You are on page 1of 17

Computers and Chemical Engineering 32 (2008) 3153–3169

Contents lists available at ScienceDirect

Computers and Chemical Engineering


journal homepage: www.elsevier.com/locate/compchemeng

Supply chain redesign through optimal asset management


and capital budgeting
P.K. Naraharisetti a , I.A. Karimi b,∗ , R. Srinivasan a,b
a
Institute of Chemical and Engineering Sciences, 1 Pesek Road, Jurong Island, Singapore 627833, Singapore
b
Department of Chemical & Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, Singapore 117576, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: Manufacturing and inventory facilities along with the materials present in them are the physical assets
Received 17 October 2006 of a company. Companies strive to maximize their shareholder values by managing these assets through
Received in revised form 27 March 2008 a variety of business decisions. In this work, we present a new mixed-integer linear programming model
Accepted 19 May 2008
for asset management and capital budgeting, which can aid the decision-making process for supply
Available online 28 May 2008
chain redesign. The decisions include facility location, relocation, investment, disinvestment, technol-
ogy upgrade, production–allocation, distribution, supply contracts, capital generation, etc. To the best
Keywords:
of our knowledge, this model is the first to address disinvestment, technology upgrade, material supply
Capacity
Planning
contracts, and loans and bonds for capital generation, while including strategic asset management and
Investment tactical planning, capacity planning, financial/regulatory factors, and production–distribution. We illus-
Disinvestment trate the impact of ignoring disinvestment and/or relocation decisions as a 14% decrease in profit for an
Supply chain design example case study.
Contracts © 2008 Elsevier Ltd. All rights reserved.

Note: If the subscript has only one entity, then it corresponds to a demand of materials, minimizing the financial losses that may
facility. If it has two entities, then the first refers to material and the arise.
second to location/facility. If it has three entities, then the first refers Asset management involves several business decisions such
to material, the second to material’s origin, and third to material’s as supply chain redesign and capital budgeting. While the for-
destination. Superscript L (U) denotes the lower (upper) limit or mer involves planning (managing) the locations, timings, and
the lowest (highest) value in a parameter vector, and superscript amounts of capacity changes in physical assets such as pro-
TU denotes technology upgrade. duction/inventory/distribution facilities, the latter involves the
allocation (management) of financial assets such as capital to
1. Introduction various physical assets. Examples of some important budgeting
decisions are allocating of capital to procure materials, produce
A company owns or employs a variety of assets, and manages products, invest in new facilities, etc. Dixit and Pindyck (1995) indi-
them efficiently to maximize shareholder value and customer sat- cate that even the disinvestment of a facility can be considered as an
isfaction. While many of these assets are physical, e.g. materials, investment, because companies do incur costs due to factors such
human resources, and financial instruments; others such as techno- as employee compensation.
logical know-how are non-physical. Technological know-how can The business environment is ever-changing due to the complex
also be viewed as an asset, because it helps companies produce dynamics of supply chains (Julka, Srinivasan, & Karimi, 2002; Julka,
better quality products at reduced costs, leading to greater cus- Karimi, & Srinivasan, 2002) including varying supplies of raw mate-
tomer satisfaction and profit. Companies with better technological rials and changing demand patterns due to the opening of new
know-how have an advantage over others and can improve the markets, introduction of new and technologically superior prod-
quality of their product at a strategic time. Similarly, in addition ucts, changes in trade barriers, product obsolescence, and business
to cash-on-hand, loans, bonds, and even contracts for the sup- competition. As markets expand with globalization, companies are
ply of raw materials can be considered as financial assets because venturing into previously unknown territories by locating new
a contract hedges risk from the uncertainties in the supply and manufacturing and distribution facilities or are partnering with
other companies in geographically distant locations, which is lead-
ing to increased complexity in the supply chain. The world trade in
∗ Corresponding author. Tel.: +65 6516 6359; fax: +65 6779 1936. merchandize for the year 2004 was $8.9 trillion in exports and $9.3
E-mail address: cheiak@nus.edu.sg (I.A. Karimi). trillion in imports (World Trade Organization: International Trade

0098-1354/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compchemeng.2008.05.008
3154 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

Nomenclature
QSre et maximum amount of material that can be pur-
Indices chased from a supplier without contract
b number of bonds RBbt rate of interest paid for a bond
e an entity (raw material supplier, raw material, RDt total revenue from disinvestments at t
inventory facility, or production facility) RFet revenue factor for an entity for disinvestment
l number of loans RLlt t the rate of installment paid at t for a loan l taken at
t period t
STle maximum number of periods for which a facility
Sets p p
may operate during PTe (STle < PTe )
FM set of materials that a given facility may hold TCIt cost for investments in all facilities in a given interval
D set of distribution centers ending at t
I set of inventory facilities TDe time before which a facility cannot be disinvested
IIN set of input inventory facilities in a production facil- TFet throughput factor for an entity at t
ity TFUet throughput factor at t for an entity that has under-
IINR set of input inventory facilities in a production facil- gone a technology upgrade
ity, which hold a recycled material TUPe initial interval during which no technology upgrade
M set of materials is allowed for a facility
MF set of facilities that a given material may occupy TXnt tax rate in a nation at t
MFF set of arcs that a given material may follow VEUet value of an entity per unit capacity
OIN set of output inventory facilities in a production WDCe et unit cost for disposing waste at an entity at t
facility YTt year corresponding to t
P set of production facilities ˛e e et unit cost of transporting a material from one entity
PI set of facilities—production and inventory to another at t
PIN set of facilities that do not exist at the beginning of ˇcet fixed unit cost of operating a storage facility at t
the horizon ˇmet variable unit cost of operating a storage facility at t
RCM set of recycled materials cet fixed unit cost of operating a production facility at t
RM set of raw materials including intermediates tet variable unit cost of operating a production facility
S set of raw material suppliers at t
UM set of materials that may undergo a technology ıcre et unit cost of material purchased in range r and (r + 1)
upgrade at time t under a contract
UMM set of material that a given material is upgraded into ıse et unit cost of material purchased at t without a con-
UP set of production facilities that may undergo a tech- tract
nology upgrade suet cost of starting production at an entity at t
UPF set of facilities that may send materials to a given sdet cost of stopping production at an entity at t
facility e et unit price of a material sold at an entity at t
˝et variable unit cost for expanding a facility
Parameters ˝eN fixed cost for adding capacity at a new location
BLb duration after which a bond matures ˝eO fixed cost for adding capacity at an existing location
BPbt principal value of a bond b issued at t  e e stoichiometry coefficient of a material that is pro-
CLet interval for which a contract may be signed cessed in a production facility
CLTe construction lead time for a facility
De et demand for a material at an entity at t Binary variables
DPet depreciation price or value of a plant on which EXet 1, if a facility is sold and the capacity is no longer
depreciation is calculated available at t
DRet rate of depreciation on a facility EYet 1, if an added capacity is available for production at
EAN e minimum capacity for a new facility at a location at t
t EZet 1, if a facility is in operation at t
EAO e minimum capacity expansion allowed for an exist- TUet 1, if a facility undergoes a technology upgrade at t
ing facility at t Vre et 1, if material is purchased through a contract in the
IvI investment interval during which there is a limit on range r and (r + 1) from an entity at t
investment amount
IRt interest rate for computing time value of money Continuous 0–1 variables
KCe et fixed cost for purchasing a raw material in the range CSe et 1, if a contract is signed for purchasing a material
r and (r + 1) from an entity at t when a contract is from an entity at t
signed EEYet 1, if an entity exists at t
LLl duration in which a loan of type l has to be paid back ESYet 1, if an entity that existed before t is sold off at t or
NEe maximum number of times that a facility may before
expand UFet 1, if a facility has undergone a technology upgrade
p
PTe interval in which routine maintenance must be done
at least once for a facility Continuous variables
QCre et maximum amount of material that may be pur- cbbt cost of paying back a bond b at time t
chased from a supplier under a contract cdet cost of depreciation at t for an entity
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3155

As new sources of raw materials and demands are identified,


a need may arise for locating new production and distribution
ciet cost for investing in a facility at t facilities, shutting down old facilities (due to technologically obso-
ciN
et fixed cost for investing in a new facility at t lete processes, high operating costs, etc.), and hence the need to
ciO
et fixed cost for investing in an existing facility at t redesign the supply chain network. Some recent examples of such
cpret cost of production at an entity at t decisions involve multinationals such as Airbus (in talks of locating
crme et cost of purchasing a raw material at an entity at t an assembly plant in China), Ford, General Motors, and Merck (shut-
cstet cost of storing at an entity at t down of plants in USA), and Samsung (new facilities in South Korea).
cswet cost of switching production at an entity at t Also, there is a growing trend of relocating production facilities to
csaet income from sales for an entity at t low-cost centers, e.g. Nike from Indonesia to China. Some examples
ctset cost of transportation for an entity at t of mergers and acquisitions are Chevron Texaco acquiring Unocal
cwet cost for disposing waste at an entity at t in 2005 ($18 billion), PetroChina (a subsidiary of China National
dcre et amount of material purchased in the range r and Petroleum Corporation International) acquiring PetroKazakhstan
(r + 1) from an entity at t under a contract in 2005 ($4.18 billion), and Sanofi-Synthelabo acquiring Aventis
dse et amount of material purchased from an entity at t to become Sanofi-Aventis in 2004 (D 54.5 billion). The worldwide
without a contract merger and acquisition volume was about $2.7 trillion in 2005, an
iait installment for a loan at t increase of about 38% from 2004, the best year since 2000, and the
shNPV net present value of the shareholder value for the third best year in history (Thomson Financial, Mergers & Acqui-
entire horizon sitions Review, Fourth Quarter, 2005). The decisions of mergers
npt net profit at t and acquisitions require integration of new assets within the exist-
rdeet revenue from disinvesting an entity at t ing supply chain, eliminating duplication or redundancy in assets,
and hence the need to redesign the supply chain. Other exam-
Positive variables ples of investments in the chemical industry include Ciba Specialty
cprtuet change in cost of production after technology Chemicals to manufacture anti-oxidants for plastics, Schenectady
upgrade at an entity at t International to manufacture alkylated phenol, and Huntsman Cor-
csuet cost of startup at an entity at t poration to manufacture polyetheramines, all on Jurong Island in
csdet cost of shutdown at an entity at t Singapore. There are also situations, where a large company spins
ctxnt amount of taxes paid in nation n at t out one or more independent offshoots, for example Zenecca from
eaet amount of capacity added by a facility at t ICI, etc. In the event of a spin-off, the company must allocate existing
ecapet capacity of an entity at t assets to each new business entity.
ede e et amount of a material e sent from e to e at t To get a perspective on the amounts of investments, consider
eee et amount of an entity present at a location at t the business of energy. It is estimated that the total investment
ibbt amount of capital raised through bonds b at t required to meet the energy needs worldwide through 2030 is $16
lalt amount of capital raised through loans l at t trillion and a considerable amount ($200 billion per year as esti-
nbbt number of bonds issued at t mated by International Energy Agency’s World Energy Investment
se et amount of material sold at an entity at t Outlook, 2003) of these investments will be in the oil and gas indus-
we et amount of material disposed as waste at an entity try. ExxonMobil invested a net (investment in property, plant, and
at t equipment) of about $100 billion in the year 2003, about 55% of
which is in the upstream oil and gas, about 40% is in the downstream
and chemicals business, the rest being ‘others’, and discontinued
Statistics, 2005). This trade is only set to increase due to the ever- operations (ExxonMobil Financial & Operating Review, 2003).
increasing global population; 6.1 billion in 2000 and estimated to Clearly, several industries and companies face the challenging
exceed 9.0 billion in 2050 (World Population Prospects: The 2004 task of making the best business decisions, and hence addressing
Revision, UN Population Division). This gives us an idea of the com- the problems of locating new facilities, discontinuing operations,
plexities that may exist in the supply chains and the need to make redesigning the supply chain, and developing a plan for implement-
efficient business decisions. The increased complexities in supply ing various decisions is of great importance. Instead of analyzing
chains are making it increasingly difficult to manage and optimize each new project individually using metrics such as the rate of
them. In order to cope with reducing profit margins, improve pro- return or payback period, a holistic analysis that considers the
ductivity, and reduce costs, businesses have identified the need for impact on the entire supply chain would be useful. In this work,
better decision support tools (Lasschuit & Thijssen, 2004; Holland, we address these issues, but first, we review the existing work.
Shaw, & Kawalek, 2005; Timpe and Kallrath, 2000).
Several resource planning systems exist to enable improved 1.1. Previous work
coordination in the movement of materials, information, and
finance across various entities in a supply chain. However, as It is generally believed that all strategic decisions are made by
Stadtler (2005) noted “It is well known that the strength of the management at the highest level in an organization. However,
the transactional enterprise resource planning (ERP) systems is we can see from Slagmulder, Bruggeman, and Wassenhove (1995),
not in the area of planning. Hence, advanced planning systems Akalu (2003), Peel and Bridge (1998), and Pirttila and Sandstrom
(APS) have been developed to fill this gap. APS are based on (1995) that contrary to this popular notion, many decisions are gen-
the principles of hierarchical planning and make extensive use erally initiated by the various departments at the other end of the
of solution approaches known as mathematical programming and hierarchy and reach the higher management after being approved
meta-heuristics”. These advanced planning systems – decision sup- at various levels or departments in the business. The project pro-
port systems or analytical information technology tools, as they posals that reach the management would have won the approval
are sometimes called (Shapiro, 2001) – are used to redesign supply of most members in the organization, and hence the management
chain and make optimal asset management and capital budgeting would very likely have a positive opinion on them. However, all
decisions. the above works concluded that the process of strategic decision-
3156 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

