System Reliability Based Topology Optimization of...
Transcript of System Reliability Based Topology Optimization of...
Junho Chun*
University of Illinois at Urbana-Champaign, USA
Junho Song
University of Illinois at Urbana-Champaign, USA
Glaucio H. Paulino
University of Illinois at Urbana-Champaign, USA
ICOSSAR 2013
11th
International Conference on Structural Safety & Reliability
System Reliability Based Topology
Optimization of Structures under
Stochastic Excitations
Columbia University, New York, NY, June 20, 2013
2
Structural engineering under Natural hazards
and risks
San Francisco Earthquake, 1907 http://www.documentingreality.com Tacoma bridge, 1940 http://failuremag.com
Structural system Courtesy of Skidmore, Owing and Merrill, LLP
3
Motivation
http://en.wikipedia.org/wiki/John_Hancock_Center
John Hancock Center Ssiger International Plaza
Courtesy of Skidmore, Owing and Merrill, LLP
Research aims to find optimal structural system under stochastic excitations
Incorporation of random vibration theories into topology optimization Topology optimization
Stromberg et al. (2011)
Stochastic excitation
0 5 10 15 20 25 30 35 40 45 50-600
-400
-200
0
200
400
600
800HYOGOKEN NANBU EQ - KOBE-JMA3.EW 1/17/1995 DT=0.02 Amax=617.14gal
x - time / DT = 0.02
Accele
rati
on
(g
al)
4
Contents
Topology optimization
Topology optimization under stochastic excitations
System reliability-based topology optimization under
stochastic excitations
Numerical example: multi-story building structure
Conclusions & future research
5
?
Topology Optimization
Topology optimization aims to identify optimal material layouts of problems
through mathematical programming while fulfilling given design constraints
min compliance
. volume constraints t
d
Applications
Nguyen at el. (2010) Stromberg et al. (2011)
?
0 void0 1
1 solid
e
e
e
Design variable : Element density
T
0( ) e
e p
e e e
K B D B
T( ) e
q
e e e ed
M N Ne
6
Discrete Representation Method
(Der Kiureghian, 2000)
Discretization of Random process
0 2 4 6 8 10Time
f(t)
0 2 4 6 8 10Time
v
Der Kiureghian, A. (2000). The geometry of random vibrations and solutions by FORM and SORM. Probabilistic Engineering Mechanics, 15(1),: 81-90.
o v: uncorrelated standard normal random variables
o s(t): deterministic basis functions based on the
spectral characteristics of the process
1
( ) ( ) ( ) ( ) ( )n
T
i i
i
f t t v s t t t
s v
W( )t
, g g Filter
White Noise 0
( )f t
Example : Filtered white noise (earthquake)
0
1 1
T
0
1
( ) ( ) ( )
( ) ( )
2π / ( ) ( )
t
n n
i i i f i
i i
n
i f i
i
f t v s t d
v s t W h t t t
t v h t t t t
s v
Magnitude of WN impulse
Unit-impulse response
function of the filter
7
Response of Linear System to Stochastic
Excitation
o hs(t) : the unit-impulse response function of the system
Response
Linear system + Gaussian process
Duhamel’s Integral
0
( ) (τ) ( τ) τ
t
su t f h t d
T
1 10
( ) ( ) ( τ) τ ( ) ( )
t n n
i i s i i
i i
u t v s h t d v a t t
a v0
( ) ( ) ( ) , 1,...,
t
i i sa t s h t d i n
Deterministic, time-dependent
- filter + structure
Random, time-independent
a(t) in FEM settings
1 1 1 0
2 1 2 2 0
1 0 1 2 1 1 0
0 1 2 1 0
( ) ( ) 0 0 0 ( )
( ) (2 ) 0 0 ( )
( ) ( ) 0 ( )
( ) ( ) ( )
n n n n
n n n n
u t u t v a t
u t u t v v a t
u t u t t v v v a t
u t u t v v v v a t
Numerical time integration Inversely computed
. . ( ) ( , ) ( ) ( , ) ( ) ( , ) ( , )e g t t t t M ρ u ρ C ρ u ρ K ρ u ρ f ρ
8
Instantaneous Failure Probability
‘Instantaneous’ failure events of a linear system
0 0 0 0
000 0 0
0 0
0 ( ) 0
ˆ( ) β , ( )
T
T
P u t u P u t P g u
tug u u t t
t t
a v
av α v
a a
0 0 0 0
0 0 0 0 0
[ β , ]
ˆβ , / ( ) *
P u t u u t
u t u t t
a α v
T0 0 0 0( )fE u t u t u a v
Failure Probability, P(Ef)
t
( )u t
0u
0t
Failure
Failure domain
Safe domain 0 0( ) 0u u t v i
v j
0 0( ) 0u u t
T
0 0( ) 0u t a v
MPP
( , )0 0β u t
*( , )0 0u tv
0ˆ ( )tα
Topology
Optimization
Reliability
Analysis
9
Topology Optimization of Structure under
Stochastic Excitations
target
min ( ( ))
. ( ) 1,...,
( ) ( , ) ( ) ( , ) ( ) ( , ) ( ))
i fi
f
