Episode Transcript
[00:00:00] Speaker A: G'day, everyone. Welcome to Lubrication experts. And today we have a really exciting kind of window into a very large topic. So surprisingly, in some 45 episodes, we have never really talked about used oil analysis, which is crazy to think when we're talking about lubricants and condition monitoring, that we've never really, really touched on oil analysis in terms of regular standard. Send my cup of oil to a lab and get it analyzed and get reports back, which is what basically 99% of industrial businesses are sort of doing at this moment to varying levels of effectiveness.
But today I've got someone very exciting here to talk to us about used oil analysis and what we can do with that data. So, Elaine hepley is the data analysis manager at Polaris Laboratories. And Elaine, thank you so much for coming on.
[00:00:56] Speaker B: Thank you for having me. Pleasure to be on.
[00:00:58] Speaker A: Awesome, awesome. Now, Elaine, we actually met in person at Stle, which was really exciting. And Elaine brings, you know, a wealth of knowledge. So we're going to get stuck right into it. And I think one thing which would actually be nice, considering this is the first podcast where we've touched on it, is a little bit of history behind oil analysis in industrial applications. So how long have people been doing it?
At what point did it sort of mature into its current form where we have a reasonably standard test slate? And in your view, at what point did it become like an essential part of the condition monitoring toolkit?
[00:01:43] Speaker B: That's a very good question. It's a loaded question.
I did my research, though. It started in 1940, 619, 48 with Denver, Colorado and the Rio Grande Railway. It started with the railroad and they started off with elemental analysis, checking the wear metals and just kind of making informed decisions based upon that. And then after that it evolved like around the 70s, it evolved to the army or the military, the Naval.
And it was at that point named Jobe. The Joint Oil Analysis Program. And it was mainly just to do their in house monitoring for their pieces of equipment. And after that, that's kind of where it kind of started to become established. Then around the 80s, we had the very first fluid analysis laboratory opened in Oakland, California. And that's where it kind of started to cement itself, where these fluid analysis laboratories started popping up. And even Fluid OEMs started having their own fluid analysis laboratories to test their products to ensure they were meeting API standards. Is this product suitable for selling and things and distributing and what have you? And then over a while, I would say I started my journey with fluid analysis in 2008. And I believe at that point it was already pretty cemented.
There's still areas in the country, in the world, better said in the world, where, like in Latin America, fluid analysis is still pretty new. There are some countries like in Colombia and Brazil, some countries in Central America where it's pretty cemented, but it's still a new technology that's not very heard of, where here in North America and Europe we're much more advanced. I mean, we've gone from already establishing a fluid analysis program and test lates to already moving on to the next thing, which is like data connect CMMS systems and sensor data and integrating all of those key pieces to giving you a bigger picture of what's going on with a component. But when large companies started losing a lot of assets and they started asking themselves, wait a minute, why are we spending so much money on repairing and replacing broken pieces of component? And we're losing money on production, we're losing a lot of assets, and cost repair, maintenance, all that labor costs were rising. I think that's when people started seeing the true value of, hey, if I spend $10 on a fluid analysis sample that's going to tell me how well my equipment is running, but I end up saving thousands of dollars, here's my return on investment. So more and more people started seeing the value of that and started investing in that versus seeing it as an additional wasteful cost or added cost.
People started believing in it and started doing it more and more.
[00:04:49] Speaker A: Yeah, that's really interesting. I think I've seen it as well. Sometimes it's not even just by geography, but by industry as well. The stratification of where people are in that sort of condition monitoring journey, they tend to be the capex, heavy ones, tend to have a little bit more in the budget and have that capacity to do a little bit more with their reliability programs. But you mentioned you started in eight.
In the recent memory. Have you seen any major changes to the way that oil analysis is done? Like, do you see anyone requesting different types of testing?
I'm thinking a little bit about some of the more intensive varnish tests. For example, have you seen maybe the interpretation change as well?
