QM ZG528 L-7 Reliability Engineering: BITS Pilani
QM ZG528 L-7 Reliability Engineering: BITS Pilani
QM ZG528 L-7 Reliability Engineering: BITS Pilani
Reliability Engineering
BITS Pilani Gajanand Gupta
Department of Mechanical Engineering
Pilani Campus
1
BITS Pilani
Pilani Campus
State-Dependent Systems
Chapter – 6
RL 2.6-2.8
Load-Sharing System
3
BITS Pilani, Pilani Campus
Load-Sharing System
... (6.13)
... (6.14)
... (6.15)
5
BITS Pilani, Pilani Campus
Load-Sharing System
... (6.16)
... (6.17)
... (6.18)
6
BITS Pilani, Pilani Campus
7
BITS Pilani, Pilani Campus
Load-Sharing System
• If we let and
... (6.19)
... (6.20)
8
BITS Pilani, Pilani Campus
Example
Q.
9
BITS Pilani, Pilani Campus
Example
Solution.
From Eq.(6.20),
1
MTTF = + 0.02 [ 1 - 1 ] = 60 days
0.02 0.08 0.10 0.02
10
BITS Pilani, Pilani Campus
Standby Systems
• Standby systems are generally much more reliable than
an active redundant system.
• The backup unit may or may not have the same failure
rate as that of the online system.
12
BITS Pilani, Pilani Campus
Standby Systems
• System of equations
... (6.21)
... (6.22)
... (6.23)
13
BITS Pilani, Pilani Campus
Standby Systems
• Solutions
... (6.24)
... (6.25)
... (6.26)
... (6.27)
14
BITS Pilani, Pilani Campus
Standby Systems
...(6.28)
...(6.29)
...(6.30)
15
BITS Pilani, Pilani Campus
Example
Q.
16
BITS Pilani, Pilani Campus
BITS Pilani
Pilani Campus
• In general,
α(x) = f (x1,x2,….xk)
where,
α- distribution parameter
x = (x1,x2,….xk) and xi = the ith covariate.
20
BITS Pilani, Pilani Campus
Proportional Hazards Models
• Models having the property that individual component
hazard rate functions are proportional to each other are
referred to as proportional hazards models.
21
BITS Pilani, Pilani Campus
Exponential Case
• For the constant failure rate model, the simplest covariate
model is given by,
k
λ(x) = a x
i 0
i i ... (7.1)
where,
ai = unknown parameters to be determined.
x0 = 1
xi = transformed variables (e.g., squares and
reciprocals), thus allowing, for example,
polynomials to be used.
22
BITS Pilani, Pilani Campus
Exponential Case
23
BITS Pilani, Pilani Campus
Exponential Case
• A popular form of the multiplicative model is obtained by
letting
... (7.2)
24
BITS Pilani, Pilani Campus
Example
25
BITS Pilani, Pilani Campus
Weibull Case
• Assume that only the characteristic lifetime depends on the
covariates, the shape parameter doesn’t.
• For the multiplicative model, let
... (7.3)
• Then,
and
... (7.4)
26
BITS Pilani, Pilani Campus
Weibull Case
... (7.5)
27
BITS Pilani, Pilani Campus
Weibull Case
• Based on Eq. (7.5) a general form of hazard rate function can
be:
... (7.6)
with
where,
λ0(t) – baseline hazard rate function when g(x) =1.
e.g.,
the exponential baseline failure rate in Eq.(7.2) is λ0(t) = e a0.
28
BITS Pilani, Pilani Campus
Example
29
BITS Pilani, Pilani Campus
Physics of Failure Models
• Limitations of statistical models
Occurrence of failures are treated as random events.
Lack of knowledge of the physical processes resulting in a
failure.
Reliability estimates are very little concerning an individual
component or failure occurrence.
Reliable reliability predictions can be made only over a
large number of failures.
e.g., For an exponential distribution, the time to failure of a
single occurrence can be any t ≥ 0. The exponential
pattern can be seen only over a large number of failures.
It doesn’t consider the effect of individual stresses and
operating conditions on components.
30
BITS Pilani, Pilani Campus
Physics of Failure Models
• Physics of failure models is an alternative approach.
Mathematically derived deterministic models based on
knowledge of the failure mechanisms and the root causes
of failures
Failure is not viewed as a stochastic event.
31
BITS Pilani, Pilani Campus
Physics of Failure Models
Failure time is estimated based on the stresses, material
properties, geometry, environmental conditions and
conditions of use.
32
BITS Pilani, Pilani Campus
Physics of Failure Models
• Disadvantages:
Models are very specific to the failure mechanism and
failure site.
33
BITS Pilani, Pilani Campus
Physics of Failure Models
• General steps for developing physical failure models:
1. Identify failure sites and mechanisms.
2. Construct mathematical models.
3. Estimate reliability for a given operating and
environmental profile and for given component
characteristics.
4. Determine dominant service life.
5. Redesign to increase service (design) life.
34
BITS Pilani, Pilani Campus
Physics of Failure Models
36
BITS Pilani, Pilani Campus
Example
37
BITS Pilani, Pilani Campus
Example
38
BITS Pilani, Pilani Campus
Example
39
BITS Pilani, Pilani Campus
Example
40
BITS Pilani, Pilani Campus
Covariate and Physics of Failure Models
Covariate Models Physics-of-Failure Models
The parameters of failure Treat the time to failure as
distributions are determined by deterministic.
the explanatory variables.
Not physical models, and do not Capture the relevant casual
necessarily show cause and effect. variables and their relationships
among one another.
Normally include physical
Do not include physical constants. constants like acceleration due to
gravity or Boltzmann’s constant.
The functional form is strongly
Simple functional forms are determined by the physical
assumed. processes generating the failures.
41
BITS Pilani, Pilani Campus
Thank You!
42
BITS Pilani, Pilani Campus