How DTU’s Rain Test Facility Predicts Erosion

Announcer: Welcome to Uptime Spotlight, shining light on wind energy’s brightest innovators. This is the progress powering tomorrow.
Rosemary Barnes: I’m here at the Technical University of Denmark, Risø campus, with Nicolai Frost-Jensen Johansen. You’re a development engineer here, working in the rain erosion test facility, right?
Nicolai Frost-Jensen Johansen: Yes. Rain erosion is the main topic I work on most days. Besides that, we’re in the section for Mechanics of Damage, so we spend a lot of time on fracture mechanics as well. A lot of those meet together in the field of rain erosion.
Rosemary Barnes: It’s a complicated field with lots of different facets, and I think it’s still developing fast. Our understanding is definitely not complete. I understand you’ve already done an episode with Allen about the rain erosion testing machine, but for people who haven’t seen that, maybe we could quickly explain what rain erosion is.
Nicolai Frost-Jensen Johansen: Our focus here is to mimic, as well as we can in a lab setting, the process of rain erosion on turbines. For those who don’t know, as turbines rotate in a rainy environment, and in some of the more severe weather in Australia, the droplets impinge on the leading edge of the blade. Over time that builds up fractures and damages the polymers, composites, or whatever protective solution is on the edge, and it erodes or falls off.
The goal is not to test on a real turbine, but to do it at lab scale. What has now become the industry standard is machines like this one, a whirling arm rain erosion tester from R&D Test Systems. I think there are about 20 of this type of machine in the world now. As you might be able to see in the background, we have a big rotor where the test specimens are mounted. We spin them underneath an artificial rain field, and as the blades spin they impact the droplets. That way we mimic, as closely as we can, the same process as on real turbines.
Rosemary Barnes: So leading edge erosion wears away the surface and causes two problems. One, the aerodynamics are affected, so you get less power. And two, it will eventually eat through the laminate in the blade.
Nicolai Frost-Jensen Johansen: Then you get into a structural repair.
Rosemary Barnes: A structural problem, right. Both of those are bad, and you want to prevent them. You don’t want to be climbing up your wind turbine every year to replace this stuff. Ideally you would have a blade that never eroded over its entire lifetime, or at least you’d minimize the number of times you have to reapply protection.
That’s why people manufacturing new leading edge protection products want to test them and see how they’ll do in the field. But if you just put a product in the field, it will take a full lifetime before you know if it lasts a full lifetime, unless it fails earlier. So they bring it here, and the goal is to accelerate that testing so you can figure out in a few days or weeks how it’s going to last.
If you want to reproduce what happens in the field, tip speed on a wind turbine blade is maybe 90 to 100 meters per second. That’s about 360 kilometers an hour, like a racing car. So you spin at that speed and dump water on it, trying to deliver a lifetime’s worth of water in one go. And you’re saying 1,600 hours of rain is about what a wind turbine would see in its lifetime?
Nicolai Frost-Jensen Johansen: It depends a little on the approach, because there are different ways of accelerating. In a normal tester it would take about 1,600 hours to simulate roughly 35 years of rain. It also depends on what rain climate you’re trying to simulate, because different places in the world have very different kinds of rain.
Rosemary Barnes: So 1,600 hours is a little more than two months, running flat out. You’d have to get an intern in here supervising at nighttime. I guess you don’t just leave it running.
Catching the Onset of Damage
Nicolai Frost-Jensen Johansen: For the lower speeds, we do at times run overnight. The machine runs more or less automatically for a set amount of time. If it’s a long test, that might be one hour. Then it turns off the rain field, spins down the rotor, and images all three blades. From those images we look for any onset of damage, what we term incubation. That’s a really small, local loss of material.
Rosemary Barnes: So you run it for an hour, stop it, wipe off the camera lens, take a few photos, and start it back up again. Every hour you’re taking photos.
