As an aerospace student, its really cool to see the "This is wrong, we know its wrong, but its close enough to work" go from slides in a lecture to an actual rocket test flight.
To be a bit more gentle: _"This is not fully accurate, we know it's an approximation, but it's close enough to work."_ (Some two+ decades ago I worked extensively with high-end finite element analysis (FEM, brother to CFD) software, so I'm well aware of "close enough")
you can only ever be So accurate. Material impurities, manufacturing defects, limits fo manufacturing tolerances, chaos theory, etc. all come into play in reality. Having a too perfect simulation is actually a bad thing, and costs additional time and money that is ultimately wasted.
@@Hydrazine1000 theres nothing more satisfying than striking the balance between assumptions and accuracy. I don't do engineering, but I am a geologist and it's very satisfying how we can make predictions about what we might find, and then 20 years later to find nearly exactly what we predicted!
TL;DR You've got it. Bridging the gap between paper and product is what engineering is. So dude when I was getting my BS in Aerospace, I "discovered" that our computer lab had Fluent CFD software, which at the time was a standalone tool not yet acquired by Ansys, so this was in 2006. I found my obsession. I made every individual, class, and team project somehow include CFD. I saw the shock collar of the Ares-1X live and spent the rest of my birthday simulating it to understand what I had just seen. My masters thesis was in CFD for computational kinematics of hydrocarbon combustion in MEMS devices. Our swiss roll micro-combustor fit inside of the D on a penny minted in Dallas. It was an angry red dot that burned JP-10 Decane fuel straight from a 1 gallon jug as RCS thrusters for station keeping on satellites. It later was made with thermo-electric metamaterials about the size of a AAA battery. It allowed service members to basically pour gasoline into their laptop on the battlefield to provide electrical power. I've even modeled my wife's prius to figure out what the optimal speed to set our cruise control on. CFD was my conduit from the back half of my JD Anderson's with all those cool pictures and concepts to reality. Fast forward 12 years since my masters and it has morphed into a data scientist and AI engineering role where I try to automate myself and 117,000 colleagues out of a job everyday. I still teach graduate CFD courses and experimental aerodynamics occasionally. I think back to how important what you just said was in my life even outside of my career. All engineering is "Good Enough Engineering" no matter how you slice it. Perfection is the antithesis of completion. It's on you to to find that connection between paper and planes. Don't trust your educators to do that for you. Many times it relies on having a start before your ready to avoid analysis paralysis mentality. But once you start once, the next time is easier and your evolution begins.
LOL, KSP is a game. Building a rocket that flies in KSP and assuming that is real rocket design is like mastering Guitar Hero and thinking your are the next Jimmy Page.
I don't make rockets. I make water go over ledge. I have hated every minute of CFD work I've done but it's hands down one of the most important tools in the modern world.
Just took a concrete maintenance and repair course where some of the damage examples were from cavitation and others were from scour of 12" cobbles/boulders permanently caught in an eddy current and grinding holes through 5' thick slabs. Crazy how much can go wrong if you don't do the CFD right.
Hi Joe: I’m a defense contractor. I understand ITAR (at least a decent part of it). The problem with ITAR is that it’s bureaucratic and obsolete before it starts. They just cant keep up with technology, and so they throw this huge wide net that includes everything but with such piss poor resolution that it both covers everything and nothing both at the same time! Case in point, you are tepid to discuss supersonic flow regimes. Guess what, the Iranians already have supersonic missiles! And so do the Chinese and so too do the Russians and so too do the Indians, and the Japanese and even the North Koreans! What the heck are we trying to protect! That cat is already out of the bag long time ago. But yeah, you will probably run into some ITAR resistance: DUMB! All this does is prevent bright new USA students from learning so that they might be the next generation of aeronautical engineers. Instead, we are dumbing ourselves down so the rest of the world will take over. I’m not saying to give out the latest stealth fighter coating recipe. There are things that should be protected, BECAUSE THEY CAN BE PROTECTED. But you can’t protect physics, and by and large rocketry is just physics. All one has to do is build a wind tunnel and one can find out everything that might be considered ITAR. worse case you put that together with photos of equipment from internet and Jane’s and you can usually easily figure out many/most things that might be considered ITAR. Why are we protecting those kinds of things? It’s a waste of time and effort on our part and just shoots us in our own foot. ITAR is run mostly by lawyers so that should tell you a lot. Enjoy your vid as usual!
Honestly, it sounds like they cast a wide net so they can investigate and catch the 'bad' people whilst allowing folks like Joe to do what they need to do (whilst probably keeping tabs in the background). If you think about it, its the only possible way for a body like ITAR to do anything at all seeing as, like you say, any specific policy on specific technology can't hope to keep up with advancement.
a state nation will always have the resources to replicate something developed 50-80-100 years ago. crazy people, terrorists, less so. it is a thin veil, but a thin cover has its uses...
What I like about these videos is they dont feel overproduced. They feel very down to earth and relatable. As a first year engineering student they're very helpful in communicating that things ARE hard, and that even people who have made a career in STEM dont always get it right or have the easiest time. Im looking forward to that Space Shot!
Something tells me when this man said "The world is your wind tunnel." he probably thought to himself, I paid for the speedometer on my car, I'm going to use the whole speedometer.
In my life, I've found the best place to go fast is a closed airport The fastest I've been pulled over for was 152 mph. It was at 3am with no other cars on the road. If you can't drive fast, why drive at all.
For my PhD i wrote a CFD for polymer liquids (like polyethylene) which are non Newtonian and remember when they've been compressed or stretched so you have to give your model the ability to remember where that bit of the liquid came from 😎. Back in 1996 this ran all weekend on a silicon graphics machine and often failed. Nice to revisit this stuff.
@@tommclean9208 yes, technically if you know the current state, you dont need memory. However its not possible to know the actual current state of complex rheological systems. For example, you have large molecules with functional groups that interact in different ways with each other and other things in the system. If you allow that system to sit, those molecules will form large structures with ofher particles and with themselves. if you apply shear to that system, those structures break down at some rate, woth different structures breaking down under different shear conditions. So of you have a memory of prior conditions, and good experimental data, you can predict how those structures have broken down and apply that to your similation. Without a memory, you cant do this. These are thixotropic fluids.
Instead of filling out the entire test matrix, you can use something called modern design of experiments. You really don't want 10x6 points, you want a formula (response surface). Knowing the general shape of the formula, you can use software to calculate which points of the matrix give you the most leverage and only measure those. In a single-shot windtunnel that can save you lots of time.
I think the best use of DOE is to maximize the data you learn from all of the points you can afford to run, not to see it as primarily a way to reduce the number of points you need to run
Published books are actually an end run around ITAR and related laws. They're covered under the first amendment. Early encryption pioneers used them to get around export controls for stuff like RSA and other important algorithms that form the bedrock of the secure Internet today.
I imagine it's also more that "if it's in a published book it's probably ok to discuss" and less "if I publish this then it's exempt" There are plenty of things that if you tried to publish them, they'd either never get published, or you'd get in big trouble for it.
Yeah remember when Defense Distributed tried that? They spent almost a decade in court and the case eventually got dropped because the state dept dropped jurisdiction to another agency.
I know I just made 2 comments, but I just wanna say that I graduated middle school today, and a new video to watch really made my day after a long day of standing and thanking people, so thank you Joey for doing what we all wanna do but don't have the time or experience and thank you for filming this journey in such hig quality cinematic production.
Same! I took classes in college as electives. I had no reason to take or basis of intelligence for understanding the professors. I just like to listen to smart people talk about things. Haha 😅I took constitutional law classes as a computer scientist.
A few years ago I was involved with a project that used explosives to cut heavy pipe in oil wells several km underground. The CFD modelling of various scenarios to predict whether the pipe would be successfully cut took a month of compute time on a specialised compute cluster. All to model something that happened in the blink of an eye.
I like how your early attempts mirrored that of early aircraft pioneers. "Eh it looks like it'll work so it must" "It just killed the test pilot. Twice." "Ah I see the problem. Must've been this random piece that somehow helps."
Dude!! I am 45 years old. Have a Third grade math education. Have no clue what all the math was. But wow that was a fun video. I just love seeing Someone so passionate. Thank you for letting us be a small part of your space shot. God bless brother.
Joe, I'm a engineer.. My main youtube is populated by Engineers/Makers and Music. You and Tom Stanton are among my favorites because you both do real science and engineering. I love watching you experiment and learn. I don't need the dirty details. I enjoy the process of development, the successes and failures. Thanks for doing this. I've learned a ton that applies to my design work. None of which involves things exploding or burning rapidly.
Just completed my Masters in AE and am currently working on CFD for the second straight summer at AFRL. Loved the bits on CFD, and got quite a few chuckles out of the "going into the weeds" bits after my in-depth CFD course this past spring. Also, after researching for a few years, ITAR is everywhere. If you want to do almost anything in AE that, say, requires a US citizenship, it's likely going to be ITAR. I respect the knowledge of both the physics and the regulations so you can keep yourself safe! Keep up the good work and I'm excited for the space shot!
