Koren Y. The Global Manufacturing Revolution: Product-Process-Business Integration and Reconfigurable Systems

Подождите немного. Документ загружается.

The system includes only four machines, and each machine is equipped with a

tool magazine that can hold up to four tools (therefore, ﬁve operations cannot be

done on a single machine).

(a) Given the two conﬁgurations in Figure 9.20, label the operations that each

machine should perform to maximize throughput. Calculate the throughput

(parts per hour) for each conﬁguration. Which conﬁguration will have a

higher throughput (assuming 100% machine reliability)?

(b) Is there another conﬁguration of the four machin es that has equal or higher

throughput? Draw it and label the operations that each machine is to

perform.

that you propos ed in item b) to scale up the system capacity. Draw the three

new conﬁgurations that you propose, and calculate the new throughput

(parts per hour) for each conﬁguration.

9.3 A manufacturer has two options for buying a new production line to produce up

to 100,000 products per year: A dedi cated line that costs $50 million or an

RMS that costs $60 million. The product is sold at a proﬁt of $150 per product.

Solve for the three scenarios described below, and draw a graph of proﬁt (or

loss) for each case and for each system during 6 years of operation. When does

the higher investment in the reconﬁgurable system become economi cally

justiﬁed?

(a) The production is 100,000 products per year, for 6 years, as was expected.

(b) After 2 years of operation the market demand grows to 150,000 units per

year. The dedicated line is not scalable and cannot meet the additional

product demand. For an addi tional investment of $15 million, the reconﬁ-

gurable system can supply the new demand for the next 4 years.

is needed (again, requiring 100,000 units per year for the new model; the old

one is discontinued). A new dedicated line could be built to produce the new

Figure 9.20 Two conﬁgurations.

PROBLEMS 251

product, or the reconﬁgurable system can be reconﬁgured for an additional

cost of $10 million, and produce the new model for 4 years.

9.4 Repeat Problem 9.3 when the interest on capital investment is 10% per year.

Do you recommend investing in the project?

REFERENCES

1. Y. Koren and A. G. Ulsoy. Reconﬁgurable Manufacturing Systems. Engineering Research

Center for Reconﬁgurable Machining Systems (ERC/RMS) Report #1, The University of

Michigan, Ann Arbor, 1997.

2. O. Garro and P. Martin. Towards new architecture of machine tools. International Journal of

Production Research, 1993, Vol. 31, No. 10, pp. 2403–2414.

3. Y. Koren and A. G. Ulsoy. Reconﬁgurable manufacturing system having a production

capacity, method for designing same, and method for changing its production capacity.

U.S. patent No. 6,349,237. Issued on February 19, 2002.

4. Y. Koren and S. Kota. Reconﬁgurable Machine Tools. U.S. Patent No. 5,943,750. Issue on

August 31, 1999.

5. Y. Koren, F. Jovane, and G. Pritschow (eds). Open Architecture Control Systems, Summary

of Global Activity, ITIA Series, Vol. 2, 1998.

6. Similar ideas for computing systems are presented in J. Villasenor and W.H. Mangione-

Smith. Conﬁgurable computing. Scientiﬁc American, pp. 66–71, June 1997.

7. G. Erixon. Modularity—the basis for Product and Factory Re-engineering, Annals of the

CIRP, 1996, Vol. 45/1, pp. 1–4.

8. H. K. T

€

onshoff, E. Menzel, H. Hinkenhuis, and E. Nitidem. Intelligence in machine tools

by conﬁguration. 7th International Conference on Production/Precision Engineering,

Chiba, Japan, 1994.

9. S. J. Hu. Stream of variations theory for automotive body assembly, Annals of the CIRP,

1997, Vol. 46/1, pp. 1–4.

10. J. Shi. Stream of Variation Modeling and Analysis for Multistage Manufacturing

Processes. CRC Press, 2006.

11. Y. Koren and R. Hill. Integrated reconﬁgurable manufacturing system. US patent

#6,920,973. Issue on July 26, 2005.

252

RECONFIGURABLE MANUFACTURING SYSTEMS

Chapter 10

System Conﬁguration Analysis

In large manufacturing systems the production is done in many stages. A product is

partially processed in one stage and then transferred to the next until all operations are

completed. The conﬁguration of a system can facilitate or impede the system’s

productivity, responsiveness, convertibility and scalability, and also impact its daily

operations. Multi-stage manufacturing systems can allow for several operational

conﬁgurations depending on the way the machines are arranged in the stages, and

connected via the material handling system. In this chapter, we offer a method

for classiﬁcation of conﬁgurations and use it to compare the attributes of various

conﬁguration classes. We will also discuss the conﬁguration of Reconﬁgurable

Manufacturing Systems (RMSs), and present a way to calculate the number of

possible RMS conﬁgurations based on the number of machines it contains.

