Forum: EXAKT

Great practical question. As a preliminary comment, before answering, I would like to make the point that building and managing predictive models is the last (and relatively easy part) of the reliability challenge in any organization, large or small.

The greater challenge is the management of the work order system as it relates to the RCM knowledge base. For purposes of analysis and continuous process improvement (CPI), work orders should be clear instances of RCM knowledge records (i.e. failure modes). That means, that when closing work orders we must reference the knowledge base, update it as necessary in the light of new information in the work order, and attribute to that instance an event type (usually one of PF, FF, or S).

Now back to your question. Let's assume the CMMS (whether it be SAP PM, Maximo, Ellipse, or whatever) operates within a living RCM process. That is to say, personnel invoke RCM thinking in their routine interactions surrounding work order planning, execution, and closure. That being the case, software will generate "samples" for analysis . A sample is a collection of life cycles of equipment, components, and failure modes. Reliability analysis (RA) requires data in the form of a sample (also called an Events table). Analysis always precedes improvement. Analysis and modeling proceed quickly when a sample is available. When the work order process can produce samples, model building and deployment can scale up to the largest operations.

Then what's the problem? It is that RCM thinking and the CMMS work order process are estranged from one another. Rather than referencing and updating the RCM knowledge base in order to compile instances of "failure modes", the CMMS uses failure "codes". Failure codes are often ambiguous or difficult to apply consistently to the varied maintenance situations encountered. Seldom does a compilation a failure code instances result in a "good" sample for analysis.

Our challenge, then, is the re-orientation of the work order process so that it links to the clear knowledge of a RCM analysis. The initial RCM knowledge base would have been developed, typically, in a RCM project way back when. After having used it to populate our PM schedules in the CMMS, we put it aside and forgot about it. It should be dusted off and blended into day-to-day CMMS activities. Thereafter, it should be built up dynamically, consistent with the principles of SAE JA 1011 (or any of the similar military standards).

The ultimate scaling up of reliability analysis can be found in Clockwork Solutions' SPAR PHM product which can apply EXAKT proportional hazard models to an entire system or plant represented by their reliability block diagrams. Using Monte Carlo Simulation, predictive what-if analyses may be run based on alternative scenarios of usage and maintenance in order to arrive at an optimal plan that will accomplish a given mission.



In the figure above, three scenarios are illustrated in the (conditional) cumulative failure probability versus working age graph. The green curve predicts reliability under a "No Maintenance" scenario. The blue curve, a partial maintenance that replaces the component most likely to fail during the mission. The red curve projects survival probability in a period that includes an extensive overhaul. The pie charts on the right guide the analyst towards an acceptable maintenance plan by pointing out the critical subsystems and their failure modes based on the current projected scenario.

Murray