You are on page 1of 21

This article was downloaded by: [Columbia University]

On: 09 December 2014, At: 02:17


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

Process flexibility with bill of material


constraints
a b a
Z. Hua , F. Huang & B. Zhang
a
School of Management, University of Science and Technology of
China , Hefei, Anhui 230026, PR China
b
School of Economics and Management, Tsinghua University ,
Beijing, 100084, PR China
Published online: 21 Nov 2007.

To cite this article: Z. Hua , F. Huang & B. Zhang (2008) Process flexibility with bill of
material constraints, International Journal of Production Research, 46:6, 1567-1586, DOI:
10.1080/00207540600969766

To link to this article: http://dx.doi.org/10.1080/00207540600969766

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

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. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
International Journal of Production Research,
Vol. 46, No. 6, 15 March 2008, 1567–1586

Process flexibility with bill of material constraints

Z. HUA*y, F. HUANGz and B. ZHANGy

ySchool of Management, University of Science and Technology of China,


Hefei, Anhui 230026, PR China
zSchool of Economics and Management, Tsinghua University, Beijing, 100084, PR China

(Revision received August 2006)


Downloaded by [Columbia University] at 02:17 09 December 2014

During the realization of process flexibility, many potential benefits will be


brought if bill of material (BOM) is considered. When we consider the BOM
constraints, most current research results about process flexibility need to be
re-appraised. This paper provides measures of process flexibility in different cases
where the BOM constraints should be satisfied. Structural properties of achieving
high process flexibility in different cases are also investigated, by which rules
for the capacity expansion problems are developed. Computation experiments
illustrate the rationality of the measures of process flexibility and effectiveness of
the developed rules for the capacity expansion problems.

Keywords: Flexibility; Capacity planning; Structural property; Bill of material

1. Introduction

The recent decades have witnessed an increasingly competitive market place


characterized by short product life cycles, demand uncertainty, product prolifera-
tion, increased customization, and quick response. The shift from mass production
of highly standardized products to small-batch production of customized products
has forced most manufacturing organizations to use some form of flexible capacity.
Flexible manufacturing systems (FMS), which can provide a variety of flexibilities
that include volume, routing, process and product mix flexibilities, are typically
adopted to meet the challenge (Chen et al. 2002, Van Hop 2004).
There are numerous research results about benefits of flexibility, source of
flexibility and methods of realizing flexibility from the standpoint of technology or
operations management. Different terms and definitions of flexibility have thus been
coined, which can be classified into three levels:
1. Basic flexibilities which include machine flexibility, material handling
flexibility, and operation flexibility.
2. System flexibilities which include volume flexibility, expansion flexibility,
routing flexibility, process flexibility and product flexibility.

*Corresponding author. Email: zshua@ustc.edu.cn

International Journal of Production Research


ISSN 0020–7543 print/ISSN 1366–588X online  2008 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/00207540600969766
1568 Z. Hua et al.

3. Aggregate flexibilities which include programme flexibility, production


flexibility and market flexibility (Sethi and Sethi 1990, Stecke and Raman
1995).
Among these kinds of flexibilities, process flexibility has drawn most attention
from researchers and practitioners. This is because when firms expand their
manufacturing scale and/or scope, how to realize process flexibility and how much
process flexibility is needed by investing multi-purpose production sources are key
issues of their manufacturing strategies.
Process flexibility results from being able to build different types of products
in the same plant or production facility at the same time (Jordan and Graves 1995).
To realize process flexibility, Jordan and Graves (1995) developed three principles
for guiding flexibility investment:
Downloaded by [Columbia University] at 02:17 09 December 2014

1. Try to equalize the capacity to which each product is directly connected.


2. Try to equalize the total expected demand to which each plant is directly
connected.
3. Try to create a chain(s) that encompass(es) as many plants and products as
possible (we term this principle as the ‘chaining policy’ in this paper).
Gavish (1994) extended the work to a two-stage supply chain and showed the
effectiveness of the chaining policy. These researches were further extended to multi-
stage manufacturing systems by Graves and Tomlin (2003), and their results showed
that the chaining policy is an effective policy of achieving high process flexibility for
each isolated stage and for the overall supply chain (the multi-stage manufacturing
systems).
In current research, process flexibility takes a product as a whole without
considering what part (component) types it consists of or which machines it is
produced by (Rajagopalan 1993, Jordan and Graves 1995, Van Mieghem 1998).
Process flexibility produces multiple products to meet customized demands, and a
customised demand is essentially reflected by a personalized bill of material (BOM)
(Yeung et al. 1999, Tseng et al. 2005), therefore we argue that it is important to
consider the impact of the BOM constraints when we try to realize process flexibility.
To elaborate this argument, we first take the PWB (printed wiring board)
assembly lines from Hua and Banerjee (2000a) as an example to show the method
of describing the BOM constraints. A PWB assembly line inserts or mounts electrical
components, e.g. resistors, transistors and capacitors, into pre-specified positions on
a PWB. A machine line makes insertions on PWBs with the only restriction that all
the insertions for a product type (PWB) be performed on the same machine line. The
placement operations for a given product type are distributed across the available
machines in the line. Figure 1 shows an illustration of such a manufacturing system,
with three assembly processes.
The manufacturing system shown in figure 1 consists of two machine lines,
and each machine line consists of three placement machines (boxes in the figures)
with the capability of placing some part types (circles in the figures) on a PWB. In
figure 1, different machine types are numbered 1 to 6. The line connecting a machine
and a part type indicates that the corresponding machine is capable of placing the
corresponding part type. A machine that can place more part types is more flexible,
and is usually more expensive. There are three part types, i.e., SC (small chips),
Process flexibility with bill of material constraints 1569

(A) Machine type Demand type (B) Machine type Demand type

1 SC 4 SC

2 OS 5 OS

SC, small chips


OS, odd-shaped
3 LC 6 LC
LC, large chips

Figure 1. Illustration of two PWB assembly lines.


