MID-SEMSTER SOLUTION

COURSE & CODE: PROCESS OPTIMIZATION CH4003D

Monsoon 2024-25

Question 1

(a) A company is redesigning its parallel flow heat exchanger to increase its heat transfer.

There are certain limitations on the design problem. The smallest available conducting

tube has a radius of 0.5 cm and all tubes must be of the same size. Further, the total cross-

sectional area of all the tubes cannot exceed 2000 cm2 to ensure adequate space inside the

outer shell. A designer has to determine the number of tubes and the radius of each tube to

maximize the surface area of the tubes in the exchanger. Formulate this as an optimization

problem. 2 Marks

Solution

We have to decide the size of the tube and the number of tubes.

Let r be the radius of the tube in cm. 0.25 Marks

Let n be the total number of tubes. 0.25 Marks

The smallest available conducting tube has a radius of 0.5 cm. i.e. r ≥ 0.5. 0.25 Marks

The total cross-sectional area of all the tubes cannot exceed 2000 cm2 .

i.e. nπr2 ≤ 2000. 0.25 Marks

The objective is to maximize the surface area of the tubes in the exchanger. i.e. n×2πrL, where,

L is the length of the tube. As L is fixed, L need not be considered in the optimization.i.e.

maximize nπr. 0.5 Marks

n has to be an integer. n ≥ 1, an integer. 0.5 Marks

1
The problem can be formulated as follows:

maximize nπr

subject to nπr2 ≤ 2000

r ≥ 0.5

n ≥ 1, an integer

π = 3.14

(b) A ventilator manufacturing company can produce R units in a month, using regular pro-

duction, at a cost of B dollars per unit. It can produce additional units at C dollars per

unit by using overtime (C > B). The firm can store units from month to month at a cost

of S dollars per unit per month. The firm forecasts the demand over the next n months to

be Di , i = 1, 2, ..., n. Formulate the problem of determining the production schedule that

minimizes cost.

3 Marks

Solution

Let xi be the number of ventilators to be produced in month-i using regular production, i =

1, 2, ..., n.

Let yi be he number of ventilators to be produced in month-i, using overtime, i = 1, 2, ..., n.

Let zi be the number of ventilators to be stored from end of month-i to the next month,

i = 1, 2, ..., n. 0.5 Marks

Pn
The objective function is the total cost, i.e.f = i=1 Bxi + Cyi + Szi . 0.5 Marks

Constraints

Assuming that the sale happens at the end of a month and considering only the next n months,

there is no need to store anything at the end of n-th month, zn = 0. Also, assuming that there

is nothing stored in the first month, z1 = 0.

Regular production limit: xi ≤ R

xi , yi and zi are all non-negative numbers and integers.

2
 0 ≤ xi ≤ R, an integer, i = 1, 2, ..., n

yi ≥ 0, an integer, i = 1, 2, ..., n

zi ≥ 0, an integer, i = 1, 2, ..., n

z0 = zn = 0

1.5 Marks

Whatever produced in first month need to be sold or stored: x1 + y1 = D1 + z1 .

Whatever produced in second month plus stored in the first month need to be sold or stored

again: z1 + x2 + y2 = D2 + z2 .

Similarly, z2 + x3 + y3 = D3 + z3 .....

zn−1 + xn + yn = Dn . 0.5 Marks

Therefore, the optimization problem gets formulated as:

n
X
minimize Bxi + Cyi + Szi
i=1

subject to zi−1 + xi + yi = Di + zi , i = 1, 2, ..., n

0 ≤ xi ≤ R, an integer, i = 1, 2, ..., n

yi ≥ 0, an integer, i = 1, 2, ..., n

zi ≥ 0, an integer, i = 1, 2, ..., n

z0 = zn = 0

3
 Question 2

(a) Is the following problem a convex programming problem? Demonstrate mathematically.

maximize f (x) = x1 − 2x22

subject to g1 (x) = −x21 − x22 ≥ −9

g2 (x) = x1 + x22 ≤ 1

g3 (x) = x1 − x2 ≥ 2

3 Marks

Solution

We need to analyse the convexity of the function and the feasible region.