[00:05:45] Speaker B: In a sense, yes. Back then it evolved with fluid OEM recommendations and their test slates. And then over time and that would be like your standard elemental analysis, viscosity acid number, base number, oxidation, nitration. Your standard test is what we call them in the industry. I've seen an evolution to filter debris analysis or root cause analysis of failure mode. Varnish analysis started off with turbines is mainly where we begun to see it. And since then it's evolved tremendously. We started off with MPC and Ruler and Rpvot, and then it has evolved to micropatch analysis, FTIR analysis, centrifuge analysis. There's a lot that you can get from all of these tests to tell you exactly what's going on or what is causing the varnish. Grease analysis has hit tremendously. Where Grease analysis started off with just what was it? Not condition monitoring, but new grease analysis, compatibility testing for Grease analysis. And since then it has evolved to in service grease analysis. And that's something that's pretty brand new. We've seen more data connect where customers would just fill out paperwork, send in their samples. We're evolving with the technology as technology, and everything is at your fingertips. We have to evolve. Change has been slow moving, I could say in some aspects of the industry. We're still kind of stuck in old ways of doing things and there are ways to improve. Change is good, it does cause a disruption, but change is good. It's better for the industry and where we're headed. Yeah, you know what I mean?
[00:07:40] Speaker A: That's a really good insight. So your specialty, well, at least your job title, is kind of data analysis. And that's one thing that I've always wanted to talk about a little bit more, because it feels like, well, everyone describes data as being like the new gold. Everyone's on this race for data.
I always think that maybe the better analogy is that data is the new oil, because until you refine oil, it's not actually very useful. It's just a store of wealth and then you have to refine it. And to me, that's like, you can have data, but without mining it for insights.
It just sits there. It doesn't do anything. So you got to interact with it.
And I see this as being a bit of a lost opportunity for a lot of businesses who do a lot of oil analysis, but then never actually look at the data. So from your perspective, a lot of the time we're generating, what, 25, up to 50 data points for every sample that we take, more or less, between all the ICP analysis, FTIR, oxidation, hydration, blah, blah, blah. So we're generating a lot of data, and sometimes over fleets, it's a lot of repeatable data too.
So it feels to me like we're not getting the full value out of that, maybe. And I think part of that is maybe the understanding. So if we could maybe start with the way that most people interact with their used oil analysis data, which is, has it hit some kind of limit?
Which is how people sort of start off in their use oil analysis program.
I think what would help a lot of people is an understanding of where those limits come from, because to a lot of practitioners, they can seem a bit arbitrary. Like, oh, my iron hit 50. Why is 50 bad? Versus 55? Or 60 seems like a very round, a suspiciously round number.
So can we get a good understanding from your perspective of how these limits are set?
And maybe sometimes where the OEM is setting a limit, how do they make the decision on what that limit is?
[00:10:05] Speaker B: Yeah, absolutely.
So there's different types of limits that are out there. That's something that is important to understand, is you have your OEM established limits. You can obtain limits by rate of change. And that's specific to the application, the operator, the industry that's serving, and that's exclusive, I would say to the customer just the end user gathering. If they have a decent amount of data that you can do that with, then you have statistical analysis. Statistical analysis is something that a lot of laboratories offer. It's something that I can only speak about what Polaris does, I'll touch on that more in detail. Basically, the way that it works is for statistical analysis you have to have a certain set of data and the criteria is going to vary based upon the data that is provided to you. So in order for us to establish sophisticated or more prescriptive feedback and flagging parameters, it is very important that we have all the information up front, right? So in order for us to get that, it's difficult. Most of the times customers don't provide us with all the information that's pertinent to give them the proper analytic feedback and recommendation that they need for their asset if they're not providing us. It's kind of like that term that you hear junk in, junk out, or garbage in, garbage out. So if you give me all the information, we can get you some data. So how it works is we take the last three years worth of data with a minimum requirement of 75 data reference sets.
We can filter that criteria based upon hydraulic, system specific, OEM and even model. So we end up with three different types of flagging criteria based on the data. But the minimum requirement is 75. Statistically, 50 reference sets is good enough to establish a flagging criteria, but we like to have more data. The more data, the more accurate has been our motto. And then based on the filter criteria, we generate these reports. We look at outliers and then remove any outliers. And the key thing there is statistically to implement limits, your average and your standard deviation have to be plus or minus five points from each other to know that that data is pretty accurate, that you can use that to establish flagging parameters. And from there it's more of a trend analysis. It's not a rate of change because we don't always for rate of change. It's ideal when you have time. Time is good for rate of change. If you don't have that, then it's difficult to kind of establish your limits at that point. Also, I don't know if you're familiar if heard of rate of wear calculations.
Those are key too. And rate of wear calculations has also been something that has been utilized by OEMs to establish their flagging criteria as well. So in my experience with OEM limits, that comes from them doing their test and data, putting that component through the wringer, changing the environmental temperature, exposing it to different operational environments, and even stress to see the rate of change and the kind of rate of wear based upon that. So we have seen where there are some OEMs that say for every 100 hours of operation, I should only see one ppm of chrome or iron rate of wear.
