RISK AND RELIABILITY OF

CIVIL INFRASTRUCTURE
SYSTEMS (CE60112)
ARITRA CHATTERJEE
DEPARTMENT OF CIVIL ENGINEERING
SPRING 2024
LECTURE 2 – RANDOM VARIABLES
RANDOM VARIABLE

 If we can map each sample point to a number line, each number on that line is a random
variable

 Mapping can be obvious, for example, what is the maximum temperature tomorrow?

 Mapping can also be based on an established rule. For example, different states of
damage to a building

 Just as we can assign probabilities to random events, the same probabilities can be
assigned to the random variables
DISCRETE AND CONTINUOUS RANDOM
VARIABLES
 Discrete random variables take a well-defined value

 Continuous random variable can lie in an interval, say between 5.2 mm and
5.3 mm

 Example of discrete RV : number of earthquakes next year, number of wind

events exceeding a threshold wind speed

 Example of continuous RV : average wind speed over next week, almost all
“measured” quantities
PROBABILITY MASS FUNCTION

pX(X) = Pr(X=x)

Since pX is a probability function, it must follow the basic rules of probabilities

Further sum(pX(X)) for all X must equal 1

This is because every realization X=x signifies a mutually exclusive event, and all
realizations of X are collectively exhaustive for the sample set

Examples – draw probability mass functions for tossing of a coin or rolling of a die
PROBABILITY DENSITY FUNCTION

 For a continuous random variable, probability that the random variable has a
specified value is 0.

 This is because the random variable can take infinite possible values,
therefore the probability of taking any one value is 0.

 In this case, probability density function fX(x) is defined as follows:

Pa x atda
f dn
a a
CUMULATIVE DISTRIBUTION FUNCTION

 CDF is defined as FX(x) = Pr (X <= x)

 For discrete random variables, this is given as:

discrete

 PMfof
af G p.cn
For continuous random variables, this is given as: FxCa
do
Ifx a
1
fx a
14
 PLOTS OF PMF, PDF AND CDF (FOR BOTH
DISCRETE AND CONTINUOUS RV)
PMF
70
P3
fx n

i
fx n i

t.lt
PLOTS OF PMF, PDF AND CDF (FOR BOTH
DISCRETE AND CONTINUOUS RV)

ix
CONDITIONING AN EVENT ON A RANDOM
VARIABLE
 Pr(E|X=x) for discrete random variable and continuous random variable

pacelx nt -

Pap EE

DISCRETE
Prp
m
conen
¾
PDF
CONDITIONING AN EVENT ON A RANDOM
VARIABLE
 Pr(E|X=x) for discrete random variable and continuous random variable

PACE X
Ʃ
R
Pa E Px a

PMF

Pa E Pa da
ElX n
fx a

PDF
CONDITIONING A RANDOM VARIABLE ON
AN EVENT
 PMF Pr(X=x|E) for discrete random variable

PNEIN Pa Cx E

 PDF fX|E(x) for continuous random variable

PapfI RfIf g
Fx as 90 adf.gr
 CDF FX|E(x) for continuous random variable

x
FxIEN 4AEI E
RELIABILITY AND HAZARD FUNCTIONS

 Consider a system whose time to failure is given by T

 Reliability function at time t, that is, R(t) is defined as Pr(t<T) = 1 – FT(t). Let us plot this…

 Given that no failures were observed upto time t0, what is the conditional CDF of T, and the
conditional reliability function? The conditioning event E is {t0 < T}

F it.at
PI ff
ttiiat
t to
t.at
EI.it
RELIABILITY AND HAZARD FUNCTIONS

 Consider a system whose time to failure is given by T

 Reliability function at time t, that is, R(t) is defined as Pr(t<T) = 1 – FT(t). Let us plot this…

 Given that no failures were observed upto time t0, what is the conditional CDF of T, and the
conditional reliability function? The conditioning event E is {t0 < T}

t t
R
PLOT OF CONDITIONAL RELIABILITY
FUNCTION VS TIME
MEAN TIME TO FAILURE


MTTF =  tf (t )dt
0
HAZARD FUNCTION

 Conditional pdf of failure function, given survival up to time T

 h(t) =fT(t)/(1-FT(t))
EEE
3
dt
 h(t) = fT(t)/RT(t)
fT(t)=-dR(t)/dt R t fact
hut at


F t 1 exp
t hL f The
Hazard function follows a bathtub curve

exp
f
 Given a hazard function, how can we compute the pdf of the failure time
function?
MULTIPLE RANDOM VARIABLE

 In many situation, we need to consider many random variables

 In such cases, we need to define not only the individual variables, but also
their dependencies

 For discrete random variables, joint PMF is defined as:

nxz.dz A Xn an

Px a Pa X a
 PROPERTIES OF JOINT RANDOM
VARIABLES (DISCRETE)
I
0 px a

Un TD
nah an Px xn in i II
Eo
an
Px
L 3
In
__ Px R

 Second condition (consistency rule) gives the marginal distribution by

summing over all possible values of one variable
PROPERTIES OF JOINT RANDOM
VARIABLES (CONTINUOUS)
 Joint PDF definition and properties

