@polyu.edu.hk S. H. Chung Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong nick.sh.chung@polyu.edu.hk B.Niu Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong ben.niu@polyu.edu.hk *The corresponding author: F. T. S. Chan ABSTRACT Environental issues caused by iproper abandoned cartridges increase obviously nowadays. Producers are perceived to be responsible for recycling products they have produced. In Hong Kong, due to abundant quantity of used cartridges, producers have to optiize their forward and reverse networks to axiize the recycling rate and their profits. In this paper, a coprehensive Closed-Loop Supply Chain (CLSC) odel is established. This odel contains eight partners in CLSC and describes the cartridge recycling situation in Hong Kong. In the literatures, any CLSC odels were established and studied, but few of the analyzed the delivery activity for different kinds of aterials extracted fro used products, and also few papers studied the situation that used products are classified. In this odel, delivery activities of different aterials are considered and the used cartridges are classified into good MLB 107
quality ones and poor quality ones. Producers will have different ethods to process the. Furtherore, this odel structures a siplified VRP to express the real situation that collection points onsite pick up used cartridges fro custoers. This proble is forulated into a linear prograing odel. Since both delivery routes and delivery quantities all over the CLSC network have to be optiized, the proble, which is NP-hard, becoes coplex to calculate. To deal with it, a odified two-stage Genetic Algorith (GA) is ipleented. The two-stage encoding in this GA reinforces the genetic searching ability in solving this kind of proble. This algorith optiizes the CLSC network and the results show a near optial solution which provide producers the axiu profits. Keyword: Closed-loop supply chain, Genetic algorith, E-waste recycling, Reverse distribution INTRODUCTION In recent years, econoy with electronic products gains a prosperity developent, and the life cycle of electronic devices turn out to be shorter and shorter, aking electronic waste the fastest growing part aong garbage strea. The environental and health risks of Waste Electric and Electronic Equipent (WEEE) becoe high, aong which, used cartridges play an iportant role. Most of the used cartridges are iproperly disposed to landfill or incineration. A toner cartridge tossed into landfill will take 450 years or ore to decopose and the toxic aterials in it such as lead and ercury will leak out to cause a great daage to the earth. Although a recycled toner cartridge can save nearly 1 kg of raw aterials like plastic and etal to reduce the environent burden efficiently, it turned out that used cartridges at the end of their lifespan still been disposed iproperly and the quantity of the has grown exponentially. Every year, about 1.2 trillion inket cartridges are used globally, but less than 30% of the being recycled. This terrible situation has been last for any years until the andatory legislation of Extended Producer Responsibility (EPR) gain ore and ore popular around the world. Producers are perceived to be responsible for the recycling of products they have produced and sold. Due to the expensive cost of third party recycling copanies, any cartridge producers choose to establish their own recycling factories, and producers also have to design product recycling networks along with product delivery networks according to specific custoer regions. In Hong Kong, any copanies start recycling of ink cartridge and toner cartridge several years ago, for exaple, Epson HK set up soe collection points and collected used cartridges fro 2007, Canon HK launched the recycling progra of ink cartridge fro 2009. Because of the abundant quantity of used products and the pressure of MLB 108
