Journal of the Operational Research Society (2012) 63, 606–619

© 2012 Operational Research Society Ltd. All rights reserved. 0160-5682/12 www.palgrave-journals.com/jors/

A novel heuristic approach to determine compromise management for end-of-life electronic products SC Lee and LH Shih National Cheng Kung University, Taiwan, ROC The European Union directive of Waste Electrical and Electronic Equipment for recycling end-of-life (EOL) products has had a significant impact on global enterprises. Recent studies have shed light on optimization of the EOL process. In addition to identifying the most economical EOL process, identifying the EOL process with the smallest environmental load is equally important, and this is thus a bi-criteria optimization problem. This study attempts to optimize EOL processes for electronic products based on a three-stage heuristic approach, which simultaneously minimizes cost and environmental impact. The proposed heuristic approach then assesses the most common disassembly and recycling processes by using the characteristics of electronic product recycling. Next, the best process for this bicriteria optimization problem is identified by using the compromise programming method. The empirical analysis is based on data for notebook, and the potential impact on best EOL processes when notebook adopt new product designs is also discussed. Journal of the Operational Research Society (2012) 63, 606–619. doi:10.1057/jors.2011.50 Published online 13 July 2011 Keywords: end-of-life; environmental load; bi-criteria optimization; heuristic approach; compromise programming

1. Introduction After the European Union (EU) enacted the Waste Electrical and Electronic Equipment (WEEE) directive in 2003, many electrical and electronic product producers worldwide have been forced to responsibility for end-of-life (EOL) products. In 2008, all EU members promulgated environmental regulations in response to the WEEE directive, and the US government has also issued its own environmental regulations (eg, the E-Waste Recycling Bill and the Electronic Waste Recycling Act) in response to the WEEE directive. Even the largest emerging market worldwide, China, has faced a dramatic increase in electronic waste. To resolve this problem, the China government issued the Administrative Measures for Prevention and Control of Environmental Pollution from Electronic Waste in response to the WEEE directive. Although the WEEE directive requires producers to recycle products, original equipment manufacturers (OEMs) have the most responsibility for EOL products. For example, most OEMs and brand-name companies in Taiwan have 

Correspondence: SC Lee, Department of Resources Engineering, National Cheng Kung University, No. 1, University Road, Tainan City 701, Taiwan, ROC. E-mail: [email protected]

seriously considered adopting cleaner production and design for recycling (DfR) and proactive EOL recycling to comply with the WEEE directive, even though Europe only accounts for roughly 25% of the global market. A common electrical and electronic product, notebook computers, contains liquid crystals in the liquid crystal display (LCD), mercury in the cold cathode fluorescent lamp (CCFL), lead, cadmium, and mercury in the battery, and halogen flame-retardant agents in plastic casing. Thus, careless disposal can cause significant environmental damage. Different recycling methods have different degrees of impact on recycling benefits and the environment. When notebooks reach the EOL processes, the primary issue for enterprises is how to optimally disassemble and recycle, such that cost and environmental impact are minimized. Restated, finding the best recycling process for EOL products is a bi-criteria optimization problem. In response to the WEEE directive, producers should be provided with a disassembly report containing a brief product introduction, connection techniques, disassembly sequence, disassembly tools, disassembly methods, disassembly time, part weights, part materials, part photographs and a Reuse, Recycling, and Recovery (3R) rate. However, producers cannot determine the recycling cost for EOL products and the environmental impact of the EOL process using only a disassembly report. As environmental

SC Lee and LH Shih—A novel heuristic approach

concerns have become an important issue, as well as a business opportunity of increasing significance, producers need to identify the best EOL process by simultaneously considering the criteria of minimizing cost and environmental impact. Therefore, how to accurately calculate cost and environmental impact of a given EOL process is an important problem for producers. Traditionally, optimizing the disassembly and recycling process has been undertaken using various methods such as mathematical programming (MP) and empirical law (eg, appropriate recycling and recovery rate). Feldmann et al (1999, 2001) has noted that optimizing the EOL strategy is even more complicated when reuse and material recovery are considered together with disassembly. Determination of a break point between disassembly and the shredding

607

process is often affected by factors like labour cost and the market prices of reclaimed materials. This study proposes a heuristic approach to combine these methods and identify the best EOL process based on the assumptions of electronic product recycling while considering both economic and environmental criteria, which are modelled as a bi-criteria optimization problem. A novel heuristic approach is applied to identify the most feasible recycling processes. Figure 1 shows the study scheme. First, this study simulates EOL product recycling benefits and environmental impacts while assuming that information for EOL processes is available. In this study, the EOL recycling process includes manual disassembly, shredding, material recovery, and waste disposal. Prior to the break point (as presented in Figure 2), the EOL process of parts for the EOL product

Figure 1 Scheme of heuristic for finding best EOL recycling process.

608

Journal of the Operational Research Society Vol. 63, No. 5

Therefore, the following assumptions are employed to obtain the major solution set.

Figure 2 Determination of the break point and the best EOL strategy.

was manual disassembly, including this part. After the break point, the EOL process is shredding (Shih and Lee, 2007). As the break point is defined between manual disassembly and shredding, the related cost and profit can be calculated. The part can be used for reuse and recycling by manual disassembly and material recovery by shredding. Profits of substances yielded from disassembly and materials reclaimed after shredding need to be considered separately, because parts and materials reclaimed from disassembly often have higher prices than the materials that are reclaimed after the shredding process. Materials reclaimed from shredding are in a bulk form and are mixed with other materials, which must be further processed prior to reuse. In addition, the environmental impact of the EOL process is evaluated by SimaPro, a European software package for lifecycle assessment. The calculation of benefits and environmental impacts are presented in Sections 3 and 4. Second, the novel heuristic approach is applied to find a ‘major’ EOL process without actually solving a combinational optimization problem. This method rapidly evaluates the major EOL processes using a simulation model and identifies a process that maximizes benefit and minimizes environmental impact. In other words, this heuristic approach helps producers to identify the best EOL process based on the assumption that EOL recyclers choose a process that minimize cost and environmental impact. Furthermore, the proposed heuristic approach is based on the characteristics of electronic product recycling, in that the process for disassembling mechanical products often differs from the process for disassembling electronic products. Since electronic products are typically small, thin, and compact, and parts are frequently soldered together instead of fastened, and these parts cannot be disassembled and reused, unlike parts in mechanical products. As the cost of manually separating electronic parts and materials is higher than that for mechanical products, most electronic subassemblies are shredded for material recovery. In addition, the Restriction of Hazardous Substances directive strives to approximate the laws of the Member States in terms of restricting the use of hazardous substances in electrical and electronic equipment, as well as contribute to the protection of human health and the environmentally sound recovery and disposal of WEEE.

