ROBOFORMING: Behind the Scenes as Machina Labs (The Future of Metalworking) - Smarter Every Day 290B

0:00 hey it's me Destin welcome to the second channel here this is smarter everyday 2

0:03 this is the channel where I like to go deep into the Weeds on the topics

0:07 like more technical detail stuff like that it's also where I I'm a little more

0:10 laxed with the editing so I hope you enjoy this we're going to be at Machina

0:14 labs and we're going to be learning about Robo forming is what they call it uh

0:17 it's also known as incremental sheet forming

0:20 but we're going to learn all about this process

0:22 and how it works if you enjoy this the second Channel I'd appreciate it if

0:25 you would consider uh supporting on patreon that'd

0:28 be amazing or just I don't know subscribing

0:30 to the second Channel this thing's been going for a while here and it's slowly

0:34 getting a larger and larger following which is

0:36 fun but if you're into that feel free

0:37 to subscribe if not no big deal enjoy the tour with Ed and bobac here

0:42 at Machina Labs let's get to it okay we're all miked up so your Ed is

0:46 that right yes Ed May Ed May okay CEO uh CEO coo CEO CEO okay

0:52 what's your background man uh engineering um so

0:56 um as academically I was a computer engineer um you know spent early days most

1:00 of these working on machine learning stuff companies like

1:03 Google Microsoft um end up going to SpaceX

1:05 eventually um and as a software engineer there

1:09 when I kind of realized all the challenges

1:11 we had manufacturing stuff and how hard it is to change geometry and design uh

1:16 after that when I saw those challenges got

1:18 really excited about 3D printing uh so joined um I think Bob's coming to oh

1:24 hey what's up hey I'm Destin Bob nice to meet you nice to meet you Bob

1:28 cool perfect pronoun actually like to I've got

1:31 a friend named bobc yeah oh yeah yeah

1:33 okay so so now we have bobc bobc what's your title I uh had the technology

1:39 Partnerships at at Machina okay so he's

1:43 also a co-founder a co-founder you're a co-founder

1:45 okay so so you guys made it happen and my understanding is uh I I like

1:50 Machining and learning about stuff like that background

1:52 mechanical airospace engineering your mechanical no computer

1:56 science what's your background materials Mater oh

1:58 we're going to talk okay sounds great so

2:01 my understanding is you guys have a way to create shaped metal by forming it

2:07 by I don't know how to say it touching it in a very special way is what

2:10 I'll say but but you you come in and and you you I don't know

2:16 let me just draw it right here so my understanding is you guys have a planer

2:22 a plane of metal and you bring two tools in and you can touch the metal

2:27 like that and depending on the position of this what do you call these tools I

2:32 call them end defectors but you know

2:34 forming styluses U but yeah end Defector forming

2:37 end defectors okay got it so these two IND defectors when they come in drop

2:41 my marker so when these two IND defectors come in and they put Force right here

2:46 you can deform uh you can deform the steel in there and if you do it

2:51 in a certain way I'm assuming I I haven't got here yet but I'm assuming

2:54 the position of this versus this matters yes okay and so you can do things

3:00 that are really interesting yeah so I think the best analogy that I can think

3:04 of you know how a Potter you know you have a clay on your you know kind

3:08 of turning table and the Potter kind

3:09 of pinches the clay with their finger and slowly

3:11 deform it that's what we're trying to do with sheet um but you know it's

3:15 a very strong sheet and you know requires

3:17 thousands of Newtons to deform it but it's

3:19 very similar so the angle between these two

3:21 tips when they vary it changes your wall angle uh the amount of pinch Force

3:25 how much force you're applying also changes how

3:27 much you're deforming the sheet um so we

3:30 something complex happens between the two tips it

3:32 can be you know defamation with the with the pinch it can be Shear it can

3:36 be stretching there's combination of different mechanisms happening

3:39 until get you a part that you want

3:41 so we're I you are playing in the really fun area of the stress strain

3:46 curve like you're you're going plast past plastic

3:49 deoration and so my understanding if if I

3:52 let's just let's just go here so like let's say steel so steel we're going

3:56 to go up uh and then we're going to yield it and then we something like

4:00 that is that right yep correct well you're perfect high you know everything no I

4:04 don't I don't know everything so when you when you're pushing on a metal you got

4:08 what which one is this this is strain right no this is stress this is stress

4:12 okay stress strain strain okay Bob back I'm

4:15 coming over and looking at you now so

4:17 stress strain curve I'm going to just give you that so when we start deforming

4:21 the metal we get to right here and we get yield and then we get permanent

4:25 deformation right there right plastic deformation

4:28 but the in theory the the plastic deformation starts even

4:35 earlier but on a macro scale starts up there and generally as a rule of thumb

4:40 you do the 0 2% let's say type of thing then you define a yield

4:47 for yourself that's one going to be

4:49 my yield strength there so just by telling me that I learned that you know a lot

4:53 about fracture mechanics right because you you

4:56 start getting yields down here and stuff right

4:59 yes yeah okay okay we're going to be

5:01 able to talk I think this is going to be really fun so so what

5:04 do you do when you when you push these two IND defectors in together so I

5:08 I imagine you have many different things you

5:10 can do you can start mushing and molding

5:13 the metal that's the Potter wheel or the Potter uh metaphor right yeah yes you

5:17 can either push them together um and then

5:20 obviously like the area underneath expands um

5:23 but they get thinner the sheet gets thinner

5:25 but also like usually depending on the what wall angle you have you usually have

5:28 the tips actually slightly let's say we started

5:32 forming right uh you usually have the tips

5:35 slightly off right you have a tip here and a tip here between the two wall

5:41 angles like they don't necessarily need to be

5:43 on top of each other depending on what wall angle you're forming and how much

5:46 your thinning you're trying to create you change

5:48 the angle between the tips I see I

5:50 see so so if okay I I understand what you're saying I have no idea how

5:54 to point to this piece of metal and explain it but I'm guessing you have

5:57 tools we can look at we can see the robots downstairs doing the same same thing

6:00 should we go do it yeah let's do it does that work for you yeah oh

6:02 yeah okay cool here's your cap y thank

6:04 you great thanks for being willing to jump

6:07 in and go for it of course this be F you should like if you

6:10 are interested this is one of those that this is a cool thing yeah you might

6:16 want to know about this you should like this is tintic service steel okay so

6:22 it's not magnetic but then when you do

6:25 the forming because you're kind of changing how

6:28 the metal is structured then the formed areas

6:32 are becoming magnetic how does that work I

6:35 I don't I guess I don't understand how

6:38 a metal becomes magnetic so the the amount of work you put into the metal

6:43 induces some phase transformation into the metal like

6:46 your structure is essentially changing and that makes

6:49 the metal magnetic really yeah can you change

6:52 that based on how you work the metal how much work you put in determines how

6:56 much transformation you have and then oh it's

7:00 it's a direct correlation correlation with that oh

7:02 wow I did not know that so that in fact if you look at some

7:06 of your kitchen sinks and you put magnets in there and if it they're good

7:10 quality and what depending on you would see that oh it's magnetic here Etc so

7:16 so it's a known phenomena so my fellow

7:19 Metals guys will say oh yeah that's strain

7:21 induced you know transformation so it's it's well-known

7:25 phenomena but now with this technology you can

7:27 actually control it oh really you can dial it in you can dial okay that's pretty

7:33 interesting but but these parts traditionally would

7:36 be made with large large uh what do

7:39 you call them molds or mandrels yeah dyes or molds dyes yeah well some of them

7:45 cannot be made using dye so you have uh the video you have around stamping

7:49 so this one if you make a mold for it then this part gets stuck

7:53 on a Dye oh I see right because then there's going to be mold behind

7:56 and then you have an overhang angle here so the D will can you is this too

8:00 heavy to pick up yeah no pick it up it's just 2 millimet sheet can you

8:04 hold it right there please this is one MIM 32,000 32,000 yeah so so I see

8:10 what you're saying so it would get stuck on a die okay stuck on so now

8:14 with our technique the robot can get

8:16 into the part and actually form in the overhangs and um that allows you to do

8:20 things that stamping cannot easily do interesting unless

8:23 you comp create a very complicated die

8:26 that falls basically opens up into pieces afterwards

8:30 I see so so you you have access

8:32 to to more interesting geometries but the trade-off

8:36 is it takes a little more time to make the part it takes a little

8:38 bit more time to make the part but if you take in account the fact

8:41 that you had to make a mold to make the part then in a lot of cases

8:46 up to even thousands of parts we

8:47 faster really right because you can start making

8:49 the part you know two hours after your design is done with the mold you have

8:53 to go through iterations on the mold and usually the mold the design of the mold

8:56 is also slightly different than the design

8:58 of the part right because the sheet has

8:59 a spring back you spam bit it comes off it moves again right so your mold

9:04 is usually different than your than your actual

9:06 geometry just not the negative of your geometry right so you have to go multiple

9:10 trials potentially on the mold there simulation software

9:12 that helps you get close to what the mold needs to look like to get a certain

9:16 part but they're not super super accurate

9:19 so for some complex geometries you have to do multiple molds so by the time

9:22 you're going through those iterations you're spending hundreds

9:25 of thousands of dollars you get your part you might be even FAS just forming

9:30 with us but then once you're forming the part

9:31 obviously we're slightly we're slower than stamping stamping

9:34 few seconds you get your part so this is great for development and low rate

9:37 initial production yes I think that's the first

9:40 area that we're going after but you'll be

9:42 surprised where where the break even point is

9:44 it's further down the the chain than than

9:47 the the chain you think you know we were talking with Tesla a few few uh

9:51 few weeks ago and they spent $150 million on a Stamping Plant in in toson

9:56 right um so $150 million meaning that like

9:59 you have a very huge facility the stamping

10:01 presses you have to have area for storage

10:03 for dyes so once you start thinking about the full cost that it takes all

10:07 the equipment machinery to die and then the equivalent

10:10 at how many cells can we buy

10:12 with our technology then the throughputs become very

10:14 close interesting right so but but you're right that it makes a lot more sense

10:19 for you know lower volumes and that's

10:21 where we start right interesting yeah I want

10:24 to see the machines yeah let's go we do it yeah let me grab my bag here

10:27 that's actually it's a funny thing we're

10:28 you're doing a special project we we're you're

10:31 going to know more about it uh in the soon but um we are forming

10:35 a hood of a car over there the hood of a car yeah yeah I actually

10:38 was reading up on this technology and my understanding

10:41 is like I think Ford helped develop

10:44 it a long time ago is that true or is a little different so there

10:47 is a there's academic Legacy probably I would

10:50 say 30 years of academic Legacy uh some

10:54 of those folks are actually advisor to our company

10:55 some of the bigger researchers in the field

10:58 like Northwestern has been working on it

11:00 uh some folks in Europe have been working

11:01 on it um but then some corporations like I think Ford Boeing Nissan did some

11:08 R&D work in this um but nothing really

11:12 commercialized out of those those efforts so we're

11:15 kind of the first people who are bringing

11:16 it commercially out there and there are few

11:18 enablers we applied like you know being able

11:21 to model process using data as a poos to relying on physics Bas simulation um

