In the long run, all machines break down and the only solution is to employ a more efficient equipment maintenance strategy. Recently, in the Reuters Events hosted PV Operations and Financial Strategies, a panel on the topic, “Reduce your O&M costs by applying proactive and predictive maintenance” touched upon the nuances of maintenance. The speakers, including Swarup Mavanoor, CEO and Co-founder, SenseHawk, Julien Glover, Control Center Manager, Cypress Creek Renewables, and Michael Eyman, Managing Director, Origis Services, spoke of their experience with and opinions on the 3 terms associated with maintenance, predictive, preventive, and preventative.
Watch the excerpt
While condition-based maintenance techniques have been well adopted because of their ability to accurately track machine performance and anticipate failures before they occur, the terms of predictive maintenance and preventive maintenance, are often mixed up. Mr. Michael Eyman, during his session, clears the air saying, “People tend to mix these terms. I hear predictive maintenance. I hear preventive maintenance. I hear proactive maintenance. And then you have preventative maintenance, which I think is the just preventive maintenance and we don’t need an extra syllable. Preventive maintenance is a scheduled maintenance, where you are putting in work orders in the system on a scheduled basis, whether annually or every two years, whatever it is for the different pieces of equipment. Preventive maintenance, corrective maintenance, and ancillary services make up one of the big buckets of costs. On the other hand, proactive maintenance is when you can see that something is going to fail. The failure might be visual on the site, you may hear humming, something may turn hot, or you may get infrared. This sort of direct observing of an issue [comes under proactive maintenance. Or it may be something like the knowledge that the breakers will go only so many cycles, and you are tracking those to proactively change it out. You are not [in proactive maintenance] waiting for them to fail. Predictive maintenance is about data, at the end of the day. It is about taking data in the aggregate, applying artificial intelligence in some cases. It involves providing data analytics, or a layer of analytical approach, [to the available data] to make predictions on when things are going to be an issue. Even if there are no indications that those things are actually happening on the site, you are intervening in advance, based on that data.”
Reactinging to Mr. Michael’s opinion, Mr. Swarup said, “Preventative maintenance is, in reality, predictive maintenance. Of all the types of failures that have already occurred, people have documented them, they understand the MTBF of those, and are doing work to prevent those, which goes into preventative maintenance. However, there are random variations on several sites that you see occurring one off and are not expected to occur thereafter. If you have all the data that led to that event, you can go in, mine the data, and see if there was anything different in that particular device that caused the event versus another equivalent device, and all the associated data. Once you find out there is a way to figure out that “This is going to occur” in the past and “This is how I can predict this event”, that is predictive maintenance. If you are hundreds of these predictive maintenance events on site, they go back and become your preventative maintenance tasks automatically. If I have 100 inverters and I see a particular pattern in 95 of them, I’m going to push it out into preventative maintenance, versus being a part of predictive maintenance. These two techniques, therefore, go hand in hand.”
Mr. Julien pitches in with a question, “Although we’ve got some clear examples of preventive and proactive maintenance, what might be a real world example of predicting that something was going to happen?” As a rejoinder, Mr. Swarup cites a real-life example saying, “As a company, one of the earliest things that we started off with was analyzing drone data. While doing this on one, we found a weird thermal signature on a bunch of modules. The asset owner made the module manufacture to replace about three thousand modules. The manufacturer came back and said that these [thermal signatures] aren’t hotspots, these are “warm spots”.. He assured us that the warm spots are no big deal, as nothing will go wrong because of them. A year later, we made the asset owner have someone walk down the site. They found burnt out contacts in the back of all these modules, a lot of which were flagged initially. So, we figured out that this weird pattern that we could not place predicted this sort of failure happening in the future. We had not seen that on any other site, we had not seen that before, nor was it documented. That is just a pattern of data we found. That really is what predictive maintenance is, and the next time we see that pattern, we know that these [hotspots] might happen in the future.”
SenseHawk has developed a data analytics platform for solar which goes provides analytics from development to O&M. We are bringing together all the data on the solar side, along with information on components, and work orders, to a single platform to use the data to drive predictive maintenance. Contact us for a free consultation.