Final Tutorial Text and Solutions PDF
Final Tutorial Text and Solutions PDF
Final Tutorial Text and Solutions PDF
Problem 1
A structure is designed with a capacity to withstand a wind velocity of 150 kph; i.e. any wind
velocity up to 150 kph will not cause damage to the structure. In a hurricane-prone area, the
maximum wind velocity during a hurricane may be modeled as an extreme Type I asymptotic
distribution with a mean velocity of 100 kph and a c.o.v. of 0.45.
(a) For a structure with the above wind-resistant capacity, what would be the probability of
wind damage during a hurricane?
(b) If the permissible damage probability were to be reduced to 1/10 of the original design, as
described in (a) above, what should be the wind-resistant capacity of the revised design
(in kph)?
(c) If the occurrence of hurricanes in the area is modeled as a Poisson process with a return
period of 200 years, what is the probability of damage to the original structure over a life
of 100 years? Also, what would be the corresponding damage probability to the revised
structure? Assume that damages between hurricanes are statistically independent.
Reminder
Asymptotic dustribution CDF FY (y) and pdf fY (y)
Distribution parameters:
π 1.282 γ
α= √ = , u = µY −
6σY σY α
where
γ ≈ 0.5772 is the Euler constant.
Problem 1 - Solution
V = wind velocity (extreme Type I asymptotic)
µV = 100 kph
c.o.v.V = 0.45 ⇒ σV = 45 kph
Distribution parameters
π
α= √ = 0.0285
6σV
γ 0.5772
u = µV − = 100 − = 79.75
α 0.0285
(a) P (V > 150) = 1 − [exp {− exp(−0.0285(150 − 79.75))}] = 0.126
x = 233 kph
P (Damage|Hurricanes) = 0.126
• N = roughness coefficient
The constant in the formula includes appropriate unit conversions. You are analyzing an
existing sewer drain, and the input values are random variables due to uncertainty in the
as-built condition relative to what you can measure. Assume the random variables are all
independent with µN = 0.015, δN = 0.15, µD = 3 ft, δD =0.02, µS = 0.005 ft/ft and δS =0.05
(a) Compute the first-order second-moment (FOSM) estimates of the mean and standard
deviation of Q
(b) What is the percentage contribution of each input random variable to the variance of Q?
(c) If you considered only the most important input random variable identified in (b), and
assumed the other two random variables to have coefficients of variation equal to zero,
what standard deviation would you estimate for Q?
Problem 2 - Solution
µN = 0.015, δN = 0.15, σN = δN · µN = 0.00225
µD = 3.0, δD = 0.02, σD = δD · µD = 0.06
µS = 0.005, δS = 0.05, σS = δS · µS = 0.00025
Let √
0.463D2.67 S
Q= = g(N, D, S) = g(X1 , X2 , X3 )
N
µQ ≈ g(µX1 , µX2 , µX3 )
3 X 3
2
X ∂g ∂g
σQ ≈ ρXi Xj σXi σXj
i=1 j=1
∂x i ∂x j
√ 2 2.67 √ 2 √ 2
0.463 · 2.67µ1.67 0.463 · µD 0.463 · µ2.67
2 D µS 2 µS 2 D µS 2
σQ ≈ σD + √ σS + − 2
σN
µN µN · 2 µS µN
2
σQ = 43.69, σQ = 6.61
Using tabulated values of the standard normal distribution, determine the numerical values of
Pf in the following two cases
Pf = P (M < 0)
0 − 0.5 m
= Φ √ = Φ(−2.24) = 1 − Φ(0.224) ≈ 0.013
0.05 m2
0 0 − 0.693
Pf = P (M < 0) = Φ √ = Φ(−2.47) = 1 − Φ(2.47) ≈ 0.007
0.0784
The two models give very different values for Pf . The reason for selecting one over the other
should be very well motivated or, at least, the results from both models should be presented.
Problem 4
The profit from your company’s last 10 projects (in units of thousands of dollars) were 25.1,
93.9, 90.3, 52.0, 12.1, 45.5, -7.4 104.2, 80.9, 65.5, where negative numbers indicate a loss. As-
sume that profit form an individual project is normally distributed. Use this data to estimate
the profit level for future projects that will be exceeded with 95% probability (or equivalently,
the profit level that you have a 5% probability of not exceeding).