Nicolai Frost-Jensen Johansen: It depends on the test, and luckily that part happens automatically. It dries out the blades, the camera takes the pictures, and it spins back up. The faster we go, the more often we take pictures. Sometimes we take a picture every five minutes just to capture the real onset. At about 160 meters per second, it’s not uncommon to see damage after the first five minutes.
Rosemary Barnes: Okay.
Nicolai Frost-Jensen Johansen: Ideally we want at least five images before the onset so our uncertainty doesn’t get too large. That’s one of the differences from normal fatigue testing. This has a lot of parallels with fatigue testing: we have a repeated load over time, and we draw curves from it. But in a fatigue test, you impose a predefined load or displacement on a single specimen and run it until it breaks, so you know exactly when it breaks and at what load.
Our specimen is relatively long. At the tip we might be at 160 meters per second, and at the root we might be at 100. We run it for a set amount of time, stop, and observe it. So all we actually know is that it failed somewhere in the time we weren’t looking, and we know the position, which corresponds to a velocity.
In an ideal world, damage starts at the tip and grows toward the root. What we more often see is that it generally happens toward the tip, but it can be somewhat random along the length of the specimen. That’s especially true of modern coatings, which are on average much better but also more unpredictable. With an old gel coat, we could get a very nice curve. Damage would start at the tip and progress toward the root. Some of the newer elastomeric systems do an order of magnitude better, but they might just randomly fail in the middle. That’s really where the tricky material science needs to come in.
Rosemary Barnes: That definitely matches my experience in the field in Australia. You don’t see a little variation one year, a bit more the next, growing evenly and predictably across the whole wind farm. It’s more patchy. In one year, 90% of the wind farm might suddenly have heaps of damage, and then nothing happens for a few years after that. Do you think that’s related to imperfections in the coating material or the substrate itself?
Nicolai Frost-Jensen Johansen: We’ve done various studies, and we’ve had PhDs look at examples in scanning tomography machines. Often we do find that we can correlate the damage back to defects. Leon in our section is famous for his modeling work, and a lot of it deals with defects acting as stress risers. Some materials can also be more sensitive to having those kinds of defects.
There’s a lot of complexity within the topic, but if there’s one trend I’ve seen, from older gel coats and tough coats up to many of the LEPs we see now, it’s that we’ve moved toward near-pure resins. These elastomeric coatings are almost pure elastomer with very little filler. If you ever look at a cross-section of an older paint system under an electron microscope, it’s not so much paint as gravel held together with a little bit of polymer. Another way to look at it is that it’s 95% defects and 5% defect-free. Those systems will fail, and damage will grow, but it’s very predictable.
Now we’re using materials with less than 5% fillers, and a lot of those fillers can be defect drivers. So we have materials that are mostly defect-free, but we struggle to get something defect-free even at a 40-centimeter specimen length. It follows that if you’re covering 40 meters at the tip of a blade with LEP, you’re not going to get it defect-free. That mirrors a lot of other things within fatigue in general.
Fatigue, Damage Tolerance, and Tip Speed
Rosemary Barnes: There are regular charts we use in fatigue testing. Fatigue is how materials behave under a load that’s too small to break the material in one go, but if you apply it enough times, it eventually breaks. You characterize materials by putting stress or strain on one axis…
Nicolai Frost-Jensen Johansen: Stress or strain, depending on the case.
Rosemary Barnes: …and then you count the number of cycles until it breaks, and you get a nice curve for the material. With rain erosion, you’re saying that on one axis you have the speed.
Nicolai Frost-Jensen Johansen: Typically velocity instead, yes.
Rosemary Barnes: The velocity of the blade section relative to the rain. And on the other axis, you’re counting how much water it has seen.
Nicolai Frost-Jensen Johansen: You could say there are two schools of solving fatigue problems. In metals, you either add tons of dislocations so any crack growth is stopped because it grows into something else. Or you go the Rolex way and make your hairspring out of a single crystal, so there are no defects and all the atoms sit exactly where they should. I think the big challenge for wind is that we cannot make anything defect-free.
Rosemary Barnes: Not at a cost.