Downloaded by [Columbia University] at 02:17 09 December 2014

Table 1. Examples of two-level BOMs.

SC OS LC

Product 1 1 6 1
Product 2 8 0 5
Product 3 2 4 8

LC (large chips) and OS (odd-shaped chips), of which different product types are
composed. Structure of a product type is described by its BOM, which defines the
number of parts of different types in a unit product. As examples, two-level BOMs of
three product types are listed in table 1. As shown in table 1, each unit of product
type 1 requires one SC, six units of OS and one LC.
By adopting the above method, BOM constraints can be described in capacity
expansion and technology selection problems. In the realization of process flexibility,
many potential benefits will be brought if BOM constraint is considered.
. Achieving higher process flexibility. Many products have their own BOMs.
Without considering these BOM constraints, current process flexibility
improvement strategy, e.g. the chaining policy, will possibly result in lower
process flexibility (we will further elaborate this phenomenon through the
example depicted in figure 2).
. Improving a firm’s capability of rapid response to customers’ needs. With the
increasingly intensive business competition, length of product life-cycle
becomes shorter and shorter, which implies a frequent BOM renovation.
If BOM is explicitly involved in the realization of process flexibility, it will
award an enterprise the ability to accommodate the rapid change of BOMs,
thus improving the firm’s capability of rapid response to customers’ needs.
. Decreasing inventory and its related costs. To meet BOM constraints, it is
required that the production of redundant components should be avoided.
Excess components that are not matches for the BOM will lead to a waste of
capacity and an increase of inventory.
. Providing a platform for supply chain coordination. BOM constraints of a
downstream manufacturer play a key role in the selection of the upstream
1570 Z. Hua et al.

(A) Machine type Demand type (B) Machine type Demand type

1 1 1 1

2 2 2 2

3 3 4 3

Figure 2. An illustrative example of the ‘chaining policy’.


Downloaded by [Columbia University] at 02:17 09 December 2014

suppliers (Yian et al. 2003), and will provide guidelines for the selected
suppliers in determining their inventory control policies (Huang et al. 2005).
Therefore, BOM constraints can work as a platform for coordination
among supply chain partners in determining their cooperative production
plans and inventory control policies.
Despite the above benefits of considering the BOM constraints in the realization
of process flexibility, problems about process flexibility (e.g. capacity planning,
technology selection) are more complex and most current research results about
process flexibility need to be re-appraised. A typical example is that when we
consider the BOM constraints, it may lead to low process flexibility if the chaining
policy is directly applied to machines and part types.
To illustrate this, we compare two machine lines that are depicted in figure 2(A)
and (B) (respectively termed as line A and line B in this example). As shown in
figure 2, both machine lines consist of three machines with the capability of
processing some part types. We assume that all machines have the same capacity of
processing C units of a part over the given planning period. As defined by Iravani
et al. (2005), process flexibility here can be measured by calculating the ‘shifted
capacity between demand (part) types’. Without considering BOM, according to the
‘chaining policy’, line B has higher process flexibility than line A.
However, if we consider the BOM constraint, e.g. the BOM constraint of product 1
listed in table 1, process flexibility of line A is higher than that of line B. As shown in
table 1, the BOM constraint of product 1 is (1, 6, 1). So the maximum amount of the
product that line A can produce is 3C/8 without any capacity idled, and the maximum
amount of the product that line B can produce is C/3 (in unit of product) with C/3
capacity (in unit of part) idled. In producing product 1, if we want to fully utilize the
capacity of machines in line B, then the machine line will produce C/3 units of
redundant parts of types 1 and/or 3 without any improvement on the throughput of
the machine line. Calculation details about figure 2 are given in appendix A.
When the BOM constraints are considered, this paper tries to find structural
properties of process flexibility: what kind of machine line structure will achieve high
process flexibility. As an application of the structural property, we develop some
rules for the capacity expansion problems, which can provide good or near optimal
solutions to the complex integer programming problems (Hua and Banerjee 2000a)
without much computation effort.
Process flexibility with bill of material constraints 1571

The organization of the remainder of this paper is as follows. In section 2, we give


a definition of process flexibility when the BOM constraints are considered;
structural properties of achieving high process flexibility are then investigated from
the simplest to the general cases. Rules for the capacity expansion problems are
developed in section 3 and numerical results illustrated in section 4. Conclusions are
presented in section 5.

2. Process flexibility with BOM constraints

We assume that a manufacturing system has L machine lines. Each machine line
consists of E machine types with the same capacity of processing C units of a part
over the given planning period. The manufacturing system is used to produce N
Downloaded by [Columbia University] at 02:17 09 December 2014

