3D Printed 64mm MiG 21
- Thread starter telnar1236
- Start date
Inq
Elite member
Do you use any other plastics even in small quantities... ABS, TPU, TPE, PC... etc?
I was thinking about a rising rate leaf spring might be useful. In ABS it would be quite resilient and PC would be able to make thinner and accomplish the same thing. They would be a whole lot lighter and than rubber bands, hinges and hooks... I would think. With your expertise in CFD, I'd imagine you have (or certainly could handle) non-linear FEM. You could set up cases for parked to any level of "bad" landing you think you'd be capable. (Me... I'd need air-bags and ejection seat) Although I think the FEM would be fun, I'd probably still just WAG it and permute it experimentally.
Obviously aspects of mains versus nose gear, lateral versus longitudinal, but you might even incorporate the gear door as part of the active spring... Something like this.
I was thinking about a rising rate leaf spring might be useful. In ABS it would be quite resilient and PC would be able to make thinner and accomplish the same thing. They would be a whole lot lighter and than rubber bands, hinges and hooks... I would think. With your expertise in CFD, I'd imagine you have (or certainly could handle) non-linear FEM. You could set up cases for parked to any level of "bad" landing you think you'd be capable. (Me... I'd need air-bags and ejection seat) Although I think the FEM would be fun, I'd probably still just WAG it and permute it experimentally.
Obviously aspects of mains versus nose gear, lateral versus longitudinal, but you might even incorporate the gear door as part of the active spring... Something like this.
Inq
Elite member
Forgot to elaborate on the TPU/TPE. They might work great in an air-spring configuration... might make a great compression spring and replace something like your coiled spring in the F-104. You could even poke holes in it after printing to act as control orifices for damping like a air-shock-absorber. I've noted my tires in TPU do that naturally as I print single wall and the layering is rather porous. You squeeze them and they whistle. Perfect as shock absorption in my tundra tires.
This same wheel has now been on three crashed planes. Hard enough that I bent 1/8" music wire legs! The ABS spokes are quite fine at 0.8mm thick and yet they've taken a lot of abuse!
I did it intentionally to test the ABS wheels... That's my story and I'm sticking to it!
This same wheel has now been on three crashed planes. Hard enough that I bent 1/8" music wire legs! The ABS spokes are quite fine at 0.8mm thick and yet they've taken a lot of abuse!
I did it intentionally to test the ABS wheels... That's my story and I'm sticking to it!
telnar1236
Elite member
telnar1236
Elite member
I have looked at designing plastic springs a bit. My initial concept for the main gear on my F-104 used ABS for a leaf spring. The problem is that the stiffness of the plastic was too high relative to its yield strength, so either the gear wouldn't deflect enough to absorb much energy on softer landings, or it would break on heavier landings. I could have made the spring longer to get around the stiffness, but then you run into packaging issues.
Right now, I'm experimenting with PETG since it is less stiff than ABS and has better layer adhesion to try and get suspension in the nose gear on my F-104 since there isn't really space to package conventional springs.
If you have any recommendations for a free FEA software, I'd love to hear them. I did some looking a while back, but the best I could find was LISA which didn't offer enough nodes in the free version to do much of anything.
telnar1236
Elite member
I want to experiment with TPU dampers, but I think you would still need a more conventional spring to return it to place (and the springs on the F-104 need an absurdly high rate for such a small footprint - I actually twisted three springs together in parallel to get it to work). Even with thicker walls, PETG is too porous so you would just lose all the air and have no spring force after a while. For this MiG, the tires are too small to do much of anything - they're most there to look cool. Pretty impressive that those wheels have survived three crashes. I hope the wheels I've designed hold up that well.Forgot to elaborate on the TPU/TPE. They might work great in an air-spring configuration... might make a great compression spring and replace something like your coiled spring in the F-104. You could even poke holes in it after printing to act as control orifices for damping like a air-shock-absorber. I've noted my tires in TPU do that naturally as I print single wall and the layering is rather porous. You squeeze them and they whistle. Perfect as shock absorption in my tundra tires.
This same wheel has now been on three crashed planes. Hard enough that I bent 1/8" music wire legs! The ABS spokes are quite fine at 0.8mm thick and yet they've taken a lot of abuse!
I did it intentionally to test the ABS wheels... That's my story and I'm sticking to it!
