Hello, everyone. Thank you for joining us today for this webinar titled Reinventing Quality and Efficiency in Aviation The Technology That Learned to Control It is brought to you by Composites World and presented by Science Expert. My name is Ginger Gartner and I am Senior Technical Editor at Composites World. In this exclusive interview style webinar, since Expert Managing Director Doctor Alexander will sit down with Jorge Blanco Fernandez, R&D engineer at Insa, the Spanish entity of Hestia, which is actually an Airbus company that focuses on MDT and inspection. They will uncover the story of how dielectric analysis, or DEA, wants a lab based measurement tool, was transformed into a comprehensive solution for process control in composites manufacturing. Discover how Testa, which is constantly seeking new technologies to improve manufacturing processes, chose dielectric measurement for its potential only to encounter the challenges of complex data analysis and hazards, expert has worked to provide a solution turning dielectric technology into a scalable system with automated data analysis, real time process adjustments, and in mold material behavior insights that directly connect to quality standards. Our agenda today will include a discussion of dielectric measurement technology, how to comply with industry regulations, the transformation of this technology to a full solution, using the sense expert for process control, and the benefits and impact on aviation manufacturing. Our presenters today are Doctor Alexander and George Blanco Fernandez. Uh, Doctor Gupta is a polymer expert who, after studying physics, started his career at the Fraunhofer Institute for Chemical Technology in Augsburg, Germany, in the Department of Thermo Physical and Chemical Analysis and Rheology. He then began working the analyzing and testing business of the next group. With over ten years of experience in R&D, sales and general management. He founded since expert in 2021, where he is running net, his first corporate venture, as Managing Director and CTO, Jorge Blanco Fernandez began his career in 2017 and jigs and tools maintenance at the Airbus factory, which produces composite wing and fuselage structures for the A350. He then worked for the engineering firm City Engineers and the research center in Dia materials, before again working for Airbus, this time on the A330 mrt production team in Getafe, Spain. In 2021, Jorge was asked to work on a project at Airbus on behalf of Nsia, the Spanish entity within testing. The main goal of that project was to achieve composites cure monitoring with dielectric sensors, which were designed by. Since expert Jorge continues to work with Insead as a research and technology engineer for composite manufacturing processes, and has a master's degree in Aeronautical Engineering from the Polytechnic University of Madrid. Before we begin, I'd like to remind you to please submit your questions by typing them into the Q&A pane on the control panel at the right of your screen. We will answer these during our Q&A session at the end of the presentation. And now I'm going to hand it over to Alex and Jorge. -Thank you very much, Ginger. Thank you for the nice introduction. And thank you for Composites World for hosting us. Thank you for being part of this webinar. I think the audience will have a very nice hour for the evening in Europe and for the morning before lunchtime in the US, where we will explain how we manage to get a measurement technology with a high potential into a real solution that can help the composite manufacturing innovation to reach new levels of efficiency. Jorge, thank you for being here. And my first question for you is what was the reason behind that? Airbus has started to watch out for sensor technologies to monitor the material behavior during production many, many years ago. Even if Airbus has been working in composites industries and manufacturing for decades already. Mhm. Yeah, I will say definitely that the main reason could be the increasing of efficiency in production, but of course ensuring the quality, as you may know, in the aircraft industry we have always in mind the, the safety and air. Airbus as other manufacturers, has the highest safety standards in the quality assurance of components. Um, and this safety is based on some protocols, uh, relying on materials data sheets by suppliers. Uh, but the experience shows that sometimes these safety times to ensure the quality of composite components could be lower, uh, a little bit, when having the chance to look into the material behavior, uh, during curing and, and having a reliable solution for doing so. So Airbus for many, many years have, uh, have been looking for this kind of solution, to look inside the behavior of the preparation or resins when they are curing. Yes to how can we reduce the manufacturing times without affecting the quality? And of course the safety of the product of the parts manufacturer by Airbus. Yeah. Great. Thank you. Thank you for this insight. Another question that comes to my mind, uh, when talking about this topic, because what we observe since many, many years when talking to customers and working together with them is they historically grown, they relied on the material datasheets the material suppliers are supplying. But in from experience, we see that customers are not sticking to this, um, cure cycles that are mentioned in the material data sheets. Sometimes, yes, but in most of the cases, not because processes have been changed, the manufacturing environment has been changed and therefore customers have different manufacturing environments and therefore they change the Q cycles by themselves. Is this also a topic why Airbus is looking into getting insights from the material behavior to really ensure that the component is really cured when not sticking to, or when not being able to stick to the material data sheet anymore? Yeah. That's it. That that could be a very, very good summary. Yeah. -Okay, great. Thank you. Yeah. Then