Rule-based recursive selective disassembly sequence planning for green design

Advanced Engineering Informatics 25 (2011) 77–87

Contents lists available at ScienceDirect

Advanced Engineering Informatics journal homepage: www.elsevier.com/locate/aei

Rule-based recursive selective disassembly sequence planning for green design Shana S. Smith *, Wei-Hsiang Chen Department of Mechanical Engineering, National Taiwan University, Taiwan, ROC

a r t i c l e

i n f o

Article history: Received 3 November 2009 Received in revised form 24 March 2010 Accepted 30 March 2010

Keywords: Selective disassembly sequence planning Green design Rule-based Recursive

a b s t r a c t Disassembly sequence planning not only reduces product lifecycle cost, but also greatly influences environmental impact. Many prior green design research studies have focused on complete disassembly of an end-of-life product to recover valuable components. However, complete disassembly is often not practical or cost effective if only a few components will be recovered and recycled from a given product. Selective disassembly sequence planning focuses on disassembling only one or more selected components from a product for reuse, recycling, remanufacturing, and maintenance. This paper presents a rule-based recursive method for finding a near-optimal heuristic selective disassembly sequence for green design. Most prior methods either enumerate all solutions or use stochastic random methods to generate solutions. Enumerative or stochastic methods often require tremendous computational resources while, at the same time, they often fail to find realistic or optimal solutions. On the contrary, the proposed method establishes certain heuristic disassembly rules to eliminate uncommon or unrealistic solutions. In addition, rather than considering geometric constraints for each pair of components, the developed method only considers geometric relationships between a part and its neighboring parts. As a result, the developed method can effectively find a near-optimal heuristic solution while greatly reducing computational time and space. Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction Due to rising environmental concerns and the need to reduce waste, manufacturers are beginning to take back their products for ‘‘recovering” [1]. A ‘‘green” product can be characterized by several factors, and design for disassembly is one of the key factors. Usually, disassembly is the first stage in the process of recovering a product. In particular, design for manufacturing and design for recycling are based on design for disassembly [2]. Seo et al. [3] showed that product disassemblability has a strong correlation with product life cycle cost. Pnueli and Zussman [4] showed that only 10–20% of product recycling cost depends upon the recycling process, the rest is determined during the product design stage. Prior research also indicated that disassembly is the last and most important process before the added-value recovery operations [5]. Thus, product dissassemblability affects the overall value and sustainability of the products [2]. Disassembly sequence planning aims to find efficient sequences for disassembling a product. Both product design and choice of disassembly sequence for material recovery affect a given product’s end-of-life value. Products that can be rapidly disassembled into component parts can be more easily remanufactured and recycled. An efficient disassembly sequence should minimize disassembly * Corresponding author. Address: No.1, Sec. 4, Roosevelt Road, Taipei, Taiwan, ROC. E-mail address: [email protected] (S.S. Smith). 1474-0346/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.aei.2010.03.002

time and both the number of disassembly steps and tools used. Many prior green design research studies have focused on complete disassembly of an end-of-life product to recover valuable components. However, complete disassembly is often not practical or cost effective if only a few components will be recovered and recycled from a given product. Selective disassembly sequence planning focuses on disassembling only one or more selected components from a product for reuse, recycling, remanufacturing, and maintenance. Thus, compared with complete disassembly, selective disassembly is a more powerful and a more efficient tool for solving de-manufacturing (DM) problems. DM involves separating certain components and materials from a product for reuse, recycling, replacement, and maintenance to reduce product lifecycle cost. Although disassembly is crucial for successful product recovery, finding an optimal selective disassembly sequence is a very difficult and complex problem when multiple factors are involved, e.g., disassembly time, cost, reorientations, tools, and environmental impacts. In addition, if a product has n parts, it will have n! possible disassembly solutions. The goals of this paper are to develop a user-friendly disassembly method and tool for quickly finding an efficient selective disassembly sequence for a product and, thereby, to increase products’ sustainability and end-of-life value, to reduce the waste sent to landfills or incinerators (by increasing product disassemblability, recyclability, and maintainability), and to reduce product life cycle cost.