environent, producers have to optiize their forward and reverse networks to axiize the recycling rate and also axiize their profits. In recent years, any studies address the issue of WEEE recycling. Tsai and Hung (2009) focused on the treatent and recycling process of the syste, they proposed a two-stage decision fraework which include treatent stage and recycling stage. Although suppliers selection was added in this fraework, it is not the optiization of the whole Closed-Loop Supply Chain (CLSC) network. Veenstra et al. (2010) suggested a Markov chain odel analyzing the flow of WEEE through the reverse chain. Gaberini et al. (2010) established a transportation network in Italy which contained vehicle routing proble. An integrated solution approach was used to solve it. Mar-Ortiz et al. (2011) optiized the design of reverse chain for the collection of WEEE. In this network, a ixed integer linear prograing was forulated to address the facility location proble, a new integer prograing was established to solve the vehicle routing proble and a siulation study is ipleented to assess the perforance of the recovery syste. Dwivedy and Mittal (2012) investigated into the WEEE flows in India and used a Markov chain to odel the business sector of WEEE trade, which including the inforal recycling of WEEE in developing countries. Aluur et al. (2012) proposed a ulti-period reverse logistics network which forulated into a ixed-integer linear prograing odel. A real case of washing achines in Gerany was ipleented to ustify this odel. Wang and Huang (2013) established a two-stage robust prograing odel to decide the recycling volue and tie in a CLSC. Fro the literature, it can be found that the focus of research about WEEE is the reverse network design and optiization, lacking the research of integrate both forward and reverse network. In this paper, a coprehensive CLSC odel based on the real situation of cartridge recycling in Hong Kong is established. It integrates both forward chain of product procedure and reverse chain of product recycling. This proposed odel contains eight partners in CLSC and describes the cartridge recycling situation in Hong Kong. In the literatures, any CLSC odels were established and studied. Olugu and Wong (2012) proposed a CLSC perforance evaluation syste to a copany fro the autootive industry which reduces the cost of the whole CLSC network proinently. Ozkir and Basligil (2012) addressed a ixed integer linear progra odel to describe a CLCS network considering three ways of recovery process. Ain and Zhang (2012) established a three-stage ulti-obective ixed-integer linear prograing odel designing the configuration and selection process of CLSC siultaneously. MLB 109
Although any CLSC odels were studied, few of the analyzed the delivery activity for different kinds of aterials extracted fro used products, and also few papers studied the situation that collected used products are classified. In this odel, delivery activities of several kinds of aterials are considered and the collected used cartridges are classified into two categories: good quality ones and poor quality ones. Producers will have different ethods to process the. Furtherore, this odel structures a siplified Vehicle Routing Proble (VRP) to express the real situation that collection points onsite pick up used cartridges fro custoers. This proble is forulated into a linear prograing odel. Since both delivery routes and delivery quantities all over the CLSC network have to be optiized, the proble, which is NP-hard, becoes coplex to calculate. To deal with this coplex calculation, a odified two-stage Genetic Algorith (GA) is ipleented. The two-stage encoding in this GA reinforces the genetic searching ability in solving this kind of proble. This algorith optiizes the CLSC network and the results show a near optial solution which provide producers the axiu profits. MODELLING This odel is based on the real case study of ink and toner cartridge delivering and recycling in Hong Kong. The proposed odel contains eight partners in the CLSC network: suppliers (S), anufacturers (M), warehouses (W), retailers (R), custoers (Cu), collection points (Co), recycling copanies (RC) and waste disposal plant (WDP). Figure 1 displays the whole CLSC network of this proposed odel. In this odel, the deands of custoers are preset. In order to fulfill the deand, anufacturers have two choices, one is to produce brand new products using raw aterials, and the other one is to reanufacture the collected used products with good condition/quality. As for the raw aterials, anufacturers have two acquisition channels: one is fro suppliers and the other is fro recycling copanies. In the proposed odel, it considers that anufacturers purchase several kinds of raw