Вы находитесь на странице: 1из 19

This article was downloaded by: [Maulana Azad National Institute of Tech] On: 25 April 2013, At: 19:39

Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Production Research


Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/tprs20

Concurrent engineering: decomposition and scheduling of design activities


ANDREW KUSIAK & KWANGHO PARK
a a b

Department of Industrial Engineering, College of Engineering, The Universityof Iowa, Iowa City, Iowa, 52242, U.S.A.
b

SAMSUNG Data Systems, Samsung Building, 50, 1-Ka, Eulchi-Ro, Chung-Ku, Seoul, Korea. Version of record first published: 27 Apr 2007.

To cite this article: ANDREW KUSIAK & KWANGHO PARK (1990): Concurrent engineering: decomposition and scheduling of design activities, International Journal of Production Research, 28:10, 1883-1900 To link to this article: http://dx.doi.org/10.1080/00207549008942841

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

INT. J. PROD. RES.,

1990,

VOL.

28,

NO.

10, 1883~1900

Concurrent engineering: decomposition and scheduling of design activities


ANDREW KUSIAKt and KWANGHO PARK:!: Design of complex products or large scale systems involves hundreds of resources (designers,analysts, computers, software systems, and procedures) and thousands of design activities. In this paper, a methodology is presented for decomposition of the design task into activities and modules. The methodology is based 01' clustering of design activities into groups that allow effective organization 0' resources required in the design process. A knowledge-based approach is used for managing design activities. The system performs analysis aiming at exploring concurrency and reducing the design project makespan.

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

1. Introduction
The traditional design of products and components has relied on an iterative approach. The main difficulty with the iterative approach is that it is time consuming and many iterations are required before a design project is completed. Prior to actual design, the product is analysed for at least two constraints: financial marketing

The design is typically analysed subject to the following constraints: quality reliability maintainability manufacturability schedulability safety In traditional design, the preceding constraints are considered sequentially. In order to reduce the design cycle, concurrent (simultaneous) design has been introduced. The basic idea of concurrent design is to shorten the time horizon in which the design constraints are introduced. Concurent engineering refers to the integration of various activities within the broad scope of the product life cycle (Nevins and Whitney 1989, and Vasilash 1987). In concurrent engineering, the product design is viewed as a strategic task that has a major effect on the subsequent production related activities. Design of products determines their quality and 70 to 80% of the final production

Received January 1990. t Department of Industrial Engineering, College of Engineering, The Universityof Iowa, Iowa City, Iowa 52242, U.s.A. t SAMSUNG Data Systems, Samsung Building, 50, l-Ka, Eulchi-Ro, Chung-Ku, Seoul, Korea.
0020-7543/90

sroo 1990 Taylor &

Francis Ltd.

1884

A. Kusiak and K. Park

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

cost (O'Grady et al. 1988). Another study included in Nevins and Whitney (1989) shows that about 70% ofthe life cycle cost of a product is determined at the conceptual design stage. Product design in concurrent engineering is viewed as a teamwork approach (Whitney 1988). It requires a great deal of co-ordination, a flexible organization, and utilization of the most advanced design and manufacturing technology. In this paper, the 'concurrency aspect of design is explored. In section 2, a decomposition concept is presented that allows a set of design activities to cluster into groups of activities. Then a knowledge-based system for managing the design project is presented in section 3. A numerical example demonstrating the system performance is presented in section 4. Conclusions are discussed in section 5.

2. System decomposition A typical product or system (mechanical, electrical, etc.) can be decomposed into subsystems and these in turn into modules. Design of each module involves a set of design activities. Similar or identical activities may be performed in the design of different modules. To clarify the proposed concept, consider a hypothetical system shown in Fig. 1. The system in the figure involves three subsystems, seven modules, and ten activities. In this example, activity 2 is involved in the design of modules I, 3, and 6. The interaction between modules and activities illustrated in Fig. I can be represented as the module-activity incidence matrix [au] in (I);

Activity

10

M
0

2
3 4 5

[au]

d u

(I)

where jfactivity j is involved in design of module i otherwise. The matrix (I) is nonstructured, which means that it has no special properties in terms of concurrency of activities during the design process. The grouping problem to be solved is to maximize the number of mutually separable clusters in the

