[00:00:05]
>> We have to do talks today to give you a little bit shorter than I have heard some of our. Heads up that. We really love him your apartment and I'll have. A sample of your receipt in your. Share what they're doing in their respective research groups so we'll start with here is associate professor of Mechanical Engineering Michigan now for about 8 years she got her B.S. probably see bowler after.

[00:00:38]
Her brother works is that he has been a factory of robotics folks a lot of bodily sense all those that were paid who. Were here of the day it was many orders the end of her career or the most recently in a S.M.B. but never discussed a whole.

[00:00:56]
All right thank you thank you. So yeah I'm very excited to be here today I'm in a actually just found out this is a short talk so I might roll a little quickly through some of those but this is actually really just to give you kind of that perspective that the way that we think about coming at these problems that we're going to talk about is we think about how to do it from a control architecture which means we look at how do we get data out of our system so how do we sense things that are going on how do we understand what's happening which is really the modeling piece and then how do we use that information and synthesize it in such a way that we can make intelligent decisions on what we want to do which is control OK And so we do that for actually a lot of different applications in my in my group so they are broadly in.

[00:01:42]
Manufacturing they seem very disconnected they actually are into when you get to the underlying sort of structure of what we're doing and how we're approaching the problems but you know you can see there is some cooperate of systems there is even from the sort of rehab robotics I mean factoring plays a big role both at kind of a system level where we look at more of the smart manufacturing ideas and then today what I want to focus on is kind of the process level and specifically additive manufacturing.

[00:02:11]
So this is the outline but we don't care about that so additive I probably don't have to motivate this for folks but you know we have been seeing more and more interest in this and especially started to look more at functional devices and that's really exciting to me because you see some of these really cool applications and we see some from lots of different kinds of 3 D. printing so at the end which infuse deposition modeling that's where a lot of folks think about like a Maker Bot this is obviously a higher end version of that all the way down to electronics but where I think that you really get a lot of play in this and where particularly for myself I'm really interested is actually when you start going into the micro a nano range and at that link scale that's when we start talking about things like sensors and actuators and looking at this from the electronics the optics even the acoustic side that's a little bit lower resolution but that's the kind of base that makes us excited in our in our group so to put some perspective on where we sort of had to start thinking about what we wanted to do in terms of the type of process.

[00:03:21]
This is a quick overview of kind of more conventional 3 D. printing technologies and where they lie in terms of late scale material diversity and those are 2 really important things for us when we want to think about how do we make functional devices at at a high resolution or really at any resolution but we tend to look more at the high resolution.

[00:03:41]
So metal based powder directed muddled up position these are going to be your structural pieces right that's what the automotive the aerospace that's what they're really interested in is this this domain there's a lot of really cool things this is where we're seeing some actually commercialized products come out not a lot of material diversity but.

[00:04:01]
A lot of work there so that fuse deposition modeling that we see a lot of that actually in some components like the brace or the robotic hand we saw you also see this in biological applications and actually I put it here but it can really stretch along this sort of whole domain here in terms of resolution size has a fair bit other material classes that it can work with and you can embed composites and other interesting things carbon fiber pieces like that into these materials to support with them.

[00:04:35]
OK stereo the fogger fi again sort of single usually classes of material but much higher resolution and there are some newer versions on the market that try to do this really quickly. Probably one of the more common ones that people are familiar with it comes out of the same process this is what you think of for your printer actually a fair bit of diversity in terms of it can do multiple materials at once can change colors can change some textures and dromedaries and so you can in the same process kind of change the physical properties of something but it can be limited in terms of the length scale it can get down to and also the class of material spits specifically when you start think about things like this cost city so it doesn't like to really work with anything that much more viscous than water.

[00:05:29]
One of the newer ones is air cell jet this is actually an area to process that takes bunch of materials sort of puts it into this air process and then it just it can do that actually a pretty high resolution down to about 10 microns maybe slightly below that but has some restrictions in terms of the types of materials because it has to be able to break these materials up into particles and so it can't really deal with anything there's a particle size greater than about $100.00 nanometers.

