integrated logistic
TRANSCRIPT
-
7/23/2019 Integrated Logistic
1/150
Int gr ted
2
Beulle
Univer
2/1/20
Lo
01
sP.
sityofSout
15
ist
5
ampton
cs
-
7/23/2019 Integrated Logistic
2/150
Integrated Logistics
2
Contents
1 Introduction .............................................................................................. 4
1.1 Defining Supply Chain Management ........................................................... 4
1.2
Understanding Supply Chain Management .................................................... 4
1.3 Measuring Supply Chain Performance ......................................................... 4
1.4 What is Integrated Logistics? ................................................................... 4
2 IL & Finance .............................................................................................. 5
2.1 Strategy and Finance ............................................................................ 5
2.2 Time value of money ............................................................................. 6
2.2.1 Net Present Value for discrete interest rates ............................................. 6
2.2.2 Net Present Value for continuous interest rates ......................................... 8
2.2.3 Annuity Stream ............................................................................... 10
2.2.4 Linear approximations of NPV and AS ..................................................... 10
2.2.5
NPV and AS of a few useful cases .......................................................... 11
2.3 Economic Order Quantity (EOQ) Harris (1913) ............................................ 14
2.4 Economic Production Quantity (EPQ) Taft (1918) ........................................ 18
2.5 EOQ with batch demand Grubbstrm (1980) .............................................. 21
2.6 Lot-for-lot production at finite rate Monahan (1984) .................................... 24
2.7 EOQ for batch demand Goyal (1976) ....................................................... 26
2.8 EPQ for batch demand Joglekar (1988) .................................................... 28
2.9 Vending machine ................................................................................ 29
2.10 Payment structures ............................................................................. 30
2.11 Consignment arrangements .................................................................... 32
2.12
The profit function of the integrated supply chain ........................................ 34
2.13 Using the NPV framework to include other cost components ............................ 35
3 IL in Buyer-Supplier Supply Chains .................................................................. 38
3.1 One-to-one shipping ............................................................................. 38
3.1.1 Shortest Path Problem ....................................................................... 38
3.1.2 Economic Transport Quantity (ETQ) ....................................................... 41
3.1.3 Maximum Economic Haulage Radius (MEHR) ............................................. 45
3.1.4 Optimal policy to order N different items from a Cross-Docking Facility (CDF) .... 46
3.2 One-to-many shipping .......................................................................... 48
3.2.1 Length of an optimal TSP tour visiting many customers ............................... 48
3.2.2
Expected length of an optimal TSP tour visiting a few customers ................... 51
3.2.3 Continuous approximation of an optimal vehicle routing solution ................... 52
3.2.4 One-to-many ETQ ............................................................................ 54
4 Strategy in the supply chain: Alliances versus Leaders .......................................... 61
4.1 Alliances .......................................................................................... 61
4.1.1 Cooperative game theory ................................................................... 61
4.1.2 Alternative methods ......................................................................... 71
4.2 Two-party alliances in the supply chain ..................................................... 71
4.2.1 Two-party alliance: example ............................................................... 71
4.2.2 Two-party alliance: example 2 ............................................................. 79
4.3 Perfect coordination ............................................................................ 82
-
7/23/2019 Integrated Logistic
3/150
Integrated Logistics
3
4.3.1 Perfect coordination schemes: examples ................................................ 83
4.4 Leaders and followers .......................................................................... 88
4.4.1 Leaders and followers: Stackelberg games ............................................... 88
4.4.2 Stackelberg leader in the supply chain: examples ...................................... 90
5 Stochastic Models ...................................................................................... 96
5.1 When is the assumption of a constant demand rate valid? ............................... 96
5.2 (r, Q) reorder point models .................................................................... 97
5.2.1 Backorder case ................................................................................ 98
5.2.2 Lost sales case .............................................................................. 103
5.2.3 Service level approach .................................................................... 103
5.2.4 Standard Tables ............................................................................ 105
5.2.5 Exercises ..................................................................................... 110
5.3 News vendor problems ........................................................................ 111
5.3.1 Example ...................................................................................... 111
5.3.2 Cost minimisation .......................................................................... 111
5.3.3 Profit maximisation ........................................................................ 112
5.3.4 Regret minimisation ....................................................................... 113
5.3.5 Exercises ..................................................................................... 114
6 MRP, JIT, and Bottlenecks .......................................................................... 116
6.1 Materials Requirements Planning ........................................................... 116
6.1.1 MRP inputs ................................................................................... 116
6.1.2 MRP outputs ................................................................................. 118
6.1.3 Lot sizing policies .......................................................................... 122
6.1.4 Remarks on MRP ............................................................................ 125
6.2
Just-In-Time .................................................................................... 130
6.2.1 Motivation ................................................................................... 131
6.2.2 Push and pull systems ..................................................................... 131
6.2.3 The Kanban System ........................................................................ 131
6.2.4 How many cards? ........................................................................... 132
6.2.5 Subcontractors .............................................................................. 134
6.2.6 Fluctuations in demand ................................................................... 134
6.2.7 Comparison JIT and MRP .................................................................. 136
6.3 Bottleneck scheduling ........................................................................ 139
6.3.1 Optimised Production Technology (OPT) ............................................... 139
6.3.2
Theory Of Constraints (TOC).............................................................. 143
-
7/23/2019 Integrated Logistic
4/150
Integrated Logistics
4
1 Introduction
Foranyquestions,email:[email protected]
Seealsothematerialonblackboard.
Thistext follows the lecturesgiven,but Imayhavearranged thesequenceofa few topics inorder toarriveatamorelogicalflow.
1.1 DefiningSupply
Chain
Management
Seeslidesonblackboardandreadthefollowingarticle(onblackboard):
MENTZER J T, DE WITT W, KEEBLER J S, MIN S, NIX N W, SMITH C D, AND ZACHARIA Z G. 2001.
DEFININGSUPPLYCHAINMANAGEMENT.JOURNALOFBUSINESSLOGISTICS22(2),125.
1.2 UnderstandingSupplyChainManagement
Seeslidesonblackboardandreadthefollowingarticle(onblackboard):CHENIJANDPAULRAJA.2004.UNDERSTANDINGSUPPLYCHAINMANAGEMENT:CRITICALRESEARCH
ANDATHEORETICALFRAMEWORK. INTERNATIONAL JOURNALOFPRODUCTIONRESEARCH42 (1),131
163.
1.3 MeasuringSupplyChainPerformance
Seeslidesonblackboardandreadthefollowingarticle(handout):SHEPHERDCANDGNTERH.2006.MEASURINGSUPPLYCHAINPERFORMANCE:CURRENTRESEARCH
AND FUTURE DIRECTIONS. INTERNATIONAL JOURNAL OF PRODUCTIVITY AND PERFORMANCE
MANAGEMENT55(3/4),242258.
1.4 Whatis
Integrated
Logistics?
Seetheslidesonblackboard.
-
7/23/2019 Integrated Logistic
5/150
Integrated Logistics
5
2 IL Finance
2.1 StrategyandFinance
Source: SILVER A., PYKE D.F., AND PETERSON R. 1998. INVENTORY MANAGEMENT AND PRODUCTION
PLANNING AND SCHEDULING, THIRD ED. JOHN WILEY & SONS, NEW YORK, CHAPTER 2: STRATEGICISSUES,P.14.
Integrated Logistics (IL) should be linked to the corporate and business strategy. The most importantobjectiveofanyfirmisarguablylongtermprofitability.Inthiscontexttheoperatingprofitisdefinedas:
ILcanaffectbothtermsontherighthandside.Byreducinge.g.aggregateinventorylevelsinthefirm,
the investmentcostcanbe reduced.Byallocatingproper inventory levelsamongdifferent items inanimprovedway,salesrevenuemayincrease.
One common aggregate performance measure in IL and inventory management is the inventory
turnover:
An increase in sales without a corresponding increase in inventory will increase the inventory
turnover,aswilladecreaseininventorywithoutadeclineinsales.Turnovercanbeausefulmeasuretocomparedivisionsofafirmorfirmsinanindustry.
Ahigherturnoverratioforthesamelevelofsalesmeansamoreprofitablebusiness,aslessmoneyis
tiedupininventories.Thedangeristhatareductionofinventorylevelsmayalsonegativelyaffectsales.Whenitisnotknownwithcertaintyhowmuchdemandthereisperperiod,anamountofsafetystockis
needed to make sure enough products are in stock in case demand would be somewhat higher than
expected.Furthermore,theright figureof inventoryturnover foryour firmdependsonthe levelof in
houseproductionversusoutsourcing.Afirmthatdoeseverything inhousewillneeda lower inventory
turnoverthan
afirm
that
is
completely
based
on
outsourcing
of
the
production.
Indeed,
the
first
firm
will
alsohaveastockofrawmaterialsandworkinprocess,whilethesecondwillonlyhaveendproductsininventory.These considerationsare importantwhen comparing inventory turnover figures of different
firms.OtherusefulperformancemeasuresforILincludethosethatmeasureallsortsof:costs;averageand
variabilityofleadtimes(oftenseenasthetimebetweeninitiationofsomesalesorderandrealisationof
thesales,sometimesalsojust thetime istakes foraproduct tomovethroughaparticularpartofthe
supply chain); product and service quality; customer satisfaction; and innovativeness (see also Section
1.3).
-
7/23/2019 Integrated Logistic
6/150
Integrated Logistics
6
2.2 Timevalueofmoney
Source (on blackboard): GRUBBSTRM, R.W. 1980. A PRINCIPLE FOR DETERMINING THE CORRECT
CAPITAL COSTS OF WORKINPROGRESS AND INVENTORY. INT. J. OF PRODUCTION ECONOMICS 18(2),259271.
2.2.1 NetPresentValuefordiscreteinterestrates
IfIinvestV()intoaprojectnow,theprojecthavingarateofreturn =0.2afteroneyear,whatwillbetheamountofmoneya()thatIwillreceiveoneyearlater?
Theanswer: 1 1.2
If Ikeepthe investmentrunningnot foronebutforTyears intotal,whatwillbemyrewardattheend?
Theanswer: 1 1 1 1 Wecanturnthisaroundandaskthequestion:whatisthecurrentvalueV()ofreceivinga()within
Tyearstimefromnow?Algebraicmanipulationoftheaboveequationgives:
1
Time
Cashflows
0 1
V
a
Time
Cashflows
0 1
V
a
T...2
-
7/23/2019 Integrated Logistic
7/150
Integrated Logistics
7
ViscalledtheNetPresentValueofa. Itthusalldependsonthevalueof,calledthediscountfactor,interestrate,orinternalrateofreturn. Companiestypicallyusevaluesofwithintherange0.1to0.3.