Concurrent engineering

1885

Subsystems

Modules
2 3 4

5
3,7,8,9

7 1,9

1,2,4,6

5,9,10

2,9

3,8,9

1,2,4

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

Figure 1. The hierarchical structure of a system. module-activity incidence matrix, subject to constraints, e.g., limited size of some or all clusters and inclusion of specific activities in some clusters. Transformation of the nonstructured matrix [aij] into a structured (possibly diagonal) matrix is a computationally complex (NP complete) problem. Cluster analysis provides an underlying theory for solving the grouping problem. Solving the problem represented in matrix (I) with the clustering algorithm presented in Kusiak and Chow (1987) results in matrix (2). Activity

10

3
6 Module 7
(2)

2
5

Two .groups of modules GM-I = {I, 3, 6, 7}, GM-2 = {4, 2, 5}, and two groups of activities GA-I = {I, 2, 4, 6}, GA-2 = p, 8, 5, 1O,7} are visible in matrix (2). Activity 9 overlaps with the two modules obtained. Another solution generated by the cluster analysis algorithm is presented in matrix (3).

1886

A. Kusiak and K. Park Activity 3 4 7 8 2 4 6

10

[: :]

Module

6 3

[:

IJ
[I I]

(3)

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

2 7

Three groups of modules GM-l = {4, 5}, GM-2 = {I, 6, 3}, GM-3 = {2} and three groups of activities GA-I = {3, 7,8}, GA-2 = {2,4,6}, GA-3 = {5, 1O} are visible in matrix (3). Activities 9 and I overlap with the three groups of modules. Introducing additional resources for performing the two activities allows decomposition of matrix (3) into three mutually separable submatrices shown in (4). Activity 3 4 7 8

9'

I'

9"

10

9'ft

1"

[:

:]

Module

6 3 2 7.
1

l:

J [I

(4)

As shown in matrix (4), activity 9 has been triplicated (activities 9', 9", and 9''') while activity I has been duplicated (activity I' and I "). Rather than multiplying the two activities they could have been left as presented in matrix (3); however, a special consideration should be given while scheduling the design project. A group of activities in the module-activity incidence matrix identifies the modules that are to be designed and analysed by the required human, hardware, and software resources. 2.1. Definition of modules To define modules, consider an example of the vehicle (system) that can be decomposed into subsystems as shown in Fig. 2. Each subsystem can be further decomposed into new subsystems and modules. For example, the carriage unit can be decomposed as presented in Fig. 3.

Concurrent engineering
System Passenger Vehicle

1887

Figure 2. The passenger vehicle and its subsystems.

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

2
I

3 I

4
I

5
I

6 I

7
I

Modules

Figure 3. The carriage unit and its decomposition into subsystems and modules.

2.2. Definition of design activities At the aggregate level, design activities and procedures involved in the process of a vehicle design (see Fig. 2) can be classified as follows:

Reliability analysis-determination of the size and form of vehicle components, testing procedures, structural analysis. Functional analysis-occupant comfort, occupant entrance and exit, travel speed analysis, acoustic and vibration analysis. Styling-form and harmony analysis. Economic analysis-weight analysis, fuel efficiency, wear lifeanalysis, inspection and maintenance analysis, repair cost analysis. Assembly analysis-in-use inspection, maintenance and repair analysis, gravimetric, and pressworking analysis. Liability and safety-stability analysis, ergonomic analysis, illumination analysis, acceleration and braking capability analysis, federal, state, and industrial standards consideration. Road factors analysis-weight distribution and dimensions analysis, road grades, surface, and curvature analysis. Accident analysis-collision analysis, injury analysis, accident and injury avoidance analysis.

