Journal of Food Engineering 70 (2005) 383–391

www.elsevier.com/locate/jfoodeng

Design of a supply chain network for pea-based novel protein foods

a,* b
Radhika K. Apaiah , Eligius M.T. Hendrix
a
Product Design and Quality Management Group, Department of Agrotechnology and Food Science, Wageningen University,
P.O. Box 8129, 6700 EV Wageningen, The Netherlands
b
Operations Research and Logistics, Department of Social Sciences, Wageningen University, P.O. Box 9101, 6700HB Wageningen, The Netherlands

Received 30 September 2003; received in revised form 27 January 2004; accepted 13 February 2004
Available online 25 November 2004

Abstract

This paper presents an operations research technique that can be used for supply chain design and applies it to create a supply
network with a goal to manufacture a pea-based NPF as cheaply as possible. The current food production and consumption pattern
has a strong impact on the environment and resources and is not sustainable. Meat production in particular is not appealing from an
environmental point of view, because of the inefficient conversion of protein in the feed into protein in the slaughtered animal. Novel
protein foods (NPFs) are non-meat protein ingredients that are designed to replace meat-based ingredients in meals. The non-meat
protein products presently available do not meet the expectations of most consumers and cannot be considered as realistic alterna-
tives to meat (http://www.profetas.nl). They are niche products and are expensive when compared to pork. The prospects for
replacing meat-derived ingredients by NPFs are more promising. The partial shift from an animal based diet to a plant, specifically
pea-based diet may be feasible only if the price of these products decreases.
A supply chain for NPFs can be divided into three major links: primary production (growing and harvesting), ingredient prep-
aration (milling and concentration of pea protein) and product processing (manufacture of the NPF). The pea-based product is
designed for the Dutch market. The peas are sourced from several locations around the world such as Canada, Ukraine, France
and the Netherlands and are transported by sea, rail, road or barge. This paper presents a study on the optimisation of the supply
network for NPFs in the Netherlands using linear programming. It focuses on finding the lowest cost at which NPFs can be man-
ufactured for a specific market demand; while deciding the location of primary production, ingredient processing and product pro-
duction areas and modes of transportation by minimising the sum of production and transportation costs.

2004 Elsevier Ltd. All rights reserved.

Keywords: Novel protein foods; Supply chains; Optimisation

1. Introduction the technological developments and open markets have

changed the production, trade and distribution (i.e. the
Consumers today demand high-quality products in supply chain) of food products beyond recognition (Tri-
various innovative forms through the entire year at com- enekens & Omta, 2001).
petitive prices. Society imposes constraints on producers A supply chain (SC) is an integrated process where
in order to economise the use of resources, ensure ani- raw materials are acquired, converted into products
mal friendly and safe production practices and restrict and then delivered to the consumer (Beamon, 1998).
environmental damage. These demands, together with The chain is characterised by a forward flow of goods
and a backward flow of information. Food supply
chains are made up of organisations that are involved
*
Corresponding author. Tel.: +31 317 482286; fax: +31 317 483669. in the production and distribution of plant and ani-
E-mail address: [email protected] (R.K. Apaiah). mal-based products (Zuurbier, Trienekens, & Ziggers,

0260-8774/$ - see front matter
2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfoodeng.2004.02.043
384 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391