product types, which all consist of P part (component) types. The structure of a
product is described by its BOM constraint srp (r ¼ 1, 2, . . . , N, p ¼ 1, 2, . . . , P),
which represents the number of parts of type p (p ¼ 1, 2, . . . , P) in a unit of product r
(r ¼ 1, 2, . . . , N). In the manufacturing system, it is required that all part (component)
types of a product be produced on the same machine line.
Without BOM constraints Iravani et al. (2005) defined process flexibility of a
manufacturing system as its ability to reallocate production to respond to changes in
demand or in source capacity. They defined a structural flexibility (SF) matrix,
whose diagonal elements are the ‘total capacity capable of processing a demand
type’, and non-diagonal elements are the ‘shifted capacity between demand types’.
Under the given structure of a manufacturing system, the elements of the SF matrix
can be determined by solving the max flow problems. Iravani et al. (2005) then
showed that process flexibility of a manufacturing system can be measured by the
mean index and the eigenvalue index, where the mean index is the mean of all
the elements in the SF matrix and the eigenvalue index is the dominant eigenvalue of
the SF matrix.
When BOM constraints are considered, we adopt the definition of process
flexibility given by Iravani et al. (2005), and define a process flexibility (PF) matrix
which is similar to the SF matrix. That is, diagonal elements of a PF matrix are the
‘total capacity capable of processing a demand type’, and non-diagonal elements are
the ‘shifted capacity between demand types’. The main difference between the SF
matrix and the PF matrix is that, demand type in the SF matrix is independent
product demand (determined by the market), while demand type in the PF matrix
can be independent product demand or dependent part demand (dependent on
product demand).
We next investigate measures of process flexibility from the simplest to the
general cases, and try to find properties of the process flexibility measures through
analysing the corresponding PF matrices.

2.1 The simplest case


The simplest case is one where each product includes only one unit of a part type,
and multiple products with independent demands are produced on a single
machine line, i.e., L ¼ 1, N41 and P ¼ 1. Here BOM constraint is actually not
considered, part demand equals product demand, and the PF matrix is equivalent
1572 Z. Hua et al.

Table 2. Definitions of the four cases.

Cases No. of machine lines (L) No. of product types (N) No. of part types (P)

Case 1 L41 N41 P¼1


Case 2 L¼1 N¼1 P41
Case 3 L¼1 N41 P41
Case 4 L41 N41 P41

to the SF matrix. This case has been well investigated by many researchers (Jordan
and Graves 1995, Hopp et al. 2004, Iravani et al. 2005). The mean and eigenvalue
indices of the PF matrix are appropriate measures for process flexibility. In this
Downloaded by [Columbia University] at 02:17 09 December 2014

case, the ‘chaining policy’ can be directly applied to achieve high process flexibility.
We start from this simplest case to study four more general cases. The four cases
are defined in table 2.
Among the four cases defined in table 1, Case 1 is directly extended from the
simplest case, in which BOM is not considered. Cases 2 and 3 have the BOM
constraints and the result of Case 2 will support the research of Case 3, and then both
of their results will further support the research of Case 4.
Case 1: L41, N41 and P ¼ 1
In this case, the PF matrix for a machine line is the same as the SF matrix. When
there are more than one machine lines in a manufacturing system, process flexibilities
of different machine lines have interactions. The following propositions summarize
this phenomenon.
Proposition 1: If L41, N41 and P ¼ 1, then the mean index measure of the process
flexibility of all machine lines as a whole is no less than the sum of the measures of all
machine lines.
Proof of this proposition appears in appendix B.
It can be found in appendix B that Proposition 1 does not hold true if we
substitute the eigenvalue index measure for the mean index measure.
Proposition 1 suggests that we should aggregate multiple machine lines into one
machine line in evaluating the process flexibility of the manufacturing system. If a
new machine will be added to a machine line in the manufacturing system, we then
have the following proposition.
Proposition 2: If L41, N41 and P ¼ 1, then there is no difference in process
flexibility of the whole manufacturing system no matter to which machine line a
machine is added.
This proposition is rather straightforward.
From the standpoint of improving process flexibility, if we take multiple machine
lines as a whole, then Case 1 is degenerated to the simplest case.
Case 2: L ¼ 1, N ¼ 1 and P41
Denote the corresponding PF matrix as MPP ¼ (mpq)PP, where mpq (p, q ¼ 1,
2, . . . , P, p 6¼ q) is the total capacity (in unit of part) that can be shifted from part type
p to part type q; mpp (p ¼ 1, 2, . . . , P) is the total capacity which can be used to
produce part type p.
Process flexibility with bill of material constraints 1573

Given the BOM constraints of the product, it is clear that the method of
computing ‘shifted capacity between demand types’ described by Iravani et al. (2005)
is not applicable, thus we need a new approach to determine the PF matrix.
Recall that each machine in the machine line can produce C units of a part type,
so we define f (C) (in unit of the product) as the maximum amount of the product the
machine line can produce, which can be computed by solving a simple linear
programming problem (Hua and Banerjee 2000b). Then mpp (p ¼ 1, 2, . . . , P) is
determined as follows

mpp ¼ s1p f ðCÞ, ð1Þ


where s1p (p ¼ 1, 2, . . . , P) is the BOM constraint of the product.
Before giving the method of computing mpq (p, q ¼ 1, 2, . . . , P, p 6¼ q), we first
Downloaded by [Columbia University] at 02:17 09 December 2014

obtain some intuitions by analysing the machine line depicted in figure 2(A). First
of all, it should be emphasised that, although machine 1 in the machine line can
produce part types 1 and 2, with the BOM constraints, it is meaningless in any
situation to shift all its capacity C from producing part type 1 to part type 2.
Actually, the maximum amount of capacity shifted from part type 1 (2) to part
type 2 (1) is
 
s12 s11
C C :
s11 þ s12 s11 þ s12
This observation gives us a hint on how to determine mpq.
For any p, q ¼ 1, 2, . . . , P, p 6¼ q, let m0pq be the total capacity (represented by unit
of part) that can be shifted from part type p to part type q without considering the
BOM constraints. m0pq (p, q ¼ 1, 2, . . . , P, p 6¼ q) can be easily determined by solving
the max flow problems (Iravani et al. 2005).
We then define the PF matrix for Case 2 as follows.
Definition 1: If L ¼ 1, N ¼ 1 and P41, then the PF matrix M can be defined as:

mpp ¼ s1p f ðCÞ, p ¼ 1, 2, . . . , P, ð2Þ


s1q
mpq ¼ m0 , p, q ¼ 1, 2, . . . , P, p 6¼ q: ð3Þ
s1p þ s1q pq
For example, if the BOM constraints of a product are s11 ¼ 1, s12 ¼ 2 and s13 ¼ 3,
then according to Definition 1, the PF matrix of the machine line depicted in
figure 2(A) is as follows:
0 1
1 2 3
B2 3 4C
B C
B1 6C
M¼B 1 C ð4Þ
B3 5C
@ A
1 4 3
4 5 3
After obtaining the PF matrix, we can use the mean and eigenvalue indexes of the
PF matrix to measure process flexibility. A structural property of the manufacturing
system is described in the following proposition.
1574 Z. Hua et al.

