Fuzzy-Petri-Net-Based Disassembly Planning ... - Semantic Scholar
718
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
Fuzzy-Petri-Net-Based Disassembly Planning Considering Human Factors Ying Tang, Member, IEEE, MengChu Zhou, Fellow, IEEE, and Meimei Gao, Member, IEEE
Abstract—Disassembly, as the process of systematic removal of desirable constituent parts from an assembly, is of growing importance due to the increasing environmental and economic pressure. Although disassembly in practice is manual and labor intensive, little attention has been paid to the human intervention in the disassembly process. This paper addresses this deficiency by developing a fuzzy attributed Petri net (FAPN) model to mathematically represent uncertainty in disassembly due to a large amount of human intervention. An algorithm based upon this model is further proposed for optimal disassembly planning with a view to making the technique more applicable to real industry settings. The benefit of the proposed model and algorithm is illustrated through the disassembly of a personal computer (PC) in a prototypical disassembly system. Index Terms—Disassembly planning, fuzzy logic, human factors, optimization, Petri nets.
I. I NTRODUCTION
D
ISASSEMBLY, as the process of systematic removal of desirable constituent parts from an assembly, is of growing importance due to increasing environmental and economic pressure. Considering disassembly as an expensive process, many researchers and industry executives have started to realize opportunities for performing disassembly in a cost-effective manner. Some methodologies are developed to identify the extent to which the disassembly of a product should be conducted to keep the process profitable and environmentally friendly [3], [10], [22]. Several heuristic approaches are proposed to generate the optimal disassembly sequence that minimizes the cost of disassembly (assuming that a certain level of disassembly is required) or obtains the best cost/benefit ratio for disassembly [5], [7], [11], [15], [17], [18]. A recent survey of disassembly modeling and planning is presented in [8] and [17]. While a significant amount of research is being conducted in this area, not much effort has been focused on human factors in disassembly. In fact, disassembly is currently Manuscript received April 27, 2004; revised August 31, 2004. This work was supported in part by the National Science Foundation under Grant 0097887, the National Natural Science Foundation of China under Grants 60228004 and 60334020, the Shandong Provincial Government under Grant 030335, and the Ministry of Science and Technology of China under Grant 2002CB312200. This paper was recommended by Associate Editor E. J. Bass. Y. Tang is with the Department of Electrical and Computer Engineering, Rowan University, Glassboro, NJ 08028 USA (e-mail: [email protected]). M. Zhou is with the Department of Electrical and Computer Engineering, New Jersey Institute of Technology, Newark, NJ 07102 USA, and also with the Laboratory of Complex Systems and the Intelligence Science Institute of Automation, Chinese Academy of Sciences, Beijing 100080, China (e-mail: [email protected]). M. Gao is with the Department of Mathematics and Computer Science, Seton Hall University, South Orange, NJ 07079 USA (e-mail: [email protected]). Digital Object Identifier 10.1109/TSMCA.2005.853508
manual/semiautomatic and labor intensive. The large amount of human intervention in disassembly provides extra flexibility, but it also gives rise to many uncertainties in the process. For instance, disassembly processing time, quality of disassembled components, and disassembly cost might change significantly due to varying skill levels of human operators. Moreover, these uncertainties contain fuzzy or imprecise information that are sometimes described by linguistic terms. There is a need to develop a formal model that allows us to analyze the influence of human intervention in disassembly and incorporate these nonprocess disturbances into disassembly planning and control. Human factors in manufacturing systems arise when designing and controlling the automated production and assembly [12]. A lesson learned from process automation is that, if human factors are not taken into consideration, even technologically state-of-the-art systems can be more problematic than beneficial [4]. Rahimi et al. raised several engineering issues as they concerned the human component of the hybrid system [13]. Fuzzy theory has been applied to explore imprecise human cognitive boundaries and judgmental behaviors in manufacturing systems. Jahan-Shahi et al. proposed multivalued fuzzy sets to model nonprocess factors (e.g., operators’ condition and working environment) in time/cost estimation of flat plate processing [6]. Taking fuzzy human behavior into account, Utkin et al. proposed a method to analyze computer integrated manufacturing systems with probability and possibility measures [19]. Several methods have been proposed to integrate fuzzy logic into Petri nets for knowledge representation and modeling of dynamics of manufacturing systems [1], [2], [9]. In view of resource constraints, Wu et al. proposed a modified fuzzy Petri net to model alternative and optimal operation planning [20]. Considering the involvement of human operators in flexible manufacturing systems (FMSs), Xu presented a fuzzy Petri net model and a two-stage fuzzy reasoning algorithm to derive the optimum schedule for the proposed FMS [21]. In a highly automated manufacturing environment, human operators are recognized as decision makers or problem solvers that interact with hardware and software. In the disassembly process, human operators are involved not only in the decision-making process but also the dismantling tasks in many situations since disassembly automation level is very low compared with manufacturing automation. Coupled with the uncertainty in product condition, this makes disassembly much more complex. Considering similarities and differences between assembly and disassembly, this paper proposes a fuzzy attributed Petri net (FAPN) model to address human factors in disassembly. An algorithm based upon the model is then presented to dynamically derive the optimal disassembly sequence for an obsolete product. The rest
1083-4427/$20.00 © 2006 IEEE
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
of this paper is organized as follows. Section II presents an FAPN model for disassembly modeling. Section III focuses on disassembly process planning. Section IV gives a case study, and Section V presents the conclusion and future research directions. II. D ISASSEMBLY M ODELING W ITH H UMAN F ACTORS Obsolete products exhibit a high level of uncertainty in their structure and condition, which may invalidate the original precedence relationship and further complicate the disassembly process. For example, products that originally have the same configuration may need different disassembly paths, and unpredictable physical or functional defects can lead to a need for destructive disassembly. When that is the case, it is always at human operators’/managers’ discretion to choose appropriate disassembly tools and orientation and decide to continue the disassembly or to send the product/subassembly directly to a shredder for disposal. A decision made by an unskilled worker might degrade disassembly productivity in terms of longer process time and a large amount of residue going to landfills. While several disturbances exist in a real disassembly cycle as the result of human intervention, ignoring them will lead to inaccurate process plan and control. This paper focuses on three human factors in the disassembly process, which are illustrated in Section II-A. Taking these nonprocess factors into account, an FAPN is proposed in Section II-B to model all feasible disassembly sequences of an obsolete product, where the impact of human operations on disassembly is mathematically represented as fuzzy attributes associated with the Petri net components that model operators. A. Impact of Human Factors on Disassembly There are two types of human operations in disassembly. In heavy-duty tasks (e.g., shredding and processing contaminated materials) usually handled by machines, a human being is involved in the process by interacting with hardware and software (e.g., setup, loading, and monitoring). Their performance is purely based on their understanding of instructions. On the other hand, human operators may participate in simple dismantling tasks (e.g., taking a part out from a product/subassembly) by making decisions regarding which tool to use and which orientation to follow. In this paper, variations in the outcome of different operators working under identical conditions are attributed to skill, defined as “proficiency at following a given method.” Under different skill levels, the time that an operator needs to complete a task, the quality of the product/decision resulting from this manual operation, and the labor cost for this operation would be quite different, which are the major aspects considered in this paper. 1) Disassembly Time: A disassembly task, referred to as simply “a task” in the context of this paper, is defined as a process handled by a machine or an operator to dismantle a product or subassembly into two or more subassemblies and/or components. It is assumed that each task can be completed by a machine or an operator within a certain fixed time on
719
the average called operation time. The operation time depends on the condition of a product/subassembly. As human being is involved in a task as either to set up a machine for the process or to completely deal with the process, extra time, called extended time, is added, depending on the skill level of the operator. The disassembly time, defined as the time taken to complete a particular task, is thus the sum of the operation time and the extended time. 2) Labor Cost: Labor cost is an important factor that partially determines the final cost for a disassembly task. The higher skill level an operator has, the more salary he/she receives. On the other hand, a higher skill operator spends less time than a lower skill operator on the same task. 3) Quality of Disassembled Subassemblies/Components: An unskilled worker may damage disassembled components by using inappropriate tools or following an incorrect disassembly orientation due to the lack of experience. This performance may further degrade the disassembly revenue in view of the resale/reuse values of disassembled components. The possibility of obtaining a quality subassembly/component from a skilled operator is generally higher than that of an unskilled one.
B. Fuzzy Attributed Petri Net As stated earlier, the impact of human intervention is usually evaluated using qualitative linguistic terms. A linguistic variable differs from a numerical variable in that its values are not numbers but words or sentences in a natural or artificial language. For example, the labor cost of an operator is a linguistic variable, and its values can be numerical or high, fair, low, and so on. Fuzzy set theory provides a good tool to represent such vague input data by formulating the values using membership functions. By taking advantage of both fuzzy logic and well-formed formalism of Petri nets, this paper proposes an FAPN model to analyze human-in-the-loop disassembly planning, where operation time for each disassembly task and the expected profit through the resale/reuse of a discarded product/subassembly/component are assumed deterministic. Definition 2.1: An FAPN is an eight-tuple model defined as FAPN = (P, T, I, O, M, τ, α, λ) where we have the following. 1) P is a nonempty set, where P = {p1 , p2 , . . . , pn } = W ∪ Q. The set of places W = {w1 , w2 , . . . , wr } represents operators, and the set of places Q = {q1 , q2 , . . . , qs } stands for a product, subassembly, or component. r + s = n: a) W ∩ Q = ∅. b) There is a unique p ∈ Q such that • p = ∅: This place is usually denoted by p1 and named as root or a product place. c) ∃ Q ⊂ Q such that Q = ∅ and ∀p ∈ Q , p• = ∅. Places in Q are called leaves. d) I(p, t) = O(p, t) ∀t ∈ T and ∀p ∈ W . 2) T = {t1 , t2 , . . . , tm } is a nonempty set of transitions representing disassembly operations. P ∩ T = ∅.