11:26 using more cost-effective systems with robots as opposed

11:30 to building very complex customade systems so there

11:32 are a few enablers we're adding to it that kind of made this possible

11:36 in a commercial software and and software yes yeah

11:39 and that's where you come in uh that's right yeah yeah yeah okay um so the yeah

11:45 the software side is is yeah the other piece of it which is one

11:49 of the core enablers uh even though I

11:51 come from a software background I have been I've

11:53 been in manufacturing for a long time right so uh before this I was at a company

11:57 called relativity space I was in in charge of a 3D printing team that oh

12:00 yeah um mayor um yeah so so been working on intersection of robotic software Ai

12:06 and Manufacturing for the past uh 10 15

12:09 years it's amazing it's beautiful like how close

12:13 can we get can we go to the frame yeah let's go let's go I

12:16 think that one is actually our design our geometry so you can show all of it

12:20 so you're running cuas uh we runas at the moment the robotic system is cuka

12:26 um but we also can use fanex so you can see on the other side

12:30 those are fanic robots y uh we are robot agnostic um those are yeah we pretty

12:35 much basically build the whole control system

12:38 for the robots from scratch so it's not dependent

12:40 on any features that the ca has it's pretty difficult to film yeah it's pretty

12:48 cool so it's like a dance can can I go talk to yeah of course gentlemen

12:54 come on look at this real quick oh

12:57 that's amazing it's different than I thought it's

13:02 actually pushing sorry I'm being rude I'm I'm

13:06 destined Mark what's up man nice to meet

13:08 you nice to meet you how you doing man doing well is it okay I

13:11 love your channel oh thanks I appreciate that I I love your robots is pretty is

13:15 pretty cool yeah pretty so you're Mark I am all right so what's your background

13:20 um my background's in composits actually um

13:23 but Bobak recruited me and found me and now

13:26 I'm a robots guy oh really you can switch just like that that it was

13:30 a steep learning curve I think but it's been a couple years um can you show

13:33 me what's happening over here yeah so

13:36 this is incremental forming yep incremental metal forming that's

13:41 right yep uh it's 2 mm aluminum sheet and there's one rope on on each side

13:45 and they're sort of pinching and uh it's a lot like um like a potter's

13:50 wheel sort of like spinning like that but here

13:53 the uh the tips are moving instead

13:55 of the work piece and it doesn't have to be round right you can sort

13:58 of push any shape you want and so layer by layer uh we're sort of stretching

14:02 and pinching and pulling uh whatever shape I want here out of this metal so

14:07 so the triangle here is coming out

14:10 of plane that's right which means this one over

14:13 here uh are there any Optical lockouts I need to be aware of nope you can

14:16 walk right up to it can I walk here yep all right thanks just try not

14:19 to get between the robot and the sheet you should be good yes sir so

14:23 so this one is pushing that's right so this R do you have a force

14:27 gauge on this IND Defector yeah there's load cells on both robots so you can

14:30 feel exactly what they're doing um can I show the that too or we keep it

14:35 on the screen here you can see over here what's going on so so I could

14:41 imagine here's the workstation here's this is

14:43 the forces that both robots are feeling right now

14:46 okay so which can you tell me what colors are what uh yeah there's uh

14:50 the Z component the inplane component and then

14:53 green is just the total okay um and so there's robot number one here is the one

14:57 that's pushing and then robot number two here is the support that's on the other

15:01 side it's pretty easy to understand the forces

15:03 pushing into the plate but the inplane forces

15:06 are the side to side forces as the thing

15:08 moves around there's what you could do is you could do this process with just

15:12 one robot and you could just sort of poke in metal and basically do a bunch

15:15 of work by stretching but the boundary conditions like it does it then you load

15:19 the whole thing up in tension right so what we're doing by having a support

15:22 robot on the other side is you pinch and so you localize all the force Just

15:26 Between the tips um which sort of reduces tension on the sheet um and makes

15:30 it so you can form a lot more

15:31 accurately that's amazing I'm sorry that I'm interrupting

15:34 you because I'm just very excited and I realized I'm being rude so I'm so

15:38 sorry all right so so if you were let's can we talk about that over here

15:43 yeah so if you were to not have this robot and only have the other

15:48 robot yeah do they have names um sometimes

15:52 uh these don't have names yet these are

15:54 one and two um but over there we have Jeffrey and Stumpy okay and over there

15:58 we have thas and princess that's fantastic okay so this is one and two so is

16:04 this two uh this is yeah this is the support robot this is robot

16:07 to the support robot yeah okay so um so one is pushing so I would expect

16:14 a positive force or compressive force on that one and this one do you if it

16:19 was just holding in one position would you

16:21 just hold the robot here until it fills

16:23 force from the other side so that you know that you are actually pinching we

16:27 get pretty fancy about about it where basically

16:30 okay should I talk about all this stuff

16:31 am I allowed to do that should just well should I let him do it since

16:35 ultimately his authority no no no I mean I I think I think Mark Mark knows

16:39 yes go ahead now continue sure yeah just

16:41 Happ yeah really there's control systems on both

16:44 of these robots where we can sort

16:46 of plan their nominal trajectory that just sort

16:48 of is the theoretical what they should be doing um they also like they feel

16:53 the forces and we use that to update

16:56 um to make corrections to the trajectory Bas continuous yeah so one of them is

17:00 correcting for its own deflection underload because they're

17:03 not infinitely stiff the other one is trying to hit a Target pinch force that we

17:07 set okay okay so it's it's an interesting

17:11 combo right you're forming parts that are like

17:13 half a supposed to be in the end like you're pinching a half a millimet sheet

17:16 sometimes I think this is a 2mm sheet so yeah you have two robots

17:20 on a pretty large envelope right so the envelope

17:22 is 12T X 5T getting two industrial robots

17:25 to be very accurate in this envelope is one task that we have to do

17:29 through a lot of kind of calibration

17:30 and proper kinematics um but then the moment they

17:34 start touching sheet the sheet were resistant so

17:36 and these robots are pretty noodle you can

17:38 kind of imagine every joint slightly deflects right

17:41 so there are controls and mechanisms that basically

17:45 uh compensate for that deflection to to so

17:47 the robot stays accurate under Dynamic load

17:50 right sometimes depending on what sheep re forming

17:53 it can be like 20,000 Newtons of force

17:56 that you're applying it's like as high as a weight of a truck on a very

17:58 very small end Defector on the forming side right so being able to and under

18:03 those loads robots sometimes deflect 6 7 mm right and you want to pinch a sheet

18:08 that's only 2 mm or half a millimeter so we have to adjust for that so

18:12 there's a control Loop that uses to force sta some of the other data

18:15 to constantly figure out how it can stay

18:17 accurate and and pinch the sheet the right

18:19 way but if I uh can you give me an IND Defector with your finger

18:23 like that just straight up so I can imagine if the two IND defectors are coming

18:27 in and touching each other do they

18:29 always stay perpendicular to each other or normal

18:32 they they depending on what all wall

18:34 angle they're forming and what is the thickness

18:35 of the sheet they might they basically they the whole um uh angle radius between

18:41 the tips is part of the part

18:43 of the forming process depending on basically what wall angle

18:46 you're trying but it is a sphere uh we have different tens tips there's a sphere

18:50 tip there's a flat tip uh depending what kind of effect you want to get

18:54 out of the part you might want

18:54 to use different end effect different radi as well

18:57 if it's a sharp crease okay so that detects the smallest feature that you can do

19:01 like you've done like quarter quar an inch

19:04 diameter or some very fine detail features

19:07 I think these are what 3/4 3/4 an inch diameter yeah and what is the thickness

19:12 of this plate again 2 mm 2 mm so like 880,000 wow can you tell

19:17 me about the uh the boundary conditions I'll call it I don't know what to call

19:21 this but the uh the frame yeah so you have a term yeah we call it

19:25 uh the fixture or the frame actually you were right we call it the frame she

19:28 filming frame um so you can see hydraulic

19:31 clamps all around that you know just creates

19:33 boundary condition for the sheet and holds it in place um we used to just use

19:39 this with clamps but we're going toward

19:41 the uh toward the future where it's automated

19:43 the robot can pick up a sheet with the suction cup put it in the frame

19:47 then the clamps automatically close hold the sheet

19:49 and then allows us for the sheet to not slip while we're trying to form

19:52 that's amazing um you can imagine in the future

19:55 you can also play with this boundary condition a little bit to let a little

19:58 bit of a slip right slippage or like uh defle I don't know there's there's

20:04 so much vocabulary here that I don't have right yeah so this is this is

20:08 a hard thing that's the boundary condition essentially

20:12 you you want to have a variable boundary

20:14 condition to adjust and accommodate for what's

20:17 happening during the forming process essentially got so

20:21 you control the amount of basic pressure you're

20:23 putting on the sheet uh from the classs

20:26 okay it what is what do you call it when if I were to take

20:29 a sheet and I were to touch in the middle I would get more deflection than

20:33 the edge is that edge effects is that what that's called or I we call

20:36 it like boundary condition effects that's what you

20:38 guys call it yeah but uh but you're

20:40 right that's something that you have to account

20:42 for um boundary condition affects the accuracy of the parts that you get so we

20:47 have a whole software stack that tries to account

20:49 for that right um so you form a part and then so you can imagine

20:53 you form apart you can then scan it actually the same robot scans it afterwards

20:57 to do quality um to figure out yes exact quality or I should just figure

21:00 out what it Formed to update the model and then it will like okay I formed

21:04 this these are the areas that we're off

21:06 that means maybe I should have pushed more in some areas and pushed Less in some

21:10 others and can manually iterate on that keep

21:13 generating data and then later create a model that will tell you okay in order

21:16 to form this geometry and this location

21:18 in the sheet you need to actually form a completely

21:20 different part and in the end you're going

21:22 to actually get the right part with all these things you talk about you might

21:25 overshoot here unders shoot here just like injection

21:28 or something all these processes have that what

21:31 is the what is the Beni of this you

21:33 know the Beni 3D print the boat that you print when you're testing a uh

21:37 you're testing a 3D printer do you guys have an equivalent or what is it

21:40 Mark like a cone a pyramid little thing just a cone yeah my favorite one is

21:44 the face now actually it just looks cool yeah it looks cool um and so you

21:49 guys will print that to test new software yeah um what are you optim we do

21:53 all kind we we can't decide we keep doing cool new ones um what are

21:56 you optimizing for like everything accuracy speed surface

22:01 finish um yeah all three at the same

22:04 time is is the is the hard part that's the Holy Grail yeah it was

22:07 hey nice talking to you man this is awesome dude I really appreciate it yeah