[00:06:01]
So what you see was that we saw that there's this big gap right here OK and so that was where we wanted to be able to play we wanted to be in the range sort of in these 2 spaces but a much greater material class and so we're going to come back to that but what I want to do is motivate why we also start to think about that we're going to probably want to think about how to put control and not just in maybe the high resolution space but really across the board and this is a extrusion based process we're showing here and issue is that this looks really nice and it's going to go up it's going to build these nice stacks and this took months of tuning and figuring out the process motors and working with the materials to get to where we could do this repeatedly and just build the stack after fact after that this is actually a.

[00:06:51]
Hijack the appetite so it's a bio material that can be used to go into things like a bone scaffold and it's naturally biodegradable. So well you know I guess maybe some grad student something to do for a while that's not really very efficient or effective and so and a big problem is that you can make these nice structures but as you're going along there can be disruptions OK and the process unfortunately doesn't know that that happened and so you know you can end up with things where your fear may be pooling material at a place maybe you have wind variation that you don't want you might have some other strange dynamics that are happening when you turn things off and on and this process has no idea that that's occurring and so all of a sudden you're going to end of something that looks really funky OK And this is not an extrusion problem this is a problem in basically every type of 3 D. printing OK is that these systems aren't typically tracking themselves and they're not self-correcting and so this presents a big problem and this is where we have to think about how do we try to increase throughput and reliability.

[00:07:59]
In these processes and that's that's really where we get to come in and think about how can we provide that sort of closed loop concept to these processes. And we're going to do that with a particular process but one of the things I want to start with is you know why is this difficult to do so why don't all these systems already have this but the problem is that actually the physics really are at the micro scale and this is true for pretty much all of these processes OK so what's happening and where the failures are occurring is really at intersections of material classes it's it set a different length scales it's often really hard to see or to monitor in some way while you're actively printing and so maybe you want to be able to use some type of across them across microscopy but that's difficult to do in real time Additionally even if you do have sensing which we do maybe you can't do it in process OK so you have to think about how can you use the potential sensors that you could integrate in in a way that's going to allow you to improve the process and so examples here you might have some kind of a camera this is a high resolution you could have a high speed camera that's going to actually the example here is from a high speed camera that's going to give you maybe information about the deposition process but how do you collect that data process it and use it in some way similarly maybe in our system for example we have something like a an A.F.M. an atomic force microscope that gives us you know some policy information but we have to think about how we would use that and put it into a control architecture OK And this is the sensing problem but there's actually also another big challenge with closing the loop for control and that comes from the modeling side and the modeling piece is really hard because we have to sort of balance several different things.

[00:09:55]
Complexity actually usually goes hand in hand in many cases with accuracy and that's kind of an a challenge for us from the control side because what we want when I say a utility for control I actually want something that's relatively straightforward I don't want it to be super complex Ideally I'd like it to be linear OK we all know that systems aren't linear but I would like to be able to represent in the least the domain that I'm going to work in a linear representation that opens up tremendously the types of control algorithms I can use I'd like it to be relatively accurate because I'm probably going to want to use a model based controller OK so any uncertainty I have in my model is going to show up in my ability to control.

[00:10:40]
And so I want to be able to do that and what we find is that there are several different classes of modeling types out there and we have to think about whether or not these are going to be useful for control so the way that a lot of additive manufacturing control started was actually with the use of static or process maps OK so these give a representation and they basically show you a mapping between a set of inputs and output behavior and so these are processed maps.

[00:11:13]
Input output behaviors that challenge with these particular classes and they do give some insights on how you might soon but they are not dynamic OK So anything that happens they need a McLean in your process you don't necessarily have information for that and your process is still not going to give you an insight on what might happen so they're very useful potentially from the design signs but not so much from the control side.

[00:11:41]
So then analytic models will that actually seems like a good direction to go we can get an overall sort of maybe moment balance or power balance or momentum balance energy balance models you can use lots of different ways of thinking about it to sort of understand how the material should be moving in and how that can connect back to the process so these actually can be really good but they can be rather complex to work with and actually from a controls perspective they tend to be.