Considerasequenceofnpaymentsofdifferentamounts , 1, , atequidistantpoints intimewithcycletimeT(seeFigurebelow).WhatistheNetPresentValueofthesepayments,if
istherateof
returnoveroneperiodT?
Theanswer:
1 Therateofreturn isafunctionofthetimeperiodoverwhich it isdefined.Forexample, if
isthe
rateofreturnoveraoneyearperiod,whatistheequivalentrateofreturndefinedoveraperiodofone
month(take
this
to
be
1/12th
of
ayear)
that
would
give
the
same
Net
Present
Value?
Answer:Letuscall theequivalentrateofreturnoveronemonth,thenwithT=12: 1 1 1 1
Mathsrefresher
ln ln1 ln1
1T ln1 ln1
Thus,iftheannualinterestrateis0.2,theequivalentmonthlyinterestratewouldbegivenby:
Time
Cashflows
0 1
Van
n...2
a2
a1
TT
-
7/23/2019 Integrated Logistic
8/150
Integrated Logistics
8
112 ln1 0 . 2 ln1
. 1 0.0153Notethatiftheinterestratewouldbedefinedoverasmalltimeperiod,thenln1 .
2.2.2 NetPresentValueforcontinuousinterestrates
In general, the conversion of interest rates corresponding to different lengths of period obeys the
formula:
1 ln1 ,
where isthe interestratepertainingtoaperiodof lengthand isthecontinuous interestratecorrespondingtothelimitlengthzero.ThedefinitionofNetPresentValueinthecontinuouscasebecomes:
,where
isthecashflowattime
plussomemultiple
ofDiracsDeltafunctionatpoints
at
which
there
are
finite
payments
.
Wewilltypicallyneedtosolveonlyspecialcases,givenbelow.
Example1Aoneofflumpsum()receivedatfuturetimeL
Thus the NPV ofacashflow received L time units in the future is the cashflowmultipliedwith its
delayfactor.Time
Cashflows
L
a
0
V
-
7/23/2019 Integrated Logistic
9/150
Integrated Logistics
9
Example2Afinitenumberofcashrevenues,eachreceivedtheirownfuturemoment
Example3Aninfinitenumberofequalcashrevenuesreceivedatequidistantmoments
Mathsrefresher
If|| 1,then For , || 1
Thus: 1
Time
Cashflows
1
an
n...2
a2
a1
T2
T1
0
V
Time
Cashflows
0 1 ...2
aa
TT
...
a
-
7/23/2019 Integrated Logistic
10/150
Integrated Logistics
10
Example4Acontinuouscashflowatarateofa(/year)receivedforeternity.
2.2.3 AnnuityStream
TheAnnuityStreamASofaseriesofcashflows istheNetPresentValueVofthesecashstreamstimestherateofreturn:
Fromexample4,itisclearthattheannuitystreamisthatcontinuousstreamofcashyieldingthesamenetpresentvalueastheoriginalseriesofcashflows.
2.2.4 LinearapproximationsofNPVandAS
Mathsrefresher
Maclaurinexpansionofanexponentialfunction ! 1 ! andconvergesfor
Usingthisresultfor ,itiseasytoseethat:
1
2
,
andit
can
also
be
proven
(after
lengthy
algebraic
manipulation):
1 1 2 12 This can be used to derive linear or quadratic approximations in of NPV or AS functions, as
illustratedinthenextsection.
Time
Cashflows
0
a ....
-
7/23/2019 Integrated Logistic
11/150
Integrated Logistics
11
2.2.5 NPVandASofafewusefulcases
Case1Aoneofflumpsuma()receivedatfuturemomentL
1 1 Foraoneofflumpsum,boththelinearapproximationoftheNPVandthequadraticapproximationin
of the AS are acceptable, and indeed would be of the same accuracy. However, the linear
approximationof
AS
would
be
insufficiently
accurate
as
indeed
the
above
shows,
cannotbetheASas a itself isnot itsNPV.The linear approximationof the AS would thus neglect the delayeffect altogether.Case2Aninfiniteseriesoflumpsumpaymentsa()receivedwithcycletimeT.
1 1 1 2 12
1 12 12
1 1 2Foraninfiniteseriesofcashflows,thelinearapproximationofASisacceptable.
Time
Cashflows
0 L
a
Time
Cashflows
0 1 ...2
aa
TT
...
a
-
7/23/2019 Integrated Logistic
12/150
Integrated Logistics
12
Case3Acontinuouscashflowatratea(/year),startingattimeL,andforeverlasting.
Itcanbeobservedthatthisresultcouldhavebeenobtainedbycombiningexample4andcase3fromabove.IndeedtheNPVattimeLofthecontinuouscashflowaisa/(example4),andthenaccountingforthedelaywithtimeL(case1)givestheaboveresult.
1 2 1
Case4 Acontinuouscashflowatratea(/year)receivedinfuturetimeperiodT,startingat
timeL.
1
Thisresultcanalsobeobtainedbyviewingthetemporarycontinuouscashflowaasthesumoftwo
infinitecontinuous
cash
flows
+a
and
a
(starting
atime
Tlater),
and
thus
by
applying
the
result
of
case
3
twice:
Time
Cashflows
0
a ....
L
Time
Cashflows
0
a
L T
-
7/23/2019 Integrated Logistic
13/150
Integrated Logistics
13
1 To find the linear approximation, it is safe to take the quadratic terms first into consideration and
approximatelateron:
1 2 1 1 2 2 1
Case5 Aninfiniteseriesofcontinuouscashflowsatratea(/year)withcycletimeLandeachtimereceivedforalengthoftimeT(T L).
The result of case 4 (without delay) can first be applied to find the NPV of the cashflow of every
periodL.Usingtheapproachasincase2thengives:
1 1 1 TofindthelinearapproximationofAS,itisagainsafetofirstconsiderthequadratictermsaswell:
1
1 1
2! 1 2
2 2
Time
Cashflows
0
a
L
a
T T
L
....
-
7/23/2019 Integrated Logistic
14/150
Integrated Logistics
14
2.3 EconomicOrderQuantity(EOQ)Harris(1913)
Aretailersellstocustomersatypeofproductatthepricepperproduct.Demandfortheproductcan
beassumedtooccurataconstantrateaccordingtoanannualdemandofproducts.Theretailerhastopurchasefromanexternalsupplieratprice
perproduct,andalsohastopaya
fixed
order
cost
for
each
order
placed
at
this
supplier
(this
could
be
the
fixed
cost
of
transport
plus
the
fixedcostofadministrationtoplaceanorder).
Figure 1. Lot-size model (EOQ).
When the retailerwouldplaceanorderorsize Q (products/order)withcycletime T,the inventory
levelovertimeattheretailerwillfollowtheclassicalsawtoothpatternasillustratedintheFigureabove.
Itcanbeobservedthat:
If there is a nonzero leadtime , and we want to make sure that the order arrives when the
inventory drops to zero, we must order in advance when the inventory level is . This is called thereorderpoint.
Thetraditionalapproachtoderivingtheoptimalpolicy
Weconsideradeterministicsysteminwhichallrelevantparametersareconstantandshortagesarenotallowed.Thepolicyusedis(r,Q).Althoughtheaimistofindoptimalvaluesforbothrandq,theoptimal
valuefor
iseasilydetermined.TheproblemthereforereducestofindingtheoptimallotsizeQ.Inthis
classiceconomiclotsizesystemthefollowingassumptionsaremade:
1. constantannualdemandratey(items/year);
w(/product)s(/order)
y(products/year)p(/product)
retailer
Q(products/order)?T(cycletime)?
Inventory
retailer
Time
...
T
Q
L
r
-
7/23/2019 Integrated Logistic
15/150
Integrated Logistics
15
2. constantinfinitereplenishmentrateR=;
3. constantunitholdingcosth(/item,year);
4. constantunitordercosts(/order);
5. noshortagesallowed;
6. constantleadtimeL=0,easilyextendedtoconstantL0(years).
7. constantreplenishmentquantityQ(items/order).
TheinventoryfluctuationsinthissystemareillustratedinFigure1.Itisclearthatweplacetheorderatexactlythatmomentsothatreplenishmentsarrivewhenonhandinventoryreacheszero:ifleadtime
L=0,weorderwhentheinventory levelI(t)=r*=0; ifL0(andL
-
7/23/2019 Integrated Logistic
16/150
Integrated Logistics
16
Figure 2. Global cost f unct ion of t he EOQ model.
yhsQ 2* 7
Andsince
y= Q/T, 8
theoptimalcycletimeis:
yh
s
y
QT 2
*
* 9
Graphically, the cost equations can be described as in Figure 2. At optimum, annual holding costs
equalannualreplenishmentcosts.
UsingtheNPVframeworktoderivetheoptimalpolicy
TheNPVframeworkcanalsobeusedtoderivetheoptimalvaluesofQandT.Therefore,thecashflows
for the retailer have to be determined. The following assumptions are adopted: Since demand isoccurring at a constant rate
, the customers pay a continuous cashflow of
to the retailer. The
retailerwill
pay
the
set
up
cost
uponreceivingeverybatch,aswellastheamount tothesupplier.ThisisillustratedinFigure3.
0 Q* Q
(Q)
Ordercost
Holdingcost
Totalcost
-
7/23/2019 Integrated Logistic
17/150
Integrated Logistics
17
Figure 3. Cash-f l ows repr esentat ion of t he EOQ model .
Theannuitystreamoftheprofitfunctionfortheretaileris:
1
Thelinearapproximation inisthus: 1 2
2
2
Since
:
2 2 The optimal value for Q can be obtained by taking the derivative of this profit function to Q, and
settingthisequalto0: 2 0
2 Thusthesameresult isobtained,and inaddition itbecomesveryclearhowtheholdingcost(per
productperyear),usedinthetraditionalinventoryframework,needstobecalculated: Thus,theholdingcostfortheretailerisrelatedtotheamountofmoneyinvestedperproduct,i.e..
Wewilllaterencounterexamplesweretheholdingcostsperproductperyearwillbedifferent.
Time
Cashflows
0
s+wyT
T
py ....
....
-
7/23/2019 Integrated Logistic
18/150
Integrated Logistics
18
2.4 EconomicProductionQuantity(EPQ)Taft(1918)
Inthissectionwewillanalysealotsizesysteminwhichthereplenishmentrateisnotnecessarilyinfinite
ashasbeenassumedintheprevioussection.Specifically,thesystemhasauniformreplenishmentrateR
(items/year),where itisobviouslynecessarythatRy.Thistypeofreplenishinggenerallyoccurswhen
the
demand
has
to
be
met
by
a
manufacturing
department
inside
the
company.
