CAD Exercise: Mustang Outer Mold Line

Help Support Homebuilt Aircraft & Kit Plane Forum:

Well-Known Member
One thing I do miss about no longer being an engineer is designing stuff in CAD...so I'm drawing up a Mustang in Solidworks for fun!

This will actually be my second attempt as my first was based on drawings from some Czech scale model magazine. This iteration will be based on some factory drawings I was able to find.

Objectives:
- Create an accurate CAD model -> within 1 inch of original.
- Investigate sheet metal flat-wrapping techniques. The Mustang was designed for flat wrapping and has a remarkable low number of compound-curved surfaces.
- Fine-tune a workflow for drawing aircraft outer mold lines (OML), which is the outer shape of an aerodynamic vehicle.
- Help other members with 3d CAD techniques.

Not Objectives:
- Aerodynamic design; willing to discuss concepts but don't want to talk numbers
- Structural design; only thing I'll say is that the Mustang has a monocoque design. So all you have to do is figure out the loads, size skin thicknesses accordingly, then add stiffeners/longerons/spars to fine tune high stress areas.

I ask that we remain on topic and limit questions to the subject at hand.

Starting off!

My first step is to gather reference material and define reference lines. I usually use the nose as the origin and the thrust line for dimensioning. Front to back profiles (X-axis) are called Fuselage Stations, Centerline to wingtip slices are called Butt Lines, and those vertical are called Waterlines.

My next step is to create master side and top drawings. Pretty much all profiles are dimensioned off of these sketches. The plus is that because Solidworks is parametric, I can merely edit these drawings later on if I want to tweak some dimensions and the entire 3D model will update itself. Beats redrawing everything and/or getting everything perfect the first time.

The Mustang confused me at first. Because the plane originally had an Allison V12 engine rather than the Merlin, adjustments were needed to tweak the aerodynamics. This included angling the thrust line 1.25 degrees down, probably to give the plane more of a pitching moment for stability. The problem is that they didn't angle it from the front of the plane, but rather 41 inches back (probably the center of mass of the engine). Makes sense, but made my original Czech based main drawings worthless.

So this is where I was before I started over. I'm drawing the P-51D model with the bubble canopy, but the D model is remarkably similar to the preceding B/C model. North American tried to reuse as many jigs as they could, so only a couple panels rear of the cockpit are different. Everything below the canopy sill is essentially the same airplane.

For the first go, I didn't do much work-flow planning. I basically drew the main drawings, drew some profiles, created some surfaces, and then repeated. While this will create a model in the end, it adds a lot of needless steps that makes the CAD file HUUUUGGEE! Not to mention it's almost impossible to edit later on.

So this time I'm making most of the drawings, then making surfaces, and finally knitting those surfaces to form a model....It's like origami, really!

BJC

Well-Known Member
HBA Supporter
One thing I do miss about no longer being an engineer is designing stuff in CAD...
Being an engineer is like being a felon; once you are one, you always will be one, whether you deny it or not.

BJC

Jay Kempf

Curmudgeon in Training (CIT)
Are you using all boundary surfaces that will be perpendicular to the center plane to control symmetry? I have had a lot of trouble getting fair surfaces that way. But the P-51 is actually has a pretty tight fineness ratio so it could work out. Will be interested to see how you deal with the normal hard parts like the cooling inlet and the fairing between the tail boom and the fin. The only other difficult part will be the interface with the canopy and screens.

The location of the thrust line tip anchor is most likely just where the most convenient place to tip the engine is located like at the front engine mount location. They probably just shimmed up the rear engine mounts. Have run into these sorts of things before screwing up an entire loft just for a thrust line change. It's annoying to blow up and re-anchor a ton of dimensions and geometry but that's why people pay me to do this stuff when their grad students can't recover from a simple change. What I have been doing recently is an old ship lofting technique. You put a centerline off of the main inboard section sketch say above the canopy a foot. And dimension every thing from there. That way the location of all bulkhead curves have a position reference to another axis. Seems like extra work but it isn't if you have to tweak something. Imagine it like a laser transit line off in space for reference. I think this is why you see some people with their zero fuselage station at the firewall. Some at the spinner plane and some at the tip of the spinner. The firewall one means you can change all the fuselage lofting independent of the cowl and engine for drafting purposes. I could debate either way.

TFF

Well-Known Member
Remember the leading edge bump out for the gear is different from the B to D. Very cool you can do it.

flyboy2160

Well-Known Member
Being an engineer is like being a felon; once you are one, you always will be one, whether you deny it or not.

BJC
Hey.....why can't we make it something positive, like being Marine.

BJC

wsimpso1

Super Moderator
Staff member
Log Member
Oh, an engineer can be an ex-engineer. You are still an engineer if you retire from an engineering job... Move far enough into management in some outfits, and something changes that makes you incapable being an engineer anymore. Funny about that, I know some managment wonks who actually got back their ability to do engineering again after they retired...

Billski

Well-Known Member
Visited WestPac, a warbird restoration shop, yesterday. I tagged along a tour and they let me run my grubby hands and get close to a P-38, B-25, and a couple P-47s.

