0% found this document useful (0 votes)
119 views31 pagesLPP PDF
Uploaded by
shivamCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
119 views31 pagesLPP PDF
Uploaded by
shivamCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 31
ee ee eee
v
ee Ee Ee eae eS eS ee eee eee eee
=o mona I= ‘Non- eget “a bvaly- hes
- _ Tidushrcal, Eng
By - eT Sage Gu @ ade case
Lineare Programming Bane CLE0)
~ Mow 2 = 62%. St > objective fate. =
Sub fect to: _
Mt
aes < al
ae
: Cy fe) as
Deaths Vouobhs
~ Papblam. because... Comrtramds —.
inction... becth. are Vnean |
_dewaton Notable - od a of 1 OU “Abe “wn, _ 4 pth = o
TG: The, Conner point ! that fos tee peat
~ | -fum cho), opal. a
lhe aE
Value Zs Le bjecice agli rate Reqiow. ‘
mer Ae points unt) Pay dhe Points ue the
a beciNs Bauer het Satddy. F fear 'dbte Acgion violates
all the towntramdr atleast . Our. Conn brant
| Waxounded, Branded :
tefeaaible negian _feamubels. Reqrew
se
Feasible Solution Defeasetle Selection
| Baar Ponoibte Bola Non- Baste aes .
efit ee a
fale
— ian ap Splichon
bint ope thaw SoluHou-
= | Mintnnze 2 = 97 +3 ‘t- _
“Ne OPTIMAL SOLUTION
— EE eee
Unbounded Solstuen. Srl Zalat
Subject to _ :
~--—- S44 % loo
3X24 % Igo
U4 Zo —
| brim" Bvaliach, chjective, oschon
He Corner Beunts .
pA toned as =20 re
— Ad (Boy >. $22) Be 1S gS pha i
A \
Gow) (4010) a
At oO; loo J = 39
Net Unbounded, Sam Aegrene wweh -
: Gusnaten” * unbounded: - BaShion. :
Binding towabrata Poe
Pt pptinal Aitatin baek te eaten
PEs hoo se 6 Sata
200 > loo o 20s [995
fr py alla pond 0 oan Cob 5
Binding. Cowat taint
om
BS
Aefrvad a4
PbS
Waximae Sa 2x43 ie
Subject to
but >, 100
3x eet Z12o
E _Lk in
Mont sje jae BSubjeck 4 mty oh
=3xt 4 23
y — Lg 72 : |
N = No_fraseble_regin- -
4 Hex Fre SOTT?
“foray —— fi OT
S - eo
mo (Con a 2a
Pb Maximise. SETAE —
Subject to gerry 24
« ety > 3 —
tiy Zo
Za jo a
p He ta 228
LEX At (03) ele .
(om Ak (om) e= 48 T) “opted
a 978) Ak (80) 2=48 mila
_o* me a
To have Mubtiate optimal Sojictiow
Oa tah hg chisin. racial he_equa-
Abe Cowshratat.
AO that thuatn aan” nok be. Agdundank
— Cows paint ' =
OTmat Comabaint WUutk be Binding, Cons paint Pb> Max 2 = Bx Hy f
Subject te _
Sx+ FH Ss S10" ant
a 4ut ft ras
Me Lg zo” —H
|] o * . on zs 0 at (070) é
pot aww 2 = 28, at [0 ¥\e
Ne [ a
| Note o ~Boundaies™ aE one
one. pes ee ie fa
Oma Panes “Dor. thyed du chtion Vadvnkleg
eS plical ~wathed Cann
[beet used. me ee thon tivo Aetiasie |
taal : . 3 Atgebrait. Meliod =>
Max 2=— xt sy
Subject to wey eS
= —\ 3x4 24 212.
| Convert - trequalit tea txto 29 waltbn. By eaad
Aroek
|
_
Lig Zo
Subject» tp =>
——4 = Lt 4S, 5 = ae
: 3h t+ Ty 4S 5120
iY SiS 7,0
Nite dhe. addition 2 : Jorratato thear
ae fou Varrdioley i du eq uations: —
scchionas we can. 'abve on
he Be Vartolees s Ao we shave. “fp pi
(leon wastes to zn Valu and we com
Caan (Chossen An Hey emmys
Won Basic Vanuaiig 1 Ths. Yorrobtes Ahak we -
. fie tesa Vode adie Called mon haste Vasuabbe
i Rar “guiikes Te vaiuablks Shak we
Aokve’ are Catled ay ‘basic Vasolates »
Banc Bofuton => The Solution. Maier 5
fist Won bare. variate ty Zero ane
l ar Beanie’ atution. Non Baric Variable | BarieVanatle ad paladin
DP s=o, S20 ter, Yes | Fes2pr4 fester
G50) KEo- Wasrarene oar,
Q Si Deo pe RS) Sa-3| 2 = —— é
Ql 20, K=0 fab, Se-)| Be —— ¢
CG S20, y=0- Key) Sis} | B= 24 €
6 Kao, Neo -| $)25, S:)2 | 220 4
Bame Arctiben whl Kali fies all the é
| Constrainta »e Knew ar arate’ ble
| Solotion (19,526) -¥
=P tin __Aatution (32 4)” dak. 5
dows, refi Valiats —vintater nogasy é
as A Yok
a ettigeae—efeeed aonb ¢
é
' | Newer oh Raace Kotohon = ¢
wheat - Vin = no: i vawahle- «
M= Yo. fe
_ : ti
AL 2 es |
da 5 det eg volition tts ea rtiona
black Janudides |
there Oe mM equatigns
Feta het (int)_tnonl baried p
Laual, 16 Pero and evabale
hae alata Ges ee ted yen |
[pales ka Abe, ee
te My
a mae 4e Sdendife ihe ophowum Lobdilon 0
Qe one : voeth the aa el noi (rain
ylang sBjecve. -Panelion
\
aligebrase na a. 7
Hs We shave Jo cvaluote alt the "ew
bane Aobitiens before we obtain the
|
|
1 The ing, One the Mwitetows dj tne
_| Opty mom +
del! Some Ae bare Solitons ae imfeoorble |
_| We Aeave "45 evaluate them also. _
3: Ano the bare Pearible 4o
we dott evalrote Y better and better
Jokutions Some dk La Aubsequont:
CValuotiaA base Y : Aduhow ean:
Rave. inferior value dk dbjeck ve. fenchon
shen, Compaard to the pel - Solution’
Shak We Theatre eid dag weld Sak
ad: Does mot evaluct. om baric trfearctale.
