CHAPTER

5
More on Optimum Design Concepts:
Optimality Conditions

5.1_________________________________________________________________________________
Answer True or False.
1. A convex programming problem always has a unique global minimum point. False

2. For a convex programming problem, KKT necessary conditions are also sufficient. True

3. The Hessian of the Lagrange function must be positive definite at constrained minimum

points. False

4. For a constrained problem, if the sufficiency condition of Theorem 5.2 is violated, the

candidate point x∗ may still be a minimum point. True

5. If the Hessian of the Lagrange function at x∗, ∇2L(x∗), is positive definite, the optimum

design problem is convex. False

6. For a constrained problem, the sufficient condition at x∗ is satisfied if there are no

feasible directions in a neighborhood of x∗ along which the cost function reduces. True

5.2 _________________________________________________________________________________
Formulate the problem of Exercise 4.84. Show that the solution point for the problem is not a regular
point. Write KKT conditions for the problem, and study the implication of the irregularity of the
solution point. Refer to solution of Exercises 4.84:

A refinery has two crude oils:

1. Crude A costs $120/barrel (bbl) and 20,000bbl are available.

2. Crude B costs $150/bbl and 30,000 are available.

The company manufactures gasoline and lube oil from the crudes. Yield and sale price barrel of the
product and markets are shown in Table E2.2. How much crude oils should the company use to
maximize its profit? Formulate the optimum design problem.

Table E2.2 Data for Refinery Operation

Arora, Introduction to Optimum Design, 4e 5-1
 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Yield/bbl Sale Price

Product Crude A Crude B per bbl ($) Market (bbl)
Gasoline 0.6 0.8 200 20,000
Lube oil 0.4 0.2 450 10,000

Solution

Note : g4 = B − 30000, g5 = −A and g6 = −B are not shown on the graph.

According to the graphical solution, the point P (20000, 10000) is the minimum point with
f* = -4,600,000.

Referring to the formulation in Exercise 2.2, we have

Minimize f = - 180A - 100B
subject to:
g1 = 0.6 A + 0.8B − 20,000 ≤ 0 , (gasoline market)
g 2 = 0.4 A + 0.2 B − 10,000 ≤ 0 (lube oil market)
g3 = A − 20,000 ≤ 0
g4 = B − 30,000 ≤ 0
g5 =− A ≤ 0; g 6 =− B ≤ 0
( )
( 180 A − 100 B ) + u1 ( 0.6 A + 0.8B − 20,000 ) + s12 + u2 0.4 A + 0.2 B − 10,000 + s22 +
L =−

Arora, Introduction to Optimum Design, 4e 5-2

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

( ) ( ) ( ) (
u3 A − 20,000 + s32 + u 4 B − 30,000 + s 42 + u 5 − A + s52 + u 6 − B + s62 )
∂L ∂A =−180 + 0.6u1 + 0.4u2 + u3 − u5 = 0
∂L ∂B =−100 + 0.8u1 + 0.2u2 + u 4 − u6 =0
g i + s=
2
i 0, ui s=i 0, ui ≥ 0; =i 1 to 6
There are 64 cases because we have six inequality constraints. We shall examine three cases. Two
of them yield solutions.
Case 1. u2 = u4 = u5 = u6 = 0, s1 = s3 = 0 . The solution is given as A = 20,000, B = 10,000,
u 1 = 125, u 3 = 105; s 2 = 0, so g 2 is also active.
Case 2. u= 3 u=4 u=5 u=
6 0, s=1 s=2 0 . The solution is given as A = 20,000, B = 10,000, u 1 =

20, u 2 = 420;s3 = 0, so g 3 is also active.

Case 3. u1 = u4 = u5 = u6 = 0, s2 = s3 = 0 . The solution is given as A = 20,000, B = 10,000,
u = 500, u3 = − 20 (violation); so this case does not give a solution.
2

The candidate minimum point derived in Case 1 and Case 2 can be verified graphically. It is seen
that (20000, 10000 ) is the optimum point. At this optimum point, g1 , g 2 and g 3 (constraints on
gasoline and lube oil markets, and limit on crude A) are all active. The optimum cost is
f = -4,600,000. . The optimum point is irregular, since there are three active constraints and two
design variables. The Lagrange multipliers are not unique.

Arora, Introduction to Optimum Design, 4e 5-3

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.4 _________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8𝑥𝑥1

subject to 𝑥𝑥1 + 𝑥𝑥2 − 4 = 0

Solution
Minimize f (x1 , x2 ) = 4 x12 + 3x22 − 5 x1 x2 − 8 x1 subject to h = x1 + x2 − 4 = 0 ;
L = 4 x12 + 3x22 − 5 x1 x2 − 8 x1 + v(x1 + x2 − 4) ;
The necessary conditions give
∂ L ∂x1 = 8 x1 − 5 x2 − 8 + v = 0; ∂ L ∂x2 = 6 x2 − 5 x1 + v = 0; h = x1 + x2 − 4 = 0
The solution of these equations is x1 = 13 / 6, x 2 = 11 / 6, v = −1 / 6.
Therefore, (13/6, 11/6) is a KKT point; f = −25 / 3
Check for regularity: Ñh = (1, 1). Since Ñh is the only vector, regularity of feasible points is
satisfied.

Referring to Exercises 4.43/4.97, the point satisfying the KKT necessary conditions is
x 1 = 2.166667, x 2 = 1.833333, v = − 0.166667, f = -8.3333. This point also satisfies second order
sufficiency condition for a local minimum. Hessian of the Lagrangian function is calculated as
8 −5
𝐇𝐇 = � �
−5 6
The Hessian is positive definite since 𝑀𝑀1 = 8 > 0 𝑎𝑎𝑎𝑎𝑎𝑎 𝑀𝑀2 = 8 × 6 − (−5) × (−5) = 23 > 0.
Therefore it is positive definite for those d that satisfy ∇ℎ ∙ 𝐝𝐝 = 0. Therefore the KKT point satisfies
second order necessary and sufficient conditions for a local minimum point.

The problem is solved graphically in Exercise 3.12. The graphical solution confirms this conclusion.

Arora, Introduction to Optimum Design, 4e 5-4

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.5 _________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8𝑥𝑥1

subject to 𝑥𝑥1 + 𝑥𝑥2 − 4 = 0

Solution
Maximize F (x1 , x 2 ) = 4 x12 + 3x 22 − 5 x1 x 2 − 8 x1 subject to h = x1 + x2 − 4 = 0 ;
L = −4 x12 − 3x 22 + 5 x1 x 2 + 8 x1 + v(x1 + x 2 − 4) ;
The necessary conditions give
∂ L ∂x1 =−8 x1 + 5 x2 + 8 + v =0; ∂ L ∂x2 =−6 x2 +5 x1 + v =0; h =x1 + x2 − 4 =0
The solution of these equations is x1 = 13 / 6, x 2 = 11 / 6, v = 1 / 6.
Therefore, (2.166667, 1.833333) is a KKT point; F = −25 / 3
Check for regularity: Ñh = (1, 1). Since Ñh is the only vector, regularity of feasible points is
satisfied.

Referring to Exercises 4.44/4.98, the point satisfying the KKT necessary conditions is
x 1 = 2.166667, x 2 = 1.833333, v = 0.166667. However this point violates second order necessary
condition. Hessian of the Lagrangian function is calculated as
−8 5
𝐇𝐇 = � �
5 −6
The Hessian is negative definite since 𝑀𝑀1 = −8 < 0 𝑎𝑎𝑎𝑎𝑎𝑎 𝑀𝑀2 = −8 × −6 − (5) × (5) = 23 > 0.
Therefore it is negative definite for those d that satisfy ∇ℎ ∙ 𝐝𝐝 = 0. Therefore the point violates
the second order necessary conditions for a local minimum of f(x), or for the local maximum of F(x).
There is no maximum point for F(x).

The problem is solved graphically in Exercise 3.12. The graphical solution confirms this conclusion.

Arora, Introduction to Optimum Design, 4e 5-5

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.6 _________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 2)2 + (𝑥𝑥2 + 1)2

subject to 2𝑥𝑥1 + 3𝑥𝑥2 − 4 = 0

Solution
Minimize f ( x1 , x2 ) = ( x1 − 2 ) + ( x2 + 1) subject to h= 2 x1 + 3 x2 − 4 =
2 2
0;
L = ( x1 − 2 ) + ( x2 + 1) + v ( 2 x1 + 3 x2 − 4 ) ;
2 2

The KKT necessary conditions give

∂ L ∂x1 = 2( x1 − 2) + 2v = 0; ∂ L ∂x2 = 2( x2 +1) + 3v = 0; h = 2 x1 + 3x2 − 4 = 0
The solution of these equations is x1 =
32 /13, x2 = −4 /13, v =
−6 /13.
Therefore, (32/13, -4/13) is a KKT point; f = 9 / 13
Check for regularity: Ñh = (2, 3). Since Ñh is the only vector, regularity of feasible point is
satisfied.

Referring to Exercises 4.45/4.99, the point satisfying the KKT necessary conditions is
x 1 = 2.46154, x 2 = -0.307692, v = -0.46154, f = 0.69231. The point satisfies second order
sufficiency condition.
5.7 _________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 9𝑥𝑥22 + 6𝑥𝑥2 − 4𝑥𝑥1 + 13

subject to 𝑥𝑥1 − 3𝑥𝑥2 + 3 = 0

Solution
Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 9𝑥𝑥22 + 6𝑥𝑥2 − 4𝑥𝑥1 + 13 subject to h = 𝑥𝑥1 − 3𝑥𝑥2 + 3 = 0;
𝐿𝐿 = 4𝑥𝑥12 + 9𝑥𝑥22 + 6𝑥𝑥2 − 4𝑥𝑥1 + 13 + 𝜈𝜈(𝑥𝑥1 − 3𝑥𝑥2 + 3)
The KKT necessary conditions give
∂L ∂L ∂L
= 8𝑥𝑥1 − 4 + 𝜈𝜈 = 0, = 18𝑥𝑥2 + 6 − 3𝜈𝜈 = 0, = 𝑥𝑥1 − 3𝑥𝑥2 + 3 = 0
∂𝑥𝑥1 ∂𝑥𝑥2 ∂𝜈𝜈
The solution of these equations is 𝑥𝑥1 = −0.4, 𝑥𝑥2 = 2.6/3, 𝜈𝜈 = 7.2
Therefore, (-0.4, 2.6/3) is a KKT point; 𝑓𝑓 = 27.2
Check for regularity: Ñh = (1,-3). Since Ñh is the only vector, regularity of feasible points is
satisfied.

Referring to Exercises 4.46/4.100, the point satisfying the KKT necessary conditions is
x 1 = -0.4, x 2 = 0.866667, v = 7.2, f = 27.2. The point satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-6

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.8 _________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 + 2)2 + (𝑥𝑥3 − 2)2

subject to 2𝑥𝑥1 + 3𝑥𝑥2 − 1 = 0
𝑥𝑥1 + 𝑥𝑥2 + 2𝑥𝑥3 − 4 = 0

Solution
f ( x1 , x2 ) = ( x1 − 1) + ( x2 + 2 ) + ( x3 − 2 ) h1= 2 x1 + 3 x2 − 1 ;
2 2 2
Minimize subject to h2=
x1 + x2 + 2 x3 − 4
L = ( x1 − 1) + ( x2 + 2 ) + ( x3 − 2 ) + v1 ( 2 x1 + 3 x2 − 1) + v2 ( x1 + x2 + 2 x3 − 4 ) ;
2 2 2

The KKT necessary conditions give

∂ L ∂x1 = 2( x1 − 1) + 2v1 + v2 = 0; ∂ L ∂x2 = 2( x2 +2) + 3v1 + v2 = 0; ∂ L ∂x3 = 2( x3 -2) + 2v2 = 0;
h1 = 2 x1 + 3 x2 − 1 = 0; h 2 = x1 + x2 + 2 x3 − 4 = 0
x1 =
1.71698, x2 =
The solution of these equations is −0.81132, x3 =
1.547170057
v1 =
−0.943396132, v2 =
0.452829943
Therefore, (1.71698066, -0.811320724, 1.547170057) is a KKT point; f = 2.1318
Check for regularity: Gradients of the constraints are linearly independent; therefore the point is
a regular point of the feasible set.

Referring to Exercises 4.47/4.101, the point satisfying the KKT necessary conditions is
x 1 = 1.71698, x 2 = -0.81132, x 3 = 1.54714, v1 = -0.943396, v2 = 0.4528299, f = 2.11318. The point
satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-7

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.9 _________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 9𝑥𝑥12 + 18𝑥𝑥1 𝑥𝑥2 + 13𝑥𝑥22 − 4

subject to 𝑥𝑥21 + 𝑥𝑥22 + 2𝑥𝑥1 − 16 = 0

Solution
Minimize ( )
f x1 , x2 = 9 x12 + 18 x1 x2 + 13x22 − 4 subject to h = x12 + x22 + 2 x1 − 16 = 0
L ( x1 , x2 , v )= 9 x12 + 18 x1 x2 + 13 x22 − 4 + v ( x12 + x22 + 2 x1 − 16 )
∂L ∂x=
1 18 x1 + 18 x2 + 2vx1 + 2v= 0; ∂L ∂x2= 18 x1 + 26 x2 + 2vx=
1 0;
h = x12 + x22 + 2 x1 − 16 = 0
These equations are nonlinear, which can be solved numerically. Using any nonlinear equation
solver, we can find the following KKT points:
1. x1 = 1.5088, x2 = 3.2720, v = −17.151503, f = 244.528
2. x1 = 2.5945, x2 = −2.0198, v = −1.4390, f = 15.291
3. x1 = −3.630, x2 = −3.1754, v = −23.2885, f = 453.154
4. x1 = −3.7322, x2 = 3.0879, v = −2.1222, f =37.877

Check for regularity: = Ñh ( 2 x1 + 2, 2 x2 ) . Since Ñh is the only vector, regularity of feasible

points is satisfied for each KKT point.

Refer to Exercises 4.48/4.102:

The points x1 = −3.7322, x2 = 3.0879 ; and x1 = 2.5945, x2 = −2.0198 satisfy second order
sufficiency conditions and are isolated local minimum points. The other two KKT points violate the
second order necessary conditions.

Arora, Introduction to Optimum Design, 4e 5-8

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.10 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 − 4 = 0

Solution
Minimize f = ( x1 − 1) + ( x2 − 1) subject to h = x1 + x2 − 4 = 0
2 2

L = (x1 − 1) + (x2 − 1) + v(x1 + x2 − 4) ; the KKT necessary conditions are

2 2

∂L ∂x1 = 2 ( x1 − 1) + v = 0; ∂L ∂x2 = 2 ( x2 − 1) + v = 0; h = x1 + x2 − 4 = 0
Solution of these equations is x1 = 2, x2 = 2, v = −2 . Therefore, (2, 2 ) is a KKT point; f = 2.
Check for regularity: Ñh = (1, 1) . Since Ñh is the only vector, regularity of feasible points is
satisfied.

Refer to Exercises 4.49/4.103

(2, 2) is a KKT point; f = 2. The point satisfies second order necessary conditions.

Arora, Introduction to Optimum Design, 4e 5-9

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.11 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8

subject to 𝑥𝑥1 + 𝑥𝑥2 = 4

Solution
Minimize f = 4 x1 + 3 x2 − 5 x1 x2 − 8 subject to
2 2
h = x1 + x2 − 4 = 0
L= 4 x12 + 3x22 − 5 x1 x2 − 8 + v ( x1 + x2 − 4 ) ; the KKT necessary conditions are
∂L ∂x1 = 8 x1 − 5 x2 + v = 0; ∂L ∂x2 = 6 x2 − 5 x1 + v = 0; h = x1 + x2 − 4 = 0
Solution of these equations is x1 = 11/ 6, x2 = 13 / 6, v = −23 / 6 .
Therefore, (11/ 6, 13 / 6 ) is a KKT point; f* = -1/3.
Check for regularity: Ñh = (1, 1) . Since Ñh is the only vector, regularity of feasible point is
satisfied.

Referring to Exercises 4.51/4.105, the point satisfying the KKT necessary conditions is
x 1 = 1.83333, x 2 = 2.16667, v = -3.83333, f = -0.33333. The point satisfies second order sufficiency
condition. The sufficiency check is same as in Exercise 5.5.

Arora, Introduction to Optimum Design, 4e 5-10

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.12 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8

subject to 𝑥𝑥1 + 𝑥𝑥2 = 4

Solution
Minimize f =
−4 x1 − 3x2 + 5 x1 x2 + 8 subject to
2 2
h = x1 + x2 − 4 = 0
L =−4 x12 − 3x22 + 5 x1 x2 + 8 + v ( x1 + x2 − 4 )
The KKT necessary conditions are
∂L ∂x1 =−8 x1 + 5 x2 + v =0; ∂L ∂x2 =−6 x2 + 5 x1 + v =0; h =x1 + x2 − 4 =0
=
Solution of these equations =
is x1 11/ 6, x2 13= / 6, v 23 / 6 .
Therefore, (11/ 6, 13 / 6 ) is a KKT point; F = -1/3.
Check for regularity: Ñh = (1, 1) . Since Ñh is the only vector, regularity of feasible point is
satisfied.

Referring to Exercises 4.52/4.106, the point satisfying the KKT necessary conditions is
x 1 = 1.83333, x 2 = 2.16667, v = 3.83333, f = 0.33333. This point violates second order necessary
condition. The check for the second order conditions is the same as in Exercise 5.6. F(x) has no
maximum points.

Arora, Introduction to Optimum Design, 4e 5-11

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.13 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8

subject to 𝑥𝑥1 + 𝑥𝑥2 ≤ 4

Solution
Minimize f =
−4 x1 − 3x2 + 5 x1 x2 + 8 subject to
2 2
g = x1 + x2 − 4 ≤ 0
L =−4 x12 − 3 x22 + 5 x1 x2 + 8 + u ( x1 + x2 − 4 + s 2 ) ; the KKT necessary conditions are
∂L ∂x1 =−8 x1 + 5 x2 + u =0; ∂L ∂x2 =−6 x2 + 5 x1 + u =0;
∂L ∂u = x1 + x2 − 4 + s 2 = 0; ∂L ∂s = 2us = 0

Case 1. u = 0; gives a KKT point as (0, 0); F * = −8 .

Case 2. s = 0 (or g = 0); gives a KKT point as (11/6, 13/6); u* = 23 / 6, F * = −1 / 3 .
Check for regularity: Ñg = (1, 1) . Since there is only one constraint, regularity is satisfied.

Referring to Exercises 4.54/4.107, the points satisfying the KKT necessary conditions are:
x 1 = 0, x 2 = 0, u = 0; x 1 = 1.83333, x 2 = 2.16667, u = 3.83329. Both points violate second order
necessary condition. The check for the second order conditions is the same as in Exercise 5.6. F(x)
has no maximum points.

Arora, Introduction to Optimum Design, 4e 5-12

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.14 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8

subject to 𝑥𝑥1 + 𝑥𝑥2 ≤ 4

Solution
Minimize f = 4 x1 + 3 x2 − 5 x1 x2 − 8 subject to
2 2
g = x1 + x2 − 4 ≤ 0
L= 4 x12 + 3 x22 − 5 x1 x2 − 8 + u ( x1 + x2 − 4 + s 2 ) ; the KKT necessary conditions are
∂L ∂x1 = 8 x1 − 5 x2 + u = 0; ∂L ∂x2 = 6 x2 − 5 x1 + u = 0;
∂L ∂u = x1 + x2 − 4 + s 2 = 0; ∂L ∂s = 2us = 0

Case 1. u = 0; gives a KKT point as (0, 0); f ( x∗ ) = −8 .

Case 2. s = 0 ; gives no candidate point. (u < 0)
Check for regularity: Ñg = (1, 1) . Since there is only one constraint, regularity is satisfied.

Referring to Exercises 4.55/4.108, the point satisfying the KKT necessary condition are:
x 1 = 0, x 2 = 0, u = 0. The point satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-13

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.15 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8𝑥𝑥1

subject to 𝑥𝑥1 + 𝑥𝑥2 ≤ 4

Solution
Minimize f =
−4 x1 − 3x2 + 5 x1 x2 + 8 x1 subject to
2 2
g = x1 + x2 − 4 ≤ 0
L= 4 x12 + 3 x22 − 5 x1 x2 − 8 x1 + u ( x1 + x2 − 4 + s 2 ) ; the KKT necessary conditions are
∂L ∂x1 =−8 x1 + 5 x2 + 8 + u =0; ∂L ∂x2 =−6 x2 + 5 x1 + u =0;
∂L ∂u = x1 + x2 − 4 + s 2 = 0; ∂L ∂s = 2us = 0

Case 1. u = 0; gives a KKT point as (48/23, 40/23); F ( x∗ ) = −192 / 23 .

Case 2. s = 0 ; gives a KKT point as (13/6, 11/6); u* = 1/6, F ( x∗ ) = −8.33333 .
Check for regularity: Ñg = (1, 1) . Since there is only one constraint, regularity is satisfied.

Referring to Exercises 4.56/4.109, the points satisfying the KKT necessary conditions are:
x 1 = 2.08696, x 2 = 1.73913, u = 0; x 1 = 2.16667, x 2 = 1.83333, u = 0.16667. Both points violate
second order necessary condition.