View attachment 234190
Inq
Elite member
telnar1236
Elite member
The MiG 21 is mostly put together. I just need to print and install the canopy and attach the servos to the elevons. Unfortunately, it's looking like it will be too windy to try and fly it tomorrow, but it should be in a flyable condition at the very least.
telnar1236
Elite member
Inq
Elite member
I've only experimented with one roll of PETG. I didn't like because nothing sticks to it... glue or paint. BUT for your situation, I think it would be one of the best... only second to PC. Ever try breaking a CD/DVD?
I used Nastran professionally. Contract required it.
For my own use, I got Mecway when it was at version 0.9 and dirt cheap. The upgrades are free for life and its now up to version 18. It's also not cheap. It does all the things I've ever needed... Static, non-linear, dynamic, modal, buckling, orthotropic, anisotropic, and several things I've never used. I've also done some very large models even non-linear. - https://forum.flitetest.com/index.php?threads/inqd-turbo-storch.71382/page-4#post-737352
I've not checked it out, but I thought Fusion 360 had a FEM module.
telnar1236
Elite member
Yeah, Mecway is definitely pricey for something I would only get to use occasionally. It looks like sticking with hand calcs and experimentation is still the way to go. Fusion 360 does have an FEM module, but you have to use the subscription version, where I use the free version.I've only experimented with one roll of PETG. I didn't like because nothing sticks to it... glue or paint. BUT for your situation, I think it would be one of the best... only second to PC. Ever try breaking a CD/DVD?
I used Nastran professionally. Contract required it.
For my own use, I got Mecway when it was at version 0.9 and dirt cheap. The upgrades are free for life and its now up to version 18. It's also not cheap. It does all the things I've ever needed... Static, non-linear, dynamic, modal, buckling, orthotropic, anisotropic, and several things I've never used. I've also done some very large models even non-linear. - https://forum.flitetest.com/index.php?threads/inqd-turbo-storch.71382/page-4#post-737352
I've not checked it out, but I thought Fusion 360 had a FEM module.
telnar1236
Elite member
Attempted to fly the MiG today and it didn't go well. Ran off the runway on my first takeoff attempt resulting in the gear breaking off, so I tried a hand launch with the results below. I think the big issue was the inefficient ducting killing a lot of the thrust.
Version 2 will be lighter with more efficient ducting.
Version 2 will be lighter with more efficient ducting.
telnar1236
Elite member
telnar1236
Elite member
Looking at my old design, the perfectionist in me kicked in and I ended up deciding to completely redesign everything. The biggest issues were weight and bad ducting. Losing weight mostly consisted on making various surfaces thinner, putting holes in them, or getting rid of them altogether, which while important, isn't that interesting visually. The duct redesign on the other hand, made use of CFD. My two goals were to minimize losses at speed and to substantially increase the effective inlet area at low speeds and high angles of attack, so the two big changes were to area rule the duct as much as possible around the "shock cone" (sudden changes in area result in large pressure losses) and to add a big NACA inlet on the bottom of the fuselage in place of the inefficient one that I put there previously. I also rerouted the duct to run along the bottom of the fuselage for much of its length to increase the space for the battery and receiver.
Inq
Elite member
With all the analytical on the front-end, do you have a plan to be able to quantify the results on the back-end? Seat of the pants is a little too qualitative... especially when the butt isn't in the airplane. As we all have experienced at one time or another putting a lot of thought/work/money/effort into something always makes us say that, "It is so much better." I'm not talking specifically about you here on this project. I see all the time... someone buys a new car (after replacing a two year old one of the same make) and its sooooo much better. It's human nature to justify large doses of thought/work/money/effort. Again... not you, in fact... I'm hoping you have more concrete plans to quantify the results because you went to this much effort on the front-end.
You've mentioned in previous posts about telemetry and even load-cells on motors. I've been thinking about pressure sensors for pitot tube and gathering data... mainly just for airspeed and altitude, but you might consider those for your projects as well... say, in the ducting area.
I really wonder if the Mathematical models you are using are really taking into account the hugely significant boundary layer based on scale. My first thought would be the layering of the 3D print would cause separation at the very beginning and you have total turbulent flow even within the shock-cone's region... giving up all hope of it being smooth somewhere inside. Although I know the CFD theory is sound, I'm dubious that modeling accuracy is not able to represent your actual airplane.