1996). Such SCs can be divided into two main types (van duction and consumption pattern has a strong impact
der Vorst, 2000): on the environment and resources and is not sustain-
able. Meat production in particular is not appealing
• SCs for fresh agricultural products: the intrinsic char- from an environmental point of view, because of the
acteristics of the product remain unchanged, and inefficient conversion of protein in the feed into protein
• SCs for processed food products: agricultural prod- in the slaughtered animal. Novel protein foods (NPFs)
ucts are used as raw materials to make processed are non-meat protein ingredients that are designed to re-
products with a higher added value. place meat- based ingredients in meals. The NPFs pres-
ently available do not meet the expectations of most
The main fact that differentiates food SCs from other consumers and thus cannot be considered realistic alter-
chains is that there is a continuous change in quality natives to meat. They are niche products and are expen-
from the time the raw materials leave the grower to sive when compared to pork. The prospects for
the time the product reaches the consumer (Tijskens, replacing meat-derived ingredients by NPFs are more
Koster, & Jonker, 2001). A food SC as defined in this promising. The partial shift from an animal based diet
paper consists of six links: primary production, ingredi- to a plant, specifically pea-based diet may be feasible
ent preparation, product processing, distribution, retail only if the price of these products decreases. This paper
and the consumer (Fig. 1). considers an OR approach that can be applied to ex-
Performance measures or goals are used to design plore possible chain designs. The interesting question
SCs or supply networks by determining the values of here is whether an NPF based on pea protein is feasible
the decision variables that yield the desired goals or per- as a price-competitive product, when all essential cost
formance levels (Apaiah, Hendrix, Meerdink, & Linne- sources are identified.
mann, 2004; Beamon, 1998). The design of the chain
or network changes with the goal for which the chain
is being designed and optimised. As consumer demand 2. Case
has to be met, it is important to ask the consumer what
attributes he/she desires in the product as these attri- NPFs based on pea proteins do not currently exist.
butes are used to select the goals to design the chain, The proposed product is designed to resemble the vege-
e.g. if the goal is quality at any cost, then technologically tarian mincemeat currently available. As mentioned ear-
advanced and consequently expensive equipment can be lier a supply chain consists of two basic processes: (1)
used to produce the product and it can be transported to Production planning; (2) Distribution and logistics plan-
the consumer by air. However, if the goal is a low priced ning (Beamon, 1998). In this study, this is modified as:
product, care has to be taken to minimise production (1) Production—this includes all the links from primary
and transportation costs. production to product processing and (2) Distribution—
PROFETAS is a research programme dedicated to the remaining links. This paper focuses on the first pro-
making food production and consumption more sus- cess. The second process of distribution and logistics
tainable (http://www.profetas.nl). The current food pro- planning, similar to that of chilled meat products, has

Product
Primary Ingredient
Processing Distributor
producer 1 preparation- Retailer 1
- 1
Plant 1
Plant 1

C
O
Product N
Primary Ingredient
Processing Distributor S
producer 2 preparation- Retailer 2
- 2 U
Plant 2
Plant 2 M
E
R

Product
Primary Ingredient
Processing Distributor
producer n preparation- Retailer n
- m
Plant k
Plant l

transport

Fig. 1. Food supply chain.

 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391 385

Fig. 2. Production scheme for NPFs (1000 kg basis).

been much researched (Chopra, 2003; Jayaraman & facture of the NPF). The product is designed for the
Ross, 2003). Designing this part of the chain will not Dutch market. Fig. 2 shows the production scheme for
lead to a distinction between the costs of NPF and of 1000 kg of the pea-based NPF. The peas are sourced
the chilled meat Production planning for pea-based from several locations around the world such as Can-
NPFs, is a new and unknown area. The aim is to gener- ada, Ukraine, France and the Netherlands and can be
ate scenarios that lead to low costs by designing supply transported by sea, rail, road or barge.
chains with the aid of OR techniques. Fig. 3 illustrates the steps in each link of production.
The supply chain for the first process is divided into In primary production (PP) plant refuse is the main by-
three major links: primary production (growing and product. In ingredient preparation (ING), the hulls and
harvesting), ingredient preparation (milling and concen- starch are by-products. Starch comprises about 70%
tration of pea protein) and product processing (manu- of the dehulled peas and therefore the selling price of

PP1 CANADA ING 1 NPF-Extrusion

Dehulling Milling + Air-
Growing Sorting Drying classification Extrusion Packing

PP2 FRANCE NPF-Extrusion

ING 2
Dehulling Milling +
Growing Sorting Drying Extrusion Packing
Air-classification

PP3 NETHERLANDS ING 3 NPF-Extrusion

Dehulling Milling +
Growing Sorting Drying Air-classification Extrusion Packing

PP4 UKRAINE ING 4 NPF-Extrusion

Dehulling Milling + Air-
Growing Sorting Drying classification Extrusion Packing

Refuse Water Hulls Oil

Starch Water
Others
transport

Fig. 3. NPF production chain.

386 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391

the starch (starch is used as a raw material in many ent processing- milling and air classification, Primary
applications) is important in the overall cost of production.
manufacture. Performance indicator: Value added at each link.
Control variables: In this supply chain design problem
the following decision variables are considered. Fig. 4
3. Model development shows the supply network with the associated variables.