22:11 and this is going to go to next week it's going to get shipped to Chicago

22:17 fapte I don't know if you ever go to fapte trade show it's like a lot

22:21 of fabrication technology the machines Fabtech it's

22:25 like a biggest manufacturing show in the United

22:26 States probably after imts you're telling me

22:29 this now and it's next week now I want

22:31 to go and I can't go the same thing is going to go on a truck

22:36 get shipped back there oh really so

22:39 no installation not no pouring concrete this whole

22:42 thing is going to go is a big deal for you guys this is going

22:45 to be a pretty big de so can you imagine all the other cells if you look

22:47 at it they're all in concrete right and we had to actually if you go

22:52 around you will see the concrete that's under

22:53 the cell looks a slightly different yeah it's

22:56 new than the rest you need a foundation for no vibration yes to just hold

23:01 these things like I said sometimes they apply

23:02 Force as high as the weight of a truck

23:04 all that for just goes into the foundation

23:06 right um and and the regular warehouses

23:09 you do like you know 5 in of a like a 3,000 psi concrete if

23:13 it's best you have to do 13 in of uh 5,000 psi concrete to just

23:18 be able to and with rebar and everything to be able to uh counter the forces

23:22 with this new platform we're completely decoupling

23:25 foundational requirements so you can put this actually

23:28 on the this actually fits on the back of a little boy and you can kind

23:32 of ship it anywhere in the United States

23:33 that's cool that's very cool yeah man this is

23:36 like this is mechanical engineer geek Heaven

23:39 this is great one thing that's interesting to me

23:42 is you can see you can see the whole sheet deform if you look very

23:52 close the whole there is like a deflection

23:56 that you see obviously but then that's where

23:59 the the controls come in to accommodate for most what is the step over

24:05 on this particular run 0.6 mm6 mm yeah the face is 0.3 that's one of the many

24:15 knobs we can turn I see I see and so I assume the the lower

24:23 the step over the the better the surface

24:25 finish that's right yeah but then it takes

24:27 longer you know it's just adaptive kind

24:29 of step over so as opposed to Machining where

24:31 you go in kind of water lines we can potentially say just fill the whole region

24:36 right oh yeah so like when you go to a lower shallower wall angles just

24:40 increase uh decrease the layer height so you

24:43 get more layers and steeper wall angles have

24:45 less just like when you're doing CNC milling and you take a bigger bite bigger

24:48 bite slower bite whatever or smaller bite that's

24:50 cool so that end effector how much uh how much engineering went into that Oo

24:57 we probably went through I'll show you some

25:00 of the designs but we went through six

25:04 seven different designs for that end Defector from just

25:07 being the static stylus to different types

25:10 of coating that goes on the stylus what

25:12 the substrates look like can it take the forces and not Bend um doesn't is it

25:17 going to not shatter if we suddenly shock

25:19 it uh because it happens sometimes and then

25:22 does it need to move do you need a some kind of a bearing so

25:26 that it reduces the friction forces uh so we went through a bunch of designs

25:31 and this is actually designed there's like two

25:33 more designs Beyond this that we have done

25:35 so this is actually three version two versions behind so it took me a second

25:38 to realize it's actually rolling isn't it yes

25:41 so this is specific one does there are

25:43 other ones that we have that roll in different ways um uh like a ballpoint pen

25:48 yes exactly so this one is just uh I can't predict where it's going Mark you

25:57 have a complicated program so yeah it's

26:00 rolling along the pointing axis I'll say yeah

26:05 as you can see like if you changes the wall angle Z in in plane Z

26:09 also changes right the robot actually moves out

26:11 of the plane and in plane please explain

26:13 that so so the robot is not always in the same plane right as as you're

26:18 going in steeper wall angles it basically

26:20 moves around the forming forming T I see

26:23 so if if you're I so what what do you you call it the wall angle

26:28 yeah so like if if with respect to the with respect to the sheet what

26:32 is the angle of the the forming angle

26:36 it's basically the complement of uh draft angle

26:40 okay right I see that's awesome and you can see at the corners where it sort

26:43 of changes wall angle where the tip sort of goes in and out at the corners

26:47 it changes yeah like right here sort of at the bottom of the part you'll

26:50 see it comes out now and then it'll go back in oh so I see

26:53 so we're moving in the Z yeah like this wall is steeper than this wall

26:58 and now it gets steeper again so the tip sort of goes in and out there

27:01 to uh basically keep the tips perpendicular

27:04 to the surface that you're forming ah I see

27:09 so even though it looks like it's a 2d operation on any one slice it's actually

27:13 a 3D operation the whole way around so it's more not a helix but it's

27:17 a uh I don't know I forget the fusion 360 term for that but does that have

27:23 a turn I don't know I don't know might be this one is spiral they're

27:27 planer slice but the tips sort of move like this sort of depending on the wall

27:30 angle that you're forming oh really and we have some fancy non planer slices too

27:34 we want to talk about show you some the parts through that are non planer so

27:37 like you're forming on a curved section yeah

27:41 so now you're not starting from a flat

27:43 sheet you're forming from a section that's

27:44 already have forming in it now your robot

27:47 is actually going into a non-plan or Surface to deform and make a uh another

27:51 feature out of it oh that's awesome sweet so this is what's going to happen yeah

27:56 we're doing the face and then this triangle is the one that's just started over

27:59 there right now this whole thing is about

28:01 90 minutes um then it'll do this hexagon

28:03 part this one's actually um a spiral rather than layers mhm so there's no um

28:09 I feel it's a 1 mm pitch but it's just a spiral the whole way yeah

28:14 um and then all of these are actually chained together as a single operation so

28:18 the robots will do one and they'll immediately

28:20 go over and start the next one without slowing

28:21 down or that's interesting cuz when I looked at it on the internet I thought

28:24 you were going to do one plane and then just move in the z-axis and do

28:28 more and more planes but you don't have to do that we can do whatever we

28:32 can imagine yeah is it faster to do it you know one feature at a time

28:37 uh you can sort of hide the seam if there's no links like right here

28:40 there's a bit of a seam if you look really closely where we transition but if

28:43 I do a spiral then you won't find that on this part you transition meaning

28:47 you move up um like this one is just planer slices so you sort of do

28:52 one Loop and then you step in and you do the next Loop oh I

28:55 see um the other one there is pathed as a spiral so it's just a continuous

29:00 um sort of going a little bit deeper the whole is it also faster um

29:04 I guess a little bit cuz you don't have to turn around uh-huh um so yeah

29:07 that one it just kind of goes the whole way that's awesome yeah you're right

29:11 the face is cool yeah I like the face oh you might want some paper towels now

29:16 it's all good it's not uh dangerous

29:18 or anything it's just I get really greasy around

29:20 here too we're good Ed you don't have to worry about it it's just right here

29:25 okay I will uh thanks sorry about that yeah of course um this is actually

29:35 an interesting thing so like you know I told you like the robots as we were

29:38 forming they apply a lot of force and that sometimes caus them to deflect so

29:43 there's a period where we actually characterize

29:45 the amount of deflection the robot gets in each

29:47 joint and that's a Tracker right measures

29:51 the deflection and that allows us to model

29:53 the deflection and it count to what what

29:55 accuracy um we're constantly improving it I think

29:58 the total accuracy depends on the parts really

30:01 because it comes down to also the spring

30:03 back of the sheet uh so there's like coupled with the material uh so what we

30:07 are measuring we haven't I don't think we

30:09 have measured recently what how accurate the robot

30:11 itself can be I think the repeatability of the robot is2 mm so that's the that's

30:15 where we want to get to um but you know combined with the sheet depending

30:20 on the geometry we can go be as good as sometimes plusus one in a very

30:24 large part uh millimeters right you just it just blew my mind a little bit

30:30 there's a lot of math here so can is there a robot around this corner there's

30:34 a robot but it's not running at the moment but you to it okay so I'm

30:38 going to leave this here so leave this here can I walk there yeah of course

30:44 it's a it's a off St so if I understand what you just said like

30:48 there's many ways to arrive at this point with this linkage yes okay yes and if

30:56 I arrive at this point and a apply 5,000 lb in that direction right I'm

31:02 going to get a certain radial deflection here

31:04 yes I'm going to get a certain radial

31:05 deflection here that joint and there so what you just said is your code is

31:12 uh I don't know what the word is I I does the word homic systems mean

31:17 anything to you yes yes okay you have to also determine how you get there

31:23 in order to anticipate the deflection yes exactly

31:27 so so so I think what you're trying

31:29 to say is that then there's two types of accuracy like get the robot to one

31:33 point but then also accurately move from one

31:35 point to another while the forces are

31:37 changing on the robot is that what we were trying to get yes that's well well

31:42 no I wasn't thinking that far ahead I was just thinking the first thing so so

31:46 it's an incredibly complicated problem right yeah

31:49 it's I think it's basically bringing in Dynamic

31:51 load into equation of kinematics right so

31:56 with the load now there's an extra are like

31:58 okay The Joint A1 not only need to be at 60° they need to be

32:01 at 60.1 or 60.2 de depending if at the end theor is facing facing like a 5,000

32:08 lb horse right and including

32:10 that in the into the kinematic calculation that's hard yeah

32:14 it's it's fun it's a fun thing that our robotic team is foring up am I

32:18 correct in saying uh I'm not sure that this is true but am I correct

32:21 in saying let's let's say I'm determining because

32:24 of uh what I'm going to do this has

32:27 to be at Z is this 0 de or 180 what is this uh the tip

32:31 right now you can say 0 de like 0° so if this is at 0

32:35 degrees and I'm pushing into that are there

32:38 many different ways to set up the robot

32:40 to get there or is there only one way it can be set up no

32:42 so because we have seven axes there is

32:45 probably there's unlimited ways to to to get

32:48 to that point uh with different poses right so what do you do do you

32:52 pick a few things and say this always has to be here this always has

32:54 to be here no so there there's so

32:57 so many cool things you can just choose randomly

32:59 choose or you can optimize for example

33:01 for stiffness in certain direction that your force is

33:03 going to be so you say Okay move the robot and a rail so the joints

33:07 are the most stiffest combination to apply

33:10 the most amount of force without deflection um

33:13 so there's like you you have a you have a function to optimize at that point

33:16 right and you can kind of part of your creatics you can figure out what

33:18 EXA you want to do you're Wicked smart dude no I'm not yeah you guys

33:24 are you guys are awesome so was this an earlier version of the uh

33:27 this was actually our first this is as first sell I think like there are parts

33:31 of this our whole our team welded this together um we were kind of just doing

33:36 back of the envelope calculations and uh

33:39 this was the literally the first sale this is

33:41 the first sale it was funny because when we started the sell um we didn't have

33:45 a lot of money so we were like we went to robot manufacturers like oh let's

33:49 sell this 6 months lead time and it's going to cost you you know hundreds

33:53 th000 so we end up buying these robots

33:56 from automotive manufacturer these are the robots they

33:58 want to throw away they're like reached the end of their lives we don't need

34:02 it anymore so we bought them very cheaply

34:04 brought them in this facility they have been

34:06 since then they have been putting too much two twice the hours that they already

34:09 had they're still working which is good

34:12 it's actually yeah this robot can actually move

34:14 on a track there's morean roves on a track uh yeah it's is full seven aess

34:17 SP of them on both sides I'm in I'm in love with this place man those holes

34:23 on the top those are like early

34:25 on we're experimenting and we were like different

34:28 part sizes Etc so we didn't know what is it like a gen generalized sort

34:33 of a size that we can use so it was like you have a like

34:36 a magnetic drill drill holes oh next part's

34:39 coming in it's smaller let's drill a hole now

34:42 now that we have like a much better system so you see yeah so so

34:47 you love seeing this this this there a lot of nostal a Swiss cheese now yeah