s t P E P i n
with t t t f t
dρ d
M ρ u ρ C ρ u ρ K ρ u ρ Ml
Probabilistic constraints on responses of a structure
(Instantaneous failure probability)
Objective function
Stochastic topology optimization framework (Chun, Song, Paulino 2012)
Stochastic excitation (filtered Gaussian process)
5m
3aP 3bP
2aP 2bP
1aP 1bP
Design domain:
Bilinear quadrilateral element, Q4
LEVEL 04
EL: 15m
LEVEL 03
EL: 10m
LEVEL 02
EL: 5m
GROUND
EL: 0m
Frame element
0 2 4 6 8 10-50
0
50
Acce
lation
(m
/s2)
Time (seconds)
Stochastic excitation
Component failure events, Ei
A B
β 1.3 2.2
Pf 9.68% 1.39%
target( ) 1,...,i fiP E P i n
Solve optimization problem
(CRBTO)
10
SYSTEM RBTO of Building Structures under
Stochastic Excitations
SRBTO : System Reliability Based Topology Optimization
Motivation : further extension of the stochastic topology optimization framework
aforementioned is expected to provide more realistic means for the prediction of
a given building’s response to stochastic excitations
- Time: the first passage probability
- Location
- Failure modes: different types of design constraints such as a target tip displacement, a target
natural frequency of a structure
0
1
( ) ( max | ( ) |) ( )n
n
sys t t i
i
P E P x X t P X t x
Matrix-based System Reliability Method (MSR)(Song and Kang 2009, Kang et
al. 2012)
General system
Statistical dependency
Parameter sensitivity
T
T
( ) ( ) dependent( ; )
independent
t t
syst
syst t
sys
f d PP E
P
ssc p s s s
Pc p
Event vector Conditional probability
vector
Probability vector Joint PDF
Song, J., and W.-H. Kang (2009). System reliability and sensitivity under statistical dependence by matrix-based system reliability method. Structural Safety,
Vol. 31(2), 148-156.
Kang, W.-H., Y.-J. Lee, J. Song, and B. Gencturk (2012). Further development of matrix-based system reliability method and applications to structural
systems. Structure and Infrastructure Engineering: Maintenance, Management, Life-cycle Design and Performance. Vol. 8(5), 441-457.
11
SYSTEM RBTO of Building Structures under
Stochastic Excitations
,
0 0
min ( ( ))
. ( , , , ) 0, 1,...,
( ) ( ) dependent ( ; )
independent
with ( ) ( , ) ( ) ( , ) ( ) ( , ) (
ti
ti
P
t
i iP
T t t
syst
sysT t t
sys
f
s t g g t u P i n
f d PP E
P
t t t
d
ss
ρ d
ρ
c p s s sP
c p
M ρ u ρ C ρ u ρ K ρ u ρ f , )t ρ
Probabilistic constraints solved by MSR method
System equation
𝑃𝑖𝑡 quantile of the i-th limit-state function
: Filtered density
: Mass matrix, : Damping matrix, : Stiffness matrix
( ) : Stochastic
System fa
process (e.g., for earthquake loading ( ) ( )) ( ))
ilure event: ,
g
sys sy
f t t u t f t
E P
ρ
M C K
f Ml Ml
target Target failure probab: ilitys
LEVEL k
EL : - m
LEVEL GN
EL : 0 m
LEVEL j
EL : - m
LEVEL i
EL : - m
0 1 2 3 4 5 6 7-1500
-1000
-500
0
500
1000
1500
t, sec
f(t)
0 1 2 3 4 5 6 7-0.1
-0.05
0
0.05
0.1
Drift r
atio
3
2
1
Stochastic excitations
Responses
Formulation
12
Sensitivity Calculation
Formulation
( ) / , ( ) / t
i e i iP d P P s s
/ , /t ti i
t
e iP Pg d g P
1
2
1
1
2 2
1 1
( ) β ( ) 1
β ( β )
β ( ) 1
β ( β )1
β ( )1 1
( β
(
)
)
1 1
t
i i i
t t t t
i i i i
mt
i ik kk
t tmi i
ikk
mt
i ik kk
tm mi
ik
i
t
i
ikk k
P P P
P
r s
r
r s
r r
P
s ss
T
0 0
T
0
β 1
β β ( β )
β ( , ) 1
β ( β )
1( , )
( β )
ti
t ti i
tP Pi
t t t t
i i i i
t
i
t
P
t
i i
t
t
i
i
g g
P
u t
P
t
g
a ρ
a ρ
1
1
00T
1 10
00
1 1
( ) ( , )( , )
( , )
( , )( , , )
e
e
t t
e
i i
nP j
j j e
t n nji i
i
j i j e
n njt i
i i
e
P
j
e
i e
g
d
P a ta t
dt
a tc t P
g
d
d
ρρ
a ρ
ρρ
( )0i
e
P
d
s
Analytical derivation
Design variables: d, 𝑃𝑖𝑡
,
0 0
min ( ( ))
. ( , , , ) 0, 1,...,
( ) ( ) dependent ( ; )
independent
with ( ) ( , ) ( ) ( , ) ( ) ( , ) (
ti
ti
P
t
i iP
T t t
syst
sysT t t
sys
f
s t g g t u P i n
f d PP E
P
t t t
d
ss
ρ d
ρ
c p s s sP
c p
M ρ u ρ C ρ u ρ K ρ u ρ f , )t ρ
Calculation of sensitivity of the terms in the constraints with respect to design
variables is an essential procedure for efficient gradient-based optimization
13
Optimization Procedure
Result: Optimal Topology
Yes
No (iterate)
Convergence Criteria
achieved?