(1) Appropriate modules and subassemblies for EOL products are defined based on the characteristics of electronic product recycling. (2) Modules are adopted as the basic disassembly unit. (3) The disassembly sequence for each module is based on observations of experienced workers. (4) Hazardous materials are the first priority in disassembly. (5) Potential break points between manual disassembly and shredding for material recovery are identified. The above five assumptions can be adopted to reduce the number of feasible solutions (possible EOL processes) searched, thereby accelerating identification of the major solution set. Furthermore, Feldmann et al (2001) and Lee et al (2001) noted that the complexity of optimizing the EOL process increases when reuse and material recovery are considered together with disassembly. Determination of a break point between disassembly and the shredding process is often affected by such factors as labour cost and market prices for reclaimed materials. Hence, this study uses the proposed heuristic approach to find the near-optimal and satisfactory best solution, and evaluates the economic and environmental performance of feasible solutions. The remainder of this paper is organized as follows. Section 2 provides a literature review. Cost and profit items included in the simulation are described in Section 3. Section 4 evaluates the environmental impact of the EOL processes. Section 5 applies the proposed heuristic approach to identify the best EOL recycling processes via bi-criteria optimization, and assumptions and the related reasoning processes are also presented. A notebook computer is used as an example illustrating the application of the proposed heuristic approach in Section 6. Section 7 discusses the impact on the EOL recycling processes when the product designs of the notebook computer in Section 6 changes. Finally, Conclusions are drawn in Section 8.

2. Literature review 2.1. Overview of the disassembly problem Dong and Arndt (2003) pointed out that disassembly is a systematic approach separating a product into its constituent parts, components, or subassemblies. The significance of disassembly can be classified into the following three broad categories. First, disassembly has a key role in the recycling of materials and components to meet the goal of environmental protection. Second, disassembly is a key process in remanufacturing, which is an economical way to hold the products’ cost by allowing selective separation of desired parts and materials. Third, disassembly sequences are studied for service purposes. Waterbury (1985) noted

SC Lee and LH Shih—A novel heuristic approach

that product assembly time and assembly cost can be reduced by using the concept of design for assembly (DfA). Mazhar et al (2007) integrated Weibull analysis and the artificial neural network model to assess the remaining useful life of components for reuse.

2.2. Disassembly sequence planning 2.2.1. MP approaches. Optimizing the disassembly sequence has traditionally been modelled as a combinatorial optimization problem and solved with various approaches such as MP and heuristic approaches (Lambert, 2003). MP approaches require modelling with a high level of abstraction. Starting with an assembly drawing or computer-aided design file, a connection diagram and set of precedence relations are derived, and this requires visual inspection or dedicated software. An important goal of this work is to minimize the effort in this formalization step. Once the problem has been formulated, disassembly graphs can be automatically generated, and optimum and sub-optimum sequences can be determined, provided that parameter values are introduced, such as the cost of each action and revenue from each subassembly. Proper modeling of the problem is thus a prerequisite for the MP approach. Once modelled, methods exist for efficiently selecting the optimal solution. These approaches are usually based on linear programming, mixed integer programming, and dynamic linear programming. Gonza´lez and Adenso-Diaz (2006), Lambert (2007), and Lambert and Gupta (2002) used a MP approach to formulate the disassembly sequencing problem as an optimization problem. In additional, Gao et al (2004), Kongar and Gupta (2006), Neuendorf et al (2001), Tang and Turowski (2007), and Zussman et al (1994) applied graph theory and the Petri net method to solve the disassembly sequencing optimization problem. Andre´s et al (2007) and Dong et al (2007) integrated the Petri net and an And/Or graph to solve optimization problems. Conversely, Galantucci et al (2004), Kongar and Gupta (2002), Liu et al (2002), and Shih et al (2006), who recognized the complexity in solving the problem using a mathematical approach, adopted artificial intelligence to derive the optimal strategy.

2.2.2. Heuristic approaches. Pu (1992) generated an assembly sequence using case-based search techniques, in which a case library is applied to re-use existing solutions to problems for a new problem. Pu also developed novel learning algorithms. Similar techniques were applied by Veerakamolmal and Gupta (2002), who investigated the disassembly of electronic products consisting of different configurations of known modules. Swaminathan and Barber (1996) represented an assembly process using three

609

graphs—a connection diagram, a mating directions graph, and an obstacle facts graph. Problems were thus separated into subproblems that were stored in a library. Gungor and Gupta (1997) developed a heuristic approach to select near-optimal disassembly sequences and applied them to personal computer disassembly. As disassembly sequence planning is an NP-hard optimization problem Navin-Chandra (1994), Gungor and Gupta (2001), Kuo et al (2000), Lee and Xiroucharkis (2004), Rosell et al (2003), and Shih and Lee (2007) suggested using a heuristic approach to simplify the calculation process, and thereby reduce optimization problem complexity and solution time when finding a near-optimal solution.

