Phong Nguyen

Structural Reliability_Makeup Problem

1

Part I: My own Problem

Question 1:
a. Euler Bernoulli Beam Theory

g Ec , Em , W
The limit state function:
4 AWL4
5 AWL4

5 AD B 2 384 AD B 2

B Ec , Em , W

h/ 2

h/ 2

Ab

L
B Ec , Em , W
400

1 z
Em dz b Ec Em J1 Embh;
2 h

Ec Em

1 z
B b Ec Em Em z dz b Ec Em J 2 ;
2 h
h /2

h /2

Db

h /2

h /2

bh 3
1 z
Em z 2 dz b Ec Em J 3 Em
;
12
2 h

Ec Em

1
1
1
J1 ; J 2 ; J 3
b 0.2; h 0.5; L 12
6
48
240
X 1 : Ec N 150e6, 0.05 x150e6 ; X 2 : Em N 70e6, 0.05 x70e6 ; X 3 : W N 12, 0.25 x12 ;

No.
Iterati
on

Variable
s

Variable
values

X1
X2
X3

150000000
70000000
12
147824249.
912
68691956.5
55
21.9574
146081907.
58

X1
2

X2
X3
X1

0.0003522
0.0004537
-0.0040297
0.0006654

Directio
Reliabilit
n
y index
cosines
0.0865
0.1115 3.35268
-0.9900
0.1569

0.0008603

0.2029

3.32944

-0.0040988
0.0006735

-0.9666
0.1566

3.32944

gradients

Phong Nguyen

Structural Reliability_Makeup Problem

2

67635876.1
07
21.654
146088824.
86
67632683.5
95
21.654

X2
X3
X1
4

X2
X3

Probability of failure (AFOSM):

0.0008736

0.2031

-0.0041562
0.0006735

-0.9665
0.1566

0.0008736

0.2032

-0.0041562

-0.9665

3.32944

pf = (3.32944 )=4.351 x 1 04
4

pf =4.432 x 10 ; COV =11

Monte Carlo Simulation:

b. Timoshenko beam theory

g Ec , Em ,W ,
The limit state function:
2
2
AD

B
L2 AH W 4

T Ec , Em ,W ,
;
4

2
H AD B 4
L

L
T Ec , E m , W ,
400

5 A
12 1

X 1 : Ec N 150e6, 0.05 x150e6 ; X 2 : Em N 70e6, 0.05 x70e6 ; X 3 : W N 12, 0.25 x12 ;

X 4 : N 0.3, 0.1x0.3 ;

No.
Iterati
on

Variabl
es

Variable
values

X1
X2
X3
X4

150000000
70000000
12
0.3
147859795.9
67
68715182.45
9
21.7868

X1
X2
X3

0.000355
0.000457
-0.004062
-1.66e-006

Directio
n
cosines
0.0866
0.1114
-0.9900
-0.0004

0.000666

0.1559

0.00086

0.2012

-0.00413

-0.9671

gradients

Reliabili
ty index

3.29522
3.27315

Phong Nguyen

X4
X1
3

X2
X3
X4
X1

X2
X3

Structural Reliability_Makeup Problem

3

0.3
146173844.5
33
67694908.15
6
21.496
0.3
146180296.4
05
67691930.20
1
21.496

X4

0.3

Probability of failure:

-3.06e-006

-0.0007

0.000674

0.1556

0.000872

0.2015

-0.004187
-3.06e-006
0.000674

-0.9671
-0.0007
0.1556

0.000872

0.2015

-0.004187
-3.06e-006

3.27315

3.27315

-0.967
1
-0.0007

pf = (3.27315 )=5.318 x 1 04

Monte Carlo Simulation:

Question 2:

pf =5.408 x 104 ; COV =10.61

g Ec , Em , c , m , p 0.7 p f

The limit state function:

No.
Iterati
on
1
2

Variabl
es

Variable
values

gradients

X1
X2
X3
X4
X5
X1

150000000
70000000
3400
2702
16
159005711.7

-0.11217
-0.14503
0.01983
0.03161
0.224
-0.10947

Directio
Limit
n
state
consines function
-0.3843
-0.4969
0.91191
0.0680
7
0.1083
0.7675
-0.3797 0.00569

Phong Nguyen

X2
X3
X4
X5
X1
3

X2
X3
X4
X5
X1

X2
X3
X4
X5

Structural Reliability_Makeup Problem

4

3
75434182.09
9
3392.781
2692.857
15.233
158951699.2
54
75325284.13
5
3392.361
2692.327
15.218
158949909.4
36
75329300.95
3392.365
2692.331
15.218

