My diaphragm piston air engine with rotary valve

I made an air engine with a diaphragm piston and a rotary valve. It was inspired by a video from Robert Murray-Smith, but improves on the design.

Here is a video clip showing my engine:

If you want the CAD files, you can get them on github.

Compared to making my Wig-Wag, this engine was a lot less work, despite the unconventional design. I went from the idea in my head to a working engine in a single day thanks to the power of 3d printing. And that includes several iterations of some parts when my first attempt didn't work.

The "diaphragm" is a finger from a rubber glove. The piston fits inside the finger, and the base of the finger is clamped to a tapered surface on the outside of the cylinder by a nut, so that it forms a perfect seal. This means the piston can be a very loose fit inside the cylinder bore without preventing it from sealing. The diaphragm rolls up and down the outside of the piston as the engine runs.

The "rotary valve" is an angled hole in the crankshaft that alternately connects the top of the cylinder to the exhaust and inlet ports.

The "pipes" are integrated into the base plate.

The first improvement of my design over Robert's is that my rotary valve is integrated in the crankshaft instead of being a separate component geared off the crankshaft.

The second improvement of my design is that there is only one pipe going from the valve to the cylinder, which saves wasting energy on pressurising the pipe that isn't connected to anything.

Although the diaphragm means that the piston seal is totally leak-free, my rotary valve doesn't work anywhere near as well, it leaks quite badly.

(A further improvement would be to move the crankshaft much closer to the cylinder so that we don't even have one pipe to pressurise - I have an idea to do this that involves putting the crankshaft at the top of the cylinder and connecting it with a funny-shaped con rod, maybe I'll try that one day).

I find that a good way to make 3d-printed flywheels is to put radial holes in them and then screw in bolts. This has the great advantage that you can screw them in/out to finely adjust the balance. (Not that I bothered for this engine).

The best way to put threads in 3d-printed parts is to design the hole to the nominal size of the tap drill for the thread (so for M6 this is a 5mm hole), then tap it with a cordless drill. The plastic shrinks back in once it's been tapped, which is useful for a light self-locking thread, or if you don't want that, let the plastic cool down and then re-tap it a couple of times to free it up.

This is less hassle in CAD than modelled threads, and less hassle in post-processing than heat-set inserts, although I admit that heat-set inserts look cooler. It feels like tapped threads would be weaker than inserts, but they seem perfectly fine in practice, and they have the benefit that you can make the threads as deep as you can tap, versus inserts that come in a fixed size.

This is what Robert's engine looks like:

We see it running in close-up in his video. He is running it off a vacuum cleaner since he doesn't have an air compressor, but the principle is the same (your supply pressure comes from the atmosphere, and your exhaust goes into the vacuum cleaner). His engine runs remarkably slowly and doesn't even appear to slow down significantly on the return stroke, mine definitely has too much friction and inefficiency to run that slowly, even with my much heavier flywheel.