The
inventory
fluctuationscanthenbedescribedgraphicallyasinFigure4.
Aproducersellstocustomersatypeofproductatthepricewperproduct.Demandfortheproduct
canbeassumedtooccurataconstantrateaccordingtoanannualdemandofproducts.Theproducerhastomaketheproductsatavariableproductioncostperproduct,andalsohasto
payafixedsetupcostforeachrunoftheseproducts. Productionoccursataconstantfiniteproductionrateequivalenttoanannualrateof(productsperyear).Thetraditionalapproachtoderivingtheoptimalpolicy
Figure 4. Manufact ur ing lot size syst em.
Theaverageamountofinventoryequals
E(I) = |bc|/2 10
Wehavethefollowingrelationships:
c(/product)s(/order)
R(products/yr)
y(products/year)w(/product)
producer
Q(products/order)?
T(cycletime)?
t
I(t)
q
0
T
Lr
R
y
b
a
c
-
7/23/2019 Integrated Logistic
19/150
Integrated Logistics
19
abacbc (geometrical relationship) 11
rL
aby (demand rate)
12
rr L
Q
L
acR (replenishment rate)
13
Hencetheaverageamountininventorycanberewrittenasafunctionoftheknownparametersyand
RandthevariableQ:
R
yQ
QR
y
QIE 122
1
)(
14
Thenumberofreplenishmentsperyearequalsy/Q.Thetotalsystemcostisthesumofholdingcosts
andreplenishmentcosts:
Q
ys
R
yhQ
Q
ysIhEQ
1
2)()(
15
ThevalueQ*whichminimisestotalcostscanbeobtainedasfollows:
012d
d2
Q
sy
R
yh
Q
(Q)
16
R
yQ
R
yh
syQ EOQ
1
1
1
2 ** 17
where
*
EOQQ
refers
to
the
economic
order
quantity
of
the
basic
EOQ
model.
The
corresponding
cost
isgivenbysubstitutionof:
R
yh
sy
ys
R
yh
sy
R
yhQ
1
21
21
2)( *
18
Rearranging:
-
7/23/2019 Integrated Logistic
20/150
Integrated Logistics
20
2
1
2
1
)( *
R
yshy
R
yshy
Q
19
R
y
R
yshyQ EOQ 112)(
** 20
where *EOQ referstothecostobtainedinbasicEOQmodel. ItiseasytoshowthattheEOQmodelis
aspecialcaseofthecontinuousrateEOQmodelbysimplysubstitutingR=.
Example.
Lety=1000items/year,R=2000units/year,h=1.6/year,ands=200/year.Then:
707500000
5.016.1
)2000)(1000(2
1
2*
R
yh
syQ units/order.
21
UsingtheNPVframeworktoderivetheoptimalpolicy
Thefollowingcashflowsareassumed:
Then,usingCase5ofSection2.2.5forthevariableproductioncosts:
1 1 2 2 2
2
1
2
Theoptimal
order
quantity
is
derived
in
the
usual
manner
from
the
first
order
conditions:
Time
Cashflows
0
s
T
wy ....
....
yT/R
cR
-
7/23/2019 Integrated Logistic
21/150
Integrated Logistics
21
21Thesameresultisthusobtainedasinthetraditionalderivationif:
2.5 EOQwithbatchdemandGrubbstrm(1980)
Weconsiderthesamesystemasintheprevioussectionbutwithtwomodifications:
1. Theproductionrateissetsuchthat ;2. Salesoccursinbatchesofsize
TheinventorylevelasafunctionoftimelookslikeamirrorimageoftheEOQsawtoothpattern:
Thetraditionalapproachtoderivingtheoptimalpolicy
ThiswouldbeexactlythesamemodelastheEOQofHarrisandwouldproduce:
Q
ys
hQ
Q
ysIhEQ
2)()(
UsingtheNPVframeworktoderivetheoptimalpolicy
FirstNPVsolution.ThisishowGrubbstrm(1980)derivedasolution:
c(/product)
s(/order)
R(products/yr)
y(products/year)
w(/product)
producer
Q(products/order)?T(cycletime)?
Time
Inventory
0
T
....
-
7/23/2019 Integrated Logistic
22/150
Henc
Ther
This
theprod
Noti
whileth
havepla
Alternat
conditio
assumpt
momen
tooearl
derivea
Thiscan
ethelinear
fore:
eansthatt
uctandnot
e,however,
estartofpr
cedabound
iveNPVsol
nisplaceda
ionwould
offirstdeli
y andhasu
noptimalpo
berewritte
pproximatio
heholdingc
hepurchase
thatwithth
ductionisk
arycondition
tion.Beulle
tthemome
akemore s
ery isnotfi
necessary s
licyfromthe
as:
n:
1
stsinthetr
price :
e increaseo
eptfixedat
attime0.TsandJanss
tintimewh
ense if the
xed,thecus
tock tooear
followingNP
Integrated Logi
12
2ditionalfra
wearedurrenttime
hisisnotnec
ns(2011)i
enthefirst
ustomerw
omermay
ly.Aprodu
Vcalculatio
stics
1 22 2
eworkaret
elayingthe0.Wecaniessarilyalwa
troducedan
atchtothe
nts this to
itherrunou
erunder th
s:
obebasedo
eliveryofo
terpretthis
ysthebesta
othermodel
ustomerha
appen, bec
tofstockor
se circumst
nthesales
derstothe
asamodel
ssumption.
wherethis
stobedeliv
use otherw
receivethe
anceswould
22
riceof
customer
herewe
oundary
red.This
ise if the
products
need to
-
7/23/2019 Integrated Logistic
23/150
Andsin
brackets
Givin
and
asallow
The
Alter
1.
i
2.
c
Wecan
Inconcl
framew
e isafixe:
g,when
otheedbythecu
oldingcosts
natively,we
heproducer
alued at hi
nterpretatio
heproducer
alled unit s
roducesare
thenrewrite
sion,it
is
in
rk.Thisbou
constant,
:
rwise.(Inth
tomerands
inthiscase
canlookatwill
have
as own inve
ofbeingba
willalsoha
uppliers re
venue,nota
theprofitfu
some
mod
darycondit
eneedto f
lattercase,
uchthatouldhence
2toderivetholding
cost
stment cost
sedoninves
eapositive
ard cost!
ction:
ls
therefore
oniscalledt
Integrated Logi
indtheopti
32
22 theproduc .etakenas:
woterms:
ofkeeping
s and this
mentsmad
ffectfromt (
32important
heAnchorP
sticsalpolicy fr
2
rhasaninc
roducts,jus
holding cos
intothepro
hebatchdeli
eullens and
2 ow
to
set
t
oint(Beullen
mthefunct
2
ntivetoma
like
in
the
t
correspon
ductplacedi
veriestothe
Janssens, 2
eboundary
sandJansse
ion insidet
ethelotsiz
raditionalE
s to the t
instock: customergi
011). Note
conditionin
ns,2011).
23
esquare
aslarge
Q,to
be
aditional;ingaso
that this
theNPV
-
7/23/2019 Integrated Logistic
24/150
Integrated Logistics
24
2.6 LotforlotproductionatfiniterateMonahan(1984)
Wecan extend the previousmodelof batchdemand by assuming that theproducer runs everybatch at some finite production rate . The producer would hence only start a production runsometime
/earlierrelativetothedeliveryofabatchtothecustomer.
Theclassiccostfunctionofthissystem(Monahan,1984)isgivenby:
2 A first solution using NPV is found from setting the Anchor Point at start of production at time 0,
assumingthecashflowsasinthefigurebelow.
Ifthestartofproductionhastooccurattime0,theannuitystreamis:
1 1 1 2 1 2 2 2
2 2 2 2 Comparingwiththeclassiccostfunction,wefindthat 2 andwehaveasanextraterm
thesuppliersrewardwith .Inthespecialcasethat weretrievethesolutionfromtheprevioussection:
2 2
Time
Inventory
0
T
....
yT/R
Time
Cashflows
0
s
T
wyT
....
....
yT/R
cR
-
7/23/2019 Integrated Logistic
25/150
Integrated Logistics
25
andtheholdingcost is then .Note,however, thatwecanalwayswrite 2 ,soinsteadofadoptingaspecialinterpretation ,weseethatwecanequallyadoptthemoregeneralinterpretation(sincevalidforany ofhaving 2 andanextraterm,thesuppliersreward,with .
WhensettingtheAnchorPointatthestartofsales,wefind:
1 Thefunctioninbetweenthesquarebracketsisrewrittenas:
1 2 2 2 2 Thus we can take
and identify, again, the suppliers reward as an extra term with
. The special case of is as seen in the previous section and can use the sameinterpretationsforand.Usingthetraditionalinventorymodelingapproach,onewouldonlyfindthefirstpart.Thisisinfact
whathappenedintheliteraturethathasfollowedMonahans(1984)model,andhencetherearemodels
intheliteratureforwhichitmaynotbeeasytoseewhethertheywillleadtoinventorypolicieswhichwill
alsomaximisetheNPVofthefutureprofitsofthefirm.SeealsoBeullens(2014).
We finishthissectionbyprovidingsome intuitionbehindthesuppliersreward.Asshownabove, it
indicatesthatthereisapositiverevenuetermintheproducerslinearisedASprofitfunction:
2 Thesuppliersrewardarisesfromthefactthatthecustomerordersinbatchratherthanoneproductatatime.Anintuitiveexplanationisthefollowing:
CaseA.Supposeyouhavetwooptionstoreceiveincome:eitherreceiving1,200atthestartofevery
year, or 100 at the start of every month, what would you choose? The logical answer would be to
choosethefirstoption.
Case B. Suppose you have two options to pay expenses: either paying 1,200 at the start of every
year, or100 at thestartof every month,what wouldyou choose? The logical answer this time is to
choosethelatteroption.
The fact that a customer orders in batch will cause inventory costs for this customer. Thedisadvantageforthecustomertoorderinbatchisextrainventoryholdingcoststobevaluedatinvested
cost , but at the same time this creates an advantage for the supplier as he receives his revenues earlier. The suppliers reward term incorporates this advantage into the suppliers profitfunction.
Homework
Derivetheoptimalorderquantityusingtheclassiccostfunctioninwhich andthenderivetheoptimalorderquantityusingthelinearisedannuitystreamfunctionunderthetwoassumptionsoftheAnchorPoint.Comparethethreeresults.
