ZDZW Webinar – Digital Twin Applications for Industrial Thermoforming

e hello everyone uh hope you can hear me if you do please write something on the chat and also I have uh given you permissions to activate the microphone as well okay great thank you Sandra um let's wait a couple of minutes since we're having a little bit of trouble uh getting atan on on board e e okay here's heran hello Iran good afternoon yeah now I can see hi guys okay hello um uh just uh to make sure uh can you try uh sharing your screen I think you you must be able to do

that let me see one second no I don't have any uh share is not enabled for me okay uh okay let me see how we can handle that because elar cannot join us and in the end they are the Ean is the uh the ones that create sessions so uh my recommendation would be this uh can you share screens I guess you can right yeah I do okay so I already uploaded the webinar into work package 10 folder okay that's great so yeah I'll just that's that's perfect yeah that would work okay okay uh just

give me a second I'm on it there it is okay I have it now all right so well just uh just to let you know a little bit of the the uh the I don't know if you were uh if you were present when we made our uh webinar but the idea is that you start presenting uh anything that you have prepared and then we'll leave uh some minutes for any questions if if anyone from the audience is is has any okay sure okay so I'll start sharing my screen uh ban uh okay I hope

you can see it now yes I can thank you okay that's great so uh well first of all let me um welcome everyone to this the second webinar of the zero defect Z waste uh project in this this time we're going to speak about digital twin applications for industrial thermal forming uh it's going to be presented by dran Turan from similar and well from now on I give the floor to you Eran so please um thank you U maybe you can also switch the modes because we just see the oh okay I'm sorry I'm sorry

yeah I know I know what to problem no problem I'm sorry that better now yeah thank youate all right now next slide please there we go all right so today uh I'll be talking about some digital T applications for industrial ter forming and this webinar will be related to Z dzw project which is uh zero defect Z8 projects supported by European union and uh this is the short summary of our uh webinar for today I'll be briefing talk about what ter form is and its application areas then a little bit of digital TN will be

explained then I'll be giving a couple of details about the termal side this is terminal modeling of uh ter forming in our project and one in use case will be given at the end of this to now maybe uh can we move to next slide please so let me talk about my first so this is Aran uh I'm the CTO and the co-founder of similars uh I have a mechanical engine background I did my PhD back in 2010 in BOS University uh I've been in etur uh Brush company as well as G Aviation I did

a lot of engineering stuff both in the Academia and the industry uh before joining simar I was focused on developing jet engine designs for Airbus and boink aircrafts and starting from uh similar uh our main focus is creating digital uh transformation Solutions as well as digital twins using our own technology so my interest are generally for industrial Computing numerical science as well as digital transformation AI next slide please uh so uh let me also talk about uh zero defect zero waste uh project so this is a project funded by Horizon uh it is related to

this T green and digital transformation call there are 27 Partners from uh different countries All European we have also smmes we have institutions we are universes combined in a single framework the duration is 3 six months and we are right at the center of this project 18 months already passed and we have a some little little bit of information and initial results from our work so similar is part of this community and the reason that uh zwz uh ZD ZW project is initiated is that there are a lot of problems with the waste uh typically

waste can occur if there is a quality problem during the production process if the manufacturing parameters are not set then we see quality issues there is one area that would like that as a project uh contributor we would like to help the community another part is uh when we do detection uh to understand how defects work sometimes we do destructive detection processes the and the idea is if you can create nondestructive detection algorithms or system it'll be really useful because we don't have to waste the part that we are trying to solve and to address

this uh challenge uh there will be a couple of work Suites or let's say inspection suit will be created uh among this uh project Works scope our goal as similar is helping the termal inspection sus while we also have like I don't know visual inspection suits or some other suits as well and you can see here from the slide as well typically uh destructive testing techniques or inspection Technologies are really time consuming uh it requires a lot of investment it is also requires cost but we would like to keep it short and we would like

to give uh a high-tech a chance to help this uh area um maybe we can move to the next sections right as I said before we will be using Advanced AI techniques on engineering uh approach to solve this problem yeah so this is a general uh overview of Z the ZW project so there are six pilots and all of these Pilots will be focused in one or two uh inspection Technologies the pilot there involved with is the pilot two where we're going to help tal Plastics manufacturing getting better using digital tools and you can see

as I said before there are inspection suits there are proed mon diagnosis you to create a digital ecosystem in which the project contributors will be creating their own applications and under a market place or a on a general storage area this will be accessible out of the members of the member communities okay so let me talk about thoring so uh typically when we produced uh when we produce parts from Plastics terop forming is not the first choice and plastic injection or blow molding is among the top uh alternatives to create Parts out of plastics but