The Hessian of the objective function is as follows:

 
0 0
H= 
0 −4

The eigen values of the Hessian are λ = 0, −4, which means that the Hessian is negative-

semidefinite. Therefore, the function f is a concave function. 0.5 Marks

Considering g1 (x), the Hessian is  

−2 0
 
0 −2

The eigen values of the Hessian are λ = −2, −2, which means that the Hessian is negative-definite.

Therefore, the function g1 (x) is strictly concave. Therefore, the set given by g1 (x) ≥ −9 will be

a convex set. 0.5 Marks

Considering g2 (x), the Hessian is  

0 0
 
0 2

The eigen values of the Hessian are λ = 0, 2, which means that the Hessian is positive-semidefinite.

Therefore, the function g2 (x) is convex. Therefore, the set given by g2 (x) ≤ 1 will be a convex

4
set. 0.5 Marks

g3 (x) is a linear function which is both convex and concave. Therefore, the set given by g3 (x) ≥ 2

will be a convex set. 0.5 Marks

The intersection of convex sets will also be a convex set. 0.5 Marks

As the problem is maximization of a concave function upon a convex set, the problem is a convex

programming problem. 0.5 Marks

(b) Find all the stationary points of the function f (x) = (x2 − 1)4 and classify them to maxi-

mum, minimum, and saddle point.

2 Marks

Solution

The stationary points are given by f ′ (x) = 0. 0.25 Marks

Here, 4(x2 − 1)3 × 2x = 0.

Therefore, the stationary points are x = 0, x = +1, x = −1 0.75 Marks

To classify them, higher-order derivatives need to be evaluated.

f ′′ (x) = 48x2 (x2 − 1)2 + 8(x2 − 1)3 0.25 Marks

f ′′ (x = 0) = −8 < 0. i.e. x = 0 is a maximum. 0.25 Marks

f ′′ (x = +1) = 0 and f ′′ (x = −1) = 0

f ′′′ (x) = 144x(x2 − 1)2 + 192x3 (x2 − 1)

f ′′′ (x = +1) = 0 and f ′′′ (x = −1) = 0

f IV (x) = 384x4 + 144(x2 − 1)2 + 1152x2 (x2 − 1)

f IV (x = +1) = 384 > 0. n = 4 is even and therefore, x = +1 is a minimum. 0.25 Marks

f IV (x = −1) = 384 > 0. n = 4 is even and therefore, x = −1 is a minimum. 0.25 Marks

(c) Write the first three terms of the Taylor series for the function f (x) = ln(x − 1) at x = 3.

1 Mark

Solution

5
The Taylor series is given by

1
f (x) = f (x0 ) + (x − x0 )f ′ (x0 ) + (x − x0 )2 f ′′ (x0 )
2!

The first term, f (x0 ) = ln(3 − 1) = ln(2) = 0.693

0.25 Marks

1
The second term, (x − x0 )f ′ (x0 ) = (x − 3) = 0.5(x − 3)
3−1

0.25 Marks

1 1 −1
The third term, (x − x0 )2 f ′′ (x0 ) = (x − 3)2 × = 0.125(x − 3)2
2! 2! (3 − 1)2

0.5 Marks

Question 3

Air is to be compressed from p1 to p3 atm pressure in a two-stage compressor. To increase

the compression efficiency, the compressed air from the first stage of compression is cooled (it is

passed through a heat exchanger) before entering the second stage of compression. For isentropic

compression of air, the total work input to a compressor (W ) can be represented by

" (k−1)/k  (k−1)/k #
p2 p3
W = Cp T1 + −2
p1 p2

where, Cp is the specific heat of air at constant pressure, k > 1 is the ratio of specific heat at

constant pressure to specific heat at constant volume of air, and T1 is the entering gas tempera-

ture. Find analytically the intermediate pressure p2 at which cooling is required to minimize the

work input to the compressor. 4 Marks

Solution

The analytical solution is obtained by applying the necessary condition, i.e. dW/dp2 = 0.