34:52 over time what do you feel when you see this cell versus the new ones I

34:56 know I think it's going to come a long way um for sure um those are

35:00 days I mean we like we were like literally sitting on the ground putting holes

35:03 in the concrete to put things on the ground um yeah so but it's kind

35:07 of the good thing is this cell is actually one of our most reliable cell even

35:10 to this day really yeah it just went through so many trials that like it just

35:14 works the smoothest you just know how you know all the things so and this is

35:18 the what are the names of these robots you ended up calling these this one

35:22 is Stumpy and that one Jeffrey because

35:24 this Stumpy is a little bit shorter than Jeffrey

35:28 what is uh can you tell me about the IND Defector here this is earlier version

35:32 right yeah so can I stand here yeah of course um so the goal is

35:36 you know we really what we want to do is even though we are doing sheet

35:39 forming really our long-term goal is to build

35:42 what we call like a robotic Craftsman Like

35:44 A system that works like a Craftsman like you know you can pick up a forming

35:47 tool form it drop the forming tool pick up a scanner scan it drop the scanner

35:53 pick up a trimming tool you can see there's a spindle there so you can

35:56 pick up a spindle this one yeah and then that one too I don't understand

36:00 it the last one oh down here at the bottom yes both of them are spin

36:04 on that one is an an angle is that a uh is that a er20

36:08 that's like a tormach head or something what um yeah so the idea is like yeah

36:14 you can easily change it so you can see a tool changer here that that can

36:18 drop a tool pick up another tool and just move on to the next operation

36:22 right and that's kind of the we're imagining

36:24 these robots like really working like a Craftsman

36:26 that can pick up tools and different

36:27 things okay so you're not you're not pushing

36:30 like oh you're applying electrical contacts and then

36:34 you have the mill here yeah so you can like yeah once you pick up it

36:38 automatically connects through the electrical contacts and now

36:40 it's a the robot now has an ending Milling end Defector and then can drop

36:43 this pick up the forming in the vector go back to forming can you is

36:48 this the the current what do you call this the wrist or um so this is

36:51 a rest of robot but this is called the master tool changer so it's a master

36:55 and that's the slave the basically the picks up this light so you have indexing

36:59 pins to align it and you also align on the plane so you have total indexing

37:06 and then what what's going on here so

37:07 those are actually the the bearings that through

37:10 air pressure they lock into the tool okay so once the once the once it pick

37:14 up the slave this thing basically through air

37:16 pressure come out and then lock into the tool

37:18 so the tool will not fall oh that's amazing and and what is this you

37:21 told me but I I forgot so there are different connection so these are pneumatic

37:24 connections and these are electrical connections got it

37:27 um yeah will you show me the tools

37:29 or is this the best place for that or should we do that somewhere else

37:31 no we can talk about the tools I think uh so you can see right now

37:34 we do mostly forming and trimming so we do mechanical trimming um so you can

37:38 see that tool is an a spindle um uh that allows us to cut the parts

37:44 after we form them and then you

37:46 have different versions of a forming end Defector

37:48 so that one doesn't have the final end Defector that goes into it okay it does

37:52 the forming the one next to it it it does you can see that the end

37:56 Defector uh is right in there you'll look and pick it up let me give

38:01 you a it's really cool don't worry about the don't worry about the grease yeah

38:07 right here um you got it um trying to get the some of the so

38:14 your your tools are they carbide or is that too brittle um so we we tried

38:20 a bunch of material um right now most of our tools are the base of it is

38:25 carbide but then on top of it we have a coating um that coating is

38:29 slightly more complicated we have tried a bunch

38:31 of different things that's proprietary yes so can

38:33 you can B that's a coating that basically allows us to do many parts for many

38:38 parts without uh without basically destroying the part

38:41 or destroying the tool it's a pretty

38:44 simple tool isn't it yes so that that the end effector is pretty simple this is

38:47 one of our designs we have multiple designs this is some of one of the design

38:50 that we're working at them but then the coating getting to that that's a secret

38:55 sauce that that that took a while it took a while like you know 700 p.m.

39:01 running to UPS you know get this out you know and so

39:06 is is the goal that uh I'm just going to ask

39:08 a business question feel free not to answer but because you have

39:12 the secret sauce on the tools do you hope to sell the tools

39:14 is that part of it yeah I mean we we basically

39:17 the tool and all the holder so that's just the end effector

39:20 of the tool there's all the all the other parts that goes

39:22 into the tool that gives it some degrees of freedom um yes so

39:26 as part of the business we also provide the end Defector

39:30 um and the perishables and the perishables and the consumables yes um

39:34 but yeah I think really I think even though that's a secret

39:36 sauce it really also comes down to the software pieces that we put

39:40 together you know without the software if you get these T robots

39:42 to do some first heuristic path like a machining Milling path the part's

39:46 going to be inches off for forgive me I don't know what

39:48 the in in this application what is the word heuristic mean like

39:52 for example if you you say get the robots and run

39:54 them in just you know get the geometry slice it in uh

39:59 water lines and start forming it the final part is going

40:03 to be inches off of what you actually want really so you have

40:07 to really account for a lot of different things we talked

40:09 a little bit about deflection of the sheet spring back of the geometry

40:14 itself um and then the fact that it can also tear so

40:17 if you don't have put right amount of compression force between the tips

40:22 you might tear the sheet right so we get to the Cracker

40:25 limit and we tear the sheet so um combination all those things

40:30 are counted in our software to get you a part that's going

40:32 to be you know uh very close to the to your final

40:35 geometry and not like inches off um so that's I think where

40:39 the real Secret Sauce is the software piece um but allistic means

40:44 like you take everything into account is that what that means no

40:46 what I meant when I said heuristic is that like if somebody buys

40:49 a robot and says okay I'm going to do a Milling path

40:52 based on my heris stics it's probably not going to give you

40:55 a right path if you just like assume that like well I'm

40:57 going to do the the most simplest thing that I can do

41:00 the part is going to be inches off you have to actually

41:02 do a very non-intuitive path I see to get to the right

41:06 part that's what I that's awesome bobc do you do finite element

41:09 analysis on this stuff at the the pinch point between the IND defectors

41:13 we we started doing those back in the at the beginning uh

41:17 we used like the you know off the shelf tools like Alisa

41:21 those type of things generally speaking because you have to track

41:25 the the history of your forming and you cannot apply the usual you

41:30 know boundary condition simplification symmetry

41:33 those type of thing your run times

41:36 are going to become very kind of long and the computationally becomes

41:40 very expensive to to run those calculations so we're doing work on you

41:46 know accelerated ways of doing

41:48 the calculations developing you know circuits models

41:53 on top of the finite element package to be able to actually

41:57 have fast running tools that we can actually run that's awesome nin

42:02 thing about our process is as you're forming you're going to gather

42:05 all the data you need uh during the process right you can

42:09 use those data to not necessarily do finite element models right just

42:16 build an empirical model right instead of doing a finite element model

42:20 and that would not be accurate and it's going to take competen

42:22 time just run AP part capture the data and then over time once

42:25 you do thousands and thousands of Parts you can just use

42:27 the data to have a high fidelity empirical model it's not based

42:31 on physics laws it's not a fin out element model but it's

42:34 probably predicting it much better than final it's based on experience yes it's

42:38 like that's what no I was going to say that's the Craftsman

42:41 that that's kind of what Ed mention about the Craftsman like back

42:45 in the day they'll like okay work as an apprentice learn how

42:49 hard do they need to hit this thing get that learning and then

42:53 actually by the time they're they're there they they know okay I

42:57 need to hit this steel this hard this aluminum not as hard

43:00 there's that that's the real sort of approach and the strength because

43:04 all that data is being collected that's rad can can we go

43:08 back over to the force gauges sure is that okay yep hey

43:12 real quick just a little break here to say thank you to everybody

43:15 that supports smarter everyday on patreon patreon is the single biggest thing

43:21 that supports smarter every day and and just makes it where everything

43:24 levels out like I don't have to worry about month to month

43:27 like do we have a sponsor how's it going smarter every day

43:30 is supported by patrons in a huge huge way and uh I'm

43:34 grateful for that stability that you bring to my life and it's

43:37 not for everybody um but if you are interested in that you

43:40 can check it out at patreon.com smarter everyday if you'd like to chip

43:45 in on what smarter everyday is and the videos we kind

43:48 of make if you feel like this brings value to your life then

43:50 I'd be grateful if you'd consider that if not no big deal

43:53 I'm just excited to be exploring Robo form room with you here

43:56 this is amazing isn't it I think it's really really cool the software

43:59 has to be incredible so let's get back to talking with Ed

44:03 and bobac and let's see what they have to say you're you're

44:06 looking at the force diagr right is it is yes is it

44:08 running there's another part yeah we're doing a hood over there I

44:15 also like just highlight that you know if you don't have like

44:20 the the like we have an on-site machine shop that like

44:26 if you don't have that sort of agility in adopting and everything

44:30 in terms of getting stuff done like uh you wouldn't be able

44:34 to move this fast so like having that integration for us is has

44:38 been very key so it's

44:39 a partnership between robotic manufacturing and traditional

44:41 manufacturing is that is that just like like tools that we need

44:44 like if we try something doesn't work we go to our machine

44:48 shop here we're like hey let's change this angle and do that do

44:52 it really quickly and rapidly as opposed to you know having

44:56 to wait to do it so having that vertical integration within the company

45:01 has been CH is is cute super fast fast all right so

45:06 this is another Force Mo Force what do you call this Force

45:08 graph yeah so we're looking at the forces the total force

45:11 that the forming robot and the support robot the forming robot is the robot

45:15 that's pushing in the direction of defamation the support robot is

45:18 the second robot so R2 uh R1 is a forming robot R2 is

45:23 a support robot you can see that one more time please I

45:25 wasn't focused can you point again so R1 you can see it's

45:28 a forming robot and R2 is a support robot uh forming robot

45:32 is the one that's pushing in the direction of the defination and support

45:36 is trying to localize the defination just to that area that you're

45:40 trying to deform so the the full sheet doesn't globally deflect so

45:44 maybe I maybe I didn't understand earlier with Mark when I thought

45:48 I did so it's always going to be positive deformation or positive

45:51 force right on both robots well it depends on the on the on the frame

45:57 of the if you're talking about the sheet frame

45:59 one robot is forming in this direction another one applying force

46:02 in this direction I see so for basically the forces from the robot

46:08 should basically match the forces from the other robot plus the forces

46:12 that goes into the clamps right so the one robot is always

46:16 applying slightly less Force than the other so it's a it's

46:20 a complicated free body diagram because

46:22 you have planer deformation involved and not

46:25 only do you have planer deformation involved you have planer deformation

46:29 with a weird looking shape right and so that's why it's so hard

46:33 yes and that's why we also rely on empirical modeling as opposed

46:36 to try to really try to understand the physics to a very

46:40 small detail we we we understand it as much as we can

46:44 but at the end of the day it comes down form the part

46:45 capture the data and empirically model what's happening as OPP po who

46:49 try to really wrap your head around all the details that's going

46:52 on because between the tips I would say all three mechanisms

46:55 of defamation is happening you you have a little bit of Shear you

46:59 have a little bit of stretching you have you have all three

47:01 right so the three being um yeah sheer uh like the tension

47:08 compression tension compression is shear tension compression sheer okay can I