aterials fro suppliers and recycling copanies to produce brand new products, which eans ulti-products is considered. In this odel, it is assued that anufactures can ake sure that reanufactured products have the sae quality as brand new products, and also sale prices in the arket are sae. After producing new products and reanufacturing used products as brand new ones, anufacturers deliver the products oversea to their warehouses in Hong Kong. Then, fro warehouses, products are transported to retailers, and, finally, the retailers sell the to custoers. Since retailers not belong to anufacturers, this part of revenue and cost is excluded fro consideration in the CLSC network optiization. MLB 110
Therefore, the deand of each retailer also has to be preset, and the su of all the retailers deand ust be equal to the su of all the custoer deand. Suppliers Manufacturers Warehouses Retailers S1 S2 M1 W1 R1 R2 S3 t=1 M2 W2 R3 S4 M3 W3 R4 S5 S6 M4 20% 72% R5 t=2 Co1 Cu1 RC1 Co2 WDP 8% RC2 Co3 Cu2 Co4 Cu3 Waste disposal plant Recycling copanies Collection points Custoer regions Figure 1 the CLSC network in the proposed odel Custoers always discard the used ink and toner cartridge at the end of their lifecycle in the corner. In HK situation, collection points will pick used products up fro patron. In each custoer area, anagers of the cartridge copany have to consider the vehicle routing proble of one or ore collection points, which will ake the CLSC network too coplex to solve as whole. For siplicity, it is assued that collection points will have a round trip in each custoer areas to fetch used products. The transportation cost for a round trip is proportional to the quantity of used products collected in this round trip. The unit transportation cost for each used products are preset as a paraeter. No atter which ethod is used to collect, collection points will pay for the used products to the custoer. In this odel, it is assued that the price collection points paid for either good quality used products or poor quality used products are the sae. After collection, collection points deliver all the used products to recycling copanies. MLB 111
In the recycling copany, all the used products are cleaned and classified according to two categories: good quality used products and poor quality used products. The good quality ones will be packaged and transported to anufacturers. For the poor quality ones, recycling copany will disasseble and sash the and extract raw aterials through further process. During the whole process, ost of the substance can be recycled as raw aterials. The reaining parts need to be disposed by the waste disposal plant using ethod of burn or landfill. The axial disposal rate is preset in this odel. The recycled raw aterial will finally be delivered to anufacturers for producing new products. The indices, paraeters and decision variables are shown as below. Indices I t J K L V M N T the nuber of suppliers supplying aterial t with i=1,2,...,i the nuber of anufacturers with =1,2,...,J the nuber of warehouses with k=1,2,...,k the nuber of retailers with l=1,2,...,l the nuber of custoers with v=1,2,...,v the nuber of collection points with =1,2,...,M the nuber of recycling copanies with n=1,2,,n the nuber of raw aterials with t=1,2,,t Paraeters s c i capacity of supplier i c capacity of anufacturer w c k d l d v capacity of warehouse k deand of retailer l deand of custoer area v c capacity of collection point co s it unit cost of transportation of raw aterial t fro each supplier i to each anufacturer k unit cost of transportation fro each anufacturer to each warehouse k w kl unit cost of transportation fro warehouse k to retailer l cu v unit cost of round trip transportation for collection point taking back used products fro custoer area v co n unit cost of transportation fro collection point to recycling copany n r nt unit cost of transportation of raw aterial t fro recycling copany n to anufacturer r n0 unit cost of transportation of used product with good quality fro MLB 112