The ITAR question has been on my mind with respect to your videos and I think you explained it exceptionally well. As an engineer with a family I can say you are taking the right approach to sharing and keeping info in confidence. If I had to choose between my engineering career and my family I would choose my family without a second thought. This video was awesome with the CFD and the normal rocketry updates, but it reminded me about being the best person I could be. For me, that is a family man. Great approach, great content, amazing insights, keep it all coming!
The secret about high speed wind tunnel testing is called Reynolds’s number similarity. You can test a subscale version of your fin and then non-dimensionalize the coefficients to then apply them to a larger fin. CFD does solve the NS equations. It solves them numerically so instead of a derivative it would use an actual number. The amount of non linearity you are capable of capturing is dependent on the turbulence model you are solving. For instance the SA, two equation model is a linearities model. However, model such as LES capture and predict much more of the non linearity of three system. Then there are models such as the RANS models which is appropriate for steady state solutions only, where LES is applicable even to time accurate simulations. The crux of the problem is that all turbulence modeling is numerically implicit. The reason it needs to converge is because the equations are implicitly defined. You have the same number of unknowns as you do equations meaning that it can’t be directly solved. So what you have to do is guess a number, calculate the solution, use that solution to calculate an updated solution, and then compare the results. The difference in these is the residual. When the differences between the current calculation and the previous is low enough, you make the assumption that the solution is converged as the solution is not changing by updating the calculated solution. There is more to it than it that is a simplified way to think about the solver. One of the elephants in the room of Cartesian solvers is the viscous shear gradient. Because the grid cannot correctly capture the geometry, it can’t calculate viscous drag. The reason your Cartesian solver overpredicted the values is because Cartesian solvers cannot correctly estimate viscous drag as they do not capture the surface as you were saying. So what they do is use wall functions ,or extrapolated friction equations to estimate the viscous drag. They generally are conservative in this method. Even the fancy solvers that use polyhedra and polyhexcore or even adaptive refinements use empirical friction models the calculate viscous drag. The are of friction modeling boils down to how the turbulent kinetic energy production rates are calculated. These are usually based on testing that has been done and is nondimensionalized and scaled to your grid. On the note of ITAR, just because it is already online does not mean that it’s ok to share. I’m sure you know that though and thanks for your awareness in this area.
Ideally you match Reynolds number AND Mach number. I'm pretty sure I'm allowed to say that, because www.grc.nasa.gov/www/k-12/airplane/airsim.html says it too How to do that in practice, I'm guessing is ITAR, unless there's old NACA (proto-NASA) papers about it
You don't need a turbulence model for CFD and it's got nothing to do with the particular discretization of NS. With DNS/ILES you solve the discretized NS eqns numerically on a fine (close to kolmogorov scale) grid. With a turbulence model you can use a coarser grid and sweep some of that under the rug using experimental correlations to relevant problems. When you have the same number of eqns as unknowns, this means you CAN solve the system (think back to basic algebra). The difficult part is your system of equations in matrix form is way too big to invert directly so you do have to use funny techniques to progressively guess solutions. I took CFD and turbulent flow in grad school.
I really don't mean to be nasty - CFD is really interesting and amazing and it's always great to have more people interested in it - but this is wrong on a lot of levels, the person above has correctly pointed out a few things. URANS = unsteady-RANS models exist and are common. All turbulence models are not accurate in time, except perhaps DNS, but many are accurate in a statistical sense, i.e. they capture the correct average velocity. Besides, turbulence is chaotic so tiny errors will lead to large scale deviations down the road (butterfly effect), so there is not really such a thing as "time accuracy". The residual is not the change between consecutive solutions, it is the amount by which the equations are not satisfied, i.e. if you're trying to find a vector x such that Ax = b (where A is a known matrix and b is a known vector), the residual vector is r = b - Ax. If you have the correct x, then it satisfies the desired equation and r=0. CFD packages spit out a single residual number instead of the full residual vector e.g. by summing the squares of each component then square-rooting (L2 norm), or by finding the maximum value (Linf norm), or by summing the absolute values (L1 norm). In fact, a lot of times the convergence stalls and the residual stays high even though the solution does not change between iterations. Just because your solver is unable to find a better solution, it doesn't mean it's solved the equations. There is so much more that could be said - I really recommend reading a solid CFD textbook like Versteeg to understand how CFD really works, it was very helpful for me when I was starting out. Again, I hope not to be mean here, I just want to avoid confusion! I have written my own direct numerical simulation (DNS) code that is being used for research in academia. Happy to point anyone towards helpful resources on CFD.
@@GOAFPilotChannel you are right, you don’t need a turbulence model for CFD. I agree with that. Not sure what point you are making though. Can you clarify? Hah you’re totally right about the number of equations too. My dislexia got me in that one and should have said If you have more unknowns than equations you can’t solve it. Hah dang.
@@pupudski8066thanks for the reply. I agree no turbulence models are strictly accurate in time, but LES has more applicability in a simulation where time accuracy is desired. Time accuracy is not typically a thing, but I do it. When using linear turbulence models such as SA, the transient solution drifts further from reality than LES. I use it to calculate dynamic aerodynamic coefficients that are compared to test data, and have done the sensitivity study comparing different turbulence models. LES is more applicable to time accurate solutions than RANS even every time I have done this for many flow regimes. I agree, that’s not exactly how residuals are calculated and even stated so. Not sure why you’re calling me out here. It was an attempt to convey the complexity of the reality with something that someone with less knowledge might understand.
CFD takes a lot of computing power; a previous employer built a CFD facility a few years ago, the computer system occupied a complete 20' ISO container and they had to have a new substation built to power it. The cooling system was used to supplement the heating of some of the buildings elsewhere on the site.
@22:30 best way to explain to us amateurs....."if i can figure it out, so can you....jus keep trying!" I absolutely LOVE your videos. And so does my daughter.... you keep us inspired to keep going even wen our models fail to take off... with continued testing....we hope to get our Starship model to lift off soon... We love you Joe !!!❤🤟
As someone who deals with information that may or may not be ITAR or SBU (Sensitive But Unclassified) data on a regular basis, I feel your pain. I build and maintain archival systems for folks who have such data. The regulations are deliberately vague and what was ITAR information 15 years ago might not be today. On the other hand, things that are 60 years old and you’d assume have been long superseded by newer information or methods, are still considered sensitive. Worse, an image that a normal human wouldn’t pay any attention to might a eureka moment to some bad guy. So you end up fuzzy up images so that certain areas just aren’t really discernible. On the other hand, unlike you, I have an approval for release process that covers my ass. Before we release something for general public, several experts have signed off on it. I’d suggest that you do something similar. Not only would it help cover your butt, but they might spot something that you think is obviously not a big deal but might offer someone an insight. The trick with imagery is to not only examine the object or objects, but to diligently look at the background. More than one person has been bit by that. But all in all, good on you for deliberately being cautious beyond the scope of the laws. Not only to keep stuff out of the hands of bad guys, but out of the hands of the eager but not quite so sophisticated beginner. This stuff is dangerous. A lot of folks don’t get how dangerous. So we have to help them out. Keep up the good work.
@@starship3812 They already have thousands of rockets, some are probably self built. Like Joe said, if you do your research you can find all of this information in publicly available books and papers.
publicly available info cannot be subject to ITAR. it's already out there. especially if the book is sold internationally. reading books to people is not a criminal act.
@@VulpeculaJoy i mean they can improve it further. Or they can develop other advance aerospace weapons not only rockets. These are some low profile information but also needs to be regulated cuz these are enough for making enough damage. Otherwise starship level technology must be regulated carefully otherwise chinese can steal it.
Up to this point I knew about the grey area, but never realised how much it applies to your channel. Let's hope you never break the speed limit in that regard
From a mech engineer who basically just does FEA all day (not CFD yet). Agencies who run HUGE models have actual servers that run their sim’s at a reasonably quick speed. And those will only give you an extra couple of percent ‘accuracy’. A seasoned design engineer will know if the design is worth Pershing from the ‘simple’ results you would get from a 1 day simulation which you could use during preliminary phases and then run a 3-5 day simulation while you do other things to just fine tune your results for software inputs.
That flight video was beautiful. Can't wrap my head around the amount of authority that small movement has on the rotation, seeing the set point and actual angle so closely tracking!!
I did a similar project in uni. Not nearly as high speed though. I derived a semi empirical formula for deflecting fins and did some corrections for low speed flight from a wind tunnel. The fin results were not that good but the air brakes ( yes I had those too) numbers were pretty good. I ended up doing some in flight testing with the air brakes and it gave me good results on the accel plots.. I have this data published on a report if you are interested
I just want to say a genuine thank you. Prior to you addressing the risks associated with running a RUclips channel dedicated to things like steerable rockets I had put little to no thought into the significance of remaining in that grey area. If you ever feel like you need to forego the inclusion of some piece of information for your own safety, don’t hesitate. Don’t get me wrong, I’d love to learn it… that being said I would much rather know this is something you enjoy and is not at all a detriment to your comfort. Keep up the good work! Thanks again!
When I was in highschool (1970's we built a wind tunnel using bits and pieces from junk yards. The fan blades were from car wrecks, the fans used for radiator cooling. Two of them and a strong electric motor. (I don't know where from, but something from the metal forming lab. A lathe or something. ) They also played a big part in making the chamber.
