IC Engine Water Pumps

[Charles Lamont]

I have been messing about since last year with some experimental water pumps. Work on these experiments is sporadic, and I am still thinking of improvements to the test setup, but I thought a quick look at some of the work might be of interest.
 
I need a water pump for my Westbury Seagull, and I want one which provide sufficient flow at tickover to allow the engine to run indefinitely. Some initial sums indicated that a flow of about 0.3L/min would be sufficient for the engine at full power. The pipework is 5/32" (4.2mm) bore and more sums indicated that an impeller diameter of 1/2" and blade height of 5/64" would be suitable.

The first test pump, (photos 1&2) has an annular water passage cut into the body. This pump was also intended to experiment with the shaft seal. I tried a stainless to PTFE face seal, but it did not work well and the pump was quite wet at higher heads. Grease lubrication of the shaft was more than the test rig motor could manage, so the pump was modified to take a ptfe bearing sleeve.

I wanted to see if a proper spiral volute would make a difference and built pump no 2 (photos 3-6). The spiral passage was milled in steps on the rotatry table. The slight scollops can be seen in the last photo. This pump uses an o-ring seal at the drive end, so that the ptfe sleeve would run wet.

Next: impeller variants   

[Charles Lamont]

In classical centrifugal pump design theory, vector diagrams are used in working out the best shape for the impeller blades. However, in very small pumps like these, friction and viscosity effects are large, and the textbooks just don't go this small. One of them does say that for small pumps, the calculations will result in blades with a steep backward rake. In general, a backward rake should produce higher pressure and lower flow rate, while forward swept blades should give relatively higher flow rate and lower pressure, with radial blades inbetween.

So far I have tried impellers with: four straight backswept blades,  six radial ones, four forward swept, and three backswept curved ones that are typical of larger pumps.

I have also spent a little while trying CFD (computational fluid dynamics) simulations with SimScale, but have not managed to get decent results, and don't think I can justify investing much more time in following that rabbit hole.

I have so far carried out over 200 runs on the test rig, and the main conclusion so far is that at this scale the blade rake makes much less difference to the performance curve than I expected, typically something like 10% on flow rate at the same speed and head. Much more significant, I have lately found by error, is the axial clearance between the blades and the cover. The spiral volute with a small clearance at the tongue or cutwater performs better than the annulus as I would expect, but the two pump bodies are not exactly comparable, so that may be unfair.

Next: a different type of pump.     

[Vixen]

Hello Charles,

Thanks for posting the results of your water pump experiments. They have appeared at just the right time for me, as I am in the middle of making the water pumps for the two Mercedes engines.

Thanks again

Mike

[Charles Lamont]

Unsure if the little centrifugal pump would have sufficient output at low speed, and with the investigation taking on a life of its own, I decided to see If I could make a regenerative pump with a 1/2" rotor. Initially I did not relish having to make a tiny woodruff-like cutter to mill the vanes. Instead, I was able to cut radial slits and Loctite little pieces of 0.015" brass shim in place, finish turning afterwards. The water passage arcs in body and cover were milled with a 1/8" ball-nose cutter.

On test, I found it does produce more pressure at low speed than the centrifugal pumps, but I think the curve may be too steep for practical use at higher speed. At 2000 rpm the head at zero flow was 45cm, while the centrifugal pumps need 4000 rpm to reach the same pressure.

This pump has been an interesting exercise to make and test, but I probably won't use this type with the Seagull.

Next: the test rig



 

[Roger B]

Interesting work     

I have used gear pumps as cooling water pumps on a couple of engines as I think in small sizes they are less effected by tolerances/clearances. I may be completely wrong   

[Rustkolector]

Charles,
I applaud your study and testing of small pumps designs for model engines. I have to agree with Roger B though that a brass gear pump solves most of small engine cooling issues. It is self priming and therefore provides flow at very low RPM’s and this feature also solves a lot of coolant piping issues. I have used the Jerry Howell designed gear pump design for single, twin, and 4 cylinder model engines with great dependability. Flow is determined by gear width and drive speed. It is reasonably easy to build with readily available spur gear stock and you can add a lip shaft seal. Drive can be direct or belt. 
Jeff

[Charles Lamont]

An early test was to estimate the head needed give the required flow through the system. With fairly short hoses, a 100mm head was enough to give 0.3L/min with cold water and 0.4L/min with water at 48°C. The flow in the first case is laminar, and transitioning towards turbulent in the second, the water viscosity being so much lower at higher temperature.

