Lecture 4: Definition of Reliability

Lecture #4
Outline

• Time-Independent Reliability
• Formal Definition
• Uncertainty Propagation and Probability Analysis

• Time-Dependent Reliability
• Formal Definition
• Failure Rate and Bathtub Curve

2
 Definition of Reliability

Formal Definition of Reliability

 The probability of a system or component to perform its required
function(s) under stated conditions (for a specified period of
time).

Time-Independent Reliability (Design)

 Typically used at the early design stage and does not consider
time-dependent degradation that occurs during system
operation.

Time-Dependent Reliability (Operation)

 Often used to represent the reliability of a system during
operation and takes into account degradation.

3
 Definition of Reliability

Time-Independent Reliability
 Definition: the probability that the system performance meets
the required design performance under various uncertainty
sources (e.g., material properties, geometric tolerances, and
loading conditions).
Performance function
R ( X) = P G ( X) ≤ 0 = 1 − P G ( X) > 0 (e.g., stress, strain, and
( ) ( ) power output)
where X = (X1, X2,…, XN)T models uncertainty sources.

 The uncertainty of X propagates and causes the uncertainty of G.

 Equating G to zero, i.e., G = 0, produces the limit-state function
(LSF), which separates the safe region ΩS = {x: G(X) ≤ 0} from the
failure region ΩF = {x: G(X) > 0}.

4
 Definition of Reliability

Time-Independent Reliability
Schematic of time-independent
 The basic idea is to compute reliability analysis in a 2-D case
the probability that X is inside

X2: operational
the safe region. G≤0
Safe region fX(x)
R ( X) = P ( G ( X) ≤ 0 )

factor
= ⋯ S f X ( x ) dx
Ω

where fX(x) denotes the joint

probability density function G>0
Failure
(PDF) of X, and the safe region region X1: manufacturing
ΩS = {x: G(x) ≤ 0}. tolerance

5
 Handling Uncertainty at Three Steps in Design

Model (PDF) Pf = Pr ( G > 0 )

Probability
Probability

Probability
Design 2

density
density

density
Data Design 1

Design variable X 0 G 0 G

Step 1: Uncertainty Step 2: Uncertainty Step 3: Design under

modeling analysis uncertainty

• Probability theory • Reliability analysis • Reliability-based

• Statistics • Robustness analysis design
• Other theories • Sensitivity analysis • Robust design

Design process

PDF: Probability density function 6

 Time-Independent Reliability

An Example
 Consider a performance function G as the difference between the
strength S of and load L on an engineered system
G = L−S
where S and L are random variables and follow normal
distributions with PDFs fS and fL.
Probability density

fS

fL

μL μS L,S
7
 Time-Independent Reliability

An Example
 A failure will occur when the load L is greater than the strength S,
or G (= L – S) > 0. Hence G > 0 indicates a failure.
 The probability of failure can therefore be defined by
Pf = Pr ( G > 0 )  R = 1 − Pf = Pr ( G ≤ 0 )

 What does the LSF look like in a 2-D plot?

L

G>0
Failure region
G=0
Limit-state function

0 S 8
 Time-Dependent Reliability

Some Basic Math

 Considers the system degradation over time and is defined based
on the failure-free operating time.
Time to failure PDF
+∞
RT ( t ) = P (T > t ) = 
t
fT ( t ) dt = 1 − FT ( t ) (e.g., exponential and
Weibull)

 Derivation of failure (or hazard) rate: Let h(t)∆t be the conditional

probability that the system will fail at some time t < T ≤ t + ∆t
given that it has not yet failed at T = t, expressed as
P{(T > t ) ∩ (T ≤ t + ∆t )}
h ( t ) ∆t = P{T ≤ t + ∆t | T > t} =
P{T > t}
P{t < T ≤ t + ∆t} fT ( t ) ∆t
= =
R(t ) R(t )
9
 Time-Dependent Reliability

Exponential Distribution
 Often used to model the time to
failure (TTF) distribution of a system
with a constant failure rate (λ)

fX(x)
 Probability density function (PDF)
λe−λ x , x≥0
f X ( x) = 
0, x<0 X

 Cumulative distribution function (CDF)

