RELIABILITY
AND
MAINTENANCE
RELIABILITY
is the measure of the ability of a
product, component, system or service
to perform its intended function under a
defined set of conditions at a specified
time.
RELIABILITY
Thus, reliability:
+ Is specified as a probability
+ Is time related
RELIABILITY
Example:
If we specify the reliability of a
component as 85% after 300 hours
of working, we mean that
+ There is a 85% probability that it
will perform as expected at that
point in time. By extension, the
probability of failure is 15%.
SPECIFYING RELIABILITY
The probability that the product,
component, system… will function
as expected:
1 At a specific point in time or
2 For a given interval of time
COMPUTING RELIABILITY
AT A POINT IN TIME
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN SERIES
Here, successful functioning means all
the components should work at the
defined time.
Even if one component fails, the
combined system is deemed to have
failed.
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN SERIES
Probability of system success
(reliability of the system)
= Product of the individual component
reliabilities
R (system) = R1 x R2 x R3 X …Rn
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN SERIES
Example
C1 C2 C3
R1 = 0.97 R2 = 0.94 R3 = 0.98
R (system) = 0.97 x 0.94 x 0.98 = 0.894
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN SERIES
C1 C2 C3
R1 = 0.97 R2 = 0.94 R3 = 0.98
R (system) = 0.97 x 0.94 x 0.98 = 0.894
Inference: The more components there are in series,
lower will be the system reliability.
(Simple systems are more reliable than
complex ones.)
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN PARALLEL
Here, successful functioning means at
least one component should work at the
defined time.
The combined system is deemed to have
failed only when all components fail
together at the defined time.
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN PARALLEL
Probability of system success
(reliability of the system)
R (system) = 1 – [(1 - R1) x (1 - R2)… x (1 – Rn)]
This is equivalent to
(1 – probability of total failure)
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN PARALLEL
C1 R1 = 0.97 Example:
R (system) =
1 – (1 – 0.97) x
(1 – 0.94) x
C2 R2 = 0.94 (1 – 0.98)
= 0.999964
C3 R3 = 0.98
COMPUTING RELIABILITY
CASE 1: COMPONENTS IN PARALLEL
C1
R1 = 0.97 Example:
R (system) = 0.999964
Inference:
C2 R2 = 0.94 Reliability rises as more
components are added in
parallel.
R3 = 0.98
C3
IMPROVING RELIABILITY
1 Enhance the reliability of each
component (R1, R2, R3…)
2 Reduce the number of components
3 Add components in parallel
(build redundancy)
Compute the reliability of this system.
Reliability = 0.98 * 0.99 * (1 – 0.9 * 0.19)
UNDERSTANDING RELIABILITY
OVER TIME INTERVALS
FAILURE OVER TIME
THE `BATHTUB CURVE’
FAILURE OVER TIME
FAILURE OVER TIME
Product withdrawal
Product launch
FAILURE OVER TIME
β<1 Indicates infant mortality
β=1 Indicates random failure
1<β<4 Indicates early wear-out failure
β>4 Indicates rapid wear-out failure
FAILURE OVER TIME
EXTENDING EQUIPMENT LIFE
COMPUTING RELIABILITY
OVER A PERIOD OF TIME
The Weibull & Exponential Distributions
THE WEIBULL DISTRIBUTION
Named after the Swedish physician,
Dr. Wallodi Weibull (1887-1979).
It is a continuous probability distribution that imitates
several other distributions under certain conditions.
THE WEIBULL DISTRIBUTION
To describe a specific Weibull distribution,
we need to specify three parameters:
1 Scale
2 Shape
3 Location
SCALE PARAMETER
Often denoted by α (alpha)
SHAPE PARAMETER
Often denoted by β (beta)
LOCATION PARAMETER
Often denoted by γ (gamma)
Importance of Shape Parameter
When β = 1, Weibull imitates Exponential.
When β = 2, Weibull imitates Rayleigh.
When β = 3.25, Weibull imitates Normal.
THE WEIBULL DISTRIBUTION
The main applications of Weibull distribution
are:
1 Study of life expectancy (life insurance)
2 Study of reliability (plant maintenance)
THE WEIBULL DISTRIBUTION
The main practical problem with applying the
Weibull distribution is data collection and
plotting across the three parameters.
Hence, the normal practice, especially in
reliability calculations, is to approximate the
Weibull distribution with the Exponential
distribution.
PROBLEM WITH USING THE EXPONENTIAL
DISTRIBUTION INSTEAD OF WEIBULL
Exponential distribution:
Probability of occurrence (λ: failure rate) is constant
over time.
Weibull distribution:
Probability of occurrence varies over time.
Reliability falls exponentially over time
Probability of failure before time T = 1 – [e ^ T/MTBF]
Probability of failure after time T = [e ^ T/MTBF]
Applying the Exponential Distribution
to Reliability on the Shop Floor
Note that the Exponential Distribution is positively skewed;
the tail is to the right.
SCHEDULING MAINTENANCE
TYPES OF MAINTENANCE SCHEDULES
1 Preventive maintenance (pro-active)
2 Repair or corrective (reactive)
3 Condition-based maintenance (in between 1 & 2)
TYPICAL MAINTENANCE EFFORT
COMPARING THE THREE
TYPE APPROACH FREQUENCY ASSESSMENT
Preventive Proactive Pre- Seldom
planning scheduled assessed
Corrective Run to Unscheduled Not
failure assessed till
problem is
evident
Condition- Repair When pre-set Perpetual
based before conditions monitoring
failure are met
COMPARING THE THREE
Build a dashboard
that senses system
deterioration and
predicts failure.
CONDITION-BASED MAINTENANCE
Ideally, the system should sense failure at the start.
WARNING SIGNALS
Note: As time progresses, cost to repair increases.
CBM: Perpetual factors to monitor
(partial list)
SUBJECTIVE OBJECTIVE
+ Visual look + Temperature
+ Sound + Pressure
+ Smell + Vibration
+ Feel (temperature, + Lubricating oil
surface) analysis
+ Power/fuel usage
PREVENTIVE MAINTENANCE IMPACT
PM impacts failures during useful life & wear out phases.