I'm impressed that a RUclipsr is consciously thinking of improper, illegal, or terrorist results of his channel video's. I applaud your determination to learn about rocketry, that so many have over the last thousand years. We would never have made it to the moon without folks like yourself persevering to achieve a goal.
At 33years old, I can’t divide fractions, hell I can’t even read out loud that well. But your videos are entirely understandable and immensely entertaining. Also props on the quality and editing. I’ve been here since day one and it’s wild how much I look forward to new content.
I love to think about where the next generation will go if we can make it not suck so hard for them. You are such an inspiring and effective science communicator and creator.
Now we're talkin 10:30 - Joe you might want to double check that. NS is *compressible* viscous unsteady flow. Unless you enforce incompressibility of course. 16:31 - Did I hear robust control? Have we finally moved on to time-domain modern control? Woohoo!!! 24:06 - Y+? You've been doing too much CFD... Have you thought about solving the Euler equations instead of the NS equations? It may reduce the computational cost for your run matrix, especially in certain regions of your run matrix. (If you mix RANS and Euler in parts of your run matrix then does that count as multi-fidelity data fusion...)
Great to see rocketry explained in a relatable and casual fashion, feels like going to engineering school again 😅 Hoping to see more about pitch and yaw control.
12:06 - All of AI is just linear algebra too basically chat GPT is just a 400 million element matrix. (oversimplification, but, still, it's all just linear equations)
The activation functions (at the neuron output) are deliberately non-linear, otherwise the whole thing would reduce to a single (huge) matrix multiplication and we want to approximate functions that can’t be approximated using that.
My (oversimplified) assumption that the line you’re not crossing here is using all of those different speed/angle/pressure simulations to create an interpolatable map of forces that in effect replace the trig function you start out with for the gimbal case. Instead of one variable [f(theta)] it’s got all of those inputs [f(theta, velocity, pressure)]. Not to different from an open loop fuel injection map. If I’m right, blink twice in your next video on trans-sonic domain. PS: I’m a “US person” so it’s cool…
PS, I'm not a "US person" so if that was it you are so royally screwed! Which is why you can't just post the actual solutions online. Oh and the US basically forced an even worse policy into nearly every other country on Earth: We aren't even legally allowed to stumble into those answers on our own, let alone talk about them with *anyone*. No one would ever dare to fly (or even simulate) a model rocket with *any* sort of active trajectory control over here.
I’m a (retired) s/w kinda guy and I know nothing about building rockets and controlling them. BUT, I learn something cool from each and every video you produce. I also laugh occasionally (well done). Keep up the great work! Thank you! PS: I once did a tour inside a very large wind tunnel. No way it was a $200/hr. 😮
Transonic flight video? Would love to see it! Just a quick side note, your videos are absolutely incredible. From the engineering to the humour.... It's just the best thing known to man. It's amazing to see the passion you've got for rocketry and aerospace engineering. I know you have heard all that before but it's got to be said again. Keep it up and stay safe (from ITAR...)!!!! 😂
Gotta love ITAR. AKA the biggest thing stopping me from putting time into my rooftop phased-array radar dream project. (In my case the challenge isn't "what can I talk about" it's what can I buy COTS or otherwise not "specially designed" so I can manufacture using cheap overseas factories vs $$$$ domestic ITAR-registered vendors or developing the capabilities in house). ITAR controlled software and gateware? Just don't put it on github, problem (mostly) solved. ITAR controlled PCB designs so I can't make them in China? 10x project cost multiplier.
re: ITAR I 100% get (and agree) with the better safe than sorry method. I work for a company that deals in ITAR regulated items (And not even "Maybe regulated" but explicitly regulated) commercially, and to avoid any kind of issues we don't allow anyone outside the US to order from our website at all, not even clearly un-regulated items like shirts, drinking glasses, etc, simply so that if anyone ever comes knocking about it we're more covered than we need to be.
You could cut the costs from a bazillion to 1/5 a bazillion by setting up 5 models at once side by side each set at one of the angle deflections. It is still 1/5 of a bazillion though. Maybe NASA would let you use one of theirs for a RUclips presentation. They have done it before. Still, CFD is possible without a supercomputer these days (I think). Navier Stokes is a bitch though. I used to work for a company that built gas turbines. They spent close to a million dollars to get time on a super computer. They wanted a high accuracy model of airflow across a new blade design. They got a deal because they only used a few cycles at a time. After six weeks, they finally completed the run. The resulting animation was .75 seconds long. The engineers were ecstatic. If you can get better than a wild guess, you are the man.
24:48 OMG it follows the Roll Setpoint so well. I had a course on PID control two semesters ago. This may look simple, but it is not. Building and tuning a PID Controller so well is amazing. Amazing Work Joe, keep it up. Love your Videos man. This is quality contend.
Regarding the legal issues - I was in a rocketry club where we built fairly high performing solid and hybrid rockets. After talking to some people we came to the conclusion that steering would be a red line even though we managed to get everything for fuel and motor development done above board. But Germany is also a bit less friendly towards rocketry... Stay safe and I wish you the best of luck! Btw the footage at the end was great! Especially the overlay :)
Even as a professional team in a semi-related field. We too often go too far with too few simulations, the cost in hindsight is always large. However, choosing the wrong simulation to do, or modeling things you don't need too accurately is also doom.
Although the information you tell is far above my understanding most of the time, but you try to make it as easy to understand for everyone including me. Telling just enough for people to get directions to do the same, but not to much that other people get overwhelmed. I love the way you explain how you tackle the issues you find, on the wing, trying to get to your ultimate goal. Seeing the progress (how small as it could be) incl. thinks you learned so far is worth watching you. Since i started watching your channel i have always found it interesting. Just show what you like to show.
one of the most important topic aboyut CFD that you forget to mention directly but still demonstrated is validating your CFD set up. Only thur validation can you make sure that your CFD setup is not total garbage.
Can I donate pc hardware to the cause? Knowing you are limited by computation speed during CFD hurts. Either shipping you parts or discussing running things remotely for you. Either way I'd love to help.
Bravo to "I'm unwilling to accept a 5% chance". No one could reasonably fault you and it gets the message across really well (like "there's no point in arguing about the regulations in the comments plzkthx")
Thanks! It's from Target - 23 bucks. I tossed a link in the video description :) www.target.com/p/men-39-s-floral-print-button-down-shirt-goodfellow-38-co-8482-white-xxl/-/A-88248707?preselect=88248705
Here is how to build the control algorithm: 1st model the forces by hand (depending on velocity, tilt angle, weight, etc.) 2nd assume error bounds +/- 0-50% (depending on what quantities, e.g. tilt force, drag or weight etc.) 3rd construct your state space model 4th programm the controller (u can try a PID if u want or look for fancier stuff) 5th make a lookup table for the control parameters with the varied model parameters 6th u monitor all past states of the flight 7th while in flight calibrate the controllers parameters to the monitored states (that means the parameters will vary while in flight!!!) 8th assume valid initial condition for the initial control parameters Now u are ready to start your rocket straight. PS: Brute force will only work with enough compute power.
Cool video! As someone with a PhD in CFD, I must reemphasize (you did a good job pointing this out) how extraordinarily challenging CFD is. While many modern tools make it look easy, the topic is so incredibly vast that it is my opinion you cannot know what you don't know without months or years of education. Even professional engineers often do not know fundamentals required to produce correct results. For example, it is necessary to use a "compressible" solver for high speed flows (roughly over Mach 0.3). Without this modification, results will likely be off by orders of magnitude by the time you get even to Mach 1. However, simulating compressible flow is a notoriously challenging exercise in CFD. It is often difficult just to achieve convergence, and much more so to get remotely correct results. There is a laundry list of crucial comparisons and checks that must be done, but notable is a mesh independence study, where you increase your cell count until metrics of interest stop changing. I am sure this video only represents the tip of the iceberg, and much study, care, and consideration went into this, but readers should be aware of how incomprehensibly deep the field of CFD is.
Yo, Joe, to make these sims faster you should use a program that I don´t remember the name of but it was used on a SalC1 vido of trying to find the pack.png picture of minecraft. What the program does is allow other people to share their computing power to run sims faster, and I think that with your comunity a lot of people would colaborate this way. (I would be the first to participate) PD: Sorry if my english is bad, its not my native language. Thanks for reading.
There is no need to share the files, it works as if it were a super computer and Joe is the only one who can manage and view the files, the other people just process it without seeing anything or having access to anything.
