Otdfbt

I'm writing an essay at school about why there is so much variation in winglet design. I am analysing all the primary types of winglets used today in the aviation industry. I'm constructing them in Solidworks (CAD software) and then running them through the in-built CFD. The winglets which gave me the lowest drag were those of the 737MAX. I then went on to design my own winglet and see how it compared - I took on a sort of compund spiroid/MAX winglet mix - according to the numbers fom Solidworks, the drag force with this winglet is 8% lower than that of the MAX winglet.

Now, OBVIOUSLY I am not better than the a team of hundreds of Boeing engineers with this design I stumbled upon, so I was wondering what the reasons would be for Boeing/an aircraft manufacturer not to use this type of winglet. I understand the consequences of weight on the wing root moment/structural reinforcements, etc... but I feel an 8% reduction in drag would outweigh the weight addition from this winglet?

Test parameters:

The wing is the same as that of a 737NG, with the winglet stuck on the end

• TAS: 280kts


• Air density: 1.2kg/m^3


• Airflow: Incident at normal to fuselage direction


• AoA: ~1.5 degrees

Images of the CFD and winglet itself are attached

The first possibility is that that the CFD in your cad isn't as sophisticated as the software used by boeing's engineers. Which means that your design may have flaws that don't show up in your software but do in Boeing's (or not even in there but do in a windtunnel).

Second is that I only saw one flight configuration being tested. Airplanes do more than just cruise at high altitude and the winglets must be good in all conditions. They must especially not affect the stall and spin behavior in a negative way.

Your winglets look kind of flimsy and I'd be worried about them getting ripped off or deforming in turbulent conditions. Those deformities would affect how they perform, possibly for the worse.

First off, awesome question and great investigation! This kind of let's-see-what-happens inquiry will take you far should you decide to pursue aerodynamics at an advanced level (and, of course, in other pursuits). Not so long ago, I had to write a similar report: lacking the resources and knowledge of the aerospace giants, I, too, wondered why I could seemingly invent designs that on the face of things appeared vastly superior to theirs. I thought that I had winglets down cold.

Then I went to work for Boeing and started talking to the aerodynamicists. I started graduate study of aeronautics. Turns out, unsurprisingly, that there's a lot you can't get out of undergraduate textbooks and publicly available data. While I obviously can't be exhaustive here—and probably won't even answer your question to the letter—I can give you a few things to think about. To be clear, I wouldn't go much further than you have with your modeling and simulation, but if you'd like some discussion points for your paper here are some in no particular order. I've made some assumptions about your level of knowledge, so please forgive me if it's patronizing and ask me if you need clarification.

The fidelity of your baseline model

The winglets...were those of the 737MAX....The wing is the same as that of a 737NG.

On what data did you base your model? The wing of a 737 is not a simple matter of an airfoil, some taper, and some twist. I notice that you did not include nacelles/pylons or flap-track fairings. The design of a production winglet is heavily tied to the integration of the overall wing design, including all of the extra components hanging off of it.

The reason that the 737 MAX winglets are effective

The 737 MAX uses what's branded as the Advanced Technology (AT) winglet. We know that a well designed wing extension is more aerodynamically efficient than a winglet. But the 737's wingspan must stay within certain limits in order to operate with the same ground infrastructure as previous models, so a winglet is a good solution. But what if we could have a little of both? Well, the AT winglet does exactly that:

The lower winglet is configured such that upward deflection of the wing under an approximate 1-g flight loading causes the lower winglet to move upwardly and outwardly from the static position to an in-flight position resulting in an effective span increase of the wing.

So to really understand the efficiency of the AT winglet, you'd need to model this deflected geometry.

The other contributing item to the effectiveness of the AT winglet is its natural laminar flow:

On previous winglets, the drag due to friction from the airflow over the winglet is one of the main detractors from efficient airflow....this is solved by Boeing using detailed design, surface materials and coatings that enable laminar – or smoother – airflow over the winglet.

The flow regime you're modeling

The AT winglets are most effective as their efficiency is aggregated over long, high-speed, high-altitude cruise legs. All you've given is a true airspeed, but for this kind of analysis of transport aircraft the Mach number is much more important. You haven't provided an air temperature, but from the density you've given it looks like this simulation is at sea level, which means your Mach number is not high enough. But this in fact might partially explain your results. Observe the drag curve:

In general, a spiroid winglet like yours reduces the induced drag at the expense of some parasitic drag. As you can see, we can afford some extra parasitic drag at lower speeds because the induced drag dominates.

If I were to make one suggestion, it would be to run your simulation at a realistic Mach number (around 0.8) and see what happens. But beware...

The limitations of your CFD software

We are getting to the point where CFD, when implemented well, is quite good for modeling aircraft performance in cruise flight. Much of the wind-tunnel testing for large aircraft these days is focused on high-lift and maneuvering conditions, where CFD falls much shorter. Of course, we always want to validate our CFD in the wind tunnel for all flight conditions, but for well-understood configurations in cruise the results often match well in terms of calculating overall performance. But the "when implemented well" caveat is key. I personally don't have experience with SOLIDWORKS Flow Simulation, but it looks like it's designed to be a general-purpose CFD software, so I would not trust its results too much for large, complex, high-speed simulations such as the ones required for this analysis.