Proposition 3: If L ¼ 1, N ¼ 1 and P41, for given s1p ( p ¼ 1, 2, . . . , P), the structure


of a machine line that satisfies
s X X
P 1p C C, p ¼ 1, 2, . . . , P ð5Þ
p s1p e ðp, eÞ2G

will produce the maximum amount of the product. In equation (5),

G ¼ fðp, eÞjmachine e can produce part type p, e ¼ 1, 2, . . . , E, p ¼ 1, 2, . . . , Pg:


Proof of this proposition appears in appendix C.
It is noteworthy that, although the structure of a machine line that satisfies
equation (5) can produce the maximum amount of the product, Proposition 3 does
not imply that a machine line that satisfies equation (5) always reaches the highest
Downloaded by [Columbia University] at 02:17 09 December 2014

process flexibility. However, Proposition 3 implies that a machine line that satisfies
equation (5) will have higher process flexibility than those lines that do not satisfy
equation (5).
Case 3: L ¼ 1, N41 and P41
Denote the PF matrix in this case as M ~ NN ¼ ðm~ rt ÞNN , where m~ rt (r, t ¼ 1,
2, . . . , N, r 6¼ t) is the total capacity (in unit of part) that can be shifted from product r
to product t; m~ rr is the total capacity (in unit of part) that can be used to produce
product r (r ¼ 1, 2, . . . , N).
Given the BOM constraints srp (r ¼ 1, 2, . . . , N, p ¼ 1, 2, . . . , P) of the N products
and structure parameters of the machine line, we next study the approach of
determining the PF matrix.
For 8r 2 f1, 2, . . . , Ng, similar to Case 2, we can determine the maximum amount
fr(C) (in unit of the product) of the product the machine line can produce by solving
a simple linear programming problem. Then we have
X
m~ rr ¼ srp fr ðCÞ: ð6Þ
p

Intuitively, the total capacity that can be shifted from product r to product t will
never be greater than the capacity for product r or the capacity for product t,
therefore we define PF matrix for Case 3 as follows.
Definition 2: If L ¼ 1, N41 and P41, then the PF matrix M ~ can be defined as
follows:
X
m~ rr ¼ srp fr ðCÞ, r ¼ 1, 2, . . . , N, ð7Þ
( p )
X X
m~ rt ¼ min srp fr ðCÞ, srp ft ðCÞ , r, t ¼ 1, 2, . . . , N, r 6¼ t: ð8Þ
p p

For example, suppose that the BOM constraints (srp)33 of three products are
0 1
1 2 3
S ¼ @ 1 1 1 A: ð9Þ
3 2 1
Process flexibility with bill of material constraints 1575

According to Definition 2, with the BOM constraints described by equation (9),


the PF matrix of the machine line depicted in figure 2(A) is
0 1
3 3 2
M ~ ¼ @ 3 3 2 A: ð10Þ
2 2 2
Based on Definition 2, a structural property of the manufacturing system is
described in the following proposition.
Proposition 4: If L ¼ 1, N41 and P41, for given srp (r ¼ 1, 2, . . . , N, p ¼ 1, 2, . . . , P),
the structure of a machine line that satisfies
s X X
P rp C C, r ¼ 1, 2, . . . , N, p ¼ 1, 2, . . . , P ð11Þ
p srp e
Downloaded by [Columbia University] at 02:17 09 December 2014

ðp, eÞ2G

will achieve the highest process flexibility. In equation (11),


G ¼ fðp, eÞjmachine e can produce part type p, e ¼ 1, 2, . . . , E, p ¼ 1, 2, . . . , Pg:
Proof of this proposition appears in appendix D.
When each machine in a line can produce all part types, the machine line is
usually termed as a total flexible machine line (Jordan and Graves 1995), which has
the highest process flexibility.
Proposition 4 implies that a machine line that satisfies equation (11) has the same
process flexibility as that of the total flexible machine line.
Case 4: L41, N41 and P41
In this case, because of the interactions between machine lines, it is hard to define
the PF matrix to integrate the dependent part type demand with the independent
product demand. Based on the distinctive feature of the total flexible machine line,
we have the following proposition.
Proposition 5: When L41, N41 and P41, for given srp (r ¼ 1, 2, . . . , N, p ¼ 1,
2, . . . , P), if the structure of each machine line satisfies the condition described by
equation (11), then the manufacturing system will achieve the highest process flexibility.
This proposition is quite intuitive. According to Proposition 4, if the structure of
a machine line satisfies equation (11), then the machine line is totally flexible. When
BOM constraints are considered, a total flexible machine line implies that capacity of
this line can be totally shifted from producing one product to any other product
without any leftover capacity. Therefore, if the structure of each machine line
satisfies the condition described by equation (11), then the manufacturing system
that consists of these machine lines is totally flexible, i.e. it reaches the highest
process flexibility.
Proposition 5 suggests that we can improve process flexibility of a
manufacturing system that consists of multiple machine lines by improving process
flexibility of each machine line. It should be emphasised that Proposition 5 just
provides one approach of improving process flexibility of the manufacturing
system. In a practical situation, it is not always necessary to apply the approach
to all machine lines in a manufacturing system because of the interaction among
machine lines.
1576 Z. Hua et al.