Problem 4 - Solution
P
xi
X = i = 56.21
n
s
P 2
i (xi − X)
SX = = 37.5
n−1
• Using t-distribution with n − 1 = 10 − 1 = 9 dofs:
Find X such that FX (x) = 0.05
x − 56.21
CDFt = 0.05
37.5
Φ(−1.85) = 0.05, so
x − 56.21
= −1.64 ⇒ x = −5.29 thousand $
37.5
Problem 5
The distribution of wave height has been suggested to follow a Rayleigh density function
h 1 2
fH (h) = 2 exp − (h/α ) , h ≥ 0
α 2
with parameter α. Suppose the following measurements on wave heights were recorded
1.5, 2.8, 2.5, 3.2, 1.9, 4.1, 3.6, 2.6, 2.9, 2.3 m
dL
We maximize L by imposing =0⇒
dα
" q h #
dL − q i
= p e 2α2 −2nα−2n−1 + α−2n − (−8α−3 )
dα 2
" q #
−
= p e 2α2 α−2n−1 [−2n + qα−2 ] = 0
⇓
Pn
q (h2i ) 80.4
qα−2 = 2n ⇒ α2 = = i=1
= = 4.02 ⇒ α = 2.005
2n 2n 20
Problem 6
Consider the following 20-year record of the annual maximum wind velocity V (in km/h) in a
given city:
Year V (km/h) Year V (km/h)
1950 78.2 1960 78.4
1951 75.8 1961 76.4
1952 81.8 1962 72.9
1953 85.2 1963 76.0
1954 75.9 1964 79.3
1955 78.2 1965 77.4
1956 72.3 1966 77.1
1957 69.3 1967 80.8
1958 76.1 1968 70.6
1959 74.8 1969 73.5
(b) Determine the 99% confidence interval for the mean velocity. Assume that the true
standard deviation of V , σV , is satisfactorily given by the sample standard deviation sV .
(c) Assume that V has a lognormal distribution; determine the point estimates for the cor-
responding parameters λV and ξV .
Problem 6 - Solution
(a)
n
1 X
Sample Mean : V = Vi = 76.5
N i=1
n
!
1 X 2
Sample Variance : s2V = V 2 − nV = 14.27 ⇒ sV = 3.78
n − 1 i=1 i
Assume sV ≈ σV :
V −µ
−2.58 ≤ √ ≤ 2.58 ⇒ 74.3 ≤ µ ≤ 78.7
sV / n
(c)
s2V
ξV2 = = ln 1 + 2 = 0.0024
ξV = 0.049
V
⇒
1 λV = 4.34
λV = ln V − ξV2
2
Problem 7
Suppose that a sample of 9 steel reinforcing bars were tested for yield strength, and the sample
mean was found to be 20 kips.
(a) What is the 90% confidence interval for the population mean, if the standard deviation
is assumed to be equal to 3 kips?
(b) How many additional bars must be tested to increase the confidence of the same interval
to 95%?
(c) If the standard deviation is not known, but the 9 measurements yielded
9
X
(xi − 20)2 = 84.5,
i=1
then what would be the 90% confidence interval for the mean yield strength? Assume
that the yield strength is a normal variate.
Reminder:
1 kip = 1000 pounds-force = 4.448 kN
Problem 7 - Solution
x = 20 kips, σ = 3 kips. 90% confidence, n = 9. Assume normal variates.
(a)
x−µ
Φ(z) = 0.95 ⇒ z = 1.64 ⇒ −1.64 ≤ √ ≤ 1.64 ⇒ 18.36 ≤ µ ≤ 21.64
σ/ n
(b)
Φ(z) = 0.975 ⇒ z = 1.96
x − 18.36
√ = 1.96 ⇒ m = 3.85
σ/ 9 + m
Need 4 more samples for a 95% confidence.