[00:12:13]
Lead to P.D.'s which have both basal and temporal information so they're more complex a to work with I'd like to work with O.D.S. so ordinary differential equations and they're non-linear and that non-linearity may be an issue but maybe I can simplify it so that's the potential class it does give me more accuracy and it's sort of midway for control OK then we have computational physics right now these are going to buy by far going to be the most accurate OK this is something in a simulation and sickles you going to model it it has the embedded physics there OK So it's actually showing what I should expect in terms of a except to do sorry with my mouse go.

[00:13:00]
If you get great so you get to see cool things like that and you can see that to dismantle go to right so I can get an idea about you know maybe how spreading my happen on the surface and stuff like that so again useful from sort of maybe an intuition standpoint it's not a mascot Akoko Quezon it's really hard for me to think about how do I go corporate jet into a control framework so.

[00:13:24]
It's not analytical it's hard to put in controls so what we find is that actually we want to start thinking about can we take some of the pieces from each of these and push them up into new models that are to be controls oriented accurate but relatively low in the complexity side OK so that's really what we're tasking ourselves with in order to build steps towards control so now I'll come back here and we're going to focus on a particular example case where we're working with a process that is meet kind of this technology gap and when I think about how can we apply these concepts of control to this particular process but they extend to other types of 3 D. printing so the process I'm going to focus on is known as electro How do they make printing really long someone call each at shorter name it's actually jet base process so in theory it sounds a lot like inkjet and it is in many ways but what the big difference is that uses a piece of actuation or thermal actuation So effectively it has pressure that comes from behind and pushes material out of a nozzle OK What that does is that sort of limits resolution and also that material class remember that I said it.

[00:14:37]
Only really likes to deal with things that are roughly up water. Now with each What we do that's a little bit different as we actually use an electric field so we connect electric field between a conductive nozzle and say a conductive surface and it actually pulls the material out of the nozzle and then breaks off at the tip here and so it creates this Taylor cone and rejecting from the Taylor cone which gives us a significant amount of resolution reduction and so rather than being stuck at maybe 30 microns we can go down to things that are sub 100 nanometers and feature sizes.

[00:15:15]
And interestingly also this even within that resolution or that sorry yeah that resolution range we can actually do that for materials that are closer to a consistency of toothpaste OK So anything from water to toothpaste it can have particles in it and we can get it to much higher resolution so it actually opens up the sort of class of materials the class of devices that you can potentially do with this process so as I mentioned it's a jet based process which means that you know it undergoes this jetting dynamic and it does this repeatedly over and over again and then you know what we want to be able to do is sort of 2 phases we've potentially want to be able to control this piece and then we also want to understand this interaction because both of those are going to play a role and how our device is functioning so what we do to control at the high level of the jetting process we use of polls.

[00:16:12]
And that allows us to shape both the volume of material that we release and the point in time in which we release it OK so if we leave it to the natural dynamics it will create its own sort of repetitive process but there are those 2 are linked to that the frequency in the volume become wings and we don't want that we want to break that and Suppose think is and I say this is a square you can use actually other shapes depending on the material you might want to do that some of the materials have vibration reactions afterwards so you might shape that input.

[00:16:48]
You know so we only pulse of the nozzle Yeah and the I mean you can actually flip polarity but what we typically do is we do the high voltage and high is relative so high can be anywhere from usually $3.00 to $700.00 volts that you're applying to the nozzle and then it hits the ground in the in the in this is just that it's a passive surface you can actually do this on a floating surface but it's the results are it's more variable but you can do that and then we have other variations of designs that allow you to actually have a lot more freedom but that's typically what you do and then say a 30 this is a 30 Micron but I would say the range is closer to $2.00 to $5.00 microns is what we typically use for nozzle opening.

[00:17:31]
And then that prints that that's the most standard one that we end up with and that prints anywhere from $500.00 nanometers to like 2 microns. Right so in the standard set up yes you deliberately want this to be pretty flat and ideally conductive in our sort of innovative designs that I'm just not going to get into today no we can have contoured flexible surfaces that don't have to be conductive and we're not this is roughly 30 microns up the surface we're now $5.00 to $20.00 millimeters off the surface.