The following approach was developed in Apaiah PPi amount of pea produced at primary production
et al. (2004). For a given product, goals and production location i
possibilities are identified and the relationships between TPIijn amount of dry pea transported from location i
the performance indicators of the goals and the control to facility j via transport mode n
variables are determined. The important relationships INGj amount of ingredient, pea protein concentrate
are identified. A quantitative model is then developed produced at facility j
and optimisation approaches and sensitivity analysis TIPjkn amount of protein concentrate transported
are used to design the chain quantitatively. from facility j to facility k via transport mode n
The paper deals with a long-term exploratory ques- NPFk amount of NPF produced at facility k
tion of the feasibility of pea-based NPFs and therefore SAj amount of starch produced at facility j
considers possible flows and quantities of products,
by-products, refuse and production schemes. This devi- Data to evaluate a specific supply chain design in-
ates from the usual set-up and locations decisions in cludes the following technical and cost coefficients:
similar logistics modelling that is used to support deci-
sions for particular companies (Jayaraman & Ross, wpci whole dried pea cost at location i (€/ton)
2003; Wouda, Beek, van der Vorst, & Heiko, 2002). tcdpijn transportation cost of dried pea from PP location
These typically lead to MILP type of models, whereas i to ING facility j via transport mode n (€/ton)
this paper considers a network flow, a linear program- ipcj pea protein cost at facility j (€/ton)
ming approach from a long-term perspective. tcppjkn transportation cost of protein concentrate from
ING facility j to NPF production facility k, via
transport mode n (€/ton)
3.1. The qualitative model ppck cost of producing the NPF at location k (€/ton)
ssj selling price of starch from facility j (€/ton)
According to the methodology presented in Apaiah stpt starch per ton of dehulled pea = 0.7 ton/ton
et al. (2004), the relevant aspects to model the underly- npfp pea protein per ton NPF = 0.376 ton/ton
ing supply chain are identified. ppdp pea protein per ton of dry transported
pea = 0.255 ton/ton
3.1.1. Product: a pea-based NPF resembling vegetarian pwp percentage of dry pea from total pea pro-
mincemeat duced = 0.805 ton/ton
Attribute (as specified by the consumer): An inexpen- demand total amount of NPF put into the mar-
sive product. This product is designed to replace pork ket = 30 744 ton
meat. The retail price of pork is about €6/kg. The cost
of manufacturing is about 38–40% of retail cost where
(http://www.ers.usda.com).
Goal: Minimise cost of manufacturing. i index for PP location (i = 1, 2, 3, . . . , I)
Chain: An important boundary condition for these j index for ING production facility (j = 1, 2,
chains is that consumer demands should be met. The 3, . . . , J)
chains are therefore traced backwards, i.e. described k index for NPF production facility (k = 1, 2,
from what the consumer wants back through to primary 3, . . . , K)
production. The links are: Consumer processing, Distri- n modes of transportation (n = sea, rail, road,
bution/retailing, Product processing- Extrusion, Ingredi- barge)

PRODUCT
PRIMARY INGREDIENT PROCESSING
PRODUCTION amount: TPI(I,J) PREPARATION amount: TIP(J,K) FACILITY K
LOCATION I FACILITY J amount: NPF(K)
amount: PP(I) cost: tcdp(I,J) amount: ING(J) cost: tcpp(J,K) cost: ppc(K)
cost: wpc (I) cost: ipc (J) Sum(NPF) = 30744
k

Fig. 4. Supply network with associated variables and cost coefficients.

