Professional Documents
Culture Documents
Adhitya 2012
Adhitya 2012
Adhitya 2012
pubs.acs.org/est
r 2011 American Chemical Society 10178 dx.doi.org/10.1021/es201763q | Environ. Sci. Technol. 2011, 45, 10178–10185
Environmental Science & Technology ARTICLE
been given to operational aspects. Some relevant work in this area Table 1. Diaper Compositions Adapted from EDANA18
includes a simulation approach to evaluate a SC operation by con-
components 1987 (g) 2005 (g)
sidering transport pollution, costs, time-to-market, and energy
usage.14 Another approach to green SC, called reverse logistics, fluff pulp 54.94 14.57
relies on collection of obsolete products or their components superabsorber (SAP) 0.74 13.65
(after use) for recycle.15 This paper describes an approach for polypropylene (PP) 4.36 7.22
holistically evaluating the SC design and operation by using a polyethylene (PE) 4.29 2.69
dynamic simulation and LCA indicators. The benefits of the
adhesive 1.34 1.76
proposed approach are demonstrated using the industrially
elastic 0.20 0.21
relevant case of diaper production SC.
others 1.14 1.85
total 67.01 41.95
2. DIAPER SUPPLY CHAIN
Disposable diapers are a huge business in developed countries. product-centric perspective without considering the effects of
In the U.S, it is estimated that up to $2000 are spent on different logistics options, inventories, distribution network
disposable diapers per baby.16 Diaper manufacturing requires a configuration, and ordering policy, although these can be sig-
highly automated process involving significant capital invest- nificant contributors to the overall environmental impacts. For
ment. Recently, high energy prices have put pressure on operat- example, LCA calculations are typically based on a static, average
ing costs. This, coupled with the focus on sustainability, has led truckload level and number of trips. In reality, these two variables
big players such as Kimberly-Clark to reorganize their SC to be are dynamic and would vary during the actual SC operation based
more efficient.17 on various exogenous factors (e.g., demand) and SC policies. As
The diaper SC involves suppliers, manufacturer, distributor, anecdotal evidence of the significance of these, ExxonMobil’s
and retailers/customers. The manufacturer and distributor be- 2010 Corporate Citizenship Report20 mentions that they “seek
long to a focal enterprise and have different departments to fill truckloads and optimize packaging to reduce the number of
performing various SC functions: receiving order, scheduling, trips while providing associated fuel and cost saving.” LCA alone,
procurement, storage, operations, and delivery. The manufac- therefore, does not provide useful insights to a SC manager
turer procures raw materials from different suppliers, manufac- intending to green a company or product’s SC. The sustainable
tures the diaper products, and sends them in a bulk packaged SC approach, on the other hand, takes a supply chain-centric
form to the distributor. The distributor operates in a push-mode; perspective by considering the scope of influence and the various
it keeps a certain level of inventory to fulfill customer (retailer) decisions that can be made by the SC manager. In this paper, we
orders. Different procurement policies for replenishment of propose a decision support approach for such sustainability
products can be adopted by the distributor. For example, under assessment of SC operations. This is done by evaluating the
the fixed interval policy, procurement is done at regular intervals LCA environmental indicators in each stage of the SC model
to bring its inventory back to a certain top-up level. The using a dynamic simulator. The scope of the SC of interest, as
distributor places orders to the manufacturer, which operates shown in Figure 1, includes all the stages in the diaper manu-
in a pull-mode, that is, the products are manufactured upon facture starting from the procurement of raw materials until
receiving the distributor’s orders. The manufacturer keeps in- delivery of the product to customers (retailers).
ventory of raw materials and has its own procurement policy for
ordering raw materials from suppliers. 3. DYNAMIC MODELING OF THE DIAPER SUPPLY
While modern disposable diapers come in a variety of styles, CHAIN
their basic raw materials are essentially the same consisting of fluff
paper, superabsorbent polymer (SAP), plastic components such Dynamics of SC operation are complex due to the multitiered
as low density polyethylene (LDPE), polypropylene (PP), as structure and numerous interactions among the entities. The
well as adhesives and elastics.18 Several LCA studies of the impact of an entity’s action on the overall SC performance, both
environmental impacts of disposable diapers have been under- economic and environmental, may not be immediately obvious.