-0.13955

-0.4840

0.020606
0.032834
0.224

0.0715
0.1139
0.7769

-0.109475

-0.3796

-0.139688

-0.4844

0.020601
0.032827
0.224

0.0714
0.1138
0.7767

-0.109477

-0.3796

-0.139683
0.020601
0.032827
0.224

-0.4843
0.0714
0.1138
0.7767

g 8.105e 12; 3.14369

At design point:
158950126.06

75329132.638

;u
x 3392.365

2692.331

15.218

Probability of failure:

1.19335

1.52261
-0.22457

-0.35783
-2.4417

pf = (3.14369 )=8.341 x 1 04

Monte Carlo Simulation:

pf =8.07 x 104 ; COV =10.9

Question 3:
a. Euler Bernoulli Beam Theory

2.2444e
-6

3.2901e
-9

Phong Nguyen

Structural Reliability_Makeup Problem

5

g Ec , Em ,W , p
The limit state function:
B Ec , Em , W , p

At design point:

L
B Ec , Em ,W , p
400

4 AWL4
5 AWL4

5 AD B 2 384 AD B 2

g 3.592e 9; 3.32204

0.1549
146142439.104062
-0.514341452791772

67634938.387808
-0.675731889197713
0.2034

x
; u
;
21.6130078583459
3.20433595278195
-0.9645

-0.0654
2.043367196296
0.216835981479993

Probability of failure:

pf = (3.32204 )=4.4681 x 10

i
Omission sensitivity factor:
b. Timoshenko beam theory

3.32944
1.0022
3.32204

g Ec , Em , W , , p
The limit state function:

At design point:

1 0.0654 2

L
T Ec , E m , W , , p
400

g 3.24425e 9; 3.26592

146231937.078044
-0.502408389594177

67694197.5606937
-0.65880069694467

x 21.4557335680745 ; u 3.15191118935816 ;

0.300069144781454
0.0023048260484783

2.04253916402407
0.212695820120341

Probability of failure:

pf = (3.26592 )=5.4555 x 1 04

0.1539

0.2017
-0.9651

-0.0007
-0.0652

Phong Nguyen

Structural Reliability_Makeup Problem

6

i
Omission sensitivity factor:

3.27315
1.0022
3.26592

1 0.0652 2

Code Matlab:

Q.1
Bernoulli beam:
%% Use AFOSM to calculate probability of failure.
% Limit state function: g=L/360-5*A*W*L^4/(384*(AD-B^2));
syms betav x1 x2 x3 x4
syms z lamda
format long
%% Input
b=0.2; h=0.5; L=12;
p=2;
muy=[150e6; 70e6; 12];
%Mean values of Ec,Em and W
cov=[0.05; 0.05; 0.25]; % coefficients of variantion of Ec,Em and W.
xichma=cov.*muy;
% Standard deviation.
%%
J1=int((z/h+0.5)^p,z,-h/2,h/2);
J2=int(z*(z/h+0.5)^p,z,-h/2,h/2);
J3=int(z^2*(z/h+0.5)^p,z,-h/2,h/2);
AA=b*(x1-x2)*J1+x2*b*h;
BB=b*(x1-x2)*J2;
DD=b*(x1-x2)*J3+x2*b*h^3/12;
gg=L/400-5/384*AA*x3*L^4/(AA*DD-BB^2);
g1=diff(gg,x1);
g2=diff(gg,x2);
g3=diff(gg,x3);
beta_trial=3;
% guessed value of raliability index;
tolerance=10^-7;
% interation error.
itermax=50;
% the maximum number of iteration
x0=muy;
% Guessed values
er=1;
i=1;
while er > tolerance
beta(i)=beta_trial;
x(:,i)=x0;
g(1,i)=subs(g1,{x1,x2,x3},{x(1,i),x(2,i),x(3,i)});
g(2,i)=subs(g2,{x1,x2,x3},{x(1,i),x(2,i),x(3,i)});
g(3,i)=subs(g3,{x1,x2,x3},{x(1,i),x(2,i),x(3,i)});
grad(:,i)=g(:,i).*xichma;
%Gradients in new coordinate system
anpha(:,i)=grad(:,i)/sqrt(sum(grad(:,i).^2));
xnew(:,i)=muy-betav*anpha(:,i).*xichma;

Phong Nguyen

Structural Reliability_Makeup Problem

7

A=b*(xnew(1,i)-xnew(2,i))*J1+xnew(2,i)*b*h;
B=b*(xnew(1,i)-xnew(2,i))*J2;
D=b*(xnew(1,i)-xnew(2,i))*J3+xnew(2,i)*b*h^3/12;
gbeta=L/400-5/384*A*xnew(3,i)*L^4/(A*D-B^2);
betar=(double(solve(gbeta,betav)));
%solve reliability index
x0=muy-min(betar(betar>0))*anpha(:,i).*xichma;
er=abs((min(betar(betar>0))-beta(i)));
% error of reliability index
beta_trial=min(betar(betar>0));
betaf(i)=min(betar(betar>0));
i=i+1;
if i>itermax, break, end
end
pf=normcdf(-betaf);