1888

A. Kusiak and K. Park

Of course, each of the preceding aggregate activities can be decomposed into a large number of detailed activities that are of concern in this paper. The total number of activities that will be grouped and scheduled may run in thousands. Efficient generation of these activities might be a subject of a separate research project.
2.3. Representation of design activities A typical representation scheme of activities is an 'activity on nodes' (AON) network. The network is assumed to contain unique dummy start and dummy finish activities, each with processing time equal to O. In AON representation, grouping activities can be viewed as breaking down of an original network into a number of subnetworks. Each subnetwork has associated nodes and arcs. Arcs connecting two different activities which belong to different subnetworks define relationships between subnetworks. 2.4. Advantages of decomposition of design activities Clustering of activities involved in the design process allows one to determine a potential group of activities that might be scheduled simultaneously. The degree to which the activities can be scheduled simultaneously depends on the quality of clusters and the nature of precedence constraints. Since the groups of activities in matrix (4) are mutually exclusive, in the absence of precedence constraints, the design activities belonging to this groups could be scheduled in parallel. The details of scheduling of design activities are presented in section 3 of this paper. Grouping of activities simplifies the process of generation of the precedence constraints among activities and duration of the activities. Since for a given group the number of activities to be considered by the expert designers and analysts is relatively small, the quality ofthe data generated is also improved. Another advantage of grouping of activities is simplification of the project scheduling and management. The activities within each group and their interaction with activities belonging to other groups are clearly defined. The number of activities for which the group management is responsible is relatively small. Application of the decomposition concept to the design problem in matrix (I) results in the following:

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

I. Separation of the overall design task into groups of modules and activities. 2. The group of modules do not have to correspond to the traditional organizational structures, for example vehicle body design group, transmission design group, etc. 3. Potential activities that might be performed simultaneously are detected. 4. Complexity of management of the design task is reduced. 5. Reduction of the design cycle. 6. Reduction of the computational time involved in scheduling of design activities. The latter advantage is discussed next. Scheduling of a number of subnetworks with a small number of activities is easier than scheduling the entire network. In order to obseve this advantage, a set of random test problems with sixteen activities was constructed. The activities and resources were decomposed into three independent subnetworks. In this set, the processing time of each activity was randomly generated in the interval [10,20]. The computation was performed on an Apple Macintosh SE microcomputer using a version of Common Lisp (Allegro Common Lisp). The algorithm used for computations is a heuristic

Concurrent engineering
Before decomposition (the entire network) Test problem I 2 3 4 5 T 92 83 110 94 103 CPU 1650 1517 1000 2117 0867

1889

After decomposition Group I T 63 70 63 68 65 CPU 0050 0050 0050 0050 0050 Group 2 T
92 83 110 94 103

Group 3 T 41 39 44 41 41 CPU 0083 0117 0150 0117 0083 Total CPU 0300 0367 0307 0350 0283

CPU 0167 0200 0107 0183 0150

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

Table 1. Comparison of CPU time. T is makespan and CPU is given as time in seconds on Macintosh SE.

algorithm developed in Kusiak and Park (1989). The CPU time and makes pan are reported in Table 1. One can see in the table that the CPU time is reduced due to the decomposition. 3. The knowledge-based system The clustering approach discussed in section 2 of this paper allows one to decompose a design project into groups of activities. For the scheduling and management of design activities, a knowledge-based approach is employed. An expert system is build based on the framework of decomposition of a large scale project. The two basic functions of the system are scheduling and network analysis. Scheduling of activities involved in a design project can be viewed as a project scheduling. Methods such as CPM (critical path method), PERT (project evaluation and review technique), and methodologies for constrained resource scheduling can be used (Kusiak and Park 1989, Badiru 1988). In order to analyse the scheduling network, the system uses a set of production rules.
3.1. The system architecture The expert system proposed in this paper has been implemented using the tandem architecture presented in Kusiak (1990). The system consists of the following components (see Fig. 4):

knowledge base database algorithm base inference engine

The basic functions of the inference engine are as follows:


1. It monitors the flow of the entire project. 2. It aids groups of activities to communicate with each other. 3. It provides the local management with the global information, for example the progress of activities within each group. 4. It makes decisions regarding resource allocations and overall performance balancing.