47:12 see what this one's doing real quick and then I'll come back

47:14 and see if I understand this better what what is the thickness

47:22 of this um so this is a 2mm aluminum sheet uh we are

47:28 forming a custom hood for a custom truck custom hood for a custom

47:35 truck whose truck is it your truck well you don't have

47:39 to say this is a secret project that's going to come up

47:43 so okay got it all is it uh am I is it

47:46 okay to show that we're doing it yeah you can show you

47:48 can show the track but I just likeing you you will uh

47:50 it's a full vehicle project that's going to that's going to be

47:53 come out at some point okay Ed question this robot's stationary this robot

48:00 has an additional axis the answer is embarrassing so this is

48:06 our second manufacturing cell why is that embarrassing well I'll tell you so

48:09 this is our second manufacturing cell uh we still didn't have enough money

48:12 to buy new robots so we bought them used and at the time

48:16 we couldn't find a rail in the market so we just

48:19 welded that and put it on a pedestal that's why it's called

48:22 Princess because it's on a pedestal princess is on a pedestal all

48:27 right you guys are you guys are likable that is massive that's

48:34 massive so this this I don't know if You' seen Titans um

48:38 this robot's called Titan uh the model of it and they can

48:43 they can apply as high as 20,000 Newtons of force really yeah

48:46 so sometimes when we do very thick sheets we're doing like 3

48:49 or 4 mm steel uh um it can be it can basically

48:53 applies up to 20,000 Newtons to deform the that's crazy so I

48:58 noticed it's going a lot slower on this thicker metal this it's

49:03 going slower because this is on an older version of the software okay

49:07 it's the same thickness as that one that you saw oh is

49:10 it it's the same thickness but in this specific she these robots

49:13 can do much thicker but uh it's uh this one is

49:17 the same thickness it's uh we are on wi1 that one is

49:20 on V2 so the controls have improved on that one so you

49:23 can go faster that's amazing that you've that you've upgraded that much

49:28 cuz it's fast over there yes it's all about like figuring out

49:32 how fast you can kind of command the joints on their Dynamic load

49:35 without over torquing them and kind of optimizing that get to the point

49:38 where we can have maximum acceleration without over toring the joints

49:43 as the load on the tip is also changing all the time

49:47 can you please say that again it's all about maximizing the acceleration

49:51 you can get at the tip at the end Defector without getting

49:56 an over torque on the motors right so in the earlier versions

50:01 We would be very conservative because we couldn't do that calculation of what

50:04 will be the maximum torque I can apply to this motor while

50:07 I maintain the accuracy so with the new version now we do

50:10 those calculations so we go as fast as we can and apply

50:13 as much as the acceleration that we can at the end

50:15 of vector basically pushing the motors to its limit before it over Tor

50:20 so does that mean that the software is robot specific because

50:24 you know the torque limits on each motor or is it parametric

50:26 and that you can parametric okay yeah the reason I keep flanking you

50:30 like this is because I'm keeping you and the robot in the frame

50:33 I'm not being awkward that's that's what I'm doing no let me

50:36 know I was I feel sometimes I feel like maybe you're asking

50:37 me to get out of the No No I was I just

50:39 figured i' just come out and tell you what I was doing

50:41 cuz it's like I'm just trying to sneak behind you or something

50:44 sorry about that so um it is parametric so you load each

50:49 individual robot and um you have torque limits on each robot

50:54 in the so you have a robot model you have a uh I guess

50:58 we call this the the part the model model has different features

51:02 yes and then do you also have to build in the I

51:05 guess you have to have two robots yes because I've seen

51:08 this done without a second robot where you just have like a rubber

51:13 pad on the bottom and then in that case it actually is

51:17 only z-axis yes is that true yes so in that case you

51:20 have only one robot and you can push it against the pad

51:23 in our experience that doesn't produce

51:26 accurate Parts you cannot properly localize

51:28 the deformation like I said it's really a tricky dance between

51:31 the two robots being able to get into the right angle with respect

51:35 to each other depending on what feature they're trying to form

51:38 and then depending on how thick the sheet is they need to counter

51:42 even with more Force the support robot need to counter even

51:45 with more force uh in a very surgical manner so we have noticed

51:49 that when you try to be like form it against um a uh

51:54 kind of rubber or something the part just doesn't end up being

51:56 very accurate because you always have a fixed force on the other

51:59 side or I mean a Rel is it hooks law applies

52:02 on the other side they're probably have some kind of I think

52:06 it's it's actually probably a little bit complicated because the the the rubber

52:09 is usually probably applying a little bit of compression on all sides

52:13 yeah I've never fully thought about but we realize that it's just

52:16 you can get very accurate what I like about you guys is

52:20 you're like you can do all the math you can do all

52:23 this crazy stuff but in the end it's going to do what

52:26 it's going going to do yes so let's just go rip something

52:28 up and and get that data and then fold it back

52:31 in and then lather rinse repeat do it that that that's like an I

52:36 would give kudos to Ed is really like okay let's do it

52:41 you know you can go and kind of try to figure out

52:45 oh let me figure out the forces let me figure but it's

52:48 so much informative when you do the actual part and kind of fix

52:52 the problems issues and then okay next one and capture those learnings

52:57 yeah I remember like early days we took the whole team

53:00 to a I don't know if you know Bobby Walden he's like

53:02 a very famous sheet shaper so he shap sheep shapes sheep shape sheets

53:08 with uh with hammer right so like like holds all the sheet

53:11 goes on their power hammer hammers it into shap does a lot

53:14 of custom panels for like like hot rods and old cars and I

53:17 remember early days um there's somebody I knew from a long time

53:21 I actually like worked in his shop for a little bit

53:23 I was like okay like we're we're forming this company we had

53:25 the early team like let's get him into his shop and these guys

53:29 Hammer sheet with their hand to really understand what's going on because

53:33 all at the end of the day I think matal does what

53:34 it does and if you think and then there's variances between batches

53:39 of the material uh some have higher strength some have lower strength

53:42 so you got to be adaptable um and a system who can just

53:46 look at what it did and constantly improved that's a system that's

53:49 going to win that's why humans still can do sheet shaping it's

53:51 very tough to was very tough for for for static systems

53:56 to do it other than humans right so you you guys just don't

53:59 you just don't suffer from analysis by paral what am I saying

54:02 that what is it analysis paralysis analysis paralysis yeah you just don't

54:07 do that you're like go hopefully I mean we want to be

54:11 intentional and thoughtful about it as well but I think there's a lot

54:14 of unknown unknowns when you step into physic physical world and you

54:18 start start getting those and understand those as you go through them

54:22 yeah that's great man I I'm I feel stupid cuz I've been

54:25 waiting my whole life to say paralysis by analysis and I boted

54:28 it but it is what it is dude this is awesome man

54:33 how long will it take to make a part like this can

54:36 I go right here yeah just um yeah I think that would

54:38 be good I think that it's not going to you can see how

54:40 much far it back comes back and then I won't uh

54:44 right here yeah I think it should oh yeah you can kind

54:46 of see the line actually I'll just let it come to me

54:50 here so depending on the so the full frame so this is

54:54 a 10t frame um right now this one is very going going

54:58 very slow so this one takes probably 15 hours to finish but we

55:04 have done a full frame with the newer version of the robot

55:07 in less than 5 6 hours so so when you get done

55:12 with this thing you're going to have a fully formed sheet

55:16 this is going to be a hood for a truck yes it'll be

55:19 done how do you get it out of that do you like

55:21 jigsaw it out or so the same robot after it's forming it

55:25 scans it and then Maps the scan to the trimming path so

55:30 it aligns the trimming path to the scan the best way picks

55:33 up a trimming end Defector and trims the part up but once

55:37 you start let's say let's say I start trimming it that's a heavy

55:40 part mhm how do I not drop it does the other

55:43 robot hold it while it so that's maybe something we can do

55:46 in the future for now what we do is we leave few

55:48 tabs in so you leave few tabs around the part and then technicians

55:52 come in and cut the few tabs and take it out

55:54 but we want to make sure the robot at least does the accurate

55:56 cutting um so that the before you remove it before you remove

55:59 it and then you leave few tabs in and then you cut

56:01 those tabs and clean those tabs up dude it's really cool yeah

56:04 do you do you give a lot of uh you give a whole

56:07 lot of Tours don't you um not actually a whole lot really

56:11 yeah I think we have been like really heads down for a while

56:14 and then I think just more recently as we are kind

56:16 of expanding the operation I think we have been you know we have

56:19 been asking people to come and see that we used to do

56:21 we did a most of our customers even our customers when we

56:24 have done parts for them when they come here they're like

56:27 this is much bigger robot than I thought in the video it looked

56:30 like it was pretty small they come here it's like 30t X

56:33 30t sou can you can you hold this yeah can we use

56:36 it I'm just going to show how big it is it's big

56:41 thanks yeah that does the scale uh bobc can you take me

56:46 to the metal like your stock like do you have sheet stock

56:50 somewhere yeah yeah there's some here they're like a distributed all over

56:55 the place but these these are the ones that have been

56:57 of these are the ones that we've cut to blank to the size

57:01 that we want to do for the customers and they're ready to go

57:04 and everything is you know labeled you know so that we can

57:08 have full traceability from the start to the end of the part

57:12 but will you tell me uh the difference is this galvanized this one

57:16 seems to be yes this is the galvanized steel yes so will

57:20 you tell me the uh the difference in working galvanized versus aluminum

57:24 versus stainless I mean are there different things that you have to take

57:28 into account so that the amount of like a pinch that you

57:31 apply the kind of the rate that you form so we've started

57:34 forming for example like in The Other Extreme like titanium and you

57:39 do that at room temperature maybe you have to like go slower

57:44 for those type of metals with the forces that are involved compared

57:48 to like that 2 mm aluminum or the one that you saw

57:50 earlier that's kind of like those type of pro the that's where

57:55 the function of the a function of the metal itself now if

57:59 you have like a galvanized steel where you have to be conscious

58:03 of the the the coating that you have on the metal then

58:06 you have to also choose about think about like what type of end

58:10 effector do you use for that so that you don't damage

58:13 the the coating as much so if you had titanium and you have

58:19 the ability to cut you have a trimming IND Defector you called

58:23 it like back in the day when they were building the sr7

58:26 71 they didn't know how to work titanium yes I saw

58:28 the glimmer in your eye to this day I don't think they can

58:31 easily form they can easily work titanium really yeah I think it's

58:34 a tough just a tough alley to work I mean we figured

58:37 out how to machine it but still forming it is very tough

58:40 unless you elevate temperatures and room temperature is really tough to form

58:43 T but in theory could you use this technology to form it

58:48 and then go in and cut the rivet holes yes and are

58:51 you very excited about that I think that's that those are

58:53 the areas that you're the most excited like enabling alloys were just impossible

58:57 before you know I think for example for Hypersonic you you said

59:00 SR71 but all the Hypersonic applications as you know the focus of attention