78

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

2. Literature review Some prior studies have utilized advanced searching algorithms to find optimal selective disassembly sequences. For example, Shyamsundar and Gadh developed a recursive method that considers both separation direction and disassembly direction to disassemble a target component [6]. They used a Gaussian Sphere to find the range of disassembly directions and separation directions. Since their method allows a component to be disassembled from any direction, the input information required is relatively complicated. Their method also requires input information concerning sequential disassembly and additional disassembly constraints, which make their method difficult to implement for general products. Srinivasan et al. improved upon Shyamsundar and Gadh’s work by applying a wave propagation method to solve selective disassembly problems [7–9]. Their method includes disassembling one component or more components, as well as total selective disassembly. However, they only evaluate each disassembly sequence by the total number of removed components. They consider the optimal selective disassembly sequence to be the sequence with the minimum number of removals. Although their evaluation function is simple, it might not satisfy the demands of many realistic product design problems. In addition, the analyses of their method are usually time-consuming and expensive. Thus, it is difficult for the wave propagation-based methods to generate an efficient and optimal sequence for selective disassembly [10]. Srinivasan and Gadh also developed a global selective disassembly method, which includes spatial constraints and user-defined constraints [11]. Kara et al. reversed and modified assembly sequences and used a liaison diagram to evaluate geometric connections to find selective disassembly sequences [12,13]. Their method requires extensive computational resources to generate a sequence diagram and to delete unfeasible sequences. Aguinaga et al. used a rapidgrowing random tree method to find optimal sequences [14,15]. However, their method generates many paths, and, thus, it takes a significant amount of time to find solutions. Using their method, neither the execution time nor the results are guaranteed to be constant. In addition, the resulting sequence is not optimal, and usually contains a component of noise. Most research discussed above solves selective disassembly problems by considering all geometric constraints and evaluating each selective disassembly sequence to find optimal solutions. On the contrary, some researchers used stochastic random search methods to simplify the searching process, for example, ant colony optimization (ACO) algorithms and genetic algorithms (GAs) [10,16–19]. Stochastic random search methods find near-optimal solutions by analyzing a small portion of the candidate solutions. Thus, they reduce a significant amount of searching time. Most selective disassembly methods use specific information in their searching processes: geometric constraints [15,11,17,19–22], topological locations [10,11,21,22], liaison relationships [13], AND/ OR graphs [15,20], precedence graphs [20], fastener accessibility [23], and component accessibility [11]. To reduce disassembly cost, most selective disassembly sequence planning methods focus on minimizing the removal of components [10,11,14,16,17,20–22], disassembly time [12,13,15,18,24], tool changes [10,18,19], and reorientation times [16,18,19]. However, finding an optimal selective disassembly sequence is a difficult problem. Most prior methods either enumerate all solutions or use stochastic methods to generate random solutions. Methods which enumerate all solutions might require a tremendous amount of computational resources. However, stochastic random methods, such as ACO and GAs, might generate solutions which are uncommon or unrealistic for use in reality [10,16–19,25,26].

To deal with the problems of requiring tremendous amount of computational resources or generating uncommon or unrealistic solutions, this paper presents a rule-based recursive method for obtaining near-optimal heuristic selective disassembly sequences for green design. The method uses certain disassembly rules to eliminate uncommon or unrealistic solutions. In addition, rather than consider geometric constraints for each pair of components, the developed method only considers the geometric relationship between a part and its neighboring parts. If a part can be disassembled, its geometric relationships with the neighboring parts will be dynamically updated. The topological information and part accessibility of a product is examined from inside to outside. As a result, the developed method can effectively find a near-optimal heuristic solution while greatly reducing computational time and space. Our method can handle 3D single-component and multiple-component disassembly problems. The evaluation criteria include disassembly reorientations and number of removed fasteners and components.

