Click the arrow below to view the video. The presentation is divided into two parts, so once you have finished Part 1, simply click the video player below to view Part 2. A written transcript is also provided.
The Achates Power engine has got advantages for fuel consumption and emissions and the main challenge with an engine that has holes cut in the side of the bore is oil consumption. At the same time, there is a sub-oil system that is full of oil and the challenge is to keep the oil from going straight out the bores. But, at Achates Power, we have the tools and the knowledge to tackle this problem and we have an instrument to measure the oil consumption and to show that it is quite low, and that’s what I’ll show you today. So, I’ll show you some tools and techniques for how we measure oil consumption, we’ll talk about some ongoing work we have to reduce oil consumption, I’ll show you some measured results and I’ll show you what we have upcoming.
So the measuring principle for here at Achates is the sulfur-tracing technique. So we start with sulfur-free fuel, which means less than 2 ppm, and the fuel sulfur concentration is measured. We send the sample to a research laboratory and get it measured to the nearest .2 parts per million. Typically, it’s around 1.0 and 1.2 ppm. The next thing is we need to make sure the fuel lines have no sulfur in them, so sulfur-free fuel lines. Basically, this means no rubber and we do that with Teflon fuel lines, all the way from the fuel tank to the engine. And, additionally, we make sure that the fuel coming out right by the engine is also still low sulfur so after it’s traveling through the fuel lines it hasn’t picked up any sulfur.
The next is we need oil with a known sulfur concentration so we take a sample of oil at the start of each test and a sample of oil at the end of each test and send those samples to the same research lab. Typically, the sulfur content is about 4,000 ppm. It has nothing to do with the engine, the technique is to draw an exhaust sample from the firing engine and the sample needs to be conditioned. The first thing is we add pure oxygen to the sample and it goes through two ovens at 1,000 degrees C, the two ovens are in series. This converts any sulfur-containing species in the sample to sulfur dioxide.
The next thing is we need to eliminate any water in the sample, not that it contaminates the reading necessarily, but water is very good at absorbing sulfur so it adds a huge transport delay. We do this by having two membrane dryers in series with dry nitrogen pumped on the opposite side of the membranes.
Finally, we need to eliminate nitrogen monoxide. The reason for that is that the nitrogen monoxide has a chemical signature that slightly overlaps sulfur dioxide. We do this by taking some of the pure oxygen and converting it to ozone, which is done with a spark generator. This ozone is pumped into the same ovens as the pure oxygen. The ozone converts any nitrogen monoxide into nitrogen dioxide. The nitrogen dioxide does not have any chemical signature that overlaps the SO2. And, we can make sure this is working simply by turning the ozone generator on and off. So, for example, during a run in a steady state, we’ll have the oil consumption going, turn the ozone generator off and the readings should increase dramatically because the nitrogen monoxide has shown up. Turn the ozone generator back on and the ratings should come back down to normal.
The next thing in the chain is to detect the sulfur dioxide and we do this by taking the gas sample in a mirrored chamber and shining an ultraviolet light into the sample and that excites the SO2 and causes it to fluoresce. Next we have a photomultiplier tube tuned to the proper frequencies that quantifies this fluorescence. So, the voltage on the photomultiplier tube is directly proportional to the concentration of sulfur dioxide in the sample.
But, if you look at the equation on the bottom right, you can see that we need to know the mass rate of fuel, that’s simply the fuel flow sensor. We need to know the mass rate of air going into the engine, that’s just the air flow sensor going into the engine. The sulfur concentration of the exhaust is what the instrument measures. The sulfur concentration of the fuel is what we have the fuel sample sent and the sulfur concentration of the oil is from the oil sample results. And finally, we know the sulfur concentration of the air. Now in our test lab, there’s no significant sulfur in the air but this might be important for somebody else who would have, say, two test cells running simultaneously and exhaust from one engine is wafting into the next test cell. But, in our case, it’s essentially zero. Run that equation and we get mass flow of oil real time. So the mass flow rate of oil.
Now, of course, this measures the oil coming out of the exhaust pipes. That would be the oil consumed by the piston ring and liner kit and also the turbocharger shaft; any sort of leakages through there. It is not sensitive, so it would not include losses from the crankcase. Now, those would be no different than a four-stroke because it’s still a crankcase with oil vaporizing from the case being recondensed into some sort of collection system no different than a four-stroke. Same challenges a four-stroke would have, we would have with our engines too.
It is not sensitive to any water or fuel dilution of the sub oil, so the typical metric would be to grab a metric or volume metric measurement. Basically, see how much the volume in the sub or the mass of liquid in the sub changes after some time. And, in fact, if you have a lot of fuel dilution, for example, that can show up in a negative oil consumption, which obviously isn’t real. So the technique we use is not sensitive to that.