59:04 a lot now with some of our customers they have a very

59:07 hard time forming skin of these aircrafts out of uh you know

59:10 high temperature Alloys titanium in canel

59:13 we're talking about refractory Alloys um

59:17 and yeah know we're excited to be able to like just form

59:18 and cut uh in one shot without having to spend like you

59:22 know you know hundreds of millions of dollars in let's say plastic

59:26 super plastic forming uh equipment yeah uh okay so do we have

59:29 a titanium part we can actually show you probably yes so let

59:32 me find yeah let's do it I do have uh I think

59:36 that's a we just want to make sure it's a part that's

59:38 not a customer part yeah that that's the but I I have

59:42 the nickel one but that's a customer part I think we have something

59:45 that might I'm pointing the cameras at the ground so I don't

59:47 see anything can I follow you here it's fine can I uh

59:52 what is this all right this is our assembly room so we

59:54 assemble the frames the electron ICS um the tool Changers so this is

1:00:00 where we assemble all the different pieces that goes into the cell

1:00:02 oh that's awesome so we actually these are the clamps so

1:00:05 these are the clamps yes so we machine them all steel yes

1:00:09 the clamps are steel water jet water jet and I think they were

1:00:13 and then they coated them the anod powder coat them and then

1:00:16 I go some titanium for you is it okay if I look

1:00:19 at this stuff yeah so you can kind of see there are

1:00:21 different end effectors um uh is that a is it a Thrust

1:00:25 bearing no that that one I think is there so that's

1:00:29 a needle bearing for a for one of my ectors if you would

1:00:35 prefer I don't show stuff just no we can we can we

1:00:37 look at it in the end I think I think people this is

1:00:39 going to be outside anyway but you can see this is

1:00:41 a bigger T you can see a needle bearing inside of it um

1:00:43 yep yeah so that's interesting I would have expected a Thrust bearing

1:00:47 there yeah yeah so I think uh maybe we should we should

1:00:49 we should talk to our uh our engineering team if you get

1:00:52 some of your advice that would be great oh no that's that's

1:00:55 interesting and you can kind of see so these are actually so

1:00:58 we know how we talked about we create different um there's different

1:01:02 versions of the tool so we kind of this is for the ballp

1:01:06 ballpoint version this is how we kind of measure and grind

1:01:08 the tool down um so we have our own grinder so we

1:01:11 build everything we said like the tools are all built here oh

1:01:14 man yeah you can make custom carbide tools yes is this that's

1:01:19 awesome so when I saw the video I didn't think you were

1:01:23 actually rolling I thought you were I thought you were hammering yeah

1:01:28 can you do that as well you could actually one of our plans

1:01:31 down the road is um as we are going with the thicker

1:01:33 sheets and higher strength sheets how can we bring the yield

1:01:37 of the sheet down one way is ultrasonic like you maybe can

1:01:41 apply some kinetic energy to bring the yield down um or you can

1:01:45 locally heat it so you can vibrate the sheet as you're working

1:01:48 it is that what you're saying yeah basically just like basically like

1:01:51 same as hammering but like much more much faster and much uh

1:01:55 uh basically get that a kind of ultrasonic um um kind of uh

1:02:00 frequency to to hit the sheet with the end Defector with the end

1:02:02 Defector golly man yeah it allows you allows you to go

1:02:06 through the higher higher um higher strength anded you were one

1:02:09 of the Most Fascinating People I've ever met no is this titanium

1:02:13 this is titani cut out the part from the customer cut out

1:02:16 but you can see kind of the rest of the sheet it's

1:02:18 actually pretty thin sheet super strong how much does a sheet of titanium

1:02:22 like this cost this one is expensive okay upset uh that's incredible

1:02:28 and so why did you why did this is all not part

1:02:32 of the part but this was kind of like the form I

1:02:35 don't know what the word is I think they can we set

1:02:37 it up here please yeah yeah and we kind of we you

1:02:40 know I think the the anal good analogy for folks who

1:02:42 are like familiar with 3D printing support is a support support structure

1:02:46 for a forign part interesting and you do that for what reason so

1:02:52 we do that so that basically we have have um enough we create

1:02:58 enough stiffness around the part that as we are forming it

1:03:01 the part is not buckling got it right okay so cuz I'm

1:03:05 pushing here okay wow that makes sense because if I'm pushing right

1:03:11 here if I'm so steeper angles give more rigid Parts yes ah

1:03:18 and the deeper it is it becomes rigid that no no I'm

1:03:21 like that that explain and the deeper it gets also you are more

1:03:24 rigid because have to overcome a little bit of plast elastic deformation

1:03:28 before you get to plastic yeah so like if you want to do

1:03:31 a very shallow part it just doesn't work you're just going

1:03:33 to you're just going to push the part in and out once you

1:03:35 go further deep then you create more rigidity you you can easily

1:03:40 overcome the initial elasticity and get to plastic regim faster right my brain's

1:03:46 on fire so can I can I try to say what

1:03:49 you said but I don't have the words you do no go

1:03:52 for it so like if if if I have a a sheet

1:03:54 like this and I put right here it's going to deflect yes because

1:03:59 of like the co what what is it cosine or tangent yes

1:04:03 is tangent right so it's going to deflect easier you have more

1:04:07 planer deformation but if I can rock that up the more I

1:04:09 go like this I'm going to be in straight up compression yes

1:04:14 but I can't do that cuz I'll Buckle the sheet yes so

1:04:17 I have to find an optimized position that's best for the part

1:04:21 yes and also you're probably doing stuff about tool path too yes

1:04:25 yes is that how you do it yep now you can start seeing

1:04:30 how this whole design of space around the part is also affects

1:04:35 the accuracy of the part right so you can start thinking about

1:04:38 how you can generatively come up with this design space using the empirical

1:04:43 model empirical model to come up with something that's going to give

1:04:46 you the best results right you want certain stiffness in certain direction

1:04:51 because the parts have its own the stiffness kind of properties so

1:04:55 the idea is like once you have all the data you can

1:04:57 combine all these things to come up exactly automatically suggests to you

1:05:00 what the what the what the scirt needs to be like have

1:05:03 you taken uh this is a dumb question but I I don't

1:05:08 really understand what it does to the material cuz the material has

1:05:11 to flow yep maybe flow is not the word but it has

1:05:13 to move flow is the right word flow is the right word

1:05:17 so if if I'm making have have you taken a plane and then

1:05:22 done a a a sweep like that and then did dog bones

1:05:26 along the way to see what it does to yes I think

1:05:30 that's work that have been doing with some of are you allowed

1:05:33 to talk about that just general terms like we were benchmarking some

1:05:38 not titanium necessar not yet but benchmarking against hydroforming um some some

1:05:44 Aerospace grade aluminums and we did form the same geometry using our process

1:05:52 our Robo formed Ser geometry and a hydrofor process and we took

1:05:57 dog bones with an independent sort of a tension test on each

1:06:02 spot tension sp sp hardness um like you said it was like

1:06:06 different wall angles different directions so you can kind of characterize how

1:06:10 much work hardening you put in into each uh depending on the direction

1:06:14 and the wall angle like you said like you know you

1:06:15 can create a kind of almost a uh what I call

1:06:19 that is that a as a geometrical name now I can I where

1:06:24 like it's constantly increasing yesal a name for I I anyway

1:06:31 but yeah getting different wall angles so you can see how much work

1:06:33 Harding you're getting based on your wall that's amazing and so uh

1:06:37 did you get any information out of it that you're willing to talk

1:06:40 about yeah yeah no it's um it's funny like early on we

1:06:43 were asked you know customers were kind of had experience with 3D

1:06:48 printing know 3D printing if you asked like if you had 3D

1:06:52 printed this part and you had asked like like hey what what's

1:06:57 the properties there I'm like I don't know where where was it

1:07:00 printed which location in the chamber all that stuff with this process

1:07:04 you know you're working the metal you're working the metal be more

1:07:08 than this so the strength is going to be higher and in metals

1:07:13 like you generally increase the strength your elongation ductility comes down so

1:07:18 the the general outcome was with our process you work the metal

1:07:23 more so the strength is higher at the end of forming let's

1:07:26 say compared to hydroforming or stamping but then your elongation is going

1:07:30 to be lower from them and that that was generally for the aluminums

1:07:34 were like you're 20 to 20 20 to 30% stronger using

1:07:40 our process to make the part obviously your elongation and the amount

1:07:44 of energy that you can absorb with the system might might be

1:07:48 lower okay so please explain hydroforming because I was shaking my head

1:07:53 but I'm not a th% sure I understand hydr forming so is

1:07:57 that a competitive process it is a yes yes and no um

1:08:03 it is a process where you do need a tool or a dye

1:08:07 to make the parts so our process you don't need any you

1:08:11 just have those in so you do need that tool whether you

1:08:15 have a bladder on top of it or not so you you

1:08:18 would have the fluid on the top with a bladder you put

1:08:21 the metal there you press and you kind of conform to the shape

1:08:25 of the the the D that to think of it stamping

1:08:28 except one side is fluid fluid pressure okay right so you're pushing

1:08:32 against the D with fluid I see um so it's a old

1:08:35 process I think it's a process that like people try to use

1:08:37 when they want to get rid of at least one D I

1:08:39 see and sometimes it gives you it allows you to do a deeper

1:08:42 draw uh depending on depending on the alloy uh and the geometry

1:08:47 um but in that in this Cas specific case the customer wanted

1:08:50 to replace our process with that so we kind of did some

1:08:54 of the tensil templ sample tests uh uh benchmarked against that I

1:08:58 see yeah so with hydroforming yes you have a die on one

1:09:02 side just like you said and then you you have that bladder

1:09:05 or I've even heard of explosive hydroforming um but you still have

1:09:09 to machine that dieye yes and what you're saying is you don't

1:09:13 have to do any of that it's all software load yes so

1:09:16 as you develop a faster routine with your software tool path you

1:09:21 have the ability to get to prototype faster is that true yes

1:09:25 so I mean you get your first part hours after your design

1:09:27 is done here with Hydro foring you still have to go through

1:09:30 die design like your part is one thing now you have

1:09:33 to design your die manufacture your die

1:09:35 go through multiple iterations because every

1:09:37 time your die is basically not giving you the right part now

1:09:40 go machine another die versus with us is just change the software parameters

1:09:43 and go again um so usually get like you know I even

1:09:47 with Hydro forming I think the fastest customers we have had we

1:09:49 worked with SpaceX for example or like one of the fastest out

1:09:52 there like Tred to really push the vendors to get to the parts

1:09:54 fast like two to three months at least before you get

1:09:57 into your parts and you're doing it in a day yeah in a day

1:09:59 you can be done I mean there was like parts that we

1:10:01 got from customers they send us an order at like 2:00 p.m.