3. Disassembly model A disassembly model is a mathematical model representing the spatial relationships between components of a product. A product needs to be translated into a mathematical disassembly model in order to carry out subsequent disassembly analysis and planning. In this section, four matrices (disassembly matrix for fasteners, disassembly matrix for components, motion constraints for components, and motion constraints for fasteners) are defined to describe the geometric relationships between parts and, thus, build a disassembly model for a product. Contrary to most prior work which does not consider fasteners, the developed method considers fastener constraints. In addition, the unique matrix designs for the motion constraints for components and motion constraints for fasteners enable the rule-based recursive selective disassembly planning to work effectively. In this paper, components are the elements providing functions in a product. Fasteners are the elements connecting components together. Parts can be either components or fasteners. 3.1. Disassembly matrix for fasteners We define a disassembly matrix for fasteners, DF, which records the disassembly directions of each fastener. Since most disassembly process occurs along the principle directions, DF includes disas-

Fig. 1. Example assembly 1.

79

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

sembly information in the +x, x, +y, y, +z, and z directions. If a fastener can be disassembled out of an assembly along its axial direction without any collision, the corresponding tuple in DF is set to be 0; otherwise, it is set to be 1. If a fastener is welded to a component, it becomes an integral part of that component. Thus, in this study, fasteners will not have parameters like (+x:x:+y: y:+z:z) = (1:1:1:1:1:1). For ease of illustration, a 2D example is used here. In Fig. 1, there are four fasteners, 5, 6, 7, and 8, and four components, 1, 2, 3, and 4. Fastener 5 can only be disassembled along its axial direction +y. Thus, DF5 (+x:x:+y:y) = (1:1:0:1). Likewise, DF6 (+x:x:+y:y) = (1:0:1:1), DF7 (+x:x:+y:y) = (0:1:1:1), and DF8 (+x:x:+y:y) = (1:1:0:1). Finally, DF = [DF5 DF6 DF7 DF8]T:

3 3 2 1 1 0 1 DF 5 6 DF 7 6 1 0 1 1 7 7 6 67 6 DF ¼ 6 7: 7¼6 4 DF 7 5 4 0 1 1 1 5 2

Fig. 2. Example assembly 2.

DF 8

1 1

0

1

3.2. Disassembly matrix for components We define a disassembly matrix for components, DC, which records the immediately touching components and fasteners which constrain the disassembly of a target component in one direction along a principle axis. Parts constrain the disassembly of a target component in both directions of a principle axis will be considered in the next session. For example, in Fig. 1, DC1 (+x: x: +y:y) = (0:0:5,8:2,4). Likewise, DC2 (+x:x:+y: y) = (3:6:1,5:0), DC3 (+x:x:+y:y) = (4,7:2,6:0:0), and DC4 (+x:x:+y:y) = (7:3:1,8:0). Finally, DC = [DC1 DC2 DC3 DC4]T:

Fig. 3. Example assembly 3.

Start

Retrieve a query component, n

Is component n fixed by any fasteners? N

Rule 5

Y

Is component n constrained by any components?

N

Y

Rule 1

Retrieve a query fastener, i

Is fastener i constrainedb y any fasteners?

Y

Rule 3

Rule 4 N Is fastener i constrainedb y any components?

Y

Rule 3

N Rule 2 Fig. 4. Flowchart of the rule-based recursive selective disassembly planning.

80

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

3 3 2 0 0 5; 8 2; 4 DC 1 6 DC 7 6 3 6 1; 5 0 7 7 6 27 6 DC ¼ 6 7 7¼6 4 DC 3 5 4 4; 7 2; 6 0 0 5 2

DC 4

7

3

1; 8

0

3.3. Motion constraint matrices

Fig. 5. Case study 1.

Fig. 6. The DC and MC matrices for the product in Fig. 5.