taken previously.18,19 Starting from raw material extraction to An integrated analysis of the impacts of decisions on the overall
production process, point of use, and final disposal, these LCAs system is thus required. This motivates the use of SC simulation
account for the consumption of natural resources and release of models, which capture the behavior of the entities and their
pollutants into air, water and soil. The overall conclusion from interactions, and indicate their direct and indirect effects on the
these studies is that with the introduction of SAP in the 1980s overall SC performance. In this paper, we integrate LCA indica-
there has been a significant reduction in the use of raw materials tors into the dynamic simulation modeling scheme of Adhitya and
and energy. As shown in Table 1, the average diaper weight has Srinivasan.21
been reduced by almost 40% in a period of 18 years from 1987 to The SC model uses a discrete-time representation, where one
2005.18 With this, the disposable diaper shows no significant day is divided into a predefined number of time ticks t. The
difference from home and commercially laundered reusable cloth location of various SC entities such as the plant, distributor, or
diaper in terms of environmental impacts.19 customer is represented through a pair of coordinates (x, y). Raw
While the LCA technique can be effectively applied to material arrival at the plant is modeled as shown:
measure the environmental consequences associated with bring-
ing a product to market, it offers limited assistance when the SC RA r ðtÞ ¼ RPr ðt LTr Þ ð1Þ
of any existing product has to be improved and issues related to
policies along the manufacturerdistributorcustomer chain where RAr(t) is the amount of raw material r arriving at the plant
come to the fore. LCA impacts have been derived from a at time t, RPr(t) is the amount of raw material r ordered from the
10179 dx.doi.org/10.1021/es201763q |Environ. Sci. Technol. 2011, 45, 10178–10185
Environmental Science & Technology ARTICLE
supplier, and LTr is the lead time between order and delivery of used for making decision to drive a SC operation to be greener as
the raw material r. illustrated by the following case studies.
Processing in the plant is modeled through the following:
RMreqr ¼ RCPr 3 CJamt ð2Þ 4. RESULTS AND DISCUSSION
Our base-case scenario is a SC network with a manufacturing
RUr ðtÞ ¼ RMreqr ð3Þ plant producing disposable diapers of year 2005 composi-
tion (see Table 1) and a single distributor, whose ordering policy
where RMreqr is the amount of raw material r required for is 1 day procurement interval. The breakdown of the overall envi-
processing a job CJ, which depends on RCPr, the recipe specify- ronmental impact to each stage in the SC is shown in Table 5. It
ing the amount of raw material r required to make one unit of can be seen that SC stages other than raw material and manu-
product, and CJamt, the amount of product to be made in job CJ. facturer (the two stages that are typically the focus of LCA
RUr(t) is the amount of raw material r needed for production at studies) have a significant contribution in some impacts. For
time t. example, the distributor contributes 62% of ozone layer deple-
The material balance on raw material inventory at the plant is tion, and the three transportation stages further add up to 20%
given by of ozone layer depletion. The following three scenarios con-
IR r ðt þ 1Þ ¼ IR r ðtÞ þ RA r ðtÞ RUr ðtÞ ð4Þ sider effects of product composition, SC network configura-
tion, and SC operation policy on economic and environmental
where IRr(t) is the inventory of raw material r at time t. impacts.
The complete model equations as summarized in Table 2 4.1. Scenario 1: Changing Diaper Composition. The first
consider all the key aspects of raw material procurement, scenario analyzes the impact of changes to the diaper composi-
logistics, manufacturing, and product distribution for the man- tions. Figure 2 compares the overall environmental impacts per
ufacturing plant and the distributor. Using this model, different carton diaper from the SCs of the two diaper compositions listed
SC policies, configurations, and also uncertainties can be simu- in Table 1. It shows a significant reduction in the water usage for
lated and their impacts on overall performance, both economic the 2005 diaper as compared to the 1987 one. This is mainly due
and environmental, quantitatively evaluated. Economic perfor- to the reduction in the quantity of fluff pulp used (14.57 g vs
mance is measured through indicators such as manufacturer 54.94 g). However, the figure also highlights an increase in the
profit and distributor profit (Table 3). Environmental perfor- other seven environmental indicators (global warming, photo-
mance is measured through 11 indicators, namely abiotic re- chemical oxidation, acidification, human toxicity, terrestrial eco-
source depletion (ARD), global warming (GWP), ozone layer toxicity, energy consumption, and abiotic resource depletion) for
depletion (ODP), photochemical oxidation (PO), acidification the 2005 diaper. This is mainly caused by the increased amount
(ACD), eutrophication (EUT), human toxicity (HT), fresh of SAP in the 2005 diaper (13.65 g vs 0.74 g). The increased
water aquatic ecotoxicity (FWAE), terrestrial ecotoxicity (TE), amount of polypropylene in the 2005 diaper (7.224 g vs 4.355 g)
water usage (WU), and energy consumption (EC), evaluated at also contributes quite significantly to the increase in the environ-
each stage of the SC (Table 4). These indicators are adapted mental impacts for these categories.