recycling copany n to anufacturer f fixed cost for operating anufacturer w f k fixed cost for operating warehouse k co f fixed cost for operating collection point r f n s x 0 x 1 x 2 x 3 p 1 p 2 t t v fixed cost for operating recycling copany n unit sorting cost for the used product unit producing cost of new products using raw aterials unit processing cost for the used product with good quality unit processing cost for the used product with poor quality unit dispose cost for the aterial which cannot be recycled the price of new products unit cost that collection point pay to custoers for the used product the percentage of good quality used products in all the recycling products the required quantity of raw aterial t to produce one new product the recycled percentage of aterial t in one used product the recycling rate of custoer area v the axial disposal rate unit weight per used product Decisions Variables s q it Aount of raw aterial t shipped fro supplier i to anufacturer q k Aount shipped fro anufacturer to warehouse k w q kl Aount shipped fro warehouse k to retailer l cu q v Aount shipped fro custoer v to collection point co q n Aount shipped fro collection point to recycling copany n r q nt Aount of raw aterial t shipped fro recycling copany n to anufacturer q Aount of used product with good quality shipped fro recycling r n0 r qnz copany n to anufacturer Aount of disposed aterials fro recycling copany n new q Aount of new produced products in anufacturer 1 if production takes place at anufacturer 0 otherwise 1 if warehouse k is opened k 0 otherwise 1 if collection point is opened 0 otherwise MLB 113
1 if recycling copany n is opened n 0 otherwise Obective function: ax TP TR TC (1) TR p1 dl (2) l TC TC TC TC 1 1 2 3 (3) TC s q + q + w q cu q s w cu it it k k kl kl v v t i k k l v co q + r q r q (4) co r r n n nt nt n0 n0 n t n n w co r TC2 f fk k f fn n (5) k n TC p q + s q x q + x ( q q )+ x q x q (6) cu cu r cu r r new 3 2 v v 1 n0 2 v n0 3 nz 0 v v n v n n The obective is to axiize the total profit which is the value of total revenue inus total cost as showed in the obective function (1). The total revenue is the revenue of sale the new product which is displayed in function (2). The total cost consists of the total transportation cost, total facility fixed cost and total processing cost as represented by function (3). Equation (4) shows the total transportation cost in the CLSC network, which consists of seven section costs in different stages shown as follow: raw aterial transportation cost fro suppliers to anufacturers, new product transportation cost fro anufacturers to warehouses, the cost of new products delivered fro warehouses to retailers, and the round trip cost of collection points to fetch the used products fro custoers, also the delivery cost for collected used products fro collection points to recycling copanies, the recycled raw aterial transportation cost fro recycling copanies to anufacturers and the delivery cost of collected good quality used products fro recycling copanies to anufacturers. Equation (5) shows the total fixed costs of the anufacturers, warehouses, collection points and recycling copanies. Equation (6) displays the su of used products obtained costs paid by collection points to custoers, used products sorting costs in recycling copanies, good quality used products processing costs in anufacturers, and poor quality used products processing costs in recycling copanies, disposed costs and the costs of newly produced products using raw aterials in anufacturers. Subect to s s q c, i (7) i i MLB 114
k n l k q c, (8) k w w q c, k (9) kl k k co co q c, (10) n co r q c, n (11) n n n w q d, l (12) kl l cu d q, v (13) v v v v n n w q q, k (14) k l kl cu co q q, (15) v n n r co q q, n, t (16) nt t n r s new q q q,, t (17) nt it i q +q = q, (18) r new n0 k k r co q n0 qn n (19) q = q - q - ( q ) n (20) r co r r nz n n0 t nt t q ( q q ) n (21) r co r nz n n0, k,, n 0,1,, k,, n (22) q, q, q, q, q, q, q, q, q N 0 i,, k, l,, n, v (23) s w cu co r r r new it k kl v n nt n0 nz Constraints (7) and (8) forulate the capacity liitation of suppliers and anufacturers. Constraint (9) represents the capacity liitation of DCs. Constraints (10) and (11) show the capacity liitation in reverse logistics for collection points and recycling copanies. Constraint (12) restrains that the retailers' deand ust be satisfied, which also eans the custoer deand ust be satisfied. Constraint (13) explains the relationship between custoer recovery and recovery rate. Constraints (14) and (15) guarantee the in-flow equal to out-flow in each warehouse and