720
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
Fig. 2.
Fig. 1. Simple example of FAPN.
3) I : P × T → {0, 1} is the input function that defines the set of ordered pairs (pi , tj ), where I(pi , tj ) = 1, if pi is an input place for tj ; otherwise, 0. 4) O : P × T → {0, 1} is an output function that defines the set of ordered pairs (pi , tj ), where O(pi , tj ) = 1, if pi is an output place for tj ; otherwise, 0. 5) M : P → {0, 1, 2, . . .} is a function that defines a marking vector, where M (pi ) represents the number of tokens in pi . An initial marking and a final marking are denoted by m0 and mf , respectively. They satisfy the following: a) m0 (p1 ) = 1, m0 (p) = 0 ∀p ∈ Q − {p1 }, and m0 (p) = 1, ∀p ∈ W . b) mf (p) = m0 (p) ∀p ∈ W . 6) τ : Q → {0} ∪ + is a revenue function assigned to a place, which is the average profit earned by resale/reuse of the product, subassembly, or component represented by p. 7) α : W → { i {(ρi (w), σρi (w))} is a mapping function that maps w to a union of two-tuple ρi (w), which is the fuzzy attribute, and σρi ∈ [0, 1], which is the corresponding truth degree. 8) λ : T → + is an operation time associated with a transition. It is measured in hours. The human factors discussed in this paper are limited to disassembly time, labor cost, and quality of disassembled subassemblies/components. Therefore, ∀w ∈ W ; three attributes are associated with it, which are the fuzzy propositions regarding the impact of the operator w on disassembly [e.g., the impact on disassembly time is small ρ1 (w), the labor cost is high ρ2 (w), and the bad influence on disassembly quality is small ρ3 (w)]. The corresponding truth degrees represent the extent to which the proposition is true. Furthermore, this paper assumes that each disassembly operation takes only one component away from an assembly/subassembly for simplicity. A simple example of FAPN is given in Fig. 1 that models the decomposition of a product into two components processed by an operator FAPN = (P, T, I, O, M, τ, α, λ) P T m0 τ
= {q1 , q2 , q3 , w1 }, where q1 is the root place; = {t1 }; = {1, 0, 0, 1}; = {0, 0.2, 0.5};
Membership function for an operator’s salary.
α(w1 ) = {(the effect on disassembly time is small, 0.75), (the labor cost is high, 0.7), (the bad effect on disassembly quality is small, 0.8)}; λ = {1/6}.
III. D ISASSEMBLY P ROCESS P LANNING Currently, disassembly automation level is very low. Because of a large amount of human intervention involved in this process, the competence and motivation of personnel have major impact on the efficiency and quality of disassembly. The skill level of human operators then becomes an imperative factor in determining the sequence of disassembly operations. In an FAPN, an operator place is introduced as an input to a transition. Considering the impact of a human operation on disassembly cost and revenue, not only are fuzzy attributes ρi (wj ) introduced to represent human factors, but also a truth degree is associated with each attribute for the strength of the impact. For a particular disassembly operation t, the disassembly time is calculated as the sum of operation time and extended time, where the latter is a function of an operator’s skill. The higher the skill of an operator, the shorter the extended time. As the truth degree drops with increasing extended time, the relationship between the truth degree and the extended time is assumed to be exponential [21]. The extended time Text (wj ) is then calculated as Text (wj ) =
− ln (σρ1 (wj )) β
(3.1)
where β is a constant whose value is chosen based on statistics, and σρ1 (wj ) is the truth degree of the attribute ρ1 (wj ). Based on (3.1), the disassembly time for the disassembly operation t Tdis (t) is Tdis (t) =
− ln (σρ1 (wj )) + λ(t), β
wj ∈• t ∩ wj ∈ W. (3.2)
Labor cost is also an important factor in determining the final disassembly cost. The more experienced operators are, the higher the salary they get. This paper assumes that the fuzzy proposition—the labor cost is high—is characterized as the function of operator’s salary. As an example, as shown in Fig. 2, the maximum and minimum salaries are $18/h and $6/h, respectively.
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
721
represents the cost/revenue when place • t ∈ Q − Q takes t for further disassembly. c(t) and r(p) are calculated using (3.3) and (3.4), respectively. After the bottom-up calculation, the FAPN is traversed, starting from p1 . Through this exhaustive search, the optimal disassembly path of a discarded product can be obtained by Algorithm 3.1. Fig. 3 summarizes the overall logic of the approach.
Fig. 3.
Logic of optimal disassembly planning.
Therefore, the final disassembly cost for a particular disassembly operation t, which is c(t), is calculated as (assuming that the charge for disassembly tools is ignored) c(t) = Tdis (t) (6 + 12σρ2 (wj ))
(3.3)
where σρ2 (wj ) is the truth degree of the attribute ρ2 (wj ). Considering the damage to a disassembled subassembly/ component made by an unskilled operator, the adjusted revenue for each place p, p ∈ Q, is also introduced and denoted as r(p) =
σρ3 (w)τ (p), τ (p),
p ∈ (w• )• p = p1
(3.4)
where σρ3 (w) is the truth degree of the attribute ρ3 (wj ). ∀p ∈ Q − {p1 }, r(p) may take different values as p results from separate disassembly operations processed by different operators w. Finally, the disassembly-planning algorithm with the consideration of human factors is proposed. The aim is to identify the order of disassembly operations (i.e., transitions firing in an FAPN) that maximizes the expected return from the processed items. This objective is then formally expressed by introducing the disassembly value function d : Q → {0} ∪ + , which is associated with the disassembly plan and maps each place p ∈ Q to a non-negative value representing the expected return obtained from the processing of the product/ subassembly/component represented by place p. Eventually, computing an optimal disassembly plan is restated as the determination of a disassembly sequence that maximizes d(p1 ). According to the acyclic structure of the FAPN model defined in Section II, the maximization of d(p1 ) can be achieved through a recursive computation that starts from the leaf places and derives the optimal d-values for every place p ∈ Q. The detailed calculation is defined as follows. Definition 3.1: The disassembly value of place p is denoted as d(p) and is calculated in a recursive manner using 1) ∀p ∈ Q , d(p) = r(p) and 2) ∀p ∈ Q − Q , d(p) = max{r(p), maxt∈p• ∆(t)}, where ∆(t) = q∈t• ∩q∈Q d(q) − c(t) is the margin benefit of transition t, t ∈ T , which
Algorithm 3.1: Step 1. For p ∈ Q , which is the set of leaves according to Definition 2.1, set d(p) = r(p). Step 2. Set L = Q − Q . While (L = ∅) do: Find a place p ∈ L such that ∀q ∈ (p• )• , q ∈ L ∩ q ∈ Q. Calculate ∆(t) = d(q) − c(t) q∈t•
∆(t) . d(p) = max r(p), max •
t∈p
Let L = L − {p}, i.e., remove p from L. End. Step 3. Set Z = {p1 } (assuming that p1 is the root node for every incoming product), DP = ∅, and C = ∅ (DP represents final disassembly path, and C represents the last components after disassembly). While (Z = ∅) do: For each node p in Z: a) Find a transition t, where t ∈ p• , ∆(t) = d(p) = r(p). b) If a) succeeds, DP = DP ∪ {t}, and Z = Z ∪ {t• }; otherwise, C = C ∪ {p}. c) Z = Z − {p}. End. Due to the deterministic features of the FAPN (i.e., the operation time associated with each transition, the revenue function associated with each place in Q, and the truth degree of each attribute associated with a worker place are assumed known a priori), the exhaustive search in Algorithm 3.1 guarantees the optimal disassembly planning. In reality, uncertainty exists in product condition and judgmental estimation of worker skill that is extremely critical in determining disassembly costs and profits. Therefore, uncertainty management methods are needed to ensure this optimization in practice, which is out of the scope of this paper and is not addressed here. Proposition 1: In an FAPN, the number of leaf places is less than or equal to n + 1, where n is the number of transitions. Proof: The number of places p ∈ Q and transitions varies from one product to another since it depends on how the components in the product are interconnected. Taking an extreme case into consideration where there are m − 1 interconnections between the m components and there is only one disassembly sequence feasible, the number of leaf places is m = n + 1, under the assumption that each transition disassembles one component at time. As the number of interconnection between
722
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
components and possible disassembly sequences increase, more transitions are introduced into the FAPN, and the number of leaf places is unchanged. Therefore, Proposition 1 is proven. Proposition 2: The algorithm complexity is O(n), where n is the number of transitions in an FAPN. Proof: Let |Y | be the cardinality of set Y . 1) In Step 1, the disassembly value of each leaf place is calculated. From Proposition 1, the number of leaf places is less than or equal to n + 1. Therefore, the computation runs at O(n). 2) In Step 2, the calculation is based on each nonleaf place. For each such place pi ∈ Q − Q , the marginal benefits of its output transitions ∆(t), t ∈ p•i are required to be calculated and compared with its adjusted revenue r(pi ). The computation runs at O(2|p•i |) in the worst case. Since leaf places do not have output transitions and ∀pi , pj ∈ • n. Therefore, Q − Q , p•i ∩ p•j = ∅, pi ∈Q−Q |pi | = the total computational complexity is O(2 pi ∈Q−Q p•i ), i.e., O(n). 3) In Step 3, in order to find the disassembly sequence with the best cost/revenue ratio, the traversal of the FAPN is needed, which runs at O(n) at worst case. Hence, the computational complexity of the algorithm is O(n). IV. C ASE S TUDY To fully understand the above concepts and algorithm, this paper designs and implements a disassembly system and uses an obsolete personal computer (PC) as a disassembly example. Section IV-A gives the design and implementation method, and Section IV-B presents the experimental results. A. Implementation Method By using the developed FAPN and algorithm, this paper proposes the following design and implementation steps in a disassembly system where a given number of machines/workers form a disassembly line to dismantle one type of products. 1) Construct an FAPN given the product information, all feasible disassembly choices, and workers’ profiles (e.g., functionality, skill level, etc.). 2) Collect and associate all the data with places and transitions in the FAPN. 3) Update the data associated with places and transitions by considering human factors through (3.3) and (3.4). 4) Integrate the developed algorithm (i.e., Algorithm 3.1) with controllers and other hardware in the disassembly system. The Pentium II PC used in this experiment consists of 11 main components, i.e., a cover case {j}, cables {k}, a power supply {h}, a motherboard {a}, expansion cards {e, f, g}, secondary storage {b, c, d}, and memory chips {i}. Its layout without the case is given in Fig. 4, showing how components are connected to each other. This PC will go through some or all of the four stages in the prototypical disassembly system
Fig. 4.