We define two motion constraint matrices; one records motion constraints for fasteners (MF), and the other records motion constraints for components (MC). Before MF and MC are defined, ‘‘first-level parts” need to be defined. First-level parts are parts which do not immediately touch a query component or fastener but which are the first parts beyond the immediately touching parts which would block movement of the query component or fastener in given moving directions. Query parts are components or fasteners which are currently under disassemblability checking. The MF matrix records both the first-level parts of each fastener and any immediately touching parts of the fastener in a given disassembly direction. For example, in Fig. 2, there are six fasteners, 7, 8, 9, 10, 11 and 12, and six components, 1, 2, 3, 4, 5, and 6. Fastener 12 can only be disassembled along the +y direction. However, since component 1 is the first component which fastener 12 would collide with, in the given disassembly direction, component 1 is a first-level component of fastener 12. Thus, MF 12 (+x:x:+y: y) = (0:0:1:0). Recall that the disassembly matrix for components, DC, only records immediately touching components and fasteners which constrain the disassembly of a query component in one direction along a principle axis. In contrast, MC records the first-level parts of a query component and immediately touching components and fasteners which constrain the disassembly of a query component in both directions of a principle axis. For example, in Fig. 2, component 5 is a first-level part of component 6, and component 6 is constrained by fastener 12 in both directions of the x-axis, while component 6 is only constrained by component 3 in one direction of the y-axis. Thus, MC6 (+x:x:+y:-y) = (12:12:5:0).

Fig. 7. User interface.

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

Fig. 8. Inputs for DC, MC, DF, MF.

Fig. 9. Exploded view of power brake [27].

81

82

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

Fig. 3 shows another example. The assembly in Fig. 3 includes two components and one fastener. For the given assembly, DC1 (+x:–x:+y:-y) = (0:0:2,3:0), MC1 (+x:–x:+y:-y) = (2,3:2,3:0:0), DC 2 (+x:–x:+y:-y) = (0:0:3:1), and MC2 (+x:–x:+y:-y) = (1,3:1,3:0:0).

4. Selective disassembly sequence planning The developed approach is applicable to multi-layer products, three principle disassembly directions, and any non-destructible fasteners. We establish five rules which define our recursive selective disassembly planning process. The rules are developed based on the analysis of the disassembly characteristics, and the corresponding unique matrix representations which are developed in the previous section. Here, a parent part is a part which has already been selected for disassemblability checking.  Rule 1: IF (there are any fasteners attached to query component n in DCn) THEN (all the fasteners need to be disassembled first)  Rule 2: IF (there are corresponding tuples which are 0 in both DFi and MFi) THEN (disassemble fastener i along the directions associated with the tuples and update DC, MC, and MF matrices by eliminating fastener i from the matrices)  Rule 3: IF (there are tuples in DFi which are 0 but the corresponding tuples in MFi are not 0) THEN (disassemble the parts, which are not the parent parts of fastener i, in the corresponding tuples in MFi, before disassembling fastener i)  Rule 4: IF (there are corresponding tuples which are 0 in both DCn and MCn) THEN (disassemble component n along the directions associated with the tuples and update DC, MC, and MF matrices by eliminating component n from the matrices)  Rule 5: IF (there are no corresponding tuples which are 0 in both DCn and MCn) THEN (disassemble the parts which are in DCn and the first-level parts which are in MCn, but not the parent parts of n, before disassembling component n) The searching process first checks if query component n is fixed by any fasteners. If so, all the fasteners need to be disassembled, before disassembling component n, according to Rule 1. If fastener i is not fixed by any fasteners or components, it can be disassembled from the assembly, and DC, MC, and MF need to be updated by eliminating fastener i from the matrices, according to Rule 2. If fastener i is constrained by any other fasteners or components, all the fasteners and components need to be disassembled first, according to Rule 3. If component n is not fixed by any fasteners or any components, it can be disassembled from the assembly, and DC, MC, and MF need to be updated by eliminating component n from the matrices, according to Rule 4. If component n is not fixed by any fasteners but is constrained by some other components, all the components need to be disassembled first, according to Rule 5. Fig. 4 shows a flowchart of the searching process. The selective disassembly planning method is recursive since Rules 3 and 5 recursively remove components and fasteners. The first input query component is the target component which will finally be reached and disassembled by the recursive disassembly process. For example, in Fig. 2, if component 2 is a target component, according to Rule 1, fasteners 7 and 8 need to be disassembled first. Since DF7 (+y) = MF7 (+y) = 0, according to Rule 2, fastener 7 can be disassembled in the +y direction, and DC, MC, and MF will be updated by eliminating 7 from them. Thus, DC2 (3,5,6:8:1,7:0) will be updated to DC2 (3,5,6:8:1:0). Likewise, after fastener 8 is disassembled in the –x direction, DC2 will be updated to DC2 (3,5,6:0:1:0). After fasteners 7 and 8 are disassembled, DC2 (x) = MC2 (x) = 0, and DC2 (y) = MC2 (y) = 0. Thus, according