Looking around at all the people who've lost their minds, it's super nice to find a group of people that haven't!! At least only in a good way! Always been an exceptional channel. Thanks BPS!! ♥🚀🚀
I wrote a 3D Navier-Stokes "windmapping" system for the last game engine I wrote that included LOD based on proximity to the camera - because it was only supposed to be an inconsequential visual effect to make particles flying around be affected by things like explosions and objects "moving through the air". It seems like whatever CFD system OP is using here should employ a similar LOD scheme. I went ahead and did something like an octree subdivision that was based on the camera's position, orientation, and view frustum, where higher resolution cells next to a lower resolution cell in the octree would sample from it and interpolate what a simulated "virtual" neighbor cell's velocity/pressure would be. I would never simulate a whole 3D volume, even with the axes squished in certain ranges (which is better than nothing) and go for something that is only high resolution around the mesh, while everything else is low resolution. It also helps to simulate across available CPU cores on a CPU that has a lot of cores - or use a compute shader on a GPU! I had my "windmapping" running realtime, along with my game engine, on an 8-core CPU. I believe there's at least one video of it in action on my channel if anyone is curious to see what that looked like. There were vector lines drawn for the cells and their length reflected the velocity while the color reflected the pressure (green was low pressure, red was high pressure). Anyway, that's my two cents :]
You’ve been doing perfectly Re: ITAR so far. Don’t change anything, leave out what you need to leave out. We understand. :) Re: What to see next, I’d love to see more camera views and flight test data! Keep on with creative angles and sharing build process vids 🙏🏼 and keep up the great work!
Your RUclips videos are the reader's digest version of building a rocket!!! Just enough info to spark a interest but not enough to build a complete rocket in our bedrooms
I've just subbed after months of watching this amazing journey because of that end talk Joe, share what You feel is safe to. The struggle, joys of steps finally achieved on difficult field, that's what this channel is most valuable for me. Wish You many good flights. This whole project just looks promising, You are down on the ground whilst up in space, have a good one and keep them coming.
22:43 I have had the thought ana occasionally wondered aloud what bits are left out . I appreciate the concerns and like the way Joe seems to be approaching the matter . Pure scholarship and learning is fun, bit mindfulness of unintended consequences to a reasonable degree is wise. Also, some elements are better for the learner to gain on their own experience, anyway. I am quite impressed with the resulting roll control footage. Excellent video all around.
I did my PhD in numerical (computational) methods (and work in aerospace) and was expecting another "RUclips aerodynamics special" ... but actually I was pretty surprised with this video. Good job.
I've had a whole semester of CFD studies in university. I don't remember much, but one thing that struck me was how small your domain is ! My teachers always said your output surface (end of the domain behind your object) should be 20-30 chord lengths away ! Input surface may remain 5-10 chord lengths ahead. Top and bottom surfaces should also be somewhat 10-20 chords away to avoid that issue you had with section decrease. So basically your domain should be 30-40 chords long, and 20-25 chords high. This ensures your boundary conditions are correct and do not influence whatever happens to your fin. Good luck !
Obviously this increases your number of cells, but then you may tune the mesh better to get large cells around the boundaries, and very small near the object. Also consider inflation layers to capture the boundary layer, and so on... Have fun 🙂
Armchain engineer here, but I feel like the way you go from test data (CFD, wind tunnel, or anything else) to a control system would be to create a sort of look up table. Essentially, your test data lets you build a big table with (speed, pressure, fin angle) as indices, and deflection forces in each cell. In preperation for flight you would use this data to create a look up table with (speed, pressure, desired deflection force) as input and fin angle as output. You would probably want to interpolate between the values to get a smoother output. I did something a little like this for sailing optimisation, using test data for sails, to pick the correct sail angle for a desired speed and heading.
As far as I know, the design for the WW2 Spitfire is still classified. (Almost 90 years later.) I wanted study the cooling system of the P51 Mustang for the use in water cooled experimental aircraft, since it supposedly it had a net negative drag. (Thrust.) I even looked at some freely available CFD programs, but got lost in the weeds of grid design.
One of those projects I keep meaning to try to build is a foam wind tunnel. The speed of sound is the speed where energy can be freely exchanged between kinetic energy and compression. Kinetic energy per unit volume is dependent on density, so speed of sound is dependent on density and compressibility. SoS in air is high because it is very light. SoS in water is very high because it is very incompressible. Mix the two and you get something intermediate density that is far more compressible than water. If you can produce a 50/50 foam the speed of sound is ~20m/s.
For CFD or physical testing, you can use a design of experiments method to reduce the number of runs. You are doing one factor at a time which gives you the best data at the cost of increased testing time.
This is so impressive and fantastic to follow along with, I love the magic 8-ball site... when it doubt play it very very safe. Stoked for more roll control!
The availability of sensitive engineering information is actually a point I never even thought about thinking about. I am from Germany and I consider basically everything accessible that is not intentionally kept a secret by some company/ government. That citizens of countries are legally responsible for what they share with me / an audience of known or unknown origin is crazy to me but in a weird way makes sense against. I guess some countries make it hard for themselves by restricting access to the Internet, but using state boundaries and citizenships as a deciding factor is weird. I am certain this doesn't actually work (to a notable extent) but the existence of such rules is new to me. I might look up if there's an equivalent in the EU so I can be like " sorry mate, can't tell you that" to my colleagues (im studying in Switzerland)
aerospace student currently working on my masters degree specializing in "autonomeous systems" here, the control loop part was really the thing i was waiting for the whole time. But i never thought about ITAR, and its a very fair point to bring up. So your decision is perfectly understandeable, i guess i will have to figure that stuff out on my own :D Great Video as always
"After every simulation, there is the experiment" - this was told me too often from more experienced rocketry guys around me and it's true as you also mentioned in your video. I would love to get a deeper look into CFD - i was always on the construction side of developing and "only" did some FEMs but never get into CFD. I tried once to simulate stability for a rocket of a friend of me but im sure, as you said, garbage in - garbage out. So plz show more CFD stuff.. I guess i was on the same way as you but i dont have more time to explore the rabbit hole.
As an aerospace student, its really cool to see the "This is wrong, we know its wrong, but its close enough to work" go from slides in a lecture to an actual rocket test flight.
To be a bit more gentle: _"This is not fully accurate, we know it's an approximation, but it's close enough to work."_
(Some two+ decades ago I worked extensively with high-end finite element analysis (FEM, brother to CFD) software, so I'm well aware of "close enough")
you can only ever be So accurate. Material impurities, manufacturing defects, limits fo manufacturing tolerances, chaos theory, etc. all come into play in reality. Having a too perfect simulation is actually a bad thing, and costs additional time and money that is ultimately wasted.
@@Hydrazine1000 theres nothing more satisfying than striking the balance between assumptions and accuracy. I don't do engineering, but I am a geologist and it's very satisfying how we can make predictions about what we might find, and then 20 years later to find nearly exactly what we predicted!
@@biscuit715 As someone said, " Models are just imperfect representations of reality."
TL;DR You've got it. Bridging the gap between paper and product is what engineering is.
So dude when I was getting my BS in Aerospace, I "discovered" that our computer lab had Fluent CFD software, which at the time was a standalone tool not yet acquired by Ansys, so this was in 2006. I found my obsession. I made every individual, class, and team project somehow include CFD. I saw the shock collar of the Ares-1X live and spent the rest of my birthday simulating it to understand what I had just seen. My masters thesis was in CFD for computational kinematics of hydrocarbon combustion in MEMS devices. Our swiss roll micro-combustor fit inside of the D on a penny minted in Dallas. It was an angry red dot that burned JP-10 Decane fuel straight from a 1 gallon jug as RCS thrusters for station keeping on satellites. It later was made with thermo-electric metamaterials about the size of a AAA battery. It allowed service members to basically pour gasoline into their laptop on the battlefield to provide electrical power. I've even modeled my wife's prius to figure out what the optimal speed to set our cruise control on.
CFD was my conduit from the back half of my JD Anderson's with all those cool pictures and concepts to reality. Fast forward 12 years since my masters and it has morphed into a data scientist and AI engineering role where I try to automate myself and 117,000 colleagues out of a job everyday. I still teach graduate CFD courses and experimental aerodynamics occasionally. I think back to how important what you just said was in my life even outside of my career. All engineering is "Good Enough Engineering" no matter how you slice it. Perfection is the antithesis of completion. It's on you to to find that connection between paper and planes. Don't trust your educators to do that for you. Many times it relies on having a start before your ready to avoid analysis paralysis mentality. But once you start once, the next time is easier and your evolution begins.
"If only there were a computer program that could give you perfect data… at the cost of a slightly higher electricity bill" -- KSP1 Menu Theme begins.
So you see, that's where the trouble began. That music. That damned music.
Do do do do...... Do.
(Sneaky adventure begins playing)
This totally should have happened. The KSP soundtrack is all royalty free.
LOL, KSP is a game. Building a rocket that flies in KSP and assuming that is real rocket design is like mastering Guitar Hero and thinking your are the next Jimmy Page.
@@connecticutaggie I hate to break it to you, but you really missed the sarcasm memo.
All models are wrong but some are useful!
Unbelievably true
The world is just a model. It's been useful at times.
that's why they're models. They are not designed to represent reality.
if u simulate the wind tunnel with a molecular simulator it can be a good approximation of the reality.. but gonna take forever to compute XD
@@sunoncream1118 it can be done in real time. See my previous comment.
I don't make rockets. I make water go over ledge. I have hated every minute of CFD work I've done but it's hands down one of the most important tools in the modern world.
Had to use CFD to design an tiny antenna for my senior design project, and it was the most excruciating thing I've ever had to do on a computer.
"I make water go over ledge" I love this pitch
I spent my entire career making water go downhill. I understand !
Just took a concrete maintenance and repair course where some of the damage examples were from cavitation and others were from scour of 12" cobbles/boulders permanently caught in an eddy current and grinding holes through 5' thick slabs. Crazy how much can go wrong if you don't do the CFD right.