Timoshenko Beam:
%% Use AFOSM to calculate probability of failure.
% Limit state function: g=L/400-[(AD-B^2)pi^2/L^2+AH]W/[H(AD-B^2)pi^4/L^4];
syms betav x1 x2 x3 x4
syms z
format long
%% Input
b=0.2; h=0.5; L=12;
p=2;
muy=[150e6; 70e6; 12; 0.3];
%Mean values of Ec,Em and W, poison ratio
cov=[0.05; 0.05; 0.25; 0.1]; % coefficients of variantion of Ec, Em, W, poison ratio.
xichma=cov.*muy;
% Standard deviation.
%%
J1=double(int((z/h+0.5)^p,z,-h/2,h/2));
J2=double(int(z*(z/h+0.5)^p,z,-h/2,h/2));
J3=double(int(z^2*(z/h+0.5)^p,z,-h/2,h/2));
AA=b*(x1-x2)*J1+x2*b*h;
BB=b*(x1-x2)*J2;
DD=b*(x1-x2)*J3+x2*b*h^3/12;
HH=(5/12)*AA/(1+x4);
gg=L/400-1.273*((AA*DD-BB^2)*pi^2/L^2+AA*HH)*x3/(HH*(AA*DD-BB^2)*pi^4/L^4);
g1=diff(gg,x1);
g2=diff(gg,x2);
g3=diff(gg,x3);
g4=diff(gg,x4);
beta_trial=3;
% guessed value of raliability index;
tolerance=10^-7;
% interation error.
itermax=10;
% the maximum number of iteration
x0=muy;
% Guessed values
er=1;
i=1;
iternum=5;
for i=1:iternum
beta(i)=beta_trial;
x(:,i)=x0;
g(1,i)=subs(g1,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});

Phong Nguyen

Structural Reliability_Makeup Problem

8

g(2,i)=subs(g2,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
g(3,i)=subs(g3,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
g(4,i)=subs(g4,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
grad(:,i)=g(:,i).*xichma;
%Gradients in new coordinate system
anpha(:,i)=(grad(:,i)/sqrt(sum(grad(:,i).^2)));
xnew(:,i)=muy-betav*anpha(:,i).*xichma;
A=b*(xnew(1,i)-xnew(2,i))*J1+xnew(2,i)*b*h;
B=b*(xnew(1,i)-xnew(2,i))*J2;
D=b*(xnew(1,i)-xnew(2,i))*J3+xnew(2,i)*b*h^3/12;
H=(5/12)*A/(1+xnew(4,i));
gbeta=L/400-1.273*((A*D-B^2)*pi^2/L^2+A*H)*xnew(3,i)/(H*(A*D- B^2)*pi^4/L^4);
%limit state function in term of beta
betar=(double(solve(gbeta,betav)));
%solve reliability index
betaf(i)=min(betar(betar>0));
x0=muy-min(betar(betar>0))*anpha(:,i).*xichma;
er=abs((min(betar(betar>0))-beta(i)));
% error of reliability index
beta_trial=min(betar(betar>0));
end
pf=normcdf(-betaf)

Q2.
%% Use New FOSM (Rackwitz-Fiessler algorithm) to calculate probability of failure.
% Limit state function: g=L/360-5*A*W*L^4/(384*(AD-B^2));
syms x1 x2 x3 x4 x5
syms z lamda
format long
%% Input
b=0.2; h=0.5; L=5;
p=2;
muy=[150e6; 70e6; 3400; 2702; 16];
%Mean values of Ec,Em and pro_c, pro_m, omega_p
cov=[0.05; 0.05; 0.01; 0.01; 0.02]; xichma=cov.*muy;
% Standard deviation.
%%
omega=1/14073748835532800*(-318954283212249450785993520220890*x3*x1-...
636365057765537158911562306801735*x4*x1-636365057765537159064790073939380*x3*x2-...
1269643098213150832808730680598670*x4*x2-1228285163894519192601457852416*x1*pi*x4-...
1228285163894519192601457852416*x2*pi*x3+1228285163894519192601457852416*x2*pi*x4+...
1228285163894519192601457852416*x1*pi*x3+...
(-775950152915851682223433388725080334688732732710789356705546240*x3*x1*x2*pi*x4-...
16897265771045921223758890041058086220520378114626105114624000*x1*pi^2*x4*x2*x3+...
7241685330448251953039524303310608380223019191982616477696000*x2^2*pi^2*x4^2+...
8448632885522960611879445020529043110260189057313052557312000*x1*pi^2*x4^2*x2-...
4827790220298834635359682868873738920148679461321744318464000*x1^2*pi^2*x4*x3-...
14483370660896503906079048606621216760446038383965232955392000*x2^2*pi^2*x3*x4+...
8448632885522960611879445020529043110260189057313052557312000*x2*pi^2*x3^2*x1+...
7241685330448251953039524303310608380223019191982616477696000*x2^2*pi^2*x3^2-...
779741890488142549074641931811401211393656294625276146525143040*x3*x1^2*pi*x4-...
779741890488142549451056717955108233004107098497169948860743680*x3^2*x1*x2*pi+...
1555692043403994231674490106680188567692839831207959305566289920*x4^2*x1*x2*pi+...

