Report on LTI Test Flight #2
|We had a very successful flight test of our new stainless porous diffuser
on Friday, 21 April. The NCAR Electra flew for about 2 hours, at altitudes
of 5.2, 2.5, and 1.6 km asl.
Fabrication of porous stainless
This was the first test of the porous stainless material, after a long period of learning how to machine it and then re-open the pores after machining. Jack Fox (of NCAR's Design and Fabrication Services) tried several approaches, including electrical discharge machining. He has now settled on traditional machining on a lathe, followed by an electro-polishing step in hot phosphoric acid that etches away the thin strips of material that have closed the pores, while leaving the bulk material unaltered. As you can imagine, this process is quite time-sensitive, to make sure one doesn't etch away the surface you want to preserve. The processes seem now to be well in hand, so we can fabricate parts of porous stainless relatively easily. However, since Mott Metallurgical has to manufacture the material for each order, it can still take several weeks to acquire the material for a cone of different porosity or inner diameter.
Flight tests and suction percentages
Timing issues forced us to fly this diffuser without first doing extensive lab tests on its performance, but it worked so well that the risk paid off. At 5.2 km we literally had difficulty finding conditions where it would become turbulent (short of shutting off the suction completely). The flow was laminar over a wide range of suction percentages (porous diffuser suction flow over total tip flow), from 20% on up.
While these low suction fractions are desirable from the standpoint of minimizing the extra pumping required for an LTI, they have a downside: much of the deceleration of the air is accomplished by the suction, rather than simply by increasing the area of the diffuser. Removing a smaller fraction implies that the velocity of sample air exiting the back of the diffuser will be much higher than desirable. A diffuser designed to slow air from 100 m/s to 5 m/s with 80% suction will have an exit velocity of 20 m/s at 20% suction. At this high exit velocity, the (expected) return to turbulent pipe flow in the tubing behind the inlet will still be capable of impacting large particles to the tubing walls. The right combination of area increase and suction percentage is required to slow the sample air enough to prevent particle losses. Since large area ratios generally increase turbulence, substantial suction percentages will be required to achieve low particle losses in practical inlets.
At lower altitudes we were still able to make the flow laminar, but the range of stable conditions decreased slightly. It is clear that this principal works best at lower pressures (higher altitudes), so a test near sea level will be required. However, the stainless diffuser worked far better than any plastic diffuser tested to date. While we can't be sure of the reason for this improvement, one possibility is that in porous plastic some beads break off unevenly during machining, leaving a rougher surface than what we can achieve with stainless.
The major problem encountered on this flight was our inability to achieve isokinetic flow. The needle valves and plumbing (including some small-diameter right angles) were adequate in the lab, where the vacuum source is a large evacuated chamber, but they simply restricted the flow too much for our carbon-vane pumps to pull enough air to get us isokinetic in flight. The plumbing has already been redone (optimized for low pressure-drops) in preparation for our next flight.
Flow alignment and leading edges
However, whenever we approached isokinetic total flow rates, the sample flow became turbulent. This tendency to degrade at high flows does not occur in the lab, even though the flight behavior at lower flows is identical to what happens in the lab. This strongly suggests that the high-flow problem is related to our failure to sample isoaxially from the aircraft. While we made a modest effort to align the probe into the mean streamlines (based on flow modeling done by Cindy Twohy and confirmed with measurements from a co-located Rosemont 858 probe), we did not adequately correct for sideslip. This mis-alignment is compounded by our use of a sharp leading edge on the inlet.
Sharp leading edges precisely define the sample flow and avoid curving streamlines, but they promote boundary layer separation and turbulence at very small degrees of misalignment. Curved leading edges are far less sensitive to misalignment (their ability to keep boundary layer flow attached is why you always see them on aircraft wings), but for us they raise issues of particles bouncing off the blunt surface and the impact of curving streamlines on particle concentrations. For now we plan to stick with the sharp leading edge, but to install a shroud to align the flow into the tip, to see if that helps us eliminate this high-flow turbulence. (We also have a curved leading-edge piece we can quickly glue in place, if the shroud is not enough.) We feel confident we can resolve the issue of attack-angle dependence in our next flight.