Nicolai Frost-Jensen Johansen: Not at a cost that makes sense.
Rosemary Barnes: Not if you want people to choose wind over coal, gas, or whatever else.
Nicolai Frost-Jensen Johansen: No. So one of the big shifts we’re trying to push in the science and the industry is to start focusing on how to make damage-tolerant materials. As I mentioned, we’re in Mechanics of Damage, with a background in fracture mechanics. The normal design criterion has been these curves, but that’s a purely empirical observation. We just say that at this point it failed. We have no real knowledge of why it failed, and it isn’t tied to the material properties.
We’re hoping to go further into developing materials that stop cracks themselves. There are various strategies you can use, and ways to define whether a material will naturally arrest a crack. Those might not always be the best-performing erosion coatings today, but they’re another avenue. Hopefully that will eventually let us do lab-scale tests other than the rain erosion test.
Rosemary Barnes: Say you want your wind turbine blade surface to last 35 years. You know how much rain it’s going to see in that time. The tip speed is 90 meters per second, so you spin it in this machine at 90 meters per second and keep putting water on it until it has seen its full lifetime of water. That’s already accelerated, because your blade doesn’t experience any dry periods.
But that might take 63 actual days, about two months, or maybe six months if you only run during business hours. That’s obviously an extremely expensive test, so you want to accelerate it. One way is to spin the blades faster, which you do. Can you tell me what speeds you run tests at to compress that time frame?
Nicolai Frost-Jensen Johansen: Under the current recommended practice, we tend to test at 160 meters per second tip speed, and also down at 100 meters per second, which is closer to real turbines. The big thing there is that these materials follow a power curve, sometimes with quite an extreme exponent. We talk about exponents sometimes up to around negative 3 or 4. That means a tiny jump in velocity can mean orders of magnitude in lifetime.
For example, a test that takes one or two days at 160 meters per second could, if we trusted the curve and extrapolated down to 100 meters per second, take 30 days or two months in some of the worst cases. It’s very sensitive to where on that curve you end up landing.
Rosemary Barnes: So when you speed it up, it doesn’t just run twice as fast, to keep the math simple.
Nicolai Frost-Jensen Johansen: No, it’s to a high exponent. A test that takes one day at 160 might take a week at 130. That’s only 30 meters per second slower. Humans are just not good at understanding power curves. Engineers see them all the time on log plots, where they look like straight lines, but the real consequence of a power curve is very counterintuitive.
Increasing the tip speed of a turbine by 1 or 2 meters per second can be the difference between having no problems with rain erosion, maybe doing spot repairs every 5 or 10 years, and suddenly repairing every year. It might have been a benefit for the drivetrain or the blade design to increase the tip speed.
Rosemary Barnes: To be fair, it is a big benefit.
Nicolai Frost-Jensen Johansen: It’s a big benefit. I just have to fight on my side of the fence and say there’s a consequence.
Rosemary Barnes: It’s been really interesting for me. I used to work in design and development, and now I work in operations and maintenance, and I’ve totally flipped on which issues I think are important. But this should be intuitive to all of us, because most of us have seen pictures of a blade with erosion. It’s at the tip. You don’t see erosion all the way down.
Obviously the tips are moving faster, but it’s only the last bit. If you keep moving inboard, there’s a point where the erosion just suddenly stops, and there’s barely any difference in speed between the eroded area and the clean area. One spot might be eroded all the way through the laminate, and a few centimeters away there’s nothing wrong except some tiny pitting. So we should have known that intuitively.
Nicolai Frost-Jensen Johansen: But on paper, what is one meter per second? It’s 3.6 kilometers an hour, a very slow walking pace. And that’s enough to suddenly lose a large part of your lifetime.
Speeding Up the Test
Rosemary Barnes: So that’s a risk, because you can accelerate the speed, but then you’re changing the behavior of the material. How much can you accelerate it? Could you go at 200 or 300 meters per second and get a really fast test? Or are you then measuring something interesting in the lab that doesn’t correlate with reality?