This almost makes you sound like a dam engineer :)
Pro tip: enrolling at an educational institution to gain access to the really expensive software is cheaper than buying the software alone!
Or just go to the Russian forums like the rest of us...
especially outside the US
@@malloott could you suggest some forum/links pls?
@@towfiqueulhaider2000check ru tracker
@@towfiqueulhaider2000 a friend tells me that Bit Torrent gives you open access to Ansys, Abaqus, NASTRAN and FLUENT.
Hi Joe: I’m a defense contractor. I understand ITAR (at least a decent part of it). The problem with ITAR is that it’s bureaucratic and obsolete before it starts. They just cant keep up with technology, and so they throw this huge wide net that includes everything but with such piss poor resolution that it both covers everything and nothing both at the same time! Case in point, you are tepid to discuss supersonic flow regimes. Guess what, the Iranians already have supersonic missiles! And so do the Chinese and so too do the Russians and so too do the Indians, and the Japanese and even the North Koreans! What the heck are we trying to protect! That cat is already out of the bag long time ago. But yeah, you will probably run into some ITAR resistance: DUMB! All this does is prevent bright new USA students from learning so that they might be the next generation of aeronautical engineers. Instead, we are dumbing ourselves down so the rest of the world will take over. I’m not saying to give out the latest stealth fighter coating recipe. There are things that should be protected, BECAUSE THEY CAN BE PROTECTED. But you can’t protect physics, and by and large rocketry is just physics. All one has to do is build a wind tunnel and one can find out everything that might be considered ITAR. worse case you put that together with photos of equipment from internet and Jane’s and you can usually easily figure out many/most things that might be considered ITAR. Why are we protecting those kinds of things? It’s a waste of time and effort on our part and just shoots us in our own foot. ITAR is run mostly by lawyers so that should tell you a lot. Enjoy your vid as usual!
Honestly, it sounds like they cast a wide net so they can investigate and catch the 'bad' people whilst allowing folks like Joe to do what they need to do (whilst probably keeping tabs in the background). If you think about it, its the only possible way for a body like ITAR to do anything at all seeing as, like you say, any specific policy on specific technology can't hope to keep up with advancement.
a state nation will always have the resources to replicate something developed 50-80-100 years ago.
crazy people, terrorists, less so.
it is a thin veil, but a thin cover has its uses...
Ah yes, bureaucrats, the worst thing an engineer can face, followed by bean counter executives
@yt45204Would certainly make the world more exciting.
@yt45204 yeah this guy has missed the point entirely lol
What I like about these videos is they dont feel overproduced. They feel very down to earth and relatable. As a first year engineering student they're very helpful in communicating that things ARE hard, and that even people who have made a career in STEM dont always get it right or have the easiest time.
Im looking forward to that Space Shot!
Joe after spending 2 hours mixing the snare for the background beat: WHAT DO YOU MEAN NOT OVERPRODUCED??
@@mju135 I mean, fair enough.
It's pretty groovy.
Something tells me when this man said "The world is your wind tunnel." he probably thought to himself, I paid for the speedometer on my car, I'm going to use the whole speedometer.
Average German thought:
@@Wurtoz9643 gotta use the entire autobahn
@@sebdapleb1523 Fahrn Farhrn Fahrn auf der Autobahn.
In my life, I've found the best place to go fast is a closed airport
The fastest I've been pulled over for was 152 mph. It was at 3am with no other cars on the road.
If you can't drive fast, why drive at all.
@@BastiVC I came here to quote Kraftwerk, too. But, you beat me to it. ;)
ITS YA BOI TONY PEPPERONI WITH ANOTHER BEPIS SPACE ROCKET REVIEW
EEEYYYYY IM TONY PEPPERONI
EYEYYYYYYYYYYYYYY
@@BPSspace The "Tony Cesaroni" one had me rolling
Quiet on set.. QUIET ON SET!!!
@@BPSspace GIVE US TONY PEPPERONI
For my PhD i wrote a CFD for polymer liquids (like polyethylene) which are non Newtonian and remember when they've been compressed or stretched so you have to give your model the ability to remember where that bit of the liquid came from 😎. Back in 1996 this ran all weekend on a silicon graphics machine and often failed.
Nice to revisit this stuff.
Damn i need some of that
why would the equations need to remember anything? surely the physics is only based on the current state
@@tommclean9208 yes, technically if you know the current state, you dont need memory. However its not possible to know the actual current state of complex rheological systems. For example, you have large molecules with functional groups that interact in different ways with each other and other things in the system. If you allow that system to sit, those molecules will form large structures with ofher particles and with themselves. if you apply shear to that system, those structures break down at some rate, woth different structures breaking down under different shear conditions. So of you have a memory of prior conditions, and good experimental data, you can predict how those structures have broken down and apply that to your similation. Without a memory, you cant do this. These are thixotropic fluids.
Instead of filling out the entire test matrix, you can use something called modern design of experiments.
You really don't want 10x6 points, you want a formula (response surface). Knowing the general shape of the formula, you can use software to calculate which points of the matrix give you the most leverage and only measure those.
In a single-shot windtunnel that can save you lots of time.
I think the best use of DOE is to maximize the data you learn from all of the points you can afford to run, not to see it as primarily a way to reduce the number of points you need to run
is that anything to do with complex mapping?
@@billynomates920 Not that I know
A basic taguchi (dunno if this applies to aerodynamics?) should make it an order of magnitude fewer tests to run
Can you elaborate how this works?
Published books are actually an end run around ITAR and related laws. They're covered under the first amendment. Early encryption pioneers used them to get around export controls for stuff like RSA and other important algorithms that form the bedrock of the secure Internet today.
wouldn't the exact same thing apply to a video?
I don’t know if I would risk 20 years of my life on this defense in court. I also wouldn’t even want it to make it to court.
I imagine it's also more that "if it's in a published book it's probably ok to discuss" and less "if I publish this then it's exempt"
There are plenty of things that if you tried to publish them, they'd either never get published, or you'd get in big trouble for it.
Remember the T-Shirt with the DVD decryption key on it?
Yeah remember when Defense Distributed tried that? They spent almost a decade in court and the case eventually got dropped because the state dept dropped jurisdiction to another agency.
I know I just made 2 comments, but I just wanna say that I graduated middle school today, and a new video to watch really made my day after a long day of standing and thanking people, so thank you Joey for doing what we all wanna do but don't have the time or experience and thank you for filming this journey in such hig quality cinematic production.
Congratulations on finishing middle school! 🚀
@BPSspace Thank You you have no clue what it mean to be talking to one of my favorite creators
Congrats dude! I hope your next school gives opportunities to learn everything you want.
I don't know you kid, but you need to know that I see you and I'm rooting for you. Never give up.
Just today finishing middle school, heading to Aerospace engineering thanks to you
90% of the time I don't understand what he's talking about, but I still enjoy listening to him. Thanks for the great content. Big fan from Kenya
Same! I took classes in college as electives. I had no reason to take or basis of intelligence for understanding the professors. I just like to listen to smart people talk about things. Haha 😅I took constitutional law classes as a computer scientist.
Joeseph Bizzlington back with another slammer
YOU KNOW WHO IT IS
Who's that
@@seargesoren9391you know who it is
@Chezburger8
Where are these nicknames coming from 😭
A few years ago I was involved with a project that used explosives to cut heavy pipe in oil wells several km underground. The CFD modelling of various scenarios to predict whether the pipe would be successfully cut took a month of compute time on a specialised compute cluster. All to model something that happened in the blink of an eye.
I like how your early attempts mirrored that of early aircraft pioneers.
"Eh it looks like it'll work so it must"
"It just killed the test pilot. Twice."
"Ah I see the problem. Must've been this random piece that somehow helps."
"It just killed the test pilot. Twice."
Stop! Stop! He's Already Dead!
“Man we should really commercialise our resurrection system…”
“Yeah but the Kerbals hold the patent on resurrection…”
Dude!! I am 45 years old. Have a Third grade math education. Have no clue what all the math was. But wow that was a fun video. I just love seeing Someone so passionate. Thank you for letting us be a small part of your space shot. God bless brother.
Joe, I'm a engineer.. My main youtube is populated by Engineers/Makers and Music. You and Tom Stanton are among my favorites because you both do real science and engineering. I love watching you experiment and learn. I don't need the dirty details. I enjoy the process of development, the successes and failures. Thanks for doing this. I've learned a ton that applies to my design work. None of which involves things exploding or burning rapidly.
Just completed my Masters in AE and am currently working on CFD for the second straight summer at AFRL. Loved the bits on CFD, and got quite a few chuckles out of the "going into the weeds" bits after my in-depth CFD course this past spring. Also, after researching for a few years, ITAR is everywhere. If you want to do almost anything in AE that, say, requires a US citizenship, it's likely going to be ITAR. I respect the knowledge of both the physics and the regulations so you can keep yourself safe! Keep up the good work and I'm excited for the space shot!
The ITAR question has been on my mind with respect to your videos and I think you explained it exceptionally well. As an engineer with a family I can say you are taking the right approach to sharing and keeping info in confidence. If I had to choose between my engineering career and my family I would choose my family without a second thought. This video was awesome with the CFD and the normal rocketry updates, but it reminded me about being the best person I could be. For me, that is a family man.