I have no formal training in CFD, but I do have quite a bit in the theory and practice using FEM. I can express it best by an example. Modeling a crack (or any high stress concentration region) in FEM can totally defeat a rookie engineer.
The top model would give a stress concentration of 2.0. The bottom would give significantly higher, but not anywhere near the theoretical value of infinity! And that is long before we'd get into non-linear plasticity effects and how they skew the results. Point being, from my position (the cheap seats) if you're not modeling the actual 3D printed layers in the model's geometry, I would suspect that you're not giving the CFD enough detail to do its job on your airplane model.
Have you done a detailed model of say... just the inlet and run the grid resolution up into the millions of elements modeling every 3D print layer and see if you keep or lose laminar flow? For all I know... it wouldn't surprise me if you get into some shark-skin phenomenon and actually have better flow than a mirror finished duct!
As we all have experienced at one time or another putting a lot of thought/work/money/effort into something always makes us say that, "It is so much better." I'm not talking specifically about you here on this project. I see all the time... someone buys a new car (after replacing a two year old one of the same make) and its sooooo much better. It's human nature to justify large doses of thought/work/money/effort. Again... not you, in fact... I'm hoping you have more concrete plans to quantify the results because you went to this much effort on the front-end. You've mentioned in previous posts about telemetry and even load-cells on motors. I've been thinking about pressure sensors for pitot tube and gathering data... mainly just for airspeed and altitude, but you might consider those for your projects as well... say, in the ducting area.
I really wonder if the Mathematical models you are using are really taking into account the hugely significant boundary layer based on scale. My first thought would be the layering of the 3D print would cause separation at the very beginning and you have total turbulent flow even within the shock-cone's region... giving up all hope of it being smooth somewhere inside. Although I know the CFD theory is sound, I'm dubious that modeling accuracy is not able to represent your actual airplane.
I have no formal training in CFD, but I do have quite a bit in the theory and practice using FEM. I can express it best by an example. Modeling a crack (or any high stress concentration region) in FEM can totally defeat a rookie engineer.
The top model would give a stress concentration of 2.0. The bottom would give significantly higher, but not anywhere near the theoretical value of infinity! And that is long before we'd get into non-linear plasticity effects and how they skew the results. Point being, from my position (the cheap seats
) if you're not modeling the actual 3D printed layers in the model's geometry, I would suspect that you're not giving the CFD enough detail to do its job on your airplane model.Have you done a detailed model of say... just the inlet and run the grid resolution up into the millions of elements modeling every 3D print layer and see if you keep or lose laminar flow? For all I know... it wouldn't surprise me if you get into some shark-skin phenomenon and actually have better flow than a mirror finished duct!
telnar1236
Elite member
First, right off the bat, you're completely right that this isn't really an adequate representation if I wanted to get great numbers. This model is not expected to be that accurate and is more for flow visualization. As a hobbyist, I don't have the resources to make it much better. You can actually see that the sealed nose cone has a predicted flow, probably due to a mesh issue, and I can't do anything about it since I'm right at the 200,000 node limit of Simflow. And the EDF unit isn't modeled at all. So while the model can help me see major issues, it isn't good enough for precise fine tuning and I haven't even tried to use it for that.With all the analytical on the front-end, do you have a plan to be able to quantify the results on the back-end? Seat of the pants is a little too qualitative... especially when the butt isn't in the airplane.
You've mentioned in previous posts about telemetry and even load-cells on motors. I've been thinking about pressure sensors for pitot tube and gathering data... mainly just for airspeed and altitude, but you might consider those for your projects as well... say, in the ducting area.
I really wonder if the Mathematical models you are using are really taking into account the hugely significant boundary layer based on scale. My first thought would be the layering of the 3D print would cause separation at the very beginning and you have total turbulent flow even within the shock-cone's region... giving up all hope of it being smooth somewhere inside. Although I know the CFD theory is sound, I'm dubious that modeling accuracy is not able to represent your actual airplane.
I have no formal training in CFD, but I do have quite a bit in the theory and practice using FEM. I can express it best by an example. Modeling a crack (or any high stress concentration region) in FEM can totally defeat a rookie engineer.