2.3. Optimal EOL strategy Lee et al (2001) used the multi-objective linear programming (MOLP) model to determine the appropriate EOL options for components in manufactured products based on the objective of maximizing profit (or minimizing cost). Bakal and Akcali (2006) applied the MP approach to determine the optimal acquisition price for the EOL product and the selling price for remanufactured parts while reflect an EOL product from which a particular part can be recovered and remanufactured for reuse, and the remainder of the product can be recycled for material recovery. Both the supply of EOL products and demand for remanufactured parts are price-sensitive. Yield of the recovery process is random and depends on the acquisition price offered for the EOL products. As environmental regulations and take-back laws are becoming more stringent around the world, OEMs are forced to consider reverse supply chain strategies to ensure the collection and processing of their used (ie, EOL) products. Flowers and Linderman (2003) developed a goal-programming approach to the waste-fuel-blending process that considers the diverse objectives of fuel managers. Karakayali et al (2007) also used the MOLP model to determine the optimal acquisition price of the EOL products and the selling price of the remanufactured parts in centralized, as well as remanufacturer- and collector-driven decentralized channels. Webster and Mitra (2007) developed a general two-period model to examine the impact of take-back laws within a manufacturer/remanufacturer competitive framework. Their analytical results demonstrated that enactment of collective WEEE take-back will result in higher manufacturer and remanufacturer profits while simultaneously spurring remanufacturing activity and reducing the tax burden on society. A negative effect is higher consumer prices in the market. Webster and Mitra (2007) also found that collective WEEE take-back introduces a structural change to the industry—creating an environment where remanufacturing becomes profitable when it is not profitable without a take-back law. In recent years, in line with

610

Journal of the Operational Research Society Vol. 63, No. 5

the growing awareness of a need for environmental protection, companies have started processing hazardous or poisonous wastes when the EOL product exceeds cost or profit considerations. The environmental impact of different recycling methods has been considered when attempting to find an optimal EOL process that balances economic benefit and environmental impact. Bufardi et al (2003), Mergias et al (2007), and Staikos and Rahimifard (2007) used the analytical hierarchy process to establish a criteria hierarchy of EOL processing and construct an evaluation model. The criteria can be classified into the following three broad categories: economic/financial, environmental, and social/technical. Ferna´ndez et al (2010) introduced the maximum dispersion territory design problem based on the greedy randomized adaptive search procedure heuristic methodology. The problem consists of designing territories for corporations that recollect and recycle different types of white goods at the end of their lifetime. Ferguson and Toktay (2006) also used the MOLP model to determine the optimal manufacturer’s recovery strategy in the face of a competitive threat on the remanufactured product market. The average cost to collect/remanufacture is modelled as an increasing function of the quantity collected/remanufactured, thus capturing a unique aspect of the remanufacturing industry that has not been explored in previous market segmentation research. Kongar and Gupta (2006) recognized that appropriate EOL process should achieve four goals simultaneously—maximize resale profit, maximize recycling profit, minimize inventory and minimize waste, and used fuzzy goal programming to solve this multi-objective problem. Chan and Tong (2007) applied grey relational analysis (GRA) to determine the relationship between product materials and EOL treatment methods, and the measurement indicators in GRA were economic/cost criteria and environmental criteria. Ravi et al (2008) utilized the analytical network process (ANP) and zero one goal programming (ZOGP) to determine the EOL process. The ANP was used to determine the degree of interdependence among criteria and candidate reverse logistics projects, while ZOGP permits consideration of resource limitations and other constraints when arriving at a solution.

3. Cost and profit estimations for EOL recycling processes An EOL recycling process generally includes manual disassembly, shredding and separation, material recycling, energy recovery, and disposal of waste and hazardous substances (Lee, 2006). The corresponding costs and profits of recycling processes can be calculated using detailed information about disassembly, shredding and material recovery, energy recovery, ordinary waste disposal, and hazardous waste disposal processes. Cost and profit can be calculated

based on the weight of an EOL product, its component materials and market prices for reclaimed substances. In summary, the cost incurred generally comprises disassembly cost, shredding and separation cost, and the fees for ordinary and hazardous waste disposal. On the other hand, profit from the EOL processes generally includes the following: (1) Profit from parts or materials yielded from disassembly. (2) Profit from reclaimed materials recovered after shredding. Thus, when a product reaches EOL processes, this study adopts an estimation model (ie, Equation (1) to simulate total benefit. By defining the parts undergoing disassembly, shredding and disposal, the cost and profit can be calculated accordingly. The first term in Equation (1) derives the cost and profit of material recycling after disassembly. The second term derives the cost and profit of materials recovered after shredding. Finally, the third term derives the cost of processing ordinary waste and hazardous waste. The unit measurement of cost and profit items is New Taiwan Dollars (NTD) per kilogram. ( ) S X CXS ¼ ½ðBDi  Wi Þ  ðCL  Ti Þ  ðCDE  Wi Þ i¼1

(

þ

n  X

 BAi  Wi  rq  ðCHE  Wi Þ

i¼Sþ1



   Cj  Wi  1  rq 

( X i2J

ðCJ  Wi Þþ

X

) )

ðCH  Wi Þ

ð1Þ

i2H

where CXs is total benefit of recycling process with break point set at part s (NTD/set); BDi is material unit price of part i from disassembly (NTD/kg); Wi is weight of part i (kg); CL is unit cost of labour for disassembly (NTD/ second); Ti is time for disassembling part i (second); CDE is unit cost of purchasing disassembly equipment (NTD/kg); BAi is material unit price of part i after shredding (NTD/ kg); rq is recovery rate, ranging from 80 to 98% depending on material types; CHE is unit cost of shredding and material recovery equipment (NTD/kg); CJ is unit disposal cost for ordinary wastes, with J as a set of ordinary wastes (NTD/kg); CH is unit disposal cost of hazardous wastes, with H as a set of hazardous wastes (NTD/kg); S is part number where break point is set; i is index for part number, i ¼ 1Bn; n is total number of parts.