to Rule 4, component 2 can be disassembled in the x or y directions, and it will be eliminated from DC, MC, and MF. The process is rule-based and recursive. The given rules reduce searching time by eliminating unrealistic and uncommon solutions. Thus, not all possible solutions are generated and checked. Experiments show that the method can generate reasonable and near-optimal heuristic solutions efficiently.

5. Cost function Our cost function for evaluating disassembly sequences includes disassembly reorientations and number of components and fasteners removed.

Cost value ¼ w1  reorientations þ w2  parts

ð1Þ

In Eq. (1), we can choose weight values w1 and w2 to establish the weighted importance of each of the cost parameters in determining the outcome of the search process. The final best or optimal selective disassembly sequence which is found has the lowest cost value.

5.1. Reorientations During a disassembly process, if the number of reorientations for disassembly directions is reduced, the total disassembly time is also reduced [16–19,25,26]. Since we only consider principle disassembly directions, each reorientation requires either a 90-deg or a 180-deg change. If the reorientation requires a 90-deg change, we increase the reorientation cost by 1. If the disassembly direction requires a 180-deg change, we increase the reorientation cost by 2. When no reorientations are needed, we set the reorientations cost to 0.

Table 1 Material list of the power brake in Fig. 9. Index No.

Component (C) or Fastener (f)

Nomenclature

Units per Assy.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

C C C C C C C C C f f f f f f f f f f f f f f f f f f f

1 1 1 2 2 2 8 2 1 10 10 10 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

29 30

f f

Housing Cover Gasket Seat Packing – neoprene Piston Packing – neoprene Link Lever – assembly Stud 1=4 ” Nut Washer Washer Spring Ball 1/4” Dia. Spring Pin Spacer 11/2 Dia. Nut 1 1/8 Pin 3/8” Dia. Spacer 1 1/8” Nut 7/8” 14 NF Capnut Washer Nut 10–32 NF Screw 1–32 NF Shaft 5/8” Nickel Steel Shaft 9/11” Dia. nickel steel Nut 5/16 Screw 3/8 Dia. 5/16

2 2

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

83

5.2. Parts

6.1. Example 1

Most prior methods include the number of removed components in their cost functions for evaluating the quality of a disassembly sequence [9,11,14,16,17,20–22]. Here, the number of removed components and fasteners are both considered. If fewer parts are removed to disassemble a target component, less time is required and the cost of the disassembly sequence is lower.

Fig. 5 shows the first example. There is no fastener in this example. Fig. 6 shows the corresponding DC and MC matrices. Fig. 7 is the user interface of the software tool. Fig. 8 shows the inputs for DC, MC, DF, and MF. The target component, or the first query component, is component 3. One disassembly sequence, 5, 4, 3, is generated, and the cost value is 3, as shown in Fig. 7. For the given example, the weight for reorientation is much higher than the weight for part number. The assembly in Fig. 5 has only five components. If an enumeration method is used, there are 5! = 120 possible disassembly sequences. If a stochastic searching method is used, some

6. Case study Three examples are used to demonstrate our rule-based recursive selective disassembly method.

Fig. 10. The DC, MC, DF, and MF matrices for the power brake.