making varies widely from organization to organization and even emphasis on financial strategies. They discuss several issues such
from one project to another project within the same organization. as international tax planning, means of financing investments,
Also, as the sizes of organizations and the complexities of their transfer pricing and uncertainties. In addition, they discuss various
supply chains increase, decentralized decision-making may give modeling and solution strategies such as simulation–optimization,
suboptimal solutions. Hence, it is important to consider the entire real options, multi-stage models, among others. Amiri (2006) and
supply chain. Often, the management relies on an overall empirical Bjorkqvist and Roslof (2005) presented a capacity expansion model
analysis, mathematical/financial analysis, or the leadership. that handles capacity increments in multiples of some discrete
Qualitative approaches for strategic decision-making include amounts (defined a priori). Bok, Lee, and Park (1998) presented a
those of Olhager, Rudberg, and Wikner (2001), Bhatnagar, Jayaram, capacity expansion model for the Korean petrochemical industry.
and Phua (2003), and Partovi (2006). Olhager et al. (2001) gave Bhutta, Huq, Frazier, and Mohamed (2003) presented an inte-
an empirical analysis of how manufacturing strategy and sales grated investment model with production–distribution. Gatica,
and operations planning are related and presented a long-term Papageorgiou, and Shah (2003), Levis and Papageorgiou (2004), and
capacity management framework. Bhatnagar et al. (2003) iden- Papageorgiou, Rotstein, and Shah (2001) presented capacity expan-
tified the relative importance of various factors affecting plant sion and production–distribution models for the pharmaceutical
location between Singapore and Malaysia and demonstrated the industry. Lanzenauer, von Eschen, and Pilz-Glombik (2002) pre-
advantages of locating a plant in Singapore versus Malaysia. Partovi sented a case study on upgrading technological capabilities of a
(2006) used an analytical hierarchy process combined with quality petrochemical firm. Lee, Lee, and Reklaitis (2000) and Rajagopalan
function deployment to make strategic decisions and demonstrated and Swaminathan (2001) presented a nonlinear model to consider
how Singapore is chosen as opposed to Bangkok or Beijing by capacity expansion, while Oh and Karimi (2004) incorporated the
an unnamed organization. Their result challenges the popular effect of regulatory factors in a strategic capacity planning model,
belief that a facility should be located in or moved to a loca- which is relevant in this era of globalization. Recently, Guillen, Mele,
tion where the operating costs are lower, and the study thus Bagajewicz, Espuna, and Puigjaner (2005) presented a multiobjec-
demonstrated the need for objective analytical approaches. In addi- tive supply chain design model considering capacity additions.
tion, two expert opinion papers (Grossmann & Westerberg, 2000; In the area of financial management, Sykuta (1996) presented
Grossmann, 2004) on the future of process systems engineering an empirical analysis of futures and contracts in the chemical
have highlighted the importance of enterprise and supply chain industries considering strategic decisions. Annupindi and Bassok
optimization. (1999) and Tsay, Nahmias, and Agrawal (1999) gave a comprehen-
Several researchers have considered the problem of sive analysis of contracts in supply chains and Fan (2000) described
production–allocation and distribution. To date, Arntzen, Brown, contracts in petrochemical industries. Romero, Badell, Bagajewicz,
Harrison, and Trafton (1995) have probably presented the most and Puigjaner (2003) considered cash management or budgeting
comprehensive model to address various issues in the supply chain models in planning and scheduling, while Barbaro and Bagajewicz
of the electronics industry. Pyke and Cohen (1994) presented a (2004) presented a mathematical formulation to study options
production–distribution model for a multiproduct system with contracts in the chemical industries for the short term planning.
uncertain demands. McDonald and Karimi (1997) considered Both these works deal with the operational/tactical issues in con-
production planning and distribution by grouping products into tracts. Sharma, Ammons, and Hartman (2007) considered contracts
families and included transportation times in the formulation. involving the use of computer and electronic equipment.
Wilkinson, Cortier, Shah, and Pantelides (1996) presented an inte- Based on our literature review, a formulation that considers
grated production and distribution model and solved it by using the issues of disinvestments (thus relocation) of facilities, supply
aggregated time periods. Bradley and Arntzen (1999) considered chain redesign, technology upgrade, strategic contracts related to
an integrated planning model considering production capacity the sales and purchase of materials, and loans and bonds for rais-
and inventory for a system with seasonal demands. Tsiakis, ing capital is not present in the literature. In addition, the following
Shah, and Pantelides (2001) presented a production–allocation issues or assumptions are common (of course, exceptions do exist)
and distribution model, while considering economies of scale in the existing literature and need further attention.
in material purchase/sales. Gjerdrum, Shah, and Papageorgiou
(2001) presented a nonlinear model to consider transfer pricing 1. Discrete amounts (multiples of a fixed amount) of investments
in the supply chain. Cohen and Lee (1989) and Goetschalckx, are assumed instead of continuous amounts within some limits.
Vidal, and Dogan (2002) presented models for global logistics 2. Lower limits on the capacity of a new construction and that of
system and considered transfer pricing in them. Reklaitis and the expansion of an existing facility are assumed to be identical.
co-workers (Subramanian, Pekny, & Reklaitis, 2000; Subramanian, 3. Existing facilities are not allowed to shutdown, so a maintenance
Pekny, & Reklaitis, 2001; Subramanian, Pekny, Reklaitis, & Blau, shutdown cannot occur.
2003) presented comprehensive strategic planning studies on 4. Lower limits on the capacity utilizations of production facilities
R&D pipeline management using a simulation–optimization are assumed to be zero (rather than nonzero) in order to avoid
framework. Perea-Lopez, Ydstie, and Grossmann (2003) presented infeasibilities in the absence of raw materials.
a model for production–distribution planning of multi-product
batch plants and applied model predictive control strategy on In this work, we consider a multi-echelon supply chain net-
a rolling horizon. Ryu and Pistikopoulos (2005) presented an work of a multi-national corporation (MNC) and present a novel
enterprise-wide production and distribution model considering deterministic asset management model, which addresses the
various operational policies. Oh and Karimi (2006) considered aforementioned issues using a mixed-integer linear programming
a production–distribution planning model considering duty (MILP) approach. We have previously presented preliminary ver-
drawbacks, which are becoming increasingly important due to sions of this model in Naraharisetti, Karimi, and Srinivasan (2005)
globalization and competition between organizations. and Naraharisetti, Karimi, and Srinivasan (2006). We also combine
Considerable literature exists on the specific issues of new both the strategic (facility planning) and tactical asset management
facility location and/or technology upgrade of an existing facility. (production–allocation–distribution) problems into an integrated
McDonald and Reklaitis (2004) have presented a comprehensive asset management, capital budgeting, and supply chain redesign
review on the design and operation of supply chains with a greater model that handles several key decisions in the end-to-end supply
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3157