And while that's great for us, we feel like when you release that component or that machinery and it's out and it's being sold and it's out in the field, it's not going to wear the same way as it did in the test stand during that testing phase. You have a different operator, you have different environments. So we gather all of that data and base it off of a pool of customers that have provided us that data. And based off of that is how we've been able to establish our limits. And we currently have, I want to say, close to 18 million data reference sets in our database. It's pretty large.
[00:14:30] Speaker A: That's a nice peek behind the curtain, actually.
So maybe you can help me, because I think one thing that a lot of people would be interested with is those wear limits. It's always struck me that I go back to in engines. For example, the flagging limit will often be iron at 100 right now because it's an absolute limit and it's not tied to the amount of time that has passed or the number of kilometers driven.
Could you please help explain why is it that an absolute number for what should be a wear limit? Effectively, how does an absolute number make sense?
[00:15:24] Speaker B: It it really doesn't. To me, it doesn't make sense. There are some OEMs that establish absolute numbers, and they haven't been updated since like 1998. I think that was 1990 was the last time I saw one OEM. I can't call them out, but the last time they updated their flagging limits and the industry has changed, the engine design has changed. The PSI on these injectors has gone up from where it used to be. So the evolution, the metal, allergy of these engines is going to continue to change because of the demand from EPA, the customers, consumers.
You can't continue to have an absolute value when you're changing and designing a complete engine and consider that to be the absolute and always be that way. It's impossible, especially if you're talking about an engine that is working under a construction application and is exposed to dirt, and if it's ingesting a lot of dirt, you can't just limit iron at 100 ppm when it has a potential for more, or start flagging at 100 ppm. In some engines, that's so high that that's already a condemning limit. It just really depends.
For us, I prefer the approach of taking it by specific OEM and even specific model, because even though Caterpillar, Cummins, Volvo, they have a plethora of different models. But just because Volvo makes them doesn't necessarily mean that every single engine is designed the same way as every single model.
[00:17:12] Speaker A: That's not the case if you head.
[00:17:14] Speaker C: Over to the website Lubrication Expert. I'm building a platform to make the job of a Lubrication expert that much easier. There's a range of application based training modules as well as certificate preparation, including ICML's MLA One, MLT One, Vim and VPR. MLA Two and MLA Three are coming later this year, as well as, hopefully, CLS. There are tools for Lubricant and viscosity selection. And I'm starting to run biweekly zoom meetings where we can all just catch up and share our experiences as Lubricant professionals. Best of all, while a range of certification courses are in the order of $1,000 each, all of this is available.
[00:17:49] Speaker A: For $100 a month. I'm so glad that you've helped to illustrate that, because to me, it's all trying to convince customers that the OEM may have set some limits, but there is the potential that the OEM doesn't know what they're doing or that it's outdated.
And the example that I always give, because I've got a stationary gas engine background, is that there are quite a few of the OEMs set an arbitrary limit for things like iron and aluminium, copper, all the rest of it. And that limit is completely independent of the sump size, because you can buy these engines with extended sumps that have, like, 500 more liters in the oil tank, and that's dilutive right to the absolute number that you'll see on iron. And so how can it be that the number is still the same regardless of how much oil is in the tank?
And so that's always been a source of frustration for me.
And you have people running up against these limits, and they use that as kind of gospel when, like you said, we have many more tools available to us to look at things like rate of change and how fast something is wearing out and whether the rate of change has changed. So now we're going to second derivative kind of thing. Okay, awesome. So now that we've established that and.
[00:19:22] Speaker B: That'S just where we haven't even talked about fluid properties.
That's just where, in a nutshell exactly. Properties and all of that. But we won't need to get into all that.
[00:19:35] Speaker A: Yeah, there's plenty of other podcasts after this one that I'll get you back for. Okay, so now we've established that's how most people interact with their data is they're getting the results back, basically. Is there a red box or I've seen in some cases with some oil analysis companies now they've gone to emojis. Very modern, so sad face. Emoji means that you need to change your oil.
But now, from my perspective, we're now sitting potentially on a huge pile of data.
Maybe people have been doing oil analysis at this point for 1520 years, and in a lot of cases, their assets are quite old. So they've built up a lot of data on their assets. In my experiences, there's not that many of my clients, for example, who will go back and review all of their past data, and maybe they've got tens of thousands or even hundreds of thousands of oil analysis samples that are just sitting there in a database. So what's the opportunity here? I would have thought that with new kind of like machine learning tools and everyone I hate to say AI, because that seems to be a bit parse at the moment and everyone's talking about it, but it's kind of like an extension of machine learning.
