Captured by Data Part 4

by Mr. Mather
Posted 8-1-06

Predictive Maintenance

As detailed in Figure 3 below, Predictive Maintenance (PTive) tasks are established to try to detect the warning signs that indicate the onset of failure, thus allowing for actions to be taken to avoid the failure. Yet there is also another aspect of PTive tasks that is often overlooked. That of the corrective, or Predicted (PTed), task once warning signs have been detected.
Immediately following the analysis, the information established at this point can be used for creating proactive whole-of-life costing models that are directly tied to performance and risk.

Whole-of-life cost of an asset, or component, subject to Predictive Maintenance tasks
= (Cost (PTive) x n) + Cost (PTed)
Where n represents the number of times the PTive task is likely to be executed. This also drives estimates of the time between installation and likely failure. It needs to be recognized that the corrective, or PTed, task is executed at a time less than end-of life. (Although small)
As time passes the amount of data that is collected on these tasks will grow, collected now in a responsible manner, and can also be used in statistical models regarding asset degradation and predictions of capital spend requirements. By the inclusion of these outputs of an RCM analysis, asset managers can use the results with increasing confidence as predictors of whole of life cost profiles, and end of life points.

Where Predictive Maintenance tasks cannot be applied, for whatever reason, the next two options on either side of the decision diagram are Preventive Maintenance tasks. These are tasks that are aimed at either restoring an assets resistance to failure (PRes), or replacing the asset at a time before the failures can occur. (PRep) Thus preventing failures. These tasks have limited use and are based on age, usage, or some other representation of time.

When applied correctly these tasks are part of the approach to maintenance that, by necessity, reduces the volume of failure data available for statistical analysis. However, with the component out of the operational environment, it can safely be tested to try to establish the extent of its remaining economically useful life.
Whole-of-life cost of an asset, or component, subject to Preventive Maintenance tasks
(Cost (PRes) - or - (Cost (PRep)
This is an additional task and one that would not be generated from the RCM analysis. Yet it represents another aspect of responsible data capture and is an important element of businesses where confidence in statistical life prediction, and whole of life costing models, are of importance.

As with predictive maintenance tasks there are actually two tasks that are being implemented here. First the detective (DTive) maintenance task, and second the detected (DTed) maintenance task. The result of this is the same as with the predictive maintenance tasks. That is, it provides further information about the likely failure rate, collected in a responsible manner, which can be used to inform decisions regarding optimizing the frequency of this task. 

Whole-of-life cost of an asset, or component, subject to Detective Maintenance tasks
= (Cost (DTive) x n) + Cost (DTed)
Where n represents the number of times the DTive task is likely to be executed. This also drives estimates of the time between installation and likely failure. It needs to be recognized that the corrective, or DTed, task is executed at a time greater than end-of life due to the characteristics of this task. As time passes the data collected can be used to inform decisions and whole-of-life models with increasing certainty.
This is particularly relevant for hidden failures, or hidden functions as they are sometimes called. When implementing the outcomes of an RCM analysis, some of the tasks are DTive tasks. That is, they are tasks put in place to detect if a failure has occurred. Often, the items being tested have not been tested for a long period of time, sometimes years. And often nobody knows if they are working or not!
So when establishing the initial DTive task frequencies, often the information used is not very certain and backed by only the experiences and memories of those involved in the exercise. Fortunately manufacturers do often have a good level of information regarding failure rates in these sorts of devices. But the result is till quite conservative and not tuned for the specific operational climate. Performing DTive tasks will immediately help the company to establish some baseline information regarding failure rates of the device.

The last policy option, aside from redesign and combinations of tasks, is that of run-to-failure. This option is for the acceptable, or low / negligible cost, failures detailed in figure 1.The EAM will allow these failures to be captured for analysis to inform whole of life cost models, spending forecasts, and to be used in reviewing maintenance policies when relevant. 

Along with the responsible data capture forced by these policy options, configuring and managing the EAM in line with RCM thinking will also allow visibility of exceptional failures.
Due to the way that RCM is, by necessity, carried out, there is the possibility that some failures may be missed. Modern methods of execution have expanded the original default method of team-based analyses to include expert analysis sources outside of the team, but there always remains the possibility that the analysis will miss a critical failure despite the best efforts of the analyst and those involved.
In these circumstances the data recorded in these exceptional failures will provide the impetus for the analyst to revisit the analysis to factor in this failure mode and to put in place a relevant management policy. It is not an area that is used for capturing data for statistical analysis and is, as the name suggests, the exception rather than the rule.
It can be seen that part of the role of the modern RCM Analyst is not only to minimize the volume of failure data that is collected for later analysis, but also to maximize the quality and usability of data that is captured via collection methods that support the principles of responsible asset stewardship. It can also be seen that advances in modern technology, combined with the growing needs of asset intensive companies, have enabled this information to be used in newer and more comprehensive ways than originally conceived of and correspondingly, not mentioned in previous work on RCM.
In particular it fuels the shift by the company away from the Static methods of life cycle costing, and towards the Proactive methods of whole-of-life costing. This is a step that enables companies to set up the data capture techniques and practices required to propel it towards the Stochastic, or probabilistic, model of whole of life costing.