Nicolai Frost-Jensen Johansen: I think most agree that at 160 meters per second we’re probably already over-accelerating it, and potentially triggering failure modes that aren’t realistic. The risk is that you end up designing materials that are good for the test but not good for real life.
One of the big complications is that this is different from normal fatigue loading. There we can use linear mechanics. Hooke’s law and Young’s modulus are valid. But with impact loading, things happen on the microsecond or sub-microsecond scale, and materials no longer necessarily behave linearly. We have to account for viscoelasticity and all of these things, which compounds the difficulty of correlating results back to the field. There can be velocities where a material trips over a threshold and starts to behave very differently.
You could say there are two main ways of solving rain erosion. Either the material is so hard that the droplet hits and shatters. That’s why airplane leading edges use a lot of metallic solutions; you can get away with it there, and the yield strength of the metal is high enough to shatter the droplet. The other option, which is what many LEPs do, is to go soft. Instead of shattering the droplet, you disperse the energy by deforming, like a trampoline, with soft elastomers.
There are some polymers that are borderline hard enough to work. If you go to a higher velocity, you can trip over that critical threshold where the material just isn’t strong enough anymore. Then you’re giving a big advantage to a soft system, where at a lower velocity it might be the opposite relationship.
Rosemary Barnes: That’s really interesting. Of course everyone realizes that an accelerated test won’t behave exactly like the field. But you would expect, or at least hope, that it would rank materials correctly. If you have ten materials in here, the ranking one through ten would match how they last in the field. But it sounds like that’s not necessarily the case, especially when comparing hard and soft materials.
Nicolai Frost-Jensen Johansen: They can be very different.
Rosemary Barnes: Okay. So if going faster and faster isn’t the solution, or at least isn’t the only solution, what’s your other option for accelerating testing?
Nicolai Frost-Jensen Johansen: The direction we’ve been going is based on some studies from about eight years back, where we looked at the effect of different drop sizes. One of the unintended consequences of that study was that we ran some tests with much more rain. When we went back through the data and normalized it, we found that even though one test had twice the amount of rain flowing into the field, we could get all the curves to collapse back onto each other.
That showed we’re maybe not at the limit of adding water. Even though what we run is like a torrential downpour, it might not be too much rain. So our tester has a second rain field. On that rain field alone, we can run twice the amount of water, and we can potentially run both rain fields together to triple it.
So far this seems to scale linearly. If you double the rain flow, you halve the exposure time. We haven’t tested it fully yet, but hopefully when we triple the rain flow, we go to one third. Then all of a sudden one and a half or two months becomes a couple of weeks, which is much more realistic. Exactly where the outer boundary lies is a good question, and that’s what we have to figure out.
Rosemary Barnes: So you’ve been adding more little needles to get more water flowing out of the machine at once.
Nicolai Frost-Jensen Johansen: For the audience, a normal tester has 600 needles sitting around a radial manifold, creating a radial rain field. We have a second set of manifolds with 1,200 needles, allowing us to flow twice the amount of water.
Rosemary Barnes: What did you say the maximum rate was?
Nicolai Frost-Jensen Johansen: If you look at one point on the blade and count the impacts, we still see an impact rate of between 0.1 and 0.3 hertz. So it’s somewhere between three and ten seconds between impacts on the exact same position, when you account for the size of the droplet.
Rosemary Barnes: And that would be a problem if a droplet hit before the material could return to its original state.
Nicolai Frost-Jensen Johansen: That’s one of the concerns. It is a random process, so you could have two or more droplets hitting at once. But for two droplets to interact, they really need to hit within microseconds of each other. Most of the energy in an impact event dissipates over a few microseconds, so it’s fairly rare that those events happen.
Rain Intensity, Temperature, and Hail
Rosemary Barnes: But the actual rain intensity here didn’t sound that high to me, compared with what I see at the sites I’ve worked with in Australia. You said 35?
Nicolai Frost-Jensen Johansen: If you put a rain gauge in the center of the rain field on a normal tester, it’s about 35 millimeters an hour.