Bolucion «
in Pacpenively. ohtoiws better and Betta.
SB Pdentifies phe eplimom Aolution the
basic feadeble Aobichows +. Oat not _evaluoted» .
re
Be _aloowdthn “thot Tan all the above.
Caractere
Daag 194)
oo. | Westra the mcbhod [7 \
—|— Moxinure 2. 6x 5p _
| __ Aubject Ad ve .
M+ So"
pong Sete gin
—— = % G20
| Convert “above LPP tla Mandard-form *_
Goa a as 22 ae
- eri = Me ama eC pve Ape)
_munrk be. equation _
Toad von- megabive TRS value |
a LPP into “Alandand fon. _
vein Now” vauokles amd ! Caserhy
Asst
; ped do — get © meon- negative € pe
[Qycaar toe LV converted “usks Dj
| el, Bani female
2 suitipl Race
- ey toed x Y o al
= Bu rey +51 : G {
: xm Y $155 %0% a att
> ee
ot : -
> | ere t 31 a | Rus | 6H
> o 13 pat =z = z :
. { 1
: ae rol ZX 4 2 pe _[Pp=4t7
Gey CEs ° © Zao
E Jd Sa| fo: a)y| 2 -%_[ fl s>
a k{ ox 4 | (ah,/] 0 a3. é
eee | a to So | eeog
- fy, 1 oad ole! 3
| 6 x |4 0 | -a 4 5.
jt IQ -2y | #'i —3 A | 220
bs to Romomborn
@Odnrtomt Varuable or ey e Varwable =>.
The _Vantable th maximum’ value dk CE
Cfon Maxim|2atba type sbyechive fumchon) _
@ p= 4 “whew -b a RMS Valu
TE ef of hoo tn evbeadig
a Coluitin' ”
Cav’ Variable — The varialele voyth
Warn, Vabus. fs @:
thee will alas be Tdowtiky matrix.
eines ponding, Ute Basie Varihbles a _
tex ' talole ’
Gnd 2) vaso baat Vamasole ws Zor - —p We osbsewe Wet Lntreart an the objective _
sumetion Vebur. we ay iteracon: io oe —
roduc ds (G- (Vy Mie’ Cortes
biting, bose and minim @
1 Vareble :
Cokes lance. to Aca ving.
Sew = Cold Cin prevent
zy le = (G2;
oe :
OPTIMALITY ConpITION => The Ly
“so Vortakle in a macmization oben br by
| the mon= barie._vyoriale Roviiga eee
WE CoetH alent Ln Ga -% S Aer) 'e Ties
Taw barker | arhttart opti _
Ww Aeathed’ at the Gate luskew all
date GBs, (neth ata: spenens boaic| Vaile.
{04 ow
> Feagewity donoiTlan =p - fox To
tha xi wu 2echion ftv wi eaten, Problowna -
| the deavang.” VOruabbe, in the baie Vania
RB assotiatd? with the Jmatleat non- :
dotio -f With Abitdeg- chesttive denon y :
__| Te | pitas a
f= e; Whee Gye
if Go do nak Comput. 0 Gro) Gi) (6,0) 7
Note Jimphox wmeteod evaluat vals rx
. Bbechve funchiow_ oud at Abe Conn
J
nts the _ =feoaible _ Aeqieu.
—} “The three important steps (or stages) inthe simplex algorithm : a
1. Thealgorithm starts with abasic feasible solution (itialization. The Right Hand Side IRHS) values:
\ are always non-negative. : }
2. The algorithm goes through the intermediate iterations w
function are found (Iteration).
3, The algorithm terminates when there is no entering variable to provi
hnere solution with better values of the objective ——
ide the optical, solution (Termination),
Initialization —pe
Letus consider the following problem:
Minimize Z= 3X, +4K,
Subjetted to 2K, 43X28 ag
5X, #2X, 2 12 a. |
t aan )
| [the fst step isto convert the inequalities ito uations. We add variables Xpand gy fa cert
glack (or surplus} variables thet hhave to be subtracted from the left hand)side values toequate to the RHS values.
| Maile Zi Bx, ta hy OK Ones, f
CI ' *
aed Dt, + BXy = Ke 28 : .
3X) +2X_ = Xqy: co.
* oy Xp G0
KZ xe — =
prop og EX Crcot be
takes 5 ARE ficial Nosable
int xr the specific pur ose of starting the si
2 mga herp oem wefan es a h s
uw denoted by al and 2, Since they are not part ofthe original pol ton
be the at “dk ephiiainn. Solution :
inplex table, These are °
dot.
| Minimize Z= 3K, 44K, 40K +0X + NM yt May .