10.1 CLASSIFICATION OF CONFIGURATIONS

The ﬁrst step toward a systematic approach to classifying conﬁgurations is to

determine the number of possibl e conﬁgurations when the daily demand, Q

(parts/day), and the total processing time for the part, t (m inute/part), are given. In

reality the processing times vary widely depending on the equipment involved; but, to

begin with, we assume that it is given.

The Global Manufacturing Revolution: Product-Process-Business Integration and Reconﬁgurable Systems

By Yoram Koren

253

The minimum number of machines, N, needed in the system is calculated by the

equation

N ¼

Q  t

Minutes=day available  machine reliability

ð10:1Þ

In the following calculations we assume 100% reliability of all pieces of equipment

(i.e., machine reliability ¼ 1). The resulting number of machines calculated by

Eq. (10.1) must be rounde d to the next larger integer. For example, if there is a need

for 500 parts per day and the processing time for each part is 9.5 minutes, for a working

time of 1000 minute/day there is a need for at least ﬁve machines in the system.

In the general case the total number of conﬁgurations for N machines is huge.

When plotted on a logarithmic scale, the number of conﬁgurations increases almost

linearly with the number of machines, as shown in Figure 10.1. We will see later in this

chapter that the number of p ossible RMS conﬁgurations is much smaller, as indicated

in the table in Figure 10.1.

Eq. (10.1) yields the minimum number of machines needed to meet the required

demand. The next questions are: What is the best way to arrange these machines and

how to connect them? Should we arrange them, for example, in a serial line, a pure

parallel system, or some combination? Which one of all possible conﬁgurations is the

most advantageous?

For example, in the case of ﬁve machines the total number of conﬁgurations that are

possible is 48. Figure 10.2 shows 32 of these conﬁgurations. As we have seen in

Figure 10.1 the number of possible conﬁgurations increased exponentially with the

number of machines. It is not unheard of in the automotive powertrain industry to have

80 or more machines coordinated into one system. How can one possibly analyze the

merits of so many possible conﬁgurations?

To begin with, we classify conﬁgurations as either symmetrical or asymmetrical,

based on whether one can draw a symmetric axis along the conﬁguration. A

conﬁguration is then evaluated by its machine arrangement and connections.For

example, conﬁgurations a and b have identical machine arrangements (one in Stage 1,

two in Stage 2, and two in Stage 3), but they are different because of different

connections among the machines—conﬁguration b has cross-coupling between

Stages 2 and 3. The type of the material handling system determines the connections

Figure 10.1 Total number of system conﬁgurations for different numbers of machines.

W. X. Zhu, an ERC-RMS Ph.D. student, did the calculation of the number of conﬁgurations.

254 SYSTEM CONFIGURATION ANALYSIS

of a conﬁguration. Altogether, a system with ﬁve machines may have 16 different

symmetric arrangements (13 of which are plotted in Figure 10.2). Fortunately, the

designer has to consider only symmetric conﬁgurations because:

Only symmetric conﬁgurations are suitable for manufacturing systems

Asymmetric conﬁgurations add immense complexity and are not viable in real

manufacturing lines as will be explained below. The number of possible asymmetric

conﬁgurations is much larger than that of symmetric conﬁgurations—a total of 30 in

the case of ﬁve machines (18 of them are plotted in Figure 10.2). It is important to note

that we deﬁne conﬁgurations d

and e

as asymmetric (although they have a symmetric

axis) because they may be positioned differently (as d and e). Similarly, the two

conﬁgurations f and g in Figure 10.2 are deﬁned when studying the Reconﬁguration

Science as asymmetric conﬁgurations, although they may be drawn as symmetric.

We would like to explain why asymmetric conﬁgurations are usually not suitable

for real manufacturing systems. Asymmetric conﬁgurations may be sub-classiﬁed as

(a) variable-process conﬁgurations and (b) single-process conﬁgurations with non-

identical machines in at least one of the stages. Corresponding examples are shown in

Figures 10. 3a and 10.3b, respectively.

Figure 10.2 Conﬁgurations with ﬁve machines.

Figure 10.3 Two classes of asymmetric conﬁgurations.

