BILKENT UNIVERSITY

PRINCIPLES OF ENGINEERING MANAGEMENT

MIDTERM 1 - Solutions
Code : IE 400 Last Name : Section #:
Acad.Year : 2023-2024 Name :
Semester : SPRING Student # :
Date : 03.03.2024 Signature :
Time : 16:00 5 QUESTIONS ON 10 PAGES
Duration : 150 min TOTAL 100 POINTS
1. (20) 2. (25) 3. (20) 4. (20) 5. (15)

Please draw a box around your answers. No calculators, cell-phones, notes, etc. allowed.
1.(20 pts) Consider the following LP problem

max 3x ` 2y
s.t. x ` 2y ě 5,
x, y ě 0

a) Convert this problem to standard form.

max 3x ` 2y
s.t. x ` 2y ´ e “ 5,
x, y, e ě 0

b) Let x be the initial basic variable. Put this system in the Simplex Tableau.
» fi
z x y e RHS
z – 1 ´3 ´2 0 0 fl
— ffi

x 0 1 2 ´1 5

c) Put the tableau in canonical form for this choice of basic variable.
» fi
z x y e RHS
z – 1 0 4 ´3 15 fl
— ffi

x 0 1 2 ´1 5

d) Is the current BFS optimal? If not, which variable is the entering variable?

There is a negative reduced cost coefficient. e will be the entering variable.

e) Can you find a leaving variable using the ratio test? Is this LP bounded?

There are no positive terms in the column of e. This shows that LP is unbounded.

f ) Verify your answer for part e) by solving the original LP using the graphical method.

Plot is trivial. Isoprofit lines cannot leave the feasible region completely. So the original LP
is unbounded.
2.(25 pts) Consider the following LP

max 2x1 ` x2 ` x3
s.t. x1 ` x3 ď 2,
x2 ` x3 ď 3,
x1 ` x2 ď 4,
x1 , x2 , x3 ě 0.
» fi´1 » fi
1 0 1 0.5 ´0.5 0.5
You are given that –0 1 1fl “ –´0.5 0.5 0.5 fl .
— ffi — ffi

1 1 0 0.5 0.5 ´0.5

a) Convert this problem to standard form.

max 2x1 ` x2 ` x3
s.t. x1 ` x3 ` x4 “ 2,
x2 ` x3 ` x5 “ 3,
x1 ` x2 ` x6 “ 4,
x1 , x2 , x3 , x4 , x5 , x6 ě 0.

b) Let x1 , x2 and x3 be the initial basic variables. Put this system in the Simplex Tableau.
» fi
z x 1 x2 x3 x4 x5 x6 RHS
— ffi
z — 1 ´2 ´1 ´1 0 0 0 0 ffi
— ffi
x1 —— 0 1 0 1 1 0 0 2 ffi
ffi
x2 – 0 0 1 1 0 1 0 3 fl
— ffi
x3 0 1 1 0 0 0 1 4

c) Explicitly write what B, N, cB , cN and b is for this choice of basic variables.

» fi » fi » fi » fi » fi
1 0 1 1 0 0 2 0 2
B “ –0 1 1fl,N “ –0 1 0fl,cB “ –1fl, cN “ –0fl and b “ –3fl.
— ffi — ffi — ffi — ffi — ffi

1 1 0 0 0 1 1 0 4
d) Put the tableau in canonical form for this choice of basic variables.
» fi
z x1 x2 x3 x4 x5 x6 RHS
— ffi
z — 1 0 0 0 1 0 1 6 ffi
— ffi
x1 —
— 0 1 0 0 0.5 ´0.5 0.5 1.5 ffi
ffi
x2 – 0 0 1 0 ´0.5 0.5 0.5 2.5 fl
— ffi
x3 0 0 0 1 0.5 0.5 ´0.5 0.5

e) After converting the tableau in canonical form explicitly write the current basic feasible
solution, objective value and the reduced cost coefficients.

» fi
1.5
—2.5ffi
— ffi
— ffi
—0.5ffi T ´1 T
” ı
x̂ “ — ffi, ẑ “ 6 and cB B N ´ cN “ 1 0 1 .
— ffi
— 0 ffi
— ffi
– 0 fl
0

f ) Is the current BFS optimal? Why?

All reduced cost coefficients are non-negative. In a maximization problem this means the
BFS is optimal.
g) Are there any alternate optimal solutions? If you think there are, explicitly state the
entering variable, the leaving variable and the pivot term to seek the alternate optimal
basic feasible solution and do the pivot operations. Write the parametric equation for the
line segment of optimal solutions.

x5 is non-basic with 0 as its reduced cost coefficient. Pick it as the entering variable and
after the ratio test x3 is the leaving variable. So we do the pivot operations

» fi
z x1 x2 x3 x4 x5 x6 RHS
— ffi
z — 1 0 0 0 1 0 1 6 ffi
— ffi
x1 —
— 0 1 0 1 1 0 0 2 ffi
ffi
x2 0 0 1 ´1 ´1 0 1 2 fl
— ffi
–
x5 0 0 0 2 1 1 ´1 1
» fi
2
—2ffi
— ffi
— ffi
—0ffi
This table is in canonical form and it is optimal. The current BFS is x̃ “ — —0ffi with z̃ “ 6.
ffi
— ffi
— ffi
–1fl
0
» fi
2 ´ 0.5λ
—2 ` 0.5λffi
— ffi
— ffi
— 0.5λ ffi
Then for any λ P r0, 1s, λx̂ ` p1 ´ λqx̃ is optimal. Explicitly, λx̂ ` p1 ´ λqx̃ “ — ffi is
—
— 0 ffi
ffi
— ffi
– 1 ´ λ fl
0
optimal.
3.(20 pts) A company is considering the development of facilities at m potential sites to meet
the needs of its n customers. The establishment of a warehouse at site j incurs a fixed cost
of fj dollars, while the transportation of products to customer i from warehouse j comes
with a unit cost of cij dollars. Each customer i possesses a specific demand Di , which can be
halved if the customer is compensated with di dollars. The initial capacity of any warehouse
at site j is Qj , which can be doubled with an additional investment of qj dollars. The goal
is to determine the most cost-effective strategy to meet customer demand that encompasses
decisions on warehouse construction, capacity expansion, customer service, and demand reduction.