Phong Nguyen

Structural Reliability_Makeup Problem

9

1555692043403994231298075320536481546082389027336065503230689280*x3*x2^2*pi*x4+...
100269203199468442259909210601902412590826076443469454441841148324*x3^2*x1^2+...
399124254938931626098556067158630038575772652424994550311433368497*x4^2*x1^2+...
395697153405116644263726859407719954223026314952595292554505684624*x3^2*x2^2+...
1575030795402264382586435519595857903740975375801713688692324159172*x4^2*x2^2+...
400099342658912552630490858778316650734268010539283991101585179884*x3*x1^2*x4+...
387416055042219638950286615584330611099178246227970242231882679952*x3^2*x1*x2+...
1541987404196006217152221930754029771503291352424995231110601756356*x4^2*x1*x2+...
1578904939673404900066249094594156085375458443565438657703918244784*x3*x2^2*x4+...
1545821047344481806722261812526463333008920294427829328267811034032*x3*x1*x4*x2+...
2413895110149417317679841434436869460074339730660872159232000*x1^2*pi^2*x3^2-...
783533628060433416302265261041429109709030660411656738680340480*x3^2*x1^2*pi+...
1563275518548575965376907192852830321102686955036932885205483520*x4^2*x1^2*pi+...
1563275518548575965753321978996537342713137758908826687541084160*x3^2*x2^2*pi-...
3118967561952570197051397299533018888795526786244892190771773440*x4^2*x2^2*pi+...
2413895110149417317679841434436869460074339730660872159232000*x1^2*pi^2*x4^2)^(1/2))/.
..
(-113701209351838406*x3^2-454249228073565621*x3*x4453693618739777618*x4^2+70368744177664*pi^2*x3^2-...
140737488355328*pi^2*x3*x4+70368744177664*pi^2*x4^2);
ggg=0.7*x5-omega^0.5;
g1=diff(ggg,x1);
g2=diff(ggg,x2);
g3=diff(ggg,x3);
g4=diff(ggg,x4);
g5=diff(ggg,x5);
beta_trial=3;
% guessed value of raliability index;
tolerance=10^-9;
% interation error.
itermax=5;
% the maximum number of iteration
x0=muy;
% Guessed values
er=1;
i=1;
while er > tolerance
x(:,i)=x0;
u(:,i)=(x(:,i)-muy)./xichma;
gg=double(subs(ggg,{x1,x2,x3,x4,x5},{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i)}));
g(1,i)=double(subs(g1,{x1,x2,x3,x4,x5},{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i)}));
g(2,i)=double(subs(g2,{x1,x2,x3,x4,x5},{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i)}));
g(3,i)=double(subs(g3,{x1,x2,x3,x4,x5},{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i)}));
g(4,i)=double(subs(g4,{x1,x2,x3,x4,x5},{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i)}));
g(5,i)=double(subs(g5,{x1,x2,x3,x4,x5},{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i)}));
grad(:,i)=g(:,i).*xichma;
%Gradients in new coordinate system
anpha(:,i)=double(grad(:,i)/sqrt(sum(grad(:,i).^2)));
u(:,i+1)=((grad(:,i)'*u(:,i)-gg)/(grad(:,i)'*grad(:,i)))*grad(:,i);
x0=u(:,i+1).*xichma + muy;
%neu design point in original coordinate
er=abs(double((subs(ggg,{x1,x2,x3,x4,x5},{x0(1),x0(2),x0(3),x0(4),x0(5)}))));
beta=sqrt(u(:,i+1)'*u(:,i+1));
betaf(i)=beta;
i=i+1;
if i>itermax, break, end
end
normcdf(-betaf);

Phong Nguyen

Structural Reliability_Makeup Problem

10

Part II: Lims Problem

Question 1:

f s 200 MPa

g X

1.57 wc db1
cf s
k s s d
1 2

The limit state function:

X 1 : wc N 0.35, 0.07 ; X 2 : k N 2.1e 5,3.15e 6 ; X 3 : N 1.25, 0.125 ;

X 4 : s1 N 80, 4 ; X 5 : s2 N 80, 4 ; X 6 : d c N 25,1.5 ; X 7 : db1 LN 40, 4 ; X 8 : c LN 1, 0.11 ;