[00:18:08]
So but for the sake of the control piece I'm just going to talk about the standard set up. OK so lots of different applications for this because we've got a lot of materials that we can print with OK Anything from sort of conductive materials down to materials and then the the substrates and so here you can see typically 5 as I mentioned you start to get to things where there contour we can actually print a 90 degree surfaces something that's completely vertical we can print up an over flexible all of that OK and then these are just some of the sort of the to stick I mentioned you can get down to sub 100 nanometer feature sizes.

[00:18:48]
It can obviously get larger This is not really true we can actually go to the Greater than millimeter scale but we generally don't for our application it is drop out of the ME and so you can control that but you can do lines and single droplets and then you can use things as markers you can do things as electrical connections there's a lot of applications OK wise it's an interesting one to start with well it's actually pretty.

[00:19:14]
Susceptible to changes that are going on and so there's dynamic interruptions that happen but a lot of these actually might be built into dynamics that we're not capturing by the way of slow this down just so we can see and of course that's right because Michigan the interesting thing about this one is that this is we're printing on the hydrogen L. which is a surface that actually has water based and so as we're printing it's changing because it's evaporating OK so the height of the surface is actually change and you can't really see it in the camera but we're still able to go along in Princeton we were printing some.

[00:19:49]
Fiber and in this one but we're printing some biological markers in here. But that variation should actually be something we can start to be predictable as we're going along and so we want to use some of the properties of what we're doing with the printing to help us think about the control of the modeling piece.

[00:20:07]
OK so let's close the loop so with Egypt specifically as I just said we just discussed there can be some of the variations in the substrate actually induce changes so if those are dynamic then that may be something that's a little bit more challenging to deal with but if it's something that is actually measurable then we should be able to adapt to that.

[00:20:31]
And so that means we have to be able to measure it of course but also another interesting thing is the progress is not the only problem for potential problem for us some of these materials may maintain an electric charge for it so if you print if you're creating electric potential and you print that onto an insulating surface it's not actually going to dissipate the charge OK that's actually a benefit in some cases you because you can purposely charge material but you have to keep that in mind when you're printing and so there are also you know instances where you're an alternator current and you going to do things like this so you actually reap your polarizing to different positive negative so that your materials don't want to separate from each other OK so you have to keep some of these things in mind what happens if we don't use any kind of control or we don't maybe are our static our static process map is a little bit off and then this is what I wanted to do and you can see it all it's doing is creating this really strange sort of like scatter so several droplets are actually releasing which what that tells me is that I thought that I was in a particular judging regime which is one where I have a controllable release of material and I've gone out of that regime into a spray regime which means that when I'm pulsing instead of a single drop of material it's actually sprayed which from my understanding of the physics means my voltage potential is not what I think it is so there's some interference that's happening there could mean I'm too close it could mean an action this case that is what it would mean OK is that I'm quotes and so I've gone out of there judging regime I think I'm in.

[00:22:05]
Also what can happen in this is interesting again from the sort of surface interaction piece is that these materials can sometimes do really strange things so I prints a material and it just wanders and moves over to another material and joins it OK and so active I'm a control perspective that's very strange if I put something here I don't expect it to walk off and that's exactly sometimes the material does so when I'm thinking about how to control these I also have to think about what are the inherent interaction properties and what can I do to maybe make that more.

[00:22:42]
Structured so that it's not stochastic I want to be something deterministic which means I understand what should happen and I can model that and therefore I can predict it and if I can do that then I can control for it OK And so we have to think that part of what we try to understand in this process so we break this down for from control we break it down into 4 pieces I'm not going to go into all for these this one you know to start off with material characterization that's a big part and that's something we always have to start with which is really just what is the compatibility between the materials or even trying to work with OK So let's now assume that this has been solved and we know that the surface we're printing onto and we're printing with are all going to work fine and we're going to bond.

[00:23:26]
And the properties should be relatively consistent So once that's true then I have to have some sense seeing and in my case I'm a news camera and then embedded atomic force microscope so I can get information back but that information is probably not going to be real time OK it's going to be layer to layer so the piece I'm going to focus on today is actually our model and that's because we have some really interesting things to happen here I have them in this generation which is temp oral so that means it happens as a function of time and I and it's a fast time but it's something that I can monitor and sort of try to model that behavior I also have to topography which also has time dynamics but there's so fast and I actually don't care that much about the time dynamics what I'm really interested in are the spatial dynamics and that means I want to understand the spreading behavior and it because it happened so fast I want to look at it steady state right once the time to has finished I want to see where how does it set itself up.