: ‘Subject to i; —
2X, +3X,-X,+a,=8 .
SX, +2X,—X,+a, = 12
Ky Xyy Xp Xp ap a 20 —-—
Providing a large positive value to the objective function coefficient of the artificial variable ensures that
the artificial variables do not appear in the optimal solution (if it exists), —
‘We define the large and positive value to the objective function coefficient of the artificial variable as M | —__—-
(big M), which is large and positive and tending to infinity. We have
'M *constant = M. 1 =
| Mo consant= i
‘Now, we begin the simplex table to start solving the problem. Since our standard problem has a maximization
= wwe multiply the coefficients of the ean function of a
inition Sblem by —1 and convert it tg a maximization problem. The simplex iterationlis “~~
jyn in Table 2.5, .
| Table 2.5 Simplex Table for Illustration 2.2
=3 = 0 0 —M_ —M_
ff X X Xs X a 8 RHs | 6 |«
cm] @ | 2, 3 a ofa 0 8 1, ———=
ial | st] 2 o}| -+ [0 1 12
Oe ed ee 0 0 ~
=f % | o 115 =t 1 25 165 | ——
3} x | 4 ast | o | 25-35] o ws | 125
G-Z] 0 |itsM+ 1415) -M |26M-3/5] 0 |-7/5M + 3/5
a fa] % fo 1 [sn] oan | sit] =a | tert :
31% {1 o ait] -att fans] ant | ont :
G-%) 0 0 f-tana] 11 | ve | -ve | -12601 ‘Tho optimum solution is X, = 20/11, X3 = 16)
. I and Z= 124/11. The simplex tablo will show
fdas value beau we have solved x maxinlantonpobles by mulityg toe
= t ‘Tho Simplex method involving big M is called the Big M method. i
7 t ‘Some important observations are ‘
|. Miseasier to introduce artifical variables when the problem does not havea visible ital
= basio feasible solution. We introduce as many artificial variables as the number of
- constraints. The number of variables increases and not the number of constraints, The big
‘Measures that tho artificial variables doraot appear in te optimal solution if one exists)
' 2. Since we have two artificial variables asthe starting basic variables, we need a minimum
of two iterations to find the optimum, since they have to leave the basis,
7 3. We need not evaluate. the C; ~ Z, values corresponding to the artificial variables at all
: because we don't want the artificial variable to enter the basis. They have been shown as
negative in Table 2.5. :
- ‘There is another method called two-phase method from which we can get an intial basic
feasible solution forthe simplex algorithm using artificial variables. This is explained below for
- the same example.
; Here; the artificial variables have an objective function coefficient of -1 (Maximization), The
' other variables have an objective function coefficient of zero. The simplex iterations are shown
ao in Table 2.6.
Table 28 Phase | of two-phase Method
a 7 Oa es0ees eta laeect
t x Xe % | | 8 RHS | 6
; 1 af % 2 3 -1 0 4 0 8
i aja {5 2 Gib ° 1 12 [4126
G-Z%) 5 5 “1 1 0 0
1 ay 0 115 “1 1 “25 16/5
of x | 4 ast | o | 25-5} 9 vs | sas [ett
: G-4] o-} ws | 1} -26 | o | -75 1 |
oO Xp oO 4 5/11 an SM at 16/11 '
o| x | 1 ° ant{ -a1 {an} “a | zon
' G-%] 0 ° ° o ft] om °
| Having obtained the optimum for the first phase with X; = 20/11 and X; 16/11, we can start
i the second phase of the simplex algorithm from the first phase by eliminating the artificial
{ ! variables. The second phase iterations are shown in Table 2.7.
i Table 2.7 Phase I! of two-phase Method : I
> ie 3 [470 [oe . t
—& % | % | % [% [Rus
ole ee Of 1 | -st] 21] t6/11
3] x, 1 | 0 | 2a} ana] aor
= G-% oO O | -14/11) -1/11] 124/19]
— ,. Forour example, we redlize that the starting basic feasible solution without artificial variables
‘at the end of the first phase is optimal. If it is not, then we proceed with the simplex iterations to
set the optixpal
solution. 4 Initialization deals with ‘n initial basic feasible solution for the gives. PO!
oot 4, Wentifying a set of basic wi ty ie a oe
{ loon pose, consider the following aspects of initialization (in the
88 sated): . :
tt. Right Hand Side (RH) values |
oz Variables »
ee
© 4 Constraints,
= ‘Let us ‘cons\ter each of them in detail
——_—_ ‘The Right Hand Side (RHS) value of
=X, ~X, where both X, and X, are 2
estar in al he cosas swell atin objec ction. .
‘maximization or minimlzation If itis minimization, we
ion. If constraint is of < type, we add
> we add a surplus variable (negative
_ iuusrranon 2.3 _—
Maximize 2 = 1X7 5X, :
+ Subject to
; . X20
{it comer the secon constant into an equation by adding anegative slack variable X,
“C7 jeauatons are:
‘The constraint coefficient matrix is
A We do not find variables with coofcients asi
{variables ay and ay 10 get
2K, 43%) +a =7
SK, +25 +a, =11
~~ Weave to start with a and ap as basic vari
ables and use the big Mf method
liuustearion 2.4
Maximize 2= 7X; + 5%, + 8 +O
—— Subject to I
2+ X= 7 {
= SX, +2%,+X,2 11
So Xv Xa Xs, X42 0
Wats example, we add sup variable Xs othe second contin o convert it to an equation,
= We get
2K, +3, +X, =7
SX +2644 X= 2 and then add the surplus vari
: ee
. If we have a2 constraint, we add a negative slack to convert it to an equation-This negative
_ solution. . : i |
WWe observe that variables X5 and X, have coeficiens ofthe identity matrix and we can start with
these as initial basic variables to hav ewe sti
in this case, ~~ .
sos to have a basic feasible solution. We need not use artificial vatiables ( <~
TE the second constraint was 5X; + 2%) + 2X,2 11, we ean write it s St, +p +42 UW ae
iable and choose X, as a siarting basic variable.
Adding artificial variables
ensure that the RHS value of every €pnstraint is non-negative.;
. fave have a < constrain, we add a Stack variable, This automatically qualifies 19 be an
initial basic variable.
ger oarant yc
slack cannot qualify to be an initial basic variable.
. In the system of equations identify whether there exist variables with coefficients
corresponding to the column of the identity matrix. Such variables qualify to be basic
variables. Add minimum artificial variables otherwise to get a starting basic feasible
\
Notew Once the anteficral variable | has
ed
Pie. the Aimplex table beciaune-
J
a { :
Slate to! eutirn sucto tre
20] Iteration
During iteration, only one ispue needs to be addressed. Let us expiai tit
i an exams
ILusTRATION =
Maximize Z= 4X, + 3X2 :
Subjyet to
’ 2X + 3X58 -
3X, + 2X2 S 12
X,%220
‘Adding slack variables X3 and X,, we get
We set up the simplex table with X3 and X; as basic
2X, +3%)+%3=8
BX, + 2X2 + Xa =/12 :
variables. The simplex iterati
are shown |
in Table *
Table ‘Simplex Table for Illustration
4 3 o o
Ge Basie | | % | %| Xe | RAS e
variables
ie Xs 2};3 ]1]° a | 4
Q Xy 34| 2 | 0] 1 42 | 447
—Te-zZ }a'fafol[o] °°
0 Xs 0 | 53,| 1 [-23} 0 04>
. 4 x 1 | 23{] 0] +3} 4 6 : :
G-% | o | | 0 |-43| 16
3 |. % © 1.1 | s5|-25| 0 -
4 xX, 1 0 |-25} 3/5] 4 :
G-% | 0 | 0 |-1}-8/5 16 -
Inthe simplex Table, there was tie for the leaving variable. We had a choice between X3
and.X,. We chose to leave
iteration. Although the: optimal solution was found in
then, We performed one more i
optimality condition.
X,. This tie resulted in a basic variable getting value zero in the next’
the second iteration, it was not apparent
functiba but'to obtain the
ut.
iteration without increasing the objective
one of the basic variable takes the value zero
‘Whenever theres a tie for the leavin variabl
y noxt iteration. This | We could perform some extra
occurs.at the optimum.
egeneracy results in
have an increase in the ob
in the next iteration, This henomenon is called de TAC)
i rT inarease the value oft te objective fi Function. In this example, degeneracy
iterations that do mot increase
iterations that do not improve the objective function value. Sirice
extra
leaves a variable with zero value in the next iteration, we do not
restive function Value (Note that the-nerease in the objective function |
— Fee roduct of the C,- Z, and and the minimum @ takes the value zero).
neracy can take place in the intermediate Herations. In such cases, if the * +
-| ‘Sometimes deget
ptimum exists, the simplex.
soos these cases, the entering column will have a zero (or negative) value against the
the optim
algorithm will come out of the degenes itself and terminate at’
~ Jeaving row and hence that Owill not be computed, resulting in a positive value of the minimum 2
breaking rule can resul
Thero is no proven way to gliminate degeneracy oF to avoid it, Sometimes a different tie
termination in fewer iterations. In this example, if we had chosen to ,
eave X; instead of Xin the first iteration, the algorithm terminates and gives the optimuin after ' feo region |
> 2 Gi
ome a x Brin
Zr yer
aan ae Jorleung XO =
oO ik Termination. : , :
i inati ‘These are: ——
There are aspects to’be addressed while discussing termination conditions.
1. Alternate optinrum
2. Unboundedness .
3. Infeasibility :
Let us.consider each throug) an example a
Juustranion ©
Maximize Z= 4%, +3X, ;
Subject to
BX, + 6X, $25
3X, +4%45 15
X, X20
f : basic
Adding slack variables X3 and X,, we can start the simplex iteration with X3 and X, as basi
variables. This is shown in Table. ».
Table Simplex Table for Illustration *
7 EI CT)
Ca | Basic | x, | x, | % | x; | RHS | @ .
variables |
2 ef} o6 | [oo] 25 | 255+—~ ——
i a 3] 4 [oo | 1] 15
alTa[olol] o
4 ¢ [sa] elo | 25 [256 —
2 0 | 74!) -3e| 4 | 45/8 | 45i145—
o | o {-v| 0 | 252 =
4 7 | 0 | 2|-a7l o7 : \
; \ 3 |. xy | O°} 1 [sits] a7] casita, vs
-~ G-% } 0 | o |-v2) o | 252 fo...