Lab tests of the stainless diffuser
Lab tests of this diffuser after the flight have shown it to be almost an order of magnitude better than the best plastic diffuser tested. It is possible to keep the flow laminar with a sample flow as large as 400 lpm with only 20% suction (although this results in an exit velocity of 24 m/s that is far too high). The stainless material is so superior to the plastic that we can consider geometries that were impossible with the plastic, including a smaller tip diameter. Reducing the tip size reduces the total pumping required, and is desirable for situations where you don't need hundreds of liters per minute of flow and cannot support the thousand-watt pumps needed to maintain those flows. Until now, however, reducing the inlet diameter below 1.12 cm has resulted in turbulent flow. Since we can control the roughness of the stainless better than plastic, we plan to revisit some of these ideas that were unsuccessful before.
Modeling of enhancements
Dave Geseler is working with Chuck Wilson at DU to model the enhancement of large particles in the LTI using Fluent. The maximum enhancement for the largest particles is of course the total tip flow divided by the sample flow. For an 80% suction ratio, this is a factor of 5 enhancement for those particles that are so large they move straight to the rear of the diffuser cone and stay in the sample flow. For the configuration already tested, enhancements of 2x - 3x are predicted for particles in the 7 um range. While this may be seen as desirable for size-resolving devices that are often limited by small numbers of large particles (the enhancement is well-behaved and can be corrected after the fact), it will cause a problem for devices like integrating nephelometers that need to measure the optical impact of the largest particles using a bulk sample. A series of pre-nephelometer impactors could help resolve this problem (if you know how much scattering is due to each size range, you can correct for it), but minimizing the enhancements is still desirable. Using a smaller suction ratio will help (at 50% suction the maximum enhancement is 2x), as will reducing the expansion angle of the cone. Both of these imply longer porous elements, which has generally made them more turbulent. Lab tests will be used to explore these approaches with the new stainless material.
Our next Electra flight is scheduled for 3 May. We plan to 1) reduce the pressure drop across the stainless diffuser to reduce pumping power requirements, 2) reduce losses in our plumbing to achieve isokinetic flow, and 3) resolve mis-alignment issues using a shroud and (if necessary) a curved leading edge for the inlet's tip.
We hope then to fly again the week of May 22-26. If we can get the material here in time, we'll try higher-porosity metal, which will have less tendency to plug up from high particle loads and will be easier to clean. We will also try out a surface-mount hot film anemometer that can alert a user when the sample flow becomes turbulent, without protruding into the flow stream as a traditional hot-film probe does. We also want to test a variety of pressure measurements to diagnose clogging of the porous material. If lab tests indicate that a smaller inlet diameter is workable, we may want to fly a different geometry as well.
Finally, in mid-June, as the installation of equipment on the C-130 for the PELTI passing efficiency tests is underway, we want to briefly flight test the final configuration we will use on PELTI. This configuration will incorporate everything we have learned from our lab and flight tests to date, as well as a transition from our current diffuser exit diameter of 1.91 cm to a tube diameter of 3.49 cm. This larger diameter is required to minimize particle losses at bends in the transfer tubing. The PELTI tests will look beyond turbulence reduction, to measure the passing efficiency of the LTI for various particle sizes in sea salt and dust. It is important that we characterize the laminar-flow regime of the PELTI inlet before gong into the field.
We continue to be encouraged that we can produce a tested, working LTI in time for ACE-Asia. There are numerous issues yet to resolve, but they are for the most part tractable engineering problems. We have clearly demonstrated the viability of the porous-diffuser principal, as well as our ability to resolve the problems we encounter.
-Barry Huebert 26 April, 2000 firstname.lastname@example.org