Great approach, great content, amazing insights, keep it all coming!
I have to say that “linked right here…” bit literally made me spit out my coffee. Best part of my day. Solid. Thank you
The secret about high speed wind tunnel testing is called Reynolds’s number similarity. You can test a subscale version of your fin and then non-dimensionalize the coefficients to then apply them to a larger fin.
CFD does solve the NS equations. It solves them numerically so instead of a derivative it would use an actual number. The amount of non linearity you are capable of capturing is dependent on the turbulence model you are solving. For instance the SA, two equation model is a linearities model. However, model such as LES capture and predict much more of the non linearity of three system. Then there are models such as the RANS models which is appropriate for steady state solutions only, where LES is applicable even to time accurate simulations. The crux of the problem is that all turbulence modeling is numerically implicit.
The reason it needs to converge is because the equations are implicitly defined. You have the same number of unknowns as you do equations meaning that it can’t be directly solved. So what you have to do is guess a number, calculate the solution, use that solution to calculate an updated solution, and then compare the results. The difference in these is the residual. When the differences between the current calculation and the previous is low enough, you make the assumption that the solution is converged as the solution is not changing by updating the calculated solution. There is more to it than it that is a simplified way to think about the solver.
One of the elephants in the room of Cartesian solvers is the viscous shear gradient. Because the grid cannot correctly capture the geometry, it can’t calculate viscous drag. The reason your Cartesian solver overpredicted the values is because Cartesian solvers cannot correctly estimate viscous drag as they do not capture the surface as you were saying. So what they do is use wall functions ,or extrapolated friction equations to estimate the viscous drag. They generally are conservative in this method. Even the fancy solvers that use polyhedra and polyhexcore or even adaptive refinements use empirical friction models the calculate viscous drag. The are of friction modeling boils down to how the turbulent kinetic energy production rates are calculated. These are usually based on testing that has been done and is nondimensionalized and scaled to your grid.
On the note of ITAR, just because it is already online does not mean that it’s ok to share. I’m sure you know that though and thanks for your awareness in this area.
Ideally you match Reynolds number AND Mach number. I'm pretty sure I'm allowed to say that, because www.grc.nasa.gov/www/k-12/airplane/airsim.html says it too
How to do that in practice, I'm guessing is ITAR, unless there's old NACA (proto-NASA) papers about it
You don't need a turbulence model for CFD and it's got nothing to do with the particular discretization of NS. With DNS/ILES you solve the discretized NS eqns numerically on a fine (close to kolmogorov scale) grid. With a turbulence model you can use a coarser grid and sweep some of that under the rug using experimental correlations to relevant problems.
When you have the same number of eqns as unknowns, this means you CAN solve the system (think back to basic algebra). The difficult part is your system of equations in matrix form is way too big to invert directly so you do have to use funny techniques to progressively guess solutions.
I took CFD and turbulent flow in grad school.
I really don't mean to be nasty - CFD is really interesting and amazing and it's always great to have more people interested in it - but this is wrong on a lot of levels, the person above has correctly pointed out a few things. URANS = unsteady-RANS models exist and are common. All turbulence models are not accurate in time, except perhaps DNS, but many are accurate in a statistical sense, i.e. they capture the correct average velocity. Besides, turbulence is chaotic so tiny errors will lead to large scale deviations down the road (butterfly effect), so there is not really such a thing as "time accuracy". The residual is not the change between consecutive solutions, it is the amount by which the equations are not satisfied, i.e. if you're trying to find a vector x such that Ax = b (where A is a known matrix and b is a known vector), the residual vector is r = b - Ax. If you have the correct x, then it satisfies the desired equation and r=0. CFD packages spit out a single residual number instead of the full residual vector e.g. by summing the squares of each component then square-rooting (L2 norm), or by finding the maximum value (Linf norm), or by summing the absolute values (L1 norm). In fact, a lot of times the convergence stalls and the residual stays high even though the solution does not change between iterations. Just because your solver is unable to find a better solution, it doesn't mean it's solved the equations. There is so much more that could be said - I really recommend reading a solid CFD textbook like Versteeg to understand how CFD really works, it was very helpful for me when I was starting out. Again, I hope not to be mean here, I just want to avoid confusion!
I have written my own direct numerical simulation (DNS) code that is being used for research in academia. Happy to point anyone towards helpful resources on CFD.
@@GOAFPilotChannel you are right, you don’t need a turbulence model for CFD. I agree with that. Not sure what point you are making though. Can you clarify?
Hah you’re totally right about the number of equations too. My dislexia got me in that one and should have said If you have more unknowns than equations you can’t solve it. Hah dang.
@@pupudski8066thanks for the reply. I agree no turbulence models are strictly accurate in time, but LES has more applicability in a simulation where time accuracy is desired. Time accuracy is not typically a thing, but I do it. When using linear turbulence models such as SA, the transient solution drifts further from reality than LES. I use it to calculate dynamic aerodynamic coefficients that are compared to test data, and have done the sensitivity study comparing different turbulence models. LES is more applicable to time accurate solutions than RANS even every time I have done this for many flow regimes.
I agree, that’s not exactly how residuals are calculated and even stated so. Not sure why you’re calling me out here. It was an attempt to convey the complexity of the reality with something that someone with less knowledge might understand.
CFD takes a lot of computing power; a previous employer built a CFD facility a few years ago, the computer system occupied a complete 20' ISO container and they had to have a new substation built to power it. The cooling system was used to supplement the heating of some of the buildings elsewhere on the site.
That footage absolutely blew me away. Some of the coolest stuff I’ve seen on youtube in a LONG time. Please never stop innovating :)
@22:30 best way to explain to us amateurs....."if i can figure it out, so can you....jus keep trying!"
I absolutely LOVE your videos. And so does my daughter.... you keep us inspired to keep going even wen our models fail to take off... with continued testing....we hope to get our Starship model to lift off soon...
We love you Joe !!!❤🤟
Love the “not teaching wrong people about building ICBM” segment. Since the Mark Rober video that has been an elephant
As someone who deals with information that may or may not be ITAR or SBU (Sensitive But Unclassified) data on a regular basis, I feel your pain. I build and maintain archival systems for folks who have such data. The regulations are deliberately vague and what was ITAR information 15 years ago might not be today. On the other hand, things that are 60 years old and you’d assume have been long superseded by newer information or methods, are still considered sensitive. Worse, an image that a normal human wouldn’t pay any attention to might a eureka moment to some bad guy. So you end up fuzzy up images so that certain areas just aren’t really discernible. On the other hand, unlike you, I have an approval for release process that covers my ass. Before we release something for general public, several experts have signed off on it. I’d suggest that you do something similar. Not only would it help cover your butt, but they might spot something that you think is obviously not a big deal but might offer someone an insight. The trick with imagery is to not only examine the object or objects, but to diligently look at the background. More than one person has been bit by that.
But all in all, good on you for deliberately being cautious beyond the scope of the laws. Not only to keep stuff out of the hands of bad guys, but out of the hands of the eager but not quite so sophisticated beginner. This stuff is dangerous. A lot of folks don’t get how dangerous. So we have to help them out. Keep up the good work.
Might be some Palestinians learn a better way to make their rockets. This may cause a havoc in Israel
@@starship3812 Do NOT let us start that topic here pls! 🙄thank you!
@@starship3812 They already have thousands of rockets, some are probably self built. Like Joe said, if you do your research you can find all of this information in publicly available books and papers.
publicly available info cannot be subject to ITAR. it's already out there. especially if the book is sold internationally. reading books to people is not a criminal act.
@@VulpeculaJoy i mean they can improve it further. Or they can develop other advance aerospace weapons not only rockets.
These are some low profile information but also needs to be regulated cuz these are enough for making enough damage. Otherwise starship level technology must be regulated carefully otherwise chinese can steal it.
joeseph bizzlington back with another slammer of a video
I have always been so amazed by how much detail you explain the ingredients of your solid rocket propellante
Up to this point I knew about the grey area, but never realised how much it applies to your channel. Let's hope you never break the speed limit in that regard
From a mech engineer who basically just does FEA all day (not CFD yet). Agencies who run HUGE models have actual servers that run their sim’s at a reasonably quick speed. And those will only give you an extra couple of percent ‘accuracy’. A seasoned design engineer will know if the design is worth Pershing from the ‘simple’ results you would get from a 1 day simulation which you could use during preliminary phases and then run a 3-5 day simulation while you do other things to just fine tune your results for software inputs.
So glad we could help find the rocket! Made for a fun end to a day at FAR :)
That flight video was beautiful. Can't wrap my head around the amount of authority that small movement has on the rotation, seeing the set point and actual angle so closely tracking!!