3. Rules for capacity expansion problems

Based on the measures of process flexibility and the structural properties described
in the previous section, we provide some rules for the capacity expansion problems
of FMS. By applying the developed rules, feasible or near optimal solutions to the
capacity expansion problems can be obtained without much computation effort.
We next develop the rules case by case.

3.1 Rules for Case 1


When market demand increases, some machines should be added to the machine
lines. The capacity expansion problems in this case are to determine which
Downloaded by [Columbia University] at 02:17 09 December 2014

machine(s) should be added to which machine line(s). Based on Propositions 1 and 2,


rules for the capacity expansion problems are simple:
. Rule 1.1. Aggregate all machines on different lines into one machine line.
. Rule 1.2. Select the machine(s) to be added to the aggregated machine line
according to the ‘chaining policy’ (Jordan and Graves 1995, Iravani et al.
2005).
. Rule 1.3. Selected machine(s) are arbitrarily added to the machine lines.
Propositions 1 and 2 suggest that we should take multiple machine lines as a
whole when evaluating the process flexibility of a manufacturing system, so we
develop Rule 1.1 to aggregate all machines on different lines into an aggregated
machine line.
After applying Rule 1.2, Case 1 is then degenerated to the simplest case.
Since the ‘chaining policy’ is effective in improving the process flexibility for the
simplest case (Jordan and Graves 1995, Iravani et al. 2005), we then directly adopt
the ‘chaining policy’ to select the machine(s) to be added to the aggregated machine
line.
According to Propositions 2, Rule 1.3 suggests that the selected machine(s) can be
arbitrarily added to the machine lines.

3.2 Rules for Case 2


To develop rules for the capacity expansion problems in this case and in cases
afterward, we define the bottleneck part type (BNPT) as follows.
Definition 3: A part type p (p ¼ 1, 2, . . . , P) of product r (r ¼ 1, 2, . . . , N) is said to be
a BNPT of the machine line if it satisfies
s X X
P rp C> C: ð12Þ
p srp e ðp, eÞ2G

In equation (12),
G ¼ fðp, eÞjmachine e can produce part type p, e ¼ 1, 2, . . . , Eg:
Definition 3 indicates that a part type of a product is a BNPT of the machine line if
capacity shortfall of the part type occurs prior to other part types.
Process flexibility with bill of material constraints 1577

Since there is only one product type produced on a machine line, rules for the
capacity expansion problems in this case are developed as follows:
. Rule 2.1. Identify the BNPT by applying equation (12).
. Rule 2.2. Select an available machine that can produce the BNPT.
Proposition 3 suggests that with the BOM constraints, the maximum amount
of the product that the machine line can produce is restricted by the capability of
producing BNPT. So the above rules determine a machine to be added through
identifying the BNPT.
After a machine is added to the line, the BNPT may change to other part types.
If any more machines are necessary to be added, then we can apply these rules
repetitively.
Downloaded by [Columbia University] at 02:17 09 December 2014

3.3 Rules for Case 3


Based on Definition 2 and Proposition 4, rules for the capacity expansion problems
in this case can be developed as follows:
. Rule 3.1. Identify the BNPTs for all N(N41) products according to
equation (12).
. Rule 3.2. Select available machine(s) that can produce all the BNPTs.
Definition 2 and Proposition 4 suggest that the mean index and the eigenvalue
index of the PF matrix are restricted by the capability of producing BNPTs, therefore
rules in this case are similar to those in Case 2.

3.4 Rules for Case 4


Based on Proposition 5, rules for the capacity expansion problems of Case 4 can be
developed as follows:
. Rule 4.1. Assign each product type to a (some) machine line(s).
. Rule 4.2. If a machine line has only one assigned product type, then apply
Rule 2.1 and Rule 2.2 to select the machine(s) to be added.
. Rule 4.3. If multiple product types are assigned to the same machine line,
then apply Rule 3.1 and Rule 3.2 to select the machine(s) to be added.
In Rule 4.1, a product type is assigned to the machine line that the idle capacity of
the machine line is minimal after meeting demand of the product. If demand of a
product cannot be met by the machine line, then additional machine lines are selected
in a similar way. If a machine line still has leftover capacity after it is assigned to a
(some) product type(s), then other product types are also likely to be assigned to this
machine line. If none of the current machine lines can meet the demand of a product
anymore, we assign the product to the machine line which needs the least investment
during its capacity expansion. After all products have been assigned to machine lines,
Rule 4.2 or Rule 4.3 will be applied to select the appropriate machine types to
be added.
In the rules described above, it can be observed that the relationship between
components defined by the BOM constraint plays an important role. The rules are
developed for selecting available machines that can produce the bottleneck part type
1578 Z. Hua et al.

1×a 2×B

2×a 1×b

Figure 3. Illustrative example of a multi-level BOM.


Downloaded by [Columbia University] at 02:17 09 December 2014

Table 3. The accumulated two-level BOM of product A.

a b B

Product A 5 2 2

(BNPT), and the BNPT is determined based on capacity shortfalls of the part types
appearing in the BOM. The BOM constraint srp (r ¼ 1, 2, . . . , N, p ¼ 1, 2, . . . , P) has
direct impact on capacity shortfall because it defines the dependent demand of each
part type.