-
7/23/2019 Integrated Logistic
26/150
Integrated Logistics
26
2.7 EOQforbatchdemandGoyal(1976)
Weconsideramodelofadistributorwhoneedstodeliverordersofbatchsizetocustomerswithanaverage demand rate . Customers pay per product to the distributor but the distributor incurs adeliverycost
perdelivery.Thedistributorcanplaceordersofsize
toitsownsupplierandhastopay
per
product
and
has
an
order
cost
of
per
order.
Itcan
be
proven
that
it
is
optimal
for
the
distributor
to
have
forsomepositiveinteger.Seealsotheabovefigure.Theclassicderivationofthecostfunctionisagainbasedontrigonometryandproducestheresult(Goyal,1976): 12
ToderivetheASprofitfunction,weusethefollowingcashflows:
NotethatplacingtheAnchorPointatstartofsalestothecustomersorplacingtheAnchorPointat
startofthefirstbatcharrivingfromthesupplierproducesthesameboundaryconditionattime0.TheASfunctionis:
c(/product) (/order) y(products/year)w(/product)(products/order)(/order)distributor
(products/order)? (cycletime)?
Time
Inventory
0
....
Time
Cashflows
0
s
wyT ........
wyT
-
7/23/2019 Integrated Logistic
27/150
Integrated Logistics
27
2 1 2 2Comparisonofthisresultwiththeclassiccostfunctionshowsthat
,butalsothatthere isan
extraterm,thesuppliersreward,with
.
Exercise
Derivetheoptimalorderquantity usingtheclassiccost function inwhich andthenderivetheoptimalorderquantityusingthelinearisedannuitystreamfunction.Weillustratetheprocedurefortheclassicfunction.Weneedtominimise:
12 bychoosingandoptimalintegervalue,say
.Thisvaluemustsatisfytwoconditions:
1 1Fromthefirstconditionwecanderive:
12 1 1 12 Hence
1 1 1
2
Or
12.Thisinequalityisaquadraticfunctionin.Thenonnegativerootisgivenby:
12 1 1 8
But the solution has to be integer: (=the largest integer not larger than always satisfies thequadraticinequality.Fromthesecondconditionwederivesimilarlythat:
1 2.Andwecanproceedasforthefirstconditiontoderivethesameasalwaysfeasible.Bothconditionstogetherimply .Hence:
12 1 1 8
-
7/23/2019 Integrated Logistic
28/150
Noteth
by subs
depend
2.8
E
We intr
producti
producti
Ifwe
thefollo
The
Setti
the timi
(Beullen
Henc
Setti
donot
d
tthederiva
itution of on.PQ
for
b
duce in th
oncostper
onlotsizef
assumetha
wingfigure(
lassiccostfu
gtheAncho
ng of cash
sandJansse
e, ,agtheAnch
erivethe
res
ionoftheo . The
tchdem
previous
roductand
rasinglepr
t thisisjustan
nctionisder
rPointatst
lows as in
s,2014):
dagainther
rPointatst
ulthere.
c(/produc (/order)R(products
timalfoeason is tha
ndJogl
odel a finitisthesetuductionrun.
s inthepre
example).
ivedinJogle
rtofsalesa
he previous
1 eistheextr
artofthefir
)
year)
(pr (cyc
Integrated Logi
rcanbet the suppli
ekar(19
production
pcostfora
.
iousmodel,
ar(1988):
sinthefigur
sections, p
2 2 termwith
stproductio
producer
ducts/orderletime)?
sticsderivedfrom
ers reward
8)
rate (witproductionr
the invento
1 eabove,an
roduces the
1 runwillpr
y(prodw(/pr
(pro
(/o
)?
thisresultf
is a constan
). Inun.Thedeci
y levelover
2
2
whenmaki
following li
2 .duceadiffe
cts/year)oduct)ucts/order)
der)
ortheclassic
t term that
this model,
sionvariable
timemay lo
gassumpti
nearised AS
2
rent
fun
28
function
does not
is theisthe
ok like in
nsabout
function
2
tion.We
-
7/23/2019 Integrated Logistic
29/150
Integrated Logistics
29
NotethatthemodelofJoglekarincorporatespreviouslyconsideredmodels:
For ,wegetGoyalsmodel; For 1 ,wegetMonahansmodel; For 1 and ,theEOQmodelwithbatchdemandofGrubbstrm.
Homework
Derivetheoptimalorderquantityusingtheclassiccostfunctioninwhich andthenderivetheoptimalorderquantityusingthelinearisedannuitystreamfunction.Comparethetworesults.2.9 Vendingmachine
WeillustratetheusefulnessoftheNPVframeworkforderivingtheprofitfunctionofavendingmachine
operator.Weconsiderasingleproductsoldatapriceinavendingmachineofcapacity.Theproducthasacostpriceforthevendor.Thevendorhasasetupcostfordeliveringabatchofproductstothevendingmachine.Uponthedeliveryofproductstothevendingmachine,theoperatorcollectsthecoins
of
the
customers.
Assume
a
constant
demand
rate
.
The inventory level over time follows the EOQ sawtooth pattern.The cashflows however differ from
thatintheEOQmodelandaregiveninthefigurebelow.
2 2
Theholdingcostforavendingmachineistobebasedon
,i.e.basedonthesumofcost
priceand
sales
price!
ThereasonwhythisresultdiffersfromtheEOQmodelisthatthecoinsputintoavendingmachinebycustomers iscashthat isnotyetaccessibletothevendoroperator.Onlyuponcollectionofthesecoinscanthevendorhaveaccesstothiscapitalforreinvestment. It isas ifthecustomersonlyexchangethe
cashwiththeoperatorthemomenttheoperatoremptiesthevendingmachinescoinsregister.
Itcanbeproventhattheoptimallotsizeis min , .ThisexampleillustratesthattheprofitfunctionintheNPVframeworkwilldependontheassumptions
wemake
about
when
cash
is
exchanged.
We
call
this
the
payment
structure
and
it
is
further
discussed
in
thenextsection.
Time
Cashflows
0
s+wyT
T
....
....
pyT
-
7/23/2019 Integrated Logistic
30/150
2.10
P
Conside
product
thatthe
If th
Convent
buyerp
Ther
The
transact
We
ignored.
Itisnot
incases
EOQm
Assume
time.assume
ayments
thetransfe
thatthebu
totalamoun
paymento
ional (C). It
ysatdiffere
CashInAdv ,withCredit(CR): 0arefurther
Paymentsin
with0 atthesuppli
Transaction
instrument
theamount
igurebelow
ion costs a.
illhencefor
Allpayment
difficulttoa
wheretheti
delwithCR
thatallcust
Thecashfl
thatthefirst
tructure
rofabatch
erhastopa
t
ispaid
ccurs in full
isnot theo
tmoments
ance (CIA): 0 .thebuyerca
orethefoll
instalment1,asCIA(ter),andthe
costsandtr
sed,thesup
andtimethe
illustrates.T
d delays.
thassume t
sfurthercon
aptallprevi
eisspeciandCIApa
merspayw
wdiagram i
deliverywill
ofproducythesuppli
orthemom
at the time
ly reasonab
intime,inclu
hebuyerpa
paythesu
owingtwoc
.The
buyer
hiscouldbe
emainder1nsactiondel
pliermayre
buyerhasm
hepayment
e call the
at transacti
sideredare
ouslyconsid
ied.Anexa
mentstru
ithaCRofti
sgivenbelo
occuratsom
Integrated Logi
tsbetween
r. Inthepr
ent
thatth
ofdelivery
leassumpti
ding:
ysat some
plierlatert
nsideration
maypay,
fo
adepositpai wiays.Duetoi
ceiveadiffe
adeapaym
oftheamo
first two p
oncostsan
enceredmodels
plefollows.
tures
mebutt.Sincewe
efuturetim
sticsbuyerand
vioussectio
batcharriv
then it is s
nwecanm
timebefore
anthetime
:
example,a
dthemome
haCRarran
nefficiencies
entamount
nt.
nt isCIA,yments
delaysare.forsituation
atyouhave
havetopay
ewith
asupplier.L
ns,wehave
satthebuy
aid that the
ake. In reali
thedelivery
ofdelivery,s
fractionof
t
twhenthe
gement.
andcostsch
andatadiff
ofisC,a an
zeroorso s
swherepay
topaythe
thesupplier.
etbetheconsistently
er.
payment st
y, itmaybe
ismade, sa
ayattime
heamount
buyerplaces
argedbythe
erenttimer
dofisCd the third
all that th
entsoccur
upplierwith
inadvance,
30
priceper
assumed
ucture is
thatthe
at time
,with
ue,hisorder
financial
lativeto
andhas
payment
ycanbe
CIAorCR
CIAwith
wemust
-
7/23/2019 Integrated Logistic
31/150
Integrated Logistics
31
Thisgives:
2 2
2 Wealsoget: 2 2 and
2 2
Note.ItiseasytoseethatifcustomerswouldpayCIAwecanusetheabovemodelbutconsidernegative
valuesfor.Likewise,wecanstudytheimpactofreceivingacreditperiodfromthesupplierbytakingintheabovemodelnegativevaluesfor.Numericalexample
The table below illustrates the impact for an example with 25/, 150/, 2000 /, 0 and 50/. We take 0.2.
The%gap isacommonmeasureforgetting insightinrelativedifferencesbetweenascenarioanda
base case scenario. In the table it is calculated as the percentage difference relative to the base case
scenarioofaconventionalpaymentofthesupplieri.e.for 0.The%gapformeasureis:% 100 00 ,where in the above table is, respectively, , , and. It can be observed that the increase in
logisticscosts ismuchsmaller thanthecorrespondingprofit loss.Forassessing the impactofdifferent
timingsofpaymentsitishencemuchsafertoconsidertheprofitfunction.
Time
Cashflows
0
s
T
py ....
....
L
wyT
-
7/23/2019 Integrated Logistic
32/150
Integrated Logistics
32
(months)
(products/order)
% gap
(-)
(/year)
% gap
(-)
(/year)
% gap
(-)
0 346 0.00 1732 0.00 48253 0.00
1 344 0.83 1747 0.84 47398 1.77
2 341 1.65 1761 1.68 46529 3.57
3 338 2.47 1776 2.53 45646 5.40
4 335 3.28 1791 3.39 44747 7.27
5 332 4.08 1806 4.25 43834 9.16
6 330 4.88 1821 5.13 42906 11.08
2.11Consignment
arrangements
Consignment arrangements are popular payment structures in some industries between suppliers and
buyers.Thesebuyersarecompaniesthemselvesandmayberetailersorproductioncompanies.Thestockheld at a buyer under this arrangement is called the consignment stock. It is hence inventory that is
physicallyheldatthepremisesofthebuyer,butfinanciallyitisstillunderthe(partial)ownershipofthe
supplier.Onlywhenaproductisremovedfromthisconsignmentstock,willthebuyerhavetocompletethepaymentfortheproducttothesupplier.