1 − e−λ x , x ≥ 0
FX ( x ) = 

FX(x)
0, x<0
Failure rate (exponential distribution)
fT ( t ) fT ( t ) λe−λx
FR ( t ) = = = −λ x
=λ X
R(t ) 1 − FT ( t ) e 10
 Time-Dependent Reliability

Weibull Distribution
 Often used to model the time to
failure (TTF) distribution of a system
with time-varying failure rate
Failure rate (Weibull)
 Probability density function (PDF)
λk ( λ x )k −1 e−( λ x) , x ≥ 0
k
fT ( t ) fT ( t )
f X ( x) =  FR ( t ) = =
R(t ) 1 − FT ( t )
0, x<0
k −1
= λk ( λt )
 Cumulative distribution function (CDF)
1 − e−( λ x) ,
k

x≥0 k = 1  Constant failure

FX ( x ) =  rate (reduces to
0, x<0 exponential distribution)

11
 Time-Dependent Reliability Analysis

An Example with Exponential Distribution

 Group discussion (4 min): The lifespan of a certain type of device
has an advertised failure rate of 0.01 per hour. The failure rate is
constant and the exponential distribution of TTF applies.
 Calculate the following: 1) the mean TTF (MTTF) and 2) the
probability that 200+ hours pass before a failure is observed.
 Solution
 The TTF distribution is (here X is TTF) fT (t) = λe−λt
 The MTTF can be computed as the expectation (mean) of T
+∞ +∞ +∞
MTTF =  t ⋅ fT (t )dt =  t ⋅ λ e− λt dt = −  tde −λt
0 0 0
+∞ 1
= −te− λt |0+∞ +  e− λt dt = 0 + = 100 hrs
0 λ
 The probability of TTF being at least 200 hours can be computed as
+∞ +∞ +∞
P ( t ≥ 200 ) =  fT (t ) dt =  λ e − λt dt = − e − λt = e −200 λ ≈ 0.1353
200 200 200

12
 Time-Dependent Reliability Analysis

An Example with Exponential Distribution

 Group discussion (4 min): The lifespan of a certain type of device
has an advertised failure rate of 0.01 per hour. The failure rate is
constant and the exponential distribution of TTF applies.
 Calculate the conditional probability that the device will operate
reliably for 200+ hours, given the device has already accumulated
400 hours of reliable operation.
 Solution
 The conditional probability can be computed as
P ( t ≥ 600 ) e −600 λ
P ( t ≥ 600 | t ≥ 400 ) = = −400 λ = e −200 λ ≈ 0.1353
P ( t ≥ 400 ) e

 Reliability is only a function of mission duration, not a function of

system age at the beginning of the mission (memoryless property
of exponential function).

13
 Time-Dependent Reliability

Bathtub Curve
 The failure rate of a large population of statistically independent
system units can often be represented by a bathtub curve
 Burn-in failures: “Infant mortality” failures of defective units.
 Constant failure rate: Random failures of non-defective units.
 Wear-out failures: Degradation (e.g., corrosion, fatigue and wear) of non-
defective units.

Decreasing Constant Increasing

Failure Rate Failure Rate Failure Rate

Early Observed
Failure Rate

“Infant Failure Rate

Mortality” Wear-Out
Failure Constant (Random)
Failures Failures

Time
14
 Time-Dependent Reliability Analysis

Two Main Approaches

 Classical approach: traditional reliability engineering
 To fit a probability distribution to TTF data (censored and uncensored) from
accelerated life testing (ALT), other lab testing, and/or field surveillance.
 Provides a population-wise, time-dependent reliability estimate that takes
the same value for the entire population of units.
 Emerging approach: prognostics and health management (PHM)
 Assesses the reliability of each individual unit under its actual use conditions
to determine the advent of a failure.
 Provides a unit-wise, time-dependent reliability estimate that is customized
for each individual unit

15
 Time-Dependent Reliability Analysis

Two Main Approaches

Population-wise (classical) Unit-wise (PHM)

Life PDF

Life PDF
True life

1 Time 1 Time
2 2
3 Unit- 3
Unit-specific
4 independent 4
Unit ID Unit ID

16