Looking at the P-47 adjusted my approach a bit. Most, if not all, of the outer panels are compound curved. Granted, the P-47 took more than 3 times the man-hours (9100 vs 2700hrs) to build and was almost twice as expensive ($113k vs$85k). The Thunderbolt was also technically a prewar design (by only a couple months), so Republic was probably not as concerned with producibility as North American was with the Mustang.

The question is if it's worth it to compound curve panels? Flat wrapped panels can have be cut, have rivet holes drilled, and then "wrapped" around formers on the jig. The process is a bit more complicated with "bulged" panels, but the process is easier if you're stamping out thousands of them on a die. Also interesting is that the P-47 was completely flush riveted while the F7F Tigercat, the fastest US Naval fighter of the war, used raised rivets on everything rear of the wing. Priorities...

Anyhow, I finished the "grid" of 2d sketches and started recreating the boundary surfaces, as depicted by the pictures below. I'm roughly back to where I was before restarting.

Are you using all boundary surfaces that will be perpendicular to the center plane to control symmetry? I have had a lot of trouble getting fair surfaces that way. But the P-51 is actually has a pretty tight fineness ratio so it could work out. Will be interested to see how you deal with the normal hard parts like the cooling inlet and the fairing between the tail boom and the fin. The only other difficult part will be the interface with the canopy and screens.

The location of the thrust line tip anchor is most likely just where the most convenient place to tip the engine is located like at the front engine mount location. They probably just shimmed up the rear engine mounts. Have run into these sorts of things before screwing up an entire loft just for a thrust line change. It's annoying to blow up and re-anchor a ton of dimensions and geometry but that's why people pay me to do this stuff when their grad students can't recover from a simple change. What I have been doing recently is an old ship lofting technique. You put a centerline off of the main inboard section sketch say above the canopy a foot. And dimension every thing from there. That way the location of all bulkhead curves have a position reference to another axis. Seems like extra work but it isn't if you have to tweak something. Imagine it like a laser transit line off in space for reference. I think this is why you see some people with their zero fuselage station at the firewall. Some at the spinner plane and some at the tip of the spinner. The firewall one means you can change all the fuselage lofting independent of the cowl and engine for drafting purposes. I could debate either way.
Yes, most of the profiles that make up the boundary surfaces are perpendicular to the center plane so I can mirror the other half of the plane once I'm done. Boundary surfaces are powerful, but can be tricky. I've found that having no more than 3-4 sketches in each direction to form a "grid" works best. Any more and the surface becomes tortured and unpredictable. After that, it's just editing the edge conditions like tangency, if the surface is attached to another, or being normal to profiles.

Jay Kempf

Curmudgeon in Training (CIT)
Boundary surface can be a bit "prickly" as one client used to say. Patching in multiples always causes ripples. I really like to get the fuselage loft, both sides in one shot. If I have to hack hunks out of it to put other parts in then I do. Most times I end up cutting the whole thing in half and then mirroring it anyway. Depends on how many bulkheads, how many guide curves, how many degrees of freedom you allow or don't. Been pretty successful making editable smooth lofts for a while now. Try making a blended wingtip with winglet sometime with twist and sweep. That's the advanced course. Cooling channels can be fun too.

Andy_RR

Well-Known Member
Are you using all boundary surfaces that will be perpendicular to the center plane to control symmetry? I have had a lot of trouble getting fair surfaces that way.
I use a normal to profile constraint for both surfaces and section splines and find that this is sufficient to yield a smooth transition over the plane of symmetry.

I do almost all my surface modeling with boundary surfaces. The thing to bear in mind is that they need to be as close to a "rectangular" patch as can be achieved. The other thing to use is the surface trim tool to generate daughter edge geometry for things like canopies etc. i.e. surface all the way through the cockpit and then cut away with a projection surface trim or some other geometry and use that as the starting point for canopy (or WHY) geometry.\

These are a few of my models here - all Solidworks:

Andy_RR

Well-Known Member
Try making a blended wingtip with winglet sometime with twist and sweep. That's the advanced course. Cooling channels can be fun too.
Here's my secret sauce for generating wingtips - all boundary surfaces. I start as follows:

1 - make a rectangular boundary surface through the chordlines with the twist distribution configured (3DSketch - all chordlines at length of longest required, or longer)
2 - surface trim to the desired planform with a 2D sketch profile
3 - thicken to trailing edge thickness
4 - delete all surfaces but the trailing edge (turns solid into a surface)
5 - profile the tip of the trailing edge surface to suit using a 2D sketch profile linked to the surface edge geometry
6 - add in all your aerofoil profiles as 2D sketches (usually based on planes set up in the original 3DSketch for the twist distribution
7 - boundary surface as much of the main panel as you can in one hit.
8 - boundary surface the tip in a curvature axis perpendicular to the main plane to generate the desired wingtip - use curvature continuous constraints for all geometry
9 - hack out a rectangular-ish notch with surface-trim at the leading edge where the wing panel and the tip panel don't quite meet
10 - boundary surface this notch based on the remaining surface edge geometry.