 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391 387
XX
The optimisation of the supply network for NPFs in TIPjkn 6 INGj for all j
the Netherlands is done using linear programming, sim- k n
ilar to work done by Wouda et al. (2002). It focuses on 4. The amount of pea protein produced at facility j can-
finding the lowest cost at which NPFs can be manufac- not exceed the amount of protein contained in the dry
tured for a specific market demand; while deciding the peas transported from all locations to facility j by all
location of primary production, ingredient processing modes of transport.
and product production areas and modes of transporta- XX
tion by minimising the sum of production and transpor- ppdp
TPIijn P INGj for all j
i n
tation costs (Appendix A).
The market demand: as mentioned earlier, pea-based 5. The amount of dry pea transported from all locations
NPFs are not currently available. The cost of manufac- i to facility j cannot exceed the amount of peas grown
turing a product depends largely on the quantity in in location i minus the harvest losses and refuse
question; the larger this amount, the more will be the ef- (expressed as a percentage of peas grown)
fect of economies of scale. The research program PRO- XX
FETAS aims to replace 20% of processed pork TPIijn 6 pwp
PPi for all i
j n
consumption by the year 2020. An amount equal to
20% of the processed pork consumed in 2000 is used 6. The amount of starch produced as a by-product at
as the market demand for this exercise (30744 MT). each facility j is equal to the percentage of starch in
the dry peas transported from all locations to loca-
3.2. The quantitative model tion j by all modes of transportation.
XX
SAj ¼ stpt
TPIijn for all j
A quantitative linear programming model was for- i n
mulated and implemented using the GAMS software
(http://www.gams.com) to generate various scenarios.
As mentioned earlier, the objective is to minimise the
sum of the production and transportation costs. 4. Data acquisition
This is:
X XXX The model provides a systematic tool to identify the
Minimise : wpci
PPi þ tcdpijn
TPIijn relevant information required to answer the question:
i i j n how cheaply can this product be manufactured. This
X XXX
þ ipcj
INGj þ tcppjkn information, called the cost coefficients (defined in Sec-
j j k n tion 3.1) are: wpci, tcdpijn, ipcj, tcppjkn, ppck and ssj.
X X The data was acquired after an extensive search that

TIPjkn þ ppck
NPFk  ssj
SAj
k j
involved personal and telephonic interviews with experts
and companies in the respective areas, internet searches
and estimations based on the above.
Constraints: The model has some constraints or restric-
In the model it is possible to specify as many areas/
tions for the supply of raw materials and ingredients and
countries as required to source/manufacture the prod-
the demand of the final product.
ucts. For the purpose of this exercise, four countries
were selected to give a diverse range of characteristics.
1. The flow in the whole chain is demand driven. The
They were: Netherlands (area of interest), France
amount of NPFs produced in all the locations should
(a major grower of peas in western Europe), Ukraine
be equal to the demand.
(major grower in eastern Europe) and Canada (large
X
NPFk ¼ demand grower in the Americas). Table 1 shows the primary pro-
k duction data for these countries. The coefficient wpci can
be calculated from here.
2. The amount of pea protein transported from all facil- As can be seen from Table 1, the yield and the areas
ities j to facility k by all modes of transportation is under cultivation differ greatly between countries. How-
equal to the amount of pea protein concentrate in ever the average protein content of the peas is about
the final NPF. 22% and the quality of the peas is similar. It may appear
XX from Table 1 that the best option is to choose the coun-
TIPjkn ¼ npfp
NPFk for all k
j n
try where the cheapest peas are available. However as
costs integrate over the entire supply chain, this may
3. The amount of pea protein transported from each not be the optimal strategy. There could be strategic rea-
facility j to all locations k cannot exceed the amount sons to obtain peas from various sources and to put a
of concentrate produced in location j. limit on the amount of peas from each country.
388 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391

Table 1
Primary production information
Total production (metric tonnes)a Total area (ha)a Yield (MT/ha) Export price (€/ton)b Import price (€/ton)b
Netherlands 4000 800 5.0 473 147
France 1,700,000 334,119 5.088 154 231
Ukraine 746,800 540,007 1.383 129 337
Canada 1,492,600 719,071 2.076 160 401
a
www.statpub.com
b
http://apps.fao.org