Assumethatthepriceforaproductthatabuyerneedstopaytothesupplieris.Wecandistinguishbetweenthefollowingthreecommonconsignmentarrangements:
FullConsignment(FC):thesupplierretainsownershipofthe inventoryatthebuyerandthis isimplementedbylettingthebuyerpaysthesupplierthepriceforaproductonlyatthemomentthatthisproductisactuallytakenoutoftheconsignmentstockatthebuyer.Thisproductisthentobeused inthebuyersproductionprocess,orinthecaseofthebuyerbeingaretailer,when
thebuyeractuallysellstheproducttooneofitsowncustomers.
PartialConsignment(PC):thesupplierispaidanamountforeachproductwhentheproductisdelivered to the buyers consignment inventory, but the remainder is onlypaid out themoment the product is taken out of the consignment stock by the buyer (for reasons asexplained above for the case of full consignment). The price
can in principle be any agreed
amount;itdoesnotneedtobethesuppliersowncostprice.
Graceperiod(GP(z)):ThisissomewhatsimilartoPCinthatthebuyermayfirstpayanamountthemomentthatproducts aredelivered to itsconsignmentstock,but the remainder isthentobepaidbackaccordingtoacreditarrangementthatdependsontheaveragecycletimebetweendeliveries.Thiscreditperiodiscalledthegraceperiodanditsmomentofpaymentisingeneralasfollows: ,where
isasuitablychosenconstant.The idea isthatwhenabuyerorders in larger lotsizes,
thatthecreditperiodthesupplieriswillingtoofferalsobecomeslonger.
Weillustratethethreepaymentstructureswiththefollowingexample.
-
7/23/2019 Integrated Logistic
33/150
Integrated Logistics
33
Abuyersuppliermodelwithconsignmentarrangements
WeconsiderabuyerhavinganEOQproblemandthesupplierdeliveringtothisbuyerthentohavethe
EOQproblemwithbatchdemand.Bothmodelswerediscussedpreviously.TheASfunctionofthebuyer,
underconventionalpaymentstructures,was:
2 2,wherethesubscriptisusedtomakeclearthatitconcernsthebuyer.ThesuppliersASfunctionwas: 2 1 2 2.
Wenowintroduceageneralisedpaymentstructurewhichhastheabovediscussedthreevariationsof
consignmentinit:
Anamount
ispaidforaproductwhenitisdeliveredtothebuyer;
Anamountispaidforaproductwhenthebuyersellstheproducttoitsowncustomers; The remainder is paid for with a grace period time units after thedeliveryoftheproducttothebuyer.Hence, for 0 , 0 and this corresponds to FC; for 0 but 0 this
correspondstoPC;andwith 0but 0 toGP(z).Themoregeneralcasehasallthreeamountsnonzero.Notethat .
CustomersofthebuyerstillpayaccordingtotheconventionalstructureC,asintheEOQmodel.
ThebuyersASfunctionchangesto:
Welinearisetheexponentialtermsinthedecisionvariableandfind:
2 1 2 2 Note that the holding cost for the buyer is based on
1 2
. For
, the
holdingcostforthe
componentiszero.For
1,theholdingcostforthe
componentbecomesa
negativecost
i.e.
will
increase
the
buyers
profit
function.
ThesuppliersASfunctionchangesto:
2 1 2 1 2 2.
Thesuppliersrewardisaffectedandnowtobebasedon
1 2 .
-
7/23/2019 Integrated Logistic
34/150
Integrated Logistics
34
2.12Theprofitfunctionoftheintegratedsupplychain
If firms in a supply chain want to identify the optimal way of coordinating, then a commonly used
benchmarkisassumingthatthefirmswouldoperateasonevirtualorganisation.Thisvirtualorganisationrepresents the integratedsupplychain.Toderive itsoptimalpolicy,wearguablyneedtohaveaprofit
function
as
well.
We
illustrate
with
an
example
what
can
be
done,
and
where
there
are
still
some
open
problems.
Consider the buyersupplier model from the previous section, but assume conventional paymentstructures.Theprofitfunctionsofbothfirmsare,respectively:
2 2, 2 1 2 2.Itistemptingtopostulatethattheprofitfunctionoftheintegratedsupplychainisfoundastheirsum:
Thisgives:
2 1 2 2However,thisfunctionnowmakesuseoftwodifferentopportunitycostsofcapital,and.Would
an integrated firm not be in a position to use one common opportunity cost of capital instead? If wewouldassumethatbothcapitalcostratesarereplacedbyonecommonrate
wewouldfind:
2 1 2 2 2 2Sucharesultwouldalsobefoundifwestartedfromthecashflowdiagramsofbothfirms,andthen
recognisingthatcashexchangedbetweenthefirmswouldcanceleachotherout.
Thisconundrumhasnotyetbeenadequatelyaddressedintheliterature.Wewillassumehenceforththat
; in this case the problem does not present itself. Under this assumption, we can
indeedsumtheprofitfunctionsof individualfirmsto findtheprofitfunctionoftheir integratedsupply
chain.
Exercise1
Derivefortheabovebuyersuppliermodelwithconventionalpaymentstructurestheorderquantitiesandthatwillmaximise.
Wetakethepartialderivativeof, toforaconstant,andfromsettingthistozerowefind:
2
Substitutionofthisresultinto :
-
7/23/2019 Integrated Logistic
35/150
Integrated Logistics
35
2 2
2 2 Theoptimalpositiveintegervalueishence 1.Therefore,wealsoget2
and
2 2 .Exercise2Considerthebuyersuppliermodelwithconsignmentpaymentstructuresderivedpreviously.Repeattheaboveanalysisforthiscase,i.e:
a. Findtheprofitfunctionoftheintegratedsupplychain;
b. Findthevaluesofandthatmaximisethisfunction.Youwillfindthattheprofitfunctionoftheintegratedsupplychainisexactlythesame,andtherefore
also the optimal values derived above apply. This gives an important insight: the payment structuresbetweenthefirmsinasupplychaindonotaffecttheirintegratedprofitfunction!
Homework
Considerthesetofmodelswehave lookedatsofar,andmakeupyourownfeasiblecombinationofa
singlesuppliersinglebuyermodelandrepeattheaboveanalysis,i.e.determinetheintegratedsupply
chainprofitfunction,andderivetheoptimalpolicythatwillmaximisethisfunction.
2.13UsingtheNPVframeworktoincludeothercostcomponents
TheNPVframeworkalsoallowsfortheconsiderationofothercostcomponents.Weillustratewithafew
examples,usingtheEOQmodelasourbasecasescenario.
Materialhandlingandinsurance
A variable material handling cost e (/product) for each product placed in inventory, paid uponreceivingthebatch
Acostdirectlyproportionaltotheaverageinventorylevel(e.g.thecostofinsuranceagainstfirefortheaverageamountofinventoryheld),tobepaidase.g.acontinuousstreamattherateof /2(/year).
TheASfunctionfortheEOQmodelisadaptedto:
2 2 1
-
7/23/2019 Integrated Logistic
36/150
Integrated Logistics
36
2 1 2
2
2
Thusthe
optimal
lot
size
becomes:
2 andtheholdingcostsaretobecalculatedasfollows: Ingeneral,wecanhenceidentifytwodifferentcomponentsofholdingcosts:
a. The
financial
cost
of
keeping
stock,
arising
from
investments
made
into
products
stored,
here
the
term ;b. Outofpocketholdingcosts,arisingasrealexpensestobemadewhenplacingproductsinstock;here.Notethatthesecostsneedtobeafunctionofthestockposition i.e.thecorrespondingannualcostneedstovarywith.
It isgenerallyaccepted that inmanypracticalsituations the financialholdingcostsaremuch largerthan theoutofpocketholdingcosts.That isalsowhy inmanymodelsoutofpocketholdingcostsare
simplyignored.
Priceelasticdemandfunctions
Finaldemand
foraproductcouldbeafunctionofprice
.Forexample,onepossible function is (for
0): where and arepositiveconstants. If the retailer with an EOQ model could decideon theprice,thenthereisthepracticalconstraint: asotherwisetheretailerwouldnotmakeanyprofits.TheretailerthushastheASprofitfunction:
, 2 2
Onewaytosolvethisproblemistakingpartialderivativestoandandsolvethenonlinearsystemoftwoequations.Amorepragmaticapproachwouldbetoapplythefollowingalgorithm:1. Input:Valuesfor, ,,andafunctionwithmaximumprice2. Forasequenceofprices , , 2, ,
2.1.Determine2.2.Usethistocalculate 2
2.3.Calculate
,
3. Retainthatpricethatgivesthehighestvaluefor ,.
-
7/23/2019 Integrated Logistic
37/150
Integrated Logistics
37
Creditelasticdemandfunctions
Sometimes firms will offer a delay of payments in order to boost demand. In the simplest case, the
completeamounthastobepaidfortimeunitsafterthepurchase.Assumethatthefunctionisknown.TheASfunctionfortheretailernowbecomes:
1 , 1 2 , 2 2 TofindtheoptimalvaluesforLandQ,thefollowingpragmaticalgorithmcouldbeused:
1. Input:Valuesfor, ,,andafunctionwithmaximumcreditperiod.2. Forasequenceofdelayvalues
0, , 2, ,
2.1.Determine
2.2.Usethistocalculate 2 2.3.Calculate ,
3. RetainthatLthatgivesthehighestvaluefor ,.References
GRUBBSTRM,R.W.1980.APRINCIPLEFORDETERMININGTHECORRECTCAPITALCOSTSOFWORKINPROGRESSANDINVENTORY.INT.J.OFPRODUCTIONECONOMICS18(2),259271.
BEULLENS, P. AND JANSSENS, G.K. 2011. HOLDING COSTS UNDER PUSH OR PULL CONDITIONS THEIMPACTOFTHEANCHORPOINT.EUROPEANJOURNALOFOPERATIONALRESEARCH215,(1),115125.
BEULLENS, P. 2014. REVISITING FOUNDATIONS IN LOT SIZING CONNECTIONS BETWEEN HARRIS,
CROWTHER,MONAHAN,ANDCLARK.INT.J.OFPRODUCTIONECONOMICS155,6881.
BEULLENS, P. AND JANSSENS, G.K. 2014. ADAPTING INVENTORY MODELS FOR HANDLING VARIOUS
PAYMENT STRUCTURES USING NET PRESENT VALUE EQUIVALENCE ANALYSIS. INT. J. OF PRODUCTIONECONOMICS157,190200.
