If β < 1 then the Reliability is improving, as in the failures comes slowly.
If β = 1 the Reliability is static, as in no improvement in failures.
If β > 1 the Reliability is deteriorating, as in the failures come faster.
Using the above formula, you can forecast the next failure.
First, reverse the above equation 
and you get 
The next failure will occur at 
Therefore, the next failure (failure number 7) will occur at 
= 431, which is 431 – 391 = 40 days into the future.
This contradicts the fact that the reliability of the equipment is improving. When continued for future failure forecasts, (such as, for t>8) the analysis for the same β and λ will show that the reliability is improving. This is because the forecasted failures are roughly within 10% of the actual cumulative times as observed from the error data. The graph might predict that the next failure will occur at 99 days but in reality, it will occur in 120 days. In short, CROW-AMSA says, "If you can predict it, it can be controlled". Therefore, it can be used for resource planning, develop strategies for preventing the next failure and to establish a predictive maintenance watch on the equipment to shut it down just before destructive events occur.
The regression method used to calculate the Crow-AMSAA is generally considered the best method, but only for small sample sizes of less than 5. Above that the IEC standard 61164 (IEC is International Electro technical Commission) method is resulting in far more accurate numbers. See chart below.
Calculate the Crow-AMSAA with the IEC method as follows:
|
Y |
X |
X Cumulative (Ti) |
T(N) / Ti |
LN( T(N) / Ti ) |
|---|---|---|---|---|
|
1 |
34 |
34 |
11.5 |
|
|
2 |
16 |
50 |
7.82 |
|
|
3 |
53 |
103 |
3.7961165 |
1.3339786 |
|
4 |
75 |
178 |
2.1966292 |
0.786924 |
|
5 |
93 |
271 |
1.4428044 |
0.3665887 |
|
6 |
120 |
|
|
|
Slope (β) = (N – 2 ) / (Σ LN( T(N) / Ti ) = 4 / 6.9865229 = 0.5725309
Intercept (λ) = N / ( T(N) β ) = 6 / (3910.5725309 ) = 0.1968113
-
Only calculate IEC values for N > 2
-
Depending on the values of Slope (β) and T(N), the intercept may be bigger than what can be stored in the database. If so, then store the maximum number allowed which is 99999999999999999999.9999999999 (=10 E20). It may be so big that the calculation runs into an overflow situation. In that case, leave the intercept blank.