I did a similar project in uni. Not nearly as high speed though. I derived a semi empirical formula for deflecting fins and did some corrections for low speed flight from a wind tunnel. The fin results were not that good but the air brakes ( yes I had those too) numbers were pretty good. I ended up doing some in flight testing with the air brakes and it gave me good results on the accel plots.. I have this data published on a report if you are interested
I just want to say a genuine thank you. Prior to you addressing the risks associated with running a RUclips channel dedicated to things like steerable rockets I had put little to no thought into the significance of remaining in that grey area. If you ever feel like you need to forego the inclusion of some piece of information for your own safety, don’t hesitate. Don’t get me wrong, I’d love to learn it… that being said I would much rather know this is something you enjoy and is not at all a detriment to your comfort. Keep up the good work! Thanks again!
It seems that Rocket Science is actually just a room full of elephants
So rocket science is a zoo? 🤔
That's engineering in general. You always make compromises, you always make simplifications, you always aim for "good enough for the intended purpose"
@@alexandredevert4935 and then it goes wrong and we do it again. I love it
I think there's a Far Side comic that alludes to that realization
@@alexandredevert4935It's always more satisfying when it lands just right though.
When I was in highschool (1970's we built a wind tunnel using bits and pieces from junk yards. The fan blades were from car wrecks, the fans used for radiator cooling. Two of them and a strong electric motor. (I don't know where from, but something from the metal forming lab. A lathe or something. ) They also played a big part in making the chamber.
That is the finest disclaimer I've ever seen.
I'm impressed that a RUclipsr is consciously thinking of improper, illegal, or terrorist results of his channel video's. I applaud your determination to learn about rocketry, that so many have over the last thousand years. We would never have made it to the moon without folks like yourself persevering to achieve a goal.
At 33years old, I can’t divide fractions, hell I can’t even read out loud that well. But your videos are entirely understandable and immensely entertaining. Also props on the quality and editing. I’ve been here since day one and it’s wild how much I look forward to new content.
I love to think about where the next generation will go if we can make it not suck so hard for them. You are such an inspiring and effective science communicator and creator.
Now we're talkin
10:30 - Joe you might want to double check that. NS is *compressible* viscous unsteady flow. Unless you enforce incompressibility of course.
16:31 - Did I hear robust control? Have we finally moved on to time-domain modern control? Woohoo!!!
24:06 - Y+? You've been doing too much CFD...
Have you thought about solving the Euler equations instead of the NS equations? It may reduce the computational cost for your run matrix, especially in certain regions of your run matrix. (If you mix RANS and Euler in parts of your run matrix then does that count as multi-fidelity data fusion...)
Great to see rocketry explained in a relatable and casual fashion, feels like going to engineering school again 😅 Hoping to see more about pitch and yaw control.
12:06 - All of AI is just linear algebra too
basically chat GPT is just a 400 million element matrix. (oversimplification, but, still, it's all just linear equations)
The activation functions (at the neuron output) are deliberately non-linear, otherwise the whole thing would reduce to a single (huge) matrix multiplication and we want to approximate functions that can’t be approximated using that.
@@peterfireflylund yeah, but his point remains. AI hasn't gone beyond giant matrix multiplications for about two decades now.
@@peterfireflylund ruclips.net/video/krixaEhLnlA/видео.html
@@peterfireflylund True. Although RELU has pretty well replaced sigmoid, etc. Which is still basically just linear.
Non-linear equation of motion? No problem. Taylor series that crap and you’re on your way!
Thanks for compressing the years of CFD and FEA I had in school into a far more digestible format!
My (oversimplified) assumption that the line you’re not crossing here is using all of those different speed/angle/pressure simulations to create an interpolatable map of forces that in effect replace the trig function you start out with for the gimbal case. Instead of one variable [f(theta)] it’s got all of those inputs [f(theta, velocity, pressure)]. Not to different from an open loop fuel injection map. If I’m right, blink twice in your next video on trans-sonic domain.
PS: I’m a “US person” so it’s cool…
PS, I'm not a "US person" so if that was it you are so royally screwed!
Which is why you can't just post the actual solutions online.
Oh and the US basically forced an even worse policy into nearly every other country on Earth: We aren't even legally allowed to stumble into those answers on our own, let alone talk about them with *anyone*. No one would ever dare to fly (or even simulate) a model rocket with *any* sort of active trajectory control over here.
I’m a (retired) s/w kinda guy and I know nothing about building rockets and controlling them. BUT, I learn something cool from each and every video you produce. I also laugh occasionally (well done). Keep up the great work! Thank you!
PS: I once did a tour inside a very large wind tunnel. No way it was a $200/hr. 😮
I feel like you made this video specifically for me after commenting your PID video to me last week on Twitter lol. I feel seen bro…seen
Solution to the ITAR problem? Convince Gaijin to add rocketry into War Thunder and wait until someone leaks the information in question.
Transonic flight video? Would love to see it! Just a quick side note, your videos are absolutely incredible. From the engineering to the humour.... It's just the best thing known to man. It's amazing to see the passion you've got for rocketry and aerospace engineering. I know you have heard all that before but it's got to be said again. Keep it up and stay safe (from ITAR...)!!!! 😂
Gotta love ITAR. AKA the biggest thing stopping me from putting time into my rooftop phased-array radar dream project.
(In my case the challenge isn't "what can I talk about" it's what can I buy COTS or otherwise not "specially designed" so I can manufacture using cheap overseas factories vs $$$$ domestic ITAR-registered vendors or developing the capabilities in house). ITAR controlled software and gateware? Just don't put it on github, problem (mostly) solved. ITAR controlled PCB designs so I can't make them in China? 10x project cost multiplier.
I’d rather make a phased array detector. ;)
re: ITAR I 100% get (and agree) with the better safe than sorry method. I work for a company that deals in ITAR regulated items (And not even "Maybe regulated" but explicitly regulated) commercially, and to avoid any kind of issues we don't allow anyone outside the US to order from our website at all, not even clearly un-regulated items like shirts, drinking glasses, etc, simply so that if anyone ever comes knocking about it we're more covered than we need to be.
You could cut the costs from a bazillion to 1/5 a bazillion by setting up 5 models at once side by side each set at one of the angle deflections. It is still 1/5 of a bazillion though. Maybe NASA would let you use one of theirs for a RUclips presentation. They have done it before. Still, CFD is possible without a supercomputer these days (I think). Navier Stokes is a bitch though. I used to work for a company that built gas turbines. They spent close to a million dollars to get time on a super computer. They wanted a high accuracy model of airflow across a new blade design. They got a deal because they only used a few cycles at a time. After six weeks, they finally completed the run. The resulting animation was .75 seconds long. The engineers were ecstatic. If you can get better than a wild guess, you are the man.
24:48
OMG it follows the Roll Setpoint so well. I had a course on PID control two semesters ago. This may look simple, but it is not. Building and tuning a PID Controller so well is amazing.
Amazing Work Joe, keep it up. Love your Videos man. This is quality contend.
I respect the desire to not be in prison. I also don’t want to be in prison.
Regarding the legal issues - I was in a rocketry club where we built fairly high performing solid and hybrid rockets. After talking to some people we came to the conclusion that steering would be a red line even though we managed to get everything for fuel and motor development done above board. But Germany is also a bit less friendly towards rocketry... Stay safe and I wish you the best of luck!
Btw the footage at the end was great! Especially the overlay :)
Even as a professional team in a semi-related field. We too often go too far with too few simulations, the cost in hindsight is always large. However, choosing the wrong simulation to do, or modeling things you don't need too accurately is also doom.
Although the information you tell is far above my understanding most of the time, but you try to make it as easy to understand for everyone including me. Telling just enough for people to get directions to do the same, but not to much that other people get overwhelmed. I love the way you explain how you tackle the issues you find, on the wing, trying to get to your ultimate goal. Seeing the progress (how small as it could be) incl. thinks you learned so far is worth watching you.
Since i started watching your channel i have always found it interesting. Just show what you like to show.
Haha I was waiting for the ITAR shoe to drop 😂😂
one of the most important topic aboyut CFD that you forget to mention directly but still demonstrated is validating your CFD set up. Only thur validation can you make sure that your CFD setup is not total garbage.
Can I donate pc hardware to the cause? Knowing you are limited by computation speed during CFD hurts. Either shipping you parts or discussing running things remotely for you. Either way I'd love to help.
I really like how you phrased the portion about keeping yourself safe from oversharing. Another well done video Joe!
I have been doing CFD since 1992 when we used a CRAY
You had access to a nice computer!
Bravo to "I'm unwilling to accept a 5% chance". No one could reasonably fault you and it gets the message across really well (like "there's no point in arguing about the regulations in the comments plzkthx")
bro. you look like Elon Musk
Elon Musk looks like this guy
When I first saw the thumbnail, I thought to myself: "yeah, that miter joint looks like something I can simulate."
5:32 not hearing wardrobe in that list! Keeping costs low. Smart.
Your determination and never-quit attitude are right up there with 2Strokestuffing. This was an excellent video.
Amazing work as always. What CFD software did you use, btw?
I have the same question. Looks like Solidworks Flow??
Grid, legend and labels indeed look like SolidWorks Flow. I don’t think Joe will say, as he called it not the best program^^
Phenomenal video explanation. I am an electronics engineer, and often miss complex math and physics from college days.
That's a really nice shirt, any chance you remember where you got it?
Thanks! It's from Target - 23 bucks. I tossed a link in the video description :) www.target.com/p/men-39-s-floral-print-button-down-shirt-goodfellow-38-co-8482-white-xxl/-/A-88248707?preselect=88248705
Here is how to build the control algorithm:
1st model the forces by hand (depending on velocity, tilt angle, weight, etc.)