-
7/23/2019 Integrated Logistic
38/150
Integrated Logistics
38
3 IL in Buyer-Supplier Supply Chains
3.1 Onetooneshipping
ThisstrategyinvolvesshippingproductsfromalocationAtoalocationB,bylettingavehicle(truck,taxi,
train,ship,airplane)takingafeasibleandoptimalroute,typicallythecheapestorfastest.Let us assume that the vehicle is dedicated to the transaction, i.e. during its trip it will not make
detours or perform other transportation duties. It is then also called Direct Shipping (DS) or linehaul
shipping.
Figure 5: One-to-one shipping (Direct shipping, line-haul shipping, FTL shipping)
Thetransittimeisthetimethegoodsareunderwayfromthesourcetothedestination.Itisingeneralthesmallestrelativetoothershippingstrategies(discussedlater).
Becauseavehicleisdedicatedtothissingletransportationjob,itis ingeneralonlyjustifiablefroma
costperspectivetotransportsmallamountswhentheproducttransportedhasahighvaluedensity(/m3
or/kg)orwhenitneedstoarrivefast(emergencyshipment).Itisnotunusual,forexample,todispatcha
privatejetorhelicopterforthetransferofhumantransplantorgansinordertosavealife,ortodelivera
singlesparepartbytaxiinordertopreventexcessivedelaysforapassengerairflight.
Forgoods
with
lower
value
density,
direct
shipping
is
only
cost
effective
when
avehicle
can
transport
a large enough amount. Therefore this strategy is sometimes also referred to as FullTruckLoad (FTL)shipping,althoughitisnotrestrictedtoroadtransport,norisitalwaysoptimaltoshipinquantitiesthatfillupthevehiclescapacity.
3.1.1 ShortestPathProblem
FindingtheoptimalroutefromAtoBcanbemodelledasaShortestPathProblem(SPP).TheSPPcallsfor
findingtheshortestpathfromanoriginnodetoadestinationnodeinaconnecteddirectedgraphG=(N,
A)withnodesetNandarcsetAandwhereeveryarc aAhasanonnegative length.ShortestpathproblemscanbeefficientlysolvedusingDijkstrasalgorithm.
Figure 6: A shortest path problem
1
2
6
4
53
4
33
3
2
2
2
AB
-
7/23/2019 Integrated Logistic
39/150
Integrated Logistics
39
Figure 6 showsan exampleof ashortest path problem. There are sixnodes in thegraph numbered
from1to6,andsevenarcs,whereeacharcsdirectionandlengthisalsoindicated. Theheadofanarcis
thenodeadjacenttothearcsarrowhead,andtheotheradjacentnodeofanarciscalledthearcstail.
DijkstrasAlgorithm
Tofindtheshortestpathfrom insuchgraphfromsomeoriginnodetosomedestinationnode,wecan
useDijkstras
algorithm:
1. Associatewithallnodesatemporarylabelwithvalue2. Startattheoriginnodebychangingitslabelto0andmakethelabelpermanent.Callit
thecurrentnode.3. Foreveryarcgoingoutofthecurrentnodethathasaheadnodewithatemporarylabel,
replacethevalueofthetemporarylabelofthisheadnodewiththevalue:
min ,
4. Amongalltemporarylabellednodes,selectonewiththesmallestlabelvalueandmakethe labelpermanent. If thisnode is thedestinationnode,stop,elsecall thisnode thecurrentnodeandgobacktoStep3(iterate).
Theoptimaltotallengthisnowequaltothelabelvalueofthedestinationnode.Todetermine
theoptimalpath,startatthedestinationnodeandworkbackwards inthegraphbyfinding
thearcwhichcostcorrespondstothedifferenceofthelabelsofitsheadandtailnodes.There
maybemorethanoneoptimalpath.
AppliedtotheprobleminFigure6,thealgorithmgivesthefollowingsequenceoflabelvaluesforeach
node
(note
that
permanent
label
values
are
indicated
with
a*,
and
the
position
in
the
sequence
correspondstothenodenumber):
[ ][0* ][0* 4 3 ][0* 4 3* ][0* 4 3* 6 ][0* 4* 3* 6 ][0* 4* 3* 7 6 ][0* 4* 3* 7 6* ]
[0*
4*
3*
7
6*
8][0* 4* 3* 7* 6* 8]
[0* 4* 3* 7* 6* 8*]
Thelengthoftheoptimalpathisthus8.Theoptimalpathisderivedstartingfromnode6andgoing
backtonode5sincethelength2ofthatarcisequaltothedifferenceofitsheadlabelandtaillabels,2=
86.Itistheneitherpossibletogotonode2,sincethearcslength2=64,ortonode3sinceitsarcslength3=63.Fromnode2wegobacktonode1.Fromnode3wewouldalsogobacktonode1.There
arethustwooptimalpaths:eitherthesequenceofnodes1256orthesequenceofnodes1356.
Note that all nodes in this examples have permanent labels at the end. This is not always so; ingeneraltherecouldbenodesthatstillcarrytemporarylabels.Notealsothatthealgorithmstillworksif
theretherearemultiplearcshavingthesametailandheadnodes,aslongasweselectinStep3ofthe
algorithmthe
shortest
arc
in
the
formula.
Dijkstras
Algorithm
-
7/23/2019 Integrated Logistic
40/150
Integrated Logistics
40
Thealgorithmfindstheoptimalsolutionsinceeverytimewemakealabelofsomenodejpermanent,
wewillhavefoundtheshortestpathfromtheoriginnodetothatnodejanditspermanentlabelvalueis
the length of this optimal path. The optimality of this path from origin toj is not depending on any
decisionsweneedtomake lateron inthealgorithm,andviceversa: iftheoptimalpathfromoriginto
destinationwouldpassnodej,thentheoptimalpathfromtheorigintojwillbecompletelypartofthe
optimalsolutionindependentofthedecisionmadeinthepathfromjtothedestination.Wecallthisthe
principleofoptimality.(Wewillfurtherdiscussconditionsunderwhichthisprincipleisnolongertrueandthereforethealgorithmwouldnotbeapplicable.)
Differentobjectivefunctions
Differentobjectivefunctionscanbeusedminimisingtotallength,timeorcostbylettingthelengthofeacharcintheSPPcorrespondtoitsdistance,expectedtransporttimetocrossthearc,ortotalcostto
useit,respectively.Itiseasytoincorporateanyfixedcostsorfixeddelaysencounteredonarcs,suchas
ontollroads,attollbridges,andattolltunnels,oratbordercrossings(administration,bordercontrol).
Tachographlegislation
Fortrucktransportoverlongerdistances,attentionhastobepaidtothesocalledtachographlegislation
which requires drivers to take breaks and limits the total number of driving hours per day. Typicalconstraints
may
include
the
following:
1. nomorethan2hoursofconsecutivedrivingisallowed;2. 45minutesofresttimeneedstobetakeneitherafter2hoursofconstantdrivingorduringthe2
hoursofdriving(inbreaksofminimum15minuteseach);
3. totalnumberofdrivinghoursperdriveranddayisrestrictedto8hours.
Companiesmayhavethechoicebetweenusingtwodriversreducingtotalroutetimeorusingasingledriver with longer total travel time. To minimise to total costs, one can run the SPP algorithm on the
graphusingonlytraveltimerelatedcostsforonedriverandthen,alsoknowingthetotaltraveltime,add
thecostsoftheextradriver,requiredbreaks,andresttimestofindthetotalcostandrealtotaltime.The
optionthat
is
the
cheapest
can
then
be
retained.
Timedependenttraveltimes
Theproblemofminimisingtotaltraveltimeingraphswithtimedependentexpectedtraveltimesonarcs
can be adequately modelled as a SPP if the start time at the origin node is given. The algorithm now
needstouseinStep3theexpectedtraveltimeofanarcbasedonthecalculatedexpectedarrivaltimeat
itstailnode.
Undertheassumptionthatvehiclesthatarrivelateratthetailnodeofanarccanneverarriveearlierattheheadofthearcthenvehiclesofthesametypethatarrivedearlieratthetailnode,thealgorithmfindstheoptimalpath.Theassumption is ingeneralrealistic forqueue induced traveldelaysonroadssincevehiclesofsometypeattheendofaqueuewillfind itverydifficulttobeatvehiclesofthesame
typeatthetopofthequeue.Theassumption isknownasthenoovertakingproperty.Notethatthis
propertyimplies
that
waiting
at
any
node
in
the
graph
is
never
optimal.
This
approach
also
works
for
when a desired arrival time at the destination node is given by letting the algorithm start at the
destinationnode,workingbacktotheoriginnode,andsubtractingtimesduringthesearch.Tofindminimumcostsolutionsfortimedependenttraveltimesforwhichthenoovertakingproperty
holds, the travel cost must be monotone increasing with travel time and a given units of driving timemustcostthesameasthesameunitsofwaitingtimeonnodes.Inthatcasethere isalwaysanoptimalsolutioninwhichnowaitingoccurs,andthusthealgorithmwillfindanoptimalsolution.
If waiting (i.e. resting) is less costly than driving, it may be optimal to wait for times with lesscongestion on roads in order to minimise costs. However, waiting at some tail node means that
opportunities for faster driving may be lost on arcs closer to the destination. Thus, the principle of
optimalitynolongerholdsandotheralgorithmsaretobeused.
-
7/23/2019 Integrated Logistic
41/150
Integrated Logistics
41
Multimodeltransportation
MultimodaltransportationproblemscanbemodelledasSPPsbyassociatingwitheacharcinthegraph
thedatarelevantforaspecificmodeandvehicletype;multimodeltransfernodesareintroducedinthe
graphwhichconnect thegraphsof thedifferent transportationmodes.Each transfernode willadd its
transfercostorexpectedtimeoftransfertotherelevantarcleavingthenode.ThenormalSPPalgorithmwillthenbeapplicable.
However,theremaybeotherconstraintsinpracticemakingthisapproachlessrealistic.Ships,trains,andairplanestypicallytravelaccordingtoprespecifiedschedulesandroutesandsomeshipsortrainsare
cheaperthanothers. Itmaythusbecheapertowaitatatransfernodeforthecheapestvehicle(train,
ship)orroute.Thishowever,violatestheprincipleofoptimality.
Note. For calculating SPP on road networks, various internetbased resources can now be used. For
example,GooglemapshasafunctiontoallowyoutoseekforthequickestroutefromAtoB,whereyoucanspecifyyourstartingtimeandwhichtakesintoaccounttimedependenttraveltimes.
