Aero 307 & Gliders

Neiman’s adventures in 307

[interferencedrag]

Libelle flight test: Computational Preparation using xflr5

I decided to make a quick investigation into performance enhancement of the design airfoil of the glider we’ll be flight testing in the coming weeks in xflr5.

Under some (hopefully reasonable) assumptions, here’s what I found:

Blue Line: Clean wing (free transition)

Yellow Line: Forced transition on lower surface only

Red Line: Forced transition on upper surface only

Green: Forced transition on both upper and lower surfaces

The forced transition points (x/c = 0.45 for the upper surface and x/c = 0.5 for the lower surface) were chosen based on the location of the evidence of a separation bubble on the pressure coefficient plots of the clean wing:

In terms of conclusions, these computations suggest that forcing a transition on the lower surface is significantly less important in these conditions than doing the same on the upper surface, particularly at lower speeds (higher angles of attack).

So this data seems to suggest that the most cost-effective course of action performance improvements is to install full-span turbulators at 45% chord on the upper surface and leave the lower surface alone (at a cost of ~$80).

Some limitations to this analysis:

1.  The airfoil as manufactured likely differs a fair bit from the design foil, which could easily effect the location and/or significance of the bubbles, both top and bottom. Contemporary gliders from the manufacturer (Glasflügel) are known to exhibit this attribute (see “A Flight Test Evaluation of the Glasflügel 604″ by Dick Johnson in Soaring). Since we’re not really interested in sanding and reprofiling the leading edges of the wings of the glider, we’ll have to make due with the airfoil sections we have.

2. I’m not totally sure the assumptions I fed into xflr are correct. The polars were all generated using the XFOIL type 2 analysis, which (roughly) holds lift constant and varies angle of attack and Reynolds number in order to model an aircraft in steady flight at varying airspeeds. If the parameters I set this up with were computed incorrectly, these results could be meaningless.

Hopefully the data we collect in the coming weeks will validate (or invalidate in an insightful way) these computations!

[interferencedrag]

Week 6, Day 2:

Today was the first day of project 3, in which we collected wind tunnel force balance data for two finite NACA 4412 wings of different aspect ratios across a sweep of angles of attack.

We added 3 tufts at spanwise locations varying from near the center to near the tip on the suction side of the wings in order to attempt to visualize the progression of stall from root to tip. On the lower aspect ratio wing, we were reasonably successful: the centermost tuft indicated a well-developed stall several degrees of AoA earlier than the tip tuft. In fact, the tip tuft on this wing nearly never gave the traditional fully separated flow stall indication due to interactions with the tip vortex.

Despite the use of the same airfoil section, the higher aspect ratio wing stalled much earlier and, based on observations of the tufts, more thoroughly than the lower aspect ratio wing. This makes sense because the manifestation of the differences in stall angle by spanwise location are a result of the finite-ness of the wing, so since the lower AR wing is in some sense “more finite” than the other these effects should be more pronounced.

[interferencedrag]

Week 6, Day 1:

Today we presented our findings to the class on our attempts to visualize and positively impact flow around the miniature supermileage car. After further analysis of the footage we obtained, we ended up concluding that we were able to delay separation by 6-8% of the length of the car using the larger trip strip configuration, which is more successful than I had expected. The smaller strip had little effect on most of the boundary layer due to the weakness of the disturbance it provided.

Overall, this was an enlightening experience and it was fun to have the opportunity to design an experiment for the water tunnel.

Of the presentations from the other groups, the one that interested me the most related to flow around a matchbox car in the water tunnel, both with and without the windows and rear window removed. They were able to show that, after dye had been left in the bed, it would circulate through the rear window and then out the other windows before entering the freestream rather than becoming entrained in the flow separated off of the roof of the car.

[interferencedrag]

Week 5, Day 2:

Today we collected data on our attempt to push back the separation location near the trailing edge of the supermileage car. While we think we have plenty of footage to analyse, insufficient analysis has taken place for us to draw conclusions as to whether or not we actually had a positive (or negative) influence on the flow features in the area of greatest interest to us from a drag reduction perspective.

After a preliminary review of the footage obtained after the installation of the 1mm trip strip, we were advised to attempt to have a greater impact on the flow by introducing a larger obstacle. The goal was to put up a wall which would energize the boundary layer to a greater depth in order to produce the turbulence necessary to keep the flow attached more consistently further downstream.