In particular, there is the issue of turbulence. Not in the sense of unstable air that jostles an airplane around, but in the sense of chaotic flow over the aircraft's surface. So chaotic, in fact, that no computer in the world can accurately model the motion with a sufficiently short computation time. Instead, we use turbulence models that try to approximate what's happening in a way that can be solved quickly enough. SOLIDWORKS uses the k-epsilon model, which is popular for general-purpose software but might not be the best choice here. In particular, notes Wilcox,

Even the [k-epsilon] model's demonstrable inadequacy for flows with adverse pressure gradient has done little to discourage its widespread use.

As flows over airfoils are quite influenced by adverse pressure gradients, I would exercise caution. I can tell you that Boeing makes good use of the Spalart–Allmaras turbulence model in conjunction with detached eddy simulation (Spalart is an employee). But choosing the correct CFD implementation for a particular problem is a nuanced process requiring a great deal of judgement and care.

First of all great analysis! I am not expert in aerodynamics but from the little I know, aircraft are a compromise. When you design a plane you have to make shapes that are possible to manufacture, don't cost too much and are strong (and compliant to the regulations). Last but not the least, you have to test the drag over multiple phases of the flight and in multiple configurations (flaps/slats): not only in cruise and analyse how this new design affects lift. Plus I don't know how accurate Solidworks CFD is: consider that the wind tunnel is still used because CFD is not perfectly accurate.

One thing I have forgotten: structures have a weight. If to make a complex shape you have to use stronger material, then the aircraft will have a higher weight that cancels out your gains in drag. What is the difference in drag between no winglet and B737 MAX winglet?

I would say the biggest thing is that the computer model has to be validated with wind tunnel results. As anyone truly familiar with the climate debate knows, computer models trying to simulate extremely complex phenomena often at some point start to diverge from reality and you still need to test in the real world to validate or falsify the model to really find out.

I would bet then that your version performs well because the modelling is inadequate at reproducing some subtle effect, and that if you tested it in a wind tunnel you'd be disappointed.

The structural analysis is a bit similar. Computerized finite element analysis done for the structure for strength and endurance has its own limitations and still has to be validated with a run in an endurance test rig, and very often the computer's predictions are wrong. This is why it's important to get the structural endurance rig testing completed as early as possible in the production program to minimize the patches that have to be done in service when testing uncovers inadequately modeled structures.

While all other points raised are valid, it is also important to go back to basics and first understand what winglets are actually supposed to do, and what drives the performance of an aircraft:

Winglets do not reduce 'profile drag', or the drag force that exists when there is no lift being produced. They more than likely will increase profile drag slightly. If your reduction in drag is coming from a reduction in profile drag, this is immediately suspect, bordering on perpetual motion machine crazy. Winglets generally reduce 'induced drag', which is the portion of your total drag force that is produced when the wing starts creating lift. How much induced drag is produced depends on the efficiency of the wing and its span, or an 'effective span' calculated by multiplying the physical span by the wing's efficiency. Winglets are supposed to increase the wing's 'effective span' without exceeding physical span limitations. This means that the reduction in drag will be higher the more lift you need to produce, or the higher your angle of attack. Because large airliners are typically designed around a few main cruise conditions at relatively low angles of attack, they typically claim improvements in the order of 2-4% on fuel efficiency from the addition of winglets.

It is important to compare drag values at equal lift values rather than at equal angles of attack. At a certain angle of attack, winglets can be effective by either reducing the drag, or increasing the lift (usually by sneakily increasing the physical span a little bit), or both. The important performance parameter that you should be plotting is the drag polar, where lift is plotted against drag. An aircraft will cruise at whatever angle of attack is required to maintain 1g, which will depend on its weight and lift characteristics. Angle of attack does have large implications on the design, but it does not drive performance. When comparing different configurations, overlaying drag polars will tell you everything you need to know, including how the profile drag values differ. You should run your simulations at differing angles of attack and plot the resulting lift coefficients against the resulting drag coefficients. A quick check on the case you have already run would be to make sure your drastic reduction in drag isn't accompanied by a drastic reduction in lift. Because your main lifting surface has not changed, this should not be the case.

Everything everyone else is saying about the occasional inadequacies of cfd, especially 'quick and easy' solutions built in to CAD packages is true, but I would still not expect such a drastic difference between two relatively similar configurations run on the same software. I would make sure absolutely everything else in the simulations of the different configurations is the same, check the changes in profile drag and lift, and do some flow visualisation to try and understand what is happening in the simulation to cause the drop. Then I would check if the same thing happens at a higher Mach number. Ideally you would want to check other software packages, but i understand this may not be possible and may be outside the scope of your essay.