thermop formics Al thoric procedure also has its own use I mean in reality if the pieces are large or they cannot really be created using typical plastic injection operations we use thoring in Shell ter forming is not a single process it has some stages and the first stage is is typically heating up the piece so it gets softer and we can give it a shape second St is where we apply a mold on vacuum or may you know open or closed chamber setup using this mold we can give the desired shape on the softened material

then we wait uh which is the cooling part and in the cooling part our goal is to keep the temperature as low as possible so the shape that gate to the material Stace as it it is and then we have this triming process where we chop off the additional Parts you can see there are two types of uh thermop forming the first one is the thing gauge it's generally used for packaging of foods or similar uh sizes and the one that is more uh of a IND industrial use case is the heavy gauge it is

used in liners both sometimes even the front section of the ATM machines we can move to the next stage next slide okay as I said before there are a couple of stages of thermop the first stage is where we heat up the material so our goal is to sof on the material it's a thermoplastic material it could be Perle plant poll strain or ABS materials uh in a normal uh heating process we use generally electrical heaters they are like ceramic or quartz heaters and using the radiative heat transfer mode we can just apply some uh

heat hea input on the surface of the plastic sheet using diffusion within this volume of the termal plastic sheet we can create a higher temperature material when it's hot enough we can move to the forming section in the in forming section as the image from Wikipedia shows showing to you we apply the mold the mold can be used within a CH with is the closed control chamber where we create a vacuum to support the moding operation or even in this simple example where we are creating a thop form based uh luggage you can see it

also works in an open environment as well uh in normal conditions people are not really focusing in the cooling section but during the forming stage we also need to wait uh for a couple of seconds even sometimes a minute or so just to make sure the temp temperature is below uh the critical temperature that we called uh glass temperature at the end a trimming just uh we just mum the additional excess Parts excess stuff uh from the part and the good thing about this is these parts can be recycled and start from the initial phase

of thoring where we create this uniform plastic sheet so it is it's not a waste of time it is really important H please next slide um before we proceed I would like to also give additionally short introduction to digital twins International digital twin uh was first uh introduced by Michael gaves early 2000s the idea is to create a real uh life objects copy in a virtual setting that means if I I don't know I have this mouse I would like to create a copy of this a virtual Mouse in which all of the Motions all

of the actions that I apply on this real device can be simulated in a computerized environment so uh in an nsh digital Twi is a combination of a mathematical modeling in which we might have chemistry we might have uh physics laws we have some engine mathematics and it also has a data point so we need to make sure data is also collected from the environment or the same part or same workflow same system that we are trying to investigate so model alone or data alone is generally not enough to create a digital t for that

reason I also put a section here uh to tell you guys what digital T is not really uh composer if you have a threedimensional image which is a fancy uh representation of the part there we are looking into it is not a digital twin it's just a three-dimensional uh object or if you collect sensor data if you have B bunch of sensors bunch of uh aers or some PLC data is flowing into a dashboard in which we can monitor what is going on in a factory in a shop or in a car there is also

another digital twin sometimes people call it Digital Shadow it's like a a mirror in which we know what's going on in our system but it's uh not a digital twin it's just a viral representation of the real life that is occurring in real time in a in a similar fashion a threedimensional result in which we have a a simulated uh mathematical model that is also another digital TW so we need to understand that digital T requires both a mathematical model and also data and next SL please all right so the digital T have a lot

of applications I just just put three examples here you can use a digital twin to create a optimized furnace I mean for a furnace you can maybe initially create a virtual model uh that that works for a perfect environment then by collecting data from the furnace you can just calibrate your system then it can it gives you the potential to optimize your furnace you can use a metal formic for instance in this uh here you're using a press machine to give a shape you a steel uh object and by collecting data from the press machine

you can also create a digital twin and improve your product quality or typically and most of the people probably familiar with this approach predictive maintenance can be also Incorporated with digital twns you have you have a data you have a model and using these you can create a forecast to understand where the maintenance should be really uh applied on the machine next slide please all right so back to thoring um in this webinar I'll be mostly focusing on a termal as of termal forming I know thoring is also the molding stage as well as the

triming stage but it is really uh up to the terminal uh stage of the process if you know how to uh control the behavior the heating behavior of the the termofor process then it really helps you to control the quality of the final product so the reason that we need heating is as I said before we would like to soften soften the material itself you can see one example here uh the PS material the one that is shaded with the blue color loses its strength like 1,000 F if you go from I don't know 50