chain of an MNC, as opposed to solving these problems individually. its additions that occur later. However, the rate is the same for all
In addition, the formulation incorporates production changeover additions. Inflation rates are period-dependent, nation-dependent,
costs, depreciation, taxes, regulatory factors, transportation costs, and pre-fixed. In addition to the above description, we assume the
storage costs, production costs, inflation rates, interest rates, loans following:
and bonds, etc. and aims to maximize the net present value of the
total assets including financial and physical ones. • Investments and disinvestments:
1. An expansion does not affect the current production and it
2. Problem description cannot begin, while a previous expansion/construction is in
progress.
The chemical supply chain of an MNC consists of raw material 2. The limit on the total amount of investment in a given interval
suppliers, production facilities, distribution centers, and inventory is known. Similarly, constraints apply on disinvestment (ex:
facilities. We take the planning horizon to consist of many periods of the amount of investment that can be made in a year is one
predefined identical length (week, month, quarter, etc.) and define billion).
interval as a set of consecutive periods. 3. A facility cannot be disinvested partially.
Each production facility represents one single process for pro- 4. Cost and income associated with each investment and disin-
ducing products and intermediates from raw materials. Several vestment are linear functions of capacity.
such processes may exist at a location and each is considered a 5. Once a facility is disinvested, it remains sold off for the entire
distinct facility. We define throughput as the total mass flow of planning horizon.
all materials into the facility and throughput factor as the ratio of 6. Specifications of a product improve on technology upgrade.
throughput to capacity. If the throughput is zero, then the facility is Hence, the upgraded material is treated as a different mate-
not in operation. Otherwise, its throughput must be between some rial. In addition, the stoichiometry may change on a technology
predefined limits. Note that although the throughput, as defined upgrade.
here, is different from the production capacity defined on the basis 7. The limits on transportation, production, and expansion poli-
of a primary product, both are equivalent as long as the materi- cies are all known. For instance, a facility may not expand more
als are always consumed in given fixed proportions. A facility can than thrice in the planning horizon.
operate for only some periods in any given interval (e.g. 11 of 12 • Production–allocation:
periods) to allow time for maintenance. A shutdown facility incurs 8. Startups and shutdowns of a facility incur costs.
a fixed cost in every period. We define construction to mean the 9. Back orders (negative inventory) are not fulfilled.
building of a new facility, expansion as the addition of capacity to 10. A facility’s throughput may change on technology upgrade. It
an existing facility, and disinvestment as a sell-off. A new facility is specified as a given percentage of the original throughput.
can be constructed only once during the horizon. Construction lead 11. Production costs are linear functions of capacity and amount
time or the time between the starts of operation and construction of material produced.
of each facility is fixed and known a priori. A facility once built or 12. Production cost of a technologically improved product can be
an existing facility may undergo multiple capacity expansions. We different from that of its parent product.
allow different lower limits on capacity changes for expansion and 13. Inventory costs are linear functions of facility capacity and
construction. A facility can undergo a technology upgrade or dis- amounts of material stored.
investment only once during the planning horizon. While capacity • Purchases and sales:
expansion is not integral to a technology upgrade, both expansion 14. Time for transporting a material from one location to
and upgrade may occur at the same time. another is negligible. This is a fair assumption, as time
Each production facility has two inventory facilities, namely period is expected to be a month or longer in this strategic
input inventory and output inventory. The former is for the storage problem.
of raw materials, while the latter for the storage of products and 15. All purchases and sales are free on board (FOB). Hence, the
intermediates. Unused materials are recycled back to their input cost, insurance, and freight (CIF) for raw materials are borne
inventories. Each production facility sends its products to distribu- by the MNC and those for product sales are borne by the cus-
tion centers and intermediates to the input inventory facilities of tomers (Karimi, Srinivasan, & Por, 2002). Taxes on purchase
downstream production facilities. All inventory facilities, whether are borne by the MNC and those on sales are borne by the
at production facilities or distribution centers, act as sales loca- customer.
tions. Thus, even the inventory facilities at production facilities act 16. Regulatory factors on import–export are included in the trans-
as pseudo-distribution centers. A production facility may produce portation costs and thus the tariffs are proportional to the
excess materials that cannot be sold. This excess material is then amount of material. There are no fixed costs for import–export
assumed to be sold at a discounted price. The limit on the disposal tariffs and transportation costs.
of this excess material is known a priori. On the other hand, the 17. If a contract exists for a material with a supplier, then that
demand and selling price for a technologically upgraded product material must be purchased under the contract and not out-
may be different from those of its parent product. side the contract.
The MNC may raise its capital from profits, external loans, or • Loans and bonds:
bonds. Payment installments of loans are predefined and may vary 18. A loan is repaid fully in small periodic installments and no
with periods. In order to reduce risks, the MNC may use long-term lump sum is paid at the end of the loan period.
contracts for both selling products (hedge risk against reduced 19. A bond is repaid in lump sum with interest at the end of its
demand or low product prices) and buying raw materials (hedge maturity.
risk against reduced raw material supply or increase in price). Con-
tracts provide quantity flexibility to afford economies of scale and 2.1. Problem statement
penalties for the under-supply of materials. The durations of con-
tracts are unknown and are treated as decision variables. With this, the problem for managing the assets of the MNC can
Depreciation rates of assets may vary with asset and period. Also, be stated as follows. Given the following for the entire planning
it may be different for a facility that exists at the beginning and horizon:
3158 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

1. capacities of all existing production and inventory facilities; In practice, a few large capacity additions would be preferred
2. demand profiles of products including the technologically over several small ones, because project handling is easier that
upgraded products for all customers; way. Therefore, it may be desired to limit the number of capacity
3. raw material availability profiles from all suppliers; additions for a facility by using,
4. price forecasts for all materials; 
5. transportation rates, insurance rates, import–export tariffs; EYet + TUe t  ≤ NEe e ∈ PI, t > CLTe , e ∈ UP, t  > TDe (2)
6. depreciation rates, interest rates for loans and bonds, inflation t
and tax rates in different nations, constraints on imports and From Eq. (F1), it is clear that a facility e cannot change its status
exports; (exists or does not) for t ≤ min[CLTe , TDe ]. Note that all equations
7. production rates of all products, their limits, inventory costs, with labels starting with ‘F’ are not written as explicit constraints,
production and changeover costs; but they are meant for fixing variables.
8. construction lead times, investment and disinvestment costs and
incomes, limits on investments and disinvestments, and costs of EEYet = EEYe,1 e ∈ PI, t ≤ min[CLTe , TDe ] (F1)
technology upgrades. If a facility does not exist at time (t − 1), but starts production due
to a capacity addition at time t, then it should be taken as existing
Determine the following plans for maximizing net present value at time t. In other words, EYet = 1 and EEYe(t−1) = 0 ⇒ EEYet = 1.
of the shareholder value (shNPV) of the MNC.
EEYet ≥ EYet e ∈ PI, t > CLTe (3)
1. raw material purchase and product sales;
2. production, material distribution, and inventory; EEYet ≤ EEYe(t−1) + EYet  − EXet  e ∈ PI, t > min[CLTe , TDe ],
3. investment, disinvestment, and technology upgrade; t  > CLTe , t  > TDe (4)
4. contracts, loans, and bonds.
Eq. (4) also ensures that a facility cannot be disinvested at t, if it
3. MILP formulation does not exist at (t − 1), and it can exist at t, only if it either existed at
(t − 1) or it adds capacity at t. Similarly, it also ensures that a facility
To formulate this integrated asset management problem, we use cannot add capacity and disinvest at the same time. Also, a facility
a uniform discrete-time representation with a pre-fixed number existing at time (t − 1) must either exist or be sold off at time t (Eq.
(H) of identical periods. Please note that the first period begins at (5)).
time 1 rather than 0 and all decisions are made at the beginning of
EEYe(t−1) ≤ EEYet + EXet  e ∈ PI, t  > TDe , t > min[TDe , CLTe ]
each period. We view the MNC’s multi-echelon supply chain as a
set of entities, which has several subsets. For instance, it has a set (5)
of material suppliers, a set of materials, a set of production facili-
ties, a set of inventories at production sites, and a set of inventories Clearly, a facility cannot exist and be sold off at the same time.
at distribution centers. These sets of assets include both current
ESYet + EEYet ≤ 1 e ∈ PI, t > TDe (6)
and future possible entities. An arc represents the transfer of mate-
rial from one facility (production, inventory, distribution center, Once a facility is disinvested, it cannot be purchased back and
etc.) to another. An arc may or may not exist between some loca- should remain sold off for the remainder of the horizon. Eqs. (6)–(8)
tions/entities. ensure this.
We view each facility and material as an entity or an asset. Enti-
ESYet ≥ EXet e ∈ PI, t > TDe (7)
ties are processed by some entities, which transform them into new
entities, e.g. raw materials to products. Each entity can be invested ESYe(t+1) ≥ ESYet e ∈ PI, H > t > TDe (8)
in or disinvested, e.g. procuring raw material, investment in a facil-
ity, disinvesting a facility, selling a product, etc. Hence, a single set Once a facility undergoes a technology upgrade
or subscript defines all the entities. However, different variables are (TUet = 1 ⇒ UFet = 1), it is taken as an upgraded facility
considered for each attribute of an entity. This representation helps (UFe(t−1) = 1 ⇒ UFet = 1). Clearly, if a facility is not undergoing
us address asset management, capital budgeting, and supply chain an upgrade at the beginning of the horizon, then it cannot be
redesign in a unified formulation. It integrates the various impor- upgraded before its construction lead time for upgrade (TUPe ).
tant elements of the supply chain network and helps in writing the However, if an upgrade is in progress at the beginning of the
formulation efficiently. horizon, then the variables must be initialized appropriately.
We develop the formulation in terms of constraints related to Recall that UP is the set of facilities that can undergo a technology
capacity, inventory, contracts, loans, bonds, costs, and incomes. upgrade.

UFet = UFe(t−1) + TUet e ∈ UP, t > TUPe (9)


3.1. Capacity change constraints
UFet = 0 e ∈ UP, t ≤ TUPe (F2)
Due to the time (CLTe ) required for construction, a facility cannot To track the capacity of a facility at time t, we write
expand (EYet = 1) more than once in any interval of length (CLTe ).
eaet ≤ ecapet − ecape(t−1) + ecapU
e EXet  e ∈ PI,

t
EYet ≤ 1 e ∈ PI, t > CLTe (1) t > min[CLTe , TDe ], t  > TDe (10)
t−CLTe +1
eaet ≥ ecapet − ecape(t−1) e ∈ PI, t > min[CLTe , TDe ] (11)
If a facility e is not under construction at the beginning of the
horizon, then no capacity addition is possible for t ≤ CLTe . However, The last term in Eq. (10) is to accommodate the disinvestment of
if it is under construction at time the beginning of the horizon, then a facility at t, which will make its capacity zero (ecapet = 0), and to
EYet can be fixed to zero accordingly. ensure that capacity expansion remains zero instead of becoming
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3159

negative. Note that disinvestment is not a negative capacity addi- (16b) accounts for technology upgrade. Subscript e3 refers to facil-
tion in our formulation. Investment and disinvestment are modeled ities that can be upgraded. If a facility undergoes a technology
separately and ‘ea’ is a positive variable. upgrade, then UFe3t = 1, else UFe3t = 0. IIN is the set of inventory
Since no capacity addition is possible during the initial period facilities associated with a production facility. PI is the set of facil-
of the horizon which is equal to the construction lead time of the ities.
facility, we write If a facility is in operation (EZe3t = 1), then the amount that it
processes must exceed the lower limit on allowable throughput
eaet = 0 e ∈ PI, t ≤ CLTe (F3)
(TFLet ecapet ). The first terms in the following equations ensure this.
The capacity increase for a facility must be zero, when no capac- However, the throughput of a facility may change after a technology
ity is added at time t, and it must also satisfy the lower limits for upgrade. The last terms in these equations account for that. P is the
addition and construction. set of production facilities.

eaet ≤ ecapU
e EYet e ∈ PI, t > CLTe (12) ede1e2e3t ≥ TFLe3t ecape3t − ecapU
e3 (1 − EZe3t )
e1e2

U U
eaet ≥ EAN − EEYe (t−1) ) + EAO + EYet − 1) e ∈ PI, −ecapU
e3 (TFe3 t  + TFUe3 t  )UFe3 t
  e3e1 ∈ FM, e1e2e3 ∈ MFF,
e (EYe t e (EEYe(t−1)
 
e ∈ PIN, t > CLTe (13) e3 ∈ P, e2e3 ∈ IIN, e3 ∈ UP, t > TUPe3 (17a)


The capacity of a facility must be zero, when the facility does not ede1e2e3t ≥ TFULe3t ecape3t
exist, and must be the same as that at the beginning of the horizon e1e2
for the initial interval, during which no investment or disinvest-
U U
ment is possible. −ecapU U
e3 (1 − EZe3t ) − ecape3 (TFe3t + TFUe3t )(1 − UFe3t )

ecapet ≤ ecapU e ∈ PI, t > min[CLTe , TDe ] e3e1 ∈ FM, e1e2e3 ∈ MFF, e2e3 ∈ IIN, e3 ∈ UP, t > TUPe3
e EEYet (14)

ecapet = ecape,1 e ∈ PI, t ≤ min[CLTe , TDe ] (F4) (17b)