2nd assume error bounds +/- 0-50% (depending on what quantities, e.g. tilt force, drag or weight etc.)
3rd construct your state space model
4th programm the controller (u can try a PID if u want or look for fancier stuff)
5th make a lookup table for the control parameters with the varied model parameters
6th u monitor all past states of the flight
7th while in flight calibrate the controllers parameters to the monitored states (that means the parameters will vary while in flight!!!)
8th assume valid initial condition for the initial control parameters
Now u are ready to start your rocket straight.
PS: Brute force will only work with enough compute power.
Worrying about going to prison over something that isn't malicious is not ok. Means the law is horribly flawed.
People can do bad things without malicious intent. That's hardly exclusive to ITAR.
Cool video! As someone with a PhD in CFD, I must reemphasize (you did a good job pointing this out) how extraordinarily challenging CFD is. While many modern tools make it look easy, the topic is so incredibly vast that it is my opinion you cannot know what you don't know without months or years of education. Even professional engineers often do not know fundamentals required to produce correct results.
For example, it is necessary to use a "compressible" solver for high speed flows (roughly over Mach 0.3). Without this modification, results will likely be off by orders of magnitude by the time you get even to Mach 1. However, simulating compressible flow is a notoriously challenging exercise in CFD. It is often difficult just to achieve convergence, and much more so to get remotely correct results.
There is a laundry list of crucial comparisons and checks that must be done, but notable is a mesh independence study, where you increase your cell count until metrics of interest stop changing.
I am sure this video only represents the tip of the iceberg, and much study, care, and consideration went into this, but readers should be aware of how incomprehensibly deep the field of CFD is.
Yo, Joe, to make these sims faster you should use a program that I don´t remember the name of but it was used on a SalC1 vido of trying to find the pack.png picture of minecraft. What the program does is allow other people to share their computing power to run sims faster, and I think that with your comunity a lot of people would colaborate this way. (I would be the first to participate)
PD: Sorry if my english is bad, its not my native language. Thanks for reading.
I like this suggestion.
But the grey zone of ITAR, and sharing fin designs with strangers around the world, the answer will probably be 'no'.
There is no need to share the files, it works as if it were a super computer and Joe is the only one who can manage and view the files, the other people just process it without seeing anything or having access to anything.
Looking around at all the people who've lost their minds, it's super nice to find a group of people that haven't!! At least only in a good way! Always been an exceptional channel. Thanks BPS!! ♥🚀🚀
I wrote a 3D Navier-Stokes "windmapping" system for the last game engine I wrote that included LOD based on proximity to the camera - because it was only supposed to be an inconsequential visual effect to make particles flying around be affected by things like explosions and objects "moving through the air". It seems like whatever CFD system OP is using here should employ a similar LOD scheme. I went ahead and did something like an octree subdivision that was based on the camera's position, orientation, and view frustum, where higher resolution cells next to a lower resolution cell in the octree would sample from it and interpolate what a simulated "virtual" neighbor cell's velocity/pressure would be. I would never simulate a whole 3D volume, even with the axes squished in certain ranges (which is better than nothing) and go for something that is only high resolution around the mesh, while everything else is low resolution. It also helps to simulate across available CPU cores on a CPU that has a lot of cores - or use a compute shader on a GPU! I had my "windmapping" running realtime, along with my game engine, on an 8-core CPU. I believe there's at least one video of it in action on my channel if anyone is curious to see what that looked like. There were vector lines drawn for the cells and their length reflected the velocity while the color reflected the pressure (green was low pressure, red was high pressure). Anyway, that's my two cents :]
You’ve been doing perfectly Re: ITAR so far. Don’t change anything, leave out what you need to leave out. We understand. :)
Re: What to see next, I’d love to see more camera views and flight test data! Keep on with creative angles and sharing build process vids 🙏🏼 and keep up the great work!
probably not first lol
WINNER WINNER WINNER 🚀🚀🚀🚀🚀🚀
Was actually first
Neat.
the amazing video resolution and those live demographics were just worth devouring by my eyes!
Your RUclips videos are the reader's digest version of building a rocket!!! Just enough info to spark a interest but not enough to build a complete rocket in our bedrooms
The onboard rocket launch footage at the end was awesome! Also the HUD-style data had some real Iron Man vibes :)
That was some kick ass flight footage. Nice work.
I've just subbed after months of watching this amazing journey because of that end talk Joe, share what You feel is safe to. The struggle, joys of steps finally achieved on difficult field, that's what this channel is most valuable for me. Wish You many good flights. This whole project just looks promising, You are down on the ground whilst up in space, have a good one and keep them coming.
22:43 I have had the thought ana occasionally wondered aloud what bits are left out .
I appreciate the concerns and like the way Joe seems to be approaching the matter .
Pure scholarship and learning is fun, bit mindfulness of unintended consequences to a reasonable degree is wise.
Also, some elements are better for the learner to gain on their own experience, anyway.
I am quite impressed with the resulting roll control footage.
Excellent video all around.
I did my PhD in numerical (computational) methods (and work in aerospace) and was expecting another "RUclips aerodynamics special" ... but actually I was pretty surprised with this video. Good job.
I've had a whole semester of CFD studies in university. I don't remember much, but one thing that struck me was how small your domain is ! My teachers always said your output surface (end of the domain behind your object) should be 20-30 chord lengths away ! Input surface may remain 5-10 chord lengths ahead. Top and bottom surfaces should also be somewhat 10-20 chords away to avoid that issue you had with section decrease. So basically your domain should be 30-40 chords long, and 20-25 chords high. This ensures your boundary conditions are correct and do not influence whatever happens to your fin. Good luck !
Obviously this increases your number of cells, but then you may tune the mesh better to get large cells around the boundaries, and very small near the object. Also consider inflation layers to capture the boundary layer, and so on... Have fun 🙂
Armchain engineer here, but I feel like the way you go from test data (CFD, wind tunnel, or anything else) to a control system would be to create a sort of look up table.
Essentially, your test data lets you build a big table with (speed, pressure, fin angle) as indices, and deflection forces in each cell. In preperation for flight you would use this data to create a look up table with (speed, pressure, desired deflection force) as input and fin angle as output. You would probably want to interpolate between the values to get a smoother output.
I did something a little like this for sailing optimisation, using test data for sails, to pick the correct sail angle for a desired speed and heading.
As far as I know, the design for the WW2 Spitfire is still classified. (Almost 90 years later.)
I wanted study the cooling system of the P51 Mustang for the use in water cooled experimental aircraft, since it supposedly it had a net negative drag. (Thrust.) I even looked at some freely available CFD programs, but got lost in the weeds of grid design.
i really love the pi=4 theoretical, its so intuitive yet so wrong
One of those projects I keep meaning to try to build is a foam wind tunnel. The speed of sound is the speed where energy can be freely exchanged between kinetic energy and compression. Kinetic energy per unit volume is dependent on density, so speed of sound is dependent on density and compressibility. SoS in air is high because it is very light. SoS in water is very high because it is very incompressible. Mix the two and you get something intermediate density that is far more compressible than water. If you can produce a 50/50 foam the speed of sound is ~20m/s.
For CFD or physical testing, you can use a design of experiments method to reduce the number of runs. You are doing one factor at a time which gives you the best data at the cost of increased testing time.
Amazing footage! Can't wait to see the next video! Simplex V2 or rocket cameras would be great - no preference
This is so impressive and fantastic to follow along with, I love the magic 8-ball site... when it doubt play it very very safe. Stoked for more roll control!
Ive been watching this channel since 2019, this is one of the best videos you made, light hearted and a bunch ot technical details
The availability of sensitive engineering information is actually a point I never even thought about thinking about. I am from Germany and I consider basically everything accessible that is not intentionally kept a secret by some company/ government. That citizens of countries are legally responsible for what they share with me / an audience of known or unknown origin is crazy to me but in a weird way makes sense against. I guess some countries make it hard for themselves by restricting access to the Internet, but using state boundaries and citizenships as a deciding factor is weird. I am certain this doesn't actually work (to a notable extent) but the existence of such rules is new to me.
I might look up if there's an equivalent in the EU so I can be like " sorry mate, can't tell you that" to my colleagues (im studying in Switzerland)
aerospace student currently working on my masters degree specializing in "autonomeous systems" here, the control loop part was really the thing i was waiting for the whole time. But i never thought about ITAR, and its a very fair point to bring up. So your decision is perfectly understandeable, i guess i will have to figure that stuff out on my own :D
Great Video as always
"After every simulation, there is the experiment" - this was told me too often from more experienced rocketry guys around me and it's true as you also mentioned in your video.
I would love to get a deeper look into CFD - i was always on the construction side of developing and "only" did some FEMs but never get into CFD. I tried once to simulate stability for a rocket of a friend of me but im sure, as you said, garbage in - garbage out. So plz show more CFD stuff.. I guess i was on the same way as you but i dont have more time to explore the rabbit hole.
Ooo that footage is soo good, the roll program and new cam looked sick. I hope everything goes well so we can enjoy that space shot