At design point:

g 9.18e 8; 0.338547

0.261404109971

2.3035855e-005
1.306523983378

80.46987981569

X
80.46987981569

25.21090854592
38.87648944960

1.052146622074

Probability of failure:

-1.265655571841

0.6463031746649
0.4521918670245

0.1174699539235

U
0.1174699539235

0.1406056972818
-0.2357322107048

0.5183411454206

pf = (1.6111) =0.05358

Question 2:
Order of importance of random variables:

0.785578

-0.40115
-0.28067

-0.07291

-0.07291

-0.08727
0.146316

-0.32173

Phong Nguyen

Structural Reliability_Makeup Problem

11

wc

s1

s2

dc

d b1

Question 3: probability of failure for different values of

fs
pf

fs

100

150

200

250

300

0.000521
5

0.00824

0.05358

0.17649

0.367475

We can see that The probability of failure increases rapidly when f s increases.

Figure 1: probability of failure for different values of fs

Code Matlab:
%% Use New FOSM (Rackwitz-Fiessler algorithm) to calculate probability of failure.
% Limit state function: g=1.57*wc*sqrt(db1)/(k*beta*sqrt(s1*s2*dc))-c*fs;
% db1,c are lognormal, others are normal.
syms x1 x2 x3 x4 x5 x6 x7 x8 % wc,k,beta,s1,s2,dc,db1,c
format long
%% Input
muy=[0.35; 2.1e-5; 1.25; 80; 80; 25; 40; 1];
%Mean values
cov=[0.2; 0.15; 0.1; 0.05; 0.05; 0.06; 0.1; 0.11]; % coefficients of variantion.
xichma=cov.*muy;
% Standard deviation.
fs=200;

Phong Nguyen

Structural Reliability_Makeup Problem

12

%% LogNormal Parameter Transformation for db1 & c

shape7 = log( (muy(7)^2) / sqrt(muy(7)^2 + xichma(7)^2));
scale7 = sqrt(log((cov(7))^2 + 1));
shape8 = log( (muy(8)^2) / sqrt(muy(8)^2 + xichma(8)^2));
scale8 = sqrt(log((cov(8))^2 + 1));
%% Limit state function and its derivatives
ggg=1.57*x1*sqrt(x7)/(x2*x3*sqrt(x4*x5*x6))-x8*fs;
g1=diff(ggg,x1);
g2=diff(ggg,x2);
g3=diff(ggg,x3);
g4=diff(ggg,x4);
g5=diff(ggg,x5);
g6=diff(ggg,x6);
g7=diff(ggg,x7);
g8=diff(ggg,x8);
beta_trial=5;
tolerance=10^-9;
itermax=20;
x0=muy;
er=1;
i=1;

%
%
%
%

guessed value of raliability index;

interation error.
the maximum number of iteration
Guessed values

while er > tolerance

x(:,i)=x0;
% values in original coordinate system
%% Transform from Lognormal to Equivalent Normal for x7 & x8
xichma7=x(7,i)*scale7;
muy7=x(7,i)*(1-log(x(7,i))+shape7);
xichma8=x(8,i)*scale8;
muy8=x(8,i)*(1-log(x(8,i))+shape8);
muy(7)=muy7;
xichma(7)=xichma7;
muy(8)=muy8;
xichma(8)=xichma8;
u(:,i)=(x(:,i)-muy)./xichma;
% values in new coordinate system
%% check limit state function and its derivatives
gg=double(1.57*x(1,i)*sqrt(x(7,i))/(x(2,i)*x(3,i)*sqrt(x(4,i)*x(5,i)*x(6,i)))x(8,i)*fs);
g(1,i)=double(subs(g1,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
original coordinate system
g(2,i)=double(subs(g2,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
g(3,i)=double(subs(g3,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
g(4,i)=double(subs(g4,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
g(5,i)=double(subs(g5,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
g(6,i)=double(subs(g6,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));

%Gradients in

Phong Nguyen

Structural Reliability_Makeup Problem

13

g(7,i)=double(subs(g7,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
g(8,i)=double(subs(g8,{x1,x2,x3,x4,x5,x6,x7,x8},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i),x(7,i),x(8,i)}));
grad(:,i)=g(:,i).*xichma;
anpha(:,i)=grad(:,i)/sqrt(sum(grad(:,i).^2));
u(:,i+1)=((grad(:,i)'*u(:,i)-gg)/(grad(:,i)'*grad(:,i)))*grad(:,i);
x0=u(:,i+1).*xichma + muy;
er(i)=abs(1.57*x0(1)*sqrt(x0(7))/(x0(2)*x0(3)*sqrt(x0(4)*x0(5)*x0(6)))-x0(8)*fs);
beta=sqrt(u(:,i+1)'*u(:,i+1));
betaf(i)=beta;
i=i+1;
if er<tolerance, break,end
end
pf=normcdf(-betaf);

Part III: Morsys Problem

The limit state function:

g X

Tult
r K
tan 2

RF
4 2 K a

r z qrb qt

qb qLL

sin cos 2 Sv

qt : X 1 LN 400, 40 ; qLL : X 2 LN 2400, 240 ; r : X 3 LN / 6, 2 /10800

K
: X 4 N 1, 0.1 ; Tult : X 5 N 2800, 280 ; RF : X 6 N 5, 0.5
Ka

Probability of failure for 2 elevations z

Methods
pf

MCS
0.00382

Z=2 ft
FORM
0.00441

SORM
0.00391

MCS
0.17909
5

Z=6 ft
FORM
0.19476
5

SORM
0.18619

The results obtained from second-order reliability method (SORM) are closer to
Monte Carlo simulation than FORM.
At the design point (z=2 ft):

Phong Nguyen

Structural Reliability_Makeup Problem

14

400.010799

2399.99967
0.51677565
X

1.23451019
2799.71046

5.27444296

0.00961275

0.00320638
-0.1985362
U

0.74158634
-0.0173031

0.38812095

-0.0111715

-0.0037263
0.2307302

-0.8618397
0.0201089

-0.4510574

Order of importance of random variables:

qt

qLL

K / Ka

Tult

RF

Code:
%% Use New FORM (Rackwitz-Fiessler algorithm) and SORM (second-order) to calculate
probability of failure.
% Limit state function: g=(Tult/RF)-tan(45-phi/2)*(K/Ka)*[gama*z+qrb+qt+
(qb+qll)/pi*(anpha1+sin(anpha1)*cos(anpha1+2*beta1))]*sv;
% qll is exponential; qt,phi are lognormal; others are normal
syms x1 x2 x3 x4 x5 x6 % qt,qll,phi,K/Ka,Tult,RF
format long
%% Input
muy=[400; 2400; pi/6; 1; 2800; 5];
%Mean values
varr=[40; 240; pi^2/60/180; 0.1; 280; 0.5]; % coefficients of variantion.
xichma=sqrt(varr) ;
% Standard deviation.
cov=xichma./muy;
sv=8/12; b=2; ab=2; gama=110; qrb=300; qb=1600;
z=6;
beta1=atan(ab/z); anpha1=atan((b+ab)/z)-beta1;
%% LogNormal Parameter Transformation for qt, qll & phi
shape1 = log( (muy(1)^2) / sqrt(muy(1)^2 + xichma(1)^2));
scale1 = sqrt(log((cov(1))^2 + 1));
shape2 = log( (muy(2)^2) / sqrt(muy(2)^2 + xichma(2)^2));
scale2 = sqrt(log((cov(2))^2 + 1));
shape3 = log( (muy(3)^2) / sqrt(muy(3)^2 + xichma(3)^2));
scale3 = sqrt(log((cov(3))^2 + 1));
%% Limit state function and first derivatives
ggg=x5/x6-(tan(45*pi/180-x3/2))^2*x4*(gama*z+qrb+x1+
(qb+x2)/pi*(anpha1+sin(anpha1)*cos(anpha1+2*beta1)))*sv;
g1=diff(ggg,x1);
g2=diff(ggg,x2);
g3=diff(ggg,x3);
g4=diff(ggg,x4);
g5=diff(ggg,x5);
g6=diff(ggg,x6);

Phong Nguyen

Structural Reliability_Makeup Problem

15

%% second derivatives
g11=diff(g1,x1);
g12=diff(g1,x2);g13=diff(g1,x3);g14=diff(g1,x4);g15=diff(g1,x5);g16=diff(g1,x6);
g22=diff(g2,x2); g23=diff(g2,x3);g24=diff(g2,x4);g25=diff(g2,x5);g26=diff(g2,x6);
g33=diff(g3,x3); g34=diff(g3,x4); g35=diff(g3,x5); g36=diff(g3,x6);
g44=diff(g4,x4); g45=diff(g4,x5); g46=diff(g4,x6);
g55=diff(g5,x5); g56=diff(g5,x6);
g66=diff(g6,x6);
%% FORM
beta_trial=5;
tolerance=10^-7;
itermax=20;
x0=muy;
er=1;
i=1;

%
%
%
%

guessed value of raliability index;

interation error.
the maximum number of iteration
Guessed values

while er > tolerance

x(:,i)=x0;
% values in original coordinate system
%% Transform from Lognormal (x1, x2 & x3) to Equivalent Normal
xichma1=x(1,i)*scale1;
muy1=x(1,i)*(1-log(x(1,i))+shape1);
xichma2=x(2,i)*scale2;
muy2=x(2,i)*(1-log(x(2,i))+shape2);
xichma3=x(3,i)*scale3;
muy3=x(3,i)*(1-log(x(3,i))+shape3);
muy(1)=muy1;
xichma(1)=xichma1;
muy(2)=muy2;
xichma(2)=xichma2;
muy(3)=muy3;
xichma(3)=xichma3;
u(:,i)=(x(:,i)-muy)./xichma;
% values in new coordinate system
%% check limit state function and its derivatives
gg=double(subs(ggg,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
g(1,i)=double(subs(g1,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
g(2,i)=double(subs(g2,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
g(3,i)=double(subs(g3,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
g(4,i)=double(subs(g4,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
g(5,i)=double(subs(g5,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
g(6,i)=double(subs(g6,{x1,x2,x3,x4,x5,x6},
{x(1,i),x(2,i),x(3,i),x(4,i),x(5,i),x(6,i)}));
grad(:,i)=g(:,i).*xichma;
%Gradients in new coordinate system
anpha(:,i)=grad(:,i)/sqrt(sum(grad(:,i).^2));
anpha2=grad(:,i)/sqrt(sum(grad(:,i).^2)); % use form SORM
u(:,i+1)=((grad(:,i)'*u(:,i)-gg)/(grad(:,i)'*grad(:,i)))*grad(:,i);
x0=u(:,i+1).*xichma + muy;
%neu design point in original coordinate

Phong Nguyen

Structural Reliability_Makeup Problem

16

er(i)=abs(x0(5)/x0(6)-(tan(45*pi/180-x0(3)/2))^2*x0(4)*(gama*z+qrb+x0(1)+
(qb+x0(2))/pi*(anpha1+sin(anpha1)*cos(anpha1+2*beta1)))*sv);
beta=sqrt(u(:,i+1)'*u(:,i+1));
gradd=grad(:,i); %% use for SORM
betaf(i)=beta;
i=i+1;
if er<tolerance, break,end
end
%% Second-order reliability method (SORM);
%% formulate matrix R
r0=[1 0 0 0 0 0;0 1 0 0 0 0;0 0 1 0 0 0;0 0 0 1 0 0;0 0 0 0 1 0;anpha2'];
%% Gram-Schmidt Orthogonalization
for ii=5:-1:1
rr=zeros(1,6);
for jj=ii+1:6
rr=rr+(r0(jj,:)*r0(ii,:)')/(r0(jj,:)*r0(jj,:)')*r0(jj,:);
end
r0(ii,:)=r0(ii,:)-rr;
end
r=r0;
for j=1:6
r(j,:)=r(j,:)./sqrt(sum(r(j,:).^2));
end
%% Formulate matrix D
%% Second derivatives
d11=double(subs(g11,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(1)*xichma(1);
d12=double(subs(g12,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(1)*xichma(2);
d13=double(subs(g13,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(1)*xichma(3);
d14=double(subs(g14,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(1)*xichma(4);
d15=double(subs(g15,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(1)*xichma(5);
d16=double(subs(g16,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(1)*xichma(6);
d22=double(subs(g22,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(2)*xichma(2);
d23=double(subs(g23,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(2)*xichma(3);
d24=double(subs(g24,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(2)*xichma(4);
d25=double(subs(g25,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(2)*xichma(5);
d26=double(subs(g26,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(2)*xichma(6);
d33=double(subs(g33,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(3)*xichma(3);
d34=double(subs(g34,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(3)*xichma(4);

Phong Nguyen

Structural Reliability_Makeup Problem

17

d35=double(subs(g35,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(3)*xichma(5);
d36=double(subs(g36,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(3)*xichma(6);
d44=double(subs(g44,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(4)*xichma(4);
d45=double(subs(g45,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(4)*xichma(5);
d46=double(subs(g46,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(4)*xichma(6);
d55=double(subs(g55,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(5)*xichma(5);
d56=double(subs(g56,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(5)*xichma(6);
d66=double(subs(g66,{x1,x2,x3,x4,x5,x6},
{x0(1),x0(2),x0(3),x0(4),x0(5),x0(6)}))*xichma(6)*xichma(6);
D=[ d11
d12
d13
d14
d15
d16

d12
d22
d23
d24
d25
d26

d13
d23
d33
d34
d35
d36

d14
d24
d34
d44
d45
d46

d15
d25
d35
d45
d55
d56

d16;
d26;
d36;
d46;
d56;
d66];

A=1/(sqrt(sum(gradd.^2)))*(r*D*r');
A(6,:)=[];
A(:,6)=[];
k=eig(A);
kk=1;
for iii=1:5
kkk=(1+beta*k(iii))^-0.5;
kk=kk*kkk;
end
pf1=normcdf(-beta);
pf2=normcdf(-beta)*kk;

Phong Nguyen

Structural Reliability_Makeup Problem

18

Part IV: Nick Kratkiewiczs problem

The limit state function:
g1=F y CSR

PL
P L3
; g2=Dmax
;
4
24 ESH

Dmax =1; R=1.05 ; L=20 ft ; H=13; E=29000 ksi ;