I
At the end of the first iteration we observe that the non-basic variables (X> and X,) have non-
positive values of C, — Z, indicating that the optimum solution has been reached. However, one ~———
of the non-basic variables X, has aC, ~ Z/value of zero. If. we enter this, we get another optimum
Jetuicn swith the same. value of the objective function, Now; noni-basic variable Xa has C3 ~Z3 =
It looks as though the simplex algorithm seems to be getting into an infinite loop although
optimum has been found. This phenomenon is called alternate optimum: This h: is when one
ot the-constrains_is_patallel to the objective function. Actually there”are inion number Of
alternate optimum solutions for this problem, but the simplex algorithm
points. Every point on the line joining the:
se two solutions is also y bevstearion
|__* Maximize. Z= 4X, + 3X) ,
© Subject to . : :
{os A }
t = X-6%s5 °
j 3 3X,s 1
: X20
| Adding slack variables X; and X,, we can start the simplex iteratic i
d splex iteration with X3 and
| Variables. This is shown in Table”. ae ce
4— Table" Simplex: Table for Iilustration
_ aT soe
Gp | Basic |X | % | % | %] RAS |e
variables
0 y]e]tfofs5 ol
ae 0 st] 0 oO 1 4 11/3,
at 3a{ofo| o
0 0 [6] 4 [=a] 43
4 A oF 0 13) 19/3.
0 | 3'[ o [4a] 440
ve observe that variable X, with C3 ~~ Z =3 can enter the basis
At the end of the first iteration, w« s
+— but we are unable to fix the leaving variable because. all the coefficients in the entering column arg
OTR phenomenon scaled wnboundedness,idiesing ht te WS ts take any
cent basic variables would ‘become infeasible. By: the nature of the
|. valtie and still none of the pres:
-~ nha db bh moe Dato
Se eee
Fout cowstraamnt,» we can observe:
——_———
Com be anarared. to an Valur - 7a
dhe lowotepin - Se <7,
eh hin
Inall simplex iterations, we enter the variable with the maximum positive value of Cy = Zy.
|" Based on this rule, we entered variable X; in the first iteration, Variable X2 also with a positive
value of 3 is a candidate and if we had decided to eriter Xp in the first iteration we would have
realized the unboundedness.at the first iteration. itself. ‘
[Though, most of the times we enter a variable based on the largest coefficient mule (largest
G,~ Z), there is no guarantee that this rule terminates with ‘minimum iterations. Any non-basic
——t variable with a positive value of C, - Z;is a candidate to enter. Other rules for entering variable are:
1. Largest increase rule: Here for every candidate for entering variable, the corresponding
minimum @is found and the increase in the objective function, ile. the product of @ and
1. C)=Zyis found. The variable with the maximum increase (product) is chosen as eftering
1° variable. {
2. First positive C, ~ Z;: The first variable with C, ~ Z;> 0 enters.
3. Random: A non-basic variable is chosen randomly and the value of CG -Z; is computed.
It becomes the entering variable if the C, ~ Z, is positive. Otherwise another variable is
chosen randomly. This is repeated till an entering variable is found.)
‘Coming back to unboundedness, we observe that unboundédness is caused when the feasible
| region is not bounded, Sometimes, the nature of the objective function can be such that even if the
feasible region is unbounded, the problem may have an optimum solution, An unbounded {-P
means that there is no finite optimum solution and the problem is unbounded, .*
thet Xe,
Tie
. The Malate ~ capable of detecting infeasibiftty but al
d Sofenn ihibitg ~ : _
kitsrranon
Maximize Z
=a —
Subject tt Ma
X44 s3 . :
3X, +X, 212 a
X20
Adding slack variables X; and X, (surplus)
al,
gorithm using the big Method Mate Sisal variable a we can start the simplex
met id a, as basic variables. This is shown in Table
le" + Simplex Table for IHustration
Ce 3 oo
. % | % [xa [Rag [6
-M «lee ols .
= +} oo [alt] & fap .
7 wes | 0 Toml o
a -M 4 1 | 0
su 3 [aio] 3
wim+t3[-omsa[ mo | >
“
Here the algorithm terminates when all the non-bas i
. wl non-basic variables Xo, X3 and X, have negative
. = (2, Xa (, have negative
fatues of G Zz, Tee optimality Sodlion seems to be Satisfied but an artificial variable is in the
is means ie problem is infeasible and does not have a feasible solution.
Infeasibilityiis indicated by the presence of at least one artificial vanable ‘after the optimum
cndition‘is satisfied. In this problem a, = 3 indicates that the second constraint should have the
RHS value reduced by 3 to get a feasible solution with ~ 3. Simplex ‘algorithm not only is
Iso shows the extent of infeasibility. ‘
‘Termination conditions (Maximization objective) a
1. Alllnon-basic variables have nbgativeyvalues of CZ, Basic variables are either decision
stables or slack variables, Algorithm terminates indicating unique optimum solution.
fasiables arc either decision variables or-slack variables, All non-basic variables
© 0, At least one non-basic variable has C, ~ 2/= 0. It indicates alternate
‘ced to find the other comer point and terminate.
5, Basic variable are either decision variables or slack variables, The algorithm identifies an
fy leaving variable because all values inthe entering
‘entering Variable but is unable to identi :
Column are < 0. It indicates unboundedsess and algorithm terminates.
Z, $0. Artificial variable still exists
2. Basic v
have G ~ Z
optimum. Proc
4. All non-basic variables have C; ~
indicates infeasiblity and algorithm terminates
in the basis. It \\
Let us consider the following linear programming problem: \ | -——
Maximize Z= 6X, + 5X,
Subject to ,
3X, +2%) $12
X, 20
Let us.assume that we do not know the optimum solution. Let Z’ be the optimum value of Z. Let
us try to find.a value higher than Z” without solving the problem. We wish to have as small a value
as possible but it should be greater than or equal to Z”.
