Professional Documents
Culture Documents
Lecture 39
Lecture 39
Lecture-39
Dr C S Manohar
Department of Civil Engineering
Professor of Structural Engineering
Indian Institute of Science
Bangalore 560 012 India
manohar@civil.iisc.ernet.in
1
Discussion on properties of processes with
Independent increments
2
Problem 35
Let X (t ) be a process with stationary independent increments;
assume t 0 & X 0 0. Show that
X t t;
Var X t 2t
Var X t X s 2 t s
Cov X t X s 2 min t , s
Note: clearly, X t 1 & 2 variance of X (t 1).
3
Let f (t ) X t
f (t ) X t X 0
f t s X t s X 0
X t s X s X s X 0
X t s X s X s X 0
X t X 0 X s X 0 stationary increments
f (t ) f ( s )
We get the functional equation f (t s ) f (t ) f ( s )
f t ct is the solution.
Given f 1 X 1 c
X t t
4
Let g (t ) Var X t
g (t ) Var X t Var X t X 0
g t s Var X t s X 0
Var X t s X s X s X 0
Var X t s X s Var X s X 0
Var X t X 0 Var X s X 0
stationary increments
g t s g t g s
g t ct
g 1 Var X 1 2 c 2
Var X t 2t
5
Let t s
Var X t Var X t X 0
Var X t X s X s X 0
Var X t X s Var X s X 0
=Var X t X s Var X s
Var X t X s Var X t Var X s
2 t s
6
Var X t X s
X t X s X t X s
2
2
X t X t X s X s
X t X t X s X s
2 2
2 X t X t X s X s
7
COV X t , X s
1
2
-Var X t X s +Var X t +Var X s
2
t s (t s ) 2
s assuming that t s
2
COV X t , X s 2 min t , s
8
Problem 34
Let X (t ) be a stationary Gaussian random process with
zero mean and PSD function of the form
X2 2
S XX exp - 2 ;-
2 2
•Determine the autocorrelation and cross correlation
functions of the processes X t and X t
Find the average rate of upcrossing of level
Find the PDF of time for first crossing of level
Find the average rate of peaks above level
Find the expected fractional occupation time above
level over a duration 0 to T
Find the PDF of extreme of X (t ) over duration 0 to T
9
Spectral moments
X2 2
S XX exp - 2 ;-
2 2
X2 2
0 exp - 2 d X
2
2 2
2
2
2 exp - 2 d X2 2
2 X
2 2
2
2
4 exp - 2 d 3 X2 4
4 X
2 2
10
Autocorrelation function
2
2
S XX exp - 2 ;-
X
2 2
1 X2 2
RXX exp - 2 exp i d
2 2 2
2 2
= X exp
2
2
d m n
RXX
t X t 1
m n n
Recall: X
d m n
11
dRXX d 2 RXX
X t X t ; X t X t
d d 2
dRXX d 2 RXX
X t X t
; X t X t ;
d d 2
XX
RXX
3 2
d R d
X t X t
; X t X t
d 3
d 2
XX
RXX
3 4
d R d
X t X t
; X t X t
d 3
d 4
12
2 2
RXX X exp
2
2
2 2
d
d
RXX X exp
2
2
2
d2
2 2
XX
2
R 2
exp 2
d 2 X
2
2 2
X exp
2
2
2
2 2
X exp
2 2
1
2 2
2
13
d2
2 2
d 2
R XX 2
X 2
exp 1
2 2
2
d3
2 2
d 3
R XX
2
X 2
exp 1
2 2
2
2
2 2
X exp
2 2
2 2
2
2 2
X exp
2 2
3 2
4 3
2
14
d3
2 2
d 3
R XX 2
X 2
exp 3 2
4
4 3
2
d4
2 2
d 4
RXX X exp
2 2
3 4
2 4 3 2
2
2 2
X exp
2 2
3 2
12
4 2
2
2 2
X exp
2 2
3
4 2
4
6 4
3 2
12
4 2
2
2 2
X exp
2 2
3 2
15
4 2
4
6 4
2
15
2 2
RXX X exp
2
2
2 2
d
d
RXX X exp
2
2
2
d2
2 2
d 2
RXX X exp
2 2
1
2 2
2
d3
2 2
d 3
RXX X exp
2 2
3 4
2 4 3
2
d4
2 2
d 4
R XX 2
X 2
exp 3 2
15
4 2
4
6 4
2
16
4
process
derivative
3
1
acf
-1
-2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
tau
17
Level crossing statistics
T
N , 0, T X t X t dt
0
18
PDF of time for first crossing of level
For high levels of crossings we can approximate
the number of times the level is crossed as a Poisson
random variable.