(c) If the standard deviation is unknown, then we turn to the t-test, bacause n < 10 (so we
cannot assume s ≈ σ):
9
1 X 84.5
2
s = (xi − x)2 = = 10.56
n − 1 i=1 8
x−µ
P −tα/2,n−1 ≤ √ ≤ tα/2,n−1 = 1 − α
s/ n
⇓
α = 0.1; α/2 = 0.05 for a 90% confidence
20 − µ
t0.05;8 = 1.86 ⇒ −1.86 ≤ r ≤ 1.86 ⇒ µ90% = (17.985; 22.015)
10.56
9
Problem 8
The following five repeated observations (measurements) were made on each of the outer and
inner radii of a circular ring:
(a) Determine the best estimates of the outer and inner radii, and corresponding standard
errors;
(b) The area between the two concentric circles is computed based on the mean values of the
measured outer and inner radii; namely
A = π r21 − r22 .
(c) Determine the standard deviation (standard error) of the computed area.
(d) If it is desired to compute the sample mean of r1 within ±0.07 cm with 99% confidence,
how many additional independent measurements must be made on r1 ? Assume that all
the measurements are independent and taken with the same care and skill.
Problem 8 - Solution
(a) !
X r1 1 X
r1 = k
= 2.5, s21 = r12k − nr21 = 0.01
k
n n−1 k
⇓
s1
Standard error σr1 ≈ √ = 0.045 cm
n
!
X r2 1 X
r2 = k
= 1.5, s22 = r22k − nr22 = 0.01
k
n n−1 k
⇓
s2
Standard error σr2 ≈ √ = 0.045 cm
n
(b)
A = π r21 − r22 = π 2.52 − 1.52 = 12.566 cm2
(c)
A = π r12 − r22 ⇒ σA2 = π 2 var(r12 ) − var(r22 ) ,
∂f
and therefore for f (x) = x2 and = 2x, one gets
∂x
σr22 = (2r1 )2 σr21 = 4(2.5)2 (0.045)2 = 0.051
1
⇓
σA2 = π 2 (0.051 + 0.0182) = 0.679 ⇒ σA = 0.82 cm
(d) For r1 − µ1 = ±0.07 cm with 99% confidence, for n < 10 and unknown σ, use t-test:
r1 − µ 0.07 √
√ = t0.995;n−1 ⇒ = n · t0.995;n−1
s/ n 0.1
Upon equating these expressions with their sample counterparts, the following system of
two equations in the two unknowns (µ, σ) are obtained:
10
X
xi
2
σ
= i=1
exp µ + = 385
2 10
(1)
10
X
x2i
exp (2µ + 2σ 2 ) = i=1
= 323.080
10
⇓
exp (2µ + 2σ 2 ) 323.080
2 ≡ exp(σ 2 ) = 2
= 2, 1797 ⇒ σ 2 = ln(2, 1797) = 0.779, σ = 0, 8827
2
(exp(µ + σ /2)) 385
By inverting Equation (1)1 one then gets
σ2
µ = ln 385 − = 5, 564.
2
⇓
µ = 5, 564 σ = 0, 8827
= 1 − Φ(0.484)
= 1 − 0.685 = 0.315.
Problem 10
The settlement of a column footing (See Figure 1) is composed of two components - the settle-
ments of the sand and clay strata. The flexibilities (that is, inches settlement per foot of strata
per ton of applied load) of the two strata, denoted FS and FC , are independent normal variates
N (0.001, 0.0002) and N (0.008, 0.002), respectively. The total column load is W , which may be
assumed to be statistically independent of FS and FC .
(a) If W = 20 tons, what is the probability that the total settlement will exceed 3 in.?
(b) Suppose the load W is also a random variable with Probability Mass Function (PMF)
given as
0.6 W = 20
fW (w) =
0.4 W = 25
Determine the mean and variance of the total settlement by first-order approximation. In
this case, what would be the probability that the settlement will exceed 3 in. assuming
that settlement follows a normal distribution?
(a)
u = W f ⇒ u ∼ N (W µf , W σf ) = N (2, 0.4079)
⇓
3 − µu
P (u > 3) = 1 − P (u < 3) = 1 − Φ
σu
3−2
= 1−Φ = 1 − Φ(2.4515)
0.4079
= 1 − 0.993 = 0.007
2 2
∂u ∂u
σu2 ≈ 2
σW + σ 2 = µ2f σW
2
+ µ2W σf2 = 0.2613
∂W mean
∂f mean f
⇓
σu = 0.511
3 − µu
P (u > 3) = 1 − P (u < 3) = 1 − Φ
σu
3 − 2.2
= 1−Φ = 1 − Φ(1.565)
0.511
= 1 − 0.9405 = 0.0595