3.2. Capacity constraints Eqs. (17a) and (17b) are meant for production facilities only, but
 one could write similar equations for inventory facilities also. Eqs.
The amount of a material supplied ( e1 ede1eet ) by a raw mate- (16b) and (17b) are written only for the facilities that could undergo
rial supplier to various input inventory facilities at production a technology upgrade.
facilities, must equal the amount of material purchased. Now, if a facility is not in operation, then its throughput must be
  zero.
ede1e2e3t = dse1e2t + dcre1e2t e3e1 ∈ FM, 
e3 r<r c ede1e2e3t ≤ (TFU U U
e3t + TFUe3t )ecape3t EZe3t e3e1 ∈ FM,
e1e2
e2 ∈ S, e1e2e3 ∈ MFF (15)
e1e2e3 ∈ MFF, e3 ∈ PI, e2e3 ∈ IIN (18)

In the above, subscript ‘e1’ refers to the raw material, ‘e2’ refers
to the supplier, and ‘e3’ refers to the receiving facility. For each con- A facility cannot operate (EZet = 0), if it does not exist (EEYet = 0).
tract, we have ranges for purchase amounts with different prices. ‘r’
EEYet ≥ EZet e∈P (19)
is the index that denotes the range in which the purchase amount
lies, and ‘rc ’ is the cardinality of ‘r’. MFF is the set of arcs that ‘e1’ A facility should shutdown (EZe(t−1) = 0) for some time to inte-
can take between ‘e2’ and ‘e3’, FM is the set of materials that e2 grate the new added capacity, before the new capacity becomes
can sell, and S is the set of raw material suppliers. Furthermore, operational (EYet = 1). We assume that this time is an integer mul-
only one term on the right hand side of Eq. (15) is non-zero at tiple of period length. Thus, if this time is equal to period length,
any given time depending on whether a contract is signed for then we can write,
that material or not. More on contracts is explained in Section
EZe(t−1) ≤ 1 − EYet e ∈ P, t > CLTe (20a)
3.4. 
The amount ( e1 ede1eet ) of materials that a facility can If it is longer than one period, then the above equation can be
receive/process cannot exceed facility throughput. suitably changed. Similarly, a facility should shutdown for integra-
 tion, when there is a technology upgrade. Assuming that time to be
ede1e2e3t ≤ TFU U U U
e3t ecape3t + ecape3 (TFe3 t  + TFUe3 t  )UFe3 t  equal to one period, we can write,
e1e2
EZe(t−1) ≤ 1 − TUet e ∈ UP, t > TUPe (20b)
e3 ∈ PI, e3 ∈ UP, e3e1 ∈ FM, t  > TUPe3 , e1e2e3 ∈ MFF (16a)

Most facilities cannot operate non-stop indefinitely, as they


 must shutdown for routine maintenance or overhaul. For exam-
ede1e2e3t ≤ TFUU
e3t ecape3t + ecapU U
e3 (TFe3t + TFUU
e3t )(1 − UFe3t )
ple, a facility may shutdown once every 18–24 months. If STle is
e1e2
the maximum number of periods that a facility can operate in an
e1e2e3 ∈ MFF, e3e1 ∈ MF, e2e3 ∈ IIN, e3 ∈ UP, t > TUPe3 p
interval of length PTe periods, then,
(16b) 
t

EZet ≤ STle
p
e ∈ P, t ≥ PTe (21)
where, throughput factor (TFU et ) is defined as the ratio of through- p
t−PTe +1
put to capacity (ecapet ). The last term in each of Eqs. (16a) and
3160 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

An inventory facility may store more than one material, e.g. before undergoing a technology upgrade.
when materials are containerized in drums, boxes, etc. and share  
e2 ∈ IIN
ede1 e2e3t e2 ∈ OIN
ede1e3e2t
an inventory facility. The amount ( e1 eee1e2t ) of materials that the + ecapU
e3 UFe3t ≥
facility holds at time t cannot exceed its holding capacity (ecapet ). e1 e3 e1e3

 
e2 ∈ IIN
ede1 e2e3t
eee1e2t ≤ ecape2t e2e1 ∈ FM, e2 ∈ I (22) ≥ − ecapU
e3 UFe3t e3 ∈ P, e2e1 ∈ FM,
e1 e3
e1
e1 ∈ UM, e2e1 ∈ FM, (25d)
Note that Eq. (22) is written only for inventory and not produc-
tion facilities, because the latter does not store any material per 
say. ede1e3e2t ≤ ecapU
e3 (1 − UFe3t ) e3 ∈ P, e3e1 ∈ FM,
e2 ∈ OIN
3.3. Mass balance constraints e1 ∈ UM (25e)

A mass balance on an inventory facility includes the holdup at  


e2 ∈ IIN
ede1e2e3t e2 ∈ IINR
ede1 e3e2t
time (t − 1), material received at time t, material shipped to other = e3 ∈ P, e3e1 ∈ FM,
facilities at time t, and sales (at both actual and reduced sales price) e1e3 e1 e3

at time t. e1 ∈ RCM, e3e1 ∈ FM (26)


 
eee1e3t = eee1e3(t−1) + ede1e2e3t − ede1e3e2t − se1e3t − we1e3t
e2 e2 Here, e2 refers to set of input or output inventory facilities that
are linked to a given production facility e3. e1 refers to the set of
× e1e3e2 ∈ MFF, e1e2e3 ∈ MFF, e3 ∈ I, e3e1 ∈ FM
materials that a given inventory facility can hold. e1e3 is the sto-
(24) ichiometry coefficient of material e1 that is processed in a given
production facility e3. OIN is the set of all output inventory facili-
ties that are linked to a given production facility. IINR is the set of
input inventory facilities that take recycled material from a given
Eq. (24) includes an allowance for selling excess material
production facility.
at a lower price than normal price or disposing it at some
negative price (or positive cost). We achieve this through vari-
3.4. Contracts
able (weet ). It avoids the situation in which a material occupies
storage unnecessarily. Initialization for holdups can be done
In our work, a material contract with a supplier specifies sev-
appropriately.
eral prices based on purchase amounts. Each price is associated
For a production facility, materials received from asso-
with a range of purchase amounts. The (rc − 1) ranges in which the
ciated input inventory facilities, processed, and sent to
material can be purchased are predefined and the total amount
output inventory facilities are in stoichiometry ratios.
of purchase must fall into one of the ranges. We use binary vari-
We write equations to relate input and output materi-
able Vre1e2t to identify purchase within a range r and (r + 1), and the
als Eqs. (25) and to relate input and recycle materials (Eq.
amount dcre1e2t of purchase in that range must fall within the limits
(26)).
of that range.
Eq. (25a) is written for all the materials other than those that
may undergo a technology upgrade. Vre1e2t QCre1e2t ≤ dcre1e2t ≤ Vre1e2t QC(r+1)e1e2t r < rc,
 
e2 ∈ IIN
ede1e2e3t e2 ∈ OIN
ede1 e3e2t e1 ∈ RM, e2 ∈ S (27)
= e3 ∈ P,
e1e3 e1 e3

e2e1 ∈ FM, e3e1 ∈ FM, e1 ∈ (M − UM − UMM) (25a) If a contract does not exist for a material with a supplier, then
the material must be purchased from outside the contract from that
supplier. The MNC may wish to limit purchases without contracts,
so the amount (dse1e2t ) of such purchases must not exceed some
Eqs. (25b) and (25c) are written for the materials that may
pre-specified limits.
undergo a technology upgrade and it corresponds to materials after  
undergoing a technology upgrade. 
 dse1e2t ≤ QSe1e2t 1− Vre1e2t e1 ∈ RM, e2 ∈ S (28)
e2 ∈ IIN
ede1 e2e3t
+ ecapU
e3 (1 − UFe3t )
r<r c
e1 e3
  A contract
 for a materialwith a supplier becomes effectual,
e2 ∈ OIN
ede1e3e2t e2 ∈ IIN
ede1 e2e3t when V
r<r c re1e2t
= 1 and V
r<r c re1e2(t−1)
= 0. Therefore,
≥ ≥ − ecapU
e3 (1 − UFe3t )
e1e3 e1 e3

CSe1e2t = Vre1e2t t = 1, e1 ∈ RM, e2 ∈ S (29a)
e3 ∈ P, e1 ∈ UMM, e2e1 ∈ FM, e2e1 ∈ FM (25b)
r<r c

 
 CSe1e2t ≥ Vre1e2t − Vre1e2(t−1) t > 1, e1 ∈ RM, e2 ∈ S,
ede1e3e2t ≤ ecapU
e3 UFe3t e3 ∈ P, e1 ∈ UMM, e2e1 ∈ FM
r<r c r<r c
e2 ∈ OIN (29b)
(25c)

Eqs. (25d) and (25e) are written for the materials that may CSe1e2t ≤ Vre1e2t t > 1, e1 ∈ RM, e2 ∈ S, (29c)
undergo a technology upgrade and it corresponds to materials r<r c
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3161


CSe1e2t ≤ 1 − Vre1e2(t−1) t > 1, e1 ∈ RM, e2 ∈ S, (29d) For a technologically upgraded facility, the production costs may
be different from that of the non-upgraded facility (UFet = 1). The
r<r c
difference in this production cost is given in Eqs. (35) and this can
In addition, there can be limits on contract durations. Once a be either positive or negative.
contract with a supplier begins, it must remain in effect for some  
minimum duration (unless it has been signed towards the end of  
TU TU
the horizon) and any material purchased from that supplier must cprtue3t ≤ ce3t ecape3t + ede1e2e3t te3t
follow the contract terms. e1 e2

CSe1e2t ≤ Vre1e2t  e1 ∈ RM, e2 ∈ S, t + CLLe2 > t  > t (30) +ecapU TU
e3 (ce3t
TU
+ te3t )(1 − UFe3t ) e3e1 ∈ FM, e3 ∈ UP,
r<r c e2 ∈ (MFF ∧ IIN), t > TUPe3 (35b)

CSe1e2t + Vre1e2t  ≤ CLU L
e2 − CLe2 + 1 e1 ∈ RM, e2 ∈ S,  
t  r<r c  
TU TU
cprtue3t ≥ ce3t ecape3t + ede1e2e3t te3t

t < H − CLU U L
e2 + 1, t + CLe2 ≥ t ≥ t + CLe2 (31) e1 e2

−ecapU TU
e3 (ce3t
TU
+ te3t )(1 − UFe3t ) e3e1 ∈ FM, e3 ∈ UP,
Note that contracts that are already in effect at the beginning
of the horizon can be enforced by initializing the variable CSe1e2t e2 ∈ (MFF ∧ IIN), t > TUPe3 (35c)
appropriately.
If the facility does not undergo a technology upgrade (UFet = 0),
3.5. Loans and bonds then the additional cost involved is zero.
TU TU
The MNC must pay back the principal and interest on each bond cprtue3t ≤ (ce3t + te3t )ecapU
e3 UFe3t e3e1 ∈ FM, e3 ∈ UP,
issued for raising capital at the time of maturity and no interest is
e2 ∈ (MFF ∧ IIN), t > TUP e3 (35d)
paid before maturity.