1890

A. Kusiak and K. Park


Knowledge base

n
Inference engine

..

n
Data base

Algorithm

base
. clustering

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

8 8

algorithm . scheduling algorithm

Figure 4. The architecture of the knowledge-based system. In performing the above functions, the interference engine exploits the domain-specific knowledge built into the system. The knowledge base containing production rules allows the inference engine to perform the scheduling analysis. The backward chaining is used as a reasoning strategy.
3.2. Network analysis 3.2.1. Type of network decomposition. In order to discuss the analysis that can be performed on the network of activities, consider a network including 16 activities. Assume that the design activities have been clustered into three groups of activities, GA-I = {al,a4,a5,a9}, GA-2 = {a2,a6,a7,a8,aI2,aI3}, and GA-3 = {aIO,all, a 14, a 15, a 16} (see Fig. 5). As a result of grouping activities, three different types of network decomposition can be identified:

1. Ideal decomposition: the original network is decoinposed into totally independent subnetworks.
A

Figure 5. Grouping activities in a project.

Concurrent engineering

1891

8
8

8
Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013
(aJ

(b)

(c)

Figure 6. Three types of decomposition of the network: (a) ideal decomposition; (b) moderate decomposition; (e) complicated decomposition.

2. Moderate decomposition: the original network is decomposed into subnetworks with moderate relationship between them (i.e. relatively low number of intergroup links). 3. Complicated decomposition: the original network is decomposed into subnetworks with strong relationship between them (i.e. large number of intergroup links).
Figure 6 shows the three types of network decomposition. In order to determine the type of network decomposition, production rules are used. Three production rules for defining the type of decomposition are presented next. Rule 1 IF

THEN
Rule 2 IF

there is no link between any pair of groups the decomposition is ideal number of links between groups GA-i and GA-} is greater than zero and less than the threshold value t the decomposition is moderate number of links between groups GA-i and GA-} is greater than or equal to the threshold value t the decomposition is complicated

THEN
Rule 3 IF

THEN

1892

A. Kusiak and K. Park

The threshold value t is a function of the group size measured with the number of activities. The knowledge-based system sets a default value for t which can be modified by the analyst.
3.2.2. Analysis of precedence constraints between groups of activities. For the ideal decomposition of a network, each group is completely independent and no analysis of relationship between groups is needed. In case of the moderate and complicated decomposition, analysis of the relationship between groups is required. Two types of intergroup relationship are identified:

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

trivial relationship: the interaction between groups is weak (trivial link) harmful relationship:the interaction between groups is strong (harmful link)

The two intergroup relationships are discussed next. Assume that each of the three groups of activities, GA-I, GA-2, GA-3, in Fig. 5 has precedence structure shown in Fig. 7. There are two intergroup links drawn with broken lines in the figure. Link rl (a2 -+ a9) is a trivial relationship because it is unlikely that delaying a2 affects the execution of a9. Link r2 (a14 -+ a3) is a harmful relationship because delay of a14 holds a3 and all of its successors, a7, a8, a 12,and a 13.

GA-l

GA-2

,,

GA-3

- - - ..

: intergroup link

Figure 7. Three groups of activities GA-l, GA-2, GA-3 and intergroup relationships.

Concurrent engineering

1893

The production rules allowing to determine the type of intergroup relationship are as follows: Rule 4 there is a link from activity k of group GA-i to activity 1of group GA-j IF the local finish time of k is greater than the local start time of 1 AND the relationship is harmful THEN Rule 5 IF there is a link from activity k of group GA-i to activity 1of group GA-j the local finish time of k is less than or equal to the local start time of 1 AND THEN the relationship is trivial
Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

The knowledge-based system provides the information regarding the entire project network, namely: 1. The schedule generated by the scheduling algorithm (earliest start time, latest start time, earliest finish time, and slack time). 2. The status of each activity (start or finish or in progress). 3. The normality of each activity (whether an activity is on time schedule. 4. The availability of resources and their utilization at a particular moment. The intergroup relationship defines communications between groups. For example, the group of activities GA-l should communicate with that of GA-2 to check whether activity a2 has been completed before activity a9 is executed. The group of activites GA-2 should communicate with that of GA-3 to check whether activity a 14 has been completed in order to execute activity a3. However, GA-l and GA-3 do not have to communicate with each other. From the analysis of relationships among groups, they are ranked based on their impact on the scheduling performance; for example a makespan. Also, within the group, activities are ranked on the same basis. In terms of the strength of precedence constraints, including intergroup links, two types of precedence constraints can be identified: strong constraint: the precedence constraint which must be obeyed weak constraint: the precedence constraint which may be removed