from a report by the UK Environmental Agency;19 more details The new composition also leads to an eight times increase in
are provided in the Supporting Information. The model has been manufacturer profit since it requires less raw material (41.95 g vs
implemented in MATLAB/Simulink,22 where flows of material, 67 g). Although the old composition uses less SAP, it uses much
information, and finance are depicted by various mathematical, more fluff pulp. The results also reveal that the environmental
logical, and algorithmic operation blocks. Such a model can be impacts from raw material transportation are 37% less in the
10180 dx.doi.org/10.1021/es201763q |Environ. Sci. Technol. 2011, 45, 10178–10185
Environmental Science & Technology ARTICLE
1 DOrdList r DSPðCO, DOrdListÞ the distributor receives a customer order CO and inserts it into its
order list DOrdList following its scheduling policy DSP, e.g. first-
come-first-serve, priority-based, etc.
2 DCO ¼ DOrdList1 order to be processed by the distributor, DCO, is the first customer
order in the order list DOrdList.
3 DPDðtÞ ¼ DCOamt if DIPðtÞ g DCOamt product of amount DCOamt will be delivered from the distributor
to the customer as DPD(t) if there is sufficient distributor inventory
at time t. DIP(t) is the product inventory at the distributor at time t.
4 DCOst ¼ t DCOst is the time at which the order is sent out from the
distributor.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
5 2 DCOtt is the transportation time for order DCO from the
DCO ¼ DTS ðDx DCOxloc Þ2 þ ðDy DCOyloc Þ
tt
distributor to the customer. DTS is the transportation speed.
The distributor is located at coordinates (Dx, Dy) and the
customer is at (DCOxloc, DCOxloc). Straight line distance is
assumed.
6 DCOdt ¼ DCOst þ DCOtt Product reaches the customer at time DCOdt.
(
7
DRPðtÞ ¼
DRT DIPðtÞ ∑ DRWðtÞ, if DRPðtÞ > 0, t ¼ c 3 DPC, c ¼ 1, 2, 3 the distributor employs the fixed interval procurement policy. It
0 , otherwise places orders with the manufacturer, DO, of amount DRP(t) at a
regular interval DPC to bring the inventory up to a certain top-up
8 DOtime ¼ t
level DRT. ∑DRW(t) is the amount of product which has been
9 DOamt ¼ DRPðtÞ ordered but is yet to arrive at the distributor, and c is the
procurement cycle index. DOtime is the time at which order DO
is placed with the manufacturer. DOamt is the amount ordered in
DO.
10 DPAðtÞ ¼ CJamt at t ¼ CJdt the distributor order is processed by the manufacturer as job CJ.
CJdt is the time at which order CJ arrives at the distributor. DPA(t)
is the quantity that arrives at the distributor at time t. CJamt is the
amount in job CJ.
11 DIPðt þ 1Þ ¼ DIPðtÞ þ DPAðtÞ DPDðtÞ material balance on product inventory at the distributor
12 JobSch r SPðDO, JobSchÞ the manufacturer receives an order DO from the distributor and
inserts it as a job into its job schedule JobSch following its
scheduling policy SP.
13 CJ ¼ JobSch1 if IR r ðtÞ g RMreqr and PSðtÞ ¼ 0 the first job in JobSch will be assigned as CJ, the job to be processed,
if there is sufficient raw material inventory and the plant is not
currently processing any other job. IRr(t) is the inventory of raw
material r at time t. RMreqr is the amount of raw material r required
for processing CJ. The processing status PS(t) indicates the amount
of time remaining to complete the current job. The job is
completed when PS(t) reaches 0.
14 RMreqr ¼ RCPr 3 CJamt RCPr is the recipe specifying the amount of raw material r required
to make one unit of product. CJamt is the amount of product to be
made in job CJ.
15 CJst ¼ t CJst is the processing start time of job CJ.
16 PSðt þ 1Þ ¼ CJ pt
the plant processing status is updated after a job is started. CJpt is
the total processing time for job CJ.
(
17 PSðtÞ 1 if PSðtÞ > 0 as time progresses, the processing status is updated to keep track of
PSðt þ 1Þ ¼
0 otherwise the amount of time remaining to complete the current job.
18 RUr ðtÞ ¼ RMreqr RUr(t) is the amount of raw material r used for production at time t.
19 CJ ¼ PR 3 CJ
pt amt
CJpt is the processing time for job CJ, based on the processing time
per unit product PR.