T
k
1
1 22 32 x 2 23
m ,t [ S exp
2 x
3
2 1 2
2
2 2 2 S 1 2
x 2 2x
exp 2 1 erf 3 ]dx
2 1 1 2 S
20
X 2 t X t X t X t X t
S X t X t X 2 t X t X t
X t X t X t X t X 2 t
1 0 2
X2 0 2 0 S 3 X6 6 1 2
2 0 3 4
12 X2
22 2 X2
32 3 X2 4
21
Expected fractional occupation time above level
over a duration 0 to T
T
1
y , T U X t dt
T 0
U X t 1 p x; t dx
X
1 x2
1 exp 2 dx
2 x 2 x
1
erf y , T
2 x
22
PDF of extreme of X (t ) over duration 0 to T
PT f t 1 exp t
pT f t exp t 0 t
1 a2
exp 2
;a
2 2 X
PX m 1 P T f T
23
FACTORS OF SAFETY
&
PROBABILITY OF FAILURE
24
Problem 35
In traditional engineering practice, uncertainties in
specifying loads and structural resistance are accounted
for by overestimating the loads and underestimating the
strucutral resistance. The factors by which these estimates
are obtained are calibrated against past experience with
existing stock of strucutres. It is of interest to relate this broad
principle with the probabilistic modeling of uncertainties.
To illustrate this let us consider an idealized situation in which
demands on the structure and supply of structural capacity
are modeled as a pair of mutually independent Gaussian
random variables.
25
The failure event is defined by exceedance of load effect over the
available capacity. If the tolerable level of probability of failure
is specified to be PF , determine the factors by which the expected
load and capacity are to be mutiplied.
26
27
R N ( R , R ); S N ( S , S ); R S
R S
R ; S
R S
Z R S Z N ( R S , R2 S2 )
PF 1 R S
R2 S2
R S
1 PF
R2 S2
R S R2 S2 1[1 PF ]
R S R2 S2
If is large, risk is small 28
2 2
R
; 0.75
S
R S
R S R S
R2 S2 R S
R S R S
R R S S
R R R S S S
R [1 R ] S [1 S ]
R [1 S ]
Central Safety factor
S [1 R ]
29
Capacity
reduction 1 R
factor
1 S
Load factor
R S
1 1 1 PF R R 1 1 1 PF S S
30
Nominal factor of safety
RN R 1 K R R 1 S 1 K R R
S N S 1 K S S 1 R 1 K S S
RN S N
1 R 1 S
;
1 K R R 1 K S S
31
Case of multiple loads
S N (S , S )
33
R S 2
R
2
S
2 2
Let R S
R S
R S R S
S R 2
S1
2
S2 2
Sn
34
S21 S22 S2n
Let nn
S1 S2 Sn
R S R nn S1 S2 Sn
R S1 S2 Sn
R nn S1 S2 Sn
(1 R ) R S1 1 nn S1 1
S2
nn S2
Sn 1 nn Sn
(1 R ) & i 1 nn Si
36
Problem 36
Figure (next slide) shows the pseudo-acceleration spectra
for a rocky site according to the IS 1893 (Part 1) : 2002
document. The PGA is taken to be 0.24g. It is of interest
to develop a random process model for the ground
acceleration that is compatible with this response spectrum.
It may be assumed that the ground acceleration can be
modeled as a zero mean, stationary Gaussian random
process. The duration of the acceleration can be taken
to be 30s and the given response spectra may be
interpreted as locus of the 84% percentile point and
damping may be taken to be 5%.
37
6
4
Acceleration m/s 2
1
Pseudo-acceleration spectra as per
IS 1893-2002
0
0 100 200 300 400 500 600 700
Frequency rad/s
38
How to generate a response spectrum compatible with a given PSD?
x 2 nn x n2 x
xg
xg t zero mean, stationary, Gaussian random process;
xg t ~ N 0, S gg
X m max x t
0t T
PX m exp T
2
x
exp 2
2 x 2 x
H S gg d &
2
with 2
x
H 2 S gg d
2
2
x
39
For a given probability p, the corresponding is given by
x 2
p exp exp 2 T
2 x 2 x
1
2 x 2
2 x ln
2
ln p
xT
Let R n , n be the given pseudo-acceleration response spectrum.
We interpret R n ,n as the p - th percentile point.