3.1.2 EconomicTransportQuantity(ETQ)
Considerthesituationthatabuyerneedsregularsupplyofsomegood.Weconstructacostmodelforthe
situationof
direct
shipping
of
the
product
from
asupplier
to
this
buyer.
We
assume
that
the
optimal
travelroutehasbeendetermined(ase.g.anSPP)andthatitscostandtotaltraveltimeareindependent
ofthetimeoftheyearatwhichitisundertaken.Wedenotebythetotaltransittime.WeassumeabuyerwithanEOQmodelandasupplierproducinglotforlotatafiniteproductionrate.
Wenowalsoconsidertheintermediatestagewheretheproductsareonavehicleintransit.WeplacetheAnchorPointatthedeliveryofthefirstbatchtothebuyeratarbitrarytime.
Time
Inventorysupplier
0
T
....
yT/R
Time
Inventoryintransit
0
T
....
L
Time
Inventorybuyer
0
T
....
L
-
7/23/2019 Integrated Logistic
42/150
Integrated Logistics
42
Figure 7: Inventory positions and cash-flows in the direct shipping model
Weassumethatthetransport isundertakenbya3PLandthatthebuyerpaysthiscompanyafixed
transportcostpershipmentandavariabletransportcostthatdependsonthelotsizebeingshipped.The
buyerpaysthesupplierfortheproductsthemomentthatabatch isdelivered.Weassumethatnexttosetupcostsforloadingandunloadingthevehicle,the3PLalsoincursatransportcostatarateforthedurationofthejourney.SeealsoFigure7.Thecapacityofthevehicleis
.
Weproceed
by
deriving
the
AS
profit
functions
of
the
three
firms
involved
from
their
cash
flow
functions.Forthebuyer,wehave:
2 2
Forthe3PL,wefind:
1
Time
Cashflowssupplier
0
s
T
wyT
....
yT/R
cR
L
Time
Cashflows
3PL
0
T
....
L
Time
Cashflowsbuyer
0
T
....
L
-
7/23/2019 Integrated Logistic
43/150
Integrated Logistics
43
Define: 1Then:
2 2 Forthesupplier,finally,wederive:
1 Thisleadsto,aftersomealgebraicmanipulation:
1 2 2 2 2 Optimalthreefirmsolution
Ifthethreefirmsareinterestedindeterminingtheoptimalshippingstrategyfortheirintegratedsupply
chain, we can derive this from their supply chain profit function. This function is found from the
summationoftheirprofitfunctions.(Asalways,assuming
).Thisproduces:
Define: 1 1 1 2
Then:
2 1 2 Thisproducesanoptimalunrestrictedlotsize:
2 1 However,sincethevehiclecapacityis
,theoptimalfeasibleEconomicTransportQuantity(ETQ)is:
min,
-
7/23/2019 Integrated Logistic
44/150
Integrated Logistics
44
DeterminingthevalueofTheparameterisanannuitystreamcostrateandhenceneedstobeexpressedin(/year).Weshowhowtoincorporatetworelevantcomponentsincasethatthevehicleisaroadvehicle:driverwagesand
fuel costs. Driver wages are typically given in (/hr), so if a driver costs (/hr), then it has to beconvertedtoacostrate
(/year)asfollows:
8760 ,since there are approximately 24365 8760 hours in one year (i.e. not counting years with anextradayinFebruary).Fuelcostsaretypicallyexpressedin(/km).Therefore,afuelcost(/km)hastobeconvertedintoarate(/year)byassuminganaveragespeedofthevehicleof(km/hr): 8760 .
Ifthesearetheonlyrelevantcosts,itwouldhenceproduce .Exercise
Determinethe
optimal
lot
size
when
the
buyer
would
independently
be
able
to
determine
.Inthatcase,thebuyerwouldaimtomaximise.Thisproduces min , where:2
Note.Observethattheunrestrictedoptimallotsizefortheintegratedthreefirmsolutionisafunctionof
thetransitleadtime.However,fortheabovesolutionforthebuyeritisindependentofthisleadtime.
Optimalsolutionwhentransportisoutsourced
If
the
3PL
works
independently
and
the
supplier
and
buyer
want
to
determine
the
optimal
shipping
strategyforthemselvesastwofirms,wecanderivethisfrom: Redefine:
1 2
Then:
2 1 2 Thisproducesanoptimalunrestrictedlotsize:
2 1
However,
since
the
vehicle
capacity
is
,theoptimalfeasibleresultis min , .
-
7/23/2019 Integrated Logistic
45/150
Integrated Logistics
45
Note
Ininternationaltransportbyseaitisoftenthatcasethatbuyerandsupplierneedtorelyona3PLforthe
shipping.Here, couldbe themaximum load of theproduct intoonecontainer.The 3PLmay chargedifferentratesbetweenshippingafullcontainerversusshippingafractionofacontainerload.
Homework
Derivethe
optimal
lot
size
when:
(1)
buyer
and
3PL
would
seek
to
find
their
integrated
optimal
solution;
(2)supplierand3PLwouldseektofindtheirintegratedoptimalsolution.
Note
Thesituationoflotforlotataninfiniteratecanberetrievedfromtheabovefunctionsbyconsideringthe
case .ThecasethatthesupplierproducesaccordingtoanEOQwithbatchdemandisretrievedbysetting .3.1.3 MaximumEconomicHaulageRadius(MEHR)
Thereisalimitastohowfaronecanreasonablytransportacertaintypeofproductusingdirectshipping.Ingeneral,thehigherthevaluedensityofaproductthefurtheritcanbetransportedinsmallquantities.
This
makes
it
reasonable
to
use
dedicated
small
transport
vehicles
such
as
small
vans,
taxis,
or
expensive
vehiclessuchasaeroplanes.Thelowerthevaluedensityofaproduct,theshorterthedistanceoverwhichitcanbeeconomically
transported.Tocoverlongerdistanceswouldneedshipmentsinlargequantitiessuchaslargetruckswithasecondtrailerorbulktransportinabarge(inlandwaterways),orcontainertransportontrains(rail)orships (sea and ocean transport) where transport costs can be shared with other goods from other
companies.
UsingtheprofitfunctiontoderivetheMEHR
KnowledgeoftheASprofitfunctionallowsustocalculatethemaximumdistanceoverwhichaproduct
can be transported. Since
, only strictly positive value for
will also produce strictly
positivevaluesforNPV.Asanyproject isnotconsideredworthwhilebyacompany ifitsrespectiveNPV
wouldbecome
negative,
the
boundary
condition
would
be
that:
, 0Note that when, 0, this doesnot mean that the company would not produce a positive
profitinaccountingterms.Itsimplymeansthatitwouldnotproducemoreprofitthanfromthenextbest
available alternative! If 0.20 then the firm would still gain a respective 20% of profits from thisactivity.
Wesketch theapproachwith the followingexample.Take the threefirm integratedprofit functionderivedpreviously:
, 2 1 2 It is sufficient to focus on the function in between the square brackets. Furthermore, we can
substituteintothisfunction,producingtheboundarycondition: 2 1 2 0,
where
1
-
7/23/2019 Integrated Logistic
46/150
Wen
CaseThelot
types of
function
wher
The
express
CaseSubstitu
It m
practicalreduces
3.1.4
Figur
owneedto
.izeisinoth
terms in th
of.
eworkingo
righthand s
din(years).
tionofther
y be difficu
approach
tozero.(Sim
Optimalpo
e 8: Cross-D
Inbou
onsidertwo
rwordsac
e boundary
orkingoutt
ttheactual
ide of this
Toconvertt
.sultfor lt to derive
ouldbe
to
ilartotheal
licytoorde
cking Facilit
ndtransport
cases.
nstant.Itca
condition: t
hisboundary
aluesfor
inequality wistoadista
derivedp 2an analytic
sean
algori
orithmspre
rNdiffere
Cross
Integrated Logi
nbeobserv
erms that a
condition,
andisleft
1 ould be th
ce,younee
reviouslywo
2 result for th
thmwhere
sentedatth
titemsfro
ockingFacili
1
stics1 1
dfromthe
re not a fun
ewillhence
,asanexercis
maximum
dtoconsider
uldproduce
e maximum
ouwould
in
endofSecti
aCrossD
Outbound
ty
2
bovefuncti
ction of ,findaformu
totheread
economic h
thetypeof
heconditio
1 possible val
creaseunon2.13).
ockingFaci
transport
onsthatwe
and terms t
laofthesor
er.Thismea
aulage radi
ehicleused.
:
0lue of
. Ho
ilthe
right
lity(CDF)
2
46
havetwo
at are a
:
sthat:
s
wever, a
andside
-
7/23/2019 Integrated Logistic
47/150
Integrated Logistics
47
A CrossDocking Facility (CDF) is typically located near the boundaries of a populated area. It receives
goods in (large) vehicles from various suppliers; these incoming goods are separated and mixed as
required at the CDF, and subsequently sent out in vehicles without being held in storage to different
destinationsinthelocalarea.
Thecrossdockingoperationsmayrequirelargeareasinthewarehousewhereinboundmaterialsare
sorted,consolidated,andstoreduntiltheoutboundshipmentiscompleteandreadytoship.IfthistakesseveraldaysorevenweeksitisnotconsideredaCDFbutawarehouse.InmostCDFs,goodsdonotstaylongerthan48hours.
OptimalorderpolicyforNitemsreceivedfromaCDF
Considertheproblemwhereyourfirmrepresentsstockingpoint2inFigure8.Yourfirmhasademandfor
N different types of items (i = 1, ..., N). For each item the uniform annual demand rate of yourcustomers is yi and your cost price is . You order each of these items from the CDF. Each time avehiclefromtheCDFvisitsyourfirm,however,youhavetopayafixedtransportcost.Whatwouldbetheoptimallotsize fororderingeachitem?
Since
you
have
constant
annual
demand
for
each
item,
you
have
an
EOQ
type
problem
for
each
item.
YourASprofitfunctionforitem,wheneachitemisorderedseparately,isthenarguablyofthefollowingform: 2 2 For each delivery, the CDF charges a setup cost , and that is why in the above function wehave .Yourtotalprofitfunctionwouldthenbethesumoverallitems:
2
2
Couldtherebeabetterwayofordering?Isitperhapsworthwhiletoordersomeitemstogetherintoonetrip?Thiswouldsafeontransportationcharges.
Property.AnoptimalpolicycontainsschedulingperiodsTiofequallength,i.e.Ti=TforalliN.