[00:24:26]
So we go back to our knowledge will and we can think about this in a couple different ways of modeling it so I can look at them and if which is the jabbing behavior in this is a time dependent process and actually we break this down into really it's 3 stages but these 2 get lumped into one process but we start with you know we apply a charge and then there's lots of things that hear this talking about different components because we're going to model this as a hybrid system.

[00:24:57]
But I'm not going to get into that piece but we start with a charge and what this does is starts to create a tailor code now so interesting Lee with our materials we just actually like things that are polarized OK they just need to have some free irons in there that I can try to pull it out if they don't we can actually still attack them but it's less control OK so they don't have a lot of restrictions on what's in here other than ideally that it polarizes will it starts to create this Taylor cone and at this point this is actually the most value that I want to hold it at because this is where now once I know the shape it's pretty quick for me to be able to pull the last bit for it to release OK so when I pulls I actually hold it at this shape when I go back I don't go back to 0 I go back to like a baseline and so I hold that shape and then when I want to release material I can pulse higher I can do something like that.

[00:25:50]
So we do some preliminary to know that what it looks like and then we model it so that we know how much the initial voltage should create. And so then we apply the electric field that actually creates now in many cases actually you can't you don't always see this happening and it looks like they're breaking up in air but in many cases when we're near it does come down and it touches the surface and releases so this process does make a short connection and then it breaks off and tracks and then goes back up to.

[00:26:26]
Too it doesn't actually usually go all the way up here if I apply the voltage try to go here but if it's in it's own cycle it will kind of continue to go through this OK and in each one of these and here in this process right here and in this process right here we can actually model dynamics for each of those and so that's why we think of this is what's called hybrid because there's like 3 stages and they actually have very different dynamic behavior and so there's a build up that we can model using some of the physics that are driving things and then we need to know when it transitions OK So we've got this to help us understand when it should be transitioning and then we go into judging now jetting is something that we typically actually.

[00:27:09]
Figured this out using data it's much harder to model this piece so we'll use a data driven approach and then relaxation is oftentimes it looks just like the reverse of this but other times relaxation will create vibrations like a ripple effect almost it's almost like I release and then it sort of oscillates the material and I don't like that because it will sometimes release another bit of material so it changes my volume yes that's just to say you can change your input signal Absolutely so you just what shape it differently to dampen that dynamic Yes.

[00:27:53]
That's right it's too fast and the resolutions actually too small so when when we show the the high speed camera images that's a 30 Micron nozzle that we're purposely using a bigger set up and we're we're targeting a larger feature there to sort of understand the dynamics but the jetting frequency the firing frequency is $25000.00 droplets per 2nd.

[00:28:18]
OK so it's actually very fast and and mutually this stream of material right here is sub one micron pressure so that's extremely hard to capture optically and that's why we try to be able to model you know and understand the physics that's driving it so that we can't monitor while it's happening but we should be able to then after we deposit we can detect what the material volume is and so that helps us sort of cordoning with what we think the models doing and that helps us improve our models Yeah.

[00:28:57]
Yeah. That's correct Now interestingly what we find is temperature and humidity play a very small role in ours so our process dynamics are actually pretty stable over a relatively large range of of temperatures and humidity so we do this out in the room temperature and room air but we can we've actually put it in chambers and done all of these these tests where we put it down to you know 0 degrees we put it up to 100 degrees and there's a you know there's a mid B. and it's not it's not the same dynamics over that whole range but there's a pretty large range of temperatures and humidities that we can work with and so you're right material though it's very material dependent so we do have to model each different material class Yes.