What can we do? Or what can maybe the oil analysis labs, how can they help their customers to go back and mine all that data for insights would be is one question. And the second question would be to your point with garbage in, garbage out, are there a lot of limitations placed by the fact that historically people have not been filling out the boxes correctly?
[00:21:29] Speaker B: Yeah, we get those requests a lot. Oh, by the last three years, I haven't supplied you with my model and engine OEM, by the way. Here it is. Can you go back and reevaluate all my reports?
[00:21:41] Speaker A: Oh, no, I'm sorry.
[00:21:43] Speaker B: That's like reading an old newspaper.
Yeah, that was the best analogy I could have ever received from one of my colleagues who told me that. I was like, you know what? That makes sense. It's like reading an old newspaper. Any maintenance action that needed to take place three years ago, that opportunity is long gone. The ship has sailed. There's no turning back. You can't go back and take action on the past.
So it is important to ensure that we're being provided, or you're supplying your fluid analysis laboratory with the correct information.
Oftentimes we see the chassis information, sometimes in certain cranes, tractors, we end up getting the whole asset model. It's difficult if you have a transmission hydraulic on there to know what the model or OEM model could be for that. So I understand not providing us with that, but being as close to providing us as much accurate information as possible is going to provide the customer with the best feedback. So if you just tell me it's a transmission, but you don't specify automatic transmission, mechanical, marine, how am I going to know where you're at with your gear system, with your gears?
How am I supposed to flag lead and attribute it to clutch pack wear versus bushing thrust wear? What's important? So that has a lot to do with it. With regards to the data having tools, there's a lot of tools nowadays that you can take your data and dump it into a software like Power Bi and utilize that to do trend analysis. You can even do pareto charts where you can take all the data from all of your assets and do the 80 20 rule, where you see 20% of your assets are going to be causing 80% of the problems and narrowing that down. So with regards to having one asset that has several histories, let's say, for example, 100 data sets, taking that data mining, it doing statistical analysis. Even in Excel, you can do that and determine what rate take the time, if you have time and take the metals and correlate that. And what you're going to be looking for is increases or spikes in the trend of wear, but also correlating other elements with it depending on what asset type or component type we're looking at. So if we're talking about an engine, just to keep it simple, if you're seeing a spike in iron wear, are you seeing an increase in silica? Because we have found correlations between those metals. Are you also seeing aluminum spike up? That can be an indication of aluminum silica being present. And the first thing that we see wear in an engine is going to be iron depending on the OEM. So taking that data, taking that time, you can see the correlations and what is the data time? Are there other elements increasing with that? Could it be possible that maybe your specific engine or that specific asset when iron reaches 50, that's an alarm. Maybe that might be low compared to other engines, but given the acid, its age and its stage, that might be something where you might be a little bit more stringent. Monitor it closely so you can prevent a catastrophic failure.
[00:25:15] Speaker A: Yeah, no, that's great. And I think what's good is that's something actionable that our listeners can take away. Because I think even if the people that interact with oil analysis, maybe they don't have the power bi skills or Excel skills, but these days you'll be able to generally find someone within your organization that does, whether it could even be in the accounting division or something right.
That you might have to reach out to. But generally someone is going to have that capacity to analyze the data or at least structure it for you in a way that you can analyze through relatively easy to use dashboards and things like that.
[00:25:57] Speaker B: Yeah, and that brings another point to mind.
A lot of fluid analysis laboratories have programs where you go and retrieve your report that also offer management tools, program management tools, for example. I can only speak about what Polaris offers, but we have a data, it's like a data extraction tool where literally you can pick and choose filter what you want and you extract metadata as much of that as you want. It's going to take a while, but then you can take that and run with it. There are several different management reports that we have. And I do know that other laboratories should offer have some level of service that they can give you. Ask inquire with your provider, whoever's doing the fluid analysis, I always say ask, like you said, if they can provide you with tools, there's also CMMS systems that offer that as well.
[00:26:52] Speaker A: Yeah, plenty of opportunity there.
Yeah, I think that's great as we always kind of start to get to the end of these. I always like to ask questions about the future and this is an opportunity to look into the crystal ball and in some cases dream up what you want to see. Right? So this is an opportunity to manifest what do you see as being the future of used oil analysis. So maybe if I throw a couple of options at you.