So here we show the manufacturer claims specifications—accuracy better than 10% when compared to the gravimetric method. However, we think our test method is even better. Repeatability day-to-day within 5% and, if calibrated, repeatability is within 1%. So we would take a two-point calibration at the beginning of every test and a two-point calibration at the end of every test and account for any drift. But, typically, there’s not much. So the two-point calibration means first of all zero would just be ambient air; there’s no SO2 in the air and that means zero.
Then we have a calibration span gas of sulfur dioxide. For example, the latest tank would be 2.104 ppm SO2 and that would be the full scale. Testing of the instrument is better than 10 ppb, but the lowest we’ve observed so far is maybe 160 ppb so that’s not challenging the machine at all. The machine has a measurement range of 10 ppb to 10,000 ppb and the Achates Power engine is nowhere close to 10,000 ppb, even at the worst case. It’s linear within the measurement range, it’s got no cross-sensitivity to any of the other exhaust constituents, except for nitrogen monoxide. But, as I mentioned before, that’s carefully removed. Now this can be checked, by the way, as the engine is running, we would turn the ozone generator off and make sure that the reading increased. Turn the ozone generator back on and show it back down to normal. So, we can do a check real time to make sure the ozone is doing its job.
Transport delay could be as short as 8 seconds if the machine is physically moved into the test cell, next to the engine. However, in the test cell, it’s about 40 feet from the engine so we have a transport delay of about 30 seconds. The machine samples every 1 second, so that means that we can get a real-time curve as shown here. So the black curves is oil consumption measured in real time. Every one second the sample is taken. If you look at the time scale it goes from 12 to 30 minutes in the middle of a test and, typically, we would shut the machine off between samples to make sure that it goes back to zero every time and, sure enough, it does here. Now shown in red would be engine speed. The point being that these are different operating conditions, everything here is the same load but we changed engine speeds—up, up, up—and at each point we would take a sample.
So where it shows transport delay with the two arrows, on the left-hand arrow you can see that when you go from a low consumption point to a high consumption point, the first bit of sample is literally the gas from the previous sample purging through the line. So the first 30 seconds represents the transport delay from sample to sample. So we make sure that that 30 seconds has gone by plus an additional 10 seconds just to be sure before we start averaging the data shown in the green—that would be a 60 second average for the data tag. Another example I’m showing on the right-hand there, from a high to a low consumption point, you can see the first part of the second sample starts high and then it transitions low and, again, that’s the transport delay.
So you can see here, we have six different operating points within just a few minutes. Here’s what an entire test looks like. So we test all speeds and all loads. So the load fraction is shown in green and you can see that we started at a certain load fraction and stepped all the way through the speeds and then increased the load fraction and through the speeds again and mapped the entire engine. So I’m showing just under 30 points here. So it maps the entire speed load range of the engine with 30 different operating points and we accomplished it in less than two hours.
This is very important because you can see high spots and low spots. If you just test one operating point in the map, you might not see a high point very close to it. If you just do a cycle average, you don’t have a good understanding of what point is causing the trouble. But, here, we can resolve that very closely and really zoom in quickly to see what the problem may be.
So some design parameters we look at to affect oil consumption. Let’s start with the basics, oil ring tension and conformability to the cylinder—more roundness; nothing unusual there. But, also the details: end gaps, land chamfers—all those machine details that you might not think about until you look under the microscope. All those are important. We look at ring dynamic stability, turbocharger shaft leakage, of course, oil evaporation (if it gets too hot), inter-ring volume and reverse spouting from the pressure dynamics between the piston rings. We look at details of leakage through the seal to the groove flank and, of course, leakage between the ring and the bore.
So here’s the basic model of how we simulate the ring dynamics. We have two software packages that we work with, fully detail, take everything into account. So this model is a flow path through the ring pack and with orifice areas and cumulative volumes. So in this diagram you can see, on the top is, what the cylinder (represented by cylinder pressure) which, of course, we measure on a crank angle resolve basis. And then down in the crankcases is the opposing pressure. We also have ports down between the oil rings and the compression rings and the pressure there is measured and put into the model.
Through each orifice area, the combinations, so take the top A1 there, that’s not just the end gap. The end gap is maybe 30%. But, there’s also the chamfer on the end gap. There are the chamfers on the lands. There’s the ring face-to-bore leakages from any distortions. There’s the ring-side groove leakage through even from any machine errors there, a twist, you know that sort of thing. And, don’t forget, the texture itself. So, we talk in terms of the RBK, like the area below the median line in the texture so the gas leaks through those micro channels. Same thing on the side face.