1
2 x 2
R n , n 2 x ln
2 2
ln p (typically p =84%)
xT
n
40
How to generate a PSD compatible with a given
response spectrum?
x 2nn x n2 x
xg
xg t zero mean, stationary, Gaussian random process;
xg t ~ N 0, S gg
To a first approximation we assume
H S gg d
2
2
x
1
2nn H n S gg n 2nn S gg n
2
2
2 nn2
x
& n
2 x
41
S gg n 2
R n ,n 2
2 4
ln ln p
2nn nT
n 3
To a first approximation we thus get
n R 2 n ,n
S gg n .
1
n ln ln p
nT
42
Steps
(1) Set iteration N=1
(2) Start with the initial guess on the PSD given by
n R 2 n ,n
S N n
1
n ln ln p
nT
(3)Evaluate and using
2
x
2
x
H S gg d &
2
2
x
N
H 2 S gg d
2
2
x
N
43
1
2 x 2
(4)Evaluate R n , n 2 x ln
N 2 2
ln p .
xT
n
(5)Obtain an improved estimate of PSD using
2
N 1 R ( )
S ( ) S ( ) N
N
R ( )
(6)Stop iterations if the PSD function has converged;
if not go to step 3.
44
0
10
Derived
Used
-1
10
PSD (m2/s 4)/(rad/s)
-2
10
-3
10
-4
10
-5
10
0 100 200 300 400 500 600 700
Frequency rad/s
45
6
Target
Derived
5
4
Acceleration m/s 2
0
0 100 200 300 400 500 600 700
Frequency rad/s
46
2
0
-2
0 5 10 15 20 25 30
2
0
-2
Acceleration m/s 2
0 5 10 15 20 25 30
2
0
-2
0 5 10 15 20 25 30
2
0
-2
0 5 10 15 20 25 30
2
0
-2
0 5 10 15 20 25 30
Time s
47
0
10
Target
Simulation
-1
10
PSD (m2/s 4)/(rad/s)
-2
10
-3
10
-4
10
-5
10
0 100 200 300 400 500 600 700
Frequency rad/s
48
7
Target
From derived PSD
6
From samples
5
Acceleration m/s 2
0
0 100 200 300 400 500 600 700
Frequency rad/s
49
Problem 37
Figure (next slide) shows the psd function of ground
acceleration which is modeled using Kanai-Tajimi's
approach with g 15rad/s, g 0.6 .
Determine the pseudo-acceleration spectra compatible
with this psd function. It may be assumed that the ground
acceleration is a zero mean, stationary Gaussian random
process. The duration of the acceleration can be taken to be
30s and the target response spectra may be interpreted as the
locus of the 84% percentile point and damping may be
taken to be 5%.
50
2
1.8
1.6
1.2
0.8
0.6
0.4
0.2
0
0 100 200 300 400 500 600
Frequency rad/s
51
40
35
30
Acceleration m/s 2
25
20
15
10
0
0 100 200 300 400 500 600
Frequency rad/s
52
2
Target
1.8 Derived
1.6
1.4
PSD (m2/s 4)/(rad/s)
1.2
0.8
0.6
0.4
0.2
0
0 100 200 300 400 500 600
Frequency rad/s
53
Discussion on outcrossing theory of random
Processes and applications to problems of
Load combination
54
Problem 38 : Load combination
Let Q t be a quasi-static load on a structure
(e.g., sustained live load). If we are interested in designing
the structure for this load, we can estimate
Qm max Q t and use that is the design.
t
P0 E N T
PF P0 E N T
56
Recall that in order to characterize the average rate
of crossing of a critical barrier by a random process,
we need the jpdf of the process and its derivative
at the same time instant.
Consider
Q t Q1 t Q2 t
Q t Q t Q t
1 2
U Q2 t
V Q t
2
pQQUV
q, q, u, v pQ1Q1Q2Q q u, q v, u, v
pQ1Q1 q u, q v pQ2Q u , v
pQQ q, q , u, v p q u, q v p u, v dudv
Q1Q1 Q2Q
57
X t RECALL
Y t
Y t U X t
Y t
Y t X t X t
Y t
Y t X t X t
N t
T
N (T ) X t X t dt
0
58
N T number of times the level t is crossed in 0, T
with positive slope
Y (t ) U Q t t
Y t Q t t Q t t
Z t Q t t Q t t U Q t t
T
N T Q t t Q t t U Q t t dt
0
0, t E Q t t Q t t U Q t t
59
t E Q t t Q t t U Q t t
q t q t U q t p
QQ q, q; t dqdq
q t pQQ t , q; t dq
t
q t pQ1Q1 t u , q v pQ2Q 2 u, v dudv dq
t
T
E N T t dt
0
60
Remark
The evaluation of
t
q t pQ1Q1 t u , q v pQ2Q 2 u, v dudv dq
t
is possible for Gaussian models for loads. A general solution
is difficult to obtain.