4. Environmental impact evaluation for recycling processes Owing to environmental regulations that require OEMs to collect and process their EOL products, there has been

SC Lee and LH Shih—A novel heuristic approach

increased interest towards in corporate activities aimed at reducing or eliminating the waste created during the production, use and/or disposal of the firm’s products. Melnyk et al (2003) conducted a questionnaire survey of North American managers to demonstrate that firms in possession of a formal environmental management system (EMS) perceive impacts well beyond pollution abatement and see a critical positive impact on many dimensions of operations performance. Analytical results also showed that firms that had gone through the EMS certification experienced a greater impact on performance than firms that had not. Therefore, environmental impact is the second criterion for selecting EOL recycling processes. Eco-indicator 99 in SimaPro software (Effting, 2006) is applied to generate lifecycle assessment data for various EOL processes. The database in SimaPro is constructed primarily using European data. Once material types, weights and manufacturing processes for all parts and their recycling and disposal methods are input, SimaPro calculates an overall impact index and individual impact indices at different lifecycle stages for material acquisition, manufacturing, transportation, recycling, and disposal. The normalized environmental impact is expressed in milli-point (mPt) by Eco-indicator 99. Lifecycle assessment needs a reference point. This study assumes no prior recycling, and any recycling will be counted as ‘saving’ environmental impact. For a given EOL recycling process, the environmental impact should be considered in terms of energy used in the material recovery process, and the disposal of ordinary and hazardous waste. When a product reaches EOL processes, Equation (2) is applied to derive total environmental impact. The first term derives the environmental impact prevented because parts are disassembled and recycled. The second term derives the environmental impact for parts going through the shredding process. Finally, the third term derives the environmental impact caused by ordinary and hazardous disposal. Although material recovery prevents any adverse environmental impact, however, for materials that cannot be recycled, the shredding process itself has an adverse environmental impact. The unit of environmental impact is ‘mPt’ per kilogram. EXS ¼

( S X

) ½ðEMi  Wi Þ þ ðECi  Wi Þ

i¼1

(

þ

n X

   ðEMi  Wi Þ þ ðECi  Wi Þ  rq

i¼Sþ1





þ½ESi  Wi  þ EJi  Wi  1  rq

) 

( " #) X X þ  ðEJi  Wi Þþ ðEHi  Wi Þ i2J

i2H

ð2Þ

611

where EXs is total environmental impact for the recycling process when the break point is set at part s (mPt/set); EMi is environmental impact at the material acquisition stage for part i (mPt/kg); Wi is weight of part i (kg); ECi is environmental impact at the manufacturing stage for part i (mPt/kg); EJi is environmental impact for treating ordinary waste, with J as a set of ordinary waste (mPt/kg); EHi is environmental impact for treating hazardous waste, with H as a set of hazardous waste (mPt/kg); ESi is environmental impact caused by the shredding process (mPt/kg); rq is recovery rate, ranging from 80 to 98% depending on material types; S is part number where break point is set; I is index for part number, i ¼ 1Bn; N is total number of parts.

5. Heuristic to identify bi-criteria best EOL recycling processes Sections 3 and 4 discussed how to calculate the benefit and environmental impact for EOL processes. In this section, a novel heuristic approach is applied to identify bi-criteria optimization of the EOL recycling processes. The following three stages are utilized: (1) Identify all combinations (feasible solutions) of recycling processes and evaluate their economic and environmental performance. The combinations of disassembly, shredding, and waste disposal are determined based on the recycling characteristics of EOL electronic products. (2) Compare all major solutions to find Pareto solutions. (3) Identify the best solution by using compromise programming. The proposed heuristic approach for identifying the best EOL recycling processes under bi-criteria uses the following nine steps: (1) Define appropriate modules and subassemblies for EOL products: According to a product’s bill of materials (BOMs), field disassembly experience of engineers, and electronic product characteristics, EOL products are disassembled into several modules and subassemblies. (2) Prioritize hazardous materials to establish the necessary disassembly sequence: Prioritize parts that contain hazardous materials so that they are disassembled first, as required by environmental regulations. The necessary disassembly sequence can be the shortest disassembly route for hazardous parts. (3) Establish the disassembly sequence for each module and collect the required information: Evaluate in a labouratory the possibility of disassembling each subassembly manually. Disassemble until final parts are obtained when possible, and record the disassembly

612

Journal of the Operational Research Society Vol. 63, No. 5

time, weight, and materials of the disassembled parts in a step-by-step manner. These data are entered into the mathematical formula in Sections 3 and 4 to estimate the corresponding benefit and environmental impact of the EOL processes. (4) Estimate the benefits and environmental impact of the disassembly–shredding break point for each module: Using Equation (1), estimate the benefit of the disassembly–shredding break point for each module, which are composed of three items, disassembly benefit, shredding benefit, and disposal cost of ordinary and hazardous waste. The recycling processes and benefits for each module can be determined according to these cost and profit items. After determining the recycling processes for each module, the environmental impact for each module can be estimated using Equation (2). (5) Identify all combinations (feasible solutions) of recycling processes: After the benefit and environmental impact of the disassembly–shredding break point for each module are calculated, all feasible solution combinations can be constructed. Assume that EOL products have six modules, with each module containing four parts; that is, each module has five benefits and five environmental impact values for the disassembly–shredding break point. Therefore, 5  5  5  5  5  5 ¼ 56 feasible solution combinations exist. (6) Construct the major feasible solutions set: Standardization was adopted to discuss the economic and environmental criteria and avoid the problem of different variable units. The major feasible solutions set can be constructed using the standardized values of feasible solutions. Equations (3) and (4) are the standardization equations, in which X and Y are the non-standardized benefit and environmental impact, and X and Y are the standardized benefit and environmental impact, respectively, while X and Y are standardized and between 0 and 1. X  ¼ ½X  ðMax benefitÞ=½ðMax benefitÞ ðMin benefitÞ

ð3Þ

Y  ¼ ½Y  ðMin environmental impactÞ= ½ðMax environmental impactÞ ðMin environmental impactÞ

ð4Þ

(7) Compare all major feasible solutions to find Pareto solutions: To illustrate the tradeoff between benefit and environmental impact, Figure 3 plots all standardized feasible solutions on the X–Y axis. An efficient frontier line denotes a Pareto boundary for decision-making involving economic and environmental criteria.

Figure 3 Schematic diagram for feasible solution set and efficiency boundary.