Random Vibration Analysis using dynamic FEA,
and compute Pf using MSR method
Sensitivity Analysis of objective and constraint
functions
Update Design Variables using Mathematical
Programming
Generating Stochastic
Excitation Model
Initial Design
ρ = ρInitial 0 1 2 3 4 5 6 7 8
-2000
-1000
0
1000
2000
t, sec
p
0 1 2 3 4 5 6 7 8-4
-2
0
2
t, sec
dis
pla
cem
en
ts
14
System Failure event:
It occurs when at least one of component
events happens
LEVEL 07
EL : 30m
GROUND
EL : 0m
LEVEL 02
EL : 5m
LEVEL 05
EL : 20m
LEVEL 06
EL : 25m
5m
LEVEL 03
EL : 10m
LEVEL 04
EL : 15m
Design domain:
Bilinear quadrilateral element, Q4
Frame element
Numerical example : Six-story building
7
2
isys f
i
E E
Failure event: Inter-story drift ratio
It occurs when inter-story drift ratio
exceeds a prescribed threshold value
Filtered Gaussian process using
Kanai-Tajimi filter 2
2
2
2
(2 1)( ) exp( ) sin( 1 )
1
exp( )2 cos( 1 )
f f
f f f f f
f
f f f f f f
h t t t
t t
,
0 0
min ( ( ))
. ( , , , ) 0, 1,...,
( ) ( ) dependent ( ; )
independent
with ( ) ( , ) ( ) ( , ) ( ) ( , ) (
ti
ti
P
t
i iP
T t t
syst
sysT t t
sys
V
s t g g t u P i n
f d PP E
P
t t t
d
ss
ρ d
ρ
c p s s sP
c p
M ρ u ρ C ρ u ρ K ρ u ρ f , )t ρ
15
Numerical example : Six-story building
LEVEL 07
EL : 30m
GROUND
EL : 0m
LEVEL 02
EL : 5m
LEVEL 05
EL : 20m
LEVEL 06
EL : 25m
LEVEL 03
EL : 10m
LEVEL 04
EL : 15m
CRBTO :
βti=1.7 (Pt=4.46%)
volfrac=28.5%
(Ptsys=6.67%)
SRBTO :
βtsys=1.502 (Pt
sys=6.67%)
volfrac=28.45%;
CRBTO:
- All target reliability index βti=1.7
SRBTO: - Psys: After CTBTO the probability that
at least one of the constraints is violated
(i.e. series system) is estimated by the
MSR method
16
Numerical example : Six-story building
0 20 40 60 80 100 1202
4
6
8
10
Vol
um
e (
m3)
0 20 40 60 80 100 1200
2
4
6
Rel
iabi
lity
ind
ex
1
2
3
4
5
6
0 20 40 60 80 100 1200.06
0.08
0.1
0.12
Fai
lure
Pro
ba
bilit
y
Iteration
Pt
sysPf
sys
0 1 2 3 4 5 6 7 8 9 10-100
0
100
f(t)
(m
/s2)
Ground acceleration
0 2 4 6 8 10-0.04-0.02
00.020.04
6/L
6
Initial system
0 2 4 6 8 10-0.04-0.02
00.020.04
5/L
5
0 2 4 6 8 10-0.04-0.02
00.020.04
4/L
40 2 4 6 8 10
-0.04-0.02
00.020.04
3/L
3
0 2 4 6 8 10-0.04-0.02
00.020.04
2/L
2
0 2 4 6 8 10-0.04-0.02
00.020.04
1/L
1
Time (seconds)
0 2 4 6 8 10-0.04-0.02
00.020.04
6/L
6
Optimized system
0 2 4 6 8 10-0.04-0.02
00.020.04
5/L
5
0 2 4 6 8 10-0.04-0.02
00.020.04
4/L
4
0 2 4 6 8 10-0.04-0.02
00.020.04
3/L
3
0 2 4 6 8 10-0.04-0.02
00.020.04
2/L
2
0 2 4 6 8 10-0.04-0.02
00.020.04
1/L
1
Time (seconds)
Convergence history (SRBTO) Dynamic responses (SRBTO)
17
Conclusions & Future Research
Concluding remarks
Development of a new framework incorporating random vibration theories
into topology optimization using a discrete representation method for
stochastic processes
Development of SRBTO under stochastic excitation considering statistical
dependency of component events using MSR method.