cbbt = BPb(t−BLb ) nbb(t−BLb ) (1 + RBbt )BLb t > BLb (32)


TU TU
cprtue3t ≥ −(ce3t + te3t )ecapU
e3 UFe3t e3e1 ∈ FM, e3 ∈ UP,
where, nbbt is the number of bonds b issued at time t, BPbt is a
bond’s principal value, and RBbt is the interest rate. We allow an e2 ∈ (MFF ∧ IIN), t > TUPe3 (35e)
upper limit on the capital that can be raised through bonds.
In contrast to a bond, a loan is repaid in installments over the
For transportation, we assumed that the costs are borne by the
loan period and no lump sum is paid at the end. We allow the loan
receiving inventory facilities.
amount (lalt ) to be variable with time and between some limits.  
With that, the total payment on loans that are active at time t is  
 ctse3t = ede1e2e3t ˛e1e2e3t e3 ∈ I, e3e1 ∈ FM,
ialt = lalt  RLlt  t (t − LLl ) < t  ≤ t (33) e1 e2
t
e2e3 ∈ UPF (36)
As with bonds, there may be an upper limit on the total amount
of loans. Installments for a loan that is taken during the end of the
horizon may not account completely for the loan. Hence, a sum The costs for an inventory facility include those for holding and
that is equivalent to the unpaid loan is paid as installment at the transferring the material.
end of the horizon. To account for this, the parameter RLlt H is given

cste2t = ˇce2t ecape2t + eee1e2t ˇme1e2t e2e1 ∈ FM,
accordingly. However, no bonds can be taken towards the end of the
e1
horizon unless the length of the bond is small enough to be repaid
before the end of the horizon. e3 ∈ I, e2 ∈ (MFF ∧ IIN) (37)

3.6. Manufacturing costs and incomes


The income from sale of material at its true price (Eq. (38)) and
at a discounted price (Eq. (39)), are given below.
The purchase of raw material in any period incurs two costs.

One is fixed and the other is variable with the amount of purchase. csae2t = e1e2t se1e2t e2e1 ∈ FM, e2 ∈ D (38a)
However, both vary with time and contract in which a purchase is
e1
made.
  The discounted price for the excess of the material produced is
crme1e2t = KCe1e2t Vre1e2t + dcre1e2t ıcre1e2t + dse1e2t ıse1e2t written in Eq. (39).
r<r c r<r c 
cwe2t = WDCe1e2t we1e2t e2e1 ∈ FM, e2 ∈ (IIN ∪ OIN) (39)
e1 ∈ RM, e2 ∈ S (34)
e1

The sales for a product cannot exceed its demand, which is dif-
The production cost for a facility is
ferent for upgraded and non-upgraded products.
 
  se1e2t ≤ De1e2t e2e1 ∈ FM, e2 ∈ D (40)
cpre3t = ce3t ecape3t + ede1e2e3t te3t e3e1 ∈ FM,
e1 e2 Startups and shutdowns of a facility involve costs.
e3 ∈ P, e2 ∈ (MF ∧ IIN) (35a) csuet ≥ suet (EZet − EZe(t−1) ) e ∈ P, t ≥ 2 (41a)
3162 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

csdet ≥ sdet (EZe(t−1) − EZet ) e ∈ P, t ≥ 2 (41b) depreciation, and other costs described above.

cswet = csuet + csdet e ∈ P, t ≥ 2 (41c)   
ctxnt ≥ TXnt se1e2t e1e2t + cwet
The cost of depreciation depends on the rate of depreciation and e2 ∈ Ine1 ∈ FM e2 ∈ In
asset value. The rate of depreciation for an existing facility may be    
different from that of a new facility or capacity addition. However, − crme1e2t − cstet −
if there are several simultaneous additions, the same rate applies. e2 ∈ Sne1 ∈ FM e ∈ In e ∈ Pn,e ∈ UPn

cdet = ecape(1) DPO O n n
et DRet + (ecapet − ecape(1) )DPet DRet e ∈ PI   
× (cpret + cprtue t ) − cswet − ctret − cdet
(42) e ∈ PIn e ∈ In e ∈ PIn

(49)
Just as the minimum capacity addition at a new location is dif-
ferent from that at existing locations, the fixed costs for these two
The net present value of partial profit, not including the value of
investments are also different.
assets, loans, and bonds at time t in a given nation is given by the
ciN N following equation.
et ≥ ˝et (EYet − EEYe(t−1) ) e ∈ PIN, t > CLTe (43a)
  
npnt = −ctxnt + se1e2t e1e2t + cwet
ciO O
et ≥ ˝et (EEYe(t−1) + EYet − 1) e ∈ PI, t > CLTe (43b)
e2 ∈ Ine1 ∈ FM e ∈ In
However, the variable costs are the same.   
− crme1e2t − cstet
TU
ciet = ˝et N O
 TUet  + cie t  + ciet  + ˝et  eaet  e ∈ PI, e ∈ PIN, e2 ∈ Sne1 ∈ FM e ∈ In
  
t > min[CLTe , TUPe ], t  > CLTe , t  > TUPe (43c) − (cpret + cprtue t ) − cswet − ctret
e ∈ PIn,e ∈ UPn e ∈ Fn e ∈ In

Eq. (43c) also involves a fixed cost associated with technology


 
+ rdeet − ciet (50)
upgrade. This fixed cost may arise from acquiring new technology,
e ∈ PIn e ∈ PIn
investing in new equipment, etc.
The revenue from disinvesting a facility is proportional to the
The net present value of the shareholder value includes the time
value of facility per unit capacity and the disinvestment factor. Dis-
value of money, values of assets at the beginning and end of the
investment factor is the ratio of revenue from disinvestment and
horizon, and transactions due to loans and bonds.
the value of the unsold facility.
 npnt
shNPV =
rdeet ≤ ecape(t−1) VEUe(t−1) RFe(t−1) e ∈ PI, t > TDe (44) (1 + IRnt )(YT(t)−1)
n t

rdeet ≤ ecapU
e VEUe(t−1) RFe(t−1) EXet e ∈ PI, t > TDe (45)   ecap VEUeH
eH
+ − ecape,1 VEUe,1
The revenue from disinvestment is obtained at the time when e ∈ e PIn
(1 + IRnH )(YT(H)−1)
production stops for the first time. The total revenue at time t is   BPbt nbbt
 +
rdeet  ≤ RDU
t
L
e ∈ PI, t > TD , t > TDe 
(46) (1 + IRnt )(YT (t)−1)
b ∈ nBt(t≤H−BLb )
e
  cbbt

The total investment at time t is given by the following and is
(1 + irnt )(yr(t)−1)
incurred at the time when the new capacity becomes available for b ∈ nBt(t>BLb )

production. The total amount of capital that can be raised is limited.  lalt  ialt
The capital for investments is raised through loans and bonds. + −
(YT (t)−1)
(1 + IRnt ) (1 + IRnt )(YT(t)−1)
   l ∈ nL t l ∈ nL t
ciet ≤ BPbt nbbt  + lalt min[CLTe , TUPe ] < t  ≤ H−BLb , (51)
e b l
where ecapet VEUet is the value of an unsold facility and it is pro-
t > min[CLTe , TUPe ] (47) portional to its capacity and unit value of the capacity. It should be
noted that IRnt is the cumulative interest rate in a nation. It can be
computed from the actual interest rate in each period.
The total amount of capital that can be invested in a given inter- This completes our complex and comprehensive model for asset
val is limited (Eq. (48)). management. Eqs. (38), (39), (41c), (42), (43c) and (50) can be
directly incorporated in Eq. (51) in the implementation of the

t
 model. In the above presentation, these equations are written sep-
ciet  ≤ TCIU
t t ≥ (IvI + min[CLTL ,
arately so as to improve the readability.
t−IvI+1 e
The model is clearly large with many variables and constraints
TUPL ]), t  > min[CLTe , TUPe ] (48) and is not easy to solve. Recall that we treated all assets (materials,
facilities, etc.) in one uniform fashion. This is in contrast to treat-
ing them or modeling them separately as different types of assets.
The tax that MNC must pay is computed based on the coun- While it does not lend itself to easy exposition, our unified formula-
try that host the facilities and this can be computed fromrevenue, tion has one advantage. Although the total numbers of constraints
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3163

Table 1
Initial capacities (ecape,1 , ton/quarter), lower limits (EAN O U
e and EAe , ton/quarter) on capacity additions, maximum capacity ecape (ton/quarter) for all, material balance data
for production and inventory facilities, production (cet and tet , $/ton) and storage costs (ˇce1t and ˇmee1t , $/ton)

Facility (nation) ecape,1 EAN


e EAO
e ecapU
e Material (inventory) or mass cet or ˇcelt tet or ˇmeelt
balance for production facility ($/ton) ($/ton)

23(1) before 6000 – 2000 20,000 m3 + 0.5m4 = 0.7m9 + 0.8m10 2000 + 50t 1500 + 50t
upgrade
23(1) after upgrade 6000 – 2000 20,000 m3 + 0.5m4 = 0.7m9 + 0.8m11 2000 + 50t 500 + 50t
24(1) 6000 – 2000 20,000 m5 + 0.4m6 = 0.5m8 + 0.8m12 + 0.1m5 5000 + 100t 5000 + 200t
25(2) 6000 – 2000 20,000 m7 + 0.7m8 = 0.9m13 + 0.8m14 2000 + 50t 1500 + 50t
26(2) 0 5000 2000 20,000 m5 + 0.4m6 = 0.5m8 + 0.8m12 + 0.1m5 2000 + 50t 1500 + 50t
15–18(1), 27–30(1), 6000 – 1000 20,000 3(15); 4(16); 5(17, 21); 6 (18, 22); 200 + 10e1 + 10t 200 + 10e1 + 10t
19–22(2), 31–34(2) 7(19); 8(20, 29, 33); 9(27); for 15–22 and for 15–22 and
10–11(28); 12(30, 34); 13(32); 27–34 27–34
14(32)
35–36(1), 37–39(2) 0 3000 1000 20,000 9–14 in 35–39 200 + 15t for 200 + 15t for
35–37; 35–37;
250 + 10t for 250 + 10t for
38–39 38–39

and variables would be the same for both approaches, our unified of these uncertainties. However, analyzing various scenarios would
formulation requires fewer equations or lines of code. A normal be of interest to the senior management. In this respect, a solu-
representation would require repetition of the same equations, for tion obtained in reasonable computation time and one that gives
example, one equation for transportation between suppliers and considerable profit would suffice. Thus, it is our belief that the inte-
input inventory facilities at production facilities and another for grality gap for this complex strategic asset management problem is
transportation between output inventory facilities at production not as important as it would be for a tactical planning/scheduling
facility and distribution center, and so on. problem. Consequently, we allow 1 day of computing time for each
A more important issue concerns the quality and speed of solu- run in this work and present our results.
tion for such a complex model. Slagmulder et al. (1995) found in
their survey that the average hurdle rate or the minimum RORI 4. Example
(rate of return on investment) for project decisions is about 22.3%
with 62.5% of respondents specifying a value greater than 15%. This We consider an MNC with four production facilities (three exist-
hurdle rate is used as a measure against any risks that an invest- ing and one potential) in two nations with different tax rates
ment decision may carry. In this respect, Smith and Winkler (2006) and import–export restrictions, five raw materials (including one
presented what they called the “optimizer’s curse”. They conclude recycle), one intermediate, six products (including the upgraded
that an optimization analyst should in general be prepared for dis- product), two suppliers who supply all the raw materials and inter-
appointment; the management would take a skeptic’s view of the mediates, eight input inventory facilities (six existing and two
analysis, because no analysis is perfect. Since a strategic asset man- potential), eight output inventory facilities (six existing and two
agement model involves several thousand parameters and nearly potential), and five inventory facilities at distribution centers (three
all of them are uncertain, a single optimal solution is not always existing and two potential). We consider a planning horizon of 40
preferable. Even if we get an optimal solution from our model, it quarters or 10 years. Using our unified representation, we assign
may not be the scenario that is actually realized in future because the following indices to these entities, namely raw material suppli-