Layout of a Pentium II PC.
according to the product condition. In Stage 1, the case and cables are separated from the PC. In Stage 2, the secondary storage is taken apart. The disassembly of the expansion cards and the memory chips and the power supply is processed in Stages 3 and 4, respectively. A worker is assigned to each stage, i.e., wj is for stage j, j = 1, 2, 3, and 4. To construct an FAPN, the entire product is labeled by p1 . All feasible ways to disassemble the product into the next-level subassemblies or components are then identified. Meanwhile, a corresponding worker place is added as an input to each transition. For instance, at p1 , only one method is recognized and represented by t1 , which means that other components can be disassembled only after unscrewing the case and taking out all cables. w1 is identified as the corresponding worker place for t1 . This process is repeated for each subassembly until no subassembly is left. Finally, the complete FAPN is obtained, as shown in Fig. 5. To simplify the presentation while keeping clarity, operator places wi (i = 1 to 4) are shown more than once. Some individual components are grouped together and represented by a single place, i.e., p9 is for {j, k}, p12 for {b, c, d}, and p13 for {e, f, g, i}. Second, the relevant data is collected and associated with places and transitions. An adjustment is then made in the light of impact of human operations. Finally, the proposed algorithm is implemented in a modular structure written in Visual C++, which includes Database, Process Planner, and Controller, and runs under the Windows 2000/XP operating system with the Access database engine. Database provides the relevant information of products (e.g., reuse/resale values and operation time for corresponding disassembly tasks) and workers (i.e., skill level). In this paper, the worker skill level is characterized by technical evaluation score and salary. The evaluation is based on a worker’s experience, performance, and knowledge. The truth degrees of the three attributes associated with each worker are then decided based on his/her evaluation score and salary through membership functions. During the implementation, this information in Database is determined according to the previous benchmark experience and created as follows. 1) In an FAPN, the revenue value associated with place p ∈ Q and the operation time for each disassembly task represented by t are randomly populated with a uniform
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
Fig. 5.
723
Example of FAPN for a PC.
distribution, which have the range [0, 1.5] and [0.01, 0.1], respectively. 2) The score of each worker is also randomly chosen with a uniform distribution between [60, 100]. 3) The worker’s salary is proportional to his/her score with the range [6, 18]. 4) β is chosen to be 100. Based on the information from Database, Process Planner presents the Controller the disassembly sequence of an obsolete product, on which Controller dispatches tasks to different disassembly stages.
Worker
Score
Salary ($/h)
w1 w2 w3 w4
86.2 66 96 70
13.8 7.5 16.8 9
2) FAPN = (P, T, I, O, M, τ, α, λ) P = {q1 , q2 . . . q13 , w1 , w2 , w3 , w4 } T = {t1 , t2 . . . t13 }
B. Experimental Results In this section, the value gain in a disassembly plan DP [16] f (DP) is introduced first. A case study is then performed to illustrate the idea of disassembly process planning via the disassembly system and the obsolete PC discussed above. Definition 4.1: The value gain in a disassembly plan DP f (DP) is f (DP) =
1) Worker information (i.e., score and salary)
p∈C
r(p) −
c(t).
(4.1)
t∈DP
The first set of experiment is focused on the functional structure of the planner and on the detailed operation of the software modules. When the obsolete PC comes to the disassembly facility, its FAPN is modeled and the corresponding data is collected from Database
m0 (p1 ) = 1;
m0 (pi ) = 0,
i = 2, . . . , 13
m0 (w1 ) = m0 (w2 ) = m0 (w3 ) = m0 (w4 ) = 1 τ = {0, 0, 0, 0, 0, 0, 0, 0, 0.9, 0.25, 0.5, 0.9, 1.2} α(w1 ) = {(the effect on disassembly time is small, 0.7) (labor cost is high, 0.65) (the bad effect on disassembly quality is small, 0.8)} α(w2 ) = {(the effect on disassembly time is small, 0.25) (labor cost is high, 0.15) (the bad effect on disassembly quality is small, 0.2)}
724
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
α(w3 ) = {(the effect on disassembly time is small, 0.9) (labor cost is high, 0.9) (the bad effect on disassembly quality is small, 1)} α(w4 ) = {(the effect on disassembly time is small, 0.3) (labor cost is high, 0.25) (the bad effect on disassembly quality is small, 0.2)} λ = {0.04, 0.06, 0.06, 0.04, 0.04, 0.05, 0.05 0.03, 0.09, 0.05, 0.07, 0.08, 0.04}. Based on the worker profiles, the disassembly cost for each transition and the revenue for each place in Q are updated c(t) = {0.6072, 0.6480, 1.0248, 0.4680, 0.4680 0.4992, 0.8568, 0.3780, 0.8112, 0.8568 0.6552, 1.3608, 0.4212} r(p) = {0, 0, 0, 0, 0, 0, 0, 0, 0.72, 0.05, 0.5 σρ3 (wj ), 0.18, 1.2} where r(p11 ) is not fixed since it may be processed by different workers wj via different disassembly sequences. While executing Algorithm 3.1 through the FAPN, the disassembly value for each p and the marginal benefit of each transition t are calculated. Finally, the optimal disassembly path for this PC is obtained. This disassembly sequence is referred to as DP1 ∆(t) = {0.456, −0.230, 0.175, 0.425, −0.418, −0.219, 0.843 − 0.228, −0.206, 0.343, 0.368, −0.161, −0.241} d(p) = {0.456, 0.343, 0.368, 0.425, 0, 0, 0.843, 0, 0.72 0.05, 0.5σρ3 (wj ), 0.18, 1.2} DP1 = {t1 , t10 } and f (DP1 ) = 0.456. If human factors were ignored in Process Planner, the reuse/resale value τ and the operation time λ would replace the adjusted r(p) and c(t) to calculate the disassembly value for each p and the marginal benefit for each transition t [22]. This would result in a different disassembly sequence referred to DP2 = {t1 , t10 , t5 , t6 }. However, t5 and t6 are handled by inexperienced workers w4 and w2 , respectively, in the real disassembly cycle. These human operations degrade the reuse/resale values of the disassembled components from t5 and t6 and causes longer disassembly time than what is used in Process Planner. Therefore, the value gain of this plan is decreased f (DP2 ) = −0.181. Fig. 6(a) gives the comparison of the disassembly value of this plan versus that of the DP1 as a function of the disassembly steps.
Fig. 6. Experimental results. (a) Process plan DP2 without considering human factors versus an adapted plan DP1 with human factors. (b) Modified process plan DP3 by enhancing the competence of a worker verse the adapted one DP1 .
Taking human factors into consideration, it is found that the efficient way to disassemble this PC is to retrieve only p9 and p13 , and leave p5 unchanged. Although there are two options to decompose p5 , t5 , and t13 , the operations handled by two unskilled workers w4 and w2 result in more cost than revenue. If the competence of operators (e.g., w2 ) would be enhanced to the same level as w1 , the extended time needed for t13 will be decreased, and the bad effects on the quality of two disassembled components ( p8 and p12 ) will be minimized. Consequently, the marginal benefit of t13 increases from −0.241 to 0.113, which results in another disassembly sequence referred to as DP3 = {t1 , t10 , t13 } with better value gain f (DP3 ) = 0.569. The comparison of disassembly value of DP3 with that of DP1 is given in Fig. 6(b). It is clear that the performance of an operator significantly affects the disassembly profit, which is a key factor in decision making.
V. C ONCLUSION Disassembly is currently labor intensive. The large amount of human intervention gives rise to many uncertainties in disassembly operations. Ignoring these human factors results in surprisingly few applications of current approaches for disassembly planning. To address this deficiency, this paper proposes a fuzzy attributed Petri net (FAPN) model where the influence of human factors on disassembly (e.g., disassembly time, quality of disassembled components, and labor cost) is mathematically represented as membership functions. An algorithm is then developed upon this model to derive the
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
cost-effective disassembly sequence of a discarded product. To the authors’ knowledge, no paper has comprehensively dealt with human factors in disassembly. This work successfully overcomes a deficiency in [16] and [22], which ignores the impact of human operations (an indispensable part of the whole disassembly system), thereby making the proposed methodology more applicable to real industrial settings. From the example, it was found that the performance of human operators has significant influence in disassembly decision making. The enhanced competence, motivation, and stability of human operators have a major role in improving productivity and efficiency of disassembly. The developed model can be adapted and matched to different operator–work–environment conditions by simply shifting its membership functions to provide more flexibility for cost/time estimators. The research can be extended in several directions. For instance, the proposed FAPN model uses the deterministic nature of the assumptions characterizing the associated revenue function and the time function. In reality, these involved costs and profits in disassembly will be determined dynamically due to the prevailing condition of discarded products. Thus, it is necessary and challenging to refine the proposed model with probabilistic parameters for product uncertainty, on which a learning-based algorithm would be developed, and exploit the past “knowledge” regarding the uncertainty towards improving future decisions [14]. More factory data (i.e., actual profit and cost figures in real-world problems) also need to be used to validate the proposed methodology in the future. The refinement of the proposed model to efficiently analyze the impact of physical, psychological, personal, and other characteristics of humans [6] in complex disassembly systems is another interesting issue.