each collection point respectively. Constraint (16) guarantees the output recycled aterials don't exceed the axiu value that each recycling copany can extract fro used products. Constraint (17) shows that for each aterial in each anufacturer, the su of the raw aterial fro both suppliers and recycling copanies can eet the deand for producing and reanufacturing needed new products. Constraint (18) restricts that for each anufacturer, the quantity of the provided products are the su of the newly produced ones and the reanufactured ones. Constraint (19) restricts that the percentage of good quality used products in each recycling copany cannot exceed the axiu value preset. Constraint (20) represents that the disposed aterials are MLB 115
reaining aterials after all of the recycling processes. Constraint (21) restricts that in each recycling copany, the total quantity of disposed aterials ust under the acceptable value, the right ite of this inequality displays the transforation fro the quantity of collected bad quality products to the weight of the aterial needs to be disposed of. Constraint (22) represents the binary variables. Constraint (23) represents the integer variables. METHODOLOGY A two-stage priority-based GA The obective of the proposed CLSC odel is to iniize the total cost of transportation and facility operation by designing an appropriate delivery route and delivery flow. In this study, a two-stage priority-based GA is developed to solve the proble described above. The proposed GA ipleents a two-stage priority-based encoding to solve this proble, which are Stage 1 - "Route Decision" and Stage 2 - "Freight Volue Decision". Firstly, Route Decision is applied to design the delivery route between each level of the supply chain network. And then, according to the results of stage 1, Freight Volue Decision is ipleented to decide the freight volue in each decided route. Stage 1 Route Decision In the first stage of encoding, the chroosoe has nine sections to represent each level of CLSC respectively. In each section, the nuber of genes equals to the product of the nuber of suppliers and the nuber of deanders. Totally in a chroosoe, the nuber of genes is I J T+ J K+ K L+V M + M N N J ( T 1). In this nuerical exaple, T=2, I 1 =3, I 2 =2, J=4, K=2. Two kinds of aterials are considered. I 1 =3 eans the nuber of suppliers supplying aterial one is three, I 2 =2 eans the nuber of suppliers supplying aterial two is two. Suppliers provide aterials to four anufacturers, and this four anufacturers will deliver finished products to two warehouses. Each gene contains a binary nuber, 1 eans the delivery route is used and 0 eans not. The chroosoe at this stage is shown in Figure 2. Obviously, the first section of the chroosoe has 3*4=12 genes, it represents the delivery route of aterial one: the first four genes represent the situation of supplier one providing aterial one to anufacturers two and four but not providing aterials to anufacturers one and three. The second four genes represent supplier two with aterial one, and the third four genes represent supplier three with aterial one. The second section of the chroosoe has also 2*4=8 genes, which represents the delivery route of aterial two. The principle is the sae as the first section. The third MLB 116
section of the chroosoe has 4*2=8 genes: The first two genes represent the situation of anufacturer one providing finished products to neither of the two warehouses, the second two genes represent anufacturer two, and so on. Material 1 Material 2 Products 0 1 0 1 0 1 1 0 0 0 0 1 0 1 0 1 0 1 1 0 0 0 0 1 1 0 1 1 Supplier 1 Supplier 2 Supplier 3 Supplier 1 Supplier 2 M1 M2 M3 M4 Figure 2 The first three sections of chroosoe in stage one Stage 2 -Freight Volue Decision After the first stage, a chroosoe representing the delivery route has been established. The second stage is to decide the freight volue of aterials or products according to the generated route. It is called freight volue decision. Figure 3 shows the chroosoe after the second stage of encoding. Material 1 Material 2 Products 0 50 0 50 0 0 300 0 0 0 0 200 0 50 0 50 0 0 300 0 0 0 0 50 300 0 200 150 Supplier 1 Supplier 2 Supplier 3 Supplier 1 Supplier 2 M1 M2 M3 M4 Figure 3 The first three sections of chroosoe in stage two The structure of the chroosoe reains the sae