Combined analysis for early predictive maintenance

In an article in Machinery Lubrication Using ‘Unscheduled’ Oil Analysis for Early Predictive Maintenance author Jim Fitch expounds from the premise that we may advance potential failure detection by conducting other tests in addition to the primary PdM method. 

Figure 1. Early Predictive Maintenance P-F Interval Scheme (Source LivingReliability Inc. June 2103)

Many RCM consultants discuss the  PF interval. But they might not be aware of the concept’s practical limitations. Nowlan and Heap who discovered the concept (although it was Moubray who coined the term) said that it (the PF interval) is a first approximation that needs to be tested and improved based on observed data. This is accomplished through a process called “Living RCM”. Here is an article discusses the PF interval:

http://www.livingreliability.com/en/posts/the-elusive-p-f-interval/

A concept in the above article concerns the condition versus age graph (i.e. the PF curve that you drew). It is also known as the survival graph. Most RCM consultants would be surprised to learn that it can be generated from real data in real time. That is the purpose of “Living RCM”.

Your survival graph is one of the forms of the age versus reliability relationship. Another form is the probability density graph. A particularly useful form is the Conditional Probability Density graph that I showed you from which you can get the remaining useful life estimate. Rather than starting at age 0 the conditional graphs start from the current moment. This is the moment of import to the maintenance engineer who must make a decision based on current asset condition as reflected by the latest CBM, age, and other relevant data. Your survival graph is 1 minus the integral of my Conditional Probability Density graph. This relationship is described in this article.

http://www.livingreliability.com/en/posts/real-meaning-of-the-six-rcm-curves/

The process I’m proposing is not utopia. It is achievable and its results are measurable. In order to get the actual graph from which the RUL is calculated (as well as the confidence intervals) we need only rectify a basic misunderstanding about how the EAM should acquire work order observations. It must capture the failure mode (usually the part) that failed or was replaced preventively. And it shoud record  whether the part ended its life cycle by failure or by suspension. This work order data constitutes essential “age data” needed for “reliability analysis”. When we combine age data with the CBM data we get your PF survival curve.