Model Turbines

I ran a test with my smallest boiler and the cast axial rotor to see if the steam output of the small boiler was part of the problem. I used this rotor since the position can be adjusted to allow space for the steam to go supersonic before contacting the rotor. The ability to do this has made this rotor perform better than my tangential flow rotors with steam even though the open pocket rotors have done better on air. I added a carefully measured ½ cup of water to the empty boiler and placed the lighted wick burner in the boiler. I started my stopwatch after the steam started flowing from the turbine and noticed it took approximately a minute to reach the maximum pressure of 25 psig. The pressure stayed at 25 psig for the total time of the run. The boiler went dry after approximately 9 ½ minutes. There was a large drop in speed for a few seconds that indicated a slug of water got into the steam line. The resultant mass flow is 1.64 lb/hr but the carried over water invalidates this result. The maximum speed obtained was approximately 18,000 rpm and the vibration was fairly high even though this is my best balanced rotor. This indicates that the bearings may be the problem. I will change the ball bearings and see if I can get a run without water getting into the steam line.

I ran several tests on steam with the same ball bearings and using Krytox GPL 105 oil. I add a small amount of the oil to the shields of both ball bearings before each test. On all tests, the maximum speed of the turbine was reached after the turbine housing got hot enough for the oil to get up to its operating temperature and the oil on the outside of the bearings to seep in. When the oil reaches the optimum amount inside the bearings, the turbine speed would stay approximately at the maximum speed for the rest of the test. This worked for every test until I tried running the second test with my Turbine 2 described in the post of 30/05/2020. This was the first time I did not add a small amount of oil to each ball bearing before starting the test. Failing to add the oil before the test and running the rotor that had the worst balance rigidly clamped to the base plate, may have ruined the ball bearings.

I got my new ball bearings and ran a test with air, the cast axial rotor, and the GWS EP 2508 propeller. The maximum speed reached was approximately 24,000 rpm. This was about a 1,000 rpm higher than I was able to get with the bearings that had several hours of running time and had never been exposed to steam. The loss in power after several runs on air is about 0.3 watts. I don’t plan on running on steam for a while since the test results with air are much more consistent and the runs are much easier to do.

In the post of 29/05/2020 at the following link https://www.model-engineer.co.uk/forums/postings.asp?th=140195&p=9 I stated: ‘The power was less running on steam from my small boiler than what I have obtained running on air with Turbine 2 (0.9 watts vs 1.4 watts)”. The following post from my Testing Models thread at https://www.model-engineer.co.uk/forums/postings.asp?th=139899&p=5 showed similar results for a small oscillating cylinder steam engine running on my small boiler. In that post I showed that the power of the Miniature Steam Models (MSM) Tyne, described below, was less running on steam than what I had obtained running on air (0.7 watts vs 1.1 watts). In both cases, the power running on steam was approximately 64% of the power running on air. The Reynold’s number losses appear to be a major part of this drop in power.

After failing to be able to show that the loss in performance running on steam instead of air was due to Reynolds number losses, I thought I would look at the losses due to moisture. I assumed the moisture losses were negligible when using my smallest boiler because I had documented that the steam coming out of the nozzle was completely invisible when I previously tested nozzles with this boiler. Since I am planning to try a larger nozzle to see the effects of running at a lower pressure, I made a nozzle that could be slipped into a hose and tested with my smallest boiler. When I ran a test with this nozzle, the steam exiting the nozzle was completely invisible even at a distance from the nozzle. No moisture was evident in the air or on the surface below the nozzle. This is what led me to believe the steam would stay dry until it passed through the turbine rotor. I didn’t think there was enough time for heat transfer to the turbine housing and turbine rotor to raise the moisture content significantly. Since the cover plate I use for running my axial turbine has the nozzle in it, I decided to run a test using the cover plate shown in the following photo. Running the steam through the cover plate only, allowed me to observe the steam exiting the nozzle and there was more mass to absorb the energy. Unlike the test with just the nozzle only, the moisture was very visible with drops of water on the cover plate and the floor. Apparently, the steam can be cooled enough to increase the moisture considerably if the cover plate and housing are not insulated. My turbine tests are a worst case for this because the turbine housing and cover plate have a lot of mass and the propeller blows cold air over the uninsulated housing. The run time to empty the boiler that contained a carefully measured ½ cup of water, was 10 minutes and 59 seconds. The run time for this same amount of water in a previous test was 10 minutes and 56 seconds. The mass flow of 1.4 lb/hr for both these tests was very consistent and the pressure stayed at 20 psig for almost the entire time of both tests. The best estimate I could make of the amount of moisture based on the nozzle size and this mass flow, is approximately 20% moisture. This amount of moisture reduces the power considerably and might explain the drop in performance running on steam instead of air.