ACKNOWLEDGMENT
725
[8] ——, “Disassembly sequencing: A survey,” Int. J. Prod. Res., vol. 41, no. 16, pp. 3721–3759, 2003. [9] X. O. Li, W. Yu, and F. Lara-Rosano, “Dynamic knowledge inference and learning under adaptive fuzzy Petri net framework,” IEEE Trans. Syst., Man, Cybern. C, vol. 30, no. 4, pp. 442–450, Nov. 2000. [10] A. Meacham, R. Uzsoy, and U. Venkatadri, “Optimal disassembly configurations for single and multiple products,” J. Manuf. Syst., vol. 18, no. 5, pp. 311–322, 1999. [11] K. E. Moore, A. Gungor, and S. M. Gupta, “Petri net approach to disassembly process planning for products with complex AND/OR precedence relationships,” Eur. J. Oper. Res., vol. 135, no. 2, pp. 428–449, 2001. [12] J. Mikler, H. Hådeby, A. Kjellberg, and G. Sohlenius, “Towards profitable persistent manufacturing human factors in overcoming disturbances in production systems,” Int. J. Adv. Manuf. Technol., vol. 15, no. 10, pp. 749–756, 1999. [13] M. Rahimi, P. A. Hancock, and A. Majchrzak, “On managing the human factors engineering of hybrid production systems,” IEEE Trans. Eng. Manage., vol. 35, no. 4, pp. 238–249, Nov. 1988. [14] S. Reveliotis, “Uncertainty management in optimal disassembly planning through learning-based strategies,” in Advances in Manufacturing, Logistics and Supply Chain Management, N. R. S. Raghavan et al., Ed. Bangalore, India: IIS, 2003, pp. 135–141. [15] K. K. Seo, J. H. Park, and D. S. Jang, “Optimal disassembly sequence using genetic algorithms considering economic and environmental aspects,” Int. J. Adv. Manuf. Technol., vol. 18, no. 5, pp. 371–380, 2001. [16] Y. Tang, M. C. Zhou, and R. Caudill, “An integrated approach to disassembly planning and demanufacturing operation,” IEEE Trans. Robot. Autom., vol. 17, no. 6, pp. 773–784, Dec. 2001. [17] Y. Tang, M. C. Zhou, E. Zussman, and R. Caudill, “Disassembly modeling, planning, and application,” J. Manuf. Syst., vol. 21, no. 3, pp. 200–217, 2002. [18] M. K. Tiwari, S. K. Mukhopadhyay, N. Sinha, S. Kumar, and R. Rai, “A Petri net based approach to determine the disassembly strategy of a product,” Int. J. Prod. Res., vol. 40, no. 5, pp. 1113–1129, 2002. [19] L. V. Utkin, S. V. Gurov, and I. B. Shubinsky, “Analysis of computer integrated manufacturing systems by fuzzy human operator behavior,” J. Qual. Maint. Eng., vol. 3, no. 3, p. 189, 1997. [20] R. R. Wu, L. Ma, J. Mathew, and G. H. Duan, “Optimal operation planning using fuzzy Petri nets with resource constraints,” Int. J. Comput. Integr. Manuf., vol. 15, no. 1, pp. 28–36, 2002. [21] J. X. Xu, “Fuzzy Petri net-based optimum scheduling of FMS with human factors,” in Proc. 35th Conf. Decision Control, Kobe, Japan, Dec. 1996, pp. 4451–4452. [22] E. Zussman and M. C. Zhou, “Design and implementation of an adaptive process planner for disassembly processes,” IEEE Trans. Robot. Autom., vol. 16, no. 2, pp. 171–179, Apr. 2000.
The authors would like to thank the comments provided by anonymous reviewers and the Associated Editor, which greatly helped the presentation of this paper. R EFERENCES [1] T. Cao and A. C. Sanderson, “Task sequence planning using fuzzy Petri nets,” IEEE Trans. Syst., Man, Cybern., vol. 25, no. 5, pp. 755–768, May 1995. [2] M. Gao, M. C. Zhou, X. Huang, and Z. Wu, “Fuzzy reasoning Petri nets,” IEEE Trans. Syst., Man, Cybern. A, vol. 33, no. 3, pp. 314–324, May 2003. [3] M. Gao, M. C. Zhou, and Y. Tang, “Intelligent decision making in disassembly process based on fuzzy reasoning Petri nets,” IEEE Trans. Syst., Man, Cybern. B, vol. 34, no. 5, pp. 2029–2034, Oct. 2004. [4] M. A. Goodrich and E. R. Boer, “Model-based human-centered task automation: A case study in ACC system design,” IEEE Trans. Syst., Man, Cybern. A, Syst., Humans, vol. 33, no. 3, pp. 325–336, May 2003. [5] A. Hula, K. Jalali, K. Hamza, S. J. Skerlos, and K. Saitou, “Multi-criteria decision-making for optimization of product disassembly under multiple situations,” Environ. Sci. Technol., vol. 37, no. 23, pp. 5303–5313, 2003. [6] H. Jahan-Shahi, E. Shayan, and S. H. Masood, “Multivalued fuzzy sets in cost/time estimation of flat plate processing,” Int. J. Adv. Manuf. Technol., vol. 17, no. 10, pp. 751–759, 2001. [7] A. J. D. Lambert, “Determining optimum disassembly sequences in electronic equipment,” Comput. Ind. Eng., vol. 43, no. 3, pp. 553–575, 2002.
Ying Tang (S’99–M’02) received the B.S. and the M.S. degrees from the Northeastern University, Shenyang, China, in 1996 and 1998, respectively, and the Ph.D. degree from the New Jersey Institute of Technology, Newark, in 2001. She is currently an Assistant Professor of electrical and computer engineering at Rowan University, Glassboro, NJ. Her research interests include modeling and scheduling of computer-integrated systems, Petri nets and applications, artificial intelligence, reconfigurable systems design, hardware and software co-design, software security, and networking and communication. She has over 30 publications in international journals, conference proceedings, and book chapter. Dr. Tang is a member of Sigma Xi.
726
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
MengChu Zhou (S’88–M’90–SM’93–F’03) received the B.S. degree from Nanjing University of Science and Technology, Nanjing, China, in 1983, the M.S. degree from Beijing Institute of Technology, Beijing, China, in 1986, and the Ph.D. degree in computer and systems engineering from Rensselaer Polytechnic Institute, Troy, NY, in 1990. He joined New Jersey Institute of Technology (NJIT), Newark, in 1990 and is currently a Professor of electrical and computer engineering. He has been invited to conduct lectures in Australia, Canada, China, France, Germany, Hong Kong, Italy, Japan, Korea, Mexico, Taiwan, and the US. His research interests are in computer-integrated systems, Petri nets, semiconductor manufacturing, multilifecycle engineering, biosignal, and sensor network. He is the author, coauthor, and editor of several books and many published articles. Dr. Zhou served as an Associate Editor of the IEEE TRANSACTIONS ON R OBOTICS AND A UTOMATION from 1997 to 2000, and currently, he is an Associate Editor of the IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND E NGINEERING and the Managing Editor of IEEE T RANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: PART C. He organized and chaired over 70 technical sessions and served on program committees for many conferences. He was the Program Chair of the 1998 and Cochair of the 2001 IEEE International Conference on Systems, Man, and Cybernetics (SMC) and the 1997 IEEE International Conference on Emerging Technologies and Factory Automation. He was also a Guest Editor of the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, the IEEE TRANSACTIONS ON S EMICONDUCTOR M ANUFACTURING , and the IEEE T RANSACTIONS ON MECHATRONICS. He is Editor-in-Chief of the International Journal of Intelligent Control and Systems since 1996. He was the General Cochair of the 2003 IEEE International Conference on Systems, Man, and Cybernetics, Washington D.C., October 5–8, 2003, and the 2004 IEEE International Conference on Networking, Sensing and Control, Taipei, Taiwan, March 21–23, 2004. He has led or participated in 26 research and education projects with a total budget of over $10M, funded by the National Science Foundation, Department of Defense, Engineering Foundation, New Jersey Science and Technology Commission, and industry. He is a member of the Board of Governors of the IEEE SMC Society. He is a life member of the Chinese Association for Science and Technology-USA and served as its President in 1999. He is a recipient of the NSF’s Research Initiation Award, CIM University-LEAD Award by the Society of Manufacturing Engineers, the Perlis Research Award by NJIT, the Humboldt Research Award for US Senior Scientists, the Leadership Award and Academic Achievement Award by the Chinese Association for Science and TechnologyUSA, the Asian American Achievement Award by the Asian American Heritage Council of New Jersey, and the Outstanding Contribution Award by the IEEE Systems, Man, and Cybernetics (SMC) Society. He was the founding Chair of the Discrete Event Systems Technical Committee of the IEEE SMC Society and the Cochair (founding) of the Semiconductor Factory Automation Technical Committee of the IEEE Robotics and Automation Society.
Meimei Gao (S’02–M’03) received the B.S. and M.S. degrees from the Northwestern Polytechnic University, Xi’an, China, in 1994 and 1997, respectively, the Ph.D. degree in control theory and engineering from Shanghai Jiao Tong University, Shanghai, China, in 2000, and the Ph.D. degree in computer engineering from New Jersey Institute of Technology, Newark, in 2003. She is currently an Assistant Professor with the Department of Mathematics and Computer Science, Seton Hall University, South Orange, NJ. Her research interests include discrete-event systems, Petri nets and applications, artificial intelligence, decision making, computer-integrated manufacturing and demanufacturing, and lifecycle engineering. She has published over 20 journal and conference proceedings papers in the above research areas. Dr. Gao is a member of ACM.