84

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

Fig. 11. Selective disassembly sequence for the power brake.

unrealistic or uncommon solutions might be generated. However, our rule-based recursive selective disassembly planning eliminates

Table 2 The selective disassembly sequence planning of power brake by Mascle and Hong [27]. Wave Wave Wave Wave Wave Wave Wave Wave Wave Wave Wave Wave

No. No. No. No. No. No. No. No. No. No. No. No.

1 2 3 4 5 6 7 8 9 10 11 12

10 11 12 2 3, 13, 15 14 4 16 17, 20 18 5 6

many unrealistic or uncommon solutions and finds near-optimal heuristic selective disassembly sequences quickly and effectively. 6.2. Example 2 Fig. 9 shows a 3D example of a power brake, given by Mascle and Zhao [27]. The material list is given in Table 1. Since the numbering system in our method places components before fasteners, we renumbered the parts in the table. However, the content of the parts are the same as Mascle and Zhao’s. We use reference [28] to define fasteners and components. Fig. 10 shows the inputs for DC, MC, DF, and MF. In this example, component 6 is the target component. A selective disassembly sequence planning for disassembling component 6 is found, as shown in Fig. 11, and it is 11, 12, 2, 15, 14, 13, 3, 4, 16, 17, 18, 20, 5, 6. There are fourteen parts to be disassembled and two reorientations, 18 to 20 and 20 to 5.

Fig. 12. Gear reducer assembly [7].

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

According to equation 1, cost value = w1 * Reorientations + w2 * Parts = 2 + 14 = 16. The cost value of the result by Mascle and Zhao [27] is 2 + 15 = 17, as shown in Table 2. The sequence found by Mascle and Zhao needs to remove fastener 10 first. However, fastener 10 cannot be removed unless fasteners 11 and 12 are removed first. In addition, after fasteners 11 and 12 are removed, component 2 can actually be removed without removing fastener 10. Therefore, our method provides a better and more reliable solution.

85

6.3. Example 3 A gear reducer assembly from Srinivasan and Gadh [7], as shown in Fig. 12, is used to test for single-target-component and multiple-target-component disassembly. Similarly, since the numbering system in our method places components before fasteners, we renumbered the parts. However, the content of the parts are the same as Srinivasan and Gadh’s. TheDC, MC, DF, and MF matrices are shown in Fig. 13.

Fig. 13. The DC, MC, DF, and MF matrices for the gear reducer assembly.

86

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

Fig. 14. Disassembly sequence for the gear reducer assembly.

6.3.1. Single-target component One disassembly sequence, 12, 13, 14, 1, 2, 3, 4, 5, is found, as shown in Fig. 14. The disassembly sequence includes three fasteners and five components. Since Srinivasan and Gadh [7] did not consider reorientations in their study, in order to compare with their results, we set 0 for the reorientation parameter in the cost function. The cost value for the sequence is: w1 * Reorientations + w2 * Parts = 0 + 8 = 8. For single-component disassembly, Srinivasan and Gadh [7] chose component 3 as the target component. However, they did not consider disassembly of fasteners. In their case, sequence 1, 2, 3 is the obvious best disassembly sequence solution. 6.3.2. Multiple-target components In the multiple-target components example, components 5 and 7 are the target components. Two situations are considered; one situation is to disassemble component 5 first and component 7 later, and the other situation is to disassemble component 7 first and component 5 later. If component 5 is the first target component, the best selective disassembly sequence is: 12, 13, 14, 1, 2, 3, 4, 5, 6, 7. There are three fasteners and seven components. The cost value = w 1 * Reorientations + w 2 * Parts = 0 + 10 = 10. Similarly, if component 7 is the first target component, the selective disassembly sequence is 15, 16, 17, 18, 11, 10, 9, 8, 7, 6, 5. The cost value is 11. Srinivasan and Gadh [7] only evaluated selective disassembly sequences by the number of removed components. They did not consider the number of fasteners or the number of reorientations. For multiple-component selective disassembly, Srinivasan and Gadh determined that 1, 2, 3, 4, 5, 6, 7 and 11, 10, 9, 8, 7, 6, 5 are two equivalent disassembly sequences for removing components 5 and 7. However, if we consider the number of removed fasteners, we find that sequence 12, 13, 14, 1, 2, 3, 4, 5, 6, 7 has a lower cost value. 7. Conclusions and future work In this paper, a rule-based recursive selective disassembly sequence planning method is presented to increase products’ disassemblability, recyclability, and maintainability. The method is based upon four matrices and five disassembly rules. The rules