1. The obvious value is infinity because itis a niaximization problem. If we ignore both the
constraints, the value will be infinity.
has to be feasible (Xj, Xs > 0) and, therefore, should satisfy 9X, + 6X2 < 36. Since
GX; + SXp S OX; + 6X2 S 36 for X;, Xp 2 0, the upper estimate of Z” is 36.
3. Let ug multiply the first constraint by 6 to get 6X; + 6X7 < 30. The optimum solution hat
to be feasible (Xi, X2 2 0) and, therefore, should satisfy 6X;.+ 6X, S30. Sinop
( 6X, + 5X, < 6X, + 6X, ¥ 30 for X;, X> 2 0, the upper estimate of 2” is 30,
4, Let us multiply the first constraint by 1 and second constraint by 2 and add them to
( ‘7X, + 5Xz $29. The optimum solution has to be feasible (X;, Xz = 6) and, therefc
should satisty 72; + 5X9 $29: Since 6X; + 5Xp 7K, + 6% $29 for Xi, $0, the upper
1s} estimate of Z" is 29. &
From the above arguments we undeistand that if we multiply the first constraint by a non-
id 2. Let us multiply the second constraint by 3 to get 9X, + 6X $36. The optimum solution
|
t
+ negative quantity @ and the second coiistraint by a non-negativé quattity and add them such that
If ets oe.
| exists, Let
ah t ae ae : |
the resultant constraint ha all the coefficients more than that in the. objective function then the,
ight hand side value is an upper bound to Z’. This is under the assumptian that the constraints are,
The lowest value that can be achieved will have g gertain value of a and b, if such a valu
Let such values of a and 8 be, ¥, and Yo :
la order to get the upper estimate of Z*,
aan 7 Y+3%26. Bl)
: F+2%, 25 G2)
. . %H20 , . @3)
For every ¥; and 3¥5, satisfying Eqs. (3.1) to (3.3), I= SY," 12Y, is an upper éstimate of Z”. The
lowest value 17 can take is given by she liféar programming problem. °
Minimize W = SY, + 12¥>
Subject to , no :
: ¥,+3%26- |
‘ Ytwoes tn t [
¥, ¥20 °
‘The above problem is called the duat of the Biven problem. The given problems i is called the )
primal. 7
3 X+Xys5 | — \ Solution: Let us bring
4X, +2 + ay =12
TX, + 5X2 + 6X3 29
8 15% +4555 0,
3X, + 7% 49% 27
‘unrestricted in sign and‘, < 0
X,20,%)
the primal to the standard form: - : : |
. Maximize = CX :
ceo. -
. Atsb
X20
Variable X3is unrestricted in sign. We replace it as the difference of two variables X5 =X; ~ 5,
where both X, and Xs > 0; Variable X; < 0 is replaged by -X¢ where X¢ > 0.-Now, the problest
becomes .
* Minimize’ Z= 8X, + 5X, -5Y,— ax,
Subject to :
AX + 2K ~ 2K, - 8X5 = 12
TX, + 5X4 ~ 5X5 ~ 6X49
8X) + 5X4 ~ 5X5 ~ 4X, < 10
3X, + 1X4 ~ 7K - 9X27
Xin Xi Xs, X20
We convert the objective function to maximization by multiplying with 1. ‘The first
Soustrant, which isan equatin, is rewritten as two constraints one with a < sign and another with
2 sign. All the 2 constraints are writen as < constraints by multiplying them with -1. Making the
+ Ay Ky Wie Xe 12 :
AX, ~ 20,4 254 845 5-12 |
IX; ~My + 5Xy + 6X5 < -9
8K, 4.5% - 5X5 - 4Ke 10 |
WN, ING kg + 9X, 277
Hy Xn XeXe2O
a —— . a
Now the Prisial is converted to the standard form for which we can write the dua.
Tntroducitg variables Y; to Ys for each of the constraints, we write the dusl ta: i] —
~ Minimize 2 = 12, ~12Y, - 9Y, + 107,~ 77, '
Subject to . 7 : : —
41 - 4%, -774 + BY, BYs2-8
2Y; - 2Y2~5Yq + SY, TY 52-5 4
“20, + 2%, + 5Yy~ SY, +525 : “ a
SY; + BFy + 6Yy~4Y4 + 97524
Ny, Ya, Ys, Yq, Ys2 0
‘Since the primal is «minimization problem, the dual should be a maximization problem.