1:10:05 and then 10:30 the part started run 10:30

1:10:07 at night the part start running the next day

1:10:09 we have the part yeah that that's something

1:10:11 that's something to be proud of um okay so so if I'm looking at this sheet right

1:10:15 here is the material the thickest it will ever

1:10:18 be right here in the plane and anytime I

1:10:21 work I'm having to move material so it's going

1:10:23 to be thinner yes here is that true yes so I think the good law you want

1:10:27 to apply is that kind of conservation of volume

1:10:30 right so as you're going on a higher wall angle the cosine of your wall angle

1:10:36 or S of your draft angle times the original

1:10:39 thickness roughly gives you the thickness you're going

1:10:40 to get in the part can you say that slower please so cosine of the wall angle

1:10:45 or the sign of the draft angle times the original

1:10:48 thickness of the sheet will give you the thickness

1:10:51 you should nominally get okay in that sitation

1:10:53 obviously depends on the material properties and how

1:10:55 it flows but roughly that gets you there

1:10:57 but this is only if you almost form the part in water lines so I can show you

1:11:02 some other parts if you form the part in stages that's what I was about to say

1:11:06 then you can play so if you were to push if I were to start right here

1:11:09 and I were just to push up yes then I would have parts of it that are

1:11:15 that would still be thick there and basically if

1:11:19 you go deep first you could make a thin

1:11:22 wall angle yes and and then you could have thicker uh features at the bottom is

1:11:27 that right yes actually let's go I'll show you on the on a I can show you

1:11:31 on the um on a whiteboard okay so so

1:11:36 you can imagine so okay so you originally have

1:11:38 a sheet right and you have certain thickness let's

1:11:41 say we call it t Okay um and then you're forming the part so that same part

1:11:47 now it has something like this right yeah so

1:11:55 now now you have this Alpha angle angle this was T and this is zc T Prime

1:12:01 mhm you can kind of if you do

1:12:03 simple trigonometry T Prime is cosine Alpha uh time

1:12:09 T okay right to your point it in reality will end up happening is this is T

1:12:15 and this is not this gradually happen so just one second let me let me look

1:12:19 at this so T Prime is this or this T

1:12:24 Prime is this thickness T is this thickness okay

1:12:28 got it yep right and and this is

1:12:31 obviously Alpha and that's if you start working it

1:12:34 from here and you're drawing it as you go it out yeah but if I were to start

1:12:37 here and push there then I would expect

1:12:39 this to remain thick so in reality even when

1:12:42 you do this you usually have a little bit of in reality you can I cut do

1:12:47 your thing so you in reality actually looks like

1:12:50 this right and then you get thin and then

1:12:53 you get thicker again at the end and this is going to become your T Prime so

1:12:57 you have a little bit of thick and thick on both sides yeah but so what I

1:13:03 was going to say but again the the just

1:13:05 law generally applies but in in instead of forming

1:13:09 if I form this this way that's what I'm going to get if I form it let's

1:13:15 say I go and form first I form in order to form that wall ball first I

1:13:20 form something like this and this is let's say Alpha and then I formed this wall

1:13:29 into this push this out and this ises so

1:13:35 you would bring the IND Defector normal to the wall and then push P it out so it

1:13:39 becomes a threedimensional brake it comes from break

1:13:42 or even like even like if I like water

1:13:44 line this out not necessarily just push in water

1:13:46 line this form it out right like start

1:13:49 from here and assume this is now my second

1:13:51 forming surface and start forming that and this is

1:13:54 becomes base then the this thickness in the end

1:13:59 is T will become T initial T time cosine Alpha cosine beta which means that you

1:14:10 can actually control the thinning depending on what

1:14:13 stages you're going to have in your forming process

1:14:15 it's it's a that's that's the first Bend

1:14:18 and that's the second Bend second B so you

1:14:19 have yes and then cosine Alpha cosine beta

1:14:22 is usually smaller than cosine Alpha plus beta so

1:14:27 now you you control thinning right you're now

1:14:29 actually thinning the material less but it depends

1:14:31 on the order that you get you got there

1:14:33 yes yes that's fascinating so so a lot we call

1:14:36 this actually restriking restriking yeah you strike the part

1:14:40 restrike it then to another shape and now

1:14:42 you're Distributing thickness a little bit better so

1:14:45 the part doesn't thin if you're doing a very

1:14:48 high wall angle so why would you want

1:14:50 to do that why why would you want a thicker

1:14:53 wall thickness um yeah what you see what I'm trying to say if you have a high

1:14:58 angle the the wall thickness probably doesn't have

1:15:01 to be incredibly the problem is so what is

1:15:04 cosine of 90 one uh cosine of0 1 cosine 90 Z 90 Z so that means that if

1:15:10 I push this to a 90° wall angle I'm going to have zero this is going

1:15:14 to tear ah I see right so so in order to be able to do 90 and overhangs

1:15:22 then you have to go in multiple stages

1:15:25 to to distribute the thickness a little bit better

1:15:29 so you're managing it's also going to determine whether

1:15:33 it's sheer or whether it's uh tension as well

1:15:35 yes yes and so you have to look at more Circle yes mhm more Circle 2D

1:15:41 or 3D more Circle this is 3D so could you do another thing could you do could I

1:15:49 restrike make a plane here and then instead

1:15:53 of having a tool right here could you ever

1:15:56 have a situation where you bring a whole

1:15:58 IND Defector over that's a like a plate yes

1:16:01 or like a or like a line and then you could actually truly do a 3D break yes

1:16:07 can you do things like that so you're actually

1:16:09 touching up on the right thing that's that's why

1:16:11 we're talking about the PHT and uh in mechanical

1:16:14 engineering so to do a proper Bend you

1:16:16 need three points you only have two points

1:16:19 so yes if you really want to constraint it

1:16:22 then you need a line or something that gives

1:16:24 you three points um if you haven't done

1:16:26 that I think our goal is really stick with the idea of like you only have two

1:16:30 points figure out the math figure out

1:16:32 the model that gets you the right thing because

1:16:35 the least amount of tools that we have

1:16:38 the more flexible the system is right if I can

1:16:40 get away from not having some custom IND

1:16:42 Defector on the other end that's going to be

1:16:44 a line or a plate then the more easy it is for me to be more flexible right

1:16:50 if you have like a tool tray of like 20 30 different tools it becomes a little

1:16:53 bit more complicated and economically starts to not

1:16:56 make sense sometimes got it but in like

1:16:58 the the far distant future I could see where

1:17:01 you would have a rivet gun somewhere on an infector

1:17:04 or you would have a way to put in Pim nuts and and things like that are

1:17:08 you working towards that or that's we're not quite there yet we do all we can do

1:17:11 different processes we already do forming and trimming

1:17:13 and scanning we can think about surface finishing we

1:17:16 think about riveting we can even do additive

1:17:19 like add a welder to it do some additive features on top of that um we started

1:17:23 from sheet metal because of I think sheal is just

1:17:25 like the largest metal processing sector today right

1:17:28 so there's a lot of need but yes no

1:17:30 we can do other processes in the end as well but as hopefully we can have tools

1:17:34 that are not very geometry specific as much

1:17:38 as we can right that's awesome so what's special

1:17:41 about how you do the titanium the titanium

1:17:44 generally to form it you do need to go

1:17:46 to elevated temperatures because titanium doesn't like

1:17:50 to be shaped the way we're doing it because

1:17:53 of the stresses that you have imposed on the metal

1:17:55 you can actually do it on at room temperature

1:17:57 so and that's a big deal it's a big deal cuz then if you go to elevated

1:18:02 temperatures you need special molds that can work

1:18:05 at high temperatures you need a lot of energy intensive

1:18:09 sort of systems all that stuff so it's

1:18:12 process optimization it's process optimization but again enabling something

1:18:18 new not like now you don't have to you can change your design but you don't have

1:18:22 to worry about like a expensive temperature dependant Services

1:18:27 that's amazing awesome and thanks Bob no thank you

1:18:30 I appreciate it so Mark has figured out how

1:18:31 to get uh my GoPro clamp on this thing

1:18:35 so we can see sneak through here if you want yeah how did you how did

1:18:38 you do I'm going put it like right on here like this okay so you can see

1:18:41 the tip so it can ride along yeah and so this would be the convex side uh

1:18:47 no concave excuse me all right so go ahead

1:18:49 and can you start the GoPro so marks figure

1:18:54 out how to uh get this on there okay great can we see the tip I think

1:18:58 so yeah okay great cool all right we'll

1:19:01 back up we'll see what that looks like all

1:19:06 right should I resume it uh let me yep H hold on I might have to like

1:19:19 manually restart this thing this is actually cool

1:19:22 you'll get to see the whole thing start sort

1:19:24 of from the beginning and he's doing another startup

1:19:31 routine real quick so does it have to you

1:19:34 have to calibrate it when it starts I'm

1:19:35 going to manually retract it and sort of bring

1:19:37 it back into the home position and then we'll

1:19:39 start the next one I'm just kind of backing

1:19:58 out of this part here so you're controlling

1:20:02 is that the Z AIS that's right yeah I'm just going in positive Z right now I'm

1:20:07 just manually driving this thing and manually I usually

1:20:12 like to go nice and slow it's just so I don't you know crash it or anything

1:20:17 I'm going to go this way a little bit this is the positive X Direction so you

1:20:21 can see it's parallel to the sheet yep might

1:20:23 retract a little bit more in Z this is

1:20:26 in Z yep um now I'm going to switch it into automatic mode I'm going to send

1:20:31 it to the home position it's going to sort of twist and go up here I can

1:20:34 stop it at any moment don't be too

1:20:36 nervous okay all right and so you're just observing

1:20:42 so you don't crash yeah I'm just watching it

1:20:46 I know what it's going to do but just to be safe safety yeah yeah yeah all

1:20:50 right so this here sort of the home position

1:20:55 all right that one's set up now I have to do the other one real quick yep

1:21:05 and that's like standard industrial robot stuff you're

1:21:07 doing right now uh yeah there's generally sort

1:21:12 of three different ways that I can operate

1:21:15 this thing I can jog it around manually like I'm

1:21:17 doing right now I'm going to back up so I can sort of see it um there's

1:21:26 how long did it take you to learn industrial robots I don't know how to learn I

1:21:32 think you know I started working in the about

1:21:35 the insur SpaceX and then that relativity I

1:21:38 think they're they're pretty intuitive I think once you

1:21:41 start asking it to do things that are outside

1:21:43 of what they're supposed to do that's going

1:21:45 to get tough right you know like for us to sync these two robots to the point

1:21:50 where they are actually accurate um that we cannot

1:21:54 really rely on the original software stack that comes

1:21:56 with it that's when it gets a little

1:21:57 bit trick but the actual robots they have done

1:21:59 a very good job I mean it's very easy good user interface yeah all right are you

1:22:04 ready for some robots to start moving yes sir

1:22:10 all right first they're going to go into sort

1:22:11 of a start position then I'm going to start

1:22:17 the path so this is going to the you're going to the next part that's right yeah

1:22:29 I'm starting on the first layer of the next

1:22:31 part so it'll sort of hover over the spot

1:22:35 and then come in and then start moving there's a lot floing the she too yeah

1:22:55 it's all right Ed what do you call

1:23:08 those dimples that I'm hearing that noise what is

1:23:13 that that's the plate moving that's the sheet

1:23:15 yeah buckling in and out because the boundary conditions

1:23:17 are loose allow that part so those are

1:23:19 the things you need to account for when you're adjusting

1:23:23 for the accuracy I see right so when you form the part you can when you scan