In analysing the intergroup relationship, the type of precedence constraints' is particularly important. In order to execute concurrently the groups of activities, the intergroup links have to be weakened. The latter needs to be applied especially for harmful links. In the scheduling process, a strong constraint, k -+ I, must be satisfied. The only way to 'weaken' the strong constraint is to speed up the execution of activity k by allocating more resources for performing this activity. The production rules dealing with these aspects are formulated as follows: Rule 6 IF AND THEN Rule 7 IF AND THEN precedence constraint k -+ 1 is strong the link is harmful reduce the duration of activity k the strategy is to reduce the duration of activity k the unused quantity of resource r which can be allocated to activity k is ur execute the function 'reduction (k, r, ur)'

1894

A. Kusiak and K. Park

In production rule 7 presented above, the function 'reduction (k, r, ur)' computes the reduction in the duration of activity k due to the atlocation of ur quantity of resource r to perform activity k, as a result of the capacity analysis. A weak constraint is a precedence constraint which can be completely removed. If an intergroup link which connects two groups is a weak constraint, then the concurrent execution of two groups would be possible. A production rule representing this scheduling knowledge is expressed as fotlows: Rule 8 IF AND THEN precedence constraint k .... / is weak the link is harmful execute concurrently activities k and / (ignore link k .... /)

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

4.

Numerical example Analyse the intergroup relationships among three GAs in Fig. 7, given durations of activities shown in Table 2. The local schedules generated by CPM are summarized in Figs 8-10. The global schedule generated by CPM for the entire network presented in Fig. 7 is summarized in Fig. I I. Note that the activities in the global schedule belong

Group GA-I GA-2 GA-I GA-I GA-2 GA-2 GA-2 GA-2

Activity al a4 a5 a9 a2 aJ a6 a7

Duration
10

Group GA-2 GA-2 GA-2 GA-J GA-J GA-J GA-J GA-3

Activity a8 al2 alJ alO all al4 al5 al6

Duration 10 15 10 15
10

15
10

15 15 10 15
10

10 15 15

Table 2. Durations of activities.

GA-l

Activity al Earliest-start-time Duration Finish-time Latest-start-time Slack-time Makespan 0 10 10 0 0 a4 10 15 25


10

a5 10 10 20 15 5

a9 25 15 40 25 0

= 40

Critical path

= a I ---. a4 ---. a9

Figure 8. Local schedule for GA-l.

Concurrent engineering

1895

GA-2

Activity a2 a3 0 10 10 15 15 a6 15 15 30 15 0 Critical path a7 15 10 25 20 5 a8 10


10

al2 30 15 45 30 0

a13 25
10

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

Earliest-start-time Duration Finish-time Latest-start-time Slack-time

0 15 15 0 0 Makespan

20 25 15

35 35
10

= 45

= a2 --+ a6 --+ al2

Figure 9.

Local schedule for GA-2.

GA-3

Activity alO Earliest-start-time Duration Finish-time Latest-start-time Slack-time Makespan 0 15 15 5 5 all 0 15 15 0 0 Critical path al4 15
10

al5 15
10

al6 15 15 30 15 0

25 20 5

25 20 5
--+ al6

= 30

= a II

Figure 10. Local schedule for GA-3.

to the same groups as the local schedules while the parameters of each local schedule have changed. The critical path for the network in Fig. 7 is alO -+ al4 -+ a3 -+ a7 -+ al2 with the corresponding makespan of 60. Figure 12 shows the Gantt chart of the global schedule. One notes in Fig. 12 that links al4 -+ a3 and a2 -+ a9 prohibit concurrent execution of the three groups of activities. From production rule 4, link al4 -+ a3 is harmful because (the local finish time of a14 = 25) is greater than (the local start time of al3 = 0). From production rule 5, link a2 -+ a9 is a trivial link because (the local finish time of a2 = 15) is less than (the local start time of a9 = 40). Thus the link a14 -+ a3 only will be the focus of the analysis presented next.