3.5 Validity of the rules for multi-level BOM


In this subsection, we will illustrate that our capacity expansion rules developed
under two-level BOM constraints are also applicable in the case of multi-level BOM.
Without loss of generality, we assume a three-level BOM. As shown in figure 3,
each product A consists of one component a and two components B; each
component B consists of two components a and one component b. As usual MRP
literatures do (Hopp and Spearman 2001), we define product A as level 0 of the BOM
(the highest level of the BOM) and sub-components of level i (i ¼ 0, 1, 2, . . . , n) as
level i þ 1 of the BOM. If the same component appears in different levels of the
BOM, we set the maximal level as the level it is in. Thus component B is in level 1,
both components a and b are in level 2.
To apply our capacity expansion rules to a multi-level BOM, we should first
convert the multi-level BOM into the accumulated two-level BOM, which can be
easily done in two steps:
Step 1: Take all different components appearing in the multi-level BOM, no
matter which level they appear in, as the components of level 1 in the accumulated
two-level BOM.
Step 2: For each component type listed in the accumulated two-level BOM,
accumulate the number of the component each unit of the product requires.
By adopting the above method the accumulated two-level BOM of product A is
shown in table 3.
Process flexibility with bill of material constraints 1579

After obtaining the accumulated two-level BOM, we can apply our capacity
expansion rules to make capacity expansion decisions. We next elucidate the reasons
why our rules can work correctly in the case of multi-level BOM.
For the first three cases (Cases 1–3), our capacity expansion rules are
developed for one machine line (in Case 1, Rule 1.1 aggregates all machines on
different lines into one machine line; Cases 2 and 3 are defined for one
machine line). Recall the basic requirement of the manufacturing system
(we make at the beginning of section 2) that all component types of a product
be produced on the same machine line. If this requirement is also satisfied in the
multi-level BOM case, then our rules can be directly applied to the accumulated
two-level BOM to obtain the right capacity expansion decisions. This conclusion
is true because the difference between a multi-level BOM and its accumulated
two-level BOM is that the former requires production order of components
Downloaded by [Columbia University] at 02:17 09 December 2014

while the latter does not, and this difference has no impact on capacity
expansion problems. For production planning or scheduling problems, we may
decide that components in lower levels of the BOM should be processed in
advance of components in higher levels. For capacity expansion problem,
however, the production order of components is not a problem, and the
accumulated demand of each component is enough because the planning horizon
of capacity expansion problems (usually a few years) is much longer than that
of production scheduling problems (usually in days or hours). Since the rules we
developed for Case 4 are based on rules for Cases 2 or 3, they are also
applicable to the accumulated two-level BOM.

4. Numerical results

To verify the rationality of the proposed measures of process flexibility and the
effectiveness of the developed rules for capacity expansion problems, results of some
computation experiments are reported. Since the shortfall index (see details in Jordan
and Graves 1995) is widely adopted as the reference in evaluating process flexibility,
so we first compare the eigenvalue and mean indexes of the PF matrix with the
shortfall index to test the proposed measures. We then assess the developed rules by
comparing the resultant solutions with the optimal solutions to capacity expansion
problems on investment and shortfall.

4.1 Results of Case 2


Suppose p ¼ 3, and seven machine types (symbolized from 1 to 7 respectively in
figure 4) are available for producing the three part types, which are characterized
by (1, 0, 0), (0, 1, 0), (0, 0, 1), (1, 1, 0), (1, 0, 1), (0, 1, 1), (1, 1, 1). The machine
characterized by (1, 0, 1) indicates that this machine type can produce part types
1 and 3 but cannot produce part type 2. It is clear that the machine characterized
by (1, 1, 1) is a total flexible machine, and it is the most expensive among seven
machine types.
Given the initial structure of the machine line depicted in figure 2(A), we do
30 trials to test the rationality of the proposed measures for this case. In each trial,
we compute flexibility measures after one available machine is added to the
1580 Z. Hua et al.

12

Flexibility measures 10

4 Shorfall index
Downloaded by [Columbia University] at 02:17 09 December 2014

Eigenvalue index
2 Mean index
0 1 2 3 4 5 6 7 8
Machine type

Figure 4. Results of adding different machine types in Case 2.

machine line. A randomly selected result from 30 trials is illustrated in figure 4. In


this selected trial, the BOM constraint of the product is (s11, s12, s13) ¼ (3, 2, 1). Recall
that the higher the mean or eigenvalue index, the more flexible the machine line. As
shown in figure 4, the proposed measures of process flexibility are quite consistent
with the shortfall index.

4.2 Results of Case 3


Let N ¼ 2, p ¼ 3. Given the initial structure of the machine line depicted in
figure 2(A), we do 30 trials to test the rationality of the proposed measures for this
case. In each trial, we compute flexibility measures after one available machine is
added to the machine line. A randomly selected result from 30 trials is illustrated in
figure 5. In this selected trial, the BOM constraints of two products are (s11, s12,
s13) ¼ (3, 2, 1) and (s21, s22, s23) ¼ (2, 3, 1) respectively. As shown in figure 5, the
proposed measures of process flexibility are also quite consistent with the shortfall
index.

4.3 Results of Case 4


In this case, we do 30 trials for products with random demands. We set p as 3 in all
30 trials, set L as 2 or 3 in every 15 trials, and set N as 2, 3, or 4 in every 10 trials.
Demands of products are independent and randomly generated from the normal
distribution with the mean of 2C and the standard deviation of C, where C is the
capacity of a machine type. BOM constraints srp (p ¼ 1, 2, . . . , P) of product r (r ¼ 1,
2, . . . , N) are randomly generated from [0, 20]. Because we cannot determine the
corresponding PF matrix in this case, we do not therefore compare the mean and
eigenvalue indices with the shortfall index. Instead, we compare the solutions
Process flexibility with bill of material constraints 1581

12

10
Flexibility measures
8

Shortfall index
2
Eigenvalue index
Downloaded by [Columbia University] at 02:17 09 December 2014

0 Mean index
0 1 2 3 4 5 6 7 8
Machine type

Figure 5. Results of adding different machine types in Case 3.