Point C (f1 , f 2 ) (as presented in Figure 3) represents the best solution under bi-criteria decision-making;  f 1 is the maximum value of economic benefit; and f 2 is the minimum value of environmental impact. Point A is a feasible solution for the maximum economic benefit, that is, point A is the value of maximized benefit in the major feasible solutions set. Similarly, point B is smallest environmental impact of all the feasible solutions, meaning that point B is the value of minimized environmental impact in major feasible solutions set. Therefore, a solid line (ie, connecting points A and B) is the efficient frontier line for this solution set. The efficient frontier line identifies which solutions are efficient, and suggests that combinations in the upper-left side of the efficient frontier are inadequate, and these are called inefficient feasible solution combinations. Among the efficient feasible solution combinations, decision-makers can choose EOL strategies based on their preferences. In other words, rational decision-makers typically compromise between economic benefit and environmental impact, and are responsible for the overall profit derived from selecting their preferred EOL recycling processes. (8) Calculate the shortest distance by using compromise programming: via compromise programming (Yu and Leitmann, 1974), two standardized criteria values are transformed into a distance function, denoted as Zp in Equation (5). Decision-makers decide which method to apply for bi-criteria decision-making (decide wi and p) to solve the shortest-distance problem, which is denoted as min Zp: ( min ZP ¼

I  X

wpi ðfi

 fij Þ

p 1=p

) ;

j ¼ 1; . . . ; J

i¼1

s:t:

x2S

ð5Þ

SC Lee and LH Shih—A novel heuristic approach

where Zp is distance function, for any p in which 0opoN; wi corresponds to a weight of a particular objective; fi  is maximum normalized value for objective category i; fij is normalized value of objective category i under decision variable level j; i is number of objectives categories, i ¼ 1BI; j is number of discrete decision variable values, j ¼ 1BJ; p is distance parameter, p ¼ 1 is Manhattan distance (block distance) and p ¼ 2 is Euclidean distance (straight line distance), 0opoN; S is feasible solution set. (9) Transform the Pareto solution into EOL recycling processes: The Pareto solution can be obtained by step 8. Therefore, recycling processes of EOL products under bi-criteria optimization can be identified using the Pareto solution.

6. Case study: best disassembly and recycling processes for a notebook computer A notebook is used as a case study to illustrate how to apply the proposed heuristic approach to find the disassembly–shredding break point for each module. Compromise programming is also adopted to find the best solution and an efficient frontier line between economic benefit maximization and environmental impact minimization. Since most consumer electronics products have only a few modules, such as notebook and LCD monitors, the feasible solution using the proposed heuristic approach based on the assumptions and characteristics of electronic product recycling is representative. Real data for the BOMs, materials, disassembly time, and related cost and profit parameters are obtained for a notebook computer (Jien, 2008). This 12.1-inch notebook computer weighs 1.9 kg. The disassembly time for each part was recorded in a labouratory and reviewed by experienced field engineers. The benefits and environmental parameters of the EOL recycling processes are obtained from Wen (2000), who conducted a study of 14 recycling plants in Taiwan. Figures 4 and 5 shows the inputs/outputs of the proposed model and the disassembly sequence diagram of the EOL notebook computer. A notebook can be divided into six modules: the upper case, under case, keyboard, printed circuit board, hard disk, and panel, and these modules have 8, 6, 12, 13, 8, 7 parts, respectively. According to the WEEE Directive, the liquid crystal modules and battery must be processed carefully, and thus we assume these modules and parts are disassembled and removed, and the six major modules are separated. For each module, 9, 7, 13, 14, 9, and 8 break points exist between manual disassembly and shredding for material recovery, that is there are 825 552 (9  7  13  14  9  8) possible break points. Figure 5 plots benefit on the X-axis, environmental impact on the Y-axis; all standardized feasible solutions, called major feasible

613

solutions in this bi-criteria optimization problem, are plotted on the X–Y axis. Compromise programming is applied to calculate the shortest distance between all standardized feasible solutions. Decision-makers then decide how to execute bi-criteria decision-making using the economic and environmental criteria, and different preferences may alter the results. For instance, when the weight of the economic criterion is larger than environmental criterion, the analytical result would have increased economic benefit. Conversely, when the weight of the environmental criterion is large, the recycling processes will have reduced environmental impact. According to the results of applying compromise programming (wi ¼ 1, p ¼ 2), the best solution is point C (as presented in Figure 5), and the corresponding recycling benefit, environmental impact, and recycling process for each module are presented in Table 1. The third column in Table 1 shows the best EOL process for each module and the corresponding recycling benefit and environmental impact values. Point C (as presented in Figure 6) represents the feasible solutions with maximized economic benefit and minimized environmental impact among all major feasible solutions. In this case, the total benefit is roughly 14.88 NTD/set, and environmental impact is roughly 275.61 mPt/set. Table 1 also presents the corresponding recycling process for each module. As Section 2 mentioned, economic benefit and environmental impact are at odds. When recycling processes with large economic benefit are adopted, the damage to the environment typically increases. Conversely, when environmentally friendly recycling processes are adopted, recycling cost generally increases. Furthermore, the findings from this case study enhance the information in the disassembly report by describing the benefit and environmental impact of EOL recycling processes. Therefore, decision-makers must compromise appropriately between these two criteria, and choose recycling processes that maximize economic benefit and minimize environmental impact.

7. Best disassembly and recycling processes for new notebook designs The previous section demonstrated how to identify the best EOL recycling processes and the efficient boundary of a notebook under bi-criteria optimization. When a product designer redesigns a notebook, the product configuration likely changes. For example, changing part materials or the product development process used in green design will affect the EOL notebook recycling processes. Therefore, this section discusses the impact on EOL recycling processes when notebook adopts new product designs. Three scenarios are simulated and discussed.

614

Journal of the Operational Research Society Vol. 63, No. 5

∼

∼

• •

•

•

•

•

•

•

• •

• • • •

• •

•

•

• •

• •

• •

• •

Figure 4 Input/output of the proposed model.

(1) Keyboard module redesign: Disassembly of the original keyboard module requires repeated manual disassembly. This takes considerable time and has high labour costs. To improve these problems, this scenario assumes the original notebook’s keyboard is changed to a single silicon thin-film keyboard and the related data are presented in Table 2. (2) Material change for case module: As consumers now seek out more stylish products, manufacturers have responded by finding new materials for notebook cases. For example, leather or wood cases are now used, and metallic aluminium alloy casing are relatively common. Therefore, this scenario assumes that the ABS casing of the original notebook is changed to aluminium alloy, and the related data are presented in Table 3.