1896

A. Kusiak and K. Park


Activity

GA-I Earliest-start-time Duration Finish-time Latest-start-time Slack-time GA-2

al 0 10 10 20 20 a2 0 15 15 15 15 alO 0 15 15 0 0

a4 10 15 25 30 20 a3 25 10 35 25 0 all 0 15 15 30 30 Critical path

a5 10 10 20 35 25 a6 15 15 30 30 15 al4 15 10 25 15 0

a9 25 15 40 45 20 a7 35 10 45 35 0 a15 15 10 25 50 35 a8 35 10 45 40 5 a16 15 15 30 45 30 a12 45 15 60 45 0 a13 45 10 55 50 5

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

Earliest-start-time Duration Finish-time Latest-start-time Slack-time GA-3 Earliest-start-time Duration Finish-time Latest-start-time Slack-time Makespan

= 60

= alO ..... a14 ..... a3 ..... a7 ..... al2

Figure 11. Global schedule for the network in Fig. 7.

GA-I

a5

I
a9

al

a4

GA-2

.. .......... ..' .
a2

....., . ..
a6

I ,
, , , ,
a7 a8 al3 al2

I
a3

..
a14
GA-3

a10 all

a15 a16

I , ,
, ,
30

Time

40

60

Figure 12. Gantt chart of the global schedule in Fig. 11.

Concurrent engineering
Activity GA-I Earliest-start-time Duration Finish-time Latest-start-time Slack-time GA-2 al 0 10 10 15 15 a2 aO 15 15 10 10 alO 0 15 15 0 0 a4 10 15 25 25 15 a3 20 10 30 20 0 all 0 15 15 25 25 Critical path a5 10 10 20 30 20 a6 15 15 30 25 10 al4 15 5 20 15 0 a9 25 15 40 40 15 a7 30 10 40 30 0 al5 15 10 25 45 30 a7 30 10 40 35 5 al6 15 15 30 40 25 al2 40 15 55 40 0

1897

al3 40 10 50 45 5

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

Earliest-start-time Duration Finish-time Latest-start-time Slack-time GA-3 Earliest-start-time Duration Finish-time Latest-start-time Slack-time Makespan

= 55

= alO ..... a14 ..... a3 ..... a7 ..... al2

Figure 13. Global schedule after production rule 6 has been applied.

In order to reduce the makespan, either production rule 6 or 8 could be applied. -+ a3 is a strong precedence constraint, then production rule 6 is applied. Production rule 6 suggests reducing the duration of activity a14. One way of reducing the duration of al4 is to allocate additional resources to a14. Assume that the duration of al4 is reduced from 10 to 5 by allocating more resources. Then the makespan for the network is reduced from 60 to 55. Figure 13 shows the global schedule after production rule 6 has been applied. Gantt chart corresponding to the global schedule is shown in Fig. 14. On the other hand, if link al4 -+ a3 is a weak precedence constraint, then production rule 8 can be applied. Production rule 8 suggests that link al4 -+ a3 be ignored. It allows activities al4 and a3 to be executed concurrently. Thus the makes pan ofthe network is reduced from 60 to 45. Figure 15shows the global schedule after rule 8 has been applied. The corresponding Gantt chart of the global schedule is shown in Fig. 16. In summary, GA-3 turns out to be the most critical group among the three groups of activities. Activities alO and al4 have a significant impact on the makespan because delaying them holds the progress of activities in GA-2 and increases the makespan. On the other hand, the GA-I has the least tight schedule Thus, resources for GA-I may be loaned to the other two groups, if needed for speeding up some activities. One can prioritize activities within a group based on their impact on the scheduling performance measure. In GA-I, all activities are equally important. In GA-2, activities {a3, a7, a8, a12, a13} have a greater impact on the makespan than activities
If link a 14

1898

A. Kusiak and K. Park

GA-l
al

as
a4

GA-2

.' . .
a2

..

.....