14

12

10
Shortfall

4
Approximate solution
2 Optimal solution
0 2 4 6 8 10 12
Trial

Figure 6. Comparison between shortfalls of optimal solutions and approximate solutions.

obtained by applying our rules with the optimal solutions to capacity expansion
problems on shortfall. The 10 randomly selected results of 30 trials are reported
in figure 6. Figure 6 shows that the gaps between shortfalls achieved by applying our
rules and those by the optimal solutions are small.
1582 Z. Hua et al.

Table 4. Comparison between the optimal solutions and solutions by applying our rules.

Cases Average jIo  Irj/Io  100% (%) Average jSo  Srj/So  100% (%)

Case 1 0.73 0.00


Case 2 1.53 0.00
Case 3 2.19 6.38
Case 4 6.78 26.17

Statistical results about the gaps between shortfalls and investment error
achieved by applying our rules and those by the optimal solutions for all cases are
listed in table 3, where we also show the average of investment errors and shortfall
errors of 30 trials. The investment error is computed by jIo  Irj/Io  100%, where
Downloaded by [Columbia University] at 02:17 09 December 2014

Io is the investment required by the optimal solution, and Ir is the investment


according to our rules. The shortfall error is computed by jSo  Srj/So  100%,
where So is the shortfall according to the optimal solution, and Sr is the shortfall
according to our rules.
Results in table 4 show that, in the first two cases, our rules can provide near
optimal solutions. In the other two cases, our rules can generate fairly good
solutions.

5. Conclusions

This paper provides measures of process flexibility when the BOM constraints are
considered. Structural properties of achieving high process flexibility in different
cases are also investigated. As an application of the described structural properties,
we develop some rules for the capacity expansion problems of FMS, which can
provide good or near optimal solutions to the complex integer programming
problem without much computation effort.
Further research effort can be invested to structural property and rules for the
capacity expansion problems in Case 4. It is also interesting to extend the research of
this paper to supply chain coordination.

Acknowledgements

The authors would like to thank the anonymous referee and the editor for
their helpful comments and suggestions which significantly improved the paper.
The authors are also grateful to Dale Simpson for reviewing and editing the paper.
This research was supported by the NSFC (Grant No.: 70571073) and the
Specialized Research Fund for the Doctoral Program of Higher Education
(Grant No.: 20050358002).

Appendix A: Calculation details about figure 2

Denote the maximum amount of the product that line A and line B (depicted in
figure 2(A and B) respectively) can produce as xA and xB respectively. Given the
Process flexibility with bill of material constraints 1583

BOM constraint of product 1, (1, 6, 1), xA can be obtained by solving the following
simple linear programming problem
max xA
s:t:
xA þ 6xA þ xA  3C, ðA1Þ
xA  C, ðA2Þ
6xA  3C, ðA3Þ
xA  2C, ðA4Þ
xA  0: ðA5Þ
Downloaded by [Columbia University] at 02:17 09 December 2014

In the above linear programming problem, constraint (A1) requires that the total
capacity used to produce the product (in units of part types) does not exceed the
total capacity of all machines in line A; constraints (A2)–(A4) require that the total
capacity used to produce each part type does not exceed the total available capacity
for the part type; constraint (A5) is the nonnegative constraint.
Similarly, xB can be obtained by solving the following linear programming
problem
max xB
s:t:
xB þ 6xB þ xB  3C
xB  2C ðA6Þ
6xB  2C
xB  2C
xB  0:
Solving the above two linear programming problems, we will have
3C C
xA ¼ and xB ¼ :
8 3
These solutions indicate that the capacity of line A has been fully utilized in
producing

3C
8
units of the product, and that line B can produce
C
3
units of the product with
C
3
capacity (in units of part types) idled.
1584 Z. Hua et al.

Appendix B: Proof of Proposition 1

Without loss of generality, we set L as 2, and structures of the two machine lines are
depicted in figure 7. Denote the SF matrices of the two machine lines as
M0PP ¼ ðm0pq ÞPP ðp, q ¼ 1, 2, . . . , PÞ and M00PP ¼ ðm00pq ÞPP ðp, q ¼ 1, 2, . . . , PÞ
respectively.
We then aggregate the two machine lines depicted in figure 7 into an aggregated
machine line (shown in figure 8), whose PF matrix is denoted as
M mix mix
PP ¼ ðmpq ÞPP ðp, q ¼ 1, 2, . . . , PÞ.
By directly observing figures 7 and 8, we have

mmix 0 00
pp ¼ mpp þ mpp , p ¼ 1, 2, . . . , P: ðB1Þ
Downloaded by [Columbia University] at 02:17 09 December 2014

If there is any wrapped non-overlapping path(s) between two different part types
(depicted as dashed lines in figure 8), then we have

mmix 0 00
pq  mpq þ mpq p 6¼ q, p, q ¼ 1, 2, . . . , P: ðB2Þ

Equations (B1) and (B2) indicate

Mmix  M00 þ M0 ðB3Þ

(A) Machine type Demand type (B) Machine type Demand type

1 1 1 1
..
2 2 e 2
.. .. .. ..
E P E P

Figure 7. Illustration of two machine lines.

Machine type Demand type

1 1 1

..
2 2 e

.. .. ..
E P E

Figure 8. Illustration of the aggregated machine line.