1:23:28 it you figure out okay these are the areas

1:23:29 that we're off these are the areas undershoot

1:23:31 these are the areas we overshoot and then

1:23:33 you adjust for it for the path and then

1:23:35 it counts for this um kind of loose movement

1:23:38 in the shoot that's awesome once you have uh

1:23:47 to your point once you actually have a little bit of depth gives it a little bit

1:23:50 of stiffness then it becomes much more stiffer

1:23:53 and doesn't move like this oh wow that's why

1:23:55 a lot of parts have a little bit of skirt as in the beginning there's a little

1:23:58 bit of portion that like you form into gives

1:24:01 it some stiffness now you have like what we

1:24:03 call a bathtub to actually form the part

1:24:05 in right so you're localizing the rigidity by making

1:24:09 a like kind of like an ie beam locally yes like a little little kind of like

1:24:14 uh yeah basically an ie beam yeah yeah

1:24:19 so it's not popping anymore yeah so he created

1:24:22 a little bit of stiffness there so no pop

1:24:27 still moves a little bit but yeah and that's

1:24:40 that takes pretty long time for that that feature

1:24:42 to be printed right that whole thing takes

1:24:45 90 minutes yeah got it and you guys are going to leave and leave all this right

1:24:51 oh yeah we typically leave things running overnight

1:24:53 all the time so you you run lights out

1:24:55 yeah yeah it's awesome sweet um you know what we could do so I can just leave

1:25:04 that GoPro here we ship it to yeah

1:25:30 [Applause] satellite half domes half domes yeah hemispheres yes

1:25:42 so what is the uh what is the surface

1:25:45 not what is the thickness across this actually

1:25:47 gets pretty thin so I think we started 2 mm uh in here you're probably 4 before

1:25:54 yeah pretty oh yeah so on the on the highest

1:25:57 angle you get get pretty pretty you see

1:25:59 this one actually fail so oh I see how did you catch that I think they ctd

1:26:07 I mean that one is visible but I think

1:26:09 many of these got CT oh that's interesting because

1:26:11 when you're doing pressure vessel calculations you assume

1:26:13 a uniform thickness and you don't have it you

1:26:15 don't have it so we need to make sure that we get to a to minimum at least

1:26:19 something that they're they're comfortable with yeah that's

1:26:23 an interesting GE there yes so the reason I

1:26:27 actually put it next to that that they can kind of see the sections in the other

1:26:31 one it's an annulus so there's like basically

1:26:33 annulus that comes together there's 16 sections that come

1:26:36 together and then you can form a tral tank

1:26:39 that is so cool yeah that's amazing this is

1:26:43 just a demonstration say it's manually TIG welded

1:26:45 so you got some like defamation as you're tigging um but um oh I can see

1:26:51 the future like if you had multiple robots holding

1:26:54 them in position and he had a robot welder yes you could then weld it after you

1:26:57 form it also our goal is to get to a point where we can do just do

1:27:00 this in two section top and bottom through

1:27:03 that restriking operation then we can also distribute the thickness

1:27:06 a little bit better so you kind of form it instead of forming in water line you

1:27:08 form it like this um and then these tanks

1:27:13 NASA used to make them with bump formers

1:27:15 back in' 60s for satellites it's like a very

1:27:18 efficient design for satellites because it's very volume efficient

1:27:21 right instead of having sperical tanks around your satellite

1:27:23 which is easy easier to make but then

1:27:25 you don't take advantage of the full volume

1:27:27 you can have a full tank around the satellite

1:27:29 really and it pass through for electronics and stuff

1:27:32 wow I wonder if that NASA is smart I wonder if it has something to do

1:27:36 with like solar loading on one side of the satellite

1:27:38 or the other so you keep the fluid

1:27:40 the same temperature they're smart yeah so but it

1:27:43 was very hard to manufacture but you did

1:27:45 it like this and you had support on each side do you call it support we call

1:27:49 it a skirt skirt but yeah it's basically same

1:27:52 as effectively as a support that's cool y so

1:27:56 you have a laser scanner on there to verify

1:27:58 the geometry you ever have a laser cutter or anything yes so we do we don't do

1:28:02 laser cutting today we might add it

1:28:04 in the future right now we do mechanical cutting so

1:28:07 using a spindle we cut the part out

1:28:09 um safer it's safer and also just preserves material

1:28:12 properties like you don't have heat affected zones

1:28:15 as you're cutting um but uh but eventually but then

1:28:19 it doesn't give you as good of an edge quality so with laser you get a very

1:28:21 good Edge quality but you might have we

1:28:23 defected edges you've done a lot in your life

1:28:26 in a short amount of time no thank you it was excited I think I was excited

1:28:30 to to all it was was lucky to be friends with a lot of smarter people that ended

1:28:36 up I got involved with a lot of projects that's a very humble way of saying

1:28:39 that where did you go to school uh USC

1:28:42 USC yeah they have a good engineering program there

1:28:44 computer science uh computer science good I think

1:28:47 but also I had like a very like a unconventional

1:28:49 but I left school so many times to go work cuz I always felt like you know

1:28:52 as long as if school is not going to you work yeah what's the point so I

1:28:56 left a few times I I left work at Microsoft for a couple years and then I worked

1:29:00 for Google for a little bit um now I went to SpaceX and got really liked it

1:29:04 so I left scratch school just just you know

1:29:07 join so you didn't really focus on the end like the piece of paper at the end

1:29:11 you focused on the actual education right and you

1:29:13 said oh well the limits of my education

1:29:15 are I'm not getting hands- on time so you

1:29:17 just went and got it yes I mean if it was an opportunity for to actually get

1:29:20 stuff done and do it and learn I

1:29:22 think I always took that opportunity it's it's just

1:29:24 you know until you end up doing like like I said always know it's like there are

1:29:27 so many things you don't know unless you start

1:29:29 doing it and that's when you kind of learn

1:29:30 it right so so what would you say to a young person right now that's coming out

1:29:35 of high school and they're interested in uh

1:29:37 robotic manufacturing I think Robo forming is a unique

1:29:41 subset of Robotics industrial robots I think they

1:29:44 should just start you know joining the projects

1:29:47 that are doing these things right I think intern

1:29:49 early just make sure in you're schooled you're intern

1:29:52 early you need to have a still have

1:29:53 a good Foundation but now you can learn Foundation

1:29:56 anywhere you can you can even watch your videos

1:29:58 and learn some of the foundation right the video

1:30:00 you just made yeah yeah exactly so I

1:30:03 think that's just the beauty of it right now

1:30:05 you can just decide what you like and go

1:30:07 learn whatever you need to learn while doing it right go find the people who are

1:30:11 doing that join them and meanwhile learn everything you

1:30:13 need to learn and you know so don't necessarily

1:30:16 get caught up in the uh traditional education system

1:30:19 education get this and get this yeah and then go to work yeah dude I'm impressed

1:30:25 with you you're an impressive dude man thank you

1:30:27 so much thank you yeah yeah okay we came to another room what is this this is

1:30:33 our post-processing room so this is where we basically

1:30:36 once the parts are done we bring it here clean it out you know we cut the tabs

1:30:41 that you might have in the part and that's what I want to see here so so

1:30:44 we've we've cut these tabs I notice you

1:30:47 insert and then you go along here what is

1:30:50 the methodology here I noticed they're they're kind

1:30:52 of interesting so is that a function of the robot

1:30:55 path there the tool path yes so basically

1:30:58 the robot comes in cuts it leaves some tabs

1:31:01 and then in the end you kind of cut some tabs you can see kind of the sheet

1:31:04 moves slightly and those are the spring back

1:31:08 comp residual stresses right so when you're forming

1:31:11 the part you actually have to form the part

1:31:13 slightly different that what you you think you're

1:31:15 going to end up getting and then in the end it kind of shrinks a little bit

1:31:19 and then you're going to get the rock

1:31:20 butt so you accommodate for this yes oh that's

1:31:24 interesting and so once you make that cut you're

1:31:27 done and so you're not going to you would have to rescan it if you had any

1:31:31 other processes after the cut but you don't care

1:31:33 you're going to come in here and jigsaw

1:31:34 that out what's exciting for till now like we couldn't

1:31:38 do with stamping is that you know you

1:31:40 couldn't do whole lot of different designs because

1:31:42 for every design you have to make a new

1:31:44 mold right so with our process now you start

1:31:46 can make very intricate designs of sheets that you

1:31:49 can kind of put them in a sandwich structure like you kind of stack them them

1:31:53 against each other maybe you're Riv it at certain

1:31:54 points and now you for example all of them together can look like a wing a shape

1:31:59 of a wing but this Wing is super performant

1:32:01 because it's very light but because of the structure

1:32:03 of those sheets that were inside of it

1:32:06 becomes really rigid right so you can optimize

1:32:08 the structure whereas you were like in an airplane

1:32:11 you used to have like you were limited by the type of material you had to create

1:32:15 a LeRon or something inside the the plane

1:32:18 you can create a customized even AI optimized

1:32:22 part that's the key right now you can actually

1:32:23 iterate on it and you say okay give

1:32:25 me five six 10 different sheets that are specifically

1:32:28 optimized for this application the robots will form it

1:32:30 and then maybe you want to change your criteria

1:32:32 and optimize it for another application and the robot

1:32:34 will uh form a different one and the computer

1:32:36 comes up with these designs I I can even imagine a future where you would build

1:32:40 a wing you would have the different structure

1:32:42 that you're describing inside and you load test it yes

1:32:46 and because you could do it in 3 days

1:32:48 do another you build four wings with different optimized

1:32:51 structures and you see which one breaks at what

1:32:52 load absolutely oh that's or you can capture

1:32:55 so much data and then train these model

1:32:59 chose the data to train the computer to come

1:33:00 up with the most optimized model too right

1:33:02 yeah yeah the old test and evaluation model model

1:33:05 test model because you can generate you can generate

1:33:08 Parts without process faster than you can simulate it

1:33:11 yes that's what I think the computation say

1:33:12 that again that was huge you can form Parts

1:33:16 faster than you could simulate it on a computer

1:33:18 right that is such a big deal yes so you're not like even reputation is a slower

1:33:25 than forming Parts in physics right so without

1:33:32 process thank you so much for watching this the second

1:33:35 channel here on smarter everyday 2 I guess is what we call this channel but um

1:33:40 yeah this Channel's pretty cool there's just I get

1:33:42 to go deep into the details it's really

1:33:44 really cool if you're watching this stuff you're obviously

1:33:47 a very intelligent person and I'm just glad

1:33:50 that you're here because I'm learning on all

1:33:52 this stuff and I'm EXC excited that you're excited

1:33:54 to learn with me so anyway if you'd like

1:33:57 to consider subscribing to smarter everyday 2 that would

1:34:00 be amazing because it's kind of like slowly

1:34:03 this channel is like the slow steady channel

1:34:05 That Could it's just like slowly growing which I think

1:34:08 is really cool so if you'd like to check

1:34:09 that out feel free to consider subscribing if

1:34:12 not no big deal I am Destin I'm just

1:34:14 grateful you're here have a good one bye [Music]