degrees to 150 degrees you can see uh an additional temperature input of 100 and 100 uh under 150 degrees gives it the potential to uh give a shape into the thermoplastic material and these are taken from uh JL tor's book you can see understanding tform is very good book by the way and you can say using this softer material now you have this so-called forming window and you can see the bottom figure there you see there's a point where we heat the material and then we form the material in order to form the material we

need to make sure this uh temperatures within this whole volume of the plastic sheet is close to each other so if you know that the up top surface bottom surface the inside of the volume is close to each other we say we are within the forming window for that reason controlling the temperature input and making sure the inner section and section are almost identical or almost uniform is useful just to make sure our uh terop formic operation is really running smoothly on top of that since the shapes that we are trying to produce are not

simple shapes they a lot of complex features some some sections that needs to be pretty good that needs to good uh a good control because it's a end user product we also apply zonal heating zonal heating means some part of the plastic is heated more than the uh more more than some part of the uh thermoplastic that means we would like to have some hot zones and cold zones and that also complicates the problem a little bit further okay we can move to the next section H one example here and you can see here both

heating process with zonal heating uh in m you can see there are some very red tiles and also some soft or a little bit dark red tiles these tiles means some part of the heating process is creating a botone and some part of the heaters are creating a c Zone and a simulated uh result is like here to the simulation you can see we we saw we are seeing two sections which are hot on the left and right section and you also see a section which is colder and you might ask why do we do

I mean what is the reason to create a non-uniform temperature profile and the reason for that is when we move to the next stage we apply the mold which is a structural forming operation these hot zones are the are the location where the form where the mold is initially contacted with thermoplastic so we need to make sure where the M mold Touches for the first time on this m is softer than the remaining part of the material for that reason we we are creating a nonuniform structure and this makes the quality of the end product

uh better um yeah we can move to the next stage next slide um so this is a very uh simplified version of the heating characteristics of thermor you can see we have this uh plastic sheet in the at the center in the color and on the top and the bottom we are applying heating through heater elements the heater elements could be camic heaters quartz heaters some of them are cheaper some of them are more expensive but for quartz are more expensive but their response time is really improved compared to typical theric heaters um within the

thermoplastic which is Theme by the way I mean at most it could be like four millim 5 millim 6 mm at most even if it's thin there is also diffusion within the body that means even if we apply heat on the surface of the thermoplastic within the uh plastic sheet conduction is occurring so I don't want to go too much of details into the physical equations but what is critical here is that radiation effect on a uh plastic sheet is nonlinear because the equation itself is a power of four of temperature and the effect of

conduction is linear so inell when we do zonal heating ad as well meaning some of the heaters are hotter some of heaters are colder adding the effect of nonar radiation this problem becomes really complex so for that reason we apply modeling we apply finite element analysis we apply some integral rules to make sure the temperature on the surface can be calculated and this is the uh core of our digital twin solution uh next slide please all right I said there are heaters bunch of heaters uh the previous example we had like 200 heater at the

top of the furnace and the 200 additional heaters at the bottom of the surface that means you have 400 different heater elements that means we need to understand how this heaters really behave you see how these heaters really look when you look closer to them uh in a in a typical setting these heaters might have 500 watts or th000 watts of power depending of their uh type uh but the critical graph here is the watt versus temperature so you see if you apply a 500 watts you get a different temperature value if you apply th000

BS you get a different temperature value that is also not linear so if you're are on the low s with a low voltage uh given onto the heater element you get like I don't know 300 or 400 uh Celsius as a temperature value but if you go above or behind like close to th000 Watts you get 800 900 or 700 depending on the type of the or the manufacturer or the heating element so you need to incorporate this change into your uh digital t design as well uh if you know how the heaters are behaving

then it's just making sure all of these pieces the Lego pieces are connected to each other to create a simulated result on the right hand side for instance you are seeing one of our use cases and you can see that the blue curve on the right hand side is similar to the one on the left hand side using this you were able to create uh digital simulation of the thero thermoplastic sheet we can move to the next slide all right um as I said before in order to create a digital twin data is also essential

so far what I explained uh to you was mostly on the M matical model side so you need to understand the heating characteristics of the heating elements you need to make sure the radiation equation is added conduction equation is added the geometric dimensions and the type of the material is added but everything is in a perfect mathematical world to make sure that the digital twin is calibrated or associated with the real machine we need to collect data from the machine so what kind of data can be collected I mean we can access to the PCS

of the machine which controls the overall production cycle of the uh parts of the machine you can get sensor data and the sensor data could be the temperature value uh of the heater elements it could be the Vault voltage value of the heater elements it could put some of the pyrometers which measures the temperature of the surface of the termal plastic sheet you would like the heat but could it could also the pressure values uh not in the termal stage but in the modern stage using all of these data and collecting real uh realtime sensor