(3) Implementation of a light-emitting diode (LED) backlight source in the panel module: Current commercially available notebook backlight modules are mostly made of CCFL, which contains mercury. Thus, this scenario assumes the original notebook’s CCFL backlight panel is changed to an LED backlight panel, and the related data are presented in Table 4. Following the procedure in Section 6, the best EOL processes can be identified. Table 5 presents the analytical results of the above three scenarios. (1) Improvement to the keyboard module: When EOL recycling processes for silicon-based keyboards are

SC Lee and LH Shih—A novel heuristic approach

615

Figure 5 Disassembly sequence diagram of EOL notebook. Table 1 Best EOL recycling process The number Bi-criteria optimization for point C (wi=1, p=2) of parts

Notebooks

Module EOL recycling process

Necessary disassembly sequence Upper case module Keyboard module Hard disk module PCB module Under case module Panel module

Benefit (NTD/set) Environmental impart (mPt/set)

executed under bi-criteria optimization, recycling processes for the keyboard module are modified to totally manual disassembly, the hard disk module is now fully shredded, and the processes for other modules remain the same. According to the analytical results for this scenario, the silicon keyboard notebook reduces recycling cost and has less environmental impact than the original design. (2) Changing the casing material: When EOL recycling processes for the aluminium alloy are executed under bi-criteria optimization, the recycling processes for the case module remain the same. However, the part material has changed, and thus the recycling benefit and environmental impact have also changed. According to the analytical results for this scenario, the aluminium alloy case notebook reduces both the

15 8 12 13 8 6 7

Disassembly Disassembly Part 1 disassembly, part 2B12 shredding Part 1 disassembly, part 2B13 shredding Part 1B6 disassembly, part 7B8 shredding Disassembly Disassembly

69

14.8788 275.6076

1.0 0.9 0.8 0.7 All major feasible solution set

A 0.6

Maximal solution of economics benefits

0.5 Efficient frontier line 0.4 0.3 C 0.2 0.1 B -1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1

Best solution under bicriteria using compromise programming Minimal solution of environmental impacts

0 Compromise programming wi = 1, p =2

Figure 6 Major feasible solutions set and efficient boundary.

616

Journal of the Operational Research Society Vol. 63, No. 5

Table 2 Related date for single silicon thin-film keyboard Parts

Parts material

Parts weight (g)

Disassembly time (second)

Silicone-based keyboard Control Integrated Circuit (IC) case Keyboard connecter wire Control IC board

Polypropylene ABS PVC þ copper PCB (resin)

105 95 10 50

10 12 8 20

Table 3 Related date for aluminum alloy case Module

Parts

Original material

New material

Necessary disassembly sequence

Random Access Memory protective cover Keyboard front cover LCD frame

ABS ABS ABS

Aluminium alloy Aluminium alloy Aluminium alloy

Upper case Lower case

Upper case Lower case

ABS ABS

Aluminium alloy Aluminium alloy

Table 4 Related date for LED panel Module

Parts

Necessary disassembly sequence

LED light bulb

1.12

Panel

Panel frame Housing

34 16

13.6 6.4

Upper case

Upper case

131.5

52.6

recycling cost and environmental impact of the EOL recycling processes. (3) Implementation of an LED backlight source: When EOL recycling processes for the LED panel are executed under bi-criteria optimization, the recycling processes remain the same as those for the original panel module. Thus, the LED panel notebook has lower costs and environmental impact in terms of EOL recycling processes than the original CCFL panel notebook. In summary, the EOL recycling processes for the notebook computer positively impact economic performance, and these processes become increasingly environmentally friendly when new product design (ie, green design) or new materials for different modules are used. A product designer can positively affect DfR and design for environment (DfE) for the environment by using appropriate economic and environmental criteria. In other words, new designs help reduce EOL recycling costs and significantly reduce environmental impact.

Original parts weight (g)

New parts weight (g) 45

8. Conclusions Owing to increased focus on extended producer responsibilities, many producers of electrical and electronic products have become interested in identifying the best EOL recycling processes that maximize benefit and minimize environmental impact. This study uses the characteristics of EOL product recycling and applies a novel heuristic approach to identify the best EOL recycling processes for the bi-criteria optimization problem. One example characteristic is that, differing from mechanical products, less effort is needed to optimize the manual disassembly sequence for recycling electronic products. The characteristics of electronic products can be used to reduce the number of feasible solutions (ie, possible EOL processes) searched, thereby accelerating identification of Pareto solutions. A nine-step process based on the proposed heuristic approach is applied to identify all major feasible solutions and Pareto solutions. Compromise programming is then employed to determine the best recycling processes with

SC Lee and LH Shih—A novel heuristic approach

617

Table 5 EOL recycling process for new notebook designs Scenarios case

Original NB

The number of parts

Bi-criteria optimization (wi=1, p=2)

Silicone-based keyboard NB

Aluminium alloy case NB

The number of parts

Bi-criteria optimization (wi=1, p=2)

The number of parts

Bi-criteria optimization (wi=1, p=2)

LED panel NB

The number of parts

Bi-criteria optimization (wi=1, p=2)

Necessary disassembly sequence Upper Module EOL case recycling module process Keyboard module

15

Disassembly

15

Disassembly

15

Disassembly

12

Disassembly

8

Disassembly

8

Disassembly

8

Disassembly

8

Disassembly

12

4

Disassembly

12

13

13

Shredding

13

Part 1 disassembly, Part 2B12 shredding Part 1 disassembly, 2B13 shredding

12

Hard disk module PCB module

8

Part 1 disassembly, Part 2B12 shredding Part 1 disassembly, Part 2B13 shredding Part 1B6 disassembly, Part 7B8 shredding

8

8

Part 1B6 disassembly, Part 7B8 shredding

8

Under case module Panel module

6

Disassembly

6

Part 1B6 disassembly, Part 7B8 shredding Disassembly

Part 1 disassembly, Part 2B12 shredding Part 1 disassembly, Part 2B13 shredding Part 1B6 disassembly, Part 7B8 shredding