are developed based on the analysis of the disassembly characteristics, and the corresponding unique matrix representations. With the given search rules, the method can eliminate unrealistic and uncommon disassembly sequences and find reasonable and nearoptimal heuristic selective disassembly sequences for complex assemblies effectively. The method can be used to solve selective disassembly sequence planning problems with only simple geometric and topological information. Only the geometric constraints of a part with its neighbouring parts are needed. Contrary to methods which consider geometric constraints between each pair of parts, the developed method greatly reduces required information storage space and searching complexity. The method can solve 3D single-target component and multiple-target component selective disassembly sequence planning problems. Compared to most existing methods, our method is much easier to implement for general products. Complete disassembly is often not practical or cost effective if only a few components will be disassembled from a given product. Using the developed method, a target component can be reached in an efficient way. Thus, it can reduce the waste sent to landfills or incinerators and increase products’ sustainability and end-of-life value. Since currently the four matrices are generated manually, human intervention is needed to determine all the geometric relationships. For a complex model, it might become a cumbersome and error-prone task. In the future, methods for generating the four matrices automatically from a CAD model will be developed. In addition, more rules can be added to the searching process to enhance the results for obtaining even better or desired solutions. Acknowledgement Support from the National Science Council, Taiwan (NSC 972221-E-002-158-MY2) is gratefully acknowledged. References [1] R. Zuidwijk, H. Krikke, Strategic response to EEE returns: product eco-design or new recovery processes?, European Journal of Operational Research 191 (3) (2008) 1206–1222 [2] J. Yi, B. Ji, Y. Guan, J. Dong, C. Li, Research on evaluation methodologies of product life cycle engineering design (LCED) and development of its tools

S.S. Smith, W.-H. Chen / Advanced Engineering Informatics 25 (2011) 77–87

[3]

[4]

[5] [6]

[7]

[8]

[9]

[10]

[11]

[12] [13]

[14]

[15]