"  (k−1)/k     (k−1)/k  

dW k−1 p2 1 1−k p3 1
= Cp T1 + c=0
dp2 k p1 p2 k p2 p2

6
Solving the above equation, we get
√
p2 = p1 p3

2 Marks
√
To check whether W is minimum at p2 = p1 p3 , second-derivative needs to be evaluated.

 
dW k−1 h (1−k)/k −1/k (k−1)/k (1−2k)/k
i
= Cp T1 (p1 ) (p2 ) − (p3 ) (p2 )
dp2 k
2
      
d W k−1 (1−k)/k −1 −1
k −1 (k−1)/k 1 − 2k ( 1−2ks −1)
= C T
p 1 (p 1 ) (p2 ) − (p3 ) (p2 ) k
dp22 k k k
2
d2 W
 
k−1 (1−3k)/2k (−k−1)/2k
= 2Cp T1 (p1 ) (p3 ) >0
dp22 p2 =√p1 p3 k

√
Therefore, W is minimum at p2 = p1 p3 . 2 Marks

Question 4

Consider the problem, maximize

f (x) = ex − 1.5x2 (1)

(a) Start with an initial interval of uncertainty [0,1] and reduce the interval of uncertainty to

less than 30% of its original size using dichotomous search with δ = 0.002. 3 Marks

(b) Determine the number of function evaluations required to reduce the interval of uncertainty

to less than 20% of its original size using (i) golden section method and (ii) Fibonacci

method. 2 Marks

Solution:

(a) The Objective function given is f (x) = ex − 1.5x2

It is also given that,

Initial interval of uncertainty, Lo = [0,1], and step size δ = 0.002. The first two experiments will

result in,
Lo Lo
x1 = 2 - 2δ , x2 = 2 + δ
2 0.5 Marks

Substituting the values of Lo and δ into the expression for x1 and x2 , we get

7
 1−0 0.002 1−0 0.002
x1 = 2 − 2 = 0.499 and x2 = 2 + 2 = 0.501. 0.5 Marks

The function value at x1 and x2 are f1 = f (x1 ) = 1.2736 and f2 = f (x2 ) = 1.2739, respectively.

Now, since f2 > f1 and the objective is to maximize, the revised interval of uncertainty, L1 is

[0.499, 1]. 0.5 Marks

The next two experiments x3 and x4 in a similar manner can be calculated as

1−0.499 1−0.499
x3 = 0.499 + 2 − 0.001 = 0.7485 and x4 = 0.499 + 2 − 0.001 = 0.7505.

0.5 Marks

The function value at x3 and x4 are f3 = 1.2735 and f4 = 1.2732. The function value at x3 is

greater than x4 . Therefore, the reduced interval of uncertainty, L2 is [0.499, 0.7505].

0.5 Marks

This interval of uncertainty is 0.2515/1 i.e. 25% of the initial interval of uncertainty. So, no

further experiments are required. 0.5 Marks

(b) (i) The final and initial interval of uncertainty for golden section method are related as

Ln
= (0.618)n−1 (2)
Lo

The optimal solution is assumed to be at the mid-point of the interval of uncertainty. Therefore,

the relation in Equation 2 can be used as described below to reduce the size of interval to less

than 20%:

20
(0.618)n−1 < 100

0.5 Marks

ln(0.618)n−1 < ln(0.2)

(n − 1) × ln(0.618) < ln(0.2)

(n − 1) × (−0.4823) < −1.6094

(n − 1) × (0.4823) > 1.6094

8
 n − 1 > 3.3369

n > 4.3369 ≈ 5

0.5 Marks

(ii) For the Fibonacci method, the relation between the final and initial interval of uncertainty

is given as,

Ln 1
= (3)
Lo Fn
Ln
As done for the previous section, the Lo < 0.2 in Equation 3, the number of function evaluation

required are,

1 20
Fn < 100

0.5 Marks

Fn > 100/20

Fn > 5

Now, the Fibonacci series is given as, 1, 1, 2, 3, 5, 8, 13, ...., and the next value of Fn is 8. Thus

n = 5 is the number of function evaluation required by this method.