Formulate the problem as a mixed integer linear program. Define the decision variables,
constraints and the objective function explicitly.
Sol:
Decision variables:
xij is$
the amount of products transported from warehouse j to customer i.
&1, warehouse j is constructed.
yj “
%0, otherwise.
$
&1, warehouse j extends its capacity.
wj “
%0, otherwise.
$
&1, customer i reduces its demand.
zi “
%0, otherwise.
řm řm řn řm řn
Objective: minimize j“1 fj yj ` j“1 i“1 cij xij ` j“1 qj wj ` i“1 di zi

Constraints:
Demand should be satisfied: m Di
ř
j“1 xij ě Di ´ 2 zi for all i “ 1, . . . , n
řn
Capacity should not be exceeded: i“1 xij ď Qj ` Qj wj for all j “ 1, . . . , m
Warehouse constraints: xij ď 2Qj yj for all i “ 1, . . . , n and j “ 1, . . . , m.

řm řm řn řm řn
min j“1 fj yj ` j“1 i“1 cij xij ` j“1 qj wj ` i“1 di zi
řm Di
s.t. xij ě Di ´ 2 zi @i “ 1, . . . , n,
řnj“1
i“1 xij ď Qj ` Qj wj @j “ 1, . . . , m,
xij ď 2Qj yj , @i “ 1, . . . , n, @j “ 1, . . . , m,
xij ě 0 @i “ 1, . . . , n, @j “ 1, . . . , m,
yj , wj , zi P t0, 1u @i “ 1, . . . , n, @j “ 1, . . . , m.
4.(20 pts) Consider the following LP problem

min 2x1 ` 3x2

s.t. 2x1 ` x2 ď 16,
x1 ` 2x2 ě 36,
x1 , x2 ě 0.
a) Convert this problem to standard form LP. Introduce artificial variable(s) to solve this
problem with Big-M method.

min 2x1 ` 3x2 ` M a2

2x1 ` x2 ` s1 “ 16,
x1 ` 2x2 ´ e2 ` a2 “ 36,
x1 , x2 , s1 , e2 , a2 ě 0.

b) Put this modified problem in the Simplex Tableau. Put the tableau in canonical form with
an appropriate choice of basic variables.
» fi » fi
z x1 x2 s 1 e 2 a2 RHS z x1 x2 s1 e2 a2 RHS
— ffi — ffi
z —
— 1 ´2 ´3 0 0 ´M 0 ffi
ffi ñ z
—
— 1 M ´ 2 2M ´ 3 0 ´M 0 36M ffi
ffi
s1 —
– 0 2 1 1 0 0 16 ffi
fl s1 —
– 0 2 1 1 0 0 16 ffi
fl
a2 0 1 2 0 ´1 1 36 a2 0 1 2 0 ´1 1 36

c) After converting the tableau in canonical form explicitly write the current basic feasible
solution and the objective value.

» fi
0
— ffi
— 0 ffi
— ffi
—16ffi with objective value 36M .
Current BFS is — ffi
– 0 fl
— ffi
36

d) Is the current BFS optimal? Why?

This is a minimization problem and there is a positive reduced cost coefficient. So this BFS
is not optimal.

e) Can you make any comments on the original LP problem?

Tableau is not optimal, we need optimality for certain comments. x2 is the entering variable
and s1 is the leaving variable:
» fi
z x1 x2 s1 e 2 a2 RHS
— ffi
z —— 1 4 ´ 3M 0 3 ´ 2M ´M 0 4M ` 48 ffi
ffi
x2 —
– 0 2 1 1 0 0 16 ffi
fl
a2 0 ´3 0 ´2 ´1 1 4

This table is in canonical form and the current bfs is optimal. Artificial variable takes
positive value in the optimal solution. This means that original LP problem is infeasible.
5.(15 pts) You estimate that there is a linear relationship between the stock price of a fastfood
restaurant and the number of hamburgers sold (in thousands) in a day:

Day Stock Price # of Hamburgers Sold

1 3 9
2 4 13
3 3 7
4 4 13
5 5 16
You predict the relationship as si “ a ˆ ni ` b, where si is the stock price and ni is the number of
hamburgers sold (in thousands) on day i “ 1, . . . , 5. For a given choice of a and b, your estimation
error on day i is |si ´ a ˆ ni ´ b| for i “ 1, . . . , 5. Choose the unknown coefficients a and b such
that the sum of estimation errors in 5 days is as small as possible.
Sol:
Decision variables: a, b and ti is defined to be the estimation error day i.
Parameters: si , ni for i “ 1, . . . , 5.
We have the following model
ř5
min i“1 ti
s.t. si ´ a ˆ ni ´ b ` ti ě 0, @i “ 1, . . . , 5,
si ´ a ˆ ni ´ b ´ ti ď 0, @i “ 1, . . . , 5,
a, b, ti free.