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
Fuzzy-Petri-Net-Based Disassembly Planning Considering Human Factors Ying Tang, Member, IEEE, MengChu Zhou, Fellow, IEEE, and Meimei Gao, Member, IEEE
Abstract—Disassembly, as the process of systematic removal of desirable constituent parts from an assembly, is of growing importance due to the increasing environmental and economic pressure. Although disassembly in practice is manual and labor intensive, little attention has been paid to the human intervention in the disassembly process. This paper addresses this deficiency by developing a fuzzy attributed Petri net (FAPN) model to mathematically represent uncertainty in disassembly due to a large amount of human intervention. An algorithm based upon this model is further proposed for optimal disassembly planning with a view to making the technique more applicable to real industry settings. The benefit of the proposed model and algorithm is illustrated through the disassembly of a personal computer (PC) in a prototypical disassembly system. Index Terms—Disassembly planning, fuzzy logic, human factors, optimization, Petri nets.
I. I NTRODUCTION
D
ISASSEMBLY, as the process of systematic removal of desirable constituent parts from an assembly, is of growing importance due to increasing environmental and economic pressure. Considering disassembly as an expensive process, many researchers and industry executives have started to realize opportunities for performing disassembly in a cost-effective manner. Some methodologies are developed to identify the extent to which the disassembly of a product should be conducted to keep the process profitable and environmentally friendly [3], [10], [22]. Several heuristic approaches are proposed to generate the optimal disassembly sequence that minimizes the cost of disassembly (assuming that a certain level of disassembly is required) or obtains the best cost/benefit ratio for disassembly [5], [7], [11], [15], [17], [18]. A recent survey of disassembly modeling and planning is presented in [8] and [17]. While a significant amount of research is being conducted in this area, not much effort has been focused on human factors in disassembly. In fact, disassembly is currently Manuscript received April 27, 2004; revised August 31, 2004. This work was supported in part by the National Science Foundation under Grant 0097887, the National Natural Science Foundation of China under Grants 60228004 and 60334020, the Shandong Provincial Government under Grant 030335, and the Ministry of Science and Technology of China under Grant 2002CB312200. This paper was recommended by Associate Editor E. J. Bass. Y. Tang is with the Department of Electrical and Computer Engineering, Rowan University, Glassboro, NJ 08028 USA (e-mail: [email protected]). M. Zhou is with the Department of Electrical and Computer Engineering, New Jersey Institute of Technology, Newark, NJ 07102 USA, and also with the Laboratory of Complex Systems and the Intelligence Science Institute of Automation, Chinese Academy of Sciences, Beijing 100080, China (e-mail: [email protected]). M. Gao is with the Department of Mathematics and Computer Science, Seton Hall University, South Orange, NJ 07079 USA (e-mail: [email protected]). Digital Object Identifier 10.1109/TSMCA.2005.853508
manual/semiautomatic and labor intensive. The large amount of human intervention in disassembly provides extra flexibility, but it also gives rise to many uncertainties in the process. For instance, disassembly processing time, quality of disassembled components, and disassembly cost might change significantly due to varying skill levels of human operators. Moreover, these uncertainties contain fuzzy or imprecise information that are sometimes described by linguistic terms. There is a need to develop a formal model that allows us to analyze the influence of human intervention in disassembly and incorporate these nonprocess disturbances into disassembly planning and control. Human factors in manufacturing systems arise when designing and controlling the automated production and assembly [12]. A lesson learned from process automation is that, if human factors are not taken into consideration, even technologically state-of-the-art systems can be more problematic than beneficial [4]. Rahimi et al. raised several engineering issues as they concerned the human component of the hybrid system [13]. Fuzzy theory has been applied to explore imprecise human cognitive boundaries and judgmental behaviors in manufacturing systems. Jahan-Shahi et al. proposed multivalued fuzzy sets to model nonprocess factors (e.g., operators’ condition and working environment) in time/cost estimation of flat plate processing [6]. Taking fuzzy human behavior into account, Utkin et al. proposed a method to analyze computer integrated manufacturing systems with probability and possibility measures [19]. Several methods have been proposed to integrate fuzzy logic into Petri nets for knowledge representation and modeling of dynamics of manufacturing systems [1], [2], [9]. In view of resource constraints, Wu et al. proposed a modified fuzzy Petri net to model alternative and optimal operation planning [20]. Considering the involvement of human operators in flexible manufacturing systems (FMSs), Xu presented a fuzzy Petri net model and a two-stage fuzzy reasoning algorithm to derive the optimum schedule for the proposed FMS [21]. In a highly automated manufacturing environment, human operators are recognized as decision makers or problem solvers that interact with hardware and software. In the disassembly process, human operators are involved not only in the decision-making process but also the dismantling tasks in many situations since disassembly automation level is very low compared with manufacturing automation. Coupled with the uncertainty in product condition, this makes disassembly much more complex. Considering similarities and differences between assembly and disassembly, this paper proposes a fuzzy attributed Petri net (FAPN) model to address human factors in disassembly. An algorithm based upon the model is then presented to dynamically derive the optimal disassembly sequence for an obsolete product. The rest
1083-4427/$20.00 © 2006 IEEE
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
of this paper is organized as follows. Section II presents an FAPN model for disassembly modeling. Section III focuses on disassembly process planning. Section IV gives a case study, and Section V presents the conclusion and future research directions. II. D ISASSEMBLY M ODELING W ITH H UMAN F ACTORS Obsolete products exhibit a high level of uncertainty in their structure and condition, which may invalidate the original precedence relationship and further complicate the disassembly process. For example, products that originally have the same configuration may need different disassembly paths, and unpredictable physical or functional defects can lead to a need for destructive disassembly. When that is the case, it is always at human operators’/managers’ discretion to choose appropriate disassembly tools and orientation and decide to continue the disassembly or to send the product/subassembly directly to a shredder for disposal. A decision made by an unskilled worker might degrade disassembly productivity in terms of longer process time and a large amount of residue going to landfills. While several disturbances exist in a real disassembly cycle as the result of human intervention, ignoring them will lead to inaccurate process plan and control. This paper focuses on three human factors in the disassembly process, which are illustrated in Section II-A. Taking these nonprocess factors into account, an FAPN is proposed in Section II-B to model all feasible disassembly sequences of an obsolete product, where the impact of human operations on disassembly is mathematically represented as fuzzy attributes associated with the Petri net components that model operators. A. Impact of Human Factors on Disassembly There are two types of human operations in disassembly. In heavy-duty tasks (e.g., shredding and processing contaminated materials) usually handled by machines, a human being is involved in the process by interacting with hardware and software (e.g., setup, loading, and monitoring). Their performance is purely based on their understanding of instructions. On the other hand, human operators may participate in simple dismantling tasks (e.g., taking a part out from a product/subassembly) by making decisions regarding which tool to use and which orientation to follow. In this paper, variations in the outcome of different operators working under identical conditions are attributed to skill, defined as “proficiency at following a given method.” Under different skill levels, the time that an operator needs to complete a task, the quality of the product/decision resulting from this manual operation, and the labor cost for this operation would be quite different, which are the major aspects considered in this paper. 1) Disassembly Time: A disassembly task, referred to as simply “a task” in the context of this paper, is defined as a process handled by a machine or an operator to dismantle a product or subassembly into two or more subassemblies and/or components. It is assumed that each task can be completed by a machine or an operator within a certain fixed time on
719
the average called operation time. The operation time depends on the condition of a product/subassembly. As human being is involved in a task as either to set up a machine for the process or to completely deal with the process, extra time, called extended time, is added, depending on the skill level of the operator. The disassembly time, defined as the time taken to complete a particular task, is thus the sum of the operation time and the extended time. 2) Labor Cost: Labor cost is an important factor that partially determines the final cost for a disassembly task. The higher skill level an operator has, the more salary he/she receives. On the other hand, a higher skill operator spends less time than a lower skill operator on the same task. 3) Quality of Disassembled Subassemblies/Components: An unskilled worker may damage disassembled components by using inappropriate tools or following an incorrect disassembly orientation due to the lack of experience. This performance may further degrade the disassembly revenue in view of the resale/reuse values of disassembled components. The possibility of obtaining a quality subassembly/component from a skilled operator is generally higher than that of an unskilled one.
B. Fuzzy Attributed Petri Net As stated earlier, the impact of human intervention is usually evaluated using qualitative linguistic terms. A linguistic variable differs from a numerical variable in that its values are not numbers but words or sentences in a natural or artificial language. For example, the labor cost of an operator is a linguistic variable, and its values can be numerical or high, fair, low, and so on. Fuzzy set theory provides a good tool to represent such vague input data by formulating the values using membership functions. By taking advantage of both fuzzy logic and well-formed formalism of Petri nets, this paper proposes an FAPN model to analyze human-in-the-loop disassembly planning, where operation time for each disassembly task and the expected profit through the resale/reuse of a discarded product/subassembly/component are assumed deterministic. Definition 2.1: An FAPN is an eight-tuple model defined as FAPN = (P, T, I, O, M, τ, α, λ) where we have the following. 1) P is a nonempty set, where P = {p1 , p2 , . . . , pn } = W ∪ Q. The set of places W = {w1 , w2 , . . . , wr } represents operators, and the set of places Q = {q1 , q2 , . . . , qs } stands for a product, subassembly, or component. r + s = n: a) W ∩ Q = ∅. b) There is a unique p ∈ Q such that • p = ∅: This place is usually denoted by p1 and named as root or a product place. c) ∃ Q ⊂ Q such that Q = ∅ and ∀p ∈ Q , p• = ∅. Places in Q are called leaves. d) I(p, t) = O(p, t) ∀t ∈ T and ∀p ∈ W . 2) T = {t1 , t2 , . . . , tm } is a nonempty set of transitions representing disassembly operations. P ∩ T = ∅.
720
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
Fig. 2.
Fig. 1. Simple example of FAPN.