[00:29:59]
This high speed into Yeah that would be great. Yeah we often if we had one on the main It would just allow the process to to be faster in terms of real time analysis of it. OK. OK OK look into that OK So this is the time to I mean there's the spatial side to right which is spreading in coalescing a behavior on the surface and so we also want to be able to understand that so the materials will you know if you think about a single material it comes in sort of looks like this you know some my circle that spreads out and sets up at some equilibrium contacting go OK and that's based on this gas the surface tension things like that and then if you're interacting with other droplets you also have to sort of understand that behavior now we're showing it time 2nd here we actually don't model we have but we don't we're not that interested in the modeling of the time domain process we actually just want to look at what is the steady state behavior that happens OK So what are some things we need from a controls perspective Well I I need to have this concept that if I put it impulse of energy into my system it outputs something OK so I need that aspect of it and that lets me do some things like system ID which allows me to get an input output relationship.

[00:31:34]
So that should should be should happen it shouldn't be that if I'm sitting here and I don't put an impulse of an energy in that the process does something OK that's what I don't want so I'd like that concept of deterministic I'd like it to be something that that I can model.

[00:31:51]
Eyes like for our case we're going to talk of this talk mostly about the spatial one I'd like it to be spatially variant which means that I assume it if I can I try to assume that the spreading behavior looks relatively consistent around OK So that means that if I'm here I'm over in this location of this location that behavior is more or less consistent now underlying topography will change that because it will react to that but if I were to have a relatively planar surface then I want to know that that it's going to look roughly the same and I can model that in sort of a spreading behavior and then the lot pieces as I said before linearity is really nice so if I can have this process be linear which means that if I put more energy in something I get more out and so I can control the sizes for example of the droplets that I release so I can go from 2 to 10 microns in any pattern that I want and I have control over that OK So from our model development side you know what we're interested in is really this behavior so how does it spread and that so that then we can predict and start to look at you know what should we be doing for changes to our input signals from the controls perspective to make it get a feature that we want.

[00:33:12]
So there's different methods that I can use if we were called the for the 1st part of the talk which is that I can do data driven I can do analytic modeling and I can do computational So for our case we're going to see what we get so I'm going to do a really quick example where we show how we close the loop I'm not going to talk much about control that really show what we did for the model so what we wanted to do was to create repetitive features OK We wanted to just be able to repeat our device fabrication over and over and over again so I use what's called additive warning control it basically means that I can learn what happened in one device I can take that knowledge and I can use it to enhance my inputs to my next system and that's all done through a learning controller OK so then there is information that goes into how do I design this is an input that's what you is in for us that's the pulse with to control the amount of material release.

[00:34:06]
And then I can design that based on what I do before and what's the air in my system OK. In this case that So for the model what we did is we went through and did actually a system identification which means I wanted to just kind of get a data driven representation for this initial study so I went through and I printed a whole range of droplets and we just did this over and over again and then statistically we looked at this information and then I said OK can I derive a model that roughly captures that so we modeled this is a actually a 2 dimensional Gaussian distribution you know some of the fits look better than others but what we can do is I can look at what the data really looks like because I can measure each one of these with an A.F.M. so I can get that 2 dimensional shape and then I can understand what the difference is between what I would predict with my model versus what it actually is and has someone certainty and I actually have a relatively broad range of inputs to which is essentially droplet size it's relatively linear and so that means that as I put more pulse with I got a bigger droplet and it scaled roughly the way that I thought it should that was great so I can capture that model.

[00:35:19]
And so then I can take my system so here's our printer right here's a substrate that we have this is a light this is a actually this case this is our high speed camera here's you know our knowledge will hold or stage back here actually this orange of the A.F.M. and then just what we were trying to do is we're trying to build something that looks like this where it merged together and I was in a measure and then finds out that an air and then I again I have some uncertainty in my system so how does this work we'll hear Jesus will start perfect OK so here is our system now we should out of here because this is actually is moving but you can't so this is a simpler example of that where I'm illustrating here but we print.

[00:36:04]
OK we're doing this repetitively so we would print our pattern in this case we're using a U.V. curable material so we could go over and U.V. cure it directly in the system we go over to an F.M. It measures everything not this fast unfortunately but that data comes to us we look at that data we process it pretty quickly actually the whole process doesn't take that long but for sure the longest is the if.

[00:36:31]
We take that information we run it through a controller we come back and the next part we build would have new inputs to our system and it was just built in so we can leave the system running for 20 hours and it will just keep running and it will actually learn itself what's the right set of parameters and so after maybe the 1st 2 or 3 it will actually make consistent parts OK just over and over and over again.