as in stage one, but the content has been changed. Aong the first four genes, the first gene eans supplier one won't delivery any raw aterial one to anufacture one, the second gene represents that supplier one will deliver 50 units of raw aterial one to anufacture two. Other sections have the close principle. The last gene in Figure 3 represents that anufacture four will deliver 150 units of products to warehouse two. The process "Cost Rank" is ipleented at the beginning of encoding to decide the priority in freight volue decision. In this process, the unit cost of each delivery route is ranked in ascending order. The results of the rank are prepared for the process of freight volue decision. Genetic operations Genetic operations play an iportant role in genetic algorith. Considering the properties of chroosoes in this proble, one-point crossover and one-point utation are ipleented to avoid draatic changes in the genetic structure and prevent rando genetic search. MLB 117
COMPUTATIONAL EXPERIMENTS To deonstrate the applicability of the proposed odel and the stability of the proposed GA, three coputational experients with different scales are ipleented: basic, iddle and large. Table 1 shows the scale of these three experients. Table 1 Scale of coputational experients Suppliers Supplier (t=1) s (t=2) Manufacturers Warehouses Retailers Basic Scale 3 3 4 3 5 Middle Scale 6 6 8 6 10 Large Scale 12 12 16 12 20 Custoers Collection Points Recycling Centers Basic Scale 3 4 2 Middle Scale 6 8 4 Large Scale 12 16 8 The dataset of basic scale experient is randoly generated. Other scales are doubled and redoubled of the basic one. Table 2 shows the capacity and fixed cost of each facility in the basic scale experient. Table 3 shows the deand of retailers and the recycling quantities of custoers. Table 4 shows the unit shipping cost ($) in the basic scale experient. Table 2 The capacity and fixed cost in the basic scale experient Manufacturers Warehouses Collection points Recycling Suppliers copanies Capacity Fixed Fixed Fixed Fixed Capacity Capacity Capacity Capacity cost cost cost cost 500 1 800 500 80 65 180 300 t=1 t=2 600 1400 900 650 90 260 450 600 500 600 1500 0 900 120 700 800 800 1800 150 140 800 850 Table 3 The deands and recycling quantities in the basic scale experient Retailers Deands 300 500 500 600 Deands 500 700 800 Custoers Recycling quantities 154 168 Recycling rate 0.20 0.22 0.21 MLB 118
Table 4 The unit shipping cost in the basic scale experient Costs M W R 1.5 2.5 2 3 10 12 15 5 6 8 7 4 3 1.5 2.5 4 15 13 14 W 8 6 5 3 5 M 2 2.5 2.5 2 9 10 14 9 4 5 6 7 S 4.5 6 7.5 6.5 10 11 12 7 6 5.5 6.5 M 5 6 6.5 6.5 9 8 7 9 t=0 Co RC 8 6 10 8 9 8 10 8 10 13 5 6 7 4 RC t=1 Cu 7 9 10 12 12 12 6 7 4 4 Co 9 10 12 10 13 14 10 13 12 15 t=2 10 15 12 14 13 10 S: Supplier, M: anufacturer, W: warehouse, R: retailer, Cu: custoer, Co: collection points, RC: recycling copany The coputational results with proposed GA of the basic scale experient are the sae as the results with Lingo. It can be verified by drawing the closed-loop supply chain out. The verification of the results is seen in Figure 4. The results provide the optial delivery route and delivery flow decision in the CLSC network, and these also give out the decision of facilities operational state. This integrated optiization can provide a reliable decision support for the printer cartridge copany. The results of iddle scale and large scale experients are also verified. All the three experients' results are displayed in Table 5 coparing with results of Lingo. These results worked out by the provided GA, using a PC with Intel(R) Core(TM) i7-2600 CPU @ 3.4GHz, 8.0G RAM. Table 5 The results and coparison between the proposed GA and Lingo 50 Ties each probles Lingo 11.0 Proposed GA (population size=) Scale Nuerical exaples 1 2 3 Optial (US$) 63709.5 125118 248833 Tie (s) 1 20 150 Min-cost (US$) 63709.5 126014 250367 Absolute difference 0-1236 -1534 Percentage difference 0-0.7% -0.62% Average-tie (s) 0.09 2.04 13.25 Percentage of tie 9% 10.2% 8.83% MLB 119