Arora, Introduction to Optimum Design, 4e 5-14

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.16 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 ≥ 4
𝑥𝑥1 − 𝑥𝑥2 − 2 = 0

Solution
Minimize f ( x ) = ( x1 − 1) + ( x2 − 1) ; subject to
2 2
h = x1 − x2 − 2 = 0; g =− x1 − x2 + 4 ≤ 0 .
L= ( x1 − 1) + ( x2 − 1) + v ( x1 − x2 − 2 ) + u ( − x1 − x2 + 4 + s 2 )
2 2

∂L ∂x1= 2 ( x1 − 1) + v − u= 0; ∂L ∂x2= 2 ( x2 − 1) − v − u= 0
h = x1 − x2 − 2 = 0; -x1 − x2 + 4 + s 2 = 0; us = 0, u ≥ 0.
Case 1. u = 0; no candidate minimum.
Case 2. s = 0; gives ( 3, 1) as a KKT point with v =
−2, u =
2, f =
4.
h (1, − 1) and Ñg =
Since Ñ= ( −1, -1) are linearly independent, regularity is satisfied.
Referring to Exercises 4.57/4.110, the point satisfying the KKT necessary condition is:
x 1 = 3, x 2 = 1, ν = -2, u = 2. The point satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-15

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.17 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 = 4
𝑥𝑥1 − 𝑥𝑥2 − 2 ≥ 0

Solution
Minimize f ( x ) = ( x1 − 1) + ( x2 − 1) ; subject to
2 2
h = x1 + x2 − 4 = 0; g =− x1 + x2 + 2 ≤ 0 .
L= ( x1 − 1) + ( x2 − 1) + v ( x1 + x2 − 4 ) + u ( − x1 + x2 + 2 + s 2 )
2 2

∂L ∂x=
1 2 ( x1 − 1) + v − u= 0; ∂L ∂x2= 2 ( x2 − 1) + v + u= 0
h = x1 + x2 − 4 = 0; -x1 + x2 + 2 + s 2 = 0; us = 0, u ≥ 0.
Case 1. u = 0; no candidate minimum.
Case 2. s = 0; gives ( 3, 1) as a KKT point with v =
−2, u =
2, f =
4.

Since Ñh = (1, 1) and Ñg = ( −1, 1) are linearly independent, regularity is satisfied.

Referring to Exercises 4.58/4.111, the point satisfying the KKT necessary condition is:
x 1 = 3, x 2 = 1, ν = -2, u = 2. The point satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-16

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.18 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 ≥ 4
𝑥𝑥1 − 𝑥𝑥2 ≥ 2

Solution
Minimize f ( x ) = ( x1 − 1) + ( x2 − 1) ; subject to
2 2
g1 =− x1 − x2 + 4 ≤ 0; g 2 =− x1 + x2 + 2 ≤ 0 .
L= ( x1 − 1) + ( x2 − 1) + u1 ( − x1 − x2 + 4 + s12 ) + u2 ( − x1 + x2 + 2 + s22 )
2 2

∂L ∂x1= 2 ( x1 − 1) − u1 − u2= 0; ∂L ∂x2= 2 ( x2 − 1) − u1 + u2= 0

− x1 − x2 + 4 +=
s12 0; -x1 + x2 + 2 +=
s22 0; u=
1 s1 0, u2=
s2 0, u1 , u2 ≥ 0.
Case 1. u 1 = 0, u 2 = 0; no candidate minimum.

Case 2. u 1 = 0, s 2 = 0; no candidate minimum.

Case 3. s 1 = 0, u 2 = 0; no candidate minimum.
Case 4. s 1 = 0, s 2 = 0; gives ( 3, 1) as a KKT point with=
u1 2,=
u2 2,=
f 4.
Since Ñg1 =( −1, − 1) and Ñg 2 = ( −1, 1) are linearly independent, regularity is satisfied.
Referring to Exercises 4.59/4.113, the point satisfying the KKT necessary condition is:
x 1 = 3, x 2 = 1, u 1 = 2, u 2 = 2. The point satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-17

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.19 _______________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥, 𝑦𝑦) = (𝑥𝑥 − 4)2 + (𝑦𝑦 − 6)2

subject to 12 ≥ 𝑥𝑥 + 𝑦𝑦
𝑥𝑥 ≥ 6, 𝑦𝑦 ≥ 0

Solution
Minimize f ( x, y ) = ( x − 4 ) + ( y − 6 ) ;
2 2

subject to g1 = x + y − 12 ≤ 0;
g 2 =− x + 6 ≤ 0 ;
g3 =− y ≤ 0
L =( x − 4 ) + ( y − 6 ) + u1 ( x + y − 12 + s12 ) + u2 ( − x + 6 + s22 ) + u3 ( − y + s32 )
2 2

∂L ∂x= 2 ( x − 4 ) + u1 − u=
2 0;
y 2 ( y − 6 ) + u1 − u=
∂L ∂= 3 0
x + y − 12 + s12 =0;
− x + 6 + s22 =0;
− y + s32 =0
=
u1s1 0,= =
u2 s2 0, u3 s3 0
u1 , u2 , u3 ≥ 0.

Case 1. u=1 u=
2 u=3 0 ; no candidate minimum.
Case 2. u=1 u=
2 0, s=3 0 ; no candidate minimum.
Case 3. u=1 u=
3 0, s=2 0 ; gives ( 6, 6 ) as a KKT=
point with u2 4,=
s1 =0, s3 =6, f 4 .
Case 4. u=
2 u=3 0, s=1 0 ; no candidate minimum.
Case 5. u=1 0, s=2 s=
3 0 ; no candidate minimum.
Case 6. u=2 0, s=1 s= 3 0 ; no candidate minimum.
Case 7. u=3 0, s=1 s= 2 0 ; gives ( 6, 6 ) as a KKT point with=
u1 0,=
u2 4,=
s3 6,=
f 4.
Case 8. s=
1 s=2 s=3 0 ; no candidate minimum.

Referring to Exercises 4.60/4.108, the point satisfying the KKT necessary condition is:
x = 6, y = 6, u 1 = 0, u 2 = 4, u 3 = 0. The point satisfies second order sufficiency condition.

Arora, Introduction to Optimum Design, 4e 5-18

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.20 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 2𝑥𝑥1 + 3𝑥𝑥2 − 𝑥𝑥13 − 2𝑥𝑥22

subject to 𝑥𝑥1 + 3𝑥𝑥2 ≤ 6
5𝑥𝑥1 + 2𝑥𝑥2 ≤ 10
𝑥𝑥1 , 𝑥𝑥2 ≥ 0

Solution
Minimize f ( x ) = 2 x1 + 3x2 − x13 − 2 x22 ;
subject to g1 = x1 + 3x2 − 6 ≤ 0;
g2 =5 x1 + 2 x2 − 10 ≤ 0; g3 =− x1 ≤ 0; g 4 =− x2 ≤ 0;
( ) ( ) (
L = 2 x1 + 3x2 − x13 − 2 x22 + u1 x1 + 3 x2 − 6 + s12 + u2 5 x1 + 2 x2 − 10 + s22 )
( ) (
+ u3 − x1 + s32 + u4 − x2 + s42 )
∂L ∂x1 = 2 − 3x12 + u1 + 5u2 − u3 = 0;
∂L ∂x2 = 3 − 4 x2 + 3u1 + 2u2 − u4 = 0;
x1 + 3x 2 − 6 + s12 = 0 ; 5 x1 + 2 x 2 − 10 + s 22 = 0 ; − x1 + s32 = 0; − x 2 + s 42 = 0 ;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).
Case 1. u1 = u2 = u3 = u4 = 0 .There are two possible solution points: (− 0.816, 0.75) and
(0.816, 0.75) . For (− 0.816, 0.75) , g 3 = 0.816 > 0 (violation). For (0.816, 0.75) , g 1 = − 2.934 < 0,
g 2 = − 4.42 < 0, g 3 = − 0.816 < 0, g 4 = − 0.75 < 0. All the KKT conditions are satisfied; therefore
(0.816,0.75) is a KKT point ( f = 2.214 ).
Case 2. u1 = u2 = u3 = 0, s4 = 0 . g 4 =0 → x2 =0 ; x = ± 0.816, u 4 = 3 > 0. x1 =
1
− 0.816 → g3 > 0
(violated). At x 1 = 0.816 and x 2 = 0, g1 = −5.184 < 0, g 2 = −5.92 < 0. All the KKT conditions
are satisfied; therefore (0.816, 0 ) is a KKT point (f = 1.0887).
Case 3. u1 = u2 = u4 = 0, s3 = 0 . KKT point: (0, 0.75) , u 3 = 2, f = 1.125.
Case 4. u1 = u3 = u4 = 0, s2 = 0 . Candidate points: (− 9.8407, 1.2317 ) and (1.5073, 1.2317 ) ; first
point violates g 3 ; the second point is a KKT point with u 2 = 0.9632; f = 0.251.
Case 5. u2 = u3 = u4 = 0, s1 = 0. Candidate points: (− 1.821, 1.655) and (1.0339, 1.655) ; first point
violates g 3 ; second is a KKT point with u 1 = 1.2067; f = 0.4496.
Case 6. u1 = u2 = 0, s3 = s4 = 0; (0, 0) is a KKT point with u 3 = 2 and u 4 = 3; f = 0.
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives (2, 0) as a KKT point with u 2 = 2, u 4 = 7; f = − 4.

Arora, Introduction to Optimum Design, 4e 5-19

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 8. u1 = u4 = 0, s2 = s3 = 0; gives no candidate point.

Case 9. u2 = u3 = 0, s1 = s4 = 0; gives no candidate point.
Case 10. u=
2 u= 4 0, s= 1 s= 3 0; gives (0, 2 ) as a KKT point with u1 = 5 3 , u3 = 11 3 , f = −2 .

Case 11. u3 = u4 = 0, s1 = s2 = 0; gives (1.386, 1.538) as a KKT point with

u1 = 0.633, u 2 = 0.626; f = − 0.007388 .

Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.

Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.
Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.
Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.
Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

From the above investigation, Cases 1, 2, 3, 4, 5, 6, 7, 10, 11 generate KKT points.

Check for regularity: For cases 1, 2, 3, 4 and 5, there is only one active constraint, so regularity is
satisfied. For case 6, Ñg3 = ( 0, − 1) . Since
( −1, 0 ) , Ñg 4 = Ñg3 and Ñg 4 are linearly independent,
regularity is satisfied. For case 7, Ñg=2 ( 5, 2 ) , Ñg=
4 ( 0, − 1) . Since Ñg 2 and Ñg 4 are linearly

independent, regularity is satisfied. For case 10, Ñg1 = (1, 3) , Ñg3 = ( −1, 0 ) . Since Ñg1 and Ñg3
are linearly independent, regularity is satisfied. For case 11, Ñg1 = (1, 3) , Ñg 2 = ( 5, 2 ) . Since Ñg1
and Ñg 2 are linearly independent regularity is satisfied.

Referring to Exercises 4.61/4.114, the points satisfying the KKT necessary conditions are

x1∗ x2∗ u1∗ u2∗ u3∗ u4∗

(1) 0.816 0.75 0 0 0 0
(2) 0.816 0 0 0 0 3
(3) 0 0.75 0 0 2 0
(4) 1.5073 1.2317 0 0.9632 0 0
(5) 1.0339 1.655 1.2067 0 0 0
(6) 0 0 0 0 2 3
(7) 2 0 0 2 0 7
(8) 0 2 1.667 0 3.667 0
(9) 1.386 1.538 0.633 0.626 0 0
For points (6), (7), (8) and (9), the number of active constraints is equal to the number of design
variables. There are no feasible directions in the neighborhood of the points that can reduce cost
function any further. So, all the points are isolated local minima.

Arora, Introduction to Optimum Design, 4e 5-20

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.21 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 4𝑥𝑥12 + 3𝑥𝑥22 − 5𝑥𝑥1 𝑥𝑥2 − 8𝑥𝑥1

subject to 𝑥𝑥1 + 𝑥𝑥2 ≤ 4

Solution
Minimize f (x ) = 4 x12 + 3x22 − 5 x1 x2 − 8 x1; subject to x1 + x2 − 4 ≤ 0 .
( )
L (x, u ) = 4 x12 + 3x 22 − 5 x1 x 2 − 8 x1 + u x1 + x 2 − 4 + s 2
∂L ∂x1 = 8 x1 − 5 x 2 − 8 + u = 0 ; ∂L ∂x 2 = 6 x 2 − 5 x1 + u = 0 ;

Case 1. u = 0; gives a KKT point as (48/23, 40/23); f x = − 192 ( )

∗
23 .
Case 2. s = 0 (or g = 0); gives no candidate point (u = − 1 6 ).
Check for regularity: Ñg = (1, 1) . Since there is only one constraint, regularity is satisfied.

Referring to Exercises 4.62/4.115, the point satisfying the KKT necessary conditions is:
= x1∗ 2=.0870, x2∗ 1= .7391, u1∗ 0 .
The constraint function is linear and the Hessian of cost function is positive definite. Therefore this
is a convex programming problem and from Theorem 4.11, the point is an isolated global minimum.

Arora, Introduction to Optimum Design, 4e 5-21

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.22 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 𝑥𝑥12 + 𝑥𝑥22 − 4𝑥𝑥1 − 2𝑥𝑥2 + 6

subject to 𝑥𝑥1 + 𝑥𝑥2 ≥ 4

Solution
Minimize f (x1 , x2 ) = x12 + x22 − 4 x1 − 2 x2 + 6 ; subject to g =− x1 − x2 + 4 ≤ 0 .
(
L = x12 + x 22 − 4 x1 − 2 x 2 + 6 + u − x1 − x 2 + 4 + s 2 )
∂L ∂x 1 = 2 x1 − 4 − u = 0 ; ∂L ∂x 2 = 2 x 2 − 2 − u = 0
− x1 − x2 + 4 + s=
2
0; us= 0, u ≥ 0
Case 1. u = 0; gives no candidate point s 2 = −1 .
Case 2. s = 0; gives (2.5, 1.5) as a KKT point with u = 1 and f = 1.5. Since only one constraint is
active, regularity is satisfied.

Reference Exercise 4.63/4.116

x* = (2.5, 1.5) , u = 1, and f = 1.5.
The constraint function is linear and the Hessian of cost function is positive definite. Therefore this
is a convex programming problem and from Theorem 4.11, the point is an isolated global minimum.

Arora, Introduction to Optimum Design, 4e 5-22

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.23 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 2𝑥𝑥12 − 6𝑥𝑥1 𝑥𝑥2 + 9𝑥𝑥22 − 18𝑥𝑥1 + 9𝑥𝑥2

subject to 𝑥𝑥1 + 2𝑥𝑥2 ≤ 10
4𝑥𝑥1 − 3𝑥𝑥2 ≤ 20
𝑥𝑥𝑖𝑖 ≥ 0; 𝑖𝑖 = 1,2

Solution
Minimize f (x ) = 2 x12 − 6 x1 x2 + 9 x22 − 18 x1 + 9 x2 , subject to g1 = x1 + 2 x2 − 10 ≤ 0,
g2 =4 x1 − 3x2 − 20 ≤ 0, g3 = − x1 ≤ 0, g 4 = − x2 ≤ 0 .
There are 16 cases, but only the case u=1 u=
3 u=
4 0, s=
2 0 yields a solution:
(6.3, 1.733), u 2 = 0.8 , f = −56.901 .
Since only one constraint is active, regularity is satisfied.

Reference Exercise 4.64/4.117

x* = (6.3, 1.733) .
The constraint functions are linear and the Hessian of cost function is positive definite. Therefore
this is a convex programming problem and from Theorem 4.11, the point is an isolated global
minimum.

Arora, Introduction to Optimum Design, 4e 5-23

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.24________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 − 4 ≤ 0

Solution
Minimize f ( x1 , x2 ) = ( x1 − 1) + ( x2 − 1) , subject to
2 2
g1 = x1 + x2 − 4 ≤ 0
(
L = (x1 − 1) + (x2 − 1) + u1 x1 + x2 − 4 + s12
2 2
)
∂L ∂x1 = 2(x1 − 1) + u1 = 0 ;
∂L ∂x 2 = 2(x 2 − 1) + u1 = 0 ;

x1 + x 2 − 4 + s12 = 0 ; u1 s1 = 0 ; u1 ≥ 0
Case 1. u1 = 0, gives (1, 1) as a KKT point ( f = 0 ) . Since no constraint is active, regularity is
satisfied.
Case 2. s1 = 0; gives no candidate point.

Reference Exercise 4.65/4.118

x* = (1, 1) and f = 0.
The constraint function is linear and the Hessian of cost function is positive definite. Therefore this
is a convex programming problem and from Theorem 4.11, the point is an isolated global minimum.

Arora, Introduction to Optimum Design, 4e 5-24

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.25________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 − 4 ≤ 0
𝑥𝑥1 − 𝑥𝑥2 − 2 ≤ 0

Solution
Minimize f ( x ) = ( x1 − 1) + ( x2 − 1) ; subject to
2 2
g1 = x1 + x2 − 4 ≤ 0; g 2 = x1 − x2 − 2 ≤ 0 .
( ) (
L = (x1 − 1) + (x 2 − 1) + u1 x1 + x 2 − 4 + s12 + u 2 x1 − x 2 − 2 + s 22
2 2
)
∂L ∂x1 = 2(x1 − 1) + u1 + u 2 = 0 ; ∂L ∂x 2 = 2(x 2 − 1) + u1 − u 2 = 0

x1 + x 2 − 4 + s12 = 0 ; x1 − x 2 − 2 + s 22 = 0 ;
u1 s1 = 0 , u 2 s 2 = 0 ; u1 ≥ 0, u 2 ≥ 0
Case 1. u1 = 0, u 2 = 0 ; gives (1, 1) as a KKT point, f = 0
Case 2. u1 = 0, s 2 = 0 ; no candidate minimum.
Case 3. s1 = 0 , u 2 = 0 ; no candidate minimum.
Case 4. s1 = 0 , s 2 = 0 ; no candidate minimum.
Only the first case gives a solution that satisfies all the KKT necessary conditions. Since no
constraint is active, regularity is satisfied.

Reference Exercise 4.66/4.119

x* = (1, 1) and f = 0.
The constraint function is linear and the Hessian of cost function is positive definite. Therefore this
is a convex programming problem and from Theorem 4.11, the point is an isolated global minimum.

Arora, Introduction to Optimum Design, 4e 5-25

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.26 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 1)2 + (𝑥𝑥2 − 1)2

subject to 𝑥𝑥1 + 𝑥𝑥2 − 4 ≤ 0
2 − 𝑥𝑥1 ≤ 0

Solution
Minimize f ( x ) = ( x1 − 1) + ( x2 − 1) ; subject to: g1
2 2
= x1 + x2 − 4 ≤ 0; g 2 = 2 − x1 ≤ 0 .
( ) (
L = (x1 − 1) + (x 2 − 1) + u1 x1 + x 2 − 4 + s12 + u 2 2 − x1 + s 22
2 2
)
∂L ∂x1 = 2 (x1 − 1) + u1 − u 2 = 0 ; ∂L ∂x 2 = 2 (x 2 − 1) + u1 = 0

x1 + x 2 − 4 + s12 = 0 ; 2 − x1 + s 22 = 0 ; u1s1 = 0 , u 2 s2 = 0 , u1 ≥ 0 , u 2 ≥ 0
u1 s1 = 0 ; u 2 s 2 = 0 ; u1 ≥ 0, u 2 ≥ 0

Case 1. u1 = 0, u2 = 0; no candidate minimum point s2 < 0 . ( 2

)
Case 2. u1 = 0, s2 = 0; gives (2, 1) as a KKT point with u2 = 2, f = 1 .
Case 3. s1 = 0, u 2 = 0 ; no candidate minimum (u1 < 0) .
Case 4. s1 = s2 = 0 ; no candidate minimum (u1 < 0) .
Case 2 yields a KKT point. This point is regular since there is only one active constraint.

Reference Exercise 4.67/4.120

x* = (2, 1) and f = 1.
The constraint function is linear and the Hessian of cost function is positive definite. Therefore this
is a convex programming problem and from Theorem 4.11, the point is an isolated global minimum.

Arora, Introduction to Optimum Design, 4e 5-26

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.27 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 9𝑥𝑥12 − 18𝑥𝑥1 𝑥𝑥2 + 13𝑥𝑥22 − 4

subject to 𝑥𝑥12 + 𝑥𝑥22 + 2𝑥𝑥1 ≥ 16

Solution
Minimize f (x1 , x2 ) = 9 x12 − 18 x1 x2 + 13x22 − 4 , subject to g1 = 16 − ( x12 + x22 + 2 x1 ) ≤ 0 .
(
L = 9 x12 − 18 x1 x 2 + 13x 22 − 4 + u1 16 − x12 − x 22 − 2 x1 + s12 )
∂L ∂x1 = 18 x1 − 18 x 2 − 2u1 x1 − 2u1 = 0; ∂L ∂x 2 = −18 x1 + 26 x 2 − 2u1 x 2 = 0

x1 + x2 − 4 + s12 = 0; 2 − x1 + s22 = 0 ; u s = 0 ,
1 1 u 2 s2 = 0 , u1 ≥ 0 , u 2 ≥ 0

− x12 − x22 − 2 x1 + 16 + s12 = 0; u1s1 = 0, u1 ≥ 0

Case 1. u1 = 0; no candidate minimum s1 < 0 . ( 2
)
Case 2. s1 = 0; Solving the nonlinear system of equations, we get the following KKT points:
( 2.5945, 2.0198) , u = 1.4390,
1 f = 15.291; (− 3.630, 3.1754), u 1 = 23.2885 , f = 215.97 ;
(1.5088, − 3.2720), u1 = 17.1503, f = 244.53 ; (− 3.7322, − 3.0879 ), u1 = 2.1222, f = 37.877.
Since only one constraint is active, regularity is satisfied.