data now you have the option to calibrate your mathematical model to the real machine using the data and the data is here uh almost representative real- time data that data can be used to capture your result and real data in a single framework and that is what we call the digital twin uh we can move to the next slide all right so this is the section where I'm going to talk shortly about uh industrial use case and AR company uh it is one of the largest appliances company in Europe and general European no Deco brand

rather than AR brand Artic is mostly in Turkey so this is a really huge company they have almost like seven billion uh Euros Revenue every year they have more than 40,000 employees and as far as I know there's like thousand of them are just the R&D uh R&D Center and in this example in this webinar I'll be focusing on uh arix refrigerator plant so this refrigerator plant is located in uh in a city called eser and it is one of the largest refrigerator plant in the area as well as AR largest refrigerator uh Factory this

Factory uh uh is using almost 50% of all their electricity for terop forming process you see if you can improve the ter offorming process you're going to not just have the company to control their cost it'll also uh support the sustainability goal of the company because if you can reduce the amount of electricity usage you can also reduce your carbon footprint goal and can we move to the next slide please all right oh can you go back y sorry no problem appre so in total uh aric has eight uh refrigerator plants and the largest one

are in in turkey and here they are producing the liners liners are the part is uh is when you when you open the refrigerator door you see this White Walls right this is a refrigerator liner all these white fs and it is produced using the thoric uh machine in total AR has 32 thoric lines and that was my last number and they uh acquired a couple of new factors probably this number is even higher now uh but what is critical here is that they they are using almost 20,000 tons of plastics every year that is

tons right that means you use a lot of no problem you use a lot of raw material right if your defects can be controlled then you reduce your use of plastics if you can control the heating process then you can control your electricity right and here we did some early test with arix uh one of one of arix thoric line and you see on the left we collected data from the P we also created an environment in which we can simulate how the temperature is occurring uh in our model and you see 9 5% accuracy

is achieved using our digital twin where we created a calibrated uh system uh you see the blue curves are what we are simulating on the left hand side and this reverse triangular shapes of the measurements right uh this is very important because uh at first I mean we believe in ourselves and so our initial model was like 90% accurate using the data we collected from the uh thoric line we calibrated the heater elements remember I was talking about the heater element curve and here Pablo can I ask you to go back a couple of slides

I'm going to tell you when to stop one more please one more yeah here okay heater heat yeah this one remember this curve so this curve on the left hand side has a particular shape it's like a parabolic shape from the Y AIS point of view and if you can calibrate the critical values on this curve you have the option to control the temperature Distribution on the heater array that also means the temperature Distribution on the thermoplastic sheet using our data uh from the machine we were able to capture the behavior of the heater elements

and using this input our digital T created a couple of iterations and at the end a calibrated model just for that production line was able to be created now you can go to the slide your left yeah no no like 17 I guess it plus 17 yeah perfect uh using that uh we were able bble to create a characteristic behavior for the heater elements so at the bottom center you see a couple of uh couple of blocks like 200 at the top 200 at the bottom which I'm calling temperature top heater and bom meter these

are calibrating using the data we collected from the machine so we initially assume a curve a characteristic curve we did our radiation conduction finite element assessment just to create the initial result then we created a uh feedback loop uh here we optimized our output based on the input from the P of the machine and at the end the right figure is where we get a two dimensional image uh of the real piece which is 95% occurring at the point of the sensor data and this is really important I mean so far what you are seeing

here on the right hand side is not a infrared camera the data it's not a sensor data it's just direct output of the digital train and using this we can now do a lot of stuff I mean if you have a different targeted temperature profile that would that you like to achieve you can create a reverse heat transfer problem to assign different values to different heater elements or let's say one of the heaters is broken or maybe a couple of them is broken and you need to change them you see it is is very hard

uh to stop a production line because you need to create refrigerator right here digital twing can help you to compensate the error if there is a defect on the machine and one of the heater elements are broken the neighboring heater elements can also be used to compensate the missing termal input so you're going to increase this temperature for the neighboring elements and at the end can get still the same temperature output on the machine with an expense of maybe more electri electricity usage but you're going to keep the production line running without stopping it for

maintenance all right I have just one last slide and then we can maybe move to questions if there are any right so this is just the teral side uh of the ter offorming process in reality the material is important because this is a thermoplastic material and if it softened it can be hyper elastic SL visco elastic so it is unfortunate also a non nonlinear material because of its polymer content so material Behavior can be also incorporate in the digital T especially for the molding stage where we where we really give the final shape of our