Rosemary Barnes: So that counts as a downpour. But I know that 100 or 150 millimeters an hour is not out of the question at some of the sites I’ve worked with in Australia. Will those sites have accelerated damage because they’re seeing too many raindrops one after the other? What do you think the difference would be?
Nicolai Frost-Jensen Johansen: I don’t necessarily think so. The big question becomes how the rain is connected to the wind. A lot of downpours happen when it isn’t windy, so the turbine isn’t at rated speed and isn’t hitting the same amount of water. That connection, whether it’s windy and raining at the same time, is the first next step to ask about. You can add a lot of other complications, like the effect of drop sizes, but from an experimental standpoint the main thing we’ve found is the amount of water you hit. That’s mainly driven by rain intensity, the velocity of the droplets, and how fast the turbine is spinning.
Rosemary Barnes: So it sounds like you’re going to keep adding more and more water until you reach some limit where it doesn’t make a difference anymore.
Nicolai Frost-Jensen Johansen: Yes. Of course, in the end we’re limited by how much our water system can push.
Rosemary Barnes: And you can only turn up the flow rate on an individual needle so much before it becomes a steady stream.
Nicolai Frost-Jensen Johansen: We might actually go to a steady stream at some point. That ties into the size of the droplet. Right now the tests run with droplets of about 2.4 millimeters in diameter, which is actually quite large compared with the majority of rain. Another option is to force it out as a spray of very small droplets, which might be more reminiscent of most of the rain we get. That still causes damage, because in many cases what matters is how much water is hitting the blade.
Rosemary Barnes: Interesting. Rain isn’t the only thing that dictates how resistant a blade is to erosion. You were telling me before that you’re looking at doing some different temperature tests. Can you tell me about anything else besides rainfall rate that you’re looking at?
Nicolai Frost-Jensen Johansen: We have an upcoming collaborative project starting in the next year where we’ll look into various meteorological drivers. We’ll test both high and low temperatures. We can heat the entire chamber and the water up to 30 degrees, which is a bit closer to tropical rain, and also go down to 8 degrees. We’ll change the average size of the droplets and maybe mix them, so one set of materials sees one drop size and another set sees another. Right now we test with one quite narrow drop size distribution, and it could be that we have to mimic more natural rain to be more realistic.
This is a bit more on the outside, but there’s also a lot of focus on what hail does. Depending on how you evaluate the severity of hail, the kinetic energy is off the charts compared with any rain droplet. The question is how that energy transfers to the blade and whether it does anything to the coating. We’re not going to be doing hail here.
Rosemary Barnes: I’d love to be here for that test, throwing ice cubes into the thing as it’s turning. You’d need safety goggles.
Nicolai Frost-Jensen Johansen: It would be on a more limited scale. One of our thoughts is not necessarily that we have hail erosion, but that hail could be one of the drivers for why you see erosion in a place where it shouldn’t be. You could imagine a hailstone compressing the coating enough to shear it or induce a defect. Over time that becomes the erosion point, and suddenly, at a random point on your blade, you have a big chunk missing. As soon as you break the coating, you’re into problems.
Rosemary Barnes: Right. We don’t have hail often in Australia, but we do get big hailstones fairly often where I live in Canberra. Every few years there’s a hailstorm where anyone whose car was outside has it totally wrecked. So that could be interesting to look into.
Nicolai Frost-Jensen Johansen: So far there’s a lot of theorizing on it, but we’re still lacking experimental verification of what it really does.
Rosemary Barnes: I would love to see how you manage to test that in the end. I think we’ve covered heaps. I’ve really enjoyed walking around the facility. There’s so much testing going on here. It’s a cool place to work.
Nicolai Frost-Jensen Johansen: There’s just a tiny corner that I’ve claimed.
Rosemary Barnes: I’m always jealous. I used to work somewhere where I got to work with my hands, figuring out clever ways to test crazy things like hailstones hitting wind turbine blades. You’ve got yourself a cool job.