61
Problem 39
Discussion of fatigue crack growth modeling under
random loads using fracture mechanics concepts
62
Fracture mechanics based approaches
Basic assumption:
there exists a crack in the structural component.
Question:
Given the geometry of the crack, loads, boundary conditions,
can we say if the crack is likely to grow?
r , u1
a r
2 cos
1 2 sin 2
2 2
x1
a r 2
u2 sin 2 2 cos
Mode I 2 2 2
u3 0 64
Stress Intensity Factor (SIF) and Critical SIF
In the expressions for stress and displacement components
the quantities and a appear together.
Can we give a name to the quantitiy a ?
Recall: EI , mv, 0.5mv 2 , x ct , Reynold's number,
Froude's number
Definition
K I lim 2 r 22 r , 0
r 0
Region I Region II
Crack Stable crack
Initiation growth
da da
C K
n
dn dn
Region III
Unstable crack
growth
log K 67
Model for Stage II crack growth
da
f K , K max , K min , K th , E , , ys , ult , i , ki
dN
i environmental variables
(temperature, humidity, salinity, etc.,)
ki other material or mechanics variables
(frequency of excitation, grain size, ...)
Dimensional analysis
da K K Ic K th ys ult
2
F R, , , , , i i , i ki
dN E K K E E
68
Paris - Erdogan model
da
C K ; K 0; a 0 a0
m
dN
da
log log C m log K
dN
Example a in m & K in Mpa m
da
6.80 10 K
12 3.00
Ferrite pearlite steel :
dN
69
69
Modeling of uncertainties
Sources
Macro-properties of specimens
(geometry, dimensions, and
material properties may differ
between specimens).
External loading.
Inhomgenous microstructure.
dN
dN
t
N
2
71
Cumulative jump models
(Reference : K Sobczyk and B F Spencer Jr., 1992,
Random fatigue from data to theory, Academic Pr ess )
Define A t , random process: length of the dominant
crack at time t.
(sample point). To be suppressed in further description.
N t
A(t ) A0 Y ;
i 1
i Yi Ai
; k 0,1, 2, ,
k!
Yi i 1 iid sequence of non-negative rvs with a
common pdf pY y
N t Yi i 1
P A t a PA a; t [PDF]
dPA a; t
p A a; t [pdf]
da
73
N t
Let A t A0 A1 t with A1 t Y
i 1
i
p A a; t exp 0t a ;a 0
k 0 k ! k 1!
1
0t
exp 0t a I1 2 0 at ; a 0
a
where I1 Bessel's function of the first order.
A t A0 A1 t p A a; t p A1 a A0 , t
76
Model for life time
Let be the critical crack length
(esimtated from the knowledge of K Ic ).
T time required for A(t ) to reach the critical length .
P T t P A t
PT t
0t exp
1 a A0
0t a A0 I1 2
0 a A0 t da
0
77
Estimation of system parameters
Model parameters: 0 associated with the process N t ;
: associated with pY y .
Idea: derive these model parameters from laws such as
the Paris law. An approximate method to achieve this
would be to modify the Paris law to allow for randomness
in applied stress and system parameters.
da p
C K ; a p 0 a0
m
dN
78
Let S (t ) be the stress field that is modeled as
a Gaussian, stationary random process.
What is meant by cycle?
d d dt 1 d
N s t
dN dt dN s dt
da p
s C S max S min a p ; a p 0 a0
m
dt
Interpret s as the average rate of peaks in S (t ).
79
s average rate of zero crossing of S t .
1
4 2
S s d
1
s
2 2
S d
s
Interpretation of ΔK
Recall: K a . Interpret S max S min mean range.
S mr S max S min 2 S rms
2
1
2
1
22 2
1 ; n S s d
n
0 4
m
da p m
s C a p2 S max S min ; a p (0) a0
m
dt 80
Interpretation of 0
If we take N (t )=number of peaks above a level s0 ,
then 0 becomes the average rate of peaks in S (t )
above level s0 . Select s0 fatigue limit of material
(that is the endurance limit).
1
1 4 2 4
0 { 1 s0
2 2 4 2
2
s
2 2 s0 0 2 }
2
4 2
81
H
o
w
t
o
f
i
n
d
?
Select such that
t*
F Ap t A t dt
2
is minimized.
Here t time required by Ap t to reach .
*
82