International, Journal of Computer Integrated Manufacturing 21 (8) (2008) 923–942. K. Seo, J. Park, D. Jang, D. Wallace, Approximate estimation of the product life cycle cost using artificial neural networks in conceptual design, International Journal of Advanced Manufacturing Technology 19 (2002) 461–471. Y. Pnueli, E. Zussman, Evaluating the end-of-life value of a product and improving it by redesign, International Journal of Production Research 35 (4) (1997) 921–942. A. Xanthopoulos, E. Iakovou, On the optimal design of the disassembly and recovery processes, Waste Management 29 (5) (2009) 1702–1711. N. Shyamsundar, R. Gadh, Selective disassembly of virtual prototypes, Proceedings of the IEEE International Conference on Systems, Man and Cybernetics 4 (1996) 3159–3164. H. Srinivasan, R. Gadh, Selective disassembly: representation and comparative analysis of wave propagation abstractions in sequence planning, in: Proceedings of the 1999 IEEE International Symposium on Assembly and Task Planning Porto, Portugal, 1999, pp. 129–134. H. Srinivasan, R. Figueroa, R. Gadh, Selective disassembly for virtual prototyping as applied to de-manufacturing, Robotics and Computer Integrated Manufacturing 15 (3) (1999) 231–245. H. Srinivasan, R. Gadh, Efficient geometric disassembly of multiple components from an assembly using wave propagation, Jouranl of Mechanical Design 122 (2) (2000) 179–184. C. Chung, Q. Peng, Evolutionary sequence planning for selective disassembly in de-manufacturing, International Journal of Computer Integrated Manufacturing 19 (3) (2006) 278–286. H. Srinivasan, R. Gadh, A non-interfering selective disassembly sequence for components with geometric constraints, IIE Transactions 34 (4) (2002) 349– 361. S. Kara, P. Pornprasitpol, H. Kaebernick, A selective disassembly methodology for end-of-life products, Assembly Automation 25 (2) (2005) 124–134. S. Kara, P. Pornprasitpol, H. Kaebernick, Selective disassembly sequencing: a methodology for the disassembly of end-of-life products, Annals of the CIRP 55 (1) (2006) 37–40. I. Aguinaga, D. Borro, L. Matey, Automatic selective disassembly and path planning for the simulation of maintenance operations, IEEE Virtual Reality Conference 10–14 (2007) 283–284. I. Aguinaga, D. Borro, L. Matey, Parallel RRT-based path planning for selective disassembly planning, International Journal of Advanced Manufacturing Technology 36 (11–12) (2008) 1221–1233.

87

[16] J.F. Wang, J.H. Liu, S.Q. Li, Y.F. Zhong, Intelligent selective disassembly using the ant colony algorithm, Artificial Intelligence for Engineering Design, Analysis and Manufacturing 17 (2003) 325–333. [17] J.F. Wang, S.Q. Li, J.H. Liu, Selective disassembly planning for product green manufacturing, Computer Integrated Manufacturing Systems 13 (6) (2007) 1097–1102. [18] J.F. Xue, C.H. Qiu, D. Xiang, Planning of selective disassembly sequence based on ant colony optimization algorithm, Journal of Computer Aided Design & Computer Graphics 1119 (16) (2007) 742–747. [19] X.H. Zhang, S.Q. Li, J.F. Wang, X. Li, Single object selective disassembly sequence planning based on ant colony algorithm, Computer Integrated Manufacturing Systems 13 (6) (2007) 1110–1114. [20] M. A. García, A. Larré, B. López, A. Oller, Reducing the complexity of geometric selective disassembly, in: Proceedings of the 2000 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2000, pp. 1474–1479. [21] C. Mascle, B.A. Balasoiu, Algorithmic selection of a disassembly sequence of a component by a wave propagation method, Robotics and Computer Integrated Manufacturing 19 (5) (2003) 439–448. [22] J. Yi, B. Yu, L. Du, C. Li, D. Hu, Research on the selectable disassembly strategy of mechanical parts based on the generalized CAD model, International Journal of Advanced Manufacturing Technology 37 (2008) 599–604. [23] C. Chung, Q. Peng, An integrated approach to selective-disassembly sequence planning, Robotics and Computer-Integrated Manufacturing 21 (2005) 475– 485. [24] C. Chung, Q. Peng, A hybrid approach to selective-disassembly sequence planning for de-manufacturing and its implementation on the Internet, International Journal of Advanced Manufacturing Technology 30 (2006) 521–529. [25] S. Smith, Using multiple genetic operators to reduce premature convergence in genetic assembly planning, Computers in Industry 54 (1) (2004) 35–49. [26] G.C. Smith, S. Smith, An enhanced genetic algorithm for automated assembly planning, Robotics and Computer-Integrated Manufacturing 18 (5-6) (2002) 355–364. [27] C. Mascle, H.P. Zhao, Integrating environmental consciousness in product/ process development based on life-cycle thinking, International Journal of Production Economics 112 (2008) 5–17. [28] Fastener World Inc., , 2009 (retrieved 29.09.09).