....
a6

a9

- ..,

, , ,
0 0

a7 a3 a8

a13 al2

I
I

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

R
GA-3

.''0'

alO all

315 a16

I
,
0 0

, , ,
4

Time

30

55

Figure 14. Gantt chart of the global schedule in Fig. 6.

Activity GA-I Earliest -start-time Duration Finish-time Latest -start-time Slack-time GA-2 Earliest-start-time Duration Finish-time Latest-start-time Slack-time GA-3 Earliest-start-time Duration Finish-time Latest-start-time Slack-time al 0 10 10 5 5 a2 0 15 15 0 0 alO 0 15 15 20 20 Makespan a4 10 15 25 15 5 a3 0 10 10 15 15 all 0 15 15 15 15 a5
10 10 20 20 10

a9 25 15 40 30 5 a6 15 10 25 20 5 al5 15 10 25 35 20 a8
10 10 20 25 15

a6 15 15 30 15 0 al4 15 10 25 35 20 Critical path

al2 30 15 45 30 0

al3 25 10 35 35 10

al6 15 15 30 30 15

= 45

= a2 --+ a6 --+ al2

Figure 15. Global schedule after production rule 8 has been applied.

Concurrent engineering

1899

GA-l

I
al

a6

I
a2

as

I I a4
a7

a9

I ,
, , ,
0

GA-2
a3

a6

I
a13

a12 :

a8

I
a14 .15 a16

I:,,
0 0

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

GA-3

alO all

I ,
,

, , , , , , ,
,
0 0

, ,

, , ,
45

Time

30

40

Figure 16. Gantt chart of the global schedule in Fig. 15.

{a2,a6}.ln GA-3, activities {alO, a14} have a greater impact on the makespan than activities{all,aI5,aI6}.
5. Conclusions In concurrent engineering, an attempt is made to consider the design constraints simultaneously rather than in the series as in the case of traditional design. This results in reduction of the duration of the design project, cost savings, and better quality of the final design. However, the concurrent product design gives rise to a large scale project which might be too difficult to manage as a whole. One way to reduce the complexity of a large scale project is to decompose it into subsystems. Advantages of project decomposition are as follows:
1. Project decomposition allows us to determine a potential group of activities that might be scheduled simultaneously. 2. The quality of data collected is such that the precedence constraints among activities and duration of activities are improved. 3. The project scheduling and management is simplified because the management of a group of activities focuses on the problems within the group. 4. Project decomposition creates an environment for improvement of effectiveness and efficiency of the project.

Although the management task of the entire project is simplified due to the project decomposition, an integrated system which co-ordinates and analyses activities in groups is required. In this paper, a knowledge-based system was used to manage the design project. The system allows to perform analysis regarding the type of decomposition and the precedence constraints between groups. Some production rules illustrating the capability of the knowledge based system were presented. The concepts introduced were demonstrated with a numerical example.

1900
, References

Concurrent engineering

Downloaded by [Maulana Azad National Institute of Tech] at 19:39 25 April 2013

BADIRU, A. 8., 1988, Project Management in Manufacturing and High Technology Operations (New York: Wiley). KUSIAK, A., 1990, Intelligent Manufacturing Systems (Englewood Cliffs, N.J.: Prentice Hall). KUSIAK, A., and CHOW, W. S., 1987, An algorithm for cluster identification. IEEE Transactions on Systems, Man, and Cybernetics, SMC-17(4), 696-699. KUSIAK, A., and PARK, K., 1989, Design of assemblies for schedulability. Working paper No. 89-14, Department of Industrial Engineering, The University of Iowa. NEVINS,1. L., and WHITNEY, D. E., 1989, Concurrent Design of Products and Processes (New York: McGraw-Hill). O'GRADY, P., RAMERS, D., and BOLSEN, 1., 1988, Artificial intelligence constraint nets applied to design for economic manufacture. Computer Integrated Manufacturing, 1(4),204-209. VASILASH, G. S., 1987, Simultaneous engineering-management's new competitiveness tool. Production, July, 41. WHITNEY, D. E., 1988, Manufacturing by design. Harvard Business Review, July / August, 87-88.

Вам также может понравиться