designed product so in molding State the pressure sensors can be in incorporate in digital twin because using the pressure values we can control how much vacuum we creating how much bubbling we can create on the System since now we have an understanding to improve the molding you can go to your initial uh product design and maybe even improve your mold design right as I said before cooling is also an essential part of the molding stage because we would like to redu the temperature below a glass temperature so the shape that we give into the product

stays the same and there using a special uh cooling channels you have the option to reduce the cooling time as well and finally I told I talk about it a couple of times during the webinar sustainability aspect is also important when we reduce defect we have the company to control their cost we have the company to reduce their raw material and electricity usage and there is consumption and that has a direct implication on reducing carbon foodprint as well so um this is all from my side uh for today uh I would like to thank you

for listening and for being here if you guys have any questions you can maybe I have on the next slide I guess there should be yeah this one yeah perfect if you have any question you can always reach me from my email or maybe from my L page uh SAR uh thanks for listening are there any questions okay um yeah I have one question if I may sure um what could be some of the limitations on the on the application of these digital twins uh in your case for for the thermal forming uh activities what

are some of these of these challenges that you've been facing yeah um thanks a good question so when we created a digital twin we are also creating a copy of the machine itself so you see I said adust uh Incorporated heating elements on the machine but in reality the machine has a lot of components it's has a steel frame and steel means it also gets heat up when we uh use termal input on the thermoplastics that part is not really added into the digital TN if you see that your answer if if you see that

the accuracy is not the 90 95% level then you might consider adding more details of the machine maybe some of the structural elements can be added into your model and you can can also assume they're also heated by the this heating array and that might be an option to improve your model uh and second uh trick is that the material that you're applying heat input is not always pure because of adding some recycled elements because of the variation of uh the raw material supplier uh the material properties vary that means density is is different the

conductivity is different also the surface finish is different so surface finish is really important because in radiation their reflection absorption so that those values are also critical uh as well as tolerances might be different you are working with a really rectangular thin sheet and it has a certain thickness but this thickness might have some tolerances uh in certaint directions and in your model since you cannot measure it you're just going to assume that the thickness level is somewhat constant and all of these contributing factors will probably prevent you to get 100% accuracy but 90 95%

accuracy is is not a big challenge you can always use this as an input because the alternative is no no model that means zero% accuracy okay thank you very much that was very insightful any other question from the audience okay oh yeah p p go ahead I'm going to allow the microphone sure hello hi thank you for the presentation I was just want to ask you about the inspection uh part of the process is the when you talk about the optimization of the process uh do you use the um uh data from the inspection as

the result of inspection because you can get different type of defects and sure and do you classify those defects and um because you said also that there are different types of materials um density and so on so different effects occurs from different reasons not because you changeed something in the process so if you could yeah sure sure I very good question take you P um so remember I was telling you that we are in the middle of the project and then in the remaining months we going to be focusing on the defects identification so there

are bunch of defects there are sometimes coloring on some section of the part there are some tears and on the uh part after the molding process uh you you also see uh some hot zones and cold zones that are in unexpected regions so you you want some area to be warmer but you see it's getting colder or vice versa and these inputs are really useful because example if you say you see a tiar on the surface after the molding process that means that section is thinner than you anticipated right and thinner means it gets elongated

more than anticipated that also means it is warmer than expected because remember when you walk on in the street in the summertime and you accidentally uh press on a gum and the gum gets longer and longer and longer because of the temperature difference but if you do the same on a winter time you don't see the gum stretching right in a similar fashion if you see such defects it is a a candidate uh I don't know value or it can could be a clue for you that the temperature was not really properly defined and for

some coloring stuff some black zones or other colors are typically material related these are not theault of of the term of uh termal part but it's also an input uh uh for the machine operator and in the remaining part of this project we're going to be Contin with semens semens is applying their uh infrared camera input and serere Camera inputs and using their data collected from the machine you're going to have some defect idea hey we have some defects here we have some cold zones here we have some warm zones here and these will be

an input to me because I know in a pure good running system that temperature shouldn't be occur I'm going to go back and say hey this heating element is not working fun properly so that's an insight for me to go back and fix the problem um but yeah of course defects are very important and associating some defects to some of the reasoning is also another part of our project is it uh what you expect yes thank you very much I look forward for your project I will thank you I appreciate it um yeah back to

you Pablo yeah no I raised my hands I think by accident I I don't have any more questions that I'm sorry I'm really sorry no problem no problem sorry okay so with that uh I think that is all uh thank you very much for your webinar it was really interesting sure and well we'll see you around thank you everyone for attending and see you next time yeah thank you guys thank you goodby a good day bye