Sunday, 13 January 2008

Testing Compressor designs

There's no way round it, I need to know what makes a good compressor and what makes a bad one. There's lots of stuff in the books about thermodynamic calculations for establishing diffuser vane angles etc., but most of it is just hindsight justification of a proven design.

I need to know exactly how important are the various parameters, such as blade angle, inlet and outlet diameters. Does it matter if I set the diffuser blades at 19 degrees instead of 20? If I increase the impeller diameter by 1mm does it make things better or worse?

This is the setup I made for testing compressor designs.

The casing is just a steel tube, same size as I use for the engine casing. The front is a mockup of the engine internals made of MDF.

You can see one of the test impellers (pink GRP) mounted on the end of the drive shaft. There is no diffuser assembly in this test configuration - I just want to get a base value (worst case) for the behaviour of the compressor.

At the rear of the test rig is a motor, mounted on the end plate. (MDF again). Note the ring of holes in the rear plate that allow compressed gas to pass out of the chamber. These holes can be progressively closed off using plugs, to give a measure of the pressure obtained at different mass flow rates. The number and size of holes roughly appproximates the engine flow cross-section

A simple u-tube manometer is connected to a nipple on the rear plate (not visible in this shot) for pressure measurement.


This first impeller is a simple plano-parallel configuration 60mm OD and 30mm ID.
Blade height is 7mm.

The motor turned the impeller at 15000 rpm and the measured pressure as each hole was plugged was as follows. (Two runs)

Note that 1mm of H2O = roughly one millibar of pressure.


Not as expected.

In theory, the compressor should 'stall' as the mass flow is reduced to zero, resulting in a drop of pressure at the closed-off end of the graph. In other words, there should be a maximum somewhere in the curve. But this test shows a steady increase in pressure as the flow was closed off, resulting in a maximum pressure differential of 100 mbar when fully closed.

Perhaps this is due to the lack of a diffuser assembly?

An alternative explanation is that running at relatively low speeds (20,000 rpm is considered idling) means that the airflow is stable and doesn't stall over the blades even at extreme back-pressures.

Later: -

I inserted a diffuser assembly and tried to repeat the expt. but I experienced trouble with the speed/pressure oscillating when the holes were closed off. Is this the 'surging' mentioned in textbooks? It was very pronounced, even though there didn't seem to be a big improvement in pressure generally.

Unfortunately, the stickytape-and-string diffuser fouled the impeller and destroyed it. I'm going to rebuild a slightly more convenient test rig (I'm having trouble getting it apart to adjust/replace the innards. )

Much Later: -

I redesigned the test rig to make it more convenient to change the impeller and or diffuser.

In this shot of the new rig you can see a test diffuser assembly on the table in front of the test rig. It's just a supporting ring of GRP with a set of vanes glued to it, Quick to make and install.

To cut a long story short, I ran a series of tests on different impellers using curved, angled, and compound blades (ones with a 'lip' to act as a primary axial impeller).

The results were frustratingly consistent. The only things that made a significant difference were the impeller diameter and shaft speed. Otherwise, the ultimate pressure obtained was the same, whatever the blade design.

The motor speed was the major factor. the pressure rises exponentially with speed so that a small change in motor loading (blade width or shape) causes a speed change that hides any effect the blade shape might be having directly on airflow.

The 'surging' effect I had noticed disappeared - I put it down to varying load on the motor causing a speed change resulting in a feedback effect.

Now what? I've run out of ideas.

No comments: