Problem-Solving Session 2: Reliability Analysis

Problem-Solving Session #2
Outline

• Reliability Analysis on Simple Performance Function

• First-Order Reliability Method (FORM)
• Monte Carlo Simulation (MCS)
• First-Order Expansion Method

2
 Problem-Solving Session 2

Reliability Analysis on Simple Performance Function

Consider the following performance function
80
G ( X1 , X 2 ) = 1 −
X12 + 8 X 2 + 5

where X1 and X2 each follow a normal distribution with the mean

4 and the standard deviation 0.6. Find the MPP using HL-RF.
Solution
The first iteration in the HL-RF method is detailed as follows:
Step 1: Set the number of iterations k = 0 and the initial values u = (0,0).
Step 2: Transform u(0) to x(0). x(0) = (4,4). Compute G(X) at x(0) as
80
G ( X1, X 2 ) = 1 − ≈ −0.5049
X12 + 8X 2 + 5

and the partial derivatives as:

∂G ∂G ∂X1 160 X1σ X1 ∂G ∂G ∂X 2
= = ≈ 0.1367, = ≈ 0.1367
∂U1 ∂X1 ∂U1 ( X 2 + 8 X + 5)2 ∂U 2 ∂X 2 ∂U 2
1 2
3
 Problem-Solving Session 2

Reliability Analysis on Simple Performance Function

Solution
Step 3: Update the search point at the current iteration as

( ) ∇UG u( )
0

u = u
(1)

( 0)
( ) ( )
• ∇UG u ( 0)
−G u ( 0) 
 2
∇ G ( u( ) )
0
U

= ( 0,0) • ( 0.1367,0.1367 ) − ( −0.5094) 

( 0.1367,0.1367) ≈ (1.8633,1.8633)
0.13672 + 0.13672

Iteration history for MPP search

∂G ∂G
Iteration U1 U2 β G( U)
∂U1 ∂U 2
0 0.0000 0.0000 0.0000 -0.5094 0.1367 0.1367
1 1.8633 1.8633 2.6351 -0.1090 0.0944 0.0738
2 2.7774 2.1707 3.5251 -0.0059 0.0860 0.0607
3 2.9225 2.0630 3.5772 0.0001 0.0863 0.0600
4 2.9366 2.0416 3.5766 0.0000 0.0864 0.0600
Optimum 2.9366 2.0416 3.5766 0.0000 0.0864 0.0600
4
 Problem-Solving Session 2

Reliability Analysis on Simple Performance Function

Graphical illustration of MPP search in the X-space

10 6
G(X) > 0
8
Failure region 5.8
G(X) ≤ 0
6 Safe region 5.6

4 5.4 x(3)
x(1)
2 Mean point x(0) 5.2
x(2)
G(X) = 0
LSF x(4)
0 5
0 2 4 6 8 10 5 5.2 5.4 5.6 5.8 6
X1 X1

5
 Problem-Solving Session 2

Reliability Analysis on Simple Performance Function

Graphical illustration of MPP search in the U-space

10
G(U) > 0
Failure region 3
5 G(U) ≤ 0
Safe region
2.5
0 u(3)
u(1)
Mean point u(0) 2
u(2)
G(U) = 0
-5 LSF u(4)
-5 0 5 10 2 2.5 3
U1 U1

6
 Problem-Solving Session 2

Reliability Analysis on Simple Performance Function

Reliability estimate by FORM, Φ 3.5766 0.99983
Reliability estimate by MCS
Scenario 1: Run MCS with 1,000 random samples for 4 times
, 1.00000, , 0.99900, , 1.00000, , 1.00000
Scenario 2: Run MCS with 100,000 random samples for 4 times
, 0.99981, , 0.99985, , 0.99983, , 0.99987

Failed sample
Safe sample

G(U) = 0
LSF

7
 Problem-Solving Session 2

Reliability Analysis on Simple Performance Function

Reliability estimate by FORM, Φ 3.5766 0.99983
Reliability estimate by MCS
Scenario 1: Run MCS with 1,000 random samples for 4 times
, 1.00000, , 0.99900, , 1.00000, , 1.00000
Scenario 2: Run MCS with 100,000 random samples for 4 times
, 0.99981, , 0.99985, , 0.99983, , 0.99987

Scenario 3: Run MCS with

10,000,000 random
Failed sample
samples for 4 times
Safe sample
, 0.99981,
, 0.99981,
, 0.99981,
, 0.99981

G(U) = 0
LSF

8
 Time-Independent Reliability Analysis

Reliability Analysis on Simple Performance Function

First-order expansion method
The first-order Taylor series expansion of G(X) at the mean point μX

(
∂G µ X1 , µ X 2 ) (
∂G µ X1 , µ X 2 )
(
G ( X) ≈ G µ X1 , µ X 2 +) ∂X1
( X1 − µ X1 + ) ∂X 2
(X 2 − µX2 )
We first compute the G function value at μX and the partial derivatives of G
with respect to X1 and X2 at μX

80
( )
G µ X1 , µ X 2 = 1 −
X12 + 8 X 2 + 5 X =µ
≈ −0.5094,
1 X1 , X 2 =µ X 2

(
∂G µ X1 , µ X 2 )= 80 ⋅ 2 X1
≈ 0.2278,
2
∂X1 (X 1
2
+ 8 X 2 + 5)
X1=µ X1 , X 2 =µ X 2

(
∂G µ X1 , µ X 2 )= 80 ⋅ 8
≈ 0.2278
2
∂X 2 (X 1
2
+ 8 X 2 + 5)
X1=µ X1 , X 2 =µ X 2 9
 Time-Independent Reliability Analysis

Reliability Analysis on Simple Performance Function

First-order expansion method
The first-order Taylor series expansion of G at μX can be expressed as
G ( X) ≈ −0.5094 + 0.2278( X1 − 4) + 0.2278( X 2 − 4)
= 0.2278X1 + 0.2278X 2 − 2.3318
= aT X + b
where a = [0.2278, 0.2278]T, b = ‒2.3318, and X = [X1, X2]T. Note that the
mean vector μX = [4, 4]T, and the covariance matrix ΣX = [0.62, 0; 0, 0.62].
We then compute the mean and variance (and standard deviation) of G

 4
µG = aTµX + b = [ 0.2278 0.2278]   − 2.3318 = −0.5094
 4
0.62 0  0.2278
σ = a ΣXa = [0.2278 0.2278] 
2
G
T
2  ≈ 0.0374
 0 0.6  0.2278
σ G = 0.0374 ≈ 0.1933
10
 Time-Independent Reliability Analysis

Example on Reliability Analysis of A Cantilever Beam

First-order expansion method
Finally, the reliability can be computed as

R = P ( G ≤ 0) Definition of reliability
 G − µG 0 − µG 
= P ≤ 
 σ G σ G 
 0 − ( −0.5094) 
= P Z ≤ = 2.6353 
 0.1933 
= Φ ( 2.6353) = 0.9958

The reliability estimate by the first-order expansion method contains a larger

error than that by FORM.

11
 Time-Independent Reliability Analysis

Example on Reliability Analysis of A Cantilever Beam

True LSF vs. LSFs approximated by FORM and first-order
expansion method

G(U) > 0
Failure region
G(U) ≤ 0
Safe region MPP by HL-RF

First-order expansion
method
G(U) = 0
Mean point (0,0) LSF FORM

12