6

Disassembly

6

Disassembly

7

Disassembly

7

Disassembly

7

Disassembly

9

Disassembly

Benefit (NTD/ set) Environmental impart (mPt/set)

13

14.8788

14.0404

5.3246

14.2086

275.6076

408.3138

392.6182

302.4995

decision-maker involvement. The best processes consisting of manual disassembly, shredding for material recovery, and waste disposal were obtained using the proposed heuristic approach. A notebook was used as an illustrative example to demonstrate the potential of the proposed approach. A Pareto boundary for identifying best recycling processes when economic and environmental criterions are involved was also presented and discussed. Using the proposed approach, the best recycling processes under bicriteria can be identified easily when the product adopts a new design or new materials. In addition, this study builds upon the authors’ previous work (Shih and Lee, 2007), extending it to include a larger case study, the use of compromise programming, and product design changes. Compromise programming is a common approach for solving multi-criteria decision-making problems in the operations research field, and can use different weights and calculate the shortest distance from all Pareto solutions to the best solution. Hence, this study suggests adopting compromise programming to find the best solution, that is, the compromise solution, and evaluates the economic and environmental performance of feasible solutions. In addition, before employing compromise programming, the proposed heuristic approach is utilized

to identify all feasible solutions. Compromise programming is then applied to find the near-ideal solution. The EU imports many Taiwanese electronics products, and thus Taiwan producers must face the ripple effects of the WEEE directive. For instance, Asus’ recycling plants in Europe and Germany have established recycling information management platforms, which monitor recycling progress and control recycle information in real time. Asus also takes recycling responsibility with its distributors collectively in the EU to adhere to the WEEE directive requirements. However, globally renowned brand-name companies are anxious to know how their EOL products will be recycled in the EU, as well as the benefit and environmental impact of doing so. In addition, how do recycling systems differ in various areas of the EU? What is a reasonable cost for contracting a European agent to handle EOL products? In the long run, the key question becomes ‘If the product is made greener, how will its benefits and environmental impact change?’ This study provides globally renowned brand-name companies with a simple and reliable method to answer the above questions. Process-related information adopted in a target area and the incurred costs are especially useful to brand-name companies outside of Europe, allowing them to make

618

Journal of the Operational Research Society Vol. 63, No. 5

better decisions on EOL strategies. Identifying the best EOL process for business executives implies that EOL recyclers work to maximize profits. As many local factors affect the EOL process, for example, labour costs, market prices for reclaimed materials and costs of treating hazardous materials, the proposed heuristic approach for identifying the best EOL process can provide different cost/ profit parameters for decision-makers. In addition, in response to the WEEE directive, producers should provide the basic information of the disassembly report, recycling cost and environmental impact for EOL products, which can also be calculated by using the proposed heuristic approach. The proposed heuristic approach is applicable for analyzing electronic products, such as notebook and LCD monitors, provided that a simple and reliable method is available that can rapidly evaluate feasible EOL processes for producers and that the best EOL strategy can be applied correctly. However, as is well known, optimizing EOL process is extremely complex. Therefore, despite its contributions, this study has several limitations. First, one electronic product, notebook, is utilized for the empirical analysis due to the data unavailability. Data for related parameters in the proposed approach, collected as labouratory data, can influence analytical reliability. However, data are often beyond the control of a researcher. The accuracy and reliability of data also affect the accuracy and application of the analyses in this study. Second, based on the assumptions of electronic product recycling, the proposed heuristic approach is only applied to electronic products. Third, the proposed approach is applied to identify the best EOL process based on assumptions of electronic product recycling while considering both economic and environmental criteria, that is, this is a bi-criteria optimization problem. Resolving these limitations is a viable avenue for further research and model improvement, such as using different elements for cost and profit estimations and environmental impacts, and using more than two criteria (multi-criteria). In addition to providing a valuable reference for EOL planning decisions, the proposed approach contributes significantly to optimal EOL strategy planning.

References Andre´s C, Lozano S and Adenso-Dı´ az B (2007). Disassembly sequence planning in a disassembly cell context. Robot ComputIntegr Manuf 23: 690–695. Bakal IS and Akcali E (2006). Effects of random yield in remanufacturing with price-sensitive supply and demand. Prod Opns Mngt 15: 407–420. Bufardi A, Sakara D, Gheorghe R, Kiritsis D and Xirouchakis P (2003). Multiple criteria decision aid for selecting the best product end of life scenario. Int J Comput Integr Manuf 16: 526–534.