Table 2
Raw material availability limits (QCre1e2t or QSe1e2t , ton/quarter), demands (De1e2t , ton/quarter), transport costs (˛e1e2e3t , $/ton), products sales price (e1e2t , $/ton), products
discounted sales price (WDCe1e2t , $/ton) and raw material purchase price (ıcre1e2t or ıse1e2t , $/ton)

e1 t(e2) QCre1e2t or QSe1e2t or De1e2t ˛e1e2e3t e1e2t or WDCe1e2t or


ıcre1e2t or ıse1e2t

3,5,7 [1,34](1) r = 1: 0; r = 2: 2500; r = 3: 75,000 300 + 10t − 10e3 + 20e2 r = 1: 1000 + 100t for
t ≤ 15, else 2500 + 50t;
r = 2: 900 + 100t for
t ≤ 15, else 2400 + 50t;
no contract: 500 + 150t
3,5,7 [35,36](1) r = 1: 0; r = 2: 1000; r = 3: 3000 300 + 10t − 10e3 + 20e2 Same as above
3,5,7 [37,40](1) r = 1–3: 0 300 + 10t − 10e3 + 20e2 Same as above
3,5,7 [1,5](2); [6,30](2); [31,40](2) 50,000; 1000; 50,000 300 + 10t − 10e3 + 20e2 500 + 150t
4,6 [1,34](1,2); [35,36](1,2); [37,40](1,2) 50,000; 1000; 50,000 290 + 11t − 10e3 + 20e2 1500 + 100t
8 [1,34](1); [35,36](1); [37,40](1) 50,000; 1000; 50,000 250 + 15t − 10e3 + 20e2 11,000 + 100t
8 [1,34](2); [35,36](2); [37,40](2); 50,000; 1000; 50,000 250 + 15t − 10e3 + 20e2 10,000 + 100t
9 [1,H] 3(200 + (700t + 20)/(t + 30)) 255 + 16t − 10e3 + 20e2 e1e2t :
700e2 + 450t − 4001,
WDCe1e2t :
200e2 + 100t − 50e1
10 [1,H] 3(250 − (100t + 20)/(t + 25)) 265 + 17t − 10e3 + 20e2 Same as above
11 [1,H] 3(200 + (300t + 20)/(t + 15)) 275 + 18t − 10e3 + 20e2 Same as above
12 [1,H] 3(200 + 10t) 285 + 19t − 10e3 + 20e2 Same as above
13 [1,20] 3(200 + (500t + 50)/(100 − 3t)) 295 + 20t − 10e3 + 20e2 Same as above
13 [21,H] 3(325 + (600t + 50)/(50 + t)) 295 + 20t − 10e3 + 20e2 Same as above
14 [1,20] 3(200 + (350t + 100)/(25 + t)) 295 + 20t − 10e3 + 20e2 Same as above
14 [21,H] 3(250 + (400t + 100)/(100 − t)) 295 + 20t − 10e3 + 20e2 Same as above
3164 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

ers (1 and 2), raw materials (3–8), products (9–14), input inventory Table 4
Model statistics and results
facilities (15–22), production facilities (23–26), output inventory
facilities (27–34), and inventory facilities at distribution centers Full model Base case
(35–39). Entities 24 (an existing production facility) and ‘26’ (poten- Continuous variables 18,134 15,418
tial production facility) manufacture the same products, hence this Binary variables 2,220 1,320
example could illustrate investment of ‘26’ or relocation of ‘24’ to Constraints 26,390 20,915
‘26’ (Fig. 1). Among the products, the material ‘10’ and ‘11’ represent Non-zeros 103,302 87,008
CPU time (h) 24.00 1.75
the same product before and after technology upgrade.
Relative gap (%) 10.84 2
We chose this case study to illustrate the relocation of one pro- NPV ($ billion) 6.75 6.08
duction facility producing the intermediate from one nation to
another nation where the intermediate is required and also the
redesign of supply chain arising from the shutdowns of inventory (2) The maximum throughput (TFU et = 1) for each inventory facility
capacities in some locations and investments at other locations. We equals its capacity, that for each distribution center equals twice
have chosen the parameters with a view to highlight all features of its capacity (TFU et = 2), and that for a production facility equals
our formulation. Also, some parameters are considered as constant its capacity.
over the entire planning horizon, although the model does allow (3) The minimum throughout (TFLet = 0) is zero for an inventory
period-dependent values. This is done in order to minimize the data facility (hence no lower limits on the holdups in inventory facil-
and be able to present them completely in this communication. ities), and that for a production facility is 60% of its capacity
Table 1 gives the initial capacities, capacity expansion and (TFLet = 0.6). A technology upgrade results in a 10% decrease in
construction limits, upper limit on capacity, mass balance for pro- throughput limits (TFUU U L L
et = 0.9 TFet and TFUet = 0.9 TFet ).
duction facilities, material storage for inventory facilities, fixed and (4) The most number of times that a facility may expand is three
variable production costs, fixed and variable costs for storage. It (NEe = 3).
should be noted that while input/output inventory facilities can (5) The construction lead time for all production facilities is 2 years
hold only certain materials, the distribution centers can store all. (CLTe = 8 periods) and that for all inventory facilities is 1 year or
Table 2 shows the raw material availability and its purchase costs 4 periods. The time before which a facility cannot be upgraded
both within and outside a contract, demand for products, its sell- is 1 year (TUPe = 4 periods). The minimum period during which
ing price and discounted sales price, and transportation costs for no disinvestment is allowed is 1 year (TDet = 4 periods).
all materials. The raw material availability and purchase costs are (6) The minimum and maximum lengths of a contract are 20 and
given according to the quantity flexibility. The excess material is 30 periods (CLLet = 20 and CLU et = 30).
sold at a discounted price and it can be sold at all the inventory (7) The revenue from the disinvestment of a facility does not exceed
facilities. Logically, all the excess material would be sold at these its current value. In this example, it is only half the current value
inventory facilities and will not be shipped to the distribution facil- (RFet = 0.5).
ities, since the distribution involves additional costs. Hence, the (8) suet = 100,00,00 − 500e + 1000t; sdet = 100,00,00 − 500e +
selling of over-produced material at the distribution centers does 1000t; RDPoet = 0.08 + 0.0005t; RDPnet = 0.06 + 0.0003t;
not occur. In this example, we did not write these equations pertain- DPet = 100 + 10t; DPet = 100 + 10t; RLlt t = 0.25 + (t − t )/LLl for
o n

ing to sale of waste at the distribution centers in order to reduce t < H; and RLlt t = 1.15(t − LLl )/(H − LLl ) for t = H.
the number of variables. In addition, we have limited the sale of (9) The principal value of a bond b is $1 (BPbt = 1). Only single types
products to the distribution centers. Transportation costs involve of loans and bonds are possible. Furthermore, the loan exists
insurance and tariffs also. It should be noted that transfers between only in nation 2 and the bonds in nation 1. Inflation is based
input/output inventory facilities and their associated production on the nations in which the loan and bonds are taken. However,
facility are not considered as transportation and hence they do not the capital raised in one nation can be used to fund investments
incur any transportation costs. The material handling costs for this in other nations.
case are included in the production costs.
Table 3 gives the cost of investing in a facility. While the fixed 4.1. Results and discussion
costs for construction and capacity additions are different, the vari-
able cost is the same for both and proportional to the capacity We consider two design plans for supply chains. In the first plan,
increase. A technology upgrade also carries a fixed cost. The value of disinvestments are allowed, and we call this as the full model. In
an unsold facility is also given. Table 4 gives the quantity flexibility the second plan, we present a base case model, where no disin-
for material contracts. Each contract offers two prices with corre- vestments are allowed. These two plans will enable us to compare
sponding ranges. There is an upper limit on the amount of material and highlight the importance of having a model that allows dis-
purchased without a contract. investments and supply chain redesign versus the one that allows
Other parameters used in the model are listed below: expansions only. The two models were solved using Cplex 9.0 in
GAMS 21.7 on a Windows XP based HP workstation with Intel Xeon
3.6 GHz dual processor. The model statistics and results are given in
(1) The inflation rates are 9 and 5%, and tax rates are 5 and 8%, for
Table 4. The full model was terminated after 24 h of CPU time, and
nations 1 and 2, respectively.
the base model after 2% relative gap. The full model gave a shNPV of
$6.75 billion versus $6.08 billion for the base model. In other words,
a comprehensive supply chain redesign plan gives a 10% increase
Table 3
in shNPV as compared to a mere capacity expansion plan. This is a
Investment costs. ˝etO
fixed costs for expansion ($), ˝et
N
fixed costs for construction
($), ˝et variable costs ($/ton), ˝et
TU
fixed technology upgrade costs ($) net result of several differences between the two plans, which are
highlighted next.
Set ˝et ˝et
O
˝et
N
VEUet Fig. 2a and b shows the capacity profiles of production facili-
I 2000 − 25e + 50t 10,00,00 + 100t 15,00,00 + 100t 100 ties for the two models. Fig. 2a (the full model) shows facility ‘24’
P 5000 + 100t 20,00,00 + 200t 25,00,00 + 200t 200 being disinvested and facility ‘26’ being invested at quarters 8 and
UP 15,00,00,00 − 10,00,00t – – –
9, respectively. Since ‘24’ and ‘26’ produce the same products, this is
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3165

Fig. 1. Schematic of a chemical supply chain of a multinational corporation. Entities 1–2 are suppliers, 15–22 are input inventory facilities, 23–26 are production facilities,
27–34 are output inventory facilities and 35–39 are distribution centers.

essentially a relocation of ‘24’ to ‘26’. In addition, facilities ‘23’, ‘25’, twice during the planning horizon. However, as we see later, the
and ‘26’ expand twice. In other words, the optimal solution involves capacity additions are lower for ‘26’ than those in the full model,
relocation (disinvestment + investment) plus capacity expansion. because of the continued presence of ‘24’.
Since the base model (Fig. 2b) allows no disinvestment, it keeps As direct consequences of the above changes to the production
‘24’ unchanged, but it does invest in a new ‘26’ producing the same facilities, the plans also show changes in the capacity profiles of
products as ‘24’. As in the full model, it expands ‘23’, ‘25’, and ‘26’ inventory facilities. Fig. 3a and b shows these profiles for input