0.5 Marks

Question 5

Consider the problem, minimize

f (x) = (x − 1)4 (4)

(a) Carry out the initial and one additional iteration of Quasi-Newton method to solve the

problem, using x0 = 0 and step step size, h = 0.5. 2 Marks

(b) Solve using secant method starting with an initial interval of uncertainty [0.95, 1.05] and

convergence criteria, ϵ = 0.001. 2 Marks

9
Solution:

(a) For the Quasi-Newton method,

h × [f (xi + h) − f (xi − h)]

xi+1 = xi − (5)
2[f (xi + h) − 2 × f (xi ) + f (xi − h)]

1 Marks

For the initial iteration,

x0 = 0, f (x0 ) = 1, f (xo + 0.5) = 0.0625, f (x0 − 0.5) = 5.0625, and using Equation 5, we get:

0.5×[0.0625−5.0625]
x1 = xo − 2×[0.0625−2×1+5.0625]

x1 = 0.4

0.5 Marks

For the additional iteration,

x1 = 0.4, f (x1 ) = 0.1296, f (x1 + 0.5) = 0.0001, f (x1 − 0.5) = 1.4641, and using Equation 5, we

get:

0.5×[0.0001−1.4641]
x2 = x1 − 2×[0.0001−2×0.1296+1.4641]

x2 = 0.7037

0.5 Marks

(b) For the Secant method, the formula is

f ′ (a) × (b − a)
xi = a − (6)
f ′ (b) − f ′ (a)

0.5 Marks

Here, a = 0.95 and b = 1.05, and f ′ (a) = −0.0005, f ′ (b) = 0.0005. 0.5 Marks

Substituting all of these values in Equation 6, we get:

−0.0005×(1.05−0.95)
x1 = 0.95 − 0.0005−(−0.0005)

x1 = 1

10
 0.5 Marks

and |f ′ (x1 )| = 0 which is less than ϵ = 0.001 and hence, the iteration is terminated.

0.5 Marks

Question 6

To find the maximum of the following function, perform one iteration of Newton’s method, using

x0 = −2
10 3 1 4 4 5 1 6
f (x) = 1 − 8x + 2x2 − x + x + x − x (7)
3 4 5 6

2 Marks

Solution:

The formula for Newton’s method is,

f ′ (xk )
xk+1 = xk − (8)
f ′′ (xk )

0.5 Marks

The first and second derivative of the function are

f ′ (x) = 8 − 4x + 10x2 − x3 − 4x4 + x5 (9)

f ′′ (x) = −4 + 20x − 3x2 − 16x3 + 5x4 (10)

The initial guess is given as, x0 = −2. The first and second derivative value at this point are

f ′ (x0 ) = 32 and f ′′ (x0 ) = −152. 0.5 + 0.5 Marks

Substituting these value in Equation 8, we get

32
x1 = −2 − −152 = −1.7895 0.5 Marks

Question 7

Approximate the function, f (x) = x5 − 5x3 − 20x + 5 as

(a) Quadratic polynomial using, x = 0, 2, 4 2 Marks

11
 (b) Cubic polynomial using, x = 0, 4 2 Marks

Solution:

(a) For the quadratic approximation, assume the given function can be represented as,

f (x) = ax2 + bx + c (11)

0.5 Marks

Now, to evaluate the value of a, b, and c, three equations will be required. These equations will

come from three initial guess of x, given as x1 = 0, x2 = 2, and x3 = 4. The function value at

these three points are f1 = f (x1 ) = 5, f2 = f (x2 ) = −43, and f3 = f (x3 ) = 629. Substituting

these values in Equation 11, we get

c=5

4a + 2b + 5 = −43

16a + 4b + 5 = 629

0.5 Marks

Solving all the three equations, simultaneously, we get

a = 90, b = −204, and c = 5. Therefore, Equation 11 becomes:

f (x) = 90x2 − 204x + 5 (12)

1 Marks

(b) For the Cubic approximation, the given function is assumed to be represented by

f (x) = ax3 + bx2 + cx + d (13)

0.5 Marks

The evaluation of values of a, b, c, d will require four equations. Now, since, there are two points

given, the four equations will involve two equation from function value and two from the first

12
derivative of the function. The first derivative of the original function is

f ′ (x) = 3ax2 + 2bx + c (14)

Now, substituting x = 0, 4 in Equation 13 and Equation 14, we get:

d=5

c = −20

64a + 16b − 80 + 5 = 629

48a + 8b − 20 = 1020

1 Marks

Solving all the above equations simultaneously, we get:

a = 43, b = −128, c = −20, and d = 5, respectively. Thus, the approximated cubic polynomial

is

f (x) = 43x3 − 128x2 − 20x + 5 (15)

0.5 Marks

13