Referring to Exercise 4.68/4.121, the points satisfying the KKT necessary conditions are

x1∗ x2∗ u∗
(1) 2.5945 2.0198 1.4390
(2) − 3.630 3.1754 23.2885
(3) 1.5088 − 3.2720 17.1503
(4) − 3.7322 − 3.0879 2.1222

The Hessian of Lagrangian, and gradient of the constraint are

18 − 2u −18   −2 x1 − 2 
Ñ 2 L = Ñ  −2 x 
26 − 2u 
; g
 −18
1
 2 

=
1. At point (1), x1 2.=
5945, x2 2= .0198, u 1.4390, Ñ 2 L is positive definite. It follows from
Theorem 5.3 that the point is an isolated local minimum.

2. At point (2), x1 =−3.630, x2 = 3.1754, u = 23.2885 , Hessian of Lagrangian is negative definite.

Therefore the point violates the second order necessary conditions.

3. At point (3), x1 =
1.5088, x2 =
−3.2720, u =
17.1503;

Arora, Introduction to Optimum Design, 4e 5-27

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

 −16.3006 −18 
Ñ2 L = 
 −18 −8.3006 
Let d = (d1 , d 2 ) . We need to find d such that Ñg1 ⋅ d = 0. This gives d = c (1, 0.7667 ) where c ≠ 0 is
( )
any constant. Q = d T Ñ 2 L d = − 48.7811 c 2 < 0. So, sufficient condition is not satisfied at this point.
It cannot be a local minimum since it violates second order necessary condition.

4. At point (4), x1 =
−3.7322, x2 =
−3.8790, u =
2.1222,
13.7556 −18 
Ñ2 L = 
 −18 21.7556 
Ñg1 ⋅ d = 0 gives d = c (1, − 0.8848
= ) . Q d=
T
( )
Ñ 2 L d 62.6402 c 2 > 0. The sufficient condition is
satisfied. Thus, the point is an isolated minimum point.

Arora, Introduction to Optimum Design, 4e 5-28

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.28 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = (𝑥𝑥1 − 3)2 + (𝑥𝑥2 − 3)2

subject to 𝑥𝑥1 + 𝑥𝑥2 ≤ 4
𝑥𝑥1 − 3𝑥𝑥2 = 1

Solution
Minimize f (x ) = (x1 − 3) + (x2 − 3) ; subject to h = x1 − 3x2 − 1 = 0; g = x1 + x2 − 4 ≤ 0 .
2 2

(
L = (x1 − 3) + (x2 − 3) + v ≤ (x1 − 3x2 − 1) + u x1 + x2 − 4 + s 2
2 2
)
∂L ∂x1 = 2(x1 − 3) + v + u = 0 ; ∂L ∂x 2 = 2(x 2 − 3) + 3v + u = 0
h = x1 − 3x2 − 1 = 0; x1 + x2 − 4 + s 2 = 0; us = 0, u ≥ 0.
Case 1. u = 0; no candidate minimum s < 0 . ( 2
)
Case 2. s = 0; gives (3.25, 0.75) as a KKT point with v = −1.25, u = 0.75, f = 5.125 .
Since Ñ=
h (1, − 3) and Ñg = (1, 1) are linearly independent, regularity is satisfied.
Referring to Exercise 4.69/4.122, the point satisfying the KKT necessary conditions is
x1∗ = 3.25, x2∗ = −3.8790, u = 0.75, v = − 1.25.
The Hessian of cost function is positive definite and both the constraints are linear. Therefore, this is
a convex problem. It follows from Theorem 4.11 that the point is an isolated global minimum.

Arora, Introduction to Optimum Design, 4e 5-29

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.29 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 𝑥𝑥13 − 16𝑥𝑥1 + 2𝑥𝑥2 − 3𝑥𝑥22

subject to 𝑥𝑥1 + 𝑥𝑥2 ≤ 3

Solution
Minimize f (x1 , x2 ) = x13 − 16 x1 + 2 x2 − 3x22 , subject to g = x1 + x2 − 3 ≤ 0 .
(
L = x13 − 16 x1 + 2 x2 − 3x22 + u x1 + x2 − 3 + s 2)
∂L ∂x1 = 3x12 − 16 + u = 0; ∂L ∂x2 = 2 − 6 x2 + u = 0; x1 + x2 − 3 + s 2 =0;=
u g 0, u ≥ 0.

( ) ( )
Case 1. u = 0; gives 4 3 , 1 3 , f = −24.3 and − 4 3 , 1 3 , f = 24.967 , as KKT points.
Case 2. s = 0; gives (0, 3) , u = 16, f = − 21 ; (2, 1) , u = 4, f = −25, as KKT points.
For both cases, there is only one active constraint; so, regularity is satisfied.

Referring to Exercise 4.70/4.123, the points satisfying the KKT necessary conditions are

x1∗ x2∗ u∗
(1) 2.3094 0.3333 0
(2) − 2.3094 0.3333 0
(3) 0 3 16
(4) 2 1 4

The Hessian of the Lagrangian and gradient of the constraint are

6 x1 0  1
=Ñ 2 L =  ; Ñg  
 0 −6  1

1. At point (1), x1 = 2.3094, x2 = 0.3333, Ñ 2 L is indefinite. Since no constraint is active, this is an

inflection point for the cost function which does not satisfy the sufficient condition.

2. At point (2), x1 = − 2.3094 , x2 = 0.3333, Hessian of Lagrangian is negative definite. Since no

constraint is active, this is local maximum point (satisfies sufficient condition for local maximum).

3. At point (3), x1 = 0, x2 = 3, u = 16, Hessian of Lagrangian is negative semidefinite, so this point

does not satisfy second order necessary condition. It cannot be a local minimum.

12 0 
4. At point (4), x1 = 2, x2 = 1, u = 4, Ñ 2 L =  .
 0 −6 
Ñg ⋅ d = 0 gives d = c (1, − 1) where c ≠ 0 is any constant.
= Q d=
T
( )
∇ 2 L d 6c 2 > 0, for c ≠ 0. Since
Q > 0, sufficient condition is satisfied. Thus the point is an isolated local minimum point.

Arora, Introduction to Optimum Design, 4e 5-30

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.30 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥1 , 𝑥𝑥2 ) = 3𝑥𝑥12 − 2𝑥𝑥1 𝑥𝑥2 + 5𝑥𝑥22 + 8𝑥𝑥2

subject to 𝑥𝑥12 − 𝑥𝑥22 + 8𝑥𝑥2 ≤ 16

Solution
Minimize f (x1 , x2 ) = 3x12 − 2 x1 x2 + 5 x22 + 8 x2 , subject to g = x12 − x22 + 8 x2 − 16 ≤ 0
(
L = 3x12 − 2 x1 x2 + 5 x22 + 8 x2 + u x12 − x22 + 8 x2 − 16 + s 2 )
∂L ∂x1 = 6 x1 − 2 x 2 + 2ux1 = 0 ; ∂L ∂x 2 = −2 x1 + 10 x 2 + 8 − 2ux 2 + 8u = 0.

x12 − x 22 + 8 x 2 − 16 + s 2 ≤ 0 ; us = 0 , u ≥ 0
Case 1. u = 0; gives (− 2 7 , − 6 7 ) as a KKT point ( f = − 24 7 ) .
Case 2. s = 0; no candidate minima (u < 0 ) .
For case 1, since there is no active constraint, the regularity is satisfied.

Referring to Exercise 4.71/4.124, the point satisfying the KKT necessary condition is
x1∗ =− 0.2857, x2∗ = − 0.8571, u ∗ =0.
The Hessian of Lagrangian is positive definite. The sufficient condition is satisfied and the point is
an isolated local minimum.

Arora, Introduction to Optimum Design, 4e 5-31

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.31 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥, 𝑦𝑦) = (𝑥𝑥 − 4)2 + (𝑦𝑦 − 6)2

subject to 𝑥𝑥 + 𝑦𝑦 ≤ 12
𝑥𝑥 ≤ 6
𝑥𝑥, 𝑦𝑦 ≥ 0

Solution
Minimize f ( x, y ) = ( x − 4 ) + ( y − 6 ) ; subject to g1 = x + y − 12 ≤ 0;
2 2

g 2 =x − 6 ≤ 0; g3 =− x ≤ 0; g 4 =− y ≤ 0;
L= (( x − 4) 2
+ ( y − 6)
2
) + u ( x + y −12 + s ) + u ( x − 6 + s )
1
2
1 2
2
2

( ) (
+u3 − x + s32 + u4 − y + s42 )
∂L ∂x= 2 ( x − 4 ) + u1 + u2 − u3 = 0; ∂L ∂y= 2 ( y − 6 ) + u1 − u4 = 0;
x + y − 12 + s12 = 0; x − 6 + s22 = 0; − x + s32 = 0; − y + s42 = 0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).

Case 1. u=1 u=2 u=3 u=4 0; gives ( 4, 6 ) as a KKT point ; f = 0.

Case 2. u= 1 u=2 u=3 0, s=4 0; gives no candidate point.
Case 3. u= 1 u=2 u=4 0, s=3 0; gives no candidate point.
Case 4. u=1 u=3 u=4 0, s=2 0; gives no candidate point.
Case 5. u= 2 u=3 u= 4 0, s=1 0; gives no candidate point.
Case 6. u1 = u2 = 0, s3 = s4 = 0; gives no candidate point.
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives no candidate point.
Case 8. u1 = u4 = 0, s2 = s3 = 0; gives no candidate point.
Case 9. u2 = u3 = 0, s1 = s4 = 0; gives no candidate point.
Case 10. u= 2 u=4 0, s= 1 s= 3 0; gives no candidate point.

Case 11. u3 = u4 = 0, s1 = s2 = 0; gives no candidate point.

Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.
Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.
Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.

Arora, Introduction to Optimum Design, 4e 5-32

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.

Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

Check for regularity: Only the first case gives a solution that satisfies all the KKT necessary
conditions. Since no constraint is active, regularity is satisfied.

Referring to Exercise 4.72/4.125, the point satisfying the KKT necessary condition is
=x 4= , y 6= , u 0.
The constraint functions are linear and the Hessian of cost function is positive definite. Therefore
this is a convex programming problem and from Theorem 4.11, the point is an isolated global
minimum.

Arora, Introduction to Optimum Design, 4e 5-33

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.32 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Minimize 𝑓𝑓(𝑥𝑥, 𝑦𝑦) = (𝑥𝑥 − 8)2 + (𝑦𝑦 − 8)2

subject to 𝑥𝑥 + 𝑦𝑦 ≤ 12
𝑥𝑥 ≤ 6
𝑥𝑥, 𝑦𝑦 ≥ 0

Solution
Minimize f ( x, y ) = ( x − 8 ) + ( y − 8 ) ; subject to g1 = x + y − 12 ≤ 0;
2 2

g 2 =x − 6 ≤ 0; g3 =− x ≤ 0; g 4 =− y ≤ 0;
L= ( ( x − 8) 2
+ ( y − 8)
2
) + u ( x + y −12 + s ) + u ( x − 6 + s )
1
2
1 2
2
2

( ) (
+u3 − x + s32 + u4 − y + s42 )
∂L ∂x= 2 ( x − 8 ) + u1 + u2 − u3 = 0; ∂L ∂y= 2 ( y − 8 ) + u1 − u4 = 0;
x + y − 12 + s12 = 0; x − 6 + s22 = 0; − x + s32 = 0; − y + s42 = 0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).

Case 1. u=1 u=2 u=3 u=4 0; gives no candidate point.

Case 2. u= 1 u=2 u=3 0, s=4 0; gives no candidate point.
Case 3. u= 1 u=2 u=4 0, s=3 0; gives no candidate point.
Case 4. u=1 u=3 u=4 0, s=2 0; gives no candidate point.
Case 5. u=
2 u=3 u=
4 0, s=1 0; gives ( 6, 6 ) as a KKT point with u 1 = 4; f = 8.
Case 6. u1 = u2 = 0, s3 = s4 = 0; gives no candidate point.
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives no candidate point.
Case 8. u1 = u4 = 0, s2 = s3 = 0; gives no candidate point.
Case 9. u2 = u3 = 0, s1 = s4 = 0; gives no candidate point.
Case 10. u=2 u=4 0, s=1 s=3 0; gives no candidate point.
Case 11. u3 = u4 = 0, s1 = s2 = 0; gives ( 6, 6 ) as a KKT point with u 1 = 4; f = 8.
Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.
Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.
Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.

Arora, Introduction to Optimum Design, 4e 5-34

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.

Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

Check for regularity: For case 5, there is only one active constraint, so regularity is satisfied. For
case 11, Ñg1 = (1, 1) , Ñg 2 = (1,0 ) . Since Ñg1 and Ñg 2 are linearly independent regularity is
satisfied.

Referring to Exercise 4.73/4.126, the point satisfying the KKT necessary condition is
=x 6= , y 6 ,= u1 4= , u2 0=
, u3 0= , u4 0 .
The constraint functions are linear and the Hessian of cost function is positive definite. Therefore
this is a convex programming problem and from Theorem 4.11, the point is an isolated global
minimum.

Arora, Introduction to Optimum Design, 4e 5-35

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.33 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑥𝑥, 𝑦𝑦) = (𝑥𝑥 − 4)2 + (𝑦𝑦 − 6)2

subject to 𝑥𝑥 + 𝑦𝑦 ≤ 12
6 ≥ 𝑥𝑥
𝑥𝑥, 𝑦𝑦 ≥ 0

Solution
Minimize f ( x, y ) =− ( x − 4 ) − ( y − 6 ) ; subject to g1 =x + y − 12 ≤ 0;
2 2

g 2 =x − 6 ≤ 0; g3 =− x ≤ 0; g 4 =− y ≤ 0;
(
L = − ( x − 4) − ( y − 6)
2 2
) + u ( x + y −12 + s ) + u ( x − 6 + s )
1
2
1 2
2
2

( ) (
+ u3 − x + s32 + u4 − y + s42 )
−2 ( x − 4 ) + u1 + u2 − u3 =
∂L ∂x = 0;
∂L ∂y =−2 ( y − 6 ) + u1 − u4 =0;
x + y − 12 + s12 =0;
x − 6 + s22 =0;
− x + s32 =0;
− y + s42 =0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).
Case 1. u= 1 u= 2 u=3 u=4 0; gives ( 4, 6 ) as a KKT point ; F = 0 .
Case 2. u= 1 u= 2 u=3 0, s=
4 0; gives ( 4, 0 ) as a KKT point with u 4 =12; F = 36 .
Case 3. u= 1 u= 2 u=4 0, s=
3 0; gives ( 0, 6 ) as a KKT point with u 3 =8; F = 16 .
Case 4. u=1 u=3 u=
4 0, s=
2 0; gives ( 6, 6 ) as a KKT point with u 2 = 4; F = 4 .
Case 5. u= 2 u= 3 u=4 0, s=1 0; gives ( 5, 7 ) as a KKT point with u 1 = 2; F = 2 .
Case 6. u1 = u2 = 0, s3 = s4 = 0; gives ( 0, 0 ) as a KKT point with u 3 = 8 and u 4 = 12; F = 52 .
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives ( 6, 0 ) as a KKT point with u 2 = 4, u 4 = 12; F = 40 .
Case 8. u1 = u4 = 0, s2 = s3 = 0; gives no candidate point.
Case 9. u2 = u3 = 0, s1 = s4 = 0; gives no candidate point.
Case 10. u=
2 u= 4 0, s= 1 s= 3 0; gives ( 0, 12 ) as a KKT point with u 1 = 12, u 3 = 20; F = 52 .

Arora, Introduction to Optimum Design, 4e 5-36

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 11. u3 = u4 = 0, s1 = s2 = 0; gives ( 6, 6 ) as a KKT point with u 1 = 0, u 2 = 4; F = 4 .

Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.
Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.
Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.
Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.
Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

Check for regularity: For cases 1, 2, 3, 4 and 5, there is only one active constraint, so regularity is
( −1, 0 ) , Ñg 4 =
satisfied. For case 6, Ñg3 = ( 0, − 1) . Since Ñg3 and Ñg 4 are linearly independent,
regularity is satisfied. For case 7, Ñg=2 (1, 0 ) , Ñg=
4 ( 0, − 1) . Since Ñg 2 and Ñg 4 are linearly

independent, regularity is satisfied. For case 10, Ñg1 = (1, 1) , Ñg3 = ( −1, 0 ) . Since Ñg1 and Ñg3
are linearly independent, regularity is satisfied. For case 11, Ñg1 = (1, 1) , Ñg 2 = (1, 0 ) . Since Ñg1
and Ñg 2 are linearly independent regularity is satisfied.

Refer to Exercise 4.74/4.127. There are seven points that satisfy the KKT necessary conditions.
Only the points (0, 0) and (6, 0) are the local maximum points. Others violate second order
sufficiency conditions.

Arora, Introduction to Optimum Design, 4e 5-37

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.34 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑟𝑟, 𝑡𝑡) = (𝑟𝑟 − 8)2 + (𝑡𝑡 − 8)2

subject to 10 ≥ 𝑟𝑟 + 𝑡𝑡
𝑡𝑡 ≤ 5
𝑟𝑟, 𝑡𝑡 ≥ 0

Solution
Minimize f ( r , t ) =− ( r − 8 ) − ( t − 8 ) ; subject to g1 =r + t − 10 ≤ 0;
2 2

g 2 =t − 5 ≤ 0; g3 =−r ≤ 0; g 4 =−t ≤ 0;
L= (
− ( r − 8) − ( t − 8)
2 2
) + u ( r + t −10 + s ) + u (t − 5 + s )
1
2
1 2
2
2

( ) (
+u3 − r + s32 + u4 −t + s42 )
∂L ∂r =−2 ( r − 8 ) + u1 − u3 =0; ∂L ∂t =−2 ( t − 8 ) + u1 + u2 − u4 =0;
r + t − 10 + s12 = 0; t − 5 + s22 = 0; − r + s32 = 0; − t + s42 = 0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).
Case 1. u= 1 u= 2 u=3 u=4 0; gives no candidate point.
Case 2. u=1 u=2 u=3 0, s=4 0; gives ( 8, 0 ) as a KKT point with u 4 = 16; F = 64 .
Case 3. u= 1 u=2 u=4 0, s=3 0; gives no candidate point.
Case 4. u=1 u=3 u=4 0, s=2 0; gives no candidate point.
Case 5. u= 2 u= 3 u= 4 0, s=1 0; gives no candidate point.
Case 6. u1 = u2 = 0, s3 = s4 = 0; gives ( 0, 0 ) as a KKT point with u 3 = 16, u 4 = 16; F =128.
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives no candidate point.
Case 8. u1 = u4 = 0, s2 = s3 = 0; gives no candidate point.
Case 9. u2 = u3 = 0, s1 = s4 = 0; gives (10, 0 ) as a KKT point with u 1 = 4, u 4 = 20; F =68.
Case 10. u=2 u=4 0, s=1 s=3 0; gives no candidate point.
Case 11. u3 = u4 = 0, s1 = s2 = 0; gives no candidate point.
Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.
Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.
Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.
Arora, Introduction to Optimum Design, 4e 5-38
 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.

Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

Check for regularity: For cases 2, there is only one active constraint, so regularity is satisfied. For
( −1, 0 ) , Ñg 4 =
case 6, Ñg3 = ( 0, − 1) . Since
Ñg3 and Ñg 4 are linearly independent, regularity is
1 (1, 1) , Ñg=
satisfied. For case 9, Ñg= 4 ( 0, − 1) . Since Ñg1 and Ñg 4 are linearly independent,
regularity is satisfied.

Refer to Exercise 4.75/4.128. There are three points that satisfy the KKT necessary conditions. Only
the points (0, 0) and (10, 0) are the local maximum points.