Proof. Suppose the theorem does not hold. Let T designate the smallest scheduling period which
happenstobe for item k, i.e. T=TkTand letthe
averageinventorycarriedofthisitembeE(Ij) .Nowsupposewedecrease itsschedulingperiodtoTj =T.Theaverage inventorywilldecrease from
E(Ij) to E(Ij ). Since the replenishment cost is independent of the quantity ordered, no additional
replenishmentcost
is
incurred
for
replenishing
item
type
jevery
Tj
=Tunits
of
time.
Hence
the
global
costwilldecrease.
SinceanoptimalpolicycontainsperiodsTofequallength,wehave:
Andweget:
2 2
-
7/23/2019 Integrated Logistic
48/150
Integrated Logistics
48
Wefind:
2
1
And:
2 2
3.2 Onetomanyshipping
3.2.1 LengthofanoptimalTSPtourvisitingmanycustomers
Drawonapieceofpaperasquare.Callthisyourserviceregion.Indicatethexandyaxis(see
Figure9).Wearbitrarilytakethelengthofthesidesequalto1m,sothattheserviceregioncoversanarea
of1x1=1m2.
Figure 9: Randomly distr ibut ed points in a rectangle.
Nowdrawarandomnumberfromtheinterval[0,1]andcallthisx1.Drawasecondrandomnumberfrom the interval [0, 1] and call this y1. Use these coordinates to draw a point (x1, y1) in your serviceregion. Repeat this procedure.Nowyouwill haveasecond point locatedatsomecoordinates (x2, y2).
Continue until you have generated n points in your service region. Figure 9 shows the exercise at four
differentstages,i.e.forn=5,n=10,n=15andn=20.WecallthesetofnpointobtainedXn:
)},(),....,,(),,{( 2211 nnn
yxyxyxX
Drawatourthroughyournpoints,visitingeachpointonlyonceandreturningtothefirstcitywhere
youstartedasinFigure10.Notethateachtimeyouleaveapointyouhavetodecidewhichpointtogoto
next.Youcanthusconstructseveraldifferenttours,allhavingadifferenttotaldistance.Infact,thereare(n1)!/2differentpossibletoursthroughnpoints.
x
100
1
(x1,
y1) (x1,y1) (x1,
y1)(x1,
y1)
-
7/23/2019 Integrated Logistic
49/150
Integrated Logistics
49
Figure 10: A TSP tour t hrough 9 point s.
Suppose we are interested in that tour of which the total distance travelled is minimal. If n grows
large,the
number
of
possible
tours
becomes
excessively
large
and
we
cant
just
find
the
best
tour
by
tryingallpossibletoursandretainthebestone.
Wecallthisproblemof findingtheshortesttourthroughnpointsthe TravellingSalesmanProblem
(TSP).TheTSPisoneoftheclassicproblemsinOperationalResearch.
Withoutknowingtheoptimaltour itself, letuscallthe lengthoftheoptimalTSPtour T*(Xn) inourexampleofrandomlydistributedpointsinasquareofarea1.ThenBeardwoodetal.(1959)provedthat
whenyoumakenverylarge,theratioT*(Xn)/ nbecomesconstant.Inotherwords,forverylargen:
nXT n )(*
wheretheconstantisbelievedtobe0.7124.
Infact,thisalsoholdsforaserviceregionofmore irregularshapes(as inFigure11). IfA isthearea
(m2)ofaserviceregionofanyfiniteshape,then
AnXT n )(*
Figure 11: Locat ions randomly di st r i buted in a serv i ce region.
-
7/23/2019 Integrated Logistic
50/150
Integrated Logistics
50
Finally,theresultisevenmoregeneralthanthat.Thepointsdonotneedtobedrawnfromauniform
distributionacrosstheservicearea,anydistributionwilldo.TheTheoremofBeardwoodetal.(1959) is
giveninFigure12.
Theorem BBH. (Bear dwood et al . , 1959) . I f T*( Xn) i s
t he l engt h of t he opt i mal t r avel l i ng sal esman t ourt hrough n poi nt s whi ch ar e i ndependent l y dr awn f r om
an i dent i cal di str i but i on over a bounded regi on aof t he Eucl i dean pl ane, t hen t here exi st s a const ant
such t hat wi t h pr obabi l i t y one
dn
XT
a
n
n
)(lim
*,
wher e the i nt egr ati on i s wi t h r espect t o theLebesgue ( area) measure. For t he uni f ormdi str i but i on over [ 0, 1] 2, the i ntegrat i on term i sequal t o one.
Note:0.7124(Johnsonetal.,1996,PercusandMartin,1996).
Figure 12:Theorem of Beardw ood et al . (1959).
Example1
Aprintedcircuitboardof20cmby10cmneeds1000littleholesdrilled in it.Drilling isperformedbya
moving pin. What is the expected distance that the pen needs to travel? Assume that the holes areuniformlydistributedacrosstheboard.
UsingtheTheorem,wegetforA=200cm2,n=1000
)1000(20071.0)(* AnXT n 318cm
Example2
In his small van, an express courier has small packages destined for 250 customers located acrossHampshire.AssumethatHampshirecovers22500km
2andcustomersareuniformlydistributed.
a)Estimatethedistancetobetravelledtodeliverallmail.b)Giveareasonableestimateofhowlongitisgoingtotakethecouriertodeliverthemail.
c)Whatwillbetheresultofdividingtheworkupinfivevans,eachcoveringanequalpartofHampshire?
a)UsingtheTheorem,wegetforA=22500km2,n=250
)250(2250071.0)(* AnXT n 1690km
b)Takinganaveragedrivingspeedof50km/hour,totaltraveltime=1690km/(50km/hr)=33.8hours.Inaddition, itmaytakeourdriverat leastoneminute foreverycustomertomakethedelivery,which
addsanother250min=4.2hourstothetotaldeliverytime.
c)Thedistancetravelledbyeachdriver,sincenowA=22500/5km2andn=250/5
)5/250)(5/22500(71.0)(* AnXT n 1690/5=338km
Takinganaveragedrivingspeedof50km/hour,totaltraveltime=338km/(50km/hr)=6.76hours.A
drivernowvisitsabout50customers,whichshouldkeeptotalworkingtimeperdriverwithin8hours.
-
7/23/2019 Integrated Logistic
51/150
Integrated Logistics
51
3.2.2 ExpectedlengthofanoptimalTSPtourvisitingafewcustomers
ComputationaltestsforuniformlygeneratedrandomcustomersinasquareareafindthattheformulaofBeardwoodetal.alsoservesasagoodpredictionfortheexpectedoraverageoptimalTSPtour lengthevenwhennisrelativelysmall(Eilonetal.,1971,seealsoHaimovichandRinnooyKan,1985).
Figure 13: Few customers uniformly distributed in a square.
Figure13showsthreeexamples.Thefirstrowshowsfourdifferentinstancesforn=5.Thelengthof
theoptimalTSPtourinallfourwillbedifferent,say*iT (i =1,,4).Theaveragelength,however,canbe
expectedtobe
571.054
1 4
1
*
AAnTi
i
Example
A pizza takeaway also makes home deliveries using scooters. On average there are 10 home delivery
ordersper
hour,
randomly
located
in
a50
km2
service
region
located
around
the
pizza
restaurant.
a) Estimate the average total distance a scooter will travel to deliver 3 orders and return to therestaurant.
b) Estimate the average total time needed for a scooter to deliver 3 orders and return to the pizza
restaurant.c) Estimate the minimum number of scooters needed when a scooter will deliver on average to 3
customerspertrip.d)Estimatethetimewhenthescooterarrivesatthethirdcustomeronitstrip(relativetoitsstartingtimeattherestaurant).
a)With threecustomersplusthe restaurant, n=3+1,andA=50,the formulagivesusanestimated
averagetimeofvisitingthreecustomersinthebestsequenceastominimisetraveldistance:
-
7/23/2019 Integrated Logistic
52/150
Integrated Logistics
52
10)4(5071.0 An km
b)Assumingtheaveragespeedofascooterisaround35km/hour,totaltraveltimeis10/35=0.28hour=
17min.Assumingittakes4minutestodeliveratacustomerlocation,itwilltakeintotalonaverage17+
12=29minutesorabouthalfanhourtomakethedeliverytrip.
c)It
takes
one
scooter
half
an
hour
to
deliver
to
3customers
and
in
the
same
time
10/2
=5new
orders
arrive.Thereforeoneneedsminimum2scooters.(Inordertobeabletoperformalldeliverieswithonescooter,theaveragenumberofdeliveryordersneedstodropto6perhour,orsmaller.)d)Thetotaltraveltimeof17minpertrip isdividedover4legs,andthethirdcustomer isreachedat
theendofthethirdleg.Thescooter isthereforeexpectedtoarriveaftertravellingatotaltimeof
(17)=12.75min.However,wehavetoaddtothisthetime istakesfordeliveringthe firstandsecond
order,intotal8min.Thereforethetimeofarrivalatthethirdcustomerisestimatedtobe12.75+821min.afterleavingtherestaurant.
3.2.3 Continuousapproximationofanoptimalvehicleroutingsolution
Inphysicaldistribution,goodsneedtobedeliveredusingvehiclesoflimitedcapacity.Intheonetomany
shipping
mode,
a
vehicle
will
deliver
to
more
than
one
customer
during
one
trip.
In
the
socalled
Capacitated Vehicle Routing Problem (CVRP) (Dantzig and Ramser, 1959), all vehicles have the samecapacity.TheCVRPconsistsoffindingasetofvehicleroutesofminimumcostsuchthat:everycustomerisservicedexactlybyonevehicle,eachroutestartsandendsatthedepotandthetotaldemandservicedby a route does not exceed vehicle capacity. The CVRP, like TSP, is a classic problem in Operational
Research.
To find an optimal set of routes for the CVRP is a difficult problem that can require a lot of
computationaltimeforlargeproblems.Oneoftenmakesuseofheuristicsi.e.methodsthatcanfindina
relativelyshorttimeagood,butnotnecessarilyoptimal,solution.Good(nearoptimal)solutionstypicallylook like in Figure 14 a: the total service region is divided into a number of districts; and customersbelongingtoonedistrictareservedbyoneandthesamevehicle,asindicatedinFigure14b.
Agoodapproximationforthedistancetravelledbyavehicleservingadistrictiis(Daganzo,1984):
iiiii nAnrD )(2*
whereAiisthesizeofthedistrict, ir istheaveragedistancefromacustomerindistrictitothedepot,
ni isthenumberofcustomers inthedistrict,andthedimensionless factor(ni)dependsonthemetric
andthenumberofcustomersinthedistrict. Forni6andtheEuclideanmetric,(ni)isaconstant=0.57,andforsmallervalue