CLASSIFICATION OF CONFIGURATIONS 255

Variable-process conﬁgurations are characterized by possible non-identical ﬂow

paths for the part, and therefore they need a prepa ration period of several process plans

and corresponding setups. For example, in the system depi cted in Figure 10.3a

possible ﬂow-paths are a-b-c-d-e, g-c-f, g-c-d-e, etc. The executed process plan

depends on the ﬂow-path of the part being processed in the system. This is absolutely

impractical because: (1) designers will not go to the effort to design multiple process

plans for the same part, and (2) different process plans and ﬂow paths increase part-

quality problems and make quality error detection more complicated.

In the second class of asymmetric conﬁgurations, although the process planning

is identical in each ﬂow path, the machines in at least one stage are different. For

example, in Figure 10.3b, the machine b in Stage 2 must be two times faster than

machines a; machine d in Stage 4 must be two time s faster than machines c. (By

contrast, in symmetric conﬁgurations, the processing times of every machine in

a stage are equal.) It is absolutely impractical to mix different types of machines in the

same manufacturing stage performing exactly the same sequence of tasks. The system

designer should not consider this class of conﬁguration, either, due to that excessive

complexity. The conclusion is

In a real manufacturing context only symmetric conﬁgurations should be consid-

ered; they are always single-process conﬁgurations with identical machines in

each stage.

Symmetric conﬁgurations may further be divided, into three basic classes as shown

in Figure 10.4.

A designer of manufacturing systems should consider only these three classes:

I Cell Conﬁgurations—Conﬁgurations consisting of several serial manu-

facturing lines (cells) arranged in parallel with no crossovers, as shown as

Class I in Figure 10.4, and also shown in Figure 10.5.

II RMS Conﬁgurations—Conﬁgurations with crossover connections after

every stage, as shown in Figure 10.6. The part from any machine in stage

i can be transferred to any machine in stage (i þ 1). All machines and

operations in any stage are identical.

III Conﬁgurations in which there are some stages with no crossovers. This class

includes combinations of the previous two classes.

Figure 10.4 Three classes of symmetric conﬁgurations.

256

SYSTEM CONFIGURATION ANALYSIS

To understand the RMS conﬁguration operations issue, study the sketch in

Figure 10.7 of a practical three-stage RMS with gantries that transport the parts.

A spine gantry transfers a part to a small cell conveyor; the part moves on the conveyor

to a position where a cell gantry can pick it up and take it for processing in one of the

machines in its stage. When the part processing is done, the cell gantry returns the part

to the conveyor, which moves the part to a position in which the next spine gantry can

pick it up for processing in the next stage, and so on.

10.2 COMPARING RMS WITH CELL CONFIGURATIONS

We compare below the two main practical conﬁgurations depicted in Figures 10.5 and

10.6 by four criteria: investment cost, line-balancing ability, scalability options, and

productivity when machines fail.

Capital Investment: The conﬁgurations in Figures 10.5 and 10.6 (or 10.7) have

identical machine arrangements—three stages with two machines in each

stage—but their connections are quite different. That means that they use

different part handling devices, each requiring a different capital investment.

The entire part handling system in Figure 10.5 is simpler and has a smaller

number of handling devices compared with the RMS in Fig. 10.7. Thus, the

capital investment in the RMS conﬁguration is higher.

Line Balancing: A major drawback of the cell conﬁguration compared with the

RMS conﬁguration is that it provides less ﬂexibility in balancing the system. For

example, if just one product is produced, the processing time in all stages of the

cell conﬁguration must be exactly equal to be perfectly balanced. By contrast,

Figure 10.5 Symmetric conﬁguration of Class I—parallel lines, or cells (If the two

marked machines fail, the system production stops).

Figure 10.6 Symmetric conﬁguration of Class II—RMS conﬁguration (If the two marked

machines fail, the system still has 50% production capacity).

COMPARING RMS WITH CELL CONFIGURATIONS 257

for a balanced RMS conﬁguration only the following relationship needs to be

satisﬁed:

¼ t

ð10:2Þ

where N

is the number of machines in stage i, and t

is the processing time per

machine in stage i. Therefore, in RMS conﬁgurations the number of machin es

per stage is not necessarily equal in all stages. The number of machines in the

various stages of RMS conﬁgurations may be adjusted to provide an accurate

line balancing, which consequently yields an improved productivity.

System Scalability: The scalability of the RMS conﬁguration is better by far than

that of the parallel line conﬁguration. Adding one machine in one of the stages

and rebalancing the system enables adds a small increment of capacity. In the

parallel line conﬁguration, a whole additional line must be added to increase the

overall system capacity. In markets of unstable demand, scalability is an

important advantage of RMS conﬁgurations.