Fig. 2. (a) Full model: Capacities of input inventory facilities. Inventories ‘17’ and Fig. 3. (a) Capacity profiles of production facilities for the full model. Facility ‘24’ is
‘18’ are disinvested at quarters 8 and 7, respectively, while the others expand. ‘21’ disinvested at quarter 8 and relocated to ‘26’ at quarter 9. Other facilities expand.
and ‘22’ are new facilities added at quarters 9 and 7, respectively. (b) Base case: (b) Capacity profiles of production facilities for the base model. Facility ‘24’ remains
Capacities of input inventory facilities. Inventories ‘17’ and ‘18’ remain unchanged. unchanged, but all other facilities expand.
3166 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

Fig. 4. (a) Full model: Capacities of output inventory facilities. Inventories ‘29’ and Fig. 5. (a) Full model: Holding capacities at distribution centers. (b) Base case:
‘30’ are disinvested at quarter 8, while the some of the others expand. ‘33’ and ‘34’ Holding capacities at distribution centers.
are added new at quarter 9. (b) Base case: Capacities of output inventory facilities.
Inventories ‘29’ and ‘30’ are not disinvested and do not expand.

inventories and Fig. 4a and b shows the same for output inventories.
For the full model (Figs. 3a and 4a), ‘17’, ‘18’, ‘29’, and ‘30’ disappear
along with but not at the same time as ‘24’ due to the presence
of surplus materials. Similarly, ‘21’, ‘22’, ‘33’, and ‘34’ appear along
with ‘26’. Inventory facilities ‘16’, ‘20’, ‘27’, and ‘28’ do not expand,
but others expand at different times in order to cope with the
increasing product demands. For the base model (Figs. 3b and 4b),
new facilities ‘21’, ‘22’, ‘33’, and ‘34’ appear as expected with ‘26’.
The disinvested facilities of the full model, namely ‘17’, ‘18’, ‘29’, and
‘30’ do not expand as expected, and ‘16’, ‘20’, ‘27’, and ‘28’ remain
unchanged, while the remaining facilities expand. Just as the capac-
ity addition at ‘26’ is lower in the base model when compared to
the full model, which is due to the continued existence of ‘24’, the
capacity additions at ‘21’, ‘22’, ‘33’ and ‘34’ are lower due to the
continued existence of ‘17’, ‘18’, ‘29’, and ‘30’. In both the models,
the associated input inventory facilities at the production facility
‘26’ appear earlier than the facility ‘26’. This is because of the lower
investment cost at an earlier time. Even though the input inven-
tory facilities appear early, they do not stock raw material until ‘26’
comes into existence (results not shown). This is however different
from the output inventory facilities, which are not disinvested due
to the availability of surplus products.
It can be seen from Fig. 5 that distribution centers ‘38’ and ‘39’
expand for both models, because the demand is always increasing
at all distribution centers. Interestingly, while production facility
‘24’ is relocated to ‘26’, the distribution centers do not relocate,
because the relocation of production is driven by changes in costs Fig. 6. (a) Full model: Capacity and utilization (UT) of production facility ‘23’. Plots 9,
rather than redistribution of demands. 10, and 11 show the desired capacities of ‘23’ to meet 100% demands (not considering
the materials in inventories) of products ‘9’, ‘10’ and ‘11’, respectively. Plot 23UT
Fig. 6 shows the production capacity profiles for facility ‘23’ for
shows the capacity utilization for ‘23’. (b) Base case: Capacity and utilization (UT)
both models. It can be observed that the facility expands in both of production facility ‘23’. Plots 9, 10, and 11 show the desired capacities of ‘23’ to
plans, but the capacity profiles are different. Though there is no meet 100% demands (not considering the materials in inventories) of products ‘9’,
demand for the technologically upgraded product during the first ‘10’ and ‘11’, respectively. Plot 23UT shows the capacity utilization for ‘23’.
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3167

Fig. 8. (a) Full model: Capacity and utilization (UT) of production facility ‘25’. Plots
13 and 14 are the desired capacities of production facility ‘25’ required for satisfy-
Fig. 7. (a) Full model: Capacity and utilization (UT) of production facilities ‘24’ and ing the 100% (not considering inventory) of demand for the products ‘13’ and ‘14’,
‘26’ (‘24 + 26’) combined. Plot 12 shows the desired capacity of production facility ‘24’ respectively. (b) Base case: Capacity and utilization (UT) of production facility ‘25’.
and ‘26’ (‘24 + 26’) combined that is required for satisfying the 100% (not consider- Plots 13 and 14 are the desired capacities of production facility ‘25’ required for sat-
ing inventory) of demand for the product ‘12’. Plots 8–13 and 8–14 show the desired isfying the 100% (not considering inventory) of demand for the products ‘13’ and
capacities of production facility ‘24’ and ‘26’ (‘24 + 26’) combined that is required for ‘14’, respectively.
satisfying the 100% demand for intermediate ‘8’ from downstream production facil-
ity ‘25’, if ‘25’ has to satisfy the 100% (not considering inventory) demand of product
‘13’ and ‘14’, respectively. Plots 24UT and 26UT show the capacity utilizations for
‘24’ and ‘26’, respectively. (b) Base case: Capacity and utilization (UT) of produc- combined capacity of about 15,000 ton/quarter in both the models.
tion facilities ‘24’ and ‘26’ (‘24 + 26’) combined. Plot 12 shows the desired capacity In both cases, ‘26’ expands in order to meet the demand for primary
of production facility ‘24’ and ‘26’ (‘24 + 26’) combined that is required for satisfy- material ‘12’. However, the demand for intermediate ‘8’ required by
ing the 100% (not considering inventory) of demand for the product ‘12’. Plots 8–13 ‘25’ is not met by ‘24 + 26’, so ‘25’ is forced to purchase ‘8’ from the
and 8–14 show the desired capacities of production facility ‘24’ and ‘26’ (‘24 + 26’)
combined that is required for satisfying the 100% demand for intermediate ‘8’ from
market. Fig. 8 shows that production facility ‘25’ expands twice dur-
downstream production facility ‘25’, if ‘25’ has to satisfy the 100% (not considering ing the planning horizon for both models, and its utilizations are
inventory) demand of product ‘13’ and ‘14’, respectively. Plots 24UT and 26UT show almost 100% during the entire planning horizon, except for mainte-
the capacity utilizations for ‘24’ and ‘26’, respectively. nance shutdowns. Facility ‘25’ expands at quarter 9 and 30 in both
the full model and base model. It can be clearly seen that both the
half of the horizon, the facility undergoes a technology upgrade expansion at ‘25’ and ‘26’ in both the models almost coincide. This is
at quarter 5, since an early investment would be cheaper than a because the downstream facility ‘25’ is partially dependent on ‘25’
late investment. In addition, the upgraded material ‘11’ is sold at a for the intermediate ‘8’ and only the shortfall is purchased from the
discounted price up to quarter 20, since there is no demand. This market.
result is counter-intuitive, because intuition may suggest that the Fig. 9a and b shows the demand profiles and fractions of
technology upgrade should be done around the time when there demands met for all products. The top part of each plot gives the
is demand for the new product. This result shows the effects of demands for products and the bottom part gives the fractional
introducing a product early into the market. It should be noted that demand that is met. Note that these figures are for the total demand
the technology upgrade results in a throughput reduction of 10% at all facilities. It can be seen from Fig. 9a that the demand for prod-
and ‘23’ expands its capacity early in order to meet the demand for uct ‘10’ is not met at all after quarter 8. This is because the facility
product ‘9’. undergoes a technology upgrade to produce material ‘11’ and this
Fig. 7 shows the production capacity profiles for ‘24’ and is sold at a discounted price. However, once there is demand for ‘11’
‘26’. In addition to the initial capacity at ‘26’ at quarter 9 of from quarter ‘21’, only the excess beyond the demand is sold at a
8000 ton/quarter, there is a capacity expansion at quarter 31, and discounted price.
almost a complete utilization of the facility to manufacture prod- It was observed that the contracts for materials ‘3’, ‘5’, and ‘7’
uct ‘12’ and intermediate ‘8’. In contrast to this, the base plan begin at quarters 7, 6, and 7 respectively for the full model, but all
(Fig. 7b) shows a capacity installation of 5000 ton/quarter at quar- at 7th quarter for the base model. The contracts last up to quarters
ter 9, because the minimum addition allowed is 5000 ton/quarter 34, 34, and 36 respectively for the base model, but up to 36th quarter
and a subsequent expansion at quarter 30. This means that base for the full model. There is no great difference in both the models
plan capacity of ‘24 + 26’ exceeds 11,000 ton/quarter, which is much with respect to the contracts. This is because the net purchase of
higher than 8000 ton/quarter for the full plan after the first addi- raw materials in both the models is the same even though the base
tion. However, the second addition results in almost a similar model has an excess capacity in ‘24 + 26’ during the initial part of the
3168 P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169

is the first to include useful features such as disinvestments, tech-


nology upgrades, supply contracts, capital-raising loans and bonds,
etc. in a supply chain redesign problem. A detailed numerical case
study shows that a supply chain redesign plan can produce 14% less
profit, if disinvestments are ignored.
While our work demonstrates the advantage of a comprehensive
supply chain redesign model, it is just the first step. One limitation
of the current model is its computational performance. Considering
that industry-scale redesign problems would be large, tremendous
opportunity exists for developing efficient decomposition-based
and/or alternative solution strategies to improve relative gap. Fur-
thermore, several important issues need attention, especially those
related to the uncertainties in business data such as those in the
demand and supply of raw materials, and those in the operational
issues of supply chain. Work is underway on these and other issues,
which we hope to report in future.

Acknowledgements

We would like to acknowledge the financial support for


this work under the project ICES/06-432001. We would like to
thank Arief Adhitya and Vivek Kumar Kumra of ICES, and Arul
Sundaramoorthy (currently a graduate student at University of Wis-
consin, Madison) for their valuable suggestions.