Arora, Introduction to Optimum Design, 4e 5-39

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.35 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑟𝑟, 𝑡𝑡) = (𝑟𝑟 − 3)2 + (𝑡𝑡 − 2)2

subject to 10 ≥ 𝑟𝑟 + 𝑡𝑡
𝑡𝑡 ≤ 5
𝑟𝑟, 𝑡𝑡 ≥ 0

Solution
Minimize f ( r , t ) =− ( r − 3) − ( t − 2 ) ; subject to g1 =r + t − 10 ≤ 0;
2 2

g 2 =t − 5 ≤ 0; g3 =−r ≤ 0; g 4 =−t ≤ 0;
L= (
− ( r − 3) − ( t − 2 )
2 2
) + u ( r + t −10 + s ) + u (t − 5 + s )
1
2
1 2
2
2

( ) (
+u3 − r + s32 + u4 −t + s42 )
∂L ∂r =−2 ( r − 3) + u1 − u3 =0; ∂L ∂t =−2 ( t − 2 ) + u1 + u2 − u4 =0;
r + t − 10 + s12 = 0; t − 5 + s22 = 0; − r + s32 = 0; − t + s42 = 0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).
Case 1. u= 1 u= 2 u=3 u=4 0; gives ( 3, 2 ) as a KKT point ; F = 0.
Case 2. u=1 u=2 u=3 0, s=4 0; gives ( 3, 0 ) as a KKT point with u 4 = 0; F = 4 .
Case 3. u= 1 u=2 u=4 0, s=3 0; gives ( 0, 2 ) as a KKT point with u 3 = 6; F = 9 .
Case 4. u=1 u=3 u=4 0, s=2 0; gives ( 3, 5) as a KKT point with u 2 = 6; F = 9 .
Case 5. u=
2 u=3 u=
4 0, s=1 0; gives ( 5.5, 4.5 ) as a KKT point with u 1 = 5; F = 12.5 .
Case 6. u1 = u2 = 0, s3 = s4 = 0; gives ( 0, 0 ) as a KKT point with u 3 = 6, u 4 = 4; F =13.
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives no candidate point.
Case 8. u1 = u4 = 0, s2 = s3 = 0; gives ( 0, 5) as a KKT point with u 2 = 6, u 3 = 6; F = 18 .
Case 9. u2 = u3 = 0, s1 = s4 = 0; gives (10, 0 ) as a KKT point with u 1 = 14, u 4 = 18; F =53.
Case 10. u=
2 u= 4 0, s= 1 s= 3 0; gives no candidate point.

Case 11. u3 = u4 = 0, s1 = s2 = 0; gives ( 5, 5) as a KKT point with u 1 = 4, u 2 = 2; F =13.

Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.
Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.
Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.

Arora, Introduction to Optimum Design, 4e 5-40

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.

Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

Check for regularity: For cases 1, 2, 3, 4 and 5, there is only one active constraint, so regularity is
( −1, 0 ) , Ñg 4 =
satisfied. For case 6, Ñg3 = ( 0, − 1) . Since Ñg3 and Ñg 4 are linearly independent,
regularity is satisfied. For case 8, Ñg 2 = ( 0, 1) , Ñg3 = ( −1,0 ) . Since Ñg 2 and Ñg3 are linearly
independent, regularity is satisfied. For case 9, Ñ= g1 (1, 1) , Ñ= g 4 ( 0, − 1) . Since Ñg1 and Ñg 4
are linearly independent, regularity is satisfied. For case 11, Ñg1 = (1, 1) , Ñg 2 = ( 0, 1) . Since Ñg1
and Ñg 2 are linearly independent regularity is satisfied.

Refer to Exercise 4.76/4.129. There are nine points that satisfy the KKT necessary conditions. Only
the points (0, 0), (0, 5), (10, 0) and (5, 5) are the local maximum points. Others violate second order
sufficiency conditions.

Arora, Introduction to Optimum Design, 4e 5-41

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.36 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑟𝑟, 𝑡𝑡) = (𝑟𝑟 − 8)2 + (𝑡𝑡 − 8)2

subject to 𝑟𝑟 + 𝑡𝑡 ≤ 10
𝑡𝑡 ≥ 0
𝑟𝑟 ≤ 0

Solution
Minimize f ( r , t ) =− ( r − 8 ) − ( t − 8 ) ; subject to g1 =r + t − 10 ≤ 0;
2 2

g 2 =r ≤ 0; g3 =−t ≤ 0;
(
L = − ( r − 8) − ( t − 8)
2 2
) + u ( r + t −10 + s ) + u ( −t + s )
1
2
1 2
2
2

(
+u3 r + s32 )
∂L ∂r =−2 ( r − 8 ) + u1 + u3 =0; ∂L ∂t =−2 ( t − 8 ) + u1 + u2 =0;
r + t − 10 + s12 = 0; − t + s22 = 0; r + s32 = 0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 3 (there are 8 cases).

Case 1. u=1 u=
2 u=
3 0 ; no candidate minimum.
Case 2. u=1 u=2 0, s=3
0 ; no candidate minimum.
Case 3. u=1 u=3 0, s=
0 ; no candidate minimum.
2

Case 4. u=2 u=3 0, s=1

0 ; no candidate minimum.
Case 5. u=1 0, s=
2 s=
0 ; no candidate minimum.
3

Case 6. u=2 0, s=1 s= 3 0 ; no candidate minimum.

Case 7. u=3 0, s=1 s= 2 0 ; no candidate minimum.
Case 8. s=
1 s=
2 s=3 0 ; no candidate minimum.

Refer to Exercise 4.77/4.130. There are no points that satisfy the KKT necessary conditions.

Arora, Introduction to Optimum Design, 4e 5-42

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.37 ________________________________________________________________________________
Solve the following problem graphically. Check necessary and sufficient conditions for candidate
local minimum points and verify them on the graph for the problem.

Maximize 𝐹𝐹(𝑟𝑟, 𝑡𝑡) = (𝑟𝑟 − 3)2 + (𝑡𝑡 − 2)2

subject to 10 ≥ 𝑟𝑟 + 𝑡𝑡
𝑡𝑡 ≥ 5
𝑟𝑟, 𝑡𝑡 ≥ 0

Solution
Minimize f ( r , t ) =− ( r − 3) − ( t − 2 ) ; subject to g1 =r + t − 10 ≤ 0;
2 2

g 2 =−t + 5 ≤ 0; g3 =−r ≤ 0; g 4 =−t ≤ 0;

(
L = − ( r − 3) − ( t − 2 )
2 2
) + u ( r + t −10 + s ) + u ( −t + 5 + s )
1
2
1 2
2
2

( ) (
+u3 − r + s32 + u4 −t + s42 )
∂L ∂r =−2 ( r − 3) + u1 − u3 =0; ∂L ∂t =−2 ( t − 2 ) + u1 − u2 − u4 =0;
r + t − 10 + s12 = 0; − t+5 + s22 = 0; − r + s32 = 0; − t + s42 = 0;
u i si = 0 ; u i ≥ 0 ; i = 1 to 4 (there are 16 cases).

Case 1. u=1 u=2 u=3 u=4 0; gives no candidate point.

Case 2. u= 1 u=2 u=3 0, s=4 0; gives no candidate point.
Case 3. u= 1 u=2 u=4 0, s=3 0; gives no candidate point.
Case 4. u=1 u=3 u=4 0, s=2 0; gives ( 3, 5) as a KKT point with u 3 = 6; F =9.
Case 5. u=
2 u=3 u=
4 0, s=1 0; gives no candidate point.
Case 6. u1 = u2 = 0, s3 = s4 = 0; gives no candidate point.
Case 7. u1 = u3 = 0, s2 = s4 = 0; gives no candidate point.
Case 8. u1 = u4 = 0, s2 = s3 = 0; gives no candidate point.
Case 9. u2 = u3 = 0, s1 = s4 = 0; gives no candidate point.
Case 10. u=
2 u= 4 0, s= 1 s= 3 0; gives ( 0, 10 ) as a KKT point with u 1 = 16, u 3 = 22; F =73.

Case 11. u3 = u4 = 0, s1 = s2 = 0; gives no candidate point.

Case 12. u1 = 0, s2 = s3 = s4 = 0; gives no candidate point.
Case 13. u2 = 0, s1 = s3 = s4 = 0; gives no candidate point.

Arora, Introduction to Optimum Design, 4e 5-43

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Case 14. u3 = 0, s1 = s2 = s4 = 0 ; gives no candidate point.

Case 15. u4 = 0, s1 = s2 = s3 = 0 ; gives no candidate point.
Case 16. s1 = s2 = s3 = s4 = 0 ; gives no candidate point.

Check for regularity: For cases 4, there is only one active constraint, so regularity is satisfied. For
case 10, Ñg1 = (1, 1) , Ñg3 = ( −1, 0 ) . Since Ñg1 and Ñg3 are linearly independent regularity is
satisfied.

Refer to Exercise 4.78/4.131. There are two points that satisfy the KKT necessary conditions; (3, 5)
and (0, 10). Only the point (0, 100) is a local maximum point.

Arora, Introduction to Optimum Design, 4e 5-44

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.38 ________________________________________________________________________________
Rewrite the formulation of Exercise 2.23, the problem is written in the standard form as (note that
some of the data used here is different from that used in Exercise 2.23):
( )
Minimize f = 12.331 Ro2 − Ri2 , subject to = ( ) ( )
g1 6.3662 × 106 Ro Ro4 − Ri4 − 2.5 × 104 ≤ 0
= (
g 2 4244.132 Ro2 + Ro Ri + Ri2 ) (R
4
o )
− Ri4 − 9000 ≤ 0,
g3 =
Ro − 20 ≤ 0; g 4 =
Ri − 20 ≤ 0; g 5 =
− Ro ≤ 0; g 6 =
− Ri ≤ 0;

The optimum solution found by the graphical method is

Ro∗ = 20, Ri∗ = 19.84, f ∗ = 79.1 kg where g1 and g 3 are active.

1. Check for necessary conditions. Since at the optimum only g1 and g 3 are active, we can assume
u2 = u4 = u5 = u6 = 0 and s1 = s3 = 0.
The KKT necessary conditions are
∂L
= 24.662 Ro + ( 2.5 × 104 ) u1  254.648 ( Ro4 − Ri4 ) − 254.648 Ro ( 4 Ro3 )  ( Ro4 − Ri4 ) + u3 =0
2
(1)
∂Ro
∂L
−24.662 Ri + ( 2.5 × 104 ) u1  −254.648 Ro ( −4 Ri3 )  ( Ro4 − Ri4 ) =
2
= 0 (2)
∂Ri
gi + si2= 0, ui s=
i 0, ui ≥ 0; =
i 1 to 6

Substituting the optimum value into (1) and (2), we get u1 = 3.5596 × 10 −3 > 0, u3 = 5.29 > 0. The other
conditions are obviously satisfied. Thus the point Ro∗ = −20, Ri∗ = 19.84 satisfies all of the necessary
conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints equals
the number of design variables, the point is indeed an isolated local minimum.

Arora, Introduction to Optimum Design, 4e 5-45

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.39 ________________________________________________________________________________
Rewriting the formulation of Exercise 2.24, we have
Minimize f = 1.44086 Rt , subject to
= g1 7957.7 Rt − 250 ≤ =
0 ; g 2 767.8 R 2 − 3 ≤ 0;
g3 = 20 − R t ≤ 0; g 4 = 20 − R ≤ 0; g5 = R − 200 ≤ 0; g 6 = 1 − t ≤ 0; g 7 = t − 10 ≤ 0
The optimum solutions found by the graphical method are the points on g 1 between ( 31.83 , 1.0 ) and
( 25.23, 1.26 ) , where f ∗ = 45.9 and g 1 is active.

1. Check for necessary conditions. Since only g 1 is active, we can set u2 = u3 = u4 = u5 = u6 = u7 = 0

and s 1 = 0.
L= 1.44086 Rt + u1 ( 7957.7 Rt − 250 ) .
The KKT necessary conditions are
( ) (
∂L ∂t= 1.44086 R + u1 −7957.7 Rt 2 = 0; ∂L ∂R= 1.44086t + u1 −7957.7 R 2t = 0 )
gi + si2= 0, ui si= 0, ui ≥ 0; i = 1 to 7
Using the relation, 1.44086Rt = 45.9 at optimum solutions, and either equation, we can obtain u 1 =
0.183670. All the other conditions are also satisfied. Thus, the solutions satisfy all of the necessary
conditions.
Ñ= 2
L Ñ 2 f + u1Ñ 2 g1
2. Check for sufficient condition.
 0 1.44086   2 ( 7957.7 ) Rt 3 7957.7 R 2 t 2 
1.44086 + ( ) 
0 
0 .1836
  7957.7 R t
2 2
2 ( 7957.7 ) R 3 t 
Substituting 1.44086Rt = 45.9 into Ñ 2 L and Ñg1 , we have
 2.88 R t 2.88   −249.8 t 
Ñ2 L =  , Ñg1  
 2.88 2.88 t R   −249.8 R 
c (1, − R t ) , c ≠ 0 . Hence,
T
Ñg1T d = 0 gives d = = Q d= T
Ñ2 L d 0
Thus, the sufficient condition is not satisfied. Hence solution is not an isolated local minimum.

Arora, Introduction to Optimum Design, 4e 5-46

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.40________________________________________________________________________________
Rewrite the formulation of Exercise 3.28, we have
Minimize f = 50.077 R t + 0.1885 ( R + t 2 ) , subject to g1 =25 − 25.038 ( R − t 2 ) ≤ 0 ;
2 2

g 2= 3.5 R t − 210 ≤ 0 ; g3= (1.41667 × 10 ) R t − 0.001 ≤ 0 ;

−5
g 4= 0.5 t − R ≤ 0 ;
g5 =− R ≤ 0; g 6 =−t ≤ 0
is R∗ 1=
The optimum solution obtained by the graphical method= .0, t ∗ 0.0167, f ∗ = 1.0 where g 1
and g 2 are active.

1. Check for necessary conditions. Since only g 1 and g 2 are active, we can set u3 = u4 = u5 = 0 and
s1 = s2 = 0.
L = 50.077 Rt + 0.1885 ( R + t 2 ) + u1  25 − 25.038 ( R + t 2 )  + u2 ( 3.5 R t − 210 )
2 2

 
∂L ∂=R 50.077t + 0.377 ( R + t 2 ) − 2u1 ( 25.038 )( R + t 2 ) + u2 ( 3.5 t=) 0 (1)
∂L = (
∂t 50.077 R + 0.1885 ( R + t 2 ) + u1 ( 25.038 )( R + t 2 ) − u2 3.5 R t=
2
0 ) (2)
gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 5 (3)
Substituting the optimum value into (1) and (2);
1.216434 − 1.9833 =
u1 + u 2 0, 50.267074 + 0.99165=
u1 − 60u 2 0
Solving for u1 and u2 , we get u1 = 0.0417 > 0 (o.k.), u2 = 4.080 × 10 −3 > 0 (o.k.). All the other
conditions in (3) are satisfied. Therefore, the point ( R ∗ = 1.0m, t ∗ = 0.0167m) satisfies all of the
necessary conditions.

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
design variables, the point ( R ∗ = 1.0, t ∗ = 0.0167) is indeed a local minimum.

Arora, Introduction to Optimum Design, 4e 5-47

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.41 ________________________________________________________________________________
Referring to the formulation of Exercise 4.80, we have
Minimize f = 2.4662 ×105 R t , subject=
to g1 795.775 Rt − 2.5 × 108 ≤ 0 ;
g 2 =5 × 104 − 6.5113 × 1010 R 3t ≤ 0 ; g3 =R t − 50 < 0 ; g 4 =− R ≤ 0 ; g5 =−t ≤ 0
=
The optimum solution found by the graphical method is R∗  0=
.0787, t ∗  0.00157,
= f ∗  30.56 kg;
where g 2 and g 3 are active.

1. Check for necessary conditions. Since only g 2 and g 3 are active, we can set u1 = u4 = u5 = 0 and
s2 = s3 = 0. The KKT necessary conditions are
∂L = ( )( )
∂R 2.4662 × 105 t + u2  −6.5113 × 1010 3R 2 t  + u3 (1= t) 0 (1)
( ) ( )
∂L ∂t = 2.4662 × 105 R + u2 − 6.5113 × 1010 R 3 + u3 − R t 3 = 0 (2)
gi + si=
2
0, ui s=
i 0, ui ≥ 0; =i 1 to 5 (3)
Substituting the optimum value into (1) and (2), we get
u2 = 3.056 × 10 −4 > 0, u3 = 0.3038 > 0 (o.k.).
(
All of the other constraints in (3) are also satisfied. Thus, the point R ∗ = 0.0787, t ∗ = 0.00157 )
satisfies the necessary conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints is equal
to the number of design variables, the point is indeed an isolated local minimum.

Arora, Introduction to Optimum Design, 4e 5-48

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.42 ________________________________________________________________________________
Referring to the formulation of Exercise 4.81, we have
( )
Minimize f = 1.2331 × 105 Ro2 − Ri2 , subject to = (
g1 1.59155×104 Ro2 − Ri2 − 2.5 × 108 ≤ 0; )
g 2 =×
5 10 − 1.62783 × 10
4 10
(R
4
o −R i
4
) ≤ 0; g3 =0.5 ( Ro + Ri ) − 50 ≤ 0; g 4 =− Ro ≤ 0; g5 =− Ri ≤ 0
=
The optimum solution found by the graphical method is Ro∗  0=
.0795, Ri∗  0=
.0779 , f *  30.56 kg
where g 2 and g 3 are active.

1. Check for necessary conditions. Since only g 2 and g 3 are active, we can set u1 = u4 = u5 = 0, and s 2
and s 3 = 0.
( ) ( )
L= 1.2331 × 105 Ro2 − Ri2 + u2 5 × 104 − 1.62783 × 1010 Ro4 − Ri4  + u3 0.5 ( Ro + Ri ) ( Ro + Ri ) − 50 
The KKT necessary conditions are
∂L = ( ) (  )( )
∂Ro 2.4662 × 105 Ro + u2  − 1.62783 × 1010 4 Ro3  + u3  − Ri ( Ro − R= i)
2
 0 (1)

∂L ∂Ri = ( ) (  )( )
− 2.4662 × 105 Ri + u2  1.62783 × 1010 4 Ri3  + u3  Ro ( Ro − Ri )  =
2


0 (2)
gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 5 (3)
Substituting the optimum value into (1) and (2), we get
u2 = 3.056 × 10 −4 > 0, u3 = 0.3055 > 0 (o.k.).
point Ro∗  0=
All other conditions in (3) are satisfied. Thus, the= .0795, Ri∗  0.0779 satisfies the
necessary conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints is equal
to the number of design variables, the point is indeed a local minimum. Sufficient condition is
deemed satisfied at this point.

Arora, Introduction to Optimum Design, 4e 5-49

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.43________________________________________________________________________________
Referring to Exercise 4.79, we have
Minimize f = π DH + π D 2 2 , subject to g=
1 400 − π D 2 H 4 ≤ 0; g=
2 3.5 − D ≤ 0;
g3 = D − 8 ≤ 0; g 4 = 8 − H ≤ 0; g5 = H − 18 ≤ 0
The optimum solution found by the graphical method is= H ∗ 8,= D∗  7.98,= f ∗  300.6 cm 2 where
g 1 and g 4 are active.
1. Check for necessary conditions. Since only g 1 and g 4 are active, we can set u2 = u3 = u5 = 0 and
s 1 = s 4 = 0.
( )
L = π DH + π D 2 2 + u1 400 − π D 2 H 4 + u4 ( 8 − H )
The KKT necessary conditions are
( )
∂L ∂D = π H + π D + u1 ( −π DH 2 ) = 0; ∂L ∂H = π D + u1 −π D 2 4 − u4 = 0
gi + s = 0, ui s=
2
i i 0, ui ≥ 0 ; i = 1 to 5
Substituting the optimum value into equations, we get u 1 = 0.5 > 0, u 4 = 0.063 > 0 (o.k.). All the
other conditions in (3) are also satisfied. Thus, the point (H ∗ = 8, D ∗ = 7.98) satisfies the necessary
conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints is equal
to the number of design variables, the point is indeed local minimum point.
5.44 ________________________________________________________________________________
Rewriting the formulation of Exercise 4.83, we have
Minimize f = 0.6h + 0.001A , subject to g1 = 20,000 − hA 3.5 ≤ 0;
g 2 =0.25 hA 3.5 + A − 10,000 ≤ 0; g3 =3.5 − h ≤ 0; g 4 =h − 21 ≤ 0; g5 =− A ≤ 0
The graphical optimum solution is A∗ = 5,000, h∗ = 14, f ∗ = 13.4, where g 1 and g 2 are active.

1. Check for necessary conditions.

L = 0.6h + 0.001A + u1 (20 ,000 − hA 3.5) + u2 (0.25hA 3.5 + A − 10,000 ) + u3 (3.5 − h )
+ u4 (h − 21) − u5 A
The KKT necessary conditions are
∂L ∂A = 0.01 + u1 (− h 3.5) + u2 (0.25h 3.5 + 1) − u5 = 0 (1)
∂L ∂h = 0.6 + u1 (− A 3.5) + u2 (0.25 A 3.5) − u3 + u4 = 0 (2)
gi ≤ 0, ui gi = 0, ui ≥ 0 ; i =1 to 5 (3)
Substituting A = 5,000, h = 14, u3 = u4 = u5 = 0 (since only g 1 and g 2 are active) into (1) and (2), we
u1 5.9 × 10− 4 > 0, u=
get = 2 6.8 × 10− 4 > 0 (o.k.). All the other conditions in (3) are satisfied.
Therefore, necessary conditions are satisfied.

2. Check for sufficient condition. Since the number of constraints active at the candidate point and
the number of design variables are the same, the point is indeed a local minimum point.

Arora, Introduction to Optimum Design, 4e 5-50

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.45 ________________________________________________________________________________
Rewriting the formulation of Exercise 3.34, we have
( ) ( )
Minimize f = 3.083 × 10 −3 x12 1 − x22 , subject to g1= 5.093 × 107 x13 1 − x24 − 275 ≤ 0;
=
g 2 6.36619 × 105 x14 (1 − x ) − 3.49066 × 10
4
2
−2
≤ 0; g3 = 2.0 × 107 − ( 4.17246 × 10 ) x (1 − x )
4 3
1 2
2.5
≤ 0;
g 4 = 20 − x1 ≤ 0; g5 = x1 − 500 ≤ 0; g 6 = 0.6 − x2 ≤ 0; g 7 = x2 − 0.999 ≤ 0
is x1∗ 103,
The optimum solution obtained by the graphical method= = x2∗ 0.955,
= f ∗ 2.9 where g 1
and g 3 are active.