Productivity: If machine reliability is low, the RMS conﬁguration offers highe r

productivity than a parallel line conﬁguration. As shown in Figure 10.5, if

machines in two different lines and at two different stages (marked with x) are

down, the entire system is down (i.e., throughput ¼ 0). Whereas, for an RMS, the

same conditions—two machines not working (marked with x in Figure 10.6)—

theRMS throughput is still at 50%. So, the RMS isa moreproductivesystem from

a machine downtime aspect. However, if one of the cell gantries in the RMS is

down, the whole system is down. By contrast, systems with parallel lines do not

contain cell gantries, and are more reliable from a material handling system

aspect. The consequent question is therefore:

When considering reliability, which conﬁguration has higher productivity?

We solved this problem analytically, but showing the full analysis is beyond the

scope of this book. In this analysis the number of machines per stage in the RMS

conﬁguration is equal in all stages. The RMS conﬁguration has a spine gantry with an

identical reliability to the conveyors in Figure 10.5. The analysis calculates tradeoffs

between cell-gantry reliability and machine reliability.

As an example we shall show just one typical result in Figure 10.8, which is plotted

for a gantry reliability (or availability) of G

¼ 0.96, and machine reliability of

Figure 10.7 A practical Reconﬁgurable Manufacturing System.

258

SYSTEM CONFIGURATION ANALYSIS

¼ 0.90. Our analysis revealed that there is a borderline based on the machine

reliability and gantry reliability (which must always be better than the mac hine

reliability). On the right-top side of the borderline the preferred system is RMS with

crossovers. For example, if the system has nine stages with four machines per stage,

the RMS conﬁguration will yield higher productivity than parallel-line one. The

parallel-line conﬁguration is preferred when, for example, the system has nine stages

with just two machines per stage (namely, two parallel lines with nine stages). The

better the machine average reliability is (e.g., 0.95), the larger the solution space for

the parallel-line conﬁguration (without crossovers), and vice versa.

The main conclusions of this analysis are

In large systems, with a large number of stages and machines per stage, the RMS

conﬁguration has higher productivity than the cell conﬁguration.

If the machine reliability is very high, the cell conﬁguration yields higher

productivity than the RMS conﬁguration.

The advantages of each conﬁguration are summarized below.

Capital

Investment Scalability

Line

Balancing Productivity

Parallel lines Lower Higher for high

machine reliability

RMS

conﬁguration

Much

better

Much

better

Higher in complex,

large systems

Figure 10.8 Productivity comparison between parallel lines and RMS conﬁgurations.

COMPARING RMS WITH CELL CONFIGURATIONS 259

10.3 CALCULATING THE NUMBER OF RMS CONFIGURATIONS

We have seen that it is easy to calculate the minimum number of machines N required

in the system by solving Eq. (10.1). However, as shown in Figure 10.1, the number

of all possible conﬁgurations with N machines is enormous. We have conduc ted

a thorough mathematical study of system conﬁgurations, and our conclusion is

Equations for calculating the number of conﬁgurations with N machines exist

only for RM S-type conﬁgurations.

The basic equations for calculating the number of possible RMS conﬁgurations are

given below. The number, K, of possible RMS conﬁgurations with N machines

arranged in up to m stages is

K ¼

m¼1

N1

m1



¼ 2

N1

ð10:3Þ

The number, K, o f possible conﬁgurations with N machines arranged in exactly m

stages is

K ¼

ðN1Þ!

ðNmÞ!ðm1Þ!



ð10:4Þ

For example, for N ¼ 7 machines arranged in up to seven stages Eq. (10.3) yields

K ¼ 64 conﬁgurations, and if arranged in exactly three stages, Eq. (10.4) yields K ¼ 15

RMS conﬁgurations. The mathematical results of these two equations for any N and m

may be arranged in a triangular format, known as a Pascal triangle, shown in

Figure 10.9.

The method of calculating the numerical value of each cell in the Pascal triangle

is as follows. The numerical value corresponding to N machines that are arranged in

m stages is calculated by

The value for N machines in m stages ¼ (The value for N1 machines in m1

stages) þ (the value for N 1 machines in m stage s)

For example, in Figure 10.9, the cell of N ¼ 5 and m ¼ 3 shows 6, which is the sum

of 3 þ 3 of the previous line of N  1 ¼ 4 machines with two and three stages.

The triangle also allows the designer to immediately visualize the number of

possible RMS conﬁgurations for N machines that are arranged in m stages. For

example, there are 15 RMS conﬁgurations when seven machines are allowed to be

arranged in exactly three stages. In addition, the Pascal triangle allows the designer to

immediately calculate the number of possible RMS conﬁgurations for N machines

that are arranged between i and j stages (i, j < N). We will use this information in the

following example.

260 SYSTEM CONFIGURATION ANALYSIS