References

Akalu, M. M. (2003). The process of investment appraisal: The experience of 10 large


British and Dutch companies. International Journal of Project Management, 21,
355–362.
Fig. 9. (a) Full model: Plots 9, 10, and 11 show the demand profiles for products Amiri, A. (2006). Designing a distribution network in a supply chain system: Formu-
lation and efficient solution procedure. European Journal of Operational Research,
‘9’, ‘10’ and ‘11’ respectively. Plots 9-S, 10-S, and 11-S show the demand fulfillment
171, 567–576.
fractions (DS) for products ‘9’, ‘10’ and ‘11’ respectively. (b) Full model: Demand
Annupindi, R., & Bassok, Y. (1999). In S. Tayur, R. Ganeshan, & M. Magazine (Eds.),
profile for products ‘12’, ‘13’ and ‘14’ (top part of the graph) and the fraction of
Quantitative models for supply chain management. Massachusetts, USA: Kluwer
demand that is met (bottom part of the graph). Plots ‘12-S’, ‘13-S’, ‘14-S’ are the Academic Publishers.
fraction of demand (DS) that is met for ‘12’, ‘13’ and ‘14’, respectively. Arntzen, B. C., Brown, G. G., Harrison, T. P., & Trafton, L. L. (1995). Global supply chain
management at digital equipment corporation. Interfaces, 25(1), 69–93.
Barbaro, A., & Bagajewicz, M. J. (2004). Use of inventory and options contracts
horizon. The full-scale model shows loans of $23m, $2.8m, $5.8m, to hedge financial risk in planning under uncertainty. AIChE Journal, 50(5),
$61m, $2.5m, $31m, $25m, $74m, $14.5m and $6m taken at quarters 990–998.
Bjorkqvist, J., & Roslof, J. (2005). Strategic investment planning in the pulp and paper
5, 6, 7, 9, 23, 27, 30, 31, 32, and 33, respectively, while the base model industry using mixed integer linear programming. In AIChE Annual Meeting,
shows only two loans of $20.9m and $26.7m at quarters 5 and 9, October 30–November 04.
respectively. The full-scale model shows bonds of 50m, 1.8m and Bhatnagar, R., Jayaram, J., & Phua, Y. C. (2003). Relative importance of plant location
factors: A cross national comparison between Singapore and Malaysia. Journal
16.7m at quarters 9, 10 and 16, respectively. In contrast, base model
of Business Logistics, 24(1), 147–170.
shows bonds of 50m and 5.3m at quarters 9 and 13, respectively. Bhutta, K. S., Huq, F., Frazier, G., & Mohamed, Z. (2003). An integrated location,
It can be seen clearly that the timings of these loans and bonds in production, distribution and investment model for a multinational corporation.
International Journal of Production Economics, 86, 201–216.
both cases coincide with the timings of the capacity expansions.
Bok, J.-K., Lee, H., & Park, S. (1998). Robust investment model for log-range capacity
The solution presented above can be refined by fixing some of expansion of chemical processing networks under uncertain demand forecast
the decision variables and resolving the MILP. We fixed some deci- scenarios. Computers & Chemical Engineering, 22(7–8), 1037–1049.
sion variables (namely CS, EEY, ESY, EX, TU and UF) to the values Bradley, J., & Arntzen, B. C. (1999). The simultaneous planning of production, capac-
ity, and inventory in seasonal demand environments. Operations Research, 47(6),
obtained in the above example and resolved the MILP to a gap of 795–806.
2% in 1 h. This refinement gave an shNPV of $6.92 billion, which is Cohen, M. A., & Lee, H. L. (1989). Resource deployment analysis of global manu-
a 2.5% improvement over the unrefined solution of $6.75 billion. facturing and distribution networks. Journal of Manufacturing and Operations
Management, 2, 81–104.
Thus, by using this solution refining procedure, the full model gave Dixit, A., & Pindyck, R. S. (1995). The options approach to capital investment. Harvard
about a 14% higher shNPV than the base model. Furthermore, we Business Review, May–June, 105–115.
tested the model extensively with various parameters and found Fan, J. P. H. (2000). Price uncertainty and vertical integration: An examination of
petrochemical firms. Journal of Corporate Finance, 6, 345–376.
little change in computational performance, so the model is valid Gatica, G., Papageorgiou, L. G., & Shah, N. (2003). Capacity planning under uncer-
and robust with respect to changes in various parameters. tainty for the pharmaceutical industry. Transactions of the Institution of Chemical
Engineers, 81, 665–678.
Gjerdrum, J., Shah, N., & Papageorgiou, L. G. (2001). Transfer prices for multienter-
5. Conclusion prise supply chain optimization. Industrial & Engineering Chemistry Research, 40,
1650–1660.
Goetschalckx, M., Vidal, C. J., & Dogan, K. (2002). Modeling and design of global
We have presented a novel and comprehensive MILP model for
logistics systems: A review of integrated strategic and tactical models and design
supply chain redesign, asset management, and capital budgeting. algorithms. European Journal of Operational research, 143, 1–18.
The model treats diverse supply chain entities such as facilities, Grossmann, I. E., & Westerberg, A. W. (2000). Research challenges in process systems
contracts, loans, bonds, etc. as assets, and allows investments, engineering. AIChE Journal, 46, 1700–1703.
Grossmann, I. E. (2004). Challenges in the new millennium: Product discovery and
disinvestments, regulatory factors, transportation costs, inflation, design, enterprise and supply chain optimization, global life cycle assessment.
shutdown for maintenance, etc. To the best of our knowledge, it Computers & Chemical Engineering, 29, 29–39.
P.K. Naraharisetti et al. / Computers and Chemical Engineering 32 (2008) 3153–3169 3169

Guillen, G., Mele, F. D., Bagajewicz, M. J., Espuna, A., & Puigjaner, L. (2005). Multiob- Peel, M. J., & Bridge, J. (1998). How planning and capital budgeting improve SME
jective supply chain design under uncertainty. Chemical Engineering Science, 60, performance. Long Range Planning, 31(6), 848–856.
1535–1553. Pirttila, T., & Sandstrom, J. (1995). Manufacturing strategy and capital budgeting
Holland, C. P., Shaw, D. R., & Kawalek, P. (2005). BP’s multi-enterprise asset manage- process. International Journal of Production Economics, 41, 335–341.
ment system. Information and Software Technology, 47, 999–1007. Pyke, D. F., & Cohen, M. A. (1994). Multiproduct integrated production–distribution
Julka, N., Srinivasan, R., & Karimi, I. A. (2002). Agent-based supply chain manage- systems. European Journal of Operational Research, 74, 18–49.
ment. 1. Framework. Computers & Chemical Engineering, 26, 1755–1769. Rajagopalan, S., & Swaminathan, J. M. (2001). A coordinated production planning
Julka, N., Karimi, I. A., & Srinivasan, R. (2002). Agent-based supply chain man- model with capacity expansion and inventory management. Management Sci-
agement. 2. A refinery application. Computers & Chemical Engineering, 26, ence, 47(11), 1562–1580.
1771–1781. Romero, J., Badell, M., Bagajewicz, M., & Puigjaner, L. (2003). Integrating budgeting
Karimi, I. A., Srinivasan, R., & Por, L. H. (2002). Unlocking supply chain improvements models into scheduling and planning models for the chemical batch industry.
through effective logistics. Chemical Engineering Progress, 98, 32. Industrial & Engineering Chemistry Research, 42, 6125–6134.
Lanzenauer, C. H., von Eschen, E., & Pilz-Glombik, K. (2002). Capacity planning in a Ryu, J.-H., & Pistikopoulos, E. N. (2005). Design and operation of an enterprise-wide
transitional setting with simulation based modeling: A case study. International process network using operation policies. 1. Design. Industrial & Engineering
Transactions in Operational Research, 9, 125–139. Chemistry Research, 44, 2174–2182.
Lasschuit, W., & Thijssen, N. (2004). Supporting supply chain planning and schedul- Sharma, M., Ammons, J. C., & Hartman, J. C. (2007). Asset management with
ing decisions in the oil and chemical industry. Computers & Chemical Engineering, reverse product flows and environmental considerations. Computers & Opera-
28, 863–870. tions Research, 34(2), 464–486.
Lee, H.-K., Lee, I.-B., & Reklaitis, G. V. (2000). Capacity expansion problem of multisite Shapiro, J. F. (2001). Modeling the supply chain. California, USA: Duxbury, Thomson
batch plants with production and distribution. Computers & Chemical Engineer- Learning.
ing, 24, 1597–1602. Smith, J. E., & Winkler, R. L. (2006). The optimizer’s curse: Skepticism and postdeci-
Levis, A. A., & Papageorgiou, L. G. (2004). A hierarchical solution approach for sion surprise in decision analysis. Management Science, 52(3), 311–322.
multi-site capacity planning under uncertainty in the pharmaceutical industry. Stadtler, H. (2005). Supply chain management and advanced planning—basics,
Computers & Chemical Engineering, 28, 707–725. overview and challenges. European Journal of Operational Research, 163,
McDonald, C. M., & Karimi, I. A. (1997). Planning and scheduling of parallel semicon- 575–588.
tinuous processes. 1. Production planning. Industrial & Engineering Chemistry Slagmulder, R., Bruggeman, W., & van Wassenhove, L. (1995). An empirical study
Research, 36, 2691–2700. of capital budgeting practices for strategic investments in CIM technologies.
McDonald, C. M., & Reklaitis, G. V. (2004). Design and operation of supply chains International Journal of production Economics, 40, 121–152.
in support of business and financial strategies. Foundations of Computer-Aided Subramanian, D., Pekny, J. F., & Reklaitis, G. V. (2000). A simulation–optimization
Process Design, FOCAPD, July 11–16, Princeton, NJ, USA. framework for addressing combinatorial and stochastic aspects of an
Naraharisetti, P. K., Karimi, I. A., & Srinivasan, R. (2005). Location allocation– R&D pipeline management problem. Computers & Chemical Engineering, 24,
production–distribution in the chemical process industries. In INFORMS Annual 1005–1011.
Meeting, November 12–16. Subramanian, D., Pekny, J. F., & Reklaitis, G. V. (2001). A simulation–
Naraharisetti, P. K., Karimi, I. A., & Srinivasan, R. (2006). Capacity management in optimization framework for research and development pipeline management.
the chemical supply chain. In International Symposium on Advanced Control of AIChE Journal, 47, 2226–2242.
Chemical Processes-ADCHEM, April 2–5. Subramanian, D., Pekny, J. F., Reklaitis, G. V., & Blau, G. (2003).
Oh, H.-C., & Karimi, I. A. (2004). Regulatory factors and capacity-expansion planning Simulation–optimization framework for stochastic optimization of R&D
in global chemical supply chains. Industrial & Engineering Chemistry Research, 43, pipeline management. AIChE Journal, 49, 96–112.
3364–3380. Sykuta, M. E. (1996). Futures trading and supply contracting in the oil refining indus-
Oh, H. C., & Karimi, I. A. (2006). Global multi product production–distribution plan- try. Journal of Corporate Finance, 2, 317–334.
ning with duty drawbacks. AIChE Journal, 52(2), 595–610. Timpe, C. H., & Kallrath, J. (2000). Optimal planning in large multi-site production
Olhager, J., Rudberg, M., & Wikner, J. (2001). Long-term capacity management: Link- networks. European Journal of Operational Research, 126, 422–435.
ing the perspectives from manufacturing strategy and sales and operations Tsay, A. A., Nahmias, S., & Agrawal, N. (1999). In S. Tayur, R. Ganeshan, & M. Magazine
planning. International Journal of Production Economics, 69, 215–225. (Eds.), Quantitative models for supply chain management. Massachusetts, USA:
Papageorgiou, L. G., Rotstein, G. E., & Shah, N. (2001). Strategic supply chain opti- Kluwer Academic Publishers.
mization for the pharmaceutical industries. Industrial & Engineering Chemistry Tsiakis, P., Shah, N., & Pantelides, C. C. (2001). Design of multi-echelon supply
Research, 40, 275–286. chain networks under demand uncertainty. Industrial & Engineering Chemistry
Perea-Lopez, E., Ydstie, B. E., & Grossmann, I. E. (2003). A model predictive control Research, 40, 3585–3604.
strategy for supply chain optimization. Computers & Chemical Engineering, 27, Wilkinson, S. J., Cortier, A., Shah, N., & Pantelides, C. C. (1996). Integrated produc-
1201–1218. tion and distribution scheduling on a Europe-wide basis. Computers & Chemical
Partovi, F. (2006). An analytic model for locating facilities strategically. Omega, 34(1), Engineering, 20, S1275–S1280.
41–55.

You might also like