1. Check for necessary conditions. Since only g 1 and g 3 are active, we can set u 2 = u 4 = u 5 = u 6 = u 7
= 0.
( ) (
L= 3.083 × 10−3 x12 1 − x22 + u1 5.093 × 107 x13 1 − x24 − 275 )
 ( )
+ u3  2.0 × 107 − 4.17246 × 104 x13 (1 − x2 ) 
2.5


The KKT necessary conditions are
∂L ∂= ( ) (
x1 6.166 × 10−3 x1 1 − x22 + u1  −3 5.093 × 107 x14 1 − x24 ) ( )
 ( )( )
+u3  − 4.17246 × 104 3 x12 (1 − x2 )  =
2.5


0 (1)

( )
∂L ∂x2 =−6.166 × 10−3 x12 x2 + u1  4 5.093 × 107 x23 x13 1 − x24  ( )
2

 

 ( )( )
+u3  − 4.17246 × 104 3 x12 (1 − x2 )  =
2.5


0 (2)
gi ≤ 0, ui gi =0, ui ≥ 0 ; i = 1 to 7 (3)
−3
Substituting the optimum values into (1) and (2), respectively, we get u1 = 4.568 × 10 > 0,
u3 3.332 × 10− 8 > 0 (o.k.). All the other conditions in (3) are also satisfied. Therefore, the point
=
(x∗
1 )
= 103, x2∗ = 0.955 satisfies all of the necessary conditions.

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point x1∗ = 103, x2∗ = 0.955 is indeed a local minimum point.

Arora, Introduction to Optimum Design, 4e 5-51

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.46 ________________________________________________________________________________
Rewriting the formulation of Exercise 3.35, we have
(
Minimize f = 3.083 × 10 −3 d o2 − d i2 , subject to g=
1 ) (
5.093 × 107 d o d o4 − di4 − 275 ≤ 0; ) ( )
g=
2 ( 6.3662 × 10 ) ( d − d ) − 3.49066 × 10
5 4
o i
4 −2
≤ 0;
= 2.0 × 10 − ( 4.17246 × 10 ) d (1 − d d )
2.5
g3 7 4 3
o i o ≤ 0; g 4 = 20 − d o ≤ 0;
g5 =d o − 500 ≤ 0; g 6 =0.6 − di d o ≤ 0; g 7 =di d o − 0.999 ≤ 0;
The optimum solution obtained by the graphical method is d o∗ = 103, d i∗ = 98.36, f ∗ = 2.9 where g1
and g 3 are active.

1. Check for necessary conditions. Since only g1 and g 3 are active, we can set u2 = u4 = u5 = u6 = u7
= 0.
( ) (
L = 3.083 × 10−3 do2 − di2 + u1  5.093 × 107 do do2 − di2 − 275
  ) ( )
 ( )
+ u3  2.0 × 107 − 4.17246 × 104 do3 (1 − di do ) 
2.5

The KKT necessary conditions are

(
) ( ) ( )
2
∂L ∂d= o 6.166 × 10−3 do + u1 5.093 × 107 1 do4 − di4 − 4do do4 − di4 
 
( )( )
− u3 4.17246 × 104  3do2 (1 − di do ) + 2.5 ( di do )(1 − di do )  =

2.5 1.5

0 (1)


( ) ( )( )
2
∂L ∂di =− 6.166 × 10−3 di + u1  5.093 × 107 do 4di3 do4 − di4 
 
(
 )
+u3  4.17246 × 104 ( 2.5 ) d02 (1 − di do )  =
1.5

0 (2)
gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1 to 7 (3)
Substituting the optimum values into (1) and (2) respectively, we get
u 1 = 4.657 × 10 −3 > 0,=u3 3.281 × 10−8 > 0 (o.k.).
All the other conditions in (3) are also satisfied. Therefore, the point obtained from graphic method
satisfies all of the necessary conditions.
2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point d o∗ = 103, d i∗ = 98.36 is indeed an isolated local minimum point. The
sufficient condition is deemed satisfied at this point.

Arora, Introduction to Optimum Design, 4e 5-52

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.47 ________________________________________________________________________________
Rewriting the formulation of Exercise 3.36, we have
Minimize f = 0.02466 R t , subject to g=
1 ( 3.1831× 106 ) ( 2R + t ) ( 4R3t + Rt 3 ) − 275 ≤ 0;
(
g 2 = 3.97886 × 104 ) ( 4 R t + Rt ) − 3.49066 × 10
3 3 −2
( )
≤ 0; g3 = 2.0 × 107 − 3.37972 × 105 ( R + 0.5t ) t 2.5 ≤ 0;
0.5

g 4 = 50 − R ≤ 0; g5 = R − 200 ≤ 0; g 6 = 2 − t ≤ 0; g 7 = t − 40 ≤ 0
The optimum solution obtained by the graphical method is R ∗ = 50.3, t ∗ = 2.34, f ∗ = 2.9 where g 1
and g 3 are active.
1. Check for necessary conditions. Since only g 1 and g 3 are active, we can set
u=2 u=4 u=
5 u=
6 u=
7 0.
( )
L= 0.02466 Rt + u1  3.1831 × 106 ( 2 R + t ) 4 R 3t + Rt 3  ( )
 ( )
+u3  2.0 × 107 − 3.37972 × 105 ( R + 0.5t ) t 2.5 
0.5


The KKT necessary conditions are
( ) ( )
L ∂R 0.02466t + u1 3.1831 × 106  2 4 R 3t + Rt 3 − ( 2 R + t ) 12 R 2 + t 3 4 R 3t + Rt 3  ( )( )
2
∂=
 
(  )
− u3 3.37972 × 105 ( 12 ) t 2.5 ( R + 0.5t )  =
0.5


0 (1)

( ) ( )
∂L ∂t 0.02466 R + u1 3.1831 × 106 1 4 R 3t + Rt 3 − ( 2 R + t ) 4 R 3 + 3Rt 2 4 R 3t + Rt 3  ( )( )
2
=
 
( )
− u3 3.37972 × 105 ( 14 ) t 2.5 ( R + 0.5 t ) + 2.5t1.5 ( ) ( R + 0.5 t ) 0.5
=

0 (2)
gi ≤ 0, ui gi =0, ui ≥ 0 ; i = 1 to 7 (3)
−3
Substituting the optimum value into (1) and (2) respectively, we obtain u 1 = 4.643 × 10 > 0, u 3 =
3.240 × 10 −8 > 0 (o.k.). All the other conditions in (3) are also satisfied. Therefore, the point obtained
from graphical method satisfies all the necessary conditions.

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point R ∗ = 50.3, t ∗ = 2.34 is indeed an isolated local minimum point. The
sufficient condition is deemed satisfied at this point.

Arora, Introduction to Optimum Design, 4e 5-53

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.48 ________________________________________________________________________________
Referring to the formulation of Exercise 3.54, we have
( ) , subject to
Minimize f = 6.59734 − 10 −3 D 2 H 2 + 4800 ) ( 12

( 2.546475 × 10 )( H + 4800 ) ( D H ) − 1.5 × 10 ≤ 0;

1
=g1 4 2 2 2 4

( )( ) ( ) ( )
1
g 2 =2.0 × 104 H 2 + 4800 2
H − 1.816774 × 106 D 4 H 2 + 4800 ≤ 0;
g3 = H − 500 ≤ 0; g 4 = 50 − H ≤ 0; g5 = D − 50 ≤ 0; g 6 = 0.5 − D ≤ 0 .
=
The optimum solution obtained by the graphical method is H ∗ 50,
= D∗  3.42,
= f ∗  6.6 , where g 2
and g 4 are active.

1. Check for necessary conditions. Since only g 2 and g 4 are active, we can set u1 = u3 = u5 = u6 = 0
and s2 = s4 = 0 .

( ) ( )
1
L= 6.59734 × 10−3 D 2 H 2 + 4800 2
+
 
( )( ) ( ) ( )
1
u2  2.0 × 104 H 2 + 4800 2 H − 1.816774 × 106 D 4 H 2 + 4800  + u4 ( 50 − H )
 
The KKT necessary conditions are

( ) ( ) ( )
1
=∂L ∂D 0.0131947 D H 2 + 4800 2+ u  −7.267096 × 106 D3 =
H 2 + 4800  0 (1)
2  

( ) ( ) ( )( )
1 1
∂L
= ∂H 6.59734 × 10−3 D 2 H H 2 + 4800 2 + u2  − 2.0 × 104 H 2 + 4800 2 H 2

2
( )( ) ( ) ( )
1
+ 2.0 × 104 H 2 + 4800 2 + 3.633548 × 106 D 4 H H 2 + 4800  − u4 = 0 (2)

gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 6 (3)
−5 −2
Substituting the optimum value into (1) and (2), we get u2 = 9.68 × 10 > 0,=
u4 4.68 × 10 >0 (o.k.)

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
(
design variables, the point H ∗ = 50, D ∗ = 3.42 is indeed an isolated local minimum point. The )
sufficient condition is deemed satisfied at this point.

Arora, Introduction to Optimum Design, 4e 5-54

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.49 ________________________________________________________________________________
Answer True or False.

1. Candidate minimum points for a constrained problem that do not satisfy second-order

sufficiency conditions can be global minimum designs. True

2. Lagrange multipliers may be used to calculate the sensitivity coefficient for the cost

function with respect to the right side parameters even if Theorem 4.7 cannot be used.

True

3. Relative magnitudes of the Lagrange multipliers provide useful information for practical

design problems. True

Arora, Introduction to Optimum Design, 4e 5-55

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.50 ________________________________________________________________________________

A circular tank that is closed at both ends is to be fabricated to have a volume of 250π m3. The

fabrication cost is found to be proportional to the surface area of the sheet metal needed for fabrication

of the tank and is $400/m2. The tank is to be housed in a shed with a sloping roof which limits the height

of the tank by the relation H≤8D, where H is the height and D is the diameter of the tank. The problem is

formulated as minimize f(D,H)=400(0.5πD2+πDH) subject to the constraints , and H≤8D.

Ignore any other constraints.

1. Check for convexity of the problem.

2. Write KKT necessary conditions.

3. Solve KKT necessary conditions for local minimum points. Check sufficient conditions

and verify the conditions graphically.

4. What will be the change in cost if the volume requirement is changed to 255π m3 in place

of 250π m3?

=
Minimize ( )
f 400 0.5π D 2 + π DH , subject to h1 = πD 2 H 4 − 250π = 0; g1 = H − 8 D ≤ 0

1. Check for convexity of the problem.

 ∂ f ∂D   400 p D + 400pH   400p 400p = M1 400 p > 0
=Ñf =   =   400p 0  M 2 =
, H ;
∂ f ∂H   400pD   −160000 p2 < 0
Since Hessian of the cost function is not positive definite, this is not a convex programming
problem.

2. Write Kuhn-Tucker necessary conditions.

= ( ) ( )
L 400 0.5πD 2 + πDH + v1 πD 2 H 4 − 250π + u1 ( H − 8 D )
∂L ∂D = 400π D + 400π H + v1 ( π DH 2 ) + u1 ( −8 )= 0 (1)

( )
∂L ∂H = 400π D + v1 π D 2 4 + u1 = 0 (2)
h1 =0, g1 ≤ 0, u1 g1 =0 and u1 ≥ 0

Arora, Introduction to Optimum Design, 4e 5-56

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

3. Solve the KKT conditions.

Case 1. g1 = 0 ; H = 8D . Solving the equations, we get D = 5, H = 40, v1 = −226.67, u1 = −1832.5
< 0. Since u 1 < 0, the KKT conditions are violated, and there is no solution for this case.
Case 2. u 1 = 0. Solving the equations, we obtain D = H = 10, v1 = −160 .
Since g1 = H − 8 D = 10 − 80 = −70 < 0, all the KKT conditions are satisfied. Thus, (10, 10) is a
candidate minimum point. f ∗ = 60,000 π

4. Change in cost.
Applying the constraint variation sensitivity theorem, we get the cost increase as
Δf = − u1e1 =160 ( 225 p − 250 p ) = 800 p

Arora, Introduction to Optimum Design, 4e 5-57

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.51 ________________________________________________________________________________
A symmetric (area of member 1 is the same as area of member 3) three-bar truss problem is

described in Section 2.10.

1. Formulate the minimum mass design problem treating A 1 and A 2 as design variables.

2. Check for convexity of the problem.

3. Write KKT necessary conditions for the problem.

4. Solve the optimum design problem using the data: P=50 kN, θ=30°, ρ=7800 kg/m3,

σa=150 MPa. Verify the solution graphically and interpret the necessary conditions on

the graph for the problem.

5. What will be the effect on the cost function if σa is increased to 152 MPa?

1. Referring to Section 2.10, the design problem is formulated as

=
Minimize (
f r l 2 2 A1 + A2 , subject = )
to g1 s1 2 A1+ s 2 2 A1 + 2 A 2 − σa ≤ 0 ( )
g2 = 2 s2 ( A1+ )
2 A2 − σa ≤ 0, g3 = − A1 ≤ 0, g 4 = − A 2 ≤ 0

2. Check for convexity. Referring to Exercise 4.150, it is shown to be a convex programming

problem.

3. Write Kuhn-Tucker necessary conditions.

( ) (
L = ρ l 2 2 A 1+ A 2 + u1  s1 2 A1 + s2 2 A1 + 2 A 2 − σa  + u2  2 s2
   ) ( A 1+ )
2 A 2 − σa 

( )
+u3 − A1 + u4 − A 2 ( )
∂L ∂A1 =
 (  )
2 2ρl + u1  − s1 2 A12 − s2 2 A1 + 2 A2 2  + u2  − 2 s2 A1 + 2 A2 2  − u3 =

0 ( )
∂L ∂A 2 =
 (
  )  (
ρ l + u1  − s2 A1 + 2 A 2 2  + u2  − 2 s2 A 1+ 2 A 2 2  − u4 =
0 )
gi ≤ 0, ui gi = 0, ui ≥ 0; i = 1, 2, 3, 4

4. Solve the design problem. Use Newton, meter and kilogram as units for force, length and mass
respectively.
P = 50 kN = 5× 10 4 N, σ a = 150 MPa = 1.5 × 108 N/m2, q = 30°, ρ = 7800 kg/m3,
l = 1 m, s1 = P cos q =4.330127 × 104 N, s2 =
P sin q =
2.5 × 104 N
Substituting the above values we have:
Minimize f = 7.8 × 103 2 2 A1 + A2 ( )
Arora, Introduction to Optimum Design, 4e 5-58
 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

=
subject to g 1 4.330127 × 104 2 A 1 + 2.5 × 104 ( )
2 A1 + 2 A 2 − 1.5 × 108 ≤ 0

(
g 2 =2 2.5 × 104 ) ( A1 + )
2 A 2 − 1.5 × 108 ≤ 0; g3 =
− A1 ≤ 0; g 4 =
−A2 ≤ 0
The KKT conditions become

( 
) ( )
2
∂L= ∂A 1 2 2 7.8 × 103 + u1  − 4.330127 × 104 2 A12 − 2.5 × 104 2 A 1+ 2 A 2 
 

( )( )
2
+u2  − 2 2.5 × 104 A1 + 2 A 2  − u3 = 0 (1)
 

( )
2
∂L ∂A 2 = 7.8 × 103 + u1  −2.5 × 104 A 1+ 2 A 2 
 

(
+u2  −2 2.5 × 104

2
)(
A 1+ 2 A 2  − u4 =

0 ) (2)

gi ≤ 0, ui gi =0, ui ≥ 0 ; i = 1, 2, 3, 4
There are totally 16 cases, since there are 4 inequality constraints. We shall solve for the case
(u2 = u3 = u4 = 0, g1 = 0) which yields a solution. The solution is also a global solution because
this is a convex programming problem.
∂L ∂A2 = 7.8 × 103 − u1 2.5 × 104 ( (
2 A1 + 2 A2 ) )=
2
0 (3)

∂L= (  )
∂A1 2 2 7.8 × 103 + u1  −4.330127 × 104 2 A12 − 2.5 × 104 2 A1 + 2 A=
2
2

0 ( ) (4)

=
g1 4.330127 × 104 2 A 1+ 2.5 × 104 2 A 1+ 2 A 2 − 1.5 ×= 108 0 ( ) (5)
Solving the nonlinear system for unknowns, we get
A1 = 2.937 × 10− 4 , A 2 =6.556 × 10− 5 , u 1=4.658 × 10− 8 , and f ∗ = 7.0
5. If σ a is increased to 152 MPa, the cost will decrease since u 1 > 0. The amount will be
Δf = (
− 4.658 × 10−8 2 × 106 =
− u1Δe = − 0.0932 . )( )

Arora, Introduction to Optimum Design, 4e 5-59

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.52 ________________________________________________________________________________

A 100 ×100m lot is available to construct a multistory office building. At least 20,000 m2 total
floor space is needed. According to a zoning ordinance, the maximum height of the building can
be only 21m, and the area for parking outside the building must be at least 25 percent of the floor
area. It has been decided to fix the height of each story at 3.5m. The cost of the building in millions
of dollars is estimated at 0.6 h +0.001 A, where A is the cross-sectional area of the building per
floor and h is the height of the building. Formulate the minimum cost design problem.

Solution

Note : g4 = −A is not shown on the graph.

According to the graphical solution, the point P (14, 5000) is minimum point.

Arora, Introduction to Optimum Design, 4e 5-60

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

Referring to the formulation in Exercise 2.1 we have

Minimize f=0.6h + 0.001A, subject to: g1 =20,000 − hA 3.5 ≤ 0;
g 2 =(1 + h 14 ) A − 10,000 ≤ 0; g 3 =3.5 − h ≤ 0; g 4 =h − 21 ≤ 0; g 5 =− A ≤ 0
L= ( 0.6 h + 0.001A) + u1 ( 20,000 − hA )
3.5 + s12 + u2 (1 + h 14 ) A − 10,000 + s22 
( ) ( ) (
+u3 3.5 − h + s32 + u4 h − 21 + s42 + u5 − A + s52 )
∂L ∂h = 0.6 − u1 ( A 3.5) + u 2 ( A 14) − u 3 + u 4 = 0
∂L ∂A = 0.01 − u1 (h 3.5) + u 2 (1 + h 14) − u 5 = 0
g i + si2= 0, ui s=i 0, ui ≥ 0; =i 1 to 5; si2 ≥ 0, Regularity is satisfied.
There are 32 cases because we have five inequality constraints. The case which yields a solution is
u3 = u4 = u5 = 0, s1 = s2 = 0. The solution is h = 14, A = 5000, u1 =
5.9 × 10− 4 , u2 =
6.8 × 10− 4 ,
f = 13.4 mil. dollars. The solution can be verified graphically. It is seen that the point obtained
using the KKT conditions is indeed a minimum point.

Referring to Exercises 2.1/4.83, we have already shown the point A = 5000 and h = 14 with
u1 =5.9 × 10− 4 , u2 =6.8 × 10− 4 satisfies the necessary and sufficient conditions. The effect of
constraint limit on cost function value is due to the Sensitivity Theorem; Δ=
f u1e1 + u2e2 where e1
and e2 are the small change of active constraints respectively.

Arora, Introduction to Optimum Design, 4e 5-61

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.53 ________________________________________________________________________________
Referring to the formulation in Exercises 2.3/4.85, we have
Minimize f = − π R 2 H , subject to g1 =2πRH − 900 ≤ 0; g 2 =5 − R ≤ 0;
g3 =−
R 20 ≤ 0; g 4 = − H ≤ 0; g5 = H − 20 ≤ 0 . The optimum solution obtained by the graphical
method is R∗ = 20, H ∗ = 7.16, f ∗ = − 9000 where g 1 and g 3 are active.

1. Check for necessary conditions. Since only g 1 and g 3 are active, the Lagrange multipliers are
all zero except for u 1 and u 3 . Therefore,
L = −π R 2 H + u1 ( 2π RH − 900 ) + u3 ( R − 20 )
The KKT necessary conditions are
∂L ∂R =−2π RH + 2π Hu1 + u3 − 0; ∂L ∂H =−π R 2 + 2π Ru1 =0 ;
gi ≤ 0, ui gi =0, ui ≥ 0 ; i =1 to 5
Substituting the optimum values R ∗ and H ∗ , we get u 1 = 10 > 0, u 3 = 449.9 > 0 (o.k.). All the other
conditions are also satisfied. Thus, the point obtained from graphical method satisfies all the necessary
conditions.
2. Check for sufficient condition ( ∗ ). Since the number of active constraints is equal to the number
of design variables, the point (R ∗ = 20, H ∗ = 7.16) is indeed an isolated local minimum point. The
sufficient condition is deemed satisfied at this point.
3. Effect of variations in constraint limits on cost function ( ∗ ∗ ). According to the Constraint
Variation Sensitivity Theorem 4.7, we have
∂f ∂e1 =− u1 =−10 and ∂f ∂e3 =− u3 =− 499.9
where e 1 and e 3 are a small variation in the R.H.S. of g 1 and g 3 respectively. Small variations in
R.H.S. of the other constraints have no effect on the cost. By using Taylor series expansion, we can
show that the change in cost function is ∆ f = − u1e1 − u3 e1 =−10 e1 − 449.9e3 .
___________________
( ∗ ) The argument used in 2. is applied to other problems in this section for checking sufficient condition.
( ∗ ∗ ) The argument used in 3. is applied to other problems in this section for studying the effect of
variations in constraint limits on cost function.