Process flexibility with bill of material constraints 1585

denote the mean index of matrix M as Ime(M), and the eigenvalue index of matrix M
as Iei(M). Equation (B3) implies that
Ime ðMmix Þ  Ime ðM0 Þ þ Ime ðM00 Þ: ðB4Þ
But equation (A3) does not suggest any order relationship among Iei (Mmix), Iei (M0 )
and Iei(M00 ).
When there are more than two machine lines, these conclusions can be proved by
applying the above process iteratively.
This completes the proof of Proposition 1. œ

Appendix C: Proof of Proposition 3


Downloaded by [Columbia University] at 02:17 09 December 2014

When L ¼ 1, N ¼ 1 and P41, under given BOM constraints s1p (p ¼ 1, 2, . . . , P),


if the structure G* of the machine line satisfies
s X X
P 1p C C, p ¼ 1, 2, . . . , P, ðC1Þ
p s1p e ðp, eÞ2G

where G* ¼ {(p, e)j machine e can produce part type p, e ¼ 1, 2, . . . , E, p ¼ 1, 2, . . . , P}.


Then it is obvious that, for 8p 2 f1, 2, . . . , Pg, the capacity allocated to produce part
type p can be
s X
P 1p C,
p s1p e

and then the total capacity used to produce the product is


X s1p X X
P C¼ C:
p p s1p e e

This implies that the capacity of the machine line under structure G* can be fully
utilized in producing the product, and thus this machine line will produce the
maximum amount of the product.
This completes the proof of Proposition 3. œ

Appendix D: Proof of Proposition 4

Because the PF matrix is real, symmetric and non-negative, the machine line will
achieve the highest process flexibility (both on the mean index and eigenvalue index)
if each element of the PF matrix arrives at its maximum value.
From definition 2, elements of the PF matrix are defined as
X
m~ rr ¼ srp fr ðCÞ, r ¼ 1, 2, . . . , N, ðD1Þ
p
( )
X X
m~ rt ¼ min srp fr ðCÞ, srp ft ðCÞ , r, t ¼ 1, 2, . . . , N, r 6¼ t: ðD2Þ
p p
1586 Z. Hua et al.

From equations (D1) and (D2), it is obvious that, each element of the PF matrix
arrives at its maximum value if and only if fr(C) is maximised for 8r 2 f1, 2, . . . , Ng.
For 8r 2 f1, 2, . . . , Ng, from Proposition 3, we know that fr(C) is maximized if
s X X
P rp C C, p ¼ 1, 2, . . . , P: ðD3Þ
p srp e ðp, eÞ2G

This completes the proof of Proposition 4. œ

References
Downloaded by [Columbia University] at 02:17 09 December 2014

Chen, Z.L., Li, S. and Tirupati, D., A scenario based stochastic programming approach for
technology and capacity planning. Comp. Oper. Res., 2002, 29, 781–806.
Gavish, R., Flexibility in multi-product multi-assembly multi-component systems. Masters
thesis, MIT, Cambridge, MA, 1994.
Graves, S.C. and Tomlin, B.T., Process flexibility in supply chains. Manage. Sci., 2003, 49,
907–919.
Hopp, W.J. and Spearman, M.L., Factory Physics, 2nd ed., 2001 (Irwin McGraw-Hill:
New York, NY).
Hopp, W.J., Tekin, E. and van Oyen, M.P., Benefits of skill chaining in serial production lines
with cross-trained workers. Manage. Sci., 2004, 50, 83–98.
Hua, Z.S. and Banerjee, P., Aggregate line capacity design for PWB assembly systems. Int. J.
Prod. Res., 2000a, 11, 2417–2441.
Hua, Z.S. and Banerjee, P., A model for line capacity design for PWB assembly systems.
Robot. Comp. Integ. Manuf., 2000b, 16, 241–257.
Huang, G.Q., Zhang, X. and Liang, L., Towards integrated optimal configuration of platform
products, manufacturing processes, and supply chains. J. Oper. Manage., 2005, 23,
267–290.
Iravani, S.M., van Oyen, M.P. and Sims, K.T., Structural flexibility: a new perspective on the
design of manufacturing and service operations. Manage. Sci., 2005, 51, 151–166.
Jordan, W.C. and Graves, S.C., Principles on the benefits of manufacturing process flexibility.
Manage. Sci., 1995, 41, 577–594.
Rajagopalan, S., Flexible versus dedicated technology: a capacity expansion model. Int. J.
Flex. Manuf. Syst., 1993, 5, 129–142.
Sethi, A.K. and Sethi, S.P., Flexibility in Manufacturing: a survey. Int. J. Flex. Manuf. Syst.,
1990, 2, 289–328.
Skipworth, H. and Harrison, A., Implications of form postponement to manufacturing: a case
study. Int. J. Prod. Res., 2004, 42, 2063–2081.
Stecke, K.E. and Raman, N., FMS planning decisions, operation flexibilities, and system
performance. IEEE Trans. on Eng. Manage., 1995, 42, 82–90.
Tseng, H.E., Chang, C.C. and Chang, S.H., Applying case-based reasoning for product
configuration in mass customization environments. Exp. Syst. with Appl., 2005, 29,
913–925.
van Hop, N., Approach to measure the mix response flexibility of manufacturing systems. Int.
J. Prod. Res., 2004, 42, 1407–1418.
van Mieghem, J.A., Investment strategies for flexible resources. Manage. Sci., 1998, 44,
1071–1078.
Yan, H., Yu, Z.X. and Cheng, T.C.E., A strategic model for supply chain design with logical
constraints: formulation and solution. Comp. Oper. Res., 2003, 30, 2135–2155.
Yeung, J.H.Y., Wong, W.C.K., Ma, L. and Law, J.S., MPS with multiple freeze fences in
multi-product multi-level MRP systems. Int. J. Prod. Res., 1999, 37, 2977–2966.

You might also like