Pages

Monday, April 15, 2013

Failure curves and P-F intervals linked and explained: Tying the two most important reliability engineering curves together to generate a better picture of failure

During the early development of what would become Reliability Centered Maintenance, Nowlan and Heap gave us six failure curves to the left. When folks first see that sixty eight percent fall into the infant mortality curve then the doubt fairy tends to show up. "Sixty eight percent of the failures in my facility are not instant or early on start up." With this thought they then discount the incredibly important failure mode data provided to us from these studies. What they are missing is the connection to the P-F curve below.
So if the failure curves show the probability of defects introduced over time based on an individual failure mode then the P-F shows the resistance to failure over time once the failure defect has occurred. Nolan and Heap did not say that sixty-eight percent of your assets will catastrophically fail on or near startup they said that sixty-eight percent will have a defect introduced that will then travel down the curve of the P-F becoming more prone to functional and catastrophic failure. This trip down the curve may take 5 days, 5 weeks, or 5 years depending on the failure mode and operating context. This means they don’t fail instantly, but they do fail prematurely because of the defects introduced during or shortly thereafter maintenance activities. So think of the six failure curves as the probability of introduction of a defect and the P-F as the path of that defect to functional and catastrophic failure. I hope this helps you send the doubt fairy packing and you can begin to better understand both curves and the additional knowledge they can provide.

6 comments:

  1. Shon, regarding the 68% of failures that fall into the infant mortality distribution it is important to remind people that these failures are not always instantaneous. As an example, a bearing if not properly aligned, loaded or lubricated can fit the infant mortality distribution, it will fail prematurely and by the way it should not fit this conditional probability of failure distirbution.

    ReplyDelete
  2. I think you have made a valuable distinction that most people don't understand. That two types of curves apply to different things and cannot be compared directly. You line " So think of the six failure curves as the probability of introduction of a defect and the P-F as the path of that defect to functional and catastrophic failure." tells it all.

    ReplyDelete
  3. Good observation. I have found that it also helps to remember that the 6 failure patterns pertain to the probability of failure of the total population of assets under evaluation, while the P-F Curve describes the deterioration associated to the failure of a single asset.

    ReplyDelete
  4. good comments, the message is the age of equipments/components just represent 11% of probability of failure, the 68 % is Infant Mortality due in most of the cases by lack of trainning, lack of supervision, wrong procedures, wrong parts or parts with defects, etc.

    ReplyDelete
  5. Remember that the 68% number that Nowlan and Heap were talking about applied to the aircraft they were analyzing. They did not go so far to say that YOUR equipment will fail with the same percentages.

    ReplyDelete
  6. Based on data we have collected and some that is posted over at http://rcmblitzblog.com/ the 68% is closer than you think with most industries of low reliability maturity.

    ReplyDelete