I think traditionally we've seen a lot of there's field testing and then there's lab testing and now obviously there's a whole range of real time sensors that are starting to make their way to the market.
Is there the potential that, for example, more, let's say non standard testing, whether that's done in the field or whether that's requested specifically of labs, is that going to become more common? Do you see from your customers requesting very specific kinds of tests rather than, like, a standard test slate?
Or in the age of data and wanting repeatable data sets, do you see value in actually more standardization so that you're always doing the same tests, so that you can look back and do that data analysis and compare, like, for like so where do you see maybe the future going?
[00:28:28] Speaker B: And from what I see, electric vehicles EV, that is taking over the transportation industry. So there's going to be a shift, there's still going to be a need. I don't necessarily see that the oil analysis in itself, like doing testing and condition monitoring is going to completely go away. It's going to be used more as a confirmation tool or a root cause analysis tool is what I foresee.
We're going to see an increase in more turbine testing because somebody has to provide electricity to keep a lot of these vehicles charged.
So I see an increase in more varnish analysis testing for sure, an increase in the gas industry, wind turbine and gas compression industry that's going to spike up. Hydrogen engines are the next hot thing that's on the market. So I think testing is going to be there. We are going to be seeing more customers have a need for using field testing or tools like sensor data, relying more on that. And if an alarm goes off or is set, that could be a trigger to take a sample, send it to the lab to see what confirmation you might get, confirmation to take action or verify. Are we really seeing what the sensor is picking up? I see a lot of that. I also see a lot of data integration coming on board. And that is something that I have seen more and more where customers from their own platform can send us samples, send us data to us, and then when we complete the data analytics of that, we feed in real time that data back to the customer. Then they receive that. And depending on how they have their flagging parameters set up in their CMMS system, that immediately is going to tell them what alarms, what things to look for, and even triggers based on specific comments and what have you. So I do see an evolution of that. It would be nice, ideally, to be even able to get to the point where we can get data fed from the sensors itself into a platform. And you're analyzing that sample and you're looking at it and you're like, wow, okay, well, I'm seeing a spike in fuel dilution. The test results are showing that. And I'm also seeing an increase in the fuel burn, you know what I mean? So things like that, that would be neat if we can ever get there. It's been a dream, but it's something that I've heard talk about, what, five years now? Maybe a little bit more than five years. And I have yet to see that come into fruition, which I would like to see come into fruition, because as a data analyst, we only get a snapshot of what's really going on with that component.
We only see the data, but that's it. We don't know what operation it was. We're not there.
The maintenance guy is there to see what's actually going on. And so sometimes that's where the customer comes into play and they're more in tune with their equipment. So seeing more tools provided to the customer is going to be helpful. And definitely I do see an increase in more field testing. The one thing that I worry about field testing is coming from a lab perspective. Quality is huge for laboratories, ensuring that we're reporting out accurate data.
You have a human running the instruments. We're not perfect, we make mistakes. So ensuring that that machine is staying calibrated and knowing when data is not being reported correctly or accurately, or finding that there's an instrument related issue, how can you tell with a field test that it is providing you accurate data? And how reliable is that? You know what I mean?
[00:32:29] Speaker A: Yeah.
[00:32:29] Speaker B: So that's the one thing I worry about with field testing, is making sure that whatever field test you have, that it's reliable, accurate, it's calibrated and spitting out good data for you.
[00:32:43] Speaker A: Yeah, there's some really good insights.
And I guess like you were sort of saying about the data integration, it feels like we're at the very beginning of that journey. Because even when you look at the other condition monitoring tools that we have available to us, whether it's thermography vibration analysis, ultrasound, I haven't seen that many instances of customers even connecting those sets of data to their used oil analysis data to get a bigger picture. Because like you say, in the used oil analysis world, we can identify that maybe where is occurring, but maybe something else can then point to, well, that's come about because of misalignment or something else that is invisible to the oil analysis report. So I think, yeah, big opportunity there to really kind of fine tune, I guess, some of the insights. That we get out of all of our condition monitoring programs, which is great, because a lot of businesses are spending a lot of money on these programs and they are getting insights, and it is contributing to better reliability. But I think there's definitely some improvements that can be made.
Elaine, thank you so much. Really appreciate it. I am definitely getting you back because we've only skimmed the surface of used all analysis. So there'll be many more podcasts to come. No, but I really appreciate your time and we'll be talking again soon.
[00:34:26] Speaker B: Sounds good. You have a good one. Thanks again. See you later. Bye.