20 CJpkgt ¼ PkgR 3 CJamt CJpkgt is the packaging time for job CJ, based on the packaging time
per unit product PkgR.
Table 2. Continued
no. equation explanation
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tt
21 2 CJ is the transportation time for job CJ from the plant to the
CJtt ¼ TS ðPx DxÞ2 þ ðPy DyÞ
distributor. TS is the transportation speed. The plant is located at
coordinates (Px, Py) and the distributor is at (Dx, Dy). Straight line
distance is assumed.
22 CJdt ¼ CJst þ CJpt þ CJpkgt þ CJtt product reaches the distributor at time CJdt.
(
23
RPr ðtÞ ¼ ∑
RTr IR r ðtÞ RW r ðtÞ, if IR r ðtÞ þ ∑ RWr ðtÞ < RR r the plant employs the reorder point procurement policy. Raw
0 , otherwise material will be purchased when its inventory falls below the
reorder point RRr. RPr(t) is the amount of raw material r
ordered by the plant to its supplier, RTr is the inventory top-up
level for raw material r, and ∑RWr(t) is the amount of raw material r
which has been ordered but is yet to arrive at the plant.
24 RA r ðtÞ ¼ RPr ðt LTr Þ RAr(t) is the amount of raw material r arriving at the plant at time t,
LTr is the lead time between purchase and arrival for raw material r.
25 IR r ðt þ 1Þ ¼ IR r ðtÞ þ RA r ðtÞ RUr ðtÞ material balance on raw material inventory at the plant
1 DCO rev
¼ DPrice 3 DCO amt
DCO rev
is the distributor’s revenue from order DCO. DPrice is the
price per unit product charged by the distributor.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2 DCOtc is the transportation cost for order DCO. DCostT is the
DCOtc ¼ DcostT 3 DCOamt 3 ðDx DCOxloc Þ2 þ ðDy DCOyloc Þ
transportation cost per unit distance per unit product from the
distributor to the customer.
0 1
3
B
∑ DCOtc
C
DProfit is the profit of the distributor, Price is the price per unit product
B
B
þ ∑
Price 3 DPAðtÞ C
C
charged by the plant and paid by the distributor, DCostI is the inventory
DProf it ¼ ∑ DCO B
rev
B
B þ
t
∑
C
DcostI 3 DIPðtÞ C
C
cost per unit product per time tick, and DCostOF is the fixed operating
@ t A cost (charged at each time tick), D is the number of days in the
þ ðDcostOF 3 D 3 TÞ simulation horizon, and T is the number of time ticks in one day.
4 CJ rev
¼ Price 3 CJ amt
CJrev is the plant’s revenue from job CJ. Price is the price per unit
product charged by the plant.
5 CJpc ¼ CostOV 3 CJpt CJpc is the variable processing cost for job CJ. CostOV is the processing
cost charged when the plant is processing a job.
6 CJpkgc ¼ CostPkg 3 CJamt CJpkgc is the packaging cost for job CJ. CostPkg is the packaging cost
per unit product.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
7 CJtc ¼ CostT 3 CJamt 3 ðPx DxÞ þ ðPy DyÞ2 CJtc is the transportation cost for job CJ. CostT is the transportation
cost per unit distance per unit product from the plant to the distributor.
0 1
8
B
∑ ðCJpc þ CJpkgc þ CJtc þ CJpen Þ
C
profit is the profit of the manufacturing plant, CostRr is the price of raw
B
B
þ ∑∑ CostR r 3 RA r ðtÞ C
C
material r, CostI is the inventory cost per unit raw material per time tick,
Prof it ¼ ∑ CJ B
rev
B
B þ
t
∑∑
r
CostI 3 IR r ðtÞ
C
C
C
and CostOF is the fixed operating cost (charged at each time tick
@ t r A regardless plant is processing or idle).
þ ðCostOF 3 D 3 TÞ
2005 case, because less raw materials need to be transported backup to the other, leading to higher customer satisfaction level.
resulting in fewer trips. Table 6 shows the changes in the environmental impacts per
4.2. Scenario 2: Changing Distribution Network Config- carton of diaper in the 2-distributor scenario as compared to the
uration. The second scenario evaluates the effects of the base case. Overall, there is an 85% increase in all environmental
distribution network by comparing a single distributor (the base impacts other than water use for the plant-to-distributor trans-
case) versus two distributors for a given geographical market. portation. The reason is that the second distributor is located
While the single distributor channel could be more cost-efficient farther from the manufacturing plant than the existing one.