Our efforts to produce an appropriately sized obstruction resulted in a half-attached strip of tape added to the downstream side of our existing strip, extending about 4mm into the flow, with the detached portion angled about 45 degrees off of the surface. Here’s a clip from the run completed in this configuration:

[interferencedrag]

Week 5, Day 1:

Today was primarily spent poring over the footage we collected last week in the water tunnel in an effort to develop the best understanding of what’s going on before we decide what to change or investigate further next class.

Once we had the video up on a big screen, the point at which laminar recirculation began became much more clear. Watching carefully, there were several occasions in the videos in which inconsistencies in the concentration of the dye could be tracked as they made their way back upstream after having been entrained in the flow in the laminar recirculation area.

Using these observations in particular, we concluded that laminar separation occured at about 5″ forward of the trailing edge of the body. The next step was to attempt to improve upon this by tripping the flow just upstream of this point with a trip strip.

(photo of trip strip)

We layered up some thin strips of aluminum tape, oriented spanwise, to about 1mm thickness at a location 5.5″ forward of the trailing edge in an effort to accomplish this. Hopefully our test session later in the week will yield positive results!

[interferencedrag]

Week 4, day 2:

We tested the miniature supermileage car today!

As always seems to be the case, we ran into some unexpected complications as we set things up. For example, we hadn’t expected the model to float in the water channel, and some quick modifications our mounting rig had to be made to accommodate this.

Also, since we’re most interested in separation off of the central 2D cross section of the model, it was crucial that the dye be introduced to the boundary layer of the model in the right place. This proved to be problematic because, in order to position the region of interest near the center of the transparent-walled test section, the front of the car had to be well in front of the beginning of the transparency. Essentially, we were flying blind when maneuvering the dye injector.

Fortunately, we got some useful footage out of the session. Here’s a still:

The grid behind grid behind the test section has 1″ resolution. In this frame, it looks like the laminar separation point is about 5″ forward of the trailing end of the model. Once the laminar boundary layer has separated, it appears to transition to turbulent flow around 1-2″ after the initial separation point, and then indications of the turbulent BL “attempting” to reattach are visible about 1″ further downstream. The turbulent reattachment seems to be weakened by the fact that, by this point, the curvature of the model seems to have created a relatively strong adverse pressure gradient.

Using this photo, we think we’ve been able to identify the laminar separation point for the cross section we’re interested in. With this information in hand, we’re going to come back next week and attempt to trip the flow upstream of this location to see if we can delay separation.

We experimented with a few different lighting set ups as the session went on. Unfortunately, the most illustrative dye positions seemed to occur while we were using the worst lighting set up. The gif at the begging of the post is from earlier in the session when the lighting was better.

Although it isn’t clear from any of this still, since it is not time-averaged, we did observe some turbulent reattachment downstream of the laminar separation. Since this process is very unsteady, time averaging is required to capture it in any quantifiable detail. 

[interferencedrag]

Week 4, day 1:

Initially, we had planned on testing the supermileage car model in the tunnel today, but instead we decided to iron out the kinks in our experimental set up and test later in the week.

The plan we settled on for securing the model in the tunnel involves suspending it from some long bolts from crossmembers across the top of the water channel. Fortunately, all of the hardware was available in the lab to make this happen.

Since the mounting points on the model are located on the bottom of the wheel fairings, the model will be upside down when we test.

We also spent some time today prepping the model by retaping over some the pressure taps in the vicinity of the region we’ll be most interested in.

[interferencedrag]

Week 3, day 2:

Today was spent processing our wake data generating a test plan for our next experiment, which we’ve now decided will involve flow visualization on a model of Cal Poly’s current competition supermileage car.

We’ll be running a dye test in the water tunnel. Since the test section is fairly small (only about 6” by 8” in cross section), the blockage will be fairly high, but hopefully this won’t interfere substantially with the features we’re looking to observe.

I think it’ll be most interesting to use the dye to determine the location of flow separation off the back spine of the car (without any side slip). Since the water tunnel operates at relatively low speeds (capping out at about 5 inches per second), the Reynolds numbers will be quite low compared to the ones we’re used to seeing in the wind tunnel. As a result, there’s a good chance that we’ll get to observe a laminar separation bubble followed by turbulent reattachment further downstream.