The second picture shows the radiator and fan under test. There are temperature sensors at inlet and outlet, fed to an Arduino for display. With water entering at 71°C and exiting at 60°, an average of 50° above ambient, and a flow rate of 0.5L/min, the system was putting out almost 400W. Plenty.

Next, an overview of the pump test rig. It is admittedly rather rudimentary, but was initially intended for comparative rather than absolute measurements and the whole business is a sideshow anyway. The supply tank (stainless can) and collecting tub are obvious. The delivery tube outlet is set by ruler to a known height above the supply tank level mark. Depending on the height, the collecting tub is packed up to avoid spills. During a run, the water level in the supply tank is maintained as constant as possible by carefully refilling it from the big jug. After a timed run, the weight of water in the delivery tank is measure on the kitchen scales. The water temperature is noted from time to time, so the density can be used to convert the weight to volume of water. This is a much more accurate measurement than any means I have of measuring the water volume directly.

After a general closeup of the pump and motor, there is a shot showing the hall sensor pickup which is triggered by a magnet in the brass drive flange. This tacho signal is read by the Arduino and used to provide a PWM output to control the motor via a mosfet. The Arduino uses a PID (proportional, integral, and differential) algorithm to set the PWM output duty (proportion of the time that it is on). The four pushbuttons enable the speed to be set in 10s, 100s, & 1000s of rpm, and allow the gain values for each of the PID control parameters to be adjusted for stable speed control.     

To prepare for a run, the supply tank (stainless can) is filled to the mark, the head set by moving the hose up or down the vertical rod. The motor speed is set, and a quick run to get the speed control roughly stable. Switching off the motor means no more pulses, so the PWM duty stays somewhere near. After draining the collecting tub back into the supply tank, a run can be started. Switch on the motor and start the stopwatch (my phone). The fiddly bit is to use the big jug to maintain a constant level in the supply tank. When enough time or water has passed, switch off the motor and stop the clock.
Weigh the water and return it to the jug and dry the tub. Record all the readings for later analysis, and repeat. I generally set the head, and do a sequence of runs at different speeds. I find a reasonable run is either about three minutes before the jug holding arm protests, or, at higher flows, there is about 1300ml in the tub. 

Numerous potential refinements are not difficult to think of, but I do get pretty consistent results. 

[Charles Lamont]

Mike, glad to be of service.

Roger, Jeff - I wanted to play with centrifugal pumps to see if I couldn't make one that would do the job. After all that is what full size engines have, and the challenge being the attraction. I agree with what you say about gear pumps, and have all-along regarded that as an easy way out backup solution.

Charles 

[Zephyrin]

wow, thanks to share all these experiments, many skills throughout the posts for the design of a pump of your "seagull" , lucky engine !

ps
BTW, the big brother of the seagull, the seal engine, do have a centrifugal pump running at camshaft speed.

[Vixen]

Hello Charles,

I have found the size and power of the radiator fan also has a great effect on the cooling system efficiency. A bigger radiator or a more powerful fan reduces the requirement on the water circulation pump. Also plain water is a better coolant than glycol antifreeze. But plain water causes more corrosion problems, so I generally use a mixed 50/50 solution of the two.

Mike

[RReid]

This is all very interesting and helpful for me as a small coolant pump is one of the anciliary bits I've given thought to but haven't yet attempted.

[MJM460]

Hi Charles, great work on those small pumps and your testing.  Nothing wrong with the old bucket and stopwatch for flow if the flow is not too high.  But I guess one of the improvements you considered was a float valve to keep the inlet head more constant.