Application of the proposed method to find optimal bracing system of the
structure satisfying probabilistic constraints defined in system level.
Future Direction
Incorporation of first passage probability in stochastic topology
optimization
Development of an efficient approach for computing sensitivity of system
reliability with a large number of component events
Karol Fellowship
National Science Foundation
Skidmore, Owings, Merrill. LLP
19
20
Reliability-Based Design Optimization
‘Deterministic’ design optimization (DO)
Reliability-based design optimization
(RBDO)
min ( )
. ( ) 0, 1,...,
obj
i c
lower upper
f
s t g i n
dd
d
d d d
target
min ( , )
. ( ) ( , ) 0 , 1,...,
,
obj,
sys i sys c
lower upper lower upper
f
s t P E P g P i n
X
Xd
X
X
d
d
d d d
Objective function
v i
v j
Limit-state functions
Safe domain
High probability of
failure
Low probability
of failure
-3
-2
-1
0
1
2
3
-3
-2
-1
0
1
2
3
0
0.05
0.1
0.15
0.2
0.25
21
System Reliability-Based Design
Optimization (SRBDO)
Component Reliability Based Design Optimization (CRBDO)
target
min ( , )
. ( , ) 0 , 1,...,
,
obj,
i f c
lower upper lower upper
f
s t P g P i n
X
Xd
X
X
d
d
d d d
target
min ( , )
. ( ) ( , ) 0 , 1,...,
,
obj,
sys i sys c
lower upper lower upper
f
s t P E P g P i n
X
Xd
X
X
d
d
d d d
Limit-state
functions
Song, J., and W.-H. Kang (2009). System reliability and sensitivity under statistical dependence by matrix-based system reliability method. Structural Safety,
Vol. 31(2), 148-156.
Kang, W.-H., Y.-J. Lee, J. Song, and B. Gencturk (2012). Further development of matrix-based system reliability method and applications to structural
systems. Structure and Infrastructure Engineering: Maintenance, Management, Life-cycle Design and Performance. Vol. 8(5), 441-457.
22
LEVEL 07
EL : 30m
GROUND
EL : 0m
LEVEL 02
EL : 5m
LEVEL 05
EL : 20m
LEVEL 06
EL : 25m
5m
LEVEL 03
EL : 10m
LEVEL 04
EL : 15m
Design domain:
Bilinear quadrilateral element, Q4
Frame element
Numerical example : Six-story building
T T
0 , 0 ,
0
T T
0 , 0 ,
0
T T
0 ( 1), 0 ( 1),
( ( , ) ( , ) )0 for 2
2
( ( , ) ( , ) )
2
( ( , ) ( , ) ) 0 for 3
2
i
i Left i Right
f i
i
i Left i Right
i
i
i Left i Right
i
t tE u i
L
t tu
L
t ti
L
a ρ a ρ v
a ρ a ρ v
a ρ a ρ v, ..., 7
7
2
isys f
i
E E
Failure event: Inter-story drift ratio
System Failure event
Filtered Gaussian process using KT filter 2
2
2
2
(2 1)( ) exp( ) sin( 1 )
1
exp( )2 cos( 1 )
f f
f f f f f
f
f f f f f f
h t t t
t t
,
0 0
min ( ( ))
. ( , , , ) 0, 1,...,
( ) ( ) dependent ( ; )
independent
with ( ) ( , ) ( ) ( , ) ( ) ( , ) (
ti
ti
P
t
i iP
T t t
syst
sysT t t
sys
V
s t g g t u P i n
f d PP E
P
t t t
d
ss
ρ d
ρ
c p s s sP
c p
M ρ u ρ C ρ u ρ K ρ u ρ f , )t ρ