I interviewed Carl Fransman who discussed Predictive Maintenance.
Can you first provide a brief background of yourself?
Good morning, Dustin.Yes, my name is Carl Fransman, I live in Belgium, and my background is mainly in software, dealing with aftermarket spare parts planning and optimization or any predictive processes associated with aftermarket.
What is predictive maintenance?
Very good question and actually a point of confusion for many people.A lot of people confuse, first of all, preventive maintenance with predictive.Where preventive is more like a pure insurance policy, you just try to determine normal failure rates and then you just set fixed time stamps or cycle stamps to perform interventions.Think of it like maintaining your car every 10,000 miles.That's preventive maintenance.An improvement on preventive maintenance is condition-based maintenance, where periodically, you're going to check on the actual condition of the asset or parts of the asset, and based on the check, you determine whether or not a maintenance is required.Some people confuse that and predictive maintenance, and it's not yet predictive maintenance.
Predictive maintenance will look not just at condition of the asset but also much larger.It'll even look outside the asset for any condition that may influence the future state of the asset.If I can be more clearer, let's say you want to predict the failure of an engine, and that engine sits outside.It's quite understandable that weather may play a role, but weather is not just temperature.It's also humidity, it's wind strength.What could also play a role is the load that's being put on that engine, so gathering all data together with the data collected on the engine allows us to do predictive forecasting, predictive analytics which then drive predictive maintenance processes.
It's quite a complex process.First, you have to forecast the future state of the equipment.Is it going to be good, or not good?And depending on that forecast, you have to determine whether or not you will intervene.Do I do a maintenance before it was planned, because I expect a failure to happen, or not?All that is risk mitigation.Your forecast, your prediction is what it is.It's just a prediction.It tells you there's so much chance that the machine will fail, but even at a 99% chance, you still have 1% chance it doesn't fail, so even in that case you have to weigh the costs of the intervention versus the potential cost of a breakdown, and that cost may be purely financial, it may include safety issues, what have you.So, it's much more complex than meets the eye, but the opportunities, done well, are huge.
Can you talk more about how it's done effectively?
Yes.Everybody has read about internet of things, machine-to-machine communications. This is the first condition.You need to be able to monitor the state of the equipment and, as we said before, of anything around the equipment that could potentially influence the equipment.So we need digital information on the status of the asset.This is now possible.What is also possible is to deal with the amounts of data that this generates.Big data technologies are readily available.We still have to be picky, because we need the performance that's required to do predictive is much higher than just business intelligence, for instance.These are the first conditions, so they are purely technical.Another condition is to really understand the business, meaning you have to understand what the costs and opportunities are in order to create a model that then allows you, with the predictions, to draw the right conclusions.
We see a lot of projects being done which focus solely on proving that prediction can be made.Well, not much surprise there.We can make predictions, and some are better than others, but making the prediction is not the hardest part.It's making the prediction operationable that's hard, so you have to interpret the meaning of the prediction in light of your business.And that's where a lot of work needs to be done still today.
Where have you seen some success?
We've seen success across different sectors.Now, in our case, we focus on heavy equipment, so we focus on that type of equipment where the pain is the highest, so where you absolutely want to avoid unplanned downtime.And we've seen success in sectors like aviation, like rail, automotive, industrial equipment.So in essence, the science is sector agnostic...and what's really—again, and I cannot stress this enough—we're really...the potential is realized is in the translation of the prediction to an operational action, and that is variable depending on the sector.If I tell you a train may have a failure on one of its breaks, does that matter?Yes, it matters, but it doesn't matter as much as if I said on an airplane, "Your engine going to fail."So there a lot of things that we have to take into consideration before we just jump to conclusions based on the predictions.
About Carl Fransman
Managing Director EMEA at Predikto