View attachment 234779
The top model would give a stress concentration of 2.0. The bottom would give significantly higher, but not anywhere near the theoretical value of infinity! And that is long before we'd get into non-linear plasticity effects and how they skew the results. Point being, from my position (the cheap seats
Have you done a detailed model of say... just the inlet and run the grid resolution up into the millions of elements modeling every 3D print layer and see if you keep or lose laminar flow? For all I know... it wouldn't surprise me if you get into some shark-skin phenomenon and actually have better flow than a mirror finished duct!
Quantification is a bit tough. The duct is too small to fit anything without substantially affecting flow characteristics. What I can do, however, is measure static thrust in and out of the duct and see how well it correlates with predicted losses. What I have been able to do (and what is a common technique with CFD) is validate the numbers for some measurable cases and then calibrate my technique to minimize error. That said, my error is still as high as about 20% for those test cases, just as a limitation of the node limit and software. I wish I could do a mesh sensitivity study, but with the node limit I have, it's hard to get it refined enough. I have done so with simpler models, and there is a difference, but it's small, and the simpler the geometry, the less significant it is. I've just lumped it into my 20% error and decided not to worry since there isn't much I can do about it.
As for the layer lines, those effects are actually pretty predictable and quantifiable. Aerodynamics tends to see much smaller effects from small surface features than fracture mechanics does, mostly because air doesn't really have an elastic region. It fills in the feature and then goes around it. Layer lines basically behave like surface roughness which is very well quantified and is typically just put as an input into the CFD software, affecting the shear and turbulence along the walls. I do have the CFD properly set up to deal with that, at least to the level that it is not going to be my primary source of error.
I do have plans in the works to build a sensor package for external aerodynamics that will carry probably 8 pressure transducers, as well as an accelerometer for all 3 axes and a GPS, but that is still a ways out and would only be for much bigger planes than this (I want to make something that is aerodynamically unstable in all axes fly, but that is a very long term project).
The short version is that the CFD is something I do for fun and isn't meant to be all that accurate. It lets me visualize the flow far better than I could without it and well enough to say that the new duct will be much better than the old, by some unknown amount.
Inq
Elite member
I've just barely started down this route. By chance... do you have any suggestions on extremely small / light weight tubing. Silicone type fish tubing is too heavy. The lightest (off-the-shelf) I've looked at is the fridge ice-maker tubing, or the small, black drip tubing. But I'd like to go smaller and lighter than those.
Cool! I look forward to that!I once had the opportunity to fly the AV-8B simulator. The guy said... now we'll simulate without the quad-redundancy computer control. I couldn't keep it level for even a half second. You have to hand it to those English pilots that flew the original Harrier every day without computer controls!
They must have trouble walking, is all I can say. I understand fully! I've come close to breaking out the FEM, but I always come down... I don't have any decent data. Strange how Dollar Tree doesn't do a full A-basis characterization of their foamboard.
The most I've done so far is simple Mechanics of Material type stuff on wood/fiberglass/carbon wing spars.
telnar1236
Elite member
I've only just started looking into it myself. This pack including a transducer, pitot tube, and tubing looks promising (Amazon.com: QWinOut PX4 Differential Airspeed Pitot Tube + Pitot Tube Airspeedometer Airspeed Sensor for Pixhawk PX4 Flight Controller : Toys & Games ) but it's too expensive if I'm getting 8, so I'm still looking myself.
Inq
Elite member
Have/can you do microcontroller stuff... Arduino, soldering and such? If so and you're not in a hurry (at least March, I'm doing the FTFC23 thing), the cost can certainly be improved. I've worked with these on things like weather stations and altimeters.
- The microprocessor can easily be programmed to process the pressure sensors (and any others you might want) store the data and/or send it to a base station (say a phone, tablet or laptop). This specific module includes WiFi and has pretty good range especially since the plane is above you in clear line of site. I have a cheap FlySky FS-i6X and apparently, this module can be linked to the receiver and send data back to the transmitters and display things like speed and altitude on the transmitter's LCD screen. These cost about $4, weighs 4 grams, and peaks at about 70 mA from your 5V BEC.
-
- The pressure sensor costs about $1.10, weighs about a 0.5 grams. You need one for the static pressure. And one for each dynamic pressures you want to monitor... say on a pitot tube, etc. The MPU can easily be programmed to take those readings and get the differentials and calculate say... speed and altitude in units of your choice.
-