3) I : P × T → {0, 1} is the input function that defines the set of ordered pairs (pi , tj ), where I(pi , tj ) = 1, if pi is an input place for tj ; otherwise, 0. 4) O : P × T → {0, 1} is an output function that defines the set of ordered pairs (pi , tj ), where O(pi , tj ) = 1, if pi is an output place for tj ; otherwise, 0. 5) M : P → {0, 1, 2, . . .} is a function that defines a marking vector, where M (pi ) represents the number of tokens in pi . An initial marking and a final marking are denoted by m0 and mf , respectively. They satisfy the following: a) m0 (p1 ) = 1, m0 (p) = 0 ∀p ∈ Q − {p1 }, and m0 (p) = 1, ∀p ∈ W . b) mf (p) = m0 (p) ∀p ∈ W . 6) τ : Q → {0} ∪ + is a revenue function assigned to a place, which is the average profit earned by resale/reuse of the product, subassembly, or component represented by p. 7) α : W → { i {(ρi (w), σρi (w))} is a mapping function that maps w to a union of two-tuple ρi (w), which is the fuzzy attribute, and σρi ∈ [0, 1], which is the corresponding truth degree. 8) λ : T → + is an operation time associated with a transition. It is measured in hours. The human factors discussed in this paper are limited to disassembly time, labor cost, and quality of disassembled subassemblies/components. Therefore, ∀w ∈ W ; three attributes are associated with it, which are the fuzzy propositions regarding the impact of the operator w on disassembly [e.g., the impact on disassembly time is small ρ1 (w), the labor cost is high ρ2 (w), and the bad influence on disassembly quality is small ρ3 (w)]. The corresponding truth degrees represent the extent to which the proposition is true. Furthermore, this paper assumes that each disassembly operation takes only one component away from an assembly/subassembly for simplicity. A simple example of FAPN is given in Fig. 1 that models the decomposition of a product into two components processed by an operator FAPN = (P, T, I, O, M, τ, α, λ) P T m0 τ
= {q1 , q2 , q3 , w1 }, where q1 is the root place; = {t1 }; = {1, 0, 0, 1}; = {0, 0.2, 0.5};
Membership function for an operator’s salary.
α(w1 ) = {(the effect on disassembly time is small, 0.75), (the labor cost is high, 0.7), (the bad effect on disassembly quality is small, 0.8)}; λ = {1/6}.
III. D ISASSEMBLY P ROCESS P LANNING Currently, disassembly automation level is very low. Because of a large amount of human intervention involved in this process, the competence and motivation of personnel have major impact on the efficiency and quality of disassembly. The skill level of human operators then becomes an imperative factor in determining the sequence of disassembly operations. In an FAPN, an operator place is introduced as an input to a transition. Considering the impact of a human operation on disassembly cost and revenue, not only are fuzzy attributes ρi (wj ) introduced to represent human factors, but also a truth degree is associated with each attribute for the strength of the impact. For a particular disassembly operation t, the disassembly time is calculated as the sum of operation time and extended time, where the latter is a function of an operator’s skill. The higher the skill of an operator, the shorter the extended time. As the truth degree drops with increasing extended time, the relationship between the truth degree and the extended time is assumed to be exponential [21]. The extended time Text (wj ) is then calculated as Text (wj ) =
− ln (σρ1 (wj )) β
(3.1)
where β is a constant whose value is chosen based on statistics, and σρ1 (wj ) is the truth degree of the attribute ρ1 (wj ). Based on (3.1), the disassembly time for the disassembly operation t Tdis (t) is Tdis (t) =
− ln (σρ1 (wj )) + λ(t), β
wj ∈• t ∩ wj ∈ W. (3.2)
Labor cost is also an important factor in determining the final disassembly cost. The more experienced operators are, the higher the salary they get. This paper assumes that the fuzzy proposition—the labor cost is high—is characterized as the function of operator’s salary. As an example, as shown in Fig. 2, the maximum and minimum salaries are $18/h and $6/h, respectively.
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
721
represents the cost/revenue when place • t ∈ Q − Q takes t for further disassembly. c(t) and r(p) are calculated using (3.3) and (3.4), respectively. After the bottom-up calculation, the FAPN is traversed, starting from p1 . Through this exhaustive search, the optimal disassembly path of a discarded product can be obtained by Algorithm 3.1. Fig. 3 summarizes the overall logic of the approach.
Fig. 3.
Logic of optimal disassembly planning.
Therefore, the final disassembly cost for a particular disassembly operation t, which is c(t), is calculated as (assuming that the charge for disassembly tools is ignored) c(t) = Tdis (t) (6 + 12σρ2 (wj ))
(3.3)
where σρ2 (wj ) is the truth degree of the attribute ρ2 (wj ). Considering the damage to a disassembled subassembly/ component made by an unskilled operator, the adjusted revenue for each place p, p ∈ Q, is also introduced and denoted as r(p) =
σρ3 (w)τ (p), τ (p),
p ∈ (w• )• p = p1
(3.4)
where σρ3 (w) is the truth degree of the attribute ρ3 (wj ). ∀p ∈ Q − {p1 }, r(p) may take different values as p results from separate disassembly operations processed by different operators w. Finally, the disassembly-planning algorithm with the consideration of human factors is proposed. The aim is to identify the order of disassembly operations (i.e., transitions firing in an FAPN) that maximizes the expected return from the processed items. This objective is then formally expressed by introducing the disassembly value function d : Q → {0} ∪ + , which is associated with the disassembly plan and maps each place p ∈ Q to a non-negative value representing the expected return obtained from the processing of the product/ subassembly/component represented by place p. Eventually, computing an optimal disassembly plan is restated as the determination of a disassembly sequence that maximizes d(p1 ). According to the acyclic structure of the FAPN model defined in Section II, the maximization of d(p1 ) can be achieved through a recursive computation that starts from the leaf places and derives the optimal d-values for every place p ∈ Q. The detailed calculation is defined as follows. Definition 3.1: The disassembly value of place p is denoted as d(p) and is calculated in a recursive manner using 1) ∀p ∈ Q , d(p) = r(p) and 2) ∀p ∈ Q − Q , d(p) = max{r(p), maxt∈p• ∆(t)}, where ∆(t) = q∈t• ∩q∈Q d(q) − c(t) is the margin benefit of transition t, t ∈ T , which
Algorithm 3.1: Step 1. For p ∈ Q , which is the set of leaves according to Definition 2.1, set d(p) = r(p). Step 2. Set L = Q − Q . While (L = ∅) do: Find a place p ∈ L such that ∀q ∈ (p• )• , q ∈ L ∩ q ∈ Q. Calculate ∆(t) = d(q) − c(t) q∈t•
∆(t) . d(p) = max r(p), max •
t∈p
Let L = L − {p}, i.e., remove p from L. End. Step 3. Set Z = {p1 } (assuming that p1 is the root node for every incoming product), DP = ∅, and C = ∅ (DP represents final disassembly path, and C represents the last components after disassembly). While (Z = ∅) do: For each node p in Z: a) Find a transition t, where t ∈ p• , ∆(t) = d(p) = r(p). b) If a) succeeds, DP = DP ∪ {t}, and Z = Z ∪ {t• }; otherwise, C = C ∪ {p}. c) Z = Z − {p}. End. Due to the deterministic features of the FAPN (i.e., the operation time associated with each transition, the revenue function associated with each place in Q, and the truth degree of each attribute associated with a worker place are assumed known a priori), the exhaustive search in Algorithm 3.1 guarantees the optimal disassembly planning. In reality, uncertainty exists in product condition and judgmental estimation of worker skill that is extremely critical in determining disassembly costs and profits. Therefore, uncertainty management methods are needed to ensure this optimization in practice, which is out of the scope of this paper and is not addressed here. Proposition 1: In an FAPN, the number of leaf places is less than or equal to n + 1, where n is the number of transitions. Proof: The number of places p ∈ Q and transitions varies from one product to another since it depends on how the components in the product are interconnected. Taking an extreme case into consideration where there are m − 1 interconnections between the m components and there is only one disassembly sequence feasible, the number of leaf places is m = n + 1, under the assumption that each transition disassembles one component at time. As the number of interconnection between
722
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
components and possible disassembly sequences increase, more transitions are introduced into the FAPN, and the number of leaf places is unchanged. Therefore, Proposition 1 is proven. Proposition 2: The algorithm complexity is O(n), where n is the number of transitions in an FAPN. Proof: Let |Y | be the cardinality of set Y . 1) In Step 1, the disassembly value of each leaf place is calculated. From Proposition 1, the number of leaf places is less than or equal to n + 1. Therefore, the computation runs at O(n). 2) In Step 2, the calculation is based on each nonleaf place. For each such place pi ∈ Q − Q , the marginal benefits of its output transitions ∆(t), t ∈ p•i are required to be calculated and compared with its adjusted revenue r(pi ). The computation runs at O(2|p•i |) in the worst case. Since leaf places do not have output transitions and ∀pi , pj ∈ • n. Therefore, Q − Q , p•i ∩ p•j = ∅, pi ∈Q−Q |pi | = the total computational complexity is O(2 pi ∈Q−Q p•i ), i.e., O(n). 3) In Step 3, in order to find the disassembly sequence with the best cost/revenue ratio, the traversal of the FAPN is needed, which runs at O(n) at worst case. Hence, the computational complexity of the algorithm is O(n). IV. C ASE S TUDY To fully understand the above concepts and algorithm, this paper designs and implements a disassembly system and uses an obsolete personal computer (PC) as a disassembly example. Section IV-A gives the design and implementation method, and Section IV-B presents the experimental results. A. Implementation Method By using the developed FAPN and algorithm, this paper proposes the following design and implementation steps in a disassembly system where a given number of machines/workers form a disassembly line to dismantle one type of products. 1) Construct an FAPN given the product information, all feasible disassembly choices, and workers’ profiles (e.g., functionality, skill level, etc.). 2) Collect and associate all the data with places and transitions in the FAPN. 3) Update the data associated with places and transitions by considering human factors through (3.3) and (3.4). 4) Integrate the developed algorithm (i.e., Algorithm 3.1) with controllers and other hardware in the disassembly system. The Pentium II PC used in this experiment consists of 11 main components, i.e., a cover case {j}, cables {k}, a power supply {h}, a motherboard {a}, expansion cards {e, f, g}, secondary storage {b, c, d}, and memory chips {i}. Its layout without the case is given in Fig. 4, showing how components are connected to each other. This PC will go through some or all of the four stages in the prototypical disassembly system
Fig. 4.