Multiplying by -1, we get
Maximize Z=~12¥, + 12¥, + 9Y3 - 10¥,+7¥s
\ Subject to = =
7 . AY, — AY, -7Yy + BY, ~ 3752-8 : |
— 2Y; = 2¥_ 5%, + 5Y4- ¥42-5
W2Y, + 2Yg + 5¥4~5¥4+ 4525 =
. BY; + 8%, 46%, ~ 41, + 9¥52 4
—_ Yu Yo Ys, Yas Y520
Defining ¥5= Yo— ¥; and substitut
we get _—
Maximize Z=12¥6+9¥;-10¥,+7¥5° : I
——— Subject to io 7 . eo
3 2 He - TY + 8%4—3¥52-8 . o =
i : =2Yg~ S¥s + 5Y,— 7¥52-5
" 2Y¢ + 5Y3-5Y,+7¥525 ee
+ 86+ 6Y3—4¥, + 9¥524
Yg unrestricted in sign, ¥3, Y, ¥52 0
+ Replacing Y, by -Yp such that ¥7< 0, we get 4
Maximize Z="12Y, + 9Y, + 10¥; + 7¥5 ao : . ———
Subject to“ a
— AY, ~ TY ~ 8%) - 3¥52-8
- -2¥g ~ SY, -5¥- W¥52 -5 —
— We + SY3+ 5¥y + 7¥525
8Y6 + 6Y; + 4¥, + 9¥524
Yo unrestricted in sign, Y3, Ys 20, ¥7 0
‘Multiplying the first and second constraints by —1. Writing constraints 2 and 3 as equatich,
ae
——1 we get the dual as: : 1
Maximize |W = 12¥¢+ 9¥3 + 10%; + 7¥s —_—
Subject to :
= AY + 1¥y + BY) + 3¥5 $8
‘ . 2Yg + S¥3 4 5¥y + T¥s=5
i 86 ¥ Fs HAY, + 9V524 ye i
Yo unrestricted in sign. Y3 20, ¥7 $0, Ys20
'
PRIMAL-DUAL RELATIONSHIPS
We observe that if the above problem wassthe primal, the original LP would be its dual. ¢
wel 1
\ In any-LPP, the ual of the dat isthe primal itsef
o 4 [sein
ont
—_o
: ‘ame bo eais* -
‘We also observe the following based on our examples. Table 3.1 shows
‘between the primal and dual problems.
Table sM_ Relationships between the Primal and Qual Problems -
‘Maximization Minimization 7
Minimization Maximization _
‘Number of variables (n) Nuiyber of constraints (7)
Number of constraints (m) Nuritoer of variables (m) Pe
PHS (b) Objective function coefficients (¢)
‘Objection function coefficients (c) HS (6) =
Constraint coetticients (A) Constraint coefficients (AT)
7
If the primal is @ maximization problem, the dual is a minimization problem and vice verse.
We have as many variables in the dual asthe number of constraints in the primal and vice versa.
‘The RHS of the primal becomes the objective function coefficients of the dual and vice versa, __
‘There is also a relationship between the types of variables
constraints (of the dual) and vice versa. This is also shown in Table 3.2
Primal-Dual Relationships for Constraints and Variables
(of the primal) and the type of
Primal (Maximization) Dual (Minimization) _
S constraint ® variable +
> constraint s variable a
equation constraint unrestricted variable
2 varable 2 constraint | | _
ne ‘Svariable Sconstraint ©
Unrestricted variable equation constraint
We interpret Table 3.2 as follows: * :
is of 2 type and so on.
LE tho first constraint of the primal (maximization) is of < typeyth frst variable of the
If the primatis a minimization problem, the table is to be interprete
dual
‘Gbrrespondingly. If the
cha? isa tninimization problem and fas {hird constraint as < type (say)'then the third variable
f thé lal is Of the < type. . ee a
Result. 2 Weak Duality Theorem ©
|For a maximization primal, every feasiblg solution to-thé- dual has’a objective function value|
greater than or equal to every feasible solution fo the primal, on)
Let us-apply this result to the following oxangple* nd
The primal is: aie 2 —
Maximize Z= 6X, +S¥jt* + AY at :
—s
aoe ena
elo ‘tt =
M+Xps5n| |
= 2 3X, +2% sig = =
ee nee 2 _
The dons! 2 Fees ee MY
af te ' -
_- .- Subject.o.
oe
wae
Mininivd Wy ey, 4 157)
Yyt W226 ate
= Yat ry BS et
iY, 202 NA Ai
a a (Let be the objective function value of solution feasibie to the dual 7 7
Sitice we have a feasi ee |
« feasible soluti
Subvttting we gat Sonn Mn Xe to the primal, we have Xj + Xp 5 and 2%; + 2% 12
We WK +X), + GX +24) 7,
‘Rearranging the terms, we get me
Since che W2X, (K+ 3%) +.% (Y, + 2¥,)
jince Y; and ¥ are feasil
Since Y, and feasible to the dual, we have Y, +3Y,> 6 and Yy+ my 5. Substituting, we
We 6X, + 5X, I
ince Xi, Xo are feasible to the primal, Z= 6X; + 5X;
, ‘+ 5X and hence W 2 Z.
4 ae weak duality theorem also addresses certain other aspécts to LP problems, For example!
pre Prunal (maximization) is unbounded, what happens to the dual? Since the pri is
abounded it coe ai has a feasible solution. If the dual has a feasible solution. eoorey :
solution sl lave an objective function val i i
leads to the conlusion that the duel should Be infeasible nn “Sh OE Pole, Tas
* What happens if the primal itself is infeasible? The dual is then unbounded.
also be infeasible (See Solved Example 3.3). eee
Result °3 Optimality Criterion Théorem
Ifthe primal and dual have feasible solution with the s. i
a ame value of the objective function then
both have optimal solutions with the same value of the objective function
‘The proof of the optimality criterion theorem comes from the weak duality theotem. Since |
both the primal and dual:have feasible solutions’ with thé’ samo value of the objective function,
they have to be optimal to the primal and dual,
they have tobe primal , respectively. Otherwise the weak duality theorem
1
a
_, \ The fundamental theorem defines that three possibilities exist for every LPP. ‘Since the
* edhstraints are a linear system of ine¢iaalities, there has to be a comer point if a feasible region
_ exists. Also, considering the alternate optima case, if there is an optimal solution, then it has a
Result “4 Main Duality Theorém
If primal and \ have feasible solutions then both have optima sohitidrs with the same value
of the objective function. cs ao
Ifthe primal and dual have feasible solutions, then they should:have pptimal solutions. Ifthe
primal (maximization) is feasible and unbounded, the dual is infeasible. Weak duality theorem
clearly states that the maximization problem cannot have Z* > W* (where Z* and W* are the
~ objective function values of the prima and dual respectively at the optimum.