(all surfaces should be curvature continuous constrained as should any intersecting spline data)

Here's the results:

Well-Known Member
Been busy with job training, so I haven't made much progress with the Mustang. It pretty much looks the way it did in the pictures above with minor tweaks.

Boundary surface can be a bit "prickly" as one client used to say. Patching in multiples always causes ripples. I really like to get the fuselage loft, both sides in one shot. If I have to hack hunks out of it to put other parts in then I do. Most times I end up cutting the whole thing in half and then mirroring it anyway. Depends on how many bulkheads, how many guide curves, how many degrees of freedom you allow or don't. Been pretty successful making editable smooth lofts for a while now. Try making a blended wingtip with winglet sometime with twist and sweep. That's the advanced course. Cooling channels can be fun too.
Boundary surfaces are pretty easy if limited to 3 to 4 sketches in each direction and define edge tangencies. So a max of 4 guide curves and 4 profiles. It's not necessary to loft both sides of the fuselage in one shot the boundary surface is set to be "normal to" the profile on the mirroring plane. Saves a lot of processing power, with minimalism being one of the goals of the project; trying to accurately draft the aircraft in less than 200 features. Also makes rebuilding after any tweak A LOT quicker, which can suck the fun out of modelling stuff.

With designing an airplane with sheet metal construction, I've found it's helpful to split up and generate the fuselage panel by panel. Takes a bit more planning, but it ensures that as many panels as possible are flat-wrapped and easy to fabricate.

For composite designs, I've found it's still helpful to split the surfaces up into smaller sections for management sake. Trying to use too many profiles and guide curves leads to a lot of unpredictability...

For the late model Bf-109 model below, which I consider sufficient to begin structural design, all of the surfaces between the engine cowling and the tail section are flat wrapped. Which means those panels can be made without sheet metal shaping equipment like English wheels

I use a normal to profile constraint for both surfaces and section splines and find that this is sufficient to yield a smooth transition over the plane of symmetry.

I do almost all my surface modeling with boundary surfaces. The thing to bear in mind is that they need to be as close to a "rectangular" patch as can be achieved. The other thing to use is the surface trim tool to generate daughter edge geometry for things like canopies etc. i.e. surface all the way through the cockpit and then cut away with a projection surface trim or some other geometry and use that as the starting point for canopy (or WHY) geometry.\

These are a few of my models here - all Solidworks:

View attachment 66669 View attachment 66670View attachment 66671
I actually rarely use splines: way too uncontrollable. Instead, I split complex curves into conical curves. The tangency is based on where the shoulder point is placed and the "curviness" is controlled by the ρ value. This practically eliminates all eyeballing in tweaking curves, something that one can spend hours on on splines.

Also helps that using conics in drafting was first used on the P-51....according to Peter Garrison. Some profile curves practically snap into place if you assume the ρ (0.5) value that North American used.

I'll maybe take a whack at a winglet during the weekend. Most likely what will happen is that I would tackle it with two boundary layers: the top/inner and the lower/outer surfaces.

Last edited:

Well-Known Member
So regretfully I've been cheating on the Mustang with an equally sexy airplane. Using the same concepts I've been using with the Mustang, I drew the whole thing up in 4-5 hours.

As with the Mustang above, please be aware that no aerodynamics design has taken place other than selecting actual airfoils for the wing and tail. Though if need be and calculations have been performed, it's as simple as tweaking incidence angles and twist.

Also, the belly pan and tail need some work.

The process:

Setting up reference images.

Creating grid of 2D drawings, all are interdependent on each other.

Another view of grid with wing surfaces.

Semi-completed model top

Semi-completed belly view

I can clarify on specifics if anyone is wondering how a particular feature is created. Now back to the Mustang!

Highplains

Well-Known Member
About twenty years ago, I had the opportunity to closely inspect a 109 that was undergoing restoration in Fort Collins Colorado. Really an incredible design from the mid-30's. They had just received a rebuilt Daimler Benz engine that had parts that were allegedly shoved out the back door of a certain automaker. As the post above implies, it was constructed of simple conic sections, as was the earlier 108. So I found this old video on the building of a 108 in the factory and other than a few castings and forged parts, rather basic in construction.

Aesquire

Well-Known Member
The Mustang was designed with conics from the start. It was the first production airplane laid out entirely as 3 dimensional coordinates, before metal was cut. Even the compound curve sections were simply defined as elliptical or conic shapes.

Construction tricks like building the fuselage in halves and installing everything possible before joining the halves together were ideas taken and refined from observing British and German practice. The lofting, otoh, was a first.

Highplains

Well-Known Member
Well the guy that wrote the book (literally) was Roy A. Liming, Head of Engineering Loft Mathematics, North American Aviation. His book "Practical Analytic Geometry with Applications to Aircraft" was published in 1944. Many of the sketches in the book are rather equine like. While I have seen it posted online, I bought a copy some time back.

Well-Known Member
Anybody with actual questions or feedback on the CAD?

Anyhow, putting off the scoop and front cowling, I pretty much finished the "turtledeck" on both the B and D models.

The plane has rather simple but elegant lines if you really think about it...

HBA Supporter
Log Member