The next information of interest is the transportation Pea hulls and starch are by-products that are important
costs (tcdpijn) from the primary production locations to because they have a high resale value: the cost of cereal
the protein production location. The details are sum- starch is €70/ton and the price of pea hulls is €108/ton
marised in Table 2 (Personal communication 1). It was (raw fibre).
not possible to obtain costs for rail transport and in The fourth coefficient of interest is the transportation
some cases for internal transport in certain countries. cost from the ingredient processing facilities to the NPF
As a result these costs were not considered in the model. production locations (tcppjkn). The costs are the same as
However they could be included later and lead to a those in Table 2. However, in reality, there may be some
reduction in total transportation costs. differences if the locations lie further apart than assumed
The next coefficient is the cost to make the pea pro- in the previous case. Moreover, in the model NPF pro-
tein (ipcj) at the different locations. The cost per ton of duction facilities were limited to the Netherlands and
pea protein are summarised in Fig. 5 (web addresses 1). France because of the time it would take to transport
The process of manufacturing pea protein involves the product from Canada or the Ukraine to the Nether-
dehulling, followed by milling and air-classification. lands. The cost of refrigerated transport was found to be
so high that those options were not explored further.
Table 2 The cost of manufacturing the NPF (ppck) in France
Transport cost in €/ton and the Netherlands are estimated as being €19.82 and
Sea Rail Barge Truck €17.84/ton respectively (web addresses 1). The differ-
a ences arise because of the higher energy costs in the
CANADA.CAN
CANADA.FRA 55.02 Netherlands. The selling price of starch (ssj) was calcu-
CANADA.UKA 55.02 lated as a world average of €70/ton.
CANADA.NLD 55.02
FRANCE.CAN 55.02
FRANCE.FRA 5. Scenarios
FRANCE.UKA 50.9 21 20
FRANCE.NLD 13.5 16 The model can be used to develop scenarios. The sce-
UKRAINE.CAN 55.02 narios that arise depend on the constraints that are part
UKRAINE.FRA 50.9 21 20
UKRAINE.UKA
of the model. The exploration was limited to two cases.
UKRAINE.NLD 50.9 21 20 In the first case, no additional capacity constraints were
NETHERLANDS.CAN 55.02 added and the optimisation reduces to a simple shortest
NETHERLANDS.FRA 13.5 16 path problem that identifies the cheapest supply chain in
NETHERLANDS.UKA 50.9 21 20 the network. In the second case, giving an upper limit
NETHERLANDS.NLD 5 10
a
for the primary production sources simulated the strate-
CANADA, FRANCE, UKRAINE, NETHERLANDS are the gic consideration of obtaining peas from several sources.
primary production locations; CAN, FRA, UKA, NLD are the
ingredient preparation locations.
Scenario 1. There are no constraints on the amount of
pea that can be sourced in each primary production
area. This therefore results in a single flow/chain with
Air-classification costs
the model choosing the cheapest route through all the
links. Fig. 6 illustrates this chain. The optimal path is
200.00 then to source the peas in the Ukraine, make the pea
150.00 protein there and then transport it by truck to the
Euro/ton

Energy Costs
Labour costs Netherlands.
100.00
Hardware-costs Scenario 2. Simple upper limit capacity constraints
50.00
were used. The model specifies the amount of pea that
0.00 can be sourced from each location. This is done to en-
Netherlands Canada Ukraine France
sure a supply from all sources so as not to be dependent
Fig. 5. Production costs. on any one country. The flow changes from a single
 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391 389

PRIMARY INGREDIENT NPF Total cost scenarios: Value added at each link
PRODUCTION: PROCESSING: PRODUCTION:
56323 MT 11560 MT 30744 MT
UKRAINE Truck UKRAINE Truck NETHERLANDS
12000000
10000000 ppc
Fig. 6. Scenario 1: Uncapacitated network.
8000000 tcpp

Euros
6000000 ipc
chain to a network (Fig. 7). The final product is made in 4000000
tcdp
the Netherlands and the pea protein is made in the Uk- 2000000
wpc
raine. The model calculated that the optimal path is to
0
source the peas in all the countries, transport them by Scenario 1 Scenario 2
various means to the Ukraine to be converted to pea
Fig. 8. Cost comparison.
protein and then transport the protein concentrate by
truck to the Netherlands. The estimated costs appar-
ently show that this is cheaper than setting up an addi- €1000 euro. Further value is added with the inclusion
tional processing facility in the Netherlands to process of the costs of the second part of the supply chain—
the pea sourced in the Netherlands. packaging, distribution and retail.
Cost comparisons. Fig. 8 illustrates the difference in As mentioned earlier, NPFs are targeted to replace
total costs for the two scenarios. Scenario 1 has lower pork meat in the consumerÕs diet. The retail cost of pork
total costs (for the first part of the supply chain) and meat is about €6/kg and the cost to make pork meat is
therefore a lower product cost. This is because there about 38–40% of this value (http://www.ers.usda.com).
are no constraints on capacities and the model chooses It can be seen from above that the cost of manufacture
the cheapest path of manufacture. Initial estimates of NPFs is much below this figure. This answers the
show: question posed at the beginning of Section 4—how
cheaply can this product be manufactured.
• Scenario 1: cost per ton of product = €216/ton
• Scenario 2: cost per ton of product = €273/ton
6. Conclusions
The cost estimations are limited to that for the main
ingredient, the pea and the concentrate made from the This paper presents a systematic method to develop/
former. The procurement costs for the other ingredients design a supply network for a particular product with
like oil, functional ingredients, flavours are not consid- a specific design goal. The OR linear programming
ered here. However it is known from preliminary calcu- model is a tool that can be used to generate and evaluate
lations, that the inclusion of these costs would still limit different scenarios that are based on differing con-
the production cost per ton of the product to below straints. The model also provides a methodical way to
collect relevant information.