Arora, Introduction to Optimum Design, 4e 5-62

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.54 ________________________________________________________________________________
Referring to the formulation in Exercises 2.4/4.86, we have:
Minimize f =− 2π NR , subject to g1 = 0.5 − R ≤ 0; g 2 =π NR 2 − 2000 ≤ 0; g3 =− N ≤ 0. The
optimum solution obtained by the graphical method is R ∗ = 0.5, N ∗ = 2550, f ∗ = 800 where g 1 and
g 2 are active.
1. Check for necessary conditions. Since only g 1 and g 2 are active, set u 3 = 0.
( )
L = −2πNR + u1 (0.5 − R ) + u2 πNR 2 − 2000
The KKT necessary conditions are
∂L ∂R =−2π N − u1 + 2π N − u1 + 2π NRu2 =0;
∂L ∂N = −2π R + π R 2u2 = 0; gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1 to 3
Substituting the optimum values, we obtain u 2 = 4 > 0, and u 1 = 16022 > 0 (o.k.). All the other
conditions are also satisfied. Therefore, the point obtained from graphical method satisfies the
necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. Effect of variations in constraint limits on cost function. Using the same argument as used in
Exercise 5.53, we can conclude that Δ f = − u1e1 − u2e2 =−16022e1 − 4e2

Arora, Introduction to Optimum Design, 4e 5-63

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.55 ________________________________________________________________________________
Referring to the formulation in Exercises 2.5/4.87, we have:
Minimize f = 200W + 100D, subject to g1 =W − 100 ≤ 0; g 2 =D − 200 ≤ 0; g3 =10,000 − WD ≤ 0;
g4 = D − 2W ≤ 0; g5 = W − 2 D ≤ 0; g6 = −W ≤ 0; g7 = − D ≤ 0. The optimum solution obtained by
the graphical method is W ∗ = 70.7, D ∗ = 141.4, f * = 28284 where g 3 and g 4 are active.
1. Check for necessary conditions. Since only g 3 and g 4 are active, set u1 = u2 = u5 = u6 = u7 = 0.
L = 200W + 100 D + u3 (10,000 − WD ) + u4 (D − 2W )
The KKT necessary conditions are:
∂L ∂W= 200 − u3 D − 2u4= 0;
∂L ∂D= 100 − u3 W + u4= 0;
gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1 to 7
Substituting the optimum values, we have u 3 = 1.414 > 0, u 4 = 0 ≥ 0 (o.k.). All the other conditions
are also satisfied. Therefore, the point obtained from graphical method satisfies the necessary
conditions.
2. Check for sufficient condition. Hessian of the Lagrangian is
 0 − u3   0 −1.414 
H=   =
 − u3 0   −1.414 0 
Since u 4 = 0, we only consider the gradient of g 3 .
 − D   −141.4 
c (1, − 2 ) , c ≠ 0.
T
Ñg3 = −W  =  −70.7  ; Ñg3 d =
T
0 gives d =
   
Q = dT H d = 5.656c 2 > 0 for c ≠ 0 (o.k.)
Therefore, the point ( W ∗ = 70.7, D ∗ = 141.4) satisfies sufficient condition.

3. Effect of variations in constraint limits on cost function. Using the same argument as in Exercise
5.53, Δ f = − u3 e3 =−1.414 e3 . Since the Lagrange multiplier u 4 = 0, a small variation in R.H.S. of
g 4 has no effect on the cost function.

Arora, Introduction to Optimum Design, 4e 5-64

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.56 ________________________________________________________________________________
Referring to the formulation in Exercises 2.9/4.91, we have
Minimize f =π r 2 + 2π rh − 600 =0; subject to h1 = πr 2 h − 600 = 0; g1 =1 − h 2r ≤ 0;
g 2 =h 2r − 1.5 ≤ 0; g3 =h − 20 ≤ 0; g 4 =−h ≤ 0; g5 =−r ≤ 0 . The optimum solution obtained by
the graphical method is r ∗ = 4.57, h ∗ = 9.14, f ∗ = 328 where h 1 and g 1 are active.
1. Check for necessary conditions. Since the active constraints are h 1 and g 1 , we can set
u2 = u3 = u4 = u5 = 0.

( )
L =π r 2 + 2π rh + v 1 π r 2 h − 600 + u 1 (1 − h 2r )
The KKT necessary conditions are:
(
∂L ∂r = 2π r + 2π h + 2π rhv1 + u 1 h 2r 2 = 0 ;)
∂L ∂r = 2π r + π r 2v1 − u 1 2r = 0; h1 = 0, gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1 to 5
Substituting the optimum values and solving for v 1 and u 1 , we get v 1 = −0.364365 , u 1 = 43.7562 > 0
(o.k.). All the other conditions are also satisfied. Therefore, the point ( r ∗ = 4.57, h ∗ = 9.14) satisfies
the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. Effect of variations in constraint limits on cost function. Using the same argument as used in
Exercise 5.53, Δ f = 0.3647b 1− 43.74e1 .
−v 1b 1− u 1e1 =

Arora, Introduction to Optimum Design, 4e 5-65

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.57 ________________________________________________________________________________
Referring to the formulation in Exercises 2.10/4.92, we have
Minimize f = 32 (b + 2h ) (15bh ) , subject to g1 =b − 10 ≤ 0; g 2 =h − 18 ≤ 0 . The optimum solution
obtained by the graphical method is b ∗ = 10, h ∗ = 18, f ∗ = 0.545 where g 1 and g 2 are active.
1. Check for necessary conditions. L = 32 (b + 2h ) (15bh ) + u1 (b − 10 ) + u2 (h − 18)
The KKT necessary conditions are:
∂L ∂b = (32 15)[bh − (b + 2h )h] (bh ) + u1 = 0;
2

( 32 15)  2bh − ( b + 2h ) b  (=
bh ) + u2
2
=
∂L ∂h 0;
gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1, 2
Substituting the optimum values, we obtain u 1 = 0.043 > 0, and u 2 = 0.0066 > 0 (o.k.). Thus, the
point (b ∗ = 10, h ∗ = 18) satisfies the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. Effect of variations in constraint limits on cost function. Using the same argument as used in
Exercise 5.53, Δ f = − u 1e1 − u 2e 2 = − 0.043e1 − 0.0066e 2 .

Arora, Introduction to Optimum Design, 4e 5-66

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.58 ________________________________________________________________________________
Referring to the formulation in Exercises 2.12/4.94, we have
=
Minimize f 200 p D 2 + 400 p DH , subject to
= h1 p D 2 H 4 −=
150 0 ; g1 = H + D 2 − 10 ≤ 0;
g2 =− H ≤ 0; g3 =− D ≤ 0 . The optimum solution obtained by the graphical method is D ∗ = 5.76,
H ∗ = 5.76, f ∗ = 62500, where h 1 is active.
1. Check for necessary conditions. Since only h 1 is active, set u 1 = u 2 = u 3 = 0.
L =200 p D 2 + 400 p DH + v1 ( p D 2 H 4 − 150 )
The KKT necessary conditions are: ∂L ∂D = 400p D + 400p H + p DH v1 2 = 0;
∂L ∂H= 400 pD + p D 2 v1 = 4 0; h=1 0, gi ≤ 0, ui g=
i 0, ui ≥ 0 ; i = 1, 3
Substituting the optimum values we obtain v 1 = − 277.6 . All other conditions are satisfied.
Therefore, the point obtained from graphical method satisfies the necessary conditions.
2. Check for sufficient condition. The Hessian of Lagrange function is
 400p + pHv1 2 400p + pDv1 2   −1255 − 1255
Ñ2 L =  =
 400p + pDv1 2 0   −1255 0 
 πDH 2   52.1 
c (1, − 2 ) , c ≠ 0.
T
Ñh 1=  2  =  26.05 ; Ñh1 d =
T
0 gives d =
 πD 4   
( )
Q = dT Ñ 2 L d = 3765c 2 > 0 if c ≠ 0

( ) ( ) (
L = 1 − P 1 + P 12 + 1 + 0.6P2 + u 1 60 − P1 − P2 )
The KKT necessary conditions are:
∂L ∂P1 =−1 + 2 P1 − u1 =0
∂L ∂P 2 = 0.6 + 2 P 2 − u1 = 0; gi = 0, ui gi = 0, ui ≥ 0 ; i = 1 to 3
Substituting the optimum value, we obtain u 1 = 59.8 > 0. Therefore, the necessary conditions are
satisfied.
2 0
2. Check for sufficient condition. The Hessian of Lagrangian is: Ñ 2 L =  
0 2
Since Hessian of Lagrangian is positive definite, the sufficient condition is satisfied.
3. Effect of variations in constraint limits on cost function. Use the same argument as in
Exercise 5.53, Δf = −u1e1 = −59.8e1 .

Arora, Introduction to Optimum Design, 4e 5-67

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.59 ________________________________________________________________________________
Referring to the formulation in Exercises 2.14/4.96, we have
( ) ( )
Minimize f = 1 − P1 + P12 + 1 + 0.6 P2 + P22 ; subject to g1 = 60 − P1 − P2 ≤ 0;
g2 =− P1 ≤ 0; g3 =− P2 ≤ 0.
=
The optimum solution obtained by the graphical method =
is P1* 30 =
.4 , P2* 29.6 , f * 1790 where
g 1 is active.

1. Check for necessary conditions. Since only g 1 is active, set u 2 = u 3 = 0.

( ) ( )
L = 1 − P1 + P12 + 1 + 0.6 P2 + P22 + u1 ( 60 − P1 − P2 )
The KKT necessary conditions are:
∂L/∂P1 =−1 + 2 P1 − u1 =0; ∂L / ∂P2 =0.6 + 2 P2 − u1 =0; gi =0; ui gi =0; ui ≥ 0; i =
1 to 3
Substituting the optimum value, we obtain u1 = 59.8 > 0. Therefore, the necessary conditions are
satisfied.
2 0
2. Check for sufficient condition. The Hessian of Lagrangian is: Ñ 2 L =  
0 2
Since Hessian of Lagrangian is positive definite, the sufficient condition is satisfied.

3. The effect of variations in constraint limits on cost function. Use the same argument as in
Exercise 5.53, Δ f = − u1e1 =− 59.8e1.

5.60 ________________________________________________________________________________
Referring to the formulation in Exercises 2.23/5.38, we have
( )
Minimize f = 12.331 Ro2 − Ri2 ; subject to =
g1 ( 6.3662 ×106 ) Ro ( Ro4 − Ri4 ) − 2.5 ×104 ≤ 0;
(
g 2 = 4244.132 Ro2 + Ro Ri + Ri2 ) ( Ro4 − Ri4 ) − 9000 ≤ 0; g3 = Ro − 20 ≤ 0; g 4 = Ri − 20 ≤ 0;

g5 =− Ro ≤ 0; g6 =− Ri ≤ 0. The optimum solution obtained by the graphical method is Ro∗ = 20,

Ri∗ = 19.84, f ∗ = 79.1 where g 1 and g 3 are active.

1. Check for necessary conditions. This point satisfies the necessary conditions which have been
checked in Exercise 5.38.
2. Check for sufficient condition. Use the same argument in Exercise 5.53.
3. The effect of variations in constraint limits on cost function. Using the same argument used in
Exercise 5.53, Δ f = − 3.5596 × 10−3 e 1− 5.29e 3.
− u 1e 1− u 3e 3 =

Arora, Introduction to Optimum Design, 4e 5-68

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.61 ________________________________________________________________________________
Formulation is given in Exercise 2.24. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.39.

Rewriting the formulation of Exercise 2.24, we have

Minimize f = 1.44086 Rt , subject to
= g1 7957.7 Rt − 250 ≤ =
0 ; g 2 767.8 R 2 − 3 ≤ 0;
g3 = 20 − R t ≤ 0; g 4 = 20 − R ≤ 0; g5 = R − 200 ≤ 0; g 6 = 1 − t ≤ 0; g 7 = t − 10 ≤ 0
The optimum solutions found by the graphical method are the points on g 1 between ( 31.83 , 1.0 ) and
( 25.23, 1.26 ) , where f ∗ = 45.9 and g 1 is active.

1. Check for necessary conditions. Since only g 1 is active, we can set u2 = u3 = u4 = u5 = u6 = u7 = 0

and s 1 = 0.
L= 1.44086 Rt + u1 ( 7957.7 Rt − 250 ) .
The KKT necessary conditions are
( ) (
∂L ∂t= 1.44086 R + u1 −7957.7 Rt 2 = 0; ∂L ∂R= 1.44086t + u1 −7957.7 R 2t = 0 )
gi + si2= 0, ui si= 0, ui ≥ 0; i = 1 to 7
Using the relation, 1.44086Rt = 45.9 at optimum solutions, and either equation, we can obtain u 1 =
0.183670. All the other conditions are also satisfied. Thus, the solutions satisfy all of the necessary
conditions.
Ñ= 2
L Ñ 2 f + u1Ñ 2 g1
2. Check for sufficient condition.
 0 1.44086   2 ( 7957.7 ) Rt 3 7957.7 R 2 t 2 
1.44086 + ( 0.1836 )  
 0   7957.7 R t
2 2
2 ( 7957.7 ) R 3 t 
Substituting 1.44086Rt = 45.9 into Ñ 2 L and Ñg1 , we have
 2.88 R t 2.88   −249.8 t 
Ñ2 L =  , Ñg1  
 2.88 2.88 t R   −249.8 R 
c (1, − R t ) , c ≠ 0 . Hence,
T
Ñg1T d = 0 gives d = = Q d= T
Ñ2 L d 0
Thus, the sufficient condition is not satisfied. Hence solution is not an isolated local minimum.
The Lagrange multiplier of active constraint is u 1 = 0.1836. Therefore,
Δ f = − u 1e 1 = − 0.1836e 1 .

Arora, Introduction to Optimum Design, 4e 5-69

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.62 ________________________________________________________________________________
Optimum solution and check of necessary and sufficient conditions can be found in Exercise 5.41.
Referring to the formulation of Exercise 4.80, we have
Minimize f = 2.4662 ×105 R t , subject=
to g1 795.775 Rt − 2.5 × 108 ≤ 0 ;
g 2 =5 × 104 − 6.5113 × 1010 R 3t ≤ 0 ; g3 =R t − 50 < 0 ; g 4 =− R ≤ 0 ; g5 =−t ≤ 0
=
The optimum solution found by the graphical method is R∗  0=
.0787, t ∗  0.00157,
= f ∗  30.56 kg;
where g 2 and g 3 are active.

1. Check for necessary conditions. Since only g 2 and g 3 are active, we can set u1 = u4 = u5 = 0 and
s2 = s3 = 0. The KKT necessary conditions are
∂L = ( )( )
∂R 2.4662 × 105 t + u2  −6.5113 × 1010 3R 2 t  + u3 (1= t) 0 (1)
( ) ( )
∂L ∂t = 2.4662 × 105 R + u2 − 6.5113 × 1010 R 3 + u3 − R t 3 = 0 (2)
gi + s = 0, ui s=
2
i i 0, ui ≥ 0; =i 1 to 5 (3)
Substituting the optimum value into (1) and (2), we get
u2 = 3.056 × 10 −4 > 0, u3 = 0.3038 > 0 (o.k.).
(
All of the other constraints in (3) are also satisfied. Thus, the point R ∗ = 0.0787, t ∗ = 0.00157 )
satisfies the necessary conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints is equal
to the number of design variables, the point is indeed an isolated local minimum.
The Lagrange multipliers of the active constraints are u 2 = 3.056 ×10− 4 , u3 = 0.3038.
Therefore, Δ f = − 3.056 × 10− 4 e 2 − 0.3038 e 6 .
− u 2e 2 − u 6e 6 =

Arora, Introduction to Optimum Design, 4e 5-70

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.63 ________________________________________________________________________________
The formulation is given in Exercise 3.24. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.42.

Referring to the formulation of Exercise 4.81, we have

( )
Minimize f = 1.2331 × 105 Ro2 − Ri2 , subject to = ( )
g1 1.59155×104 Ro2 − Ri2 − 2.5 × 108 ≤ 0;
g 2 =× (
5 104 − 1.62783 × 1010 Ro4 − Ri4 ≤ 0; g3 =)
0.5 ( Ro + Ri ) − 50 ≤ 0; g 4 =− Ro ≤ 0; g5 =− Ri ≤ 0
=
The optimum solution found by the graphical method is Ro∗  0=
.0795, Ri∗  0=
.0779 , f *  30.56 kg
where g 2 and g 3 are active.

1. Check for necessary conditions. Since only g 2 and g 3 are active, we can set u1 = u4 = u5 = 0, and s 2
and s 3 = 0.
( ) ( )
L= 1.2331 × 105 Ro2 − Ri2 + u2 5 × 104 − 1.62783 × 1010 Ro4 − Ri4  + u3 0.5 ( Ro + Ri ) ( Ro + Ri ) − 50 
The KKT necessary conditions are
∂L = ( ) ( )( )
∂Ro 2.4662 × 105 Ro + u2  − 1.62783 × 1010 4 Ro3  + u3  − Ri ( Ro − R=
 i)
2
 0 (1)

∂L ∂Ri = ( ) ( )( )
− 2.4662 × 105 Ri + u2  1.62783 × 1010 4 Ri3  + u3  Ro ( Ro − Ri )  =

2


0 (2)
gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 5 (3)
Substituting the optimum value into (1) and (2), we get
u2 = 3.056 × 10 −4 > 0, u3 = 0.3055 > 0 (o.k.).
point Ro∗  0=
All other conditions in (3) are satisfied. Thus, the= .0795, Ri∗  0.0779 satisfies the
necessary conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints is
equal to the number of design variables, the point is indeed a local minimum. Sufficient
condition is deemed satisfied at this point.

The Lagrange multipliers of active constraints are u 2 = 3.056 ×10− 4 , u3 = 0.3055. Therefore,
Δf = −3.056 × 10−4 e 2− 0.3038e 3 .
− u 2e 2 − u 4e 4 =

Arora, Introduction to Optimum Design, 4e 5-71

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.64 ________________________________________________________________________________
The formulation is given in Exercise 4.79. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.43.

Referring to Exercise 4.79, we have

Minimize f = π DH + π D 2 2 , subject to g=
1 400 − π D 2 H 4 ≤ 0; g=
2 3.5 − D ≤ 0;
g3 = D − 8 ≤ 0; g 4 = 8 − H ≤ 0; g5 = H − 18 ≤ 0
The optimum solution found by the graphical method is= H ∗ 8,= D∗  7.98,= f ∗  300.6 cm 2 where
g 1 and g 4 are active.
1. Check for necessary conditions. Since only g 1 and g 4 are active, we can set u2 = u3 = u5 = 0 and
s 1 = s 4 = 0.
( )
L = π DH + π D 2 2 + u1 400 − π D 2 H 4 + u4 ( 8 − H )
The KKT necessary conditions are
( )
∂L ∂D = π H + π D + u1 ( −π DH 2 ) = 0; ∂L ∂H = π D + u1 −π D 2 4 − u4 = 0
gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 5
Substituting the optimum value into equations, we get u 1 = 0.5 > 0, u 4 = 0.063 > 0 (o.k.). All the
other conditions in (3) are also satisfied. Thus, the point (H ∗ = 8, D ∗ = 7.98) satisfies the necessary
conditions.

2. Check for sufficient condition. Since this is the case that the number of active constraints is equal
to the number of design variables, the point is indeed local minimum point.

The Lagrange multipliers of active constraints are u1 = 0.5, u4 = 0.063 . Therefore,

Δf= − 0.5e 1− 0.063e 4 .
− u 1e 1− u 4e 4 =

Arora, Introduction to Optimum Design, 4e 5-72

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.65 ________________________________________________________________________________
The formulation is given in Exercise 3.28. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.40.

Rewrite the formulation of Exercise 3.28, we have

Minimize f = 50.077 R t + 0.1885 ( R + t 2 ) , subject to g1 =25 − 25.038 ( R − t 2 ) ≤ 0 ;
2 2

g 2= 3.5 R t − 210 ≤ 0 ; g3= (1.41667 × 10 ) R t − 0.001 ≤ 0 ;

−5
g 4= 0.5 t − R ≤ 0 ;
g5 =− R ≤ 0; g 6 =−t ≤ 0
is R∗ 1=
The optimum solution obtained by the graphical method= .0, t ∗ 0.0167, f ∗ = 1.0 where g 1
and g 2 are active.

1. Check for necessary conditions. Since only g 1 and g 2 are active, we can set u3 = u4 = u5 = 0 and
s1 = s2 = 0.
L = 50.077 Rt + 0.1885 ( R + t 2 ) + u1  25 − 25.038 ( R + t 2 )  + u2 ( 3.5 R t − 210 )
2 2

 
∂L ∂=R 50.077t + 0.377 ( R + t 2 ) − 2u1 ( 25.038 )( R + t 2 ) + u2 ( 3.5 t=) 0 (1)
∂L = (
∂t 50.077 R + 0.1885 ( R + t 2 ) + u1 ( 25.038 )( R + t 2 ) − u2 3.5 R t=
2
0 ) (2)
gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 5 (3)
Substituting the optimum value into (1) and (2);
1.216434 − 1.9833 =
u1 + u 2 0, 50.267074 + 0.99165=
u1 − 60u 2 0
Solving for u1 and u2 , we get u1 = 0.0417 > 0 (o.k.), u2 = 4.080 × 10 −3 > 0 (o.k.). All the other
conditions in (3) are satisfied. Therefore, the point ( R ∗ = 1.0m, t ∗ = 0.0167m) satisfies all of the
necessary conditions.