[00:36:53]
Now we're going to see again OK so now here we're purposely playing with some different things this just was to show that we could tune based on control actually how we did and. These are really weird images but what we can see here is that you know we don't know much about our system we actually pretty terrible and that's why modeling is really important and so we want to be able to model some things that are important for us from a manufacturing do we converge quickly that means how many parts do I have to do poorly before my system actually behaves the way I want.

[00:37:29]
I also know that if I tune the controller as well I'm going to have different behavior so I want to understand enough about my system that I can tune that controller and I tune it based on how much noise do I think I have how uncertain am I about my model and uncertainty is in my models really important because if I'm wrong if I'm really wrong actually the system will just go unstable and it'll just keep building things that look terrible OK And that's why I don't want.

[00:37:57]
So I think actually we're out of time so just to give you kind of a hit were you know what we want to do is of course move towards multi-dimensional multi-layered and that becomes a really nice fun problem because now you have all the underlying typology that you have to think about but it's also built into your model because now you're not just putting things down on a flat surface you're putting things down on conduit surfaces that have dimensions and different behavior and I have to understand how the material is going to set up to that and all of that you know also depends on the miniscule part too so when I'm doing this surface interaction I'm assuming my volume is consistent it's not so now I need to work on that control piece too and then put them together so that the whole process becomes a controlled process OK that's it thank.

[00:38:51]
You. Yeah they're mostly. Just spent you know most of them are suspensions actually. Especially the conductive ones right there nanoparticles suspensions so. You know what we're trying for so definitely. Yeah electronic devices Sharon says yeah so for the drop of. A certain amount of things so so that's a really good question for what we're going for I think we're going to have to do some work on looking at is because of their properties could we be able to identify say you know given a certain electric potential does that poll before it releases a certain sort of consistent concentration with the ones that we have basically the particle suspensions are pretty well dispersed and we actually sonic ate them before we use them.

[00:39:51]
That we don't actually control for that but what we find is that it works I guess relatively well but there are other times where it's there separating and that's still I think to be figured out in some of that comes from the ink side and then the other is to try to again sort of build understanding's of you know if I jacks this amount and the material has these properties what's going to come out that's on the 1st guess what you drop the device now then the device yes you do it yep exactly how much warmer a challenge that So that's it then different problem right then that switches over to another problem which is a more of a robotics problem so now you want to go register you want to find these devices right so there's lots of actually different techniques and things we're thinking about there I mean optical is potentially one but if that's not you can I mean there is there's different I think sensing techniques you can do to go in and sort of process your the earth surface after you put in.

[00:40:51]
Down so you understand where things are and what the lay of the land is and then that gets program back into your system to go back over and just Or for example if you wanted to print on top of something again it's not science this is this is wrong yes yes yes.

[00:41:10]
Sorry. So. Interesting we know the conductive material doesn't actually usually charge up so the outer part of the nozzle is what is what's conductive and not the material itself now it will sometimes less actually with the conductive materials we find like with our silver and things that could potentially be that they don't truly become conductive to we've centered or cured them and so they're they're sort of in this poem or suspense not really polymer but they're in a suspension so they're not they're not charging really in our material we now that being said we do arc Yes So we do have to be careful about the voltage levels and there are certain environments where that's more common and then we have to be careful about that and what that does is we have jetting regimes and it narrows the band of the regime that you can control and work within So you just have to understand where those boundaries lie so that you found your control or your system in general to not go into those but it does and then it sort of shatters the nozzle if it's a big problem to.

[00:42:32]
Share. Yeah yeah. Yes. Yes Absolutely yeah so I didn't get a chance to show this but. Maybe I actually was back in the simulation methods but. I'm not going to get them to. Actually here so we do a lot a fair bit of modeling where we start to look at mergers and how those the material merges together and it's driven by the substrate it's driven by the material property of the ink itself.

[00:43:17]
Even you know what you might induce if you induce some potential to that conductive don't put some of the materials will pick up some of that charge you have to kind of take all those into account when you model that. And these are mine but we do surfaces too so we do a full surface and it's very important they're OK You talk.