Layout of a Pentium II PC.
according to the product condition. In Stage 1, the case and cables are separated from the PC. In Stage 2, the secondary storage is taken apart. The disassembly of the expansion cards and the memory chips and the power supply is processed in Stages 3 and 4, respectively. A worker is assigned to each stage, i.e., wj is for stage j, j = 1, 2, 3, and 4. To construct an FAPN, the entire product is labeled by p1 . All feasible ways to disassemble the product into the next-level subassemblies or components are then identified. Meanwhile, a corresponding worker place is added as an input to each transition. For instance, at p1 , only one method is recognized and represented by t1 , which means that other components can be disassembled only after unscrewing the case and taking out all cables. w1 is identified as the corresponding worker place for t1 . This process is repeated for each subassembly until no subassembly is left. Finally, the complete FAPN is obtained, as shown in Fig. 5. To simplify the presentation while keeping clarity, operator places wi (i = 1 to 4) are shown more than once. Some individual components are grouped together and represented by a single place, i.e., p9 is for {j, k}, p12 for {b, c, d}, and p13 for {e, f, g, i}. Second, the relevant data is collected and associated with places and transitions. An adjustment is then made in the light of impact of human operations. Finally, the proposed algorithm is implemented in a modular structure written in Visual C++, which includes Database, Process Planner, and Controller, and runs under the Windows 2000/XP operating system with the Access database engine. Database provides the relevant information of products (e.g., reuse/resale values and operation time for corresponding disassembly tasks) and workers (i.e., skill level). In this paper, the worker skill level is characterized by technical evaluation score and salary. The evaluation is based on a worker’s experience, performance, and knowledge. The truth degrees of the three attributes associated with each worker are then decided based on his/her evaluation score and salary through membership functions. During the implementation, this information in Database is determined according to the previous benchmark experience and created as follows. 1) In an FAPN, the revenue value associated with place p ∈ Q and the operation time for each disassembly task represented by t are randomly populated with a uniform
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
Fig. 5.
723
Example of FAPN for a PC.
distribution, which have the range [0, 1.5] and [0.01, 0.1], respectively. 2) The score of each worker is also randomly chosen with a uniform distribution between [60, 100]. 3) The worker’s salary is proportional to his/her score with the range [6, 18]. 4) β is chosen to be 100. Based on the information from Database, Process Planner presents the Controller the disassembly sequence of an obsolete product, on which Controller dispatches tasks to different disassembly stages.
Worker
Score
Salary ($/h)
w1 w2 w3 w4
86.2 66 96 70
13.8 7.5 16.8 9
2) FAPN = (P, T, I, O, M, τ, α, λ) P = {q1 , q2 . . . q13 , w1 , w2 , w3 , w4 } T = {t1 , t2 . . . t13 }
B. Experimental Results In this section, the value gain in a disassembly plan DP [16] f (DP) is introduced first. A case study is then performed to illustrate the idea of disassembly process planning via the disassembly system and the obsolete PC discussed above. Definition 4.1: The value gain in a disassembly plan DP f (DP) is f (DP) =
1) Worker information (i.e., score and salary)
p∈C
r(p) −
c(t).
(4.1)
t∈DP
The first set of experiment is focused on the functional structure of the planner and on the detailed operation of the software modules. When the obsolete PC comes to the disassembly facility, its FAPN is modeled and the corresponding data is collected from Database
m0 (p1 ) = 1;
m0 (pi ) = 0,
i = 2, . . . , 13
m0 (w1 ) = m0 (w2 ) = m0 (w3 ) = m0 (w4 ) = 1 τ = {0, 0, 0, 0, 0, 0, 0, 0, 0.9, 0.25, 0.5, 0.9, 1.2} α(w1 ) = {(the effect on disassembly time is small, 0.7) (labor cost is high, 0.65) (the bad effect on disassembly quality is small, 0.8)} α(w2 ) = {(the effect on disassembly time is small, 0.25) (labor cost is high, 0.15) (the bad effect on disassembly quality is small, 0.2)}
724
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
α(w3 ) = {(the effect on disassembly time is small, 0.9) (labor cost is high, 0.9) (the bad effect on disassembly quality is small, 1)} α(w4 ) = {(the effect on disassembly time is small, 0.3) (labor cost is high, 0.25) (the bad effect on disassembly quality is small, 0.2)} λ = {0.04, 0.06, 0.06, 0.04, 0.04, 0.05, 0.05 0.03, 0.09, 0.05, 0.07, 0.08, 0.04}. Based on the worker profiles, the disassembly cost for each transition and the revenue for each place in Q are updated c(t) = {0.6072, 0.6480, 1.0248, 0.4680, 0.4680 0.4992, 0.8568, 0.3780, 0.8112, 0.8568 0.6552, 1.3608, 0.4212} r(p) = {0, 0, 0, 0, 0, 0, 0, 0, 0.72, 0.05, 0.5 σρ3 (wj ), 0.18, 1.2} where r(p11 ) is not fixed since it may be processed by different workers wj via different disassembly sequences. While executing Algorithm 3.1 through the FAPN, the disassembly value for each p and the marginal benefit of each transition t are calculated. Finally, the optimal disassembly path for this PC is obtained. This disassembly sequence is referred to as DP1 ∆(t) = {0.456, −0.230, 0.175, 0.425, −0.418, −0.219, 0.843 − 0.228, −0.206, 0.343, 0.368, −0.161, −0.241} d(p) = {0.456, 0.343, 0.368, 0.425, 0, 0, 0.843, 0, 0.72 0.05, 0.5σρ3 (wj ), 0.18, 1.2} DP1 = {t1 , t10 } and f (DP1 ) = 0.456. If human factors were ignored in Process Planner, the reuse/resale value τ and the operation time λ would replace the adjusted r(p) and c(t) to calculate the disassembly value for each p and the marginal benefit for each transition t [22]. This would result in a different disassembly sequence referred to DP2 = {t1 , t10 , t5 , t6 }. However, t5 and t6 are handled by inexperienced workers w4 and w2 , respectively, in the real disassembly cycle. These human operations degrade the reuse/resale values of the disassembled components from t5 and t6 and causes longer disassembly time than what is used in Process Planner. Therefore, the value gain of this plan is decreased f (DP2 ) = −0.181. Fig. 6(a) gives the comparison of the disassembly value of this plan versus that of the DP1 as a function of the disassembly steps.
Fig. 6. Experimental results. (a) Process plan DP2 without considering human factors versus an adapted plan DP1 with human factors. (b) Modified process plan DP3 by enhancing the competence of a worker verse the adapted one DP1 .
Taking human factors into consideration, it is found that the efficient way to disassemble this PC is to retrieve only p9 and p13 , and leave p5 unchanged. Although there are two options to decompose p5 , t5 , and t13 , the operations handled by two unskilled workers w4 and w2 result in more cost than revenue. If the competence of operators (e.g., w2 ) would be enhanced to the same level as w1 , the extended time needed for t13 will be decreased, and the bad effects on the quality of two disassembled components ( p8 and p12 ) will be minimized. Consequently, the marginal benefit of t13 increases from −0.241 to 0.113, which results in another disassembly sequence referred to as DP3 = {t1 , t10 , t13 } with better value gain f (DP3 ) = 0.569. The comparison of disassembly value of DP3 with that of DP1 is given in Fig. 6(b). It is clear that the performance of an operator significantly affects the disassembly profit, which is a key factor in decision making.
V. C ONCLUSION Disassembly is currently labor intensive. The large amount of human intervention gives rise to many uncertainties in disassembly operations. Ignoring these human factors results in surprisingly few applications of current approaches for disassembly planning. To address this deficiency, this paper proposes a fuzzy attributed Petri net (FAPN) model where the influence of human factors on disassembly (e.g., disassembly time, quality of disassembled components, and labor cost) is mathematically represented as membership functions. An algorithm is then developed upon this model to derive the
TANG et al.: FUZZY-PETRI-NET-BASED DISASSEMBLY PLANNING CONSIDERING HUMAN FACTORS
cost-effective disassembly sequence of a discarded product. To the authors’ knowledge, no paper has comprehensively dealt with human factors in disassembly. This work successfully overcomes a deficiency in [16] and [22], which ignores the impact of human operations (an indispensable part of the whole disassembly system), thereby making the proposed methodology more applicable to real industrial settings. From the example, it was found that the performance of human operators has significant influence in disassembly decision making. The enhanced competence, motivation, and stability of human operators have a major role in improving productivity and efficiency of disassembly. The developed model can be adapted and matched to different operator–work–environment conditions by simply shifting its membership functions to provide more flexibility for cost/time estimators. The research can be extended in several directions. For instance, the proposed FAPN model uses the deterministic nature of the assumptions characterizing the associated revenue function and the time function. In reality, these involved costs and profits in disassembly will be determined dynamically due to the prevailing condition of discarded products. Thus, it is necessary and challenging to refine the proposed model with probabilistic parameters for product uncertainty, on which a learning-based algorithm would be developed, and exploit the past “knowledge” regarding the uncertainty towards improving future decisions [14]. More factory data (i.e., actual profit and cost figures in real-world problems) also need to be used to validate the proposed methodology in the future. The refinement of the proposed model to efficiently analyze the impact of physical, psychological, personal, and other characteristics of humans [6] in complex disassembly systems is another interesting issue.
