M4. Bathtub Curve
M4. Bathtub Curve
M4. Bathtub Curve
Failure Rate
+Rate at which a component suffers faults
+Depends on age, ambient temperature, voltage or physical
shocks that it suffers, and technology
+Dependence on age or time produce a graph between failure
rate and time/age
+In the 1950s, a group known as AGREE (Advisory Group for
the Reliability of Electronic Equipment) discovered that the
failure rate of electronic equipment had a pattern similar to
the death rate of people in a closed system.
Specifically, they noted that the failure rate of electronic
components and systems follow the classical bathtub curve.
Failure Rate vs Time
Because of the
characteristic shape, this
is commonly known as the
Bathtub Curve.
Burn-in or
debugging
period
Useful life period
Old age
period
(or cycles)
There are 3 important phase
+Early Failures substandard components,
manufacturing faults.
+Random Failures this is the useful lifetime
of the item. Reliability is predictable in this
region.
+End-of-Life Failures items reaching the
end of their useful life. Also called the
wear-out period.
Early-life Period
+ Sebuah "kematian bayi" atau "fase pertumbuhan keandalan" atau
"fase awal kehidupan" ditandai dengan tingkat kegagalan yang
menurun(Phase 1).
+Kegagalan terjadi selama periode ini tidak acak terhadap
waktu melainkan hasil buruk dari komponen standar dan
kurangnya kontrol yang memadai dalam proses manufaktur.
+Parts fail at a high but decreasing rate.
+Young component high failure rate
- Good chance that some defective units slipped through
manufacturing quality control and were released
+Caused by undetected hardware/software defects that
are being fixed resulting in reliability growth
Steady-state Period
+Later after infant mortality - bad units weeded out remaining
units have a fairly constant failure rate
+A useful life period where electronics have a relatively constant
failure rate caused by randomly occurring defects and stresses
(Phase 2). This corresponds to a normal wear and tear period
where failures are caused by unexpected and sudden over stress
conditions. Most reliability analyses pertaining to electronic
systems are concerned with lowering the failure frequency (i.e.,
const shown in the Figure) during this period.
+Failure rate much lower than in early-life period
+Either constant (age independent) or slowly varying failure rate
+Failures caused by environmental shocks
+Arrival process of environmental shocks can be assumed to be a
Poisson process
+Hence time between two shocks has the exponential distribution
+A wear out period where the failure rate increases due
to critical parts wearing out (Phase 3). As they wear
out, it takes less stress to cause failure and the overall
system failure rate increases, accordingly failures do not
occur randomly in time. As component becomes very old,
aging effects cause the failure rate to rise again
+Failure rate increases rapidly with age
+Properly qualified electronic hardware do not exhibit wear
out failure during its intended service life (Motorola)
+Applicable for mechanical and other systems
+Weibull Failure Model can be used
Wear out Period
DFR IFR
Decreasing failure rate
Increasing fail. rate
h(t)
t
CFR
(useful life)
(burn-in-period)
(wear-out-phase)
Bathtub curve
DFR phase: Initial design, constant bug fixes
CFR phase: Normal operational phase
IFR phase: Aging behavior
Because of burn-in
failures and/or
inadequate quality
assurance
practices, the
failure rate is
initially high, but
gradually decreases
during the infant
period.
During the useful
life period, the
failure rate remains
constant, reflecting
randomly occurring
failures.
Later, the failure
rate begins to
increase because of
wear-out failures.
Useful Lifetime
t
m
R e
=
Reliability is predictable.
R = reliability.
t = time for which equipment is run.
m = MTBF
Note that R has no units. The prediction yields a
number <1.
Closer to 1 = greater reliability.
Examples
If an item of equipment has MTBF of 500hrs, then the reliability for 100hrs
operation is :-
100
500
e
= ) (
t
e t F
=1 ) (
Mean Time to Failure (MTTF)
+MTTF - expected value of the lifetime T
+Two ways of calculating MTTF
+First way:
+Second way:
+If the failure rate is a constant
dt t f t T E MTTF
}
= =
0
) ( ] [
} } }
= + = =
0 0 0
0
) ( ) ( | ) ( / ) ( dt t R dt t R t tR dt dt t dR t MTTF
t
e t R
= ) (
1
0 0
=
=
} }
dt e dt e t MTTF
t t
) ( / ) ( t f dt t dR =
Weibull Distribution
+Most calculations of reliability assume that a
module has a constant failure rate (or
equivalently - an exponential distribution for the
module lifetime T)
+Example - during the infant mortality and
wear-out phases of the bathtub curve
+Weibull distribution for the lifetime T can be
used instead
Weibull Distribution
+Frequently used to model fatigue failure, ball bearing failure
etc. (very long tails)
+Reliability:
+Weibull distribution is capable of modeling DFR ( < 1), CFR
( = 1) and IFR ( >1) behavior.
+ is called the shape parameter and is the scale parameter
( ) 0 > =
t e t R
t
o
Failure rate of the weibull distribution with
various values of o and = 1
5.0
1.0 2.0 3.0 4.0
Weibull distribution - Equation
+The Weibull distribution has two parameters,
and |
+The density function of the component lifetime T:
+The failure rate for the Weibull distribution is
(t) is decreasing with time for |<1, increasing
with time for |>1, constant for |=1,
appropriate for infant mortality, wearout and
middle phases, respectively
|
|
|
t
e t f t
=
1
) (
1
) (
=
|
| t t
Reliability and MTTF for Weibull
Distribution
+ Reliability for Weibull distribution is
+MTTF for Weibull distribution is
( I(x) is the Gamma function )
+The special case | = 1 is the exponential
distribution with a constant failure rate
|
t
e t R
= ) (
) ( ) / 1 (
/ 1
/
|
| | I = MTTF
Plotting Cumulative Failure on
Weibull graph paper
+Given a certain form for cumulative failture:
- F(t) = 1 exp [-(t/q)
b
],
+we can rearrange and take natural logarithms and get:
- Log log {1/[1-F(t)]} = b (log t log q)
+If we plot Log log {1/[1-F(t)]} vs log t,
the result is a straight line
+Special graph paper exists that does these
transformations
F
Time (t)
Weibull Cumulative Failure Probability
Plotted on Weibull Graph Paper
.99
b > 1
b < 1
b = 1
t = q
(
S
c
a
l
e
b
a
s
e
d
o
n
l
o
g
l
o
g
1
/
1
-
F
)
(log scale)
.63
.01
b < 1
Implies infant mortality
b = 1
Implies failures are random
An old part is as good as a new part
1 < b < 4
Occurs for:
- Low cycle fatigue
- Most bearing and gear failures
- Corrosion or Erosion
b > 4
Implies rapid wear out in old age
Weibull Interpretation
Gaussian (Normal) Distribution
+Bell shaped pdf
+Central Limit Theorem: mean of a large number of
mutually independent rvs (having arbitrary distributions)
starts following Normal distribution as n
+: mean, : std. deviation,
2
: variance (N(,
2
))
+ and completely describe the statistics. This is
significant in statistical estimation/signal
processing/communication theory etc.
Normal Distribution (contd.)
+N(0,1) is called normalized Guassian.
+N(0,1) is symmetric i.e.
- f(x)=f(-x)
- F(z) = 1-F(z).
+
+Failure rate h(t) follows IFR behavior.
- Hence, N( ) is suitable for modeling long-term
wear or aging related failure phenomena.