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
design variables, the point ( R ∗ = 1.0, t ∗ = 0.0167) is indeed a local minimum.

The Lagrange multipliers of active constraints are

u 1 = 0.0417, u 2 = 4.080 × 10 −3 . Therefore, Δ f = − 0.0417e 1− 4.080 × 10−3 e 2 .
− u 1e 1 − u 2e 2 =

Arora, Introduction to Optimum Design, 4e 5-73

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.66 ________________________________________________________________________________
The formulation is given in Exercise 3.34*. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.45.

Rewriting the formulation of Exercise 3.34, we have

( ) ( )
Minimize f = 3.083 × 10 −3 x12 1 − x22 , subject to g1= 5.093 × 107 x13 1 − x24 − 275 ≤ 0;
=
g 2 6.36619 × 105 x14 (1 − x ) − 3.49066 × 10
4
2
−2
≤ 0; g3 = 2.0 × 107 − ( 4.17246 × 10 ) x (1 − x )
4 3
1 2
2.5
≤ 0;
g 4 = 20 − x1 ≤ 0; g5 = x1 − 500 ≤ 0; g 6 = 0.6 − x2 ≤ 0; g 7 = x2 − 0.999 ≤ 0
is x1∗ 103,
The optimum solution obtained by the graphical method= = x2∗ 0.955,
= f ∗ 2.9 where g 1
and g 3 are active.

1. Check for necessary conditions. Since only g 1 and g 3 are active, we can set
u 2 = u 4 = u 5 = u 6 = u 7 = 0.
( ) (
L= 3.083 × 10−3 x12 1 − x22 + u1 5.093 × 107 x13 1 − x24 − 275 )
 ( )
+ u3  2.0 × 107 − 4.17246 × 104 x13 (1 − x2 ) 
2.5


The KKT necessary conditions are
∂L ∂= ( ) (
x1 6.166 × 10−3 x1 1 − x22 + u1  −3 5.093 × 107 x14 1 − x24 ) ( )
 ( )( )
+u3  − 4.17246 × 104 3 x12 (1 − x2 )  =
2.5


0 (1)

( )
∂L ∂x2 =−6.166 × 10−3 x12 x2 + u1  4 5.093 × 107 x23 x13 1 − x24  ( )
2

 

 ( )( )
+u3  − 4.17246 × 104 3 x12 (1 − x2 )  =
2.5


0 (2)
gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1 to 7 (3)
−3
Substituting the optimum values into (1) and (2), respectively, we get u1 = 4.568 × 10 > 0,
u3 3.332 × 10− 8 > 0 (o.k.). All the other conditions in (3) are also satisfied. Therefore, the point
=
(x∗
1 )
= 103, x2∗ = 0.955 satisfies all of the necessary conditions.

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point x1∗ = 103, x2∗ = 0.955 is indeed a local minimum point.

The Lagrange multipliers of active constraints are u 1 = 4.568 ×10− 3 , u 3 = 3.332 ×10− 8 . Therefore,
Δf = − 4.568 × 10− 3 e1 − 3.332 × 10− 8 e 3 .
− u 1e 1− u 3e 3 =

Arora, Introduction to Optimum Design, 4e 5-74

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.67 ________________________________________________________________________________
The formulation is given in Exercise 3.35*. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.46.

Rewriting the formulation of Exercise 3.35, we have

(
Minimize f = 3.083 × 10 −3 d o2 − d i2 , subject to g=
1 )
5.093 × 107 d o ( ) (d 4
o )
− di4 − 275 ≤ 0;
g=
2 ( 6.3662 × 10 ) ( d − d ) − 3.49066 × 10
5 4
o i
4 −2
≤ 0;
= 2.0 × 10 − ( 4.17246 × 10 ) d (1 − d d )
2.5
g3 7 4 3
o i o ≤ 0; g 4 = 20 − d o ≤ 0;
g5 =d o − 500 ≤ 0; g 6 =0.6 − di d o ≤ 0; g 7 =di d o − 0.999 ≤ 0;
The optimum solution obtained by the graphical method is d o∗ = 103, d i∗ = 98.36, f ∗ = 2.9 where g1
and g 3 are active.

1. Check for necessary conditions. Since only g1 and g 3 are active, we can set u2 = u4 = u5 = u6 = u7
= 0.
( ) ( ) (
L = 3.083 × 10−3 do2 − di2 + u1  5.093 × 107 do do2 − di2 − 275
  )
+ u3  2.0 × 107 − ( 4.17246 × 104 ) do3 (1 − di do ) 
2.5
 
The KKT necessary conditions are

( 
) ( ) ( )
2
∂L ∂d= o 6.166 × 10−3 do + u1 5.093 × 107 1 do4 − di4 − 4do do4 − di4 
 
( )( )
− u3 4.17246 × 104  3do2 (1 − di do ) + 2.5 ( di do )(1 − di do )  =
2.5 1.5

0 (1)


( ) ( )( )
2
∂L ∂di =− 6.166 × 10−3 di + u1  5.093 × 107 do 4di3 do4 − di4 
 

( )
+u3  4.17246 × 104 ( 2.5 ) d02 (1 − di do )  =
1.5

0 (2)
gi ≤ 0, ui gi = 0, ui ≥ 0 ; i = 1 to 7 (3)
Substituting the optimum values into (1) and (2) respectively, we get
u3 3.281 × 10−8 > 0 (o.k.).
u 1 = 4.657 × 10 −3 > 0,=
All the other conditions in (3) are also satisfied. Therefore, the point obtained from graphic method
satisfies all of the necessary conditions.
2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point d o∗ = 103, d i∗ = 98.36 is indeed an isolated local minimum point. The
sufficient condition is deemed satisfied at this point.

The Lagrange multipliers of active constraints are

u 1 = 4.657 ×10− 3 , u 3 = 3.281 ×10− 8 .
Therefore, Δ f = − 4.657 × 10−3 e 1 − 3.281 × 10−8 e 3 .
− u 1e1 − u 3e 3 =

Arora, Introduction to Optimum Design, 4e 5-75

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.68 ________________________________________________________________________________
The formulation is given in Exercise 3.36*. Optimum solution and check of necessary and sufficient
conditions can be found in Exercise 5.47.

Rewriting the formulation of Exercise 3.36, we have

Minimize f = 0.02466 R t , subject to g=
1 ( 3.1831× 106 ) ( 2R + t ) ( 4R3t + Rt 3 ) − 275 ≤ 0;
(
g 2 = 3.97886 × 104 ) ( 4 R t + Rt ) − 3.49066 × 10
3 3 −2
( )
≤ 0; g3 = 2.0 × 107 − 3.37972 × 105 ( R + 0.5t ) t 2.5 ≤ 0;
0.5

g 4 = 50 − R ≤ 0; g5 = R − 200 ≤ 0; g 6 = 2 − t ≤ 0; g 7 = t − 40 ≤ 0
The optimum solution obtained by the graphical method is R ∗ = 50.3, t ∗ = 2.34, f ∗ = 2.9 where g 1
and g 3 are active.
1. Check for necessary conditions. Since only g 1 and g 3 are active, we can set
u=2 u=4 u=
5 u=
6 u=
7 0.
( )
L= 0.02466 Rt + u1  3.1831 × 106 ( 2 R + t ) 4 R 3t + Rt 3  ( )
 ( )
+u3  2.0 × 107 − 3.37972 × 105 ( R + 0.5t ) t 2.5 
0.5


The KKT necessary conditions are
( ) ( )
L ∂R 0.02466t + u1 3.1831 × 106  2 4 R 3t + Rt 3 − ( 2 R + t ) 12 R 2 + t 3 ( ) ( 4R t + Rt ) 
2
∂= 3 3

(  )
− u3 3.37972 × 105 ( 12 ) t 2.5 ( R + 0.5t )
0.5
=

0 (1)

( ) (
∂L ∂t 0.02466 R + u1 3.1831 × 106 1 4 R 3t + Rt 3 − ( 2 R + t ) 4 R 3 + 3Rt 2) ( ) ( 4R t + Rt ) 
2
= 3 3

( )
− u3 3.37972 × 105 ( 14 ) t 2.5 ( R + 0.5 t ) + 2.5t1.5 ( ) ( R + 0.5 t ) 0.5
=

0 (2)
gi ≤ 0, ui gi =0, ui ≥ 0 ; i = 1 to 7 (3)
−3
Substituting the optimum value into (1) and (2) respectively, we obtain u 1 = 4.643 × 10 > 0, u 3 =
3.240 × 10 −8 > 0 (o.k.). All the other conditions in (3) are also satisfied. Therefore, the point obtained
from graphical method satisfies all the necessary conditions.

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point R ∗ = 50.3, t ∗ = 2.34 is indeed an isolated local minimum point. The
sufficient condition is deemed satisfied at this point.

The Lagrange multipliers of active constraints are

u 1 = 4.643 ×10− 3 , u 3 = 3.240 ×10− 8 .
Therefore, Δ f = − 4.643 × 10−3 e 1− 3.240 × 10−8 e 3 .
− u 1e 1− u 3e 3=

Arora, Introduction to Optimum Design, 4e 5-76

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.69 ________________________________________________________________________________
(a) Referring to the formulation in Exercise 3.39 (3.23), we have
Minimize f = 0.2466Rt, subject= to g 1 7957.5 ( Rt ) − 250 ≤ 0; g 2 =×
5 104 − 0.26045 R3t ≤ 0;
g3 = R t − 50 ≤ 0; g 4 = 10 − R ≤ 0; g5 = R − 1000 ≤ 0; g6 = 5 − t ≤ 0; g7 = t − 200 ≤ 0
The optimum solution is obtained by the graphical method is R ∗ = 33.7, t ∗ = 5.0, f ∗ = 41.6 where
g 2 and g 6 are active.
1. Check for necessary conditions.
Since only g 2 and g 6 are active, we can set u 1 = u 3 = u 4 = u 5 = u 7 = 0.
=L 0.2466 Rt + u2 5 × 104 − 0.26045 R3t  + u6 ( 5 − t )
 
The KKT necessary conditions are:
∂R 0.266t − u2 ( 0.78135 ) R=
∂L = 2
t 0; ( )
∂L ∂t 0.2466 R − u2 ( 0.26045 R 3=
= ) − u6 0;
gi ≤ 0, ui gi =0, ui ≥ 0 ; i =1 to 7
Substituting the optimum values, we obtain u 2 = 2.779 ×10− 4 > 0, u 6 = 5.540 > 0 (o.k.). All the
other conditions are also satisfied. Thus, the point (R ∗ = 33.7, t ∗ = 5.0) satisfies the necessary
conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δf = − u2e2 − u6e6 =− 2.779 × 10− 4 e2 − 5.54e6
(b) Referring to the formulation in Exercise 3.42 (3.24), we have
( )
Minimize f = 0.1233 Ro2 − Ri2 ; subject = to g1 15915.5 ( Ro2 − Ri2 ) − 250 ≤ 0;
g 2 =5 × 105 − 0.06511( Ro4 − Ri4 ) ≤ 0; g3 =( Ro − Ri )  2 ( Ro − Ri )  − 50 ≤ 0; g 4 = Ri − Ro + 5 ≤ 0;
g5 = Ro − Ri − 200 ≤ 0; g 6 = − 0.5 ( Ro + Ri ) + 10 ≤ 0; g=7 0.5 ( Ro + Ri ) − 1000 ≤ 0
The optimum solution obtained by the graphical method is Ro∗ = 36.0, Ri∗ = 31.0, f ∗ = 41.3 where g 2
and g 4 are active.
1. Check for necessary conditions.
Since only g 2 and g 4 are active, we can set u 1 = u 3 = u 5 = u 6 = u 7 = 0.
L = 0.1233 ( Ro2 − Ri2 ) + u2 5 × 105 − 0.06511( Ro4 − Ri4 )  + u4 ( Ri − Ro + 5 )
The KKT necessary conditions are:
( )
∂L ∂Ro = 0.2466 Ro − u2 0.026044 Ro3 − u4 = 0; ∂L ∂Ri =−0.2466 Ri + u2 ( 0.26044 Ri3 ) + u4 =0;
gi ≤ 0, ui gi =0, ui ≥ 0; i = 1 to 8
Substituting the optimum values, we get u 2 = 2.807 2.807 × 10 −4 > 0, u 4 = 5.467 > 0 (o.k.). All the
( )
other conditions are also satisfied. Thus, the point Ro∗ = 36.0, Ri∗ = 31.0 satisfies the necessary
conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δf = − 2.807 × 10− 4 e2 − 5.467e4
− u2 e2 − u4 e4 =

Arora, Introduction to Optimum Design, 4e 5-77

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.70 ________________________________________________________________________________
(a) Referring to the formulation in Exercise 3.40 (3.23), we have
Minimize f = 0.2466Rt, subject= to g1 7957.7 ( Rt ) − 250 ≤ 0; g 2 =×
5 104 − 1.04181R 3t ≤ 0;
g3 =− R t − 50 ≤ 0; g 4 =
10 − R ≤ 0; g5 = R − 1000 ≤ 0; g 6 = 5 − t ≤ 0; g 7 = t − 200 ≤ 0
The optimum solution obtained by the graphical method is R ∗ = 21.3, t ∗ = 5.0, f ∗ = 26.0 where g 2
and g 6 are active.
1. Check for necessary condition.
Since only g 2 and g 6 are active, we can set u 1 = u 3 = u 4 = u 5 = u 7 = 0.
L = 0.2466Rt + u 2 (5 × 10 4 − 1.04181R 3t ) + u 6 (5 − t )
The KKT necessary conditions are:
( )
∂L ∂R = 0.2466t − u2 3.12543R 2t = 0;
∂L ∂t = 0.2466 R − u (1.04181R ) − u
2
3
6 = 0;
g i ≤ 0, ui g i = 0 , ui ≥ 0; i = 1 to 7
Substituting the optimum values, we obtain u 2 = 1.739 ×10− 4 > 0, u 6 = 3.491 > 0 (o.k.). All the other
conditions are also satisfied. Thus, the point (R * = 21.3, t * = 5.0) satisfies the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δf = − u2 e2 − u6 e6 =−1.739 × 10− 4 e2 − 3.491e6

(b) Referring to the formulation in Exercise 3.43(3.24), we have

Minimize f = 0.1233( Ro2 − Ri2 ), subject to g 1 = 15915.5/( Ro2 − Ri2 ) - 250 ≤ 0;
g 2 = 5 ×104 − 0.26045 ( Ro4 − Ri4 ) ≤ 0; g3 =  2 ( Ro − Ri )  − 50 ≤ 0;
( Ro + Ri )
g 4 = Ri − Ro + 5 ≤ 0; g5 =Ro − Ri − 200 ≤ 0; g 6 =− 0.5 ( Ro + Ri ) + 10 ≤ 0;
g 7 = 0.5 (Ro + Ri ) − 1000 ≤ 0;
The optimum solution obtained by the graphical method is Ro∗ = 24.0 , Ri∗ = 19.0 , f * = 26.0 where g 2
and g 4 are active.
1. Check for necessary conditions.
Since only g 2 and g 4 are active, we can set u 1 = u 3 = u 5 = u 6 = u 7 = 0.
( )
L = 0.1233( Ro2 − Ri2 ) + u 2 [5 × 10 4 − 0.26045 R o4 − Ri4 ] + u 4 (R i – R o +5)
The KKT necessary conditions are:
( )
∂L ∂Ro = 0.2466 R o − u2 1.0418 Ro3 − u4 = 0;
∂L ∂R=
i 0.2466 R i + u (1.0418 R ) + u=
2 i
3
4 0; gi ≤ 0, ui g=
i 0, ui ≥ 0; =
i 1 to 8

Substituting the optimum values, we get u 2 = 1.699 ×10− 4 > 0, u4 = 3.471 > 0 (o.k.). All the other
conditions are also satisfied. Thus, the point ( Ro∗ = 24.0 , Ri∗ = 19.0 ) satisfies the necessary
conditions.
2. Check for sufficient conditions. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δf = −u2 e2 − u4 e4 = −1.699 × 10− 4 e2 − 3.471e4

Arora, Introduction to Optimum Design, 4e 5-78

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.71 ________________________________________________________________________________
(a) Referring to the formulation in Exercise 3.41 (3.23), we have
Minimize f = 0.2466Rt, subject to g 1 = 7957.7/(Rt) – 250 ≤ 0;
g 2 =5 × 104 − 0.52091R 3 t ≤ 0; g 3 =R t − 50 ≤ 0; g 4 =10 − R ≤ 0;
g 5 = R − 1000 ≤ 0; g 6 = 5 − t ≤ 0; g 7 = t − 200 ≤ 0;
The optimum solution obtained by the graphical method is R * = 27.0, t * = 5.0, f * = 33.0 where g 2
and g 6 are active.
1. Check for necessary conditions.
Since only g 2 and g 6 are active, we can set u 1 = u 3 = u 4 = u 5 = u 7 = 0.
L = 0.2466Rt + u 2 (5 × 10 4 − 0.52091R3t) + u 6 (5 – t)
The KKT necessary conditions are:
∂L ∂R = 0.2466t – u 2 (1.56273R3t) = 0; = ∂L ∂t 0.2466 R − u2 ( 0.52091R 3= ) − u6 0;
gi ≤ 0, ui gi =0, ui ≥ 0; i =1 to 7
Substituting the optimum values, we obtain u 2 = 2.165 × 10 −4 > 0, u 6 = 4.439 > 0 (o.k.). All the other
conditions are also satisfied. Thus, the point (R* = 27.0, t* = 5.0) satisfies the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δf = −u2e2 − u6e6 = −2.065 × 10−4 e2 − 4.439e6 .

(b) Referring to the formulation in Exercise 3.43 (3.24), we have

( )
Minimize f = 0.1233 Ro2 − Ri2 , subject to g 1 = 15915.5 / ( Ro2 − Ri2 ) − 250 ≤ 0;
5 104 − 0.13022 ( Ro4 − Ri4 ) ≤ 0; g3 =
g 2 =× ( Ro + Ri )
 2 ( Ro − Ri )  − 50 ≤ 0;
g 4 =( Ri − Ro ) + 5 ≤ 0; g5 =Ro − Ri − 200 ≤ 0; g 6 =− 0.5 ( Ro − Ri ) + 10 ≤ 0; = g 7 0.5 ( Ro + Ri )
The optimum solution obtained by the graphical method is R ∗o = 29.5, R ∗i = 24.5, f * = 33.0 where g 2
and g 4 are active.
1. Check for necessary conditions.
Since only g 2 and g 4 are active, we can set u 1 = u 3 = u 5 = u 6 = u 7 = 0.
=L 0.1233 ( Ro2 − Ri2 ) + u2 5 × 104 − 0.13022  + u4 ( Ri − Ro + 5 )
The KKT necessary conditions are:
(
∂L ∂Ro = 0.2466 Ro = 0.2466 R o − u2 0.52088 Ro3 − u4 = 0 )
∂L ∂R i =− 0.2466 Ri + u2 ( 0.52088 R 3
i )+u 4 =0;
gi ≤ 0, ui gi =0, ui ≥ 0; i = 1 to 8
Substituting the optimum values, we get u 2 = 2.1586 × 10 −4 > 0, u 4 = 4.388 > 0 (o.k.). All the other
conditions are also satisfied. Thus, the point ( Ro∗ = 29.5, Ri∗ = 24.5 ) satisfies the necessary
conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δ f = − u2 e2 − u4 e4 = −2.1586 × 10 −4 e2 − 4.388e4 .

Arora, Introduction to Optimum Design, 4e 5-79

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.72 _______________________________________________________________________________
Referring to the formulation in Exercise 3.46, we have
Minimize f = 9.8125 × 10 −3 (A 1 + A 2 ),
subject to g 1 = (202.97A 1 +512.95A 2 )/( A 1 A 2 ) – 50 ≤ 0; g 2 = 103420/A 1 – 250 ≤ 0;
g 3 = A 1 – 5000 ≤ 0; g 4 = 40925/ A 2 – 250 ≤ 0; g 5 = A 2 – 5000 ≤ 0;
The optimum solution obtained by the graphical method is A 1∗ = 413.68, A ∗2 = 163.7, f * = 5.7 where
g 2 and g 4 are active.

