Complexity in ISFA (in-service fluid analysis): Part XXXI

Jack Poley | TLT On Condition Monitoring March 2017

Holistic condition monitoring in the 21st Century: Part III
 

LAST COLUMN I INDICATED SOME OF THE HURDLES involved in implementing holistic condition monitoring (CM). Two of them are particularly important to consider and understand:

Data proliferation issue: There will be a mountain of data types with the advent of holistic CM and the varied techniques and disciplines utilized. Powerful intelligent agents (IAs) will be needed.
o Meaning: Manual evaluation would be almost impossible and add unnecessary risk of missing subtle patterns.
Timing issue: Combining episodic (sensor) data with bench testing (static) data necessitates assessing the indications provided by sensors where the data are always in real time, but the last bench testing occurred prior to the most current sensors.
o Meaning: The further in the past that the last bench tests (consisting of data not available via sensors) are, the more risky conclusions can be.

Data proliferation can be daunting, but it need not be suffocating or overwhelming. Simple, available calculations and/or determinations to using the data effectively, while minimizing the amount actually gathered for evaluation (and not necessarily permanently stored once information is extracted), include:

Mean: Knowing the average value sets a reference basis.
Data bandwidth: Excursions, both high and low, are of keen interest because they can be, and often are, evidence of potentially traumatic events. At the very least they announce a change in environmental or operating conditions. These moments are where sensors provide their payoffs.
o Excursion duration is key.
o Repeating patterns are key and may be indicative of harmless operational variance (e.g., duty cycles) rather than trauma.
o Feedback (from work order-generated diagnostics) is vital to identifying and differentiating actual trauma from routine (expected) operational variance. Since good, consistent feedback reporting has never been a hallmark of CM, at least for ISFA, the sensor era has a chance to help mitigate that neglect. This isn’t simply a statement about historical negligence—maintenance systems cannot be left wanting for validation in today’s monetized world. This is the adjudicating information that justifies CM by providing a credible path to a true proposed ROI as derived from one’s CM program. CMMS systems can readily demand and track such feedback.
Applying the IA: This is part of 21st Century CM. There are far too many pieces of information for a human to quickly and most advantageously link together—this is the domain of software. As I’ve stated earlier, a few qualified IAs exist and, of course, can always be improved (or more written, as particular purposes arise) to accomplish the critical task of evaluation, the end product of CM—what to do and why.

The timing issue, that of resolving the effect of old data from static lab testing versus real time data, is tricky but can be logically addressed.

Figure 1 was presented a few years ago to suggest at least four possibilities that will routinely present themselves. There are a few more, along with nuances, to be sure, but developing an approach to each scenario is the key. Once feedback is available (and if it’s not, consider a different profession hereafter rather than stubbornly hope for it as I do), adjustments ranging for tweaking to restructuring the logic can be accomplished. There is no secret to doing this—it requires domain expertise in the logical mix, i.e., in the software code, and is also teachable with useful feedback.


Figure 1.

As an aside, sensor proliferation will shift the application of domain expertise where it, perhaps, ultimately belongs—to those who know the machinery best because they see it daily. It is not going to be feasible for bench evaluators and diagnosticians to learn all the nuances that sensors will begin to expose, thus presenting new avenues to deeper diagnostics.

Next column I will suggest some ways in which the IA can approach the contribution of sensors in the context of ongoing fluid analysis bench testing, expanding on the example above.


Jack Poley is managing partner of Condition Monitoring International (CMI), Miami, consultants in fluid analysis. You can reach him at jpoley@conditionmonitoringintl.com. For more information about CMI, visit www.conditionmonitoringintl.com.