F y =X 1 N ( 52,3 ) ; C=X 2 N ( 0.75,0 .0025 ) ; S= X 3 N ( 4,0.1 ) ; P= X 4 U ( 2.4,2 .6 ) ;

For Monte Carlo Simulation:

For AFORM:

pf =P ( g 1 0 g 2 0 ) =0.2065 ;

pf 1=P ( g1 0 )=0.1319; p f 1=P ( g2 0 ) =0.1109 ;

In case of uni-modal: 0.1319 p f 1 (10.1319 ) ( 10.1109 )=0.2282

For bi-modal:
1=1.11716 ; 1,14= [ 0.7641 ; 0.4265 ; 0.3176 ;0.3652 ]
2=1.22157 ; 2,3 4= [ 0.7506 ;0.6607 ]

Phong Nguyen

Structural Reliability_Makeup Problem

19

=0.31760.7506+ (0.3652 )(0.6607 )=0.4797 ;

So,

P ( A )= pf 1 (0.78145 )=0.02866 ; P ( B ) =p f 2 (0.78145 )=0.02409;

So, 0.19005=0.1319+ 0.11090.05275 p f 0.1319+0.11090.02866=0.21414

Code Matlab:
Monte Cartlo Simulation:
syms x1 x2 x3 x4 x5 x6 % Fy, C, S, P
format long
%% Input
muy=[52; 0.75; 4];
%Mean values
xichma=[3; 0.025; 0.1]; % Standard deviation..
r=1.05; L=20*12; D=1; H=13;E=29000;
%% Monte Carlo Simulation
Ns=1000;
Nset=25;
for i=1:Nset
pf=0;
for j=1:Ns
% uniform random number
u = rand(1,3);
x1 = muy(1) + xichma(1)*norminv(u(:,1),0,1);
x2 = muy(2) + xichma(2)*norminv(u(:,2),0,1);
x3 = muy(3) + xichma(3)*norminv(u(:,3),0,1);
aa=5;
x4=2.35+0.05*ceil(aa.*rand(1,1));
% choose randomly x4 from [ 2.4; 2.45;
2.5; 2.55; 2.6]
g1=double(x1*x2*x3*r-x4*L/4);
g2=double(D-x4*L^3/24/E/x3/H);
if g1>0 && g2>0
pff=0;
else
pff=1/Ns;
end
pf=pf+pff;
end
Pf(i)=pf;
end
pf=mean(Pf);
cov=std(Pf)/pf;

AFORM for function g1

syms x1 x2 x3 x4 betav % Fy, C, S, P
format long
%% Input
muy=[52; 0.75; 4; 2.5];
%Mean values

Phong Nguyen

Structural Reliability_Makeup Problem

20

xichma=[3; 0.025; 0.1; 0.0577]; % Standard deviation..

r=1.05; L=20*12; D=1; H=13;E=29000;
%% Limit state function and its derivatives
gg1=x1*x2*x3*r-x4*L/4;
g1=diff(gg1,x1);
g2=diff(gg1,x2);
g3=diff(gg1,x3);
g4=diff(gg1,x4);
beta_trial=5;
tolerance=10^-8;
itermax=20;
x0=muy;
er=1;
i=1;

%
%
%
%

guessed value of raliability index;

interation error.
the maximum number of iteration
Guessed values

while er > tolerance

beta(i)=beta_trial;
x(:,i)=x0;
%% Transform from Uniform to Equivalent Normal for x4
xichma4=normpdf(norminv((x(4,i)/0.2-2.4/0.2)))/5;
muy4=x(4,i)-norminv((x(4,i)/0.2-2.4/0.2))*xichma4;
xichma(4)=xichma4;
muy(4)=muy4;
g(1,i)=subs(g1,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
g(2,i)=subs(g2,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
g(3,i)=subs(g3,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
g(4,i)=subs(g4,{x1,x2,x3,x4},{x(1,i),x(2,i),x(3,i),x(4,i)});
grad(:,i)=g(:,i).*xichma;
%Gradients in new coordinate system
anpha(:,i)=double(grad(:,i)/sqrt(sum(grad(:,i).^2)));
xnew(:,i)=muy-betav*anpha(:,i).*xichma;
gbeta=xnew(1,i)*xnew(2,i)*xnew(3,i)*r-xnew(4,i)*L/4;
betar=(double(solve(gbeta,betav)));
%solve reliability index
x0=double(muy-min(betar(betar>0))*anpha(:,i).*xichma);
er=abs((min(betar(betar>0))-beta(i)));
% error of reliability index
beta_trial=min(betar(betar>0));
betaf(i)=min(betar(betar>0));
i=i+1;
if i>itermax, break, end

end
pf=normcdf(-betaf);