Pm a Eaton
On
f a
In Xn ant oan
I
0
fx a

fxn.in can and an fx xn.fm

If a dx 1
PROPERTIES OF JOINT RANDOM
VARIABLES (CONTINUOUS)
 Joint CDF definition and properties a Pa X n n Xz n
Xn an

0 Fx a

An 1 D Fx
Fxi xz Xn
x 22

0 D 1
Fx X Xn
RELATIONSHIP BETWEEN JOINT PDF AND
JOINT CDF

Fx Ca
fxlasd s.int
CONDITIONAL PMF OF TWO DISCRETE
RANDOM VARIABLES

Paix la la 8
524 1 kind
o

0 Px Ca 2 0
RELATIONSHIP BETWEEN CONDITIONAL
PMF AND JOINT PMF
a 122 Px 62
Pxix N1 N2 Pxyx

For STAT IND R Vs

Pxixz 21,22 Px Ca Pxz x2

EXTENSION TO ‘n’ DISCRETE RANDOM
VARIABLES
2n
x dk dkti
Xn
PX Xxl kt

Un
0 Px xn Nkt
STATISTICAL INDEPENDENCE OF‘n’
DISCRETE RANDOM VARIABLES

2n
Px xn
a
an
Px Ca Pxn
In
for all values of x1
CONDITION PDF OF TWO CONTINUOUS
RANDOM VARIABLES

fx.me talk
t.y.fi ian
FO
O IF fx
RELATIONSHIP BETWEEN CONDITIONAL
PDF AND JOINT PDF
22 Fx me 2 fake
Fxix 21 z

fix 62124 fx Ca

are
statistically ind
IF I Xz

fxixzmnd fx.cm fear

for all 1 21 X2 22
EXTENSION TO ‘n’ CONTINUOUS RANDOM
VARIABLES
I Nkt1 An
a
k kt n

n
in

0 IF DEN 0
STATISTICAL INDEPENDENCE OF ‘n’
CONTINUOUS RANDOM VARIABLES

Rn
x
xn

fx.la fx Ca fxn an

ALL 1 21 X2 22
For
Xn Rn
EXPECTATIONS AND MOMENTS OF SINGLE
RANDOM VARIABLE
 Expectation of a function g(X) of random variable X, E[g(X}] is defined as:

96 ffa
do Ex n
is a linear
 If g(X)=Xm, E[g(X}] is known as the mth moment of X E Ʃ9 X
E Ig xD
 If g(X)=(X-m)m, where m = E[X] is the first moment of X, or the expectation of
X (also know as mean), g(X) is known as the mth central moment of X
MOMENTS OF A RANDOM VARIABLE
E doe U MEAN
a
f a

E x2 da MEAN SQUARE

α
22 f n

centroid n't
EE
w
represents
0 axis

ECF represents moment

9
MOMENTS OF A RANDOM VARIABLE

MEDIAN Roy FI 0.5

INVERSE CDF

Outcome at which
MODE
is max
Pdf
Ñ any maxfx n
MOMENTS OF A RANDOM VARIABLE
CENTRAL MOMENTS

E x mm
de
f a a
fin
D
moment
is mᵗʰ central of
MOMENTS OF A RANDOM VARIABLE
O PROVE
MOMENT
1ˢᵗ CENTRAL

MOMENT
VARIANCE 02
2ⁿᵈ CENTRAL

X M
E STANDARD DEVIATION
VARIANCE T
to radius
is analogous
T
NOTE
of gyration
C a V coeff variations 181
MOMENTS OF A RANDOM VARIABLE
VARIANCE
M
E X
M M
E x2 2

X m
E x 2 ME

x2 2 E ECD
E

E x2 EIX
 MULTIPLE
MOMENTS OF A RANDOM VARIABLES

OF PRODUCT
E XIX MEAN
M COVARIANCE
E Xi Mi X
Xi
OF Xj
COV Exi Xj

IT CAN BE SHOWN THAT

Cov x X Efx X EEXITELI

PROVE THIS
MOMENTS OF A RANDOM VARIABLE

Correlation
Pi
CYYYXI.it COEFFICIENT

Cov Xi X Pij Tir

Pi I
1
measure of
Tij is a

linear dependence
MOMENTS OF A RANDOM VARIABLE

Xi
7.0
Pu 7.0 f
LINEAR
FEES
a EFFIGIES
DEPENDENT
BE
 n.fi Exx

ARE

WHY
COVARIANCE
MATRIX Varian
covern

DX Q
DIAGONAL
MATRIX OF
e

SENSE S 0 g

Rxx

MATRIX f

en en
 E CX Mx Cx Mx
Ex
XXT EEX E XD
E

Dx Rxx DX
PROVE
THIS

Rxx
NOTE Exx and
aresymmetric
the
positive
definite if are
not
random variables
related
linearly
singular
Exx Rxx are
the variables
if any of
are
linearly related
MATRIX
M
DEFINITE
POSIT WE
all x IR
Mx 70 for an
where M is