Global Facilitator & Consultant
Living RCM Practitioner
The following is a quote from a well-known consultant:
"The biggest problem we have is lack of data… The data we do have in most cases is subject to different conditions as well. So if I have managed to collect 10 failures of a pump in 15 years, I am unlikely to be able to determine anything with much confidence. That is my point. Now if I have hundreds of components – let’s say light globes that turn on and off on the same cycle, then I have lots of data and can use it – trouble is we usually have one of everything and data is so scarce we can’t use it. Nothing at all to do with the method or the assumptions applied."
The foregoing argument is often made by RCM consultants responding to the general lack of success in achieving verifiable reliability improvement from data.
In the table is some age data collected over about 3 years in a very small fleet of four haul trucks. An optimal CBM decision model can indeed be drawn from this relatively scant amount of data. (Not shown is the CBM data prior to and concurrent with these events.) Most fleets contain 20 to 80 trucks and more. Hence abundant good age data can be acquired.  This table refers to transmissions that failed (EF = ending by failure) and transmissions that were rebuilt because their overhaul interval timed out (ES = ending by suspension). The transmissions whose lives were suspended, according to the technician’s observation. still had an indefinite amount of life left in them. Reliability analysis handles the uncertainty (of when the suspended transmissions would have failed) properly. The events labeled “OC” are oil changes.
Now let’s look at the optimal maintenance decision model and its cost-benefit analysis drawn from this meager data set:
The Proportional Hazard Model result table to the right tells you that the extended Weibull shape parameter is “Significant” (indicated in the third column as “Y” for yes). In other words, the item’s age is a significant risk factor. The results also tell you that Iron and Lead are statistically correlated with failure. Furthermore, the degree or strength of the correlation is indicated by the parameter “Estimate” in the second column. Various statistical test results in the other columns support the reliability model’s findings.
Next, we add a predictive model built from condition data transitions occurring over the time period. Finally, we add the relevant business data, specifically the cost of an unplanned failure relative to the cost of overhauling the transmission preventively. Then the model’s expected performance is given in the following table:
Summary of Cost Analysis
The above analysis tells you that, if you adhere to the optimal CBM policy (shown visually below as a red/green/yellow graph) by intervening when the transmission is in the red (potential failure) state, the savings related to that failure mode will be $1.50 per hour, 75% better than a run-to-failure or a poorly executed non-optimized CBM program.
While it is arguable that the “biggest problem” for reliability analysis is the lack of good data this shortcoming is not intrinsic to the nature of failure, nor is it an insurmountable problem. The single obstacle to data-driven decisions in maintenance is a lack of use of the EAM system’s capability to track, with accuracy, a part’s working-age so that it is known at failure or suspension. The Living RCM method for work order completion taps this vital EAM data capability and generates samples of life histories useable for reliability analysis, optimal decision modeling, and remaining useful life (RUL) prediction.