Did you test the whole curve for the pumps?  Or just shutoff and maximum flow?  Did you see the different curve shapes for the forward and backward curved vanes that are usually seen on larger size machines?

Waiting with interest on the next steps.

MJM460

[Charles Lamont]

Yes, a float valve is pretty high on the list of improvements. I did try a chicken feeder but it could not cope with high flow rates and was too much of a faff anyway.

Here is a set of curves for centrifugal pump 2 so far.

Things to note:

there are a few obviously anomalous points,

there is not a great deal of difference between the impellers except the poor showing by no 4,

I think impeller 4, the curved bladed one, (orange curves) ought to do better, and I think there is something wrong somewhere,

the three 4000 rpm curves (blue) for impeller 1 (straight backswept) show the effect of axial clearance ( but I can't remember how much "large" is)

the curves do show that the forward swept blades (green curves) give higher flow at low head but that it loses out at high heads, as expected,

the green and orange 4000 rpm points have been curve fitted to a parabola, other curves are splines done by the drawing prog,

in all these tests, there is a considerable length of tube involved and it has significant flow resistance. This is OK for comparative tests, but for truer results I should have manometers near the suction and discharge of the pump. I have tried this once so far, and the water in the suction manometer disappeared completely as soon as I switched on. More thinking required. I have also done head-flow tests on the silicone rubber tube, but can't make sense of the results.   

[Charles Lamont]

Here is a graph for the regenerative (aka turbine) pump:

[MJM460]

Hi Charles, that is an excellent exploration of the pump performance.  The regenerative pump is particularly impressive. 

I wonder if the issue with the curved blades is that you have only three blades on this impeller, compared with the four blades on the others.  Mind you, I can understand why you did this, but it would be expected to reduce the performance.  Otherwise the performance curves look very much as expected.

Have to agree on the need for the manometers.  With the inlet side, you seem to have plenty of NPSH, so  I wonder if it is acceleration loss as you had a pretty rapid start.  If you seal the top of the manometer with your finger until flow has started, that would eliminate the measurement of the acceleration loss, and might give a better result.  Otherwise, it is possible that the friction in the inlet tube is too high, so a larger inlet nozzle and tube might work well.  Interesting to think about, but I appreciate it is a lot of work to run the tests.

Thanks for posting all your work, it’s most interesting, and has resulted in a quite effective pump design in a very small size.

MJM460

[Roger B]

That's some very interesting research    Thank you for posting it   

[Charles Lamont]

MJM460, you raise several interesting points.

I agree the pressure produced by the regenerative pump is impressive, especially as I think I am running these little pumps at speeds way below their optimum. However it does not match the reqirements of the cooling curcuit. This is something I have not said much about yet. From the tests I have done, the circuit needs greater flow at lower pressure than this particular pump offers. The centrifugal pumps appear to be better matched.

Friction is an important consideration in pumps this small, and I think that the problem with the curved blade impeller may be that the flow is too restricted near the eye. Observe also that, except at zero flow, the brown curves for the six bladed radial impeller are slightly worse that for either of the four bladed ones, when one would otherwise expect the progression of the rake angles to give it a performance intermediate between the two. I ought really to make an impeller with shorter curved blades, and one with just three or four radial blades. (I could easily remachine the existing ones, but then I would not have the originals for any further testing)

In view of the close clearance requirements and friction, I do want to try some time an impeller with a front shroud (though probably not a wear ring!).

I think the inlet manometer problem is one of too much resistance in the inlet tube. Probably the easiest way to get more suction head is to raise the inlet tank, but a bigger bore hose fitting at the tank outlet might be more practical.

On the matter of the gear pump as a fallback option; not long before things got nasty, I acquired on Ebay from a Russian, a full set of 0.6 module gear cutters. They were cheap enough to buy against an idea of building a clock, if I ever get the time (pun intended). Aaanyway, 0.6mod, 12T, and a 3mm face width would give the output I think I need.