Introduction To Basic Reliability Statistics
Introduction To Basic Reliability Statistics
Introduction To Basic Reliability Statistics
Objectives
• Arithmetic Mean
• Standard Deviation
• Correlation Coefficient
• Estimating MTBF
– Type I Censoring
– Type II Censoring
• Exponential Distribution
• Reliability Predictions
• Weibull Curves and Intro to Weibull Analysis
• Basic System Reliability
– Series System
– Active Parallel Systems
Arithmetic Mean
Arithmetic Mean
Example
N
1 1
x = ∑x i x = (24 + 37 + 16 + 21 )
N i =1 4
x1 = 24 1
x 2 = 37 x = (98)
4
x 3 = 16
x 4 = 21
98
x=
4
1 4
x = ∑ xi
4 i=1 x = 24.5
Copyright 2006 Allied Reliability, Inc.
Introduction to Basic Reliability Statistics
Standard Deviation
Standard Deviation
Example
x 1 = 24
x 2 = 37
1 N 1 N
x 3 = 16 σ= ∑ i ( x − x )2
σ= ∑ i ( x − x )2
x 4 = 21 N i =1 N − 1 i =1
For known population size Estimate for unknown population size
1
σ = [( 24 − 24 . 5 ) 2 + ( 37 − 24 . 5 ) 2 + (16 − 24 . 5 ) 2 + ( 21 − 24 . 5 ) 2 ]
4
σ = 7.76208
Correlation Coefficient
• Is the likelihood that 2 sets of numbers are related
– The closer the correlation value gets to 1.0, the more linear the
relationship between the 2 sets of numbers.
• It is based on calculations of slope (m), y-intercept (b)
and correlation (r)
x y
1 2.6 6
2.3 2.8
3.1 3.1 5
4.8 4.7
5.6 5.1 4
6.3 5.3
3
Slope 0.584
Y-Intercept 1.684 2
Correlation 0.974
1
0
0 2 4 6 8
Correlation Coefficient
• Using the X & Y values, there is a non-graphical method for
calculating slope (m), y intercept (b) and correlation (r).
n Σ ( xy ) − Σ x Σ y
m =
nΣ ( x 2 ) − (Σ x ) 2
Σy − mΣx
b=
n
nΣ( xy ) − ΣxΣy
r=
[nΣ( x ) − (Σx) ][nΣ( y ) − (Σy) ]
2 2 2 2
Correlation Coefficient
• Why do I need this? How can I use it?
• Other examples:
– For a pump, are motor amps and gallons per minute
perfectly linear?
Equipment on PM (%)
Source: 1997 Benchmarking Study in Chemical Processing industry, John Schultz to be featured in
Ron Moore’s new book What Tool? When? Selecting the Right Manufacturing Improvement Strategies and Tools
Estimating MTBF
Type I Censoring
• a.k.a. Time/Cycle Truncated Censoring
• Test is halted at a given number of hours.
• Failures during the test are immediately repaired and the
test continues
nt
Θ̂ =
r
Where: Θ̂ = estimate of MTBF
n = number of items on test
t = total test time per unit
r = # of failures occurring during the test
Estimating MTBF
Type II Censoring
• a.k.a. Failure Truncated Censoring
• Test is halted at a given number of failures
• Failures during the test are immediately repaired and the
test continues
r
∑ y + (n − r ) y
i r
ˆ =
Θ i =1
r
Θ̂ = estimate of MTBF
yi = time to failure ith item
yr = time to failure of the unit at which time is truncated
n = Total number of assets in test
r = Total number of failures
Reliability Predictions
• If I know a little bit about the MTBF for a
particular asset…
Reliability Predictions
• Q is the probability of failure.
• Q=1–R
Exponential Distribution
1.2
0.8
Reliability
0.6
R(t)
0.4
0.2
t Time
Reliability Predictions
The reliability for a given time (t) during the random failure period
can be calculated with the formula:
− λt
R( t ) = e
Where:
e = base of the natural logarithms which is 2.718281828…
λ = failure rate (1/MTBF)
t = time
1+ 1 + 1 + 1 + 1 + ...
1! 2! 3! 4!
1 1.00000000000000
+ 1/1! 1.00000000000000
+ 1/2! 0.50000000000000
+ 1/3! 0.16666666666667
+ 1/4! 0.04166666666667
1 1 1 1 + 1/5! 0.00833333333333
1+ + + + + ...
1 1⋅ 2 1⋅ 2 ⋅ 3 1⋅ 2 ⋅ 3 ⋅ 4 + 1/6! 0.00138888888889
+ 1/7! 0.00019841269841
+ 1/8! 0.00002480158730
+ 1/9! 0.00000275573192
+ 1/10! 0.00000027557319
e 2.71828180114638
Reliability Predictions
Example
λ = 0.00025 or 1/4,000
t = 1,500
e − λ t = e − ( 0 . 00025 )( 1 , 500 )
= e − 0 .375 = 0 . 68728
Reliability Predictions
If the reliability for a given time (t) during the random failure period
can be calculated with the formula:
− λt
R(t ) = e
Then what is the equation when I am not in the random failure period?
What if I in the infant mortality period or wear-out period?
Weibull Shapes
Individual Curves
Initial Break-in period
Bathtub Pattern D = 7%
Pattern A = 4%
Time Time
Random
Pattern E = 14%
Wear out
Pattern B = 2%
Fatigue
Pattern C = 5% Infant Mortality
Pattern F = 68%
Weibull Analysis
Failure Curve
β
β <1 β =1 β >1 t
−
η
R(t ) = e
Failure Rate
γ1 γ2 γ3
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time
System Reliability
• Rarely do assets work alone
• Typically they are a part of a system
• Systems can many different configurations
– Series
– Active Parallel
– “Hot” Standby Parallel
– “Warm” Standby Parallel
– “Cold” Standby Parallel
• Reliability calculations for each of these is
slightly different
Rs = R1 x R2 x R3
Rs = 0.93 x 0.91 x 0.80 = 0.677 or 67.7%
− Σλiti
Rs ( t ) = e
− Σ ( 0.0001+ 0.00005+ 0.00001)(1500 )
Rs (1500 ) = e
− ( 0.00016 )(1500 )
Rs (1500 ) = e
Rs (1500 ) = e − ( 0.24 )
Rs (1500) = 78.6%
Rs (1500) = 0.7866
Copyright 2006 Allied Reliability, Inc.
Introduction to Basic Reliability Statistics
Rs = R1 + R2 – R1R2
Rs = 0.93 + 0.80 – 0.93 x 0.80 = 0.986 or 98.6%
0.0001/hr
I O
0.0004/hr
− λ1t − λ2t − ( λ1 + λ2 ) t
R(t ) = e +e −e
− ( 0.0001)(1500 ) − ( 0.0004 )(1500 ) − ( 0.0001+ 0.0004 )(1500 )
R(1500 ) = e +e −e
R(1500) = 0.9372
R(1500) = 93.72%
Copyright 2006 Allied Reliability, Inc.
Introduction to Basic Reliability Statistics
Questions?
Thanks!