If this is the case, our stretch goal is to add some trip tape just upstream of the observed laminar separation point to see if we can delay separation. Exciting stuff!

[interferencedrag]

Week 3, day 1

Successful day of testing in the tunnel! We collected some good-looking total energy wake data using the traverse located about 1/4 of the chord length behind the trailing edge.

Since we knew we only had time to make about 3 thorough sweeps, we decided it would be best to run at the same Reynolds number which we used in our wing pressure port tests (Re = 500,000). Our stated interest from the beginning has been performance at large negative angles of attack, so we chose to test at 0, -5, and -7 degrees.

In the interest of time, we used probe position spacings which were larger around the edges of the wake, with a minimum spacing of 1 mm near the area of greatest pressure differential at the center of the wake.

Preliminarily, it looks like the total drag trend based on these 3 data points is something approximating what the simulations and Theory of Wing Sections seem to predict, which is to say that the Cd at 0 and -5 are similar but there’s a noticeable step up in Cd at -7 degrees .

[interferencedrag]

Monday was testing day for the first experiment with the infinite span 4412. We got some reasonably good data, and think we’ve nailed down the location of the inverted stall to within 1 degree AoA. To help visually confirm the stall point, we taped a tuft to the lower surface of the wing at around 25% chord. Unlike the positive stall, we expected that the separation associated with the negative stall would be sharp, and, watching the behavior of the tuft as the wing was reoriented from 9 to 10 degrees AoA agreed with this.

[interferencedrag]

Inverted Flight with the NACA 4412: Week 1b

In order to process our raw pressure tap data into useful lift and drag data, I wrote a panel code. The panel endpoints were placed halfway between the pressure taps along the airfoil section, which, in some areas, results in very asymmetric panels. There’s likely a better way to do this, but, for our purposes, the error introduced by this technique shouldn’t be overwhelming.

One complication which needed to be addressed was the lack of a pressure tap at the trailing edge. Clearly, the two panels surrounding the trailing edge will make meaningful contributions to lift and drag. The workaround we utilized was assuming that the pressure at the trailing edge was the average of the pressures at the two nearest taps.

Due to the particulars of our panel placement algorithm, placing a single (false) port at the trailing edge results in one short, nearly vertical panel in place of the trailing edge. Since the section only has twenty ports, it was desirable to develop a mechanism to replace this single erroneous panel with two panels of proper size and orientation. Rather than writing an exception into the panel placement code, I found that duplicating the port at the trailing edge accomplished this. 

(With this strategy, the pressure along each of the panels closest to the trailing edge is then assumed to be the pressure at the trailing edge, which, itself, is an assumption based on the nearest two pressure ports.)

Hopefully this scheme will return reasonable lift and drag coefficients once that portion of the code is implemented!

[interferencedrag]

Inverted Flight with the NACA 4412: Week 1a

We’ve been tasked with designing a low-speed wind tunnel experiment involving an “infinite span” (wall-to-wall) NACA 4412 section. 

The 4412 was a staple of mid-20th century aircraft design, particularly in the incompressible regime. This is likely related to the wealth of knowledge surrounding the section’s performances and handling properties. In particular, the section found substantial use in the single-engine high wing and biplane segment, appearing on such popular aircraft as the Aeronca Champion & Chief, Stinson 108, Stolp Starduster, and American Champion Citabria (an aerobatic development of the Champion).

The Citabria is a two-seat, tandem, high-wing aerobatic trainer which has been produced (in some form or another) since 1965, boasting a production run of over 5,000 units. In its role as an aerobatic trainer, with a sustained load factor certification of 5 positive and 3 negative g’s, the aircraft often provides a platform for pilots to learn the difficult mental exercise of inverting control inputs in order to sustain inverted flight. 

In the fall, we had the opportunity to test the 4412 from small angles of attack up to positive stall. In an effort to broaden our understanding of the performance of the section, with particular interest in the application of the airfoil to the Citabria, we decided to investigate the location of negative stall and the lift (downforce) generated up to this point.

With this idea in hand, we settled on a testing plan which would best utilize our tunnel time while collecting useful data in this regime.