Chan JWK and Tong TKL (2007). Multi-criteria material selections and end-of-life product strategy: Grey relational analysis approach. Mater Des 28: 1539–1546. Dong J and Arndt G (2003). A review of current research on disassembly sequence generation and computer aided design for disassembly. Proc Inst Mech Eng Part B-J Eng Manuf 217: 299–312. Dong J, Gibson P and Arndt G (2007). Disassembly sequence generation in recycling based on parts accessibility and endof-life strategy. Proc Inst Mech Eng Part B-J Eng Manuf 221: 1079–1085. Effting S (2006). Wizard manual-Introduction into programming of wizards in SimaPro 7. Version 1.2. PRe´ Consultants B. V. Feldmann K, Trautner S, Lohrmann H and Melzer K (2001). Computer-based product structure analysis for technical goods regarding optimal end-of-life strategies. Proc Inst Mech Eng Part B-J Eng Manuf 215: 683–693. Feldmann K, Trautner S and Meedt O (1999). Innovative disassembly strategies based on flexible partial destructive tools. Annu Rev Control 23: 159–164. Ferguson ME and Toktay LB (2006). The effect of competition on recovery strategies. Prod Opns Mngt 15: 351–368. Ferna´ndez E, Kalcsics J, Nickel S and Rı´ os-Mercado RZ (2010). A novel maximum dispersion territory design model arising in the implementation of the WEEE-directive. J Opl Res Soc 61: 503–514. Flowers AD and Linderman K (2003). Hazardous waste disposal: A waste-fuel blending approach. Prod Opns Mngt 12: 307–319. Galantucci LM, Percoco G and Spina R (2004). Assembly and disassembly planning by using fuzzy logic & genetic algorithms. Int J Adv Robot Syst 1: 67–74. Gao M, Zhou MC and Tang Y (2004). Intelligent decision making in disassembly process based on fuzzy reasoning Petri nets. IEEE Trans Syst Man Cybern Part B-Cybern 34: 2029–2034. Gonza´lez B and Adenso-Diaz B (2006). A scatter search approach to the optimum disassembly sequence problem. Comput Opl Res 33: 1776–1793. Gungor A and Gupta SM (1997). An evaluation methodology for disassembly processes. Comput Ind Eng 33: 329–332. Gungor A and Gupta SM (2001). Disassembly sequence plan generation using a branch-and-bound algorithm. Int J Prod Res 39: 481–509. Jien CK (2008). Multi-criteria optimal process for recycling EOL laptops and its decision support system. MS thesis, Department of Resources Engineering, National Cheng Kung University, Taiwan, ROC. Karakayali I, Emir-Farinas H and Akcali E (2007). An analysis of decentralized collection and processing of end-of-life products. J Opns Mngt 25: 1161–1183. Kongar E and Gupta SM (2002). A genetic algorithm for disassembly process planning. Proc SPIE 4569: 54–62. Kongar E and Gupta SM (2006). Disassembly to order system under uncertainty. Int J Mngt Sci 34: 550–561. Kuo TC, Zhang HC and Huang SH (2000). Disassembly analysis for electromechanical products: A graph-based heuristic approach. Int J Prod Res 38: 993–1007. Lambert AJD (2003). Disassembly sequencing: A survey. Int J Prod Res 41: 3721–3759. Lambert AJD (2007). Optimizing disassembly processes subjected to sequence-dependent cost. Comput Opl Res 34: 536–551. Lambert AJD and Gupta SM (2002). Demand-driven disassembly optimization for electronic products package reliability. J Electron Manuf 11: 121–135. Lee D-H and Xiroucharkis P (2004). A two-stage heuristic for disassembly scheduling with assembly product structure. J Opl Res Soc 55: 287–297.

SC Lee and LH Shih—A novel heuristic approach

Lee SC (2006). Multi-criteria optimization for recycling process for EOL LCD type products. MS thesis, Department of Resources Engineering, National Cheng Kung University, Taiwan, ROC. Lee SG, Lye SW and Khoo MK (2001). A multi-objective methodology for evaluating product end-of-life options and disassembly. Int J Adv Manuf Technol 18: 148–156. Liu ZF, Liu XP, Wang SW and Liu GF (2002). Recycling strategy and a recyclability assessment model based on an artificial neural network. J Mater Process Technol 129: 500–506. Mazhar MI, Kara S and Kaebernick H (2007). Remaining life estimation of used components in consumer products: life cycle data analysis by Weibull and artificial neural networks. J Opns Mngt 25: 1184–1193. Melnyk SA, Sroufe RP and Calantone R (2003). Assessing the impact of environmental management systems on corporate and environmental performance. J Opns Mngt 21: 329–351. Mergias I, Moustakas K, Papadopoulos A and Loizidou M (2007). Multi-criteria decision aid approach for the selection of the best compromise management scheme for ELVs: The case of Cyprus. J Hazard Mater 147: 706–717. Navin-Chandra D (1994). The recovery problem in product design. J Eng Des 5: 65–86. Neuendorf K-P, Lee D-H, Kiritsis D and Xirouchakis P (2001). Disassembly scheduling with parts commonality using Petri nets with timestamps. Fundam Inform 47: 295–306. Pu P (1992). An assembly sequence generation algorithm using case-based search techniques. Proceedings of IEEE International Conference on Robotics and Automation, 12–14 May 1992, Vol. 3 pp 2425–2430, doi:10.1109/ROBOT.1992.220101. Ravi V, Shankar R and Tiwan MK (2008). Selection of a reverse logistics project for end-of-life computers: ANP and goal programming approach. Int J Prod Res 46: 4849–4870. Rosell J, Mun˜oz N and Gambin A (2003). Robot tasks sequence planning using Petri nets. Proceedings of the IEEE International Symposium on Assembly and Task Planning, 10–11 July 2003, pp 24–29, doi:10.1109/ISATP.2003.1217182.

619

Shih LH, Chang YS and Lin YT (2006). Intelligent evaluation approach for electronic product recycling via case-based reasoning. Adv Eng Inform 20: 137–145. Shih LH and Lee SC (2007). Optimizing disassembly and recycling process for EOL LCD-type products: A heuristic method. IEEE Trans Electron Packag Manuf 30: 213–220. Staikos T and Rahimifard S (2007). An end-of-life decision support tool for product recovery considerations in the footwear industry. Int J Comput Integr Manuf 20: 602–615. Swaminathan A and Barber KS (1996). An experience-based assembly sequence planner for mechanical assemblies. IEEE Trans Robot Autom 12: 252–267. Tang Y and Turowski M (2007). Adaptive fuzzy system for disassembly process planning with uncertainty. J Chin Inst Chem Eng 24: 20–29. Veerakamolmal P and Gupta SM (2002). A case-based reasoning approach for automating disassembly process planning. J Intell Manuf 13: 47–60. Waterbury R (1985). Designing parts for automated assembly. Assem Eng 24: 211–217. Webster S and Mitra S (2007). Competitive strategy in remanufacturing and the impact of take-back laws. J Opns Mngt 25: 1123–1140. Wen LC (2000). A Study of Waste Recycling-and-treatment Fee Collection for Scrap Personal Computers. Environmental Protection Administration, Taiwan, ROC. Yu PL and Leitmann G (1974). Compromise solutions, domination structures and Salukvadze’s solution. J Optim Theory Appl 13: 362–378. Zussman E, Kriwet A and Seliger G (1994). Disassembly-oriented assessment methodology to support design for recycling. CIRP Ann-Manuf Technol 43: 9–14.

Received February 2010; accepted March 2011