‘We explain later the complimentary slackness conditions, which arise out of the Kukin Tucker
Conditions (described later in Chaptef*12) that establish that if the primal and dual are feasible,
they have feasible solutions with the stime value ofthe objective function. The optimality criteripn
theorem also states that ifthe primal and dual have feasible solution with the same vale of the
objective function, then both have optimal solutions with the same value of the objective functibn,
‘The duality theorems help us generalize the fundamental theorem of linear programming.
Py - ~~ Result 25 i
If the primal (maximization) is unbounded, the dual is infeasible. rr
If the primal is infeasible, the dual is unbounded or infeasible. i
Result °\6 Fupdamenial Theorem of Linear Prograinming
Every linear programming problem is either feasible or unbounded or infeasible it has a
feasible solution then it has a basic feasible solution. If it hag an optimal solution least one
comer point solution is optimal. a
." comer point optimal solution. Unless it is unbounded, it will have at least one corner pot
t
rit
solution that is ptimal. : t = ren 2
‘ + “Result =.77 Complimentary Slackness Theo! yf and V' seth]
——\_— ex and ¥" are the optimal solutions to the primal and dua, respectively and U” #7
(| values of the primal and dual slack variables at the optimum then oa
X'Vv'+¥'U'=0
sas jables aie greater than or equal 10 —
‘Since X" Y* U* V"* are vectors and all decision and slack varial
— , it reduces to every x)= yu, = 0. . 5 2 ~
Lett aemelr He ‘complimentary slackness conditions to the earlier example: =
_|—— tuusreariow #2 _
|__| The primal is: x . 3
Maximize Z = 6X, +5X2 _
Subject to oe :
— | : XytXytuy=5
= a. 3X, + 2%) + ug = 12 .
A Xp Xz, uy tz 2 0 (where u; and uz are slack variables)
—|-— The dual is: - :
Minimize W = 5Y, + 12Y, :
“T= Subject to .
to} 2 ata Vy +3%,-yy=6
¥, +2%)—w = 5
—— Ny, Fv 20 ae
|) The optitnal Solution to the primal is: °
a Ae Xin 2,X}=3,Z=27
|: The optimal solution to the dual is: +. Dig
—~ k, ee, W933, B= 1,W=27 7
plying the complimentary slackness conditions we verify that since Xy'and X are the basic.
* ,Variables at the optim, v; and v are non-basic and take zero value. Similarly, since ¥, and Y, are
+ basic variables of the dual at the optimum, we verify that uy and uz are non-basig with zero value.
or : if & :
huustrarion "3 sit : ; |
—|- Maximize Z=3X,+4¥> : i 3
* Subject to Joy
“+ . “ MAM S12 |. .
\; - 2X +BXys 30,
Xp aXe $36
1: : Nkp2.0
The dualis: °° FS ohn
7 Minimize Z=12¥,+30¥,+36¥, 9 - =
|. Subject to ae
1+2n+7,23°
+ . Vy + 3Yp + 4 2h en :
artsy et ’ Tyo Ks 20h tt
(Variables;¥y.my aud’ w3:are primal slack variablesiwhile variables v; and vp are dual slack
J. variables)ay v5 ne mae : 2
The optima}-splution to the pringal is: .
Race - XL = 6X3 = 6,3 = 62M? :
- The opis solution tothe dua is: 7? : Jf
Z WAL Yai, W =42 00 (
Since 1, Xe are basic variables of the primal atthe optimum, dual slack’variables 9, and a
fon-basic atthe optimum to the dual with zero value, Since variable. is basic atthe optimura to the pri inblo Ys i i :
; L tte prima vrai Flow basic pi ult etn Variables Y, and
= al optima, sk vale an enone whe orion | .
oo i it ween asian ann tei cpinnieig
aouaey eves rom the optimum solution, X; = 6,3 ~ 6, ui = 6,2" = 42, we serve
neler ae basic atthe optimum tothe dual (because slack variables wand vz ane
ia optimam to the primal), Also variable Y is non-basic atthe optimum tothe dual
Becta yi ass to the primal tthe opium,
It is enou; i Si
7 eg fae Be Faz ad = mae a
__ Using the solution Yj =1, Y= 1, W* = 42, we can compute the optimum solution to the
| primal). ‘
% MATHEMATICAL EXPLANATION TO THE DUAL t
Now, we discuss the dual mathematically ising the following problem: #,
—|——— Maximize Z = 6X; + 5% ~
Sabject to =
Xt+Hs5 —
—+——- \ 3X, +2X, $12
oo \ Ow 6 %20 7
‘The optimal solution to the primal is: .
= = . : Xp=2xXpeh 2-27 i
—}}— The optimal solution to the dual is: :
_|. Yo3,geL adr ae . =
‘efirst constraint such that the resource available now is ~~
| Let us add a small quantity 5 to the!
“Pq 5 + 6 Assuming that X; and will remain as basic variables a he eptimum, and solying for %1
isl 4 pod Nisin gat = 1-280, RE “Be Oe
‘The increase in objective function value at the optimum
___ eoistraint (resource) is 36, where 3 is the value of the first dual variabl
value of the dual variable at the of te of change of objective’
for a small increase Sof the first __
fe ¥, at the optimum, The
function for a small. —
ed as.the changéin the objective function for
yptimum is the rat
__|-