and easier to manage, two distribution channels would have the However, this is partially offset by lower distributor-to-customer
benefit of being at closer proximity to customers. Another transportation impacts (other than water use), which decrease by
advantage of the two distributor channels is that robustness 24%, since customers are served by the nearer distributor
increases since in the event of disruptions one can serve as a resulting in less travel distance. The table also shows comparable
10182 dx.doi.org/10.1021/es201763q |Environ. Sci. Technol. 2011, 45, 10178–10185
Environmental Science & Technology ARTICLE
1 TEIRMi ¼ ∑t ∑r EIRMi, r 3 RAr ðtÞ TEIRMi is the total environmental impact i from raw materials, where i = ARD, GWP,
ODP, PO, ACD, EUT, HT, FWAE, TU, WC, or EC. EIRMi,r is the environmental
impact i per unit raw material r.
2 TEITSPi ¼ ∑t ∑r EITRMi 3 RAr ðtÞ 3 SPDistðtÞ TEITSPi is the total environmental impact i from raw material transportation from
suppliers to the plant. EITRMi is the environmental impact i from raw material
transportation per unit raw material per unit distance. SPDist(t) is the distance
covered for raw material transportation from suppliers to the plant.
3 TEIMi ¼ ∑t EIMi 3 DPAðtÞ TEIMi is the total environmental impact i from manufacturing. EIMi is the
environmental impact i from manufacturing per unit product.
4 TEIPkgi ¼ ∑t EIPkgi 3 DPAðtÞ TEIPkgi is the total environmental impact i from packaging. EIPkgi is the environmental
impact i from packaging per unit product.
5 TEITPDi ¼ ∑t EITPi 3 PDDistðtÞ TEITPDi is the total environmental impact i from product transportation from the plant
to the distributor. EITPi is the environmental impact i from product transportation per
unit distance. PDDist(t) is the distance covered for product transportation from the
plant to the distributor.
6 TEIDi ¼ ∑t EIDi 3 DPAðtÞ TEIDi is the total environmental impact i from the distributor operation. EIDi is the
environmental impact i from distributor operation per unit product.
7 TEITDCi ¼ ∑t EITPi 3 DCDistðtÞ TEITDCi is the total environmental impact i from product transportation from the
distributor to customers. DCDist(t) is the distance covered for product transportation
from the distributor to customers.
8 TEIi ¼ TEIRMi þ TEIPkgi þ TEIMi þ TEIDi TEIi is the total environmental impact i from the whole supply chain.
þ TEITSPi þ TEITPDi þ TEITDCi
raw materials 83% 80% 96% 87% 84% 45% 12% 85% 99% 41%
supplier to plant 4% 1% 1% 1% 1%
manufacturer 7% 12% 15% 2% 4% 5% 33% 57% 10% 29%
packaging 6% 2% 3% 3% 1% 1% 1% 1% 1%
plant to distributor 1% 8% 1% 2% 1% 1% 1%
distributor 3% 5% 62% 5% 4% 5% 18% 27% 5% 26%
distributor to customer 1% 8% 1% 2% 1% 1% 1%
Table 6. Comparison of Environmental Impact Per Carton Diaper in 2-Distributor and 1-Distributor Scenariosa
stage ARD GWP ODP PO ACD EUT HT FWAE TE WU EC
raw materials ∼ ∼ = ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼
supplier to plant ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ = ∼
manufacturer = = = = = = = = = = =
packaging = = = = = = = = = = =
plant to distributor +85% +85% +85% +85% +85% +85% +85% +85% +85% = +85%
distributor ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼
distributor to customer 24% 24% 24% 24% 24% 24% 24% 24% 24% = 24%
total ∼ ∼ +5% ∼ ∼ +1% +1% +1% ∼ ∼ +1%
a
Legend: = Equal value; ∼ Less than 0.5% difference.
Table 7. Comparison of Total Environmental Impact in 2-Day and 1-Day Procurement Interval Scenariosa
stage ARD GWP ODP PO ACD EUT HT FWAE TE WU EC
raw materials ∼ ∼ = ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼
supplier to plant ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ = ∼
manufacturer 1% 1% 1% 1% 1% 1% 1% 1% 1% 1% 1%
packaging 1% 1% 1% = 1% 1% 1% 1% 1% 1% 1%
plant to distributor 6% 6% 6% 6% 6% 6% 6% 6% 6% = 6%
distributor = = = = = = = = = = =
distributor to customer ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ = ∼
total ∼ ∼ 1% ∼ ∼ 1% ∼ ∼ ∼ ∼ ∼
a
Legend: = Equal value; ∼ Less than 0.5% difference.