We can be confident steam should produce more power than compressed air because it contains much more energy. Something is wrong, a few possibilities, perhaps acting together:

1. Is your steam really steam at the boiler? First thing to come off a boiler is a mix of water droplets and water vapour. It has low energy content due to being not hot enough. With rather a lot more heat water changes state into true steam.  At low temperatures and pressures still wet, only at high temperatures is steam completely dry. Unless the boiler has a super-heater, the steam will always be a bit wet, and a poor boiler will produce very wet steam.
2. The pipework and valves between boiler and turbine leak heat, causing dry steam to convert back into wet steam or actual water. Short well-lagged pipes are needed.
3. The turbine is a hefty heat sink. Even if well-insulated, it has to be warmed up to steam temperature before running a test otherwise it will remove energy from the steam and gum the turbine up with water.  Rather than outputting power to the shaft, the turbine becomes a water-pump.  Although the propeller dynamometer suggests low output, the turbine's actually doing much more work than it did on compressed air, sadly it's all going to waste.
4. Possibly the turbine suffers extra friction due to differential expansion.  If the rotor is hotter than the case, the two may foul. The rotor's certainly likely to hit solid water if any condensation occurs.  Again, the cure is to make sure the turbine is hot enough to prevent condensation before running the test.

I think your turbine will perform much better after it's been thoroughly warmed up and fed dryer steam. A water trap in the pipeline would stop slugs of water getting into the turbine.

Hotter the better.  Steam coming out of the turbine should still be well above boiling point. The trick is to extract the remaining energy, which is significant, by deliberately condensing it. A proper condenser creates a vacuum equivalent to up to an extra 14psi on the input, ie your 25psi boiler, looks like a 39psi energy source to the turbine.

The advantage of testing with compressed air is it avoids a host of wet problems. Steam may be far more energetic but it's harder to tame.

Dave

Edited By SillyOldDuffer on 05/09/2020 13:57:15

Hi Byron,

Your thread is a good example of the type getting lots of reads but few comments. Probably because what you're doing is interesting but specialised. I always read it.

One comment to show I've been paying attention relates to using the rpm of a propeller as a Dynamometer. They work most accurately at high rpm, but become less sensitive at low speeds. It shows on the graph as a long tail where power doesn't increase much over a wide range of low speeds, and then curves up nicely at the high end. There's a high-end limit too due to propeller tips going super-sonic but I don't think it's a problem. But power measured by a propeller at ordinary steam engine speeds won't be accurate in the way it does a good job at turbine rpm.

Ages ago I did a MEW article describing a Dynamometer designed to measure the very low power output of a Ridder's Coffee-cup Stirling engine. I adapted a technique used to measure the power output of 19th century marine engines. Builders of new steamships were contracted to show the engine produced an agreed power output at the propeller (Shaft Horse Power). They couldn't use your method because propellers weren't calibrated as is possible today. Instead, the ship was steamed at contract speed while the torsional twist of the propeller shaft between engine and propeller bearing was measured. The force necessary to twist the shaft as a giant spring had previously been measured in dry-dock with weights and a lever. When cruising the shaft twists slightly between engine and load (the propeller) and this was measured end-to-end with a telescope. SHP can be calculated accurately from rpm and knowing how much force is needed to twist the shaft.

For comparing turbines your calibrated propeller technique is fine I think, but it becomes unreliable when applied to slow engines. For them, safer to use a different type of dynamometer; my torsion shaft can be done with a microcontroller and optical sensors and is adaptable to a wide range of power outputs, but for most model engines I suggest a small conventional brake dynamometer is easier to make and use.

Dave

My recall of the science involved in air-resistance dynamometers is that the power required to drive a (subsonic) fan etc increases with the cube of the speed (rpm). This relates well to the effort required to shove a vehicle along on a flat road.

As the cube of next to nothing is next to nothing, this explains why the power required at very low speed is itself next to nothing. Any bicyclist will confirm this.

Cheers, Tim

Hi Dave,

I highlighted the section of your post that I found uplifting. I try to show from my testing things to avoid, things that make improvements, and the approximate performance you can expect from model steam engines and steam turbines. I sometimes wonder if anyone cares about these details because I get so few responses. It is nice to know that this is of interest to at least a few people.

I totally agree with your comments on propellers. I should point out more often, that my primary goal with using the propellers is to find if a change makes an improvement. In that respect, If the propeller speed increases or decreases after the change, it shows the change helped or hurt the performance. This should hold true even when the propellers are run at speeds outside their designed and/or tested operating range. You are correct that the actual performance of the propeller should only be used in the speed range it is designed for and/or was tested at.

Hi Tim,

I agree with you that the power of miniature turbines is VERY small at low speeds.

Thanks for the responses,

Byron

I thought about the comments you made copied above and went back and looked at some of my charts for the steam engines (especially in my Testing Models thread). I should not have extrapolated the performance based on the APC propellers to speeds below 1000 rpm. Even the APC chart for the 22" diameter propeller I used testing my largest steam engine does not show performance below speeds of 1000 rpm. I am going to copy these comments and add them in a post on my Testing Models thread.

Thanks for pointing this out,

Byron

Hi Dave

I'd like to look at the article you mentioned in which you describe your shaft torsion dynamometer. Can you please tell me the issue/year of MEW in which it appeared? I am wondering if it would be a more suitable alternative to the water-brake dyno that I have partly built. I am thinking it would enable me to have automated data collection without having to tackle the electronics needed for analysing a load cell. I feel happier about using optical sensors and microcontrollers since I've used them before.

Are there any problems with that type compared to the water brake type? e.g. calibrating the torsion shaft which I guess is done using static weights hanging off of a lever?

I'm not sure, but I think the first torsion dynamometer was the one invented by Charles Parsons in order to investigate lack of power in the Turbinia. His dynamometer is on display here in the Discovery Museum in Newcastle and led to the Denny Brothers of Dumbarton developing the shaft torsion dynamometer for use in the ships they built.

Regards

Mike

Thank you all for your responses. I am interested in better ways of testing and appreciate all the ideas that are being presented. I will probably have more to comment on the dynamometers after I do a little research.

Byron

Edited By Turbine Guy on 07/09/2020 19:54:18

Now you're asking. I'm disorganised! I'll have to look through my mags to see if I can find it. Fairly sure it was the first half of 2017 but I might be wrong.

I'll have to check the old book I got the info from: I think it predates Parsons but could be wrong. Calibration as you say by hanging weights on a lever, though that's not necessary if you have the steel's specification.

There's also a mostly complete Arduino project developed in collaboration with Duncan Webster intended for a Dynamometer car he's been working on. The Arduino captures load cell values on each wheel revolution and the time of each revolution. The data is copied to an SD card or streamed to a connected PC. A program on the PC calculates rpm and power etc from the raw data and draws graphs. Been tested with a simulator but not on the track.

Probably the main disadvantage of the torsion type is the long 'spring'. Not as compact as other types. Easy enough to set up and calibrate though. The Arduino and photosensors measure the angular twist by detecting the time difference between two normally aligned holes in discs at each end other spring. The disc next to the load lags behind the one at the engine end.

Dave

Thanks Dave.

On thinking about the torsion type dyno I am not sure how I would easily measure the angular twist accurately when the shaft is stationary - while keeping the loaded lever dead horizontal. Maybe with a dividing head to hold and rotate one end of the rod while the weighted lever is attached to the other end?? A finely divided rotary encoder disc might do the job but would add complexity. These thoughts together with what you say about the length of the shaft make me think that the load-cell approach might not be so bad after all. Nevertheless I'd still be very interested to read your article. I looked for an index to MEW but the only one I can find does not cover anything more recent than 2016.