PRIMARY
PRODUCTION
NETHERLANDS
Acknowledgments
truck
4000MT

The authors would like to thank Joris Sprakel and

Bas Toonen for all their help in data acquisition. They
PRIMARY
truck INGREDIENT would also like to thank Dr. A.R. Linnemann and
PRODUCTION PROCESSING: NPF
FRANCE 11560 MT PRODUCTION: Dr. M.A.J.S van Boekel for their comments and
15000 MT
UKRAINE
truck 30744 MT
NETHERLANDS
suggestions.

truck Appendix A. GAMS Model formulation

PRIMARY
PRODUCTION
UKRAINE A.1. Model 1: Uncapacitated
20000MT

Sets
sea

PRIMARY
PRODUCTION
I primary production locations/CANADA,
CANADA FRANCE, UKRAINE, NETHERLANDS/
17292 MT
J Ingredient preparation facilities/CAN, FRA,
Fig. 7. Scenario 2: Capacitated network. UKA, NLD/
390 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391

K NPF production facilities/fran, neth/ demand total amount of NPF put into the market/
N Modes of transportation/sea, rail, barge, truck/; 30744/

PARAMETER Variables

wpc(I) cost of production of dry pea at location I in PP(I) amount of pea produced at primary production
euro per ton/CANADA 160, FRANCE 154, location i
UKRAINE 129, NETHERLANDS 147/ TPI(I, J, N) amount of dehulled pea transported from
ipc(J) cost of making the pea protein ingredient at location i to facility j
location J in euro per ton/CAN 86.7, FRA ING(J) amount of ingredient pea protein concentrate
145.7, UKA 31.69, NLD 161.9/ produced at facility j
ss(J) selling price of starch in euro per ton/CAN 70, TIP(J, K, N) amount of pea protein concentrate trans-
FRA 70, UKA 70, NLD 70/ ported from facility j to facility k
ppc (K) cost of producing the NPF at location K in euro NPF(K) amount of NPF produced at facility k
per ton/fran 19.82, neth 17.84/ SA(J) amount of starch produced at J
Z total costs
Table tcdp(I, J, N) transport cost in euro per ton
Positive variable: PP, TPI, ING, TIP, NPF;
Sea Rail Barge Truck Equations
CANADA.CAN
cost define objective function
CANADA.FRA 55.02
supply(K) observe supply limit at j
CANADA.UKA 55.02
demand satisfy demand at market j
CANADA.NLD 55.02
sup(J) supply from location j
FRANCE.CAN 55.02
supl(I) supply from location I
FRANCE.FRA
sta(J) starch limit
FRANCE.UKA 50.9 21 20
deli(J) delivery to ingredient production
FRANCE.NLD 13.5 16
UKRAINE.CAN 55.02
cost .. z E = sum(I,wpc(I)*PP(I)) + sum((I,J,N),
UKRAINE.FRA 50.9 21 20
tcdp(I,J,N)*TPI(I,J,N)) + sum(J,ipc(J)*ING(J))
UKRAINE.UKA
+ sum((J,K,N), tcpp(J,K,N)*TIP(J,K,N)) +
UKRAINE.NLD 50.9 21 20
sum(K,ppc(K)*NPF(K))  sum(J,ss(J)*SA(J));
NETHERLANDS.CAN 55.02
supply(K) sum((J,N)$(tcpp(J,K,N) gt 0), TIP(J,K,N))
NETHERLANDS.FRA 13.5 16
= e = npfp*NPF(K);
NETHERLANDS.UKA 50.9 21 37.5
demand sum(K, NPF(K)) = e = demand;
NETHERLANDS.NLD 5 10
sup(J) sum((K,N)$(tcpp(J,K,N) gt 0), TIP(J,K,N)) =
l = ING(J);
Table tcpp(J,K,N) transport cost in euro per ton
deli(J) ppdp*sum((I,N)$(tcdp(I,J,N) gt 0), TPI(I,J,N))
= g = ING(J);
Sea Rail Barge Truck
supl (I) sum ((J,N)$(tcdp(I,J,N) gt 0), TPI(I,J,N)) =
CAN.fran 55.02 l = pwp*PP(I);
CAN.neth 55.02 sta(J) SA(J) = e = stpt*sum((I,N)$(tcdp(I,J,N) gt 0),
FRA.fran TPI(I,J,N));
FRA.neth 13.5 16.0
UKA.fran 50.9 21 20 Model model1/all/;
UKA.neth 50.9 21 20 Solve model 1 using lp minimizing z;
NLD.fran 13.5 16 Display PP.l, pp.m,TPI.L, TPI.M, ING.L,
NLD.neth 5.0 10.0 ING.M, TIP.L, TIP.L, NPF.L, NPF.M; Capacitated
model:
Scalar

