You are on page 1of 4

Optimization in Engineering Homework 1 Solution

Problem 1.
(a) f = exp[x − 1(1 − x2)]1/2
2
f is only defined for 0 < x ≤ 1 & x ≤ −1 because 1−xx ≥ 0
df ,
1−x2 1 + x2
= −e x .
dx 2x 1−x2
x
, . 2
1−x2
1−x2
Because e x
, , x , 1 + x > 0,
2 2 df
< 0 for 0 < x ≤ 1 & x ≤ −1
x dx
⇒ f (x−) for 0 < x ≤ 1 & x ≤ −1

(b) f = sin(x2 + ln y−1)


Since the sin function is periodic, f is not monotonic for x > 0 and y > 0.
2 −1
(c) f = exp(x)/ exp(1/x) = exp( x x
) defined for all x except for x = 0

df 1 + x2 x2−1
= e x
dx x2

dx > 0 for all x except for x = 0


f (x+) for all x except for x = 0
∫ b
(d) f = a exp(−xt)dt define for all x

i) if t > 0, i.e. b > a > 0, then for any x2 > x1, we have e−x2t < e−x1t. Then f (x−)

ii) if t < 0, i.e. a < b < 0, then for any x2 > x1, we have e−x2t > e−x1t. Then f (x+)

iii) if a < 0 < b, then monotonicity is unknown

1
Optimization in Engineering Homework 1 Solution

Problem 2.
The problem can be rewritten as

minimize − f (x− 1 , x2 ) = −x1 + x2


+

s.t. g1(x1+, x2+) = 2x1 + 3x2 − 10 ≤ 0


1 , x2 ) = −5x1 − 2x2 + 2 ≤ 0
g2 (x− −

1 , x 2 ) = −2x1 + 7x2 − 8 ≤ 0
g3 (x− +

By MP1, with respect to x1, g1 is active. By MP1, with respect to x2, g2 is active.
Then, solving g1 = 0 and g2 = 0, we get
14 46
x1 = − , x2 =
11 11
But this is a infeasible point because it violates g3. Note that the Monotonicity
Principles are applicable only for a well-constrained minimization problem(See p.100 in
the text). We solve this as follows For x1, g1 is active, this eliminate x1 by g1 = 0.

minimize − f (x2+) = 2.5x2 − 5


s.t. g2(x2+) = 5.5x2 − 23 ≤ 0
g3(x2+) = 10x2 − 18 ≤ 0

Since we don’t have a lower bound on x2, the problem is unbounded.

2
Optimization in Engineering Homework 1 Solution

Problem 3.

minimize f (x+, x−)


3 4
s.t. g1(x + 3, x − 4) ≤ 4
g2(x3+, x4−) ≤ 3
h1(x1, x2, x3, x4) = 2
g3(x+2, x−3 ) ≤ 2
h2(x2, x3) = 6
g4(x3+, x4+) ≤ 7
g5(x3+, x4+) ≤ 5

i) Using MP2 w.r.t. x1, h1 must be inactive and can be discarded from the problem,
and x1 is irrelevant x2, x3 and x4 are relevant.
ii) Using MP1 w.r.t. x4, we obtain g4, g5 is a conditionally critical set.
By MP2 w.r.t. x2, we can direct h2 = 6.
Then, by MP1 w.r.t. x3, g3, g2 is a conditionally critical set, too.
iii) By MP1 w.r.t. x3 and x4, g1 and g2 are uncritical.
iv) g4 is dominated by g1, g2 and g5
v) Eliminating x3 by h2 = 6, we obtain

minf (x− 2 ; x4 ) = 12 − 2x2 − x4


s.t.g1 (x−2 , x4 ) = 2 − x2 − 3x4 ≤ 0


g2 (x−2 , x4 ) = 15 − 3x2 − 2x4 ≤ 0


g3(x2+) = 2x2 − 8 ≤ 0
g5 (x−2 , x 4 ) = 1 − x2 + 3x4 ≤ 0
+

By MP1 w.r.t. x2 we find g3 critical.


By MP1 w.r.t. x4 we find g5 critical.
vi) Two variables and two active constraints means that the problem is constraint bound.

vii) No multiply critical constraints.

viii) Solving g3 = 0 and g1 = 0, we obtain x∗2 = 4, x∗4 = 1. But this point violate
g2 = 15 − 12 − 2 = 1 > 0. Therefore, this problem has no feasible solution.

Note that Monotonicity Analysis is based on assuming the problem is well-bounded or it


is a necessary condition for model boundness but not sufficient. So, everytime solving a
problem by Monotonicity Analysis, we always have to check feasibility of solution

3
Optimization in Engineering Homework 1 Solution

Problem 4.
(a) By MP1 w.r.t l, g1 is critical constraint. Replace l by g1 = 0, then construct table
agian. Loop the process, we got the result that g1, g6, g4, g2 is critical to l, h, t, b.

Table 1: Monotonicity table

h l b t
f + + + +
g1 − −
g2 − −
g3 + −
g4 − −
g5 − −
g6 −

(b) The given minimizing design violate constraint.

You might also like