I'll ask Duncan about the dynamometer car.

I'd be interested to hear if you find that book. Parsons designed his dynamometer very soon after the first trials of Turbinia had been carried out in 1894.

Regards

Mike

Can't find the magazine or the book, which I'm sure is much later than 1894 so Parsons is the man. Talented chap!

Anyway this diagram shows how I measured the elasticity of the torsion bar (a length of piano wire):

One end of the bar is clamped while the other terminates in a free to rotate disc. A balance bar is glued to the disc and adjusted so the torsion bar is unstressed, ie the two ends of the balance bar are at equal height. Then weights are added at point W causing the torsion bar to twist and the balance bar to dip. The angle can be measured directly with a home-made protractor, but I calculated it with trigonometry.

It gives the elasticity of the spring, from which the force needed to deflect through a measured angle can be calculated. From unreliable memory I think the torsion bar must be at least 40 times longer than its diameter.

Still looking for the magazine. As I can't find any mags containing my articles, I must have put them in a safe place and forgotten where it is.

Dave

Thanks for that - I see I was making things much too complicated. I guess you used trigonometry to calculate the the reduced torque as the arm with the weight dipped down below horizontal - in addition to calculating the actual angle.

I think I've identified the issue that your article appeared in - by looking at images of mag. covers. The issue for June 2017 (No 255) has a list of 5 projects for the workshop, one of which was "Arduino dynamometer". Does that sound like the one? If so I'll buy a copy.

Mike

Werner Jeggli shows in the post of 18/06/2019 the picture of the system he uses to measure the power of his steam turbines. He uses the brushless DC servomotor as a generator and can change the resistance to alter the load on the turbine. This allows him to run the turbine at any speed and measure the maximum power at that speed. This is the biggest short coming of using propellers for a load. The torque of the turbine determines the speed it can spin the propeller. What makes this even worse, is the very small number of propellers available with test data showing their performance at low powers and high speeds. Because of this and his distrust of extrapolating results, Werner has described my method of testing as “woefully inadequate”. I believe that if you are primarily interested in finding the maximum performance at a particulate speed or range of speeds, you do need a much more sophisticated test method. Either Werner’s method or the use of a torsion type of dynamometer would substantially improve the testing. I do believe that using the performance of a propeller given at a speed within a tested speed range is valid and if it meets your needs, is far simpler and less costly.

Edited By Turbine Guy on 10/09/2020 17:03:19

Edited By Turbine Guy on 10/09/2020 17:08:43

Edited By Turbine Guy on 10/09/2020 17:11:49

Don't rush to beat yourself up. The propeller method has several advantages provided it's used within a suitable speed range and a decent calibration curve is available. Cheap and easy. But it can't be safely extrapolated over too wide a speed range, and they become ever more inaccurate as RPM falls.

A torsion dynamometer has a long springy shaft highly likely to whip at high speeds. They're best suited to low rpm engines. I don't think I'd risk putting one on a turbine, something else is needed.

Werner's approach is my favourite of the three but it too has disadvantages. Biggest issue perhaps is the dynamo introduces a second layer of uncertainties. It applies a load between the turbine output and the dynamo's output, notably the bearings, so power is lost before it gets to the rotor. Then the mechanical to electrical conversion efficiency of the dynamo has to be found over a range of speeds if accuracy is very important. And care has to be taken measuring volts and amps. Nothing awful but it all needs attention. Once calibrated Werner's setup is straightforward to use but it's the most expensive solution to build - far more elaborate than a simple propeller, and less accurate than a Brake Dynamometer in that it doesn't measure power directly. Still a jolly good thing though: worth building one to progress your interest in turbines I think.

Boils down to horses for courses again. Like many technologies the propeller gives useful power indication up to a point; the fun bit is recognising when results are suspect and a different approach is needed. The history of science is littered with examples of clever chaps gradually improving step by step; took about 250 years for experimental technique to improve enough to measure the mechanical equivalent of heat consistently to high accuracy.

Keep up the good work.

Dave