1. Check for necessary conditions. Since only g 2 and g 4 are active, we can set u 1 = u 3 = u 5 = 0.
( ) (
L = 9.8125 ×10− 3 A1 + A2 + u2 (103420 A1 − 250 ) + u4 40925 A2 − 250 )
The KKT necessary conditions are:
( )
∂L ∂A1 = 9.8125 × 10 −3 − u2 103420 A12 = 0;
( )
∂L ∂A1 = 9.8125 × 10 −3 − u2 103420 A12 = 0;
g i ≤ 0, ui g i ≤ 0, ui ≥ 0 ; i =1 to 5
From the equations, we get u 2 = 1.6237 × 10−2 , u4 = 6.425 × 10 −3 > 0 (o.k.). All the conditions are
satisfied. Thus, the point (A 1∗ = 413.68, A ∗2 = 163.7) satisfies the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function

Arora, Introduction to Optimum Design, 4e 5-80

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.73 _______________________________________________________________________________
Referring to the formulation in Exercise 3.47, we have
Minimize f = 0.1233Rt, subject to g 1 = 16459.5/(Rt) – 250 ≤ 0 ;
g 2 = 1.034188 × 105 − 4.16726 R 3t ≤ 0; g3 = R t − 50 ≤ 0;
g 4 = 20 − R ≤ 0; g5 = R − 400 ≤ 0; g 6 = 2 − t ≤ 0; g 7 = t − 40 ≤ 0
The optimum solutions obtained by the graphical method are the points on the line segment between
"a" and "b". We shall check the point "a" where R* = 20. , t* = 3.3 together with g 1 and g 4 active.
1. Check for necessary conditions.
Since only g 1 and g 4 are active, we can set u 2 = u 3 = u 5 = u 6 = u 7 = 0.
L = 0.1233Rt + u 1 [16459.5/(Rt) – 250] + u 4 (20 – R)
The KKT necessary conditions are:
∂L= ∂R 0.1233t − u1 16459.5 R 2t =− u4 0;
L ∂t 0.123R − u1 16459.5 =
∂= Rt 2  0;
gi ≤ 0, ui gi = 0; ui ≥ 0 ; i = 1 to 7
Substituting the optimum values, we get u 1 = 0.03263 > 0, u 4 = 0 ≥ 0 (o.k.). All the other conditions
are also satisfied. Thus, the point (R* = 20, t* = 3.3) satisfies the necessary conditions.
2. Check for sufficient condition. We shall consider the general case where only g 1 is active. The
Hessian of Lagrangian is
 u1 (16459.5 )( 2 ) u1 (16459.5 ) 
 0.1233 + 
 R3t R2t 2 
Ñ L=
2

 u1 (16459.5 ) u1 (16459.5 )( 2 ) 
0.1233 + 
 R2t 2 Rt 3 
 16459.5 R t 
2
c (1, − t R ) , c ≠ 0. Hence Q = dT Ñ 2 L d = 0. Thus,
T
Ñg1 =  2
; Ñg1T d = 0 gives d =
 −16359.5 Rt 
the sufficient condition is not satisfied.
3. The effect of variations in constraint limits on cost function.
Δf = − u1 e1 =0.03263 e1

Arora, Introduction to Optimum Design, 4e 5-81

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.74 _______________________________________________________________________________
Referring to the formulation in Exercise 3.48, we have
Minimize f = 1.57 × 10 −5 A(h2+5.625 ×10 5 )1/2,
1
subject to =
g1 ( 250 ) ( h 2
+ 5.625 × 10 ) (100 h + 0.2309 )
5 2
A − 250 ≤ 0;
1

( )
g 2 50000 h 2 + 5.625 × 10 5 2 (100 h + 0.2309 ) A − 250 ≤ 0 . The optimum solution obtained by the
=
graphical method is A* = 390, h* = 500, f * = 5.5 where g 1 and g 7 are active.
Check for necessary conditions. Since only g 1 and g 7 are active, we can set
u 2 = u 3 = u 4 = u 6 = u 8 = 0.
 
( ) ( )
1 1
L = 1.57 ×10 − 5 A h 2 + 5.625 × 10 5 2 + u1 ( 250 ) h 2 + 5.625 × 10 5 2 (100 h + 0.2309 ) A − 250 
 
+ u 7 (500 − h )
The KKT necessary conditions are:
 
( ) ( )
1 1
∂L ∂A =1.57 × 10−5 h 2 + 5.625 × 10−5 2 − u1 ( 250 ) h 2 + 5.625 × 105 2 (100 h + 0.2309 ) A2  =0
 

( )
−1
= 1.57 × 10 − 5 Ah h 2 + 5.625 × 10 5
∂L ∂h 2
+ u1 ( 250 )
 
( ) ( ) ( )
−1 1

 h h 2
+ 5 . 625 × 10 5 2
(100 h + 0 . 2309 ) A + h 2
+ 5 .625 × 10 5 2
−100 h 2 A − u7 =
0
 
g i ≤ 0, u i g i = 0, u i ≥ 0; i = 1 to 8
Substituting the optimum values, we obtain u 1 = 2.21610 × 10 −2 > 0, u 7 > 0, u 7 = 1.67 × 10 −3 (o.k.).
All other conditions are satisfied. Thus, the point (A* = 390, h* = 500) satisfies the necessary condition.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function
Δf = − u1e1 − u7 e7 = − 2.216 × 10−2 e1 − 1.67 × 10−3 e7

Arora, Introduction to Optimum Design, 4e 5-82

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.75________________________________________________________________________________
Referring to the formulation in Exercise 3.49, we have
Minimize f = 1.57 ×10−5 A ( 0.25s 2 + 106 ) ,
0.5

( 250 ) ( 0.25s 2 + 106 ) ( 0.1 + 346.4 s )

0.5
=
Subject to g 1 A − 250 ≤ 0;

(106 + 0.25s 2 ) ( 25 + 86602.5 s ) − ( 2.072617 × 107 ) A2 (10 + 0.25s 2 ) ≤ 0;

0.5
g2 = 6

( 50 ) ( 0.25s 2 + 106 ) ( 60.62 As ) − ( 50 ) ≤ 0;

1.5
=g3 2

( 50 ) ( 0.25s 2 + 10 ) ( 4.2 × 10 A ) − ( 50 ) ≤ 0; g 5 =
1.5
=
g4 6 8
100 − A ≤ 0; g 6 =−
A 5000 ≤ 0;
g 7= 500 − s ≤ 0; g8 = s − 4000 ≤ 0 . The optimum solution obtained by the graphical method is A*
= 415, s* = 1480, f * = 8.1 where g 1 is active.
1. Check for necessary conditions.
The only active constraint is g 1 , so we can set u 2 = u 3 = u 4 = u 5 = u 6 = u 7 = u 8 = 0.

( ) ( )
1
+ u1 ( 250 ) 0.25s 2 + 10 6 ( 0.1 + 346.4 s ) A − 250 
0.5
L=1.57 × 10 − 5 A 0.25s 2 + 10 6 2
 
The KKT necessary conditions are:
(
∂L ∂A= 1.57 × 10− 5 ( 0.25s 2 + 10 6 ) − u1 ( 250 ) 0.25s 2 + 106 ( 0.1 + 346.4 s ) A2  = )
0.5 0.5
0
 
( ) (
+ u1 ( 250 ) 0.25s 0.25s 2 + 106 )
− 0.5 − 0.5
s 1.57 × 10−5 A ( 0.25s ) 0.25s 2 + 106
∂L ∂=

( 0.1 + 346.4 s ) ( ) ( 346.4 s ) A =
0.5
A − 0.25s 2 + 106 2
0
g i ≤ 0, ui gi = 0, ui ≥ 0; i = 1 to 8
Substituting the optimum values, we obtain u 1 = 0.03252 > 0, u 1 = 0.03304 > 0 (o.k.). The
difference is acceptable since the optimum values are obtained from graphical method which has
only limited accuracy. For the following calculation, we assume u 1 = 0.0328. All the other
conditions are also satisfied. Thus, the point (A* = 415, s* = 1480) satisfies the necessary conditions.
2. Check for sufficient condition.
− 6  9.535 9.420   − 2.413 × 10− 3 
∇ L 10 = 9.420 5.097  ; ∇g1 ( 250 ) 
2
−4 
   − 2.346 × 10 
Ñg1T d = 0 gives d = c (1, − 10.29 ), c ≠ 0. Hence Q = dT ( Ñ 2 L=
) d 4.41× 10− 4 c2 > 0. Thus, the
sufficient condition is satisfied.
3. The effect of variations in constraint limits on cost function
Δf = − 0.0328 e1
− u1 e1 =

Arora, Introduction to Optimum Design, 4e 5-83

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.76________________________________________________________________________________
Referring to the formulation in Exercise 3.50, we have
( )
Minimize f = 7.85 × 10 − 3 2 2 A 1 + A 2 ,

subject=
to g1 30618.6 A + 17677.65 ( A +
1 1 )
2 A2 − 150 ≤= ( )
0; g 2 35355.3 A1 + 2 A2 − 150 ≤ 0;
g3 = 50 − A1 ≤ 0 ; g 4 =A1 − 5000 ≤ 0 ; g5 =
50 − A2 ≤ 0 ; g 6 =A2 − 5000 ≤ 0 .The optimum solution
=
obtained by the graphical method is A1* 300=
, A2* 50= , f * 7.0 where g1 and g 5 are active.
1. Check for necessary conditions. Since only g 5 are active, we can set u 2 = u 3 = u 4 = u 6 = 0.

( ) (  )
L = 7.85 × 10−3 2 2 A1 + A2 + u1 30618.6 A1 + 17677.65 A1 + 2 A2 − 150  + u5 ( 50 − A2 )
The KKT necessary conditions are:
( )
∂L ∂A1 = 7.85 × 10 −3 2 2 + u1 − 30618.6 A12 − 17677.65 A1 + 2 A2  = 0

2
(

)
∂L ∂A2= 7.85 × 10−3 + u1  −17677.65 2 ( )(
A1 + 2 A2  − u5= 0 )
2

 
gi ≤ 0, ui gi =0 , ui =0; i = 1 to 6
Substituting the optimum values, we obtain u1 = 0.0473 > 0, u5 =
0 ≥ 0 (o.k.) . All the other
=
conditions in (3) are also satisfied. Thus, the point =
A1* 300,( )
A2* 50. satisfies the necessary
conditions.
2. Check for sufficient condition
1.541 × 10− 4 9.322 × 10− 6 
Ñ L=
2
−6 
9.322 × 10 1.318 × 10− 5 

(
 − 204.1 A2 − 117.9 A + 2 A 2 
)  − 3.126 × 10−3 
(150 )  
(150 ) 
1 1 2
Ñg1 =  −3 
( )( )  −1.213 × 10 
2
 −117.9 2 A1 + 2 A2 
 
Ñg1T d =0 gives d =c (1, − 2.58 ) , c ≠ 0. Hence Q =
T
dT Ñ2 L d = ( )
1.937 × 10− 4 c 2 > 0,
if c ≠ 0. Thus, sufficient condition is satisfied.
3. The effect of variations in constraint limits on cost function
∆f = − u1 e1 =− 0.0473 e1 .

Arora, Introduction to Optimum Design, 4e 5-84

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.77________________________________________________________________________________
Referring to the formulation in Exercise 3.51, we have
Minimize f = 153.7R t,
4.045 × 106 5.257 × 10 ( 2 R + t ) 1.649 × 10 ( 2 R + t )
3 10

=
subject to g 1 + + − 1 ≤ 0;
R 3 t + Rt 3 4 R 3 t + Rt 3 4 ( )
2
R 3
t + Rt 3
4
g 2 = 70 − 2 R + t ≤ 0; g3 = 2 R t − 91 ≤ 0; g=
4 4.4794 × 107 R 3t + Rt 3 4 − 20 ≤ 0; ( )
g5 = R − 0.5t − 250 ≤ 0 ; g 6 = 35 − R + 0.5t ≤ 0 ; g 7 = t − 40 ≤ 0; g8 = 1 − t ≤ 0 .
The optimum solution obtained by the graphical method is R* = 130, t* =2.86, f * = 57000 where g 1
and g 3 are active.

1. Check for necessary conditions.

Since only g 1 and g 3 are active, we can set u 2 = u 4 = u 5 = u 6 = u 7 = u 8 = 0.
 
4.045 × 106 5.257 × 10 ( 2 R + t ) 1.649 × 10 ( 2 R + t )   2R 
3 10

L = 153.7R t + u 1  3 + + − 1 + u3  − 91
 R t + Rt 4 R t + Rt 4 ( R t + Rt 4 )   t 
3 3 3 2
3 3
 
The KKT necessary conditions are:
∂L
= 153.7t + u1 [
( )
− 4.045× 106 3R 2t + t 3 4 5.257× 10 3 ( 2 ) 5.257 × 103 ( 2 R + t ) 3R 2t + t 3 4
+ −
( )
∂R ( ) ( )
2 2
R 3t + Rt 3 4 R 3t + Rt 3 4 R 3t + Rt 3 4

+ −
2
(
1.649 × 1010 ( 2 ) 1.649 × 10 ( 2 R + t ) R t + Rt 4 3R t + t 4
10 3 3 3

] + u3 ( 2 t ) =
0
)( ) (1)
(R 3t + Rt 3 4 ) R 3t + Rt 3 4 (
2
)
∂L
= 153.7 R + u1[
−4.045 × 106 3R 2t + t 3 4
+
(
5.257 × 103
−
)
5.257 × 103 ( 2 R + t ) R 3t + 3Rt 2 4 ( )
∂t R 3t + Rt 3 4
2
( )
R 3t + Rt 3 4 ( )
R 3t + Rt 3 4 ( )
1.649 × 1010 1.649 × 1010 ( 2 R + t )( 2 ) ( R t + Rt 4 )( R
3 3 3
+ 3 Rt 2 4)
+ − ]+u ( −2R t ) = 0
2
(2)
( R t + Rt 4 ) ( R t + Rt 4 )
2 4 3
3 3 3 3

gi ≤ 0, ui gi = 0, ui ≥ 0; i = 1 to 8 (3)
Substituting the optimum values into (1) and (2), we get u1 = 2.817 × 10 > 0, u 3 = 294.07 > 0 4

(o.k.). All the other conditions are also satisfied. Therefore, the point R * = 130, t * = 2.86 satisfies ( )
the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function.
∆f = − u1e1 − u3e3 =− 2.817 × 104 e1 − 294.07e3 .

Arora, Introduction to Optimum Design, 4e 5-85

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.78________________________________________________________________________________
Referring to the formulation in Exercise 3.52, we have
1.42603 × 108 d o
( = )
Minimize f = 6.126 d o2 − di2 , subject to g1 − 1.65 × 104 ≤ 0;
d o − di
4 4
( )
=g2
(
40743.7 d o2 + d o di + di2 ) − 5000 ≤=
0; g
3.719 × 106
− 10 ≤ 0 ;
(d − d ) 4
o i
4 3
(d 4
o − di4 )
( d + d ) − 60 ≤ 0 ; (d − di ) − 2 ≤ 0;
g4 =
o i
g5 =
o
g 6 = 0.5 − d o + di ( ) 2 ≤ 0;
(d − d )
o i
2
g 7 =d o − 50 ≤ 0 ; g8 =5 − d o ≤ 0 ; g9 =di − 45 ≤ 0 ; g10 =4 − di ≤ 0.
The optimum solution obtained by the graphical method is d o* = 41.6, d *i = 40.2, f ∗ = 680 where g 3
and g 4 are active.
1. Check for necessary conditions.
Since only g 3 and g 4 are active, we can set u 1 = u 2 = u 5 = u 6 = u 7 = u 8 = u 9 = u 10 = 0.
 3.719 × 106   ( d + di ) 
L = 6.126 ( d o2 − di2 ) + u3  − 10  + u4  o − 60 
 ( d o − di )  ( d o − di )
 4 4
 

The KKT necessary conditions are:
 −3.719 × 106 4 d 3
∂L
= 12.252 d o + u3 
( ) o − 10 + u  ( do − di ) − ( do= + di ) 
 0
∂d o 
( d o − di )  ( )
2 4 2
4 4

 d − d 

  o i

 3.719 × 106 4 d 3 
∂L
= −12.252di + u3  ( ) i
− 10  + u  ( d o − di ) + ( d o + di )  = 0
 
∂di 
( 
) ( d o − di )
2 4 2

 d o4 − di4   
gi ≤ 0, ui gi = 0 , ui ≥ 0; i = 1 to 10
Substituting the optimum values, we get u 3 = 35.71 > 0, u 4 = 6.067 > 0 (o.k.). All other conditions
( )
are satisfied. Thus, the point d o* = 41.6, d i* = 40.2 satisfies the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations in constraint limits on cost function.
∆f = − u3e3 − u4e4 =− 35.71e3 − 6.067e4 .

Arora, Introduction to Optimum Design, 4e 5-86

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.79________________________________________________________________________________
Referring to the formulation in Exercise 3.53, we have
Minimize f = 5.027t ( d o − t ) ,
1.688 × 109 4.098 × 107 do 1.435 × 1017 do
=
subject to g1 + + −1 ≤ 0
t ( do − t )  do + ( do − 2t )  do − ( do − 2t )
 4 4 4 2
2
 d − ( d − 2t ) 
4 4
   o o 
2.466 × 1013  do4 − ( do − 2t )  − 100 ≤ 0 ;
4
g= do t − 9 ≤ 0; g=
2 3
 
g 4 = 250 − do ≤ 0; g5 = do − 1500 ≤ 0; g6 = 5 − t ≤ 0 ;
g7 =
t − 100 ≤ 0 ;
=
The optimum solution obtained by the graphical method =
is do*  1310, t ∗  14.2, f ∗ = 92500 where
g 2 and g 3 are active.

1. Check for necessary conditions.

Since only g 2 and g 3 are active, we can set u 1 = u 2 = u 4 = u 5 = u 6 = u 7 = 0.

 ( )
L = 5.027t ( do − t ) + u2 [ do t − 92] + u3  2.466 × 1013 do4 − ( do − 2t ) − 100 
4

The KKT necessary conditions are:

( )( )( )
2
∂L ∂d o = 5.027 + u2 t − u3  4 2.466 × 1013 d o3 − ( d o − 2t ) d o4 − ( d o − 2t ) = 0
3 4


( ) 
( ) ( )
∂L ∂t= 5.027 + ( d o − 2t ) + u2 −d o t 2 − u3 8 2.466 × 1013 ( d o − 2t ) d o4 − ( d o − 2t ) 2  = 0
3 4


gi ≤ 0, ui gi = 0, ui ≥ 0; i = 1 to 7
Substituting the optimum values, we get u 2 = 5.08 ×10 2 > 0 , u 3 = 4.624 ×10 2 > 0 (o.k.). All the other
conditions are also satisfied. Thus, the point satisfies the necessary conditions.
2. Check for sufficient condition. Use the same argument as in Exercise 5.53.
3. The effect of variations of constraint limits on cost function.
∆f = − u1e1 − u2e2 =− 5.082 × 102 e2 − 4.624 × 102 e3 .

Arora, Introduction to Optimum Design, 4e 5-87

 Chapter 5 More on Optimum Design Concepts: Optimality Conditions

5.80________________________________________________________________________________
The formulation is given in Exercise 3.54.
Optimum solution and check of necessary and sufficient conditions can be found in Exercise 5.48.

Referring to the formulation of Exercise 3.54, we have

( ) (
) , subject to
Minimize f = 6.59734 − 10 −3 D 2 H 2 + 4800
12

( 2.546475 × 10 )( H + 4800 ) ( D H ) − 1.5 × 10 ≤ 0;

1
=g1 4 2 2 2 4

( )( ) ( ) ( )
1
g 2 =2.0 × 104 H 2 + 4800 2
H − 1.816774 × 106 D 4 H 2 + 4800 ≤ 0;
g3 = H − 500 ≤ 0; g 4 = 50 − H ≤ 0; g5 = D − 50 ≤ 0; g 6 = 0.5 − D ≤ 0 .
=
The optimum solution obtained by the graphical method is H ∗ 50,
= D∗  3.42,
= f ∗  6.6 , where g 2
and g 4 are active.

1. Check for necessary conditions. Since only g 2 and g 4 are active, we can set u1 = u3 = u5 = u6 = 0
and s2 = s4 = 0 .

( 6.59734 × 10 ) D ( H )
1
−3
L= 2 2
+ 4800 2
+
 
( )( ) ( ) ( )
1
u2  2.0 × 104 H 2 + 4800 2 H − 1.816774 × 106 D 4 H 2 + 4800  + u4 ( 50 − H )
 
The KKT necessary conditions are

( ) ( ) ( )
1
=∂L ∂D 0.0131947 D H 2 + 4800  −7.267096 × 106 D3 = H 2 + 4800  0
2+ u (1)
2 


( ) ( ) ( )( )
1 1
∂L
= ∂H 6.59734 × 10−3 D 2 H H 2 + 4800 2 + u2  − 2.0 × 104 H 2 + 4800 2 H 2

2
( )( ) ( ) ( )
1
+ 2.0 × 104 H 2 + 4800 2 + 3.633548 × 106 D 4 H H 2 + 4800  − u4 = 0 (2)

gi + si2= 0, ui s=
i 0, ui ≥ 0 ; i = 1 to 6 (3)
−5 −2
Substituting the optimum value into (1) and (2), we get u2 = 9.68 × 10 > 0,=
u4 4.68 × 10 >0 (o.k.)

2. Check for sufficient condition. Since the number of active constraints is equal to the number of
( )
design variables, the point H ∗ = 50, D ∗ = 3.42 is indeed an isolated local minimum point. The
sufficient condition is deemed satisfied at this point.

9.68 × 10−5 , u4 =
The Lagrange multipliers of active constraints are u2 = 4.68 × 10−2.
Therefore, ∆ f =− u2 e2 − u4 e4 =− 9.68 × 10 −5 e2 − 4.68 × 10 −2 e4 .

Arora, Introduction to Optimum Design, 4e 5-88