ADS-79D-HDBK Downloaded from https://www.everyspec.com I.4.3.3 CBM Thresholds and Treatment of CI Data The same set of CI values is used to demonstrate reliability at the green-yellow threshold, X G , and at the yellow-red threshold, X R . However, CI values from true yellow items are treated differently at X G and X R . At X G , the Weibull analysis treats CI values from true green items as right-censored data, while CI values from true yellow (optional maintenance) or true red (mandatory maintenance) items are treated as faulted data.
At X R , using the same data, the Weibull analysis treats CI values from true green or yellow items as right-censored data, while only CI values from true red items are treated as faulted data. (Censored data exists when the value of an observation is only partially known. A right- censored CI value is the CI value for an item that has yet to incur a fault. We do not know at what CI value a fault will occur; we only know the most recent healthy CI value.) For either threshold, if a Weibull distribution is found to provide an acceptable fit, then the value of the lower confidence bound at 90% reliability can be read directly from the Weibull plot. This value is the threshold.
If the threshold is too low, an unacceptable number of false positives will occur. To remedy this, we can select a higher CI value as a potential threshold, setting it equal to X 0 in Equation 12 . Once a CI value to test to is chosen ( x ), we can use Equation 12 to determine the minimum sample size required to demonstrate 90% reliability at the potential threshold. Many monitored components have yet to experience a true yellow or red item. When there are only CI values from unconfirmed green items, a Weibayes analysis may provide useful insight into the selection of the green threshold. A Weibayes analysis is the same as a Weibull analysis, but with an assumed β slope parameter.
A known β from a similar CI, or a known β from the same CI for a similar component, may provide a reasonable estimate of β for our component/CI being analyzed. However, if no basis for an estimate exists, assume β to be equal to 1.1. When Weibayes analysis is applied to reliability demonstration and minimum sample size calculations, assuming β = 1.1 is considered best practice. Using this value for β acknowledges that a positive correlation should exist between CI value and probability of a fault, while at the same time provides conservatism to the calculation of the minimum sample size required.
To perform a Weibayes analysis at the green threshold, there will only be right-censored data (the CI values from unconfirmed green items) and our assumed β . The value of the lower confidence bound at 90% reliability can be read directly from the Weibayes plot. As green items in the fleet continue to age, it is expected this lower confidence bound will increase. While it is ill-advised to establish the green threshold based solely on Weibayes results, this technique does provide useful information about the fleet of healthy items. Finally, nothing can be ascertained about the red threshold with only green items.
However, if green and yellow items exist, then we can apply Weibayes in a similar manner to provide useful information about the fleet as items begin to wear to the point where faults are present. I.4.4 Case Study Using Sample Size Method 3: Analysis of the Input Data Algorithm 1 (DA1) CI for the Apache Nose Gearbox (NGB) The Apache NGB is not a candidate for on 169