Table 5 shows the coparison between Lingo 11.0 and the proposed GA with three scales. The row of "Absolute difference" expresses the value that results of Lingo inus results of proposed GA. The inus sign eans disparity. Take Scale 2 for exaple, the value of absolute difference is -1236, and the percentage difference is -0.7%. That eans the result of proposed GA is 0.7% disadvantage copare with that of Lingo. However, the average tie of the proposed GA is 2.04 s, which is only 10.2% of the tie with Lingo. In Scale 3, the result of proposed GA is 0.62% disadvantage copare with that of Lingo, the running tie is only 8.83% of Lingo's. Suppliers Manufacturers Warehouses Retailers S1 600 S2 700 S3 800 t=1 S4 500 S5 800 S6 850 t=2 WDP 348 500 714 52 564 52 236 329 48 400 138 Disposal M1 500 M2 600 M3 600 M4 800 135 36 86 500 RC1 180 RC2 260 600 300 300 116 92 200 30 50 150 W1 800 W2 900 W3 0 Co1 80 Co2 Co3 120 Co4 150 600 400 500 200 62 18 92 150 R1 R2 300 R3 500 R4 500 R5 600 Cu1 500 Cu2 700 154 Cu3 800 168 Waste disposal plant Recycling copanies Collection points Custoer regions Products Material 1 Material 2 Figure 4 Result of basic scale experient MLB 120
CONCLUSIONS With the booing developent of electronic industry, the exponential growth of e-waste has polluted the environent seriously. Producers are perceived to be responsible for the recycling of products they have produced. However, few anufacturers adopted proper easures to deal with this because of expensive costs. To solve this proble, a coprehensive CLSC odel of cartridge recycling is established in this paper. This odel, which based on the real situation of cartridge recycle in Hong Kong, integrates both forward and reverse flow of products and contains eight partners in CLSC. In this research area, although any researchers discuss CLSC odels, few of the odels analyze the delivery activity for different kinds of aterials that extracted fro used products, and also few studies classify the collected used products according to the quality. The odel proposed in this paper can address these issues. Moreover, this proble is forulated into a linear prograing odel and solved with a odified two-stage priority-based encoding GA which enhances the genetic searching ability. The adopted algorith optiizes the CLSC network and the results show a near optial solution, which can provide producers a reliable decision support. Acknowledgeents The work described in this paper was partially supported by The Hong Kong Polytechnic University Research Coittee (Account code: RT3D); and a grant fro the Research Grants Council of the Hong Kong Special Adinistrative Region, China (Proect No. PolyU 510311). REFERENCES Aluur, S. A., Nickel, S., Saldanha-da-Gaa, F., & Verter, V. (2012). Multi-period reverse logistics network design. European Journal of Operational Research, 220(1), 67-78. Ain, S. H., & Zhang, G. Q. (2012). A three-stage odel for closed-loop supply chain configuration under uncertainty. International Journal of Production Research, 51(5), 1405-1425. Dwivedy, M., & Mittal, R. K. (2012). An investigation into e-waste flows in india. Journal of Cleaner Production, 37, 229-242. Gaberini, R., Gebennini, E., Manzini, R., & Ziveri, A. (2010). On the integration of planning and environental ipact assessent for a WEEE transportation network-a case study. Resources Conservation and Recycling, 54(11), 937-951. MLB 121
Mar-Ortiz, J., Adenso-Diaz, B., & Gonzalez-Velarde, J. L. (2011). Design of a recovery network for WEEE, collection: The case of galicia, spain. Journal of the Operational Research Society, 62(8), 1471-1484. Olugu, E. U., & Wong, K. Y. (2012). An expert fuzzy rule-based syste for closed-loop supply chain perforance assessent in the autootive industry. Expert Systes with Applications, 39(1), 375-384. Ozkir, V., & Basligil, H. (2012). Modeling product-recovery processes in closed-loop supply chain network design. International Journal of Production Research, 50(8), 2218-2233. Tsai, W. -H., & Hung, S. (2009). Treatent and recycling syste optiisation with activity-based costing in WEEE reverse logistics anageent: An environental supply chain perspective. International Journal of Production Research, 47(19), 5391-5420. Veenstra, A., Wang, C., Fan, W., & Ru, Y. (2010). An analysis of E-waste flows in china. International Journal of Advanced Manufacturing Technology, 47(5-8), 449-459. Wang, H. F., & Huang, Y. S. (2013). A two-stage robust prograing approach to deand-driven disassebly planning for a closed-loop supply chain syste. International Journal of Production Research, 51(8), 2414-2432. MLB 122