ACKNOWLEDGMENT
725
[8] ——, “Disassembly sequencing: A survey,” Int. J. Prod. Res., vol. 41, no. 16, pp. 3721–3759, 2003. [9] X. O. Li, W. Yu, and F. Lara-Rosano, “Dynamic knowledge inference and learning under adaptive fuzzy Petri net framework,” IEEE Trans. Syst., Man, Cybern. C, vol. 30, no. 4, pp. 442–450, Nov. 2000. [10] A. Meacham, R. Uzsoy, and U. Venkatadri, “Optimal disassembly configurations for single and multiple products,” J. Manuf. Syst., vol. 18, no. 5, pp. 311–322, 1999. [11] K. E. Moore, A. Gungor, and S. M. Gupta, “Petri net approach to disassembly process planning for products with complex AND/OR precedence relationships,” Eur. J. Oper. Res., vol. 135, no. 2, pp. 428–449, 2001. [12] J. Mikler, H. Hådeby, A. Kjellberg, and G. Sohlenius, “Towards profitable persistent manufacturing human factors in overcoming disturbances in production systems,” Int. J. Adv. Manuf. Technol., vol. 15, no. 10, pp. 749–756, 1999. [13] M. Rahimi, P. A. Hancock, and A. Majchrzak, “On managing the human factors engineering of hybrid production systems,” IEEE Trans. Eng. Manage., vol. 35, no. 4, pp. 238–249, Nov. 1988. [14] S. Reveliotis, “Uncertainty management in optimal disassembly planning through learning-based strategies,” in Advances in Manufacturing, Logistics and Supply Chain Management, N. R. S. Raghavan et al., Ed. Bangalore, India: IIS, 2003, pp. 135–141. [15] K. K. Seo, J. H. Park, and D. S. Jang, “Optimal disassembly sequence using genetic algorithms considering economic and environmental aspects,” Int. J. Adv. Manuf. Technol., vol. 18, no. 5, pp. 371–380, 2001. [16] Y. Tang, M. C. Zhou, and R. Caudill, “An integrated approach to disassembly planning and demanufacturing operation,” IEEE Trans. Robot. Autom., vol. 17, no. 6, pp. 773–784, Dec. 2001. [17] Y. Tang, M. C. Zhou, E. Zussman, and R. Caudill, “Disassembly modeling, planning, and application,” J. Manuf. Syst., vol. 21, no. 3, pp. 200–217, 2002. [18] M. K. Tiwari, S. K. Mukhopadhyay, N. Sinha, S. Kumar, and R. Rai, “A Petri net based approach to determine the disassembly strategy of a product,” Int. J. Prod. Res., vol. 40, no. 5, pp. 1113–1129, 2002. [19] L. V. Utkin, S. V. Gurov, and I. B. Shubinsky, “Analysis of computer integrated manufacturing systems by fuzzy human operator behavior,” J. Qual. Maint. Eng., vol. 3, no. 3, p. 189, 1997. [20] R. R. Wu, L. Ma, J. Mathew, and G. H. Duan, “Optimal operation planning using fuzzy Petri nets with resource constraints,” Int. J. Comput. Integr. Manuf., vol. 15, no. 1, pp. 28–36, 2002. [21] J. X. Xu, “Fuzzy Petri net-based optimum scheduling of FMS with human factors,” in Proc. 35th Conf. Decision Control, Kobe, Japan, Dec. 1996, pp. 4451–4452. [22] E. Zussman and M. C. Zhou, “Design and implementation of an adaptive process planner for disassembly processes,” IEEE Trans. Robot. Autom., vol. 16, no. 2, pp. 171–179, Apr. 2000.
The authors would like to thank the comments provided by anonymous reviewers and the Associated Editor, which greatly helped the presentation of this paper. R EFERENCES [1] T. Cao and A. C. Sanderson, “Task sequence planning using fuzzy Petri nets,” IEEE Trans. Syst., Man, Cybern., vol. 25, no. 5, pp. 755–768, May 1995. [2] M. Gao, M. C. Zhou, X. Huang, and Z. Wu, “Fuzzy reasoning Petri nets,” IEEE Trans. Syst., Man, Cybern. A, vol. 33, no. 3, pp. 314–324, May 2003. [3] M. Gao, M. C. Zhou, and Y. Tang, “Intelligent decision making in disassembly process based on fuzzy reasoning Petri nets,” IEEE Trans. Syst., Man, Cybern. B, vol. 34, no. 5, pp. 2029–2034, Oct. 2004. [4] M. A. Goodrich and E. R. Boer, “Model-based human-centered task automation: A case study in ACC system design,” IEEE Trans. Syst., Man, Cybern. A, Syst., Humans, vol. 33, no. 3, pp. 325–336, May 2003. [5] A. Hula, K. Jalali, K. Hamza, S. J. Skerlos, and K. Saitou, “Multi-criteria decision-making for optimization of product disassembly under multiple situations,” Environ. Sci. Technol., vol. 37, no. 23, pp. 5303–5313, 2003. [6] H. Jahan-Shahi, E. Shayan, and S. H. Masood, “Multivalued fuzzy sets in cost/time estimation of flat plate processing,” Int. J. Adv. Manuf. Technol., vol. 17, no. 10, pp. 751–759, 2001. [7] A. J. D. Lambert, “Determining optimum disassembly sequences in electronic equipment,” Comput. Ind. Eng., vol. 43, no. 3, pp. 553–575, 2002.
Ying Tang (S’99–M’02) received the B.S. and the M.S. degrees from the Northeastern University, Shenyang, China, in 1996 and 1998, respectively, and the Ph.D. degree from the New Jersey Institute of Technology, Newark, in 2001. She is currently an Assistant Professor of electrical and computer engineering at Rowan University, Glassboro, NJ. Her research interests include modeling and scheduling of computer-integrated systems, Petri nets and applications, artificial intelligence, reconfigurable systems design, hardware and software co-design, software security, and networking and communication. She has over 30 publications in international journals, conference proceedings, and book chapter. Dr. Tang is a member of Sigma Xi.
726
IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 36, NO. 4, JULY 2006
MengChu Zhou (S’88–M’90–SM’93–F’03) received the B.S. degree from Nanjing University of Science and Technology, Nanjing, China, in 1983, the M.S. degree from Beijing Institute of Technology, Beijing, China, in 1986, and the Ph.D. degree in computer and systems engineering from Rensselaer Polytechnic Institute, Troy, NY, in 1990. He joined New Jersey Institute of Technology (NJIT), Newark, in 1990 and is currently a Professor of electrical and computer engineering. He has been invited to conduct lectures in Australia, Canada, China, France, Germany, Hong Kong, Italy, Japan, Korea, Mexico, Taiwan, and the US. His research interests are in computer-integrated systems, Petri nets, semiconductor manufacturing, multilifecycle engineering, biosignal, and sensor network. He is the author, coauthor, and editor of several books and many published articles. Dr. Zhou served as an Associate Editor of the IEEE TRANSACTIONS ON R OBOTICS AND A UTOMATION from 1997 to 2000, and currently, he is an Associate Editor of the IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND E NGINEERING and the Managing Editor of IEEE T RANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: PART C. He organized and chaired over 70 technical sessions and served on program committees for many conferences. He was the Program Chair of the 1998 and Cochair of the 2001 IEEE International Conference on Systems, Man, and Cybernetics (SMC) and the 1997 IEEE International Conference on Emerging Technologies and Factory Automation. He was also a Guest Editor of the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, the IEEE TRANSACTIONS ON S EMICONDUCTOR M ANUFACTURING , and the IEEE T RANSACTIONS ON MECHATRONICS. He is Editor-in-Chief of the International Journal of Intelligent Control and Systems since 1996. He was the General Cochair of the 2003 IEEE International Conference on Systems, Man, and Cybernetics, Washington D.C., October 5–8, 2003, and the 2004 IEEE International Conference on Networking, Sensing and Control, Taipei, Taiwan, March 21–23, 2004. He has led or participated in 26 research and education projects with a total budget of over $10M, funded by the National Science Foundation, Department of Defense, Engineering Foundation, New Jersey Science and Technology Commission, and industry. He is a member of the Board of Governors of the IEEE SMC Society. He is a life member of the Chinese Association for Science and Technology-USA and served as its President in 1999. He is a recipient of the NSF’s Research Initiation Award, CIM University-LEAD Award by the Society of Manufacturing Engineers, the Perlis Research Award by NJIT, the Humboldt Research Award for US Senior Scientists, the Leadership Award and Academic Achievement Award by the Chinese Association for Science and TechnologyUSA, the Asian American Achievement Award by the Asian American Heritage Council of New Jersey, and the Outstanding Contribution Award by the IEEE Systems, Man, and Cybernetics (SMC) Society. He was the founding Chair of the Discrete Event Systems Technical Committee of the IEEE SMC Society and the Cochair (founding) of the Semiconductor Factory Automation Technical Committee of the IEEE Robotics and Automation Society.
Meimei Gao (S’02–M’03) received the B.S. and M.S. degrees from the Northwestern Polytechnic University, Xi’an, China, in 1994 and 1997, respectively, the Ph.D. degree in control theory and engineering from Shanghai Jiao Tong University, Shanghai, China, in 2000, and the Ph.D. degree in computer engineering from New Jersey Institute of Technology, Newark, in 2003. She is currently an Assistant Professor with the Department of Mathematics and Computer Science, Seton Hall University, South Orange, NJ. Her research interests include discrete-event systems, Petri nets and applications, artificial intelligence, decision making, computer-integrated manufacturing and demanufacturing, and lifecycle engineering. She has published over 20 journal and conference proceedings papers in the above research areas. Dr. Gao is a member of ACM.