Now the volume would be between the two rings so if there’s any accumulated volume or even just the clearance from the land to the bore and all that, all that goes in. So, take for example, if P2 is bigger than P1, at any one point in the cycle, you can get reverse spouting, which means that the gas flow reverses and flows backwards through the ring. Worse yet, it could actually loft the ring off the side face and it destroys the ceiling at that point. So if they ever have a situation where the pressures are still high and the ring is lofted, that can cause some instabilities.
One more point is that all of these volumes are filled with oil aerosol to varying extents. The crankcase would be a heavy concentration of oil aerosol and the cylinder, you know, would be some very small concentration. And we send the cylinder out to get the oil concentration measured with the aerosol test fixture. What you’ll see is that the crankcase has got the heavy concentration between the second and third ring—it would be a lighter concentration. First and second ring it would be lighter yet—finally, up into the cylinders.
Now, just because the concentration is lighter between the top rings doesn’t mean it can be ignored because the pressures are higher—so both go into account.
So again, with the measured cylinder pressure trace, measured port pressure traces, multi-intake exhaust, also lighter temperature profile to understand how the oil might be evaporating and also thinning out, all this is put into the detailed model and it can predict inter-ring pressure traces, the friction of each ring, the reverse blow light and, of course, oil consumption.
So here’s some measured results and some progress that we’ve made here at Achates Power. Now oil consumption can be plotted on a contour plot just like brake-specific emissions, for example, or BSNOx or BSFC. What’s shown in these two graphs are the contour plots. Now they show fuel-specific oil consumption, which means how much oil is burned per mass of fuel as a percentage.
So, on the left you can see there are some high spots and low spots. This is early in the development, but through multiple efforts, we’ve cleaned those up. To the map on the right, and a nice, low flat map.
So shown here is a hardware change. We did an A/B/A test to where A is a cylinder kit. So we installed the hardware, ran the engine and got a baseline map and then removed that hardware and installed the B cylinder kit hardware re-ran the test and then re-installed the A kit and ran the test one more time. So, first thing to notice, is from the initial A test to the final A test, we have good repeatability—the high spots are high, low spots are low and generally the characteristic of the map is identical. Now the B kit, in this case, was interesting because we expected the oil consumption to go up. But, the interesting thing was, the low spots went up but the high spots did not. That kind of implies two separate phenomena are going on. In addition, if you notice the bottom right part of the map, at 2600 rpm and light load. That was a high spot to begin with and it got higher again. So that implies a third phenomenon. Now all this shows up with the techniques that we have because with the test where the entire map is run and sort of lumped into one result, these nuances don’t show up. But, we can excise this very quickly and zoom in to understand how to fix it.
Here we show our achievements to date. Fuel-specific oil consumption is less than .10% across most of the operating map, including the B50 point and 2000 rpm and moderate loads there. Also, I show oil consumption in two different ways. First of all, just simply oil consumption in grams per hour. Some manufacturers might be familiar with this and you can see that it’s about 3 grams per hour or less in most of the map. When I think in terms of brake-specific oil consumption, which means how much oil is consumed per useful power of the engine, and that’s good across most of the map too. Of course, at very, very light loads it looks bad because the engine is not producing much power. And, in fact, in idle, by definition, there is no useful power produced until you get an infinite number there. But, fuel-specific oil consumption is a nice measure because if you know the fuel consumption you know what the oil consumption is.
So, just as a reference point, if you have 0.10% fuel-specific oil consumption in a vehicle that consumes 30 MPG of fuel, well that means the oil consumption is 30,000 miles per gallon of oil—4 quarts to the gallon so that means 7500 miles per quart and that’s state-of-the-art with a typical modern automobile, for example.
So the Achates Power engine has achieved fuel-specific oil consumption of less than 0.10% across most of the operating map. Some upcoming work: we’re going to continue to analyze and optimize the ring dynamics and its effects on oil consumption and friction at the same time. Generally, what’s good for oil consumption is also good for friction, by the way. We’re going to continue testing the basics: ring tension, bore cylindricity, geometry of the rings and so forth. We’ve got a couple different oil ring configurations to analyze. The opposed-piston two-stroke allows some radically different concepts there. And, then once we get oil consumption low in every step, we will continue our durability testing to make sure that the wear and scuff resistance has not been degraded at all. One important thing, from our point of view, is to determine the customer’s drive cycle and what their oil consumption target might be.
So, in summary, we have measured results showing that oil consumption in the Achates Power engine is low across most of the map and this leads to a better engine.