stpt starch per ton/0.7/ A.2. Model 2: Capacitated

npfp pea protein per ton npf/0.376/
ppdp pea protein per ton of dry transported pea/0.255 The capacitated model is similar to the uncapacitated
pwp percentage of dry pea from total pea produced/ model. The capacity constraints added are for the pri-
0.805/ mary production of peas.
 R.K. Apaiah, E.M.T. Hendrix / Journal of Food Engineering 70 (2005) 383–391 391

The extra parameter is: van der Vorst, J. G. A. J. (2000). Effective food supply chains.
PARAMETER Generating, modeling and evaluating supply chain scenarios. Wagen-
ingen: Wageningen University.
ppu(I) upper limit for primary production in tons/ Wouda, F. H. E., Beek, P., van der Vorst, J. G. A. J., & Heiko, T.
CANADA 20000, FRANCE 15000, UKRAINE (2002). An application of mixed-integer linear programming
20000, NETHERLANDS 4000/; models on the redesign of the supply network of Nutricia
The upperbound for the production is also declared Dairy & Drinks Group in Hungary. OR Spectrum, 24,
as a positive variable. 449–465.
Zuurbier, P. J. P., Trienekens, J. H., & Ziggers, G. W. (1996). Verticale
Positive Variable PP, TPI, ING, TIP, NPF; Samenwerking. Deventer: Kluwer Bedrijfswetenschappen.
PP.UP(I) = ppu(I);

References Personal communication 1

Apaiah, R. K., Hendrix, E. M. T., Meerdink, G., & Linnemann, A. R. Product-agency Pulses, cereals and seeds, Netherlands
(2004). Qualitative methodology for efficient food chain design.
MSC Shipping Rotterdam, Netherlands
Trends in Food Science and Technology, accepted.
Beamon, B. M. (1998). Supply chain design and analysis: Models and Shortsea, Netherlands
methods. International Journal of Production Economics, 55, Vd Bosch, Erp, Netherlands
281–294. Vd Linde Rotterdam, Netherlands
Chopra, S. (2003). Designing the distribution network in a supply Zijdehand Moerkappelle BV, Netherlands
chain. Transportation Research Part E, 39, 123–140.
Jayaraman, V. A., & Ross, A. (2003). A simulated annealing
methodology to distribution network design and management.
European Journal of Operational Research, 144, 629–645.
Web addresses 1
Tijskens, L. M. M., Koster, A. C., & Jonker, J. M. E. (2001). Concepts
of chain management and chain optimization. In Tijskens, L. M. http://www.fao.org
M., Hertog, M. L. A. T. M., & Nicolai, B. M. (Eds.), Food process http://www.alberta-canada.com
modeling. Woodhead. http://www.euireland.ie
Trienekens, J. H., & Omta, S. W. F. (2001). Paradoxes in food chains
and networks. In Proceedings of the Fifth International Conference http://www.cbs.nl
on Chain and Network Management in Agribusiness and the Food http://www.nuon.nl
Industry. The Netherlands: Noordwijk. http://www.edf.fr