second question. Why did you decide for working with the dielectric analysis. So, um, I guess you have looked into other technologies as well. What was the decision that, um, made the way that you decided for dielectric analysis? -Uh, I, I will say that the main reason behind we select the, um, your technology was your capability of measuring, uh, the killing of thermal sets. And, of course, we also show that you have implemented your technology in other fields or in other industries, and that was something very good for us, and that was very attractive because we wanted something that could be implemented in, in manufacturing, not only in labs. So I will say that the main reason is your capability to to of measuring the curing of thermostats, and after that that you have a little bit of experience implementing that technology in, in the manufacturing process of other industries. -Great. And the next question. There are other technologies as well that are capable of measuring the curing. What was the reason in terms of sensors or measurement principle that made the decision for for the day? Besides the experience we have with the network and the knowledge we have since? Yeah, now it's 2024, since more than 25 years within the next group. So what was the decision for? For the dielectric sensors? Um, above others? Um, well, besides the possibility to look into the material behavior, uh, I think it will be the, the easy integration of those sensors into two links. Okay. And the fact that the Da can look through the component thickness up to a specific limit, and that is given by the component layer and the robustness with which the sensor can be manufactured. -Okay. Great. Thank you. You're welcome. So something else okay. Uh, no, I think that's everything I will say. Uh, I think that, um, the your capability of not or not only measure the, the resin, but also the bricks, because you know that the effect of carbon fiber is important. And when we are talking about the electric sensors and that that could be another benefit from your from your sensors and is the capability of not just measuring the resin, but also fabrics with the carbon fiber on it. Yeah, yeah, yeah. That's true, that's true. Um, there I can tell a story. Um, it wasn't planned, so. But, um, because you mentioned that, uh, to even measure the carbon fibers for me, it's, it's standard at the moment. But when I started my journey with the dielectric analysis, it was not standard. So when I started to work with dielectric analysis back in 2012 at the Fraunhofer Institute in Augsburg, there we even had to coat the sensors, whether, uh, dry fabric or whatever, where the resin can flow through. But the sensor was not allowed to get in contact with electrically conductive fillers like carbon fibers. And we made our way. So that time I started a collaboration with analyzing and testing already to coat the sensor and to give the coating the possibility to really get in touch with the material to be mechanical resistive so that the sensor can really be treated like the entire mode. So you can dry eyes clean. You can use soft metals to clean if some material is sticking on top. And yeah, exactly. That was the way I was directly involved with and we really made it happen. After I joined niche analyzing and testing in 2016, where we then really developed the first series manufacturing coated sensors, um, we, I think we launched them in 2018, if I'm right. Yes. Uh, I think. -So, because before that we use like a glass fiber to to. -Yeah, exactly. To measure the carbon fiber. -Yeah. Exactly. And I think that's still, still one of the USPS we have that we have a sensor technology that really can get in touch with. Um, yeah. Electrically conductive. Uh, within the resin, we even, we even have use cases where there are just 7 to 9% of resin inside and natural stone and stuff, and there is a lot of water inside and you can still see the behavior there. Great. But now we have talked about the technology and why the decision has been made. Now for the audience, I will jump back a little bit and will explain the dielectric measurement principles. So what is really happening within. So you can see here on this picture it's a carbon fiber component. And what we do is we can look into the microscopic mobility of a material. So what we are using is that the polymer from um from the properties they have partially or they are partially charged, they have partial charges. So the carbon carbon backbone and the side groups around, they define which kind of polymer do we have in front of us. If it's an epoxy, if it's a polyurethane or whatever. And these side groups are on the carbon carbon backbone, are usually partially charged. And in addition, smaller molecule chains that are not connected to each other, they even have partial charges. And exactly these charges. This is what we use for measuring. So we penetrate the polymer with an alternating electric field. And we give therefore the dipoles the possibility to orient with the electric field, but also the smaller molecules with the partial charge they can move with the electric field. And this mobility is what we measure. And this mobility is decreasing as soon as a polymer is solidifying or curing. And this gives us the direct insight into the micro molecule mobility. And we can relate this information to the behavior of a material that even means while doing so, while monitoring with the dielectric analysis in inside manufacturing processes, we can detect quality criteria like degree of cure, glass transition temperature and the flow behavior. But there are even quality criterias customers have that are going into the direction of component geometry. So instead I mentioned a stability given or the mechanical performance like the E modulus and many other things. So we have proven in specific industries already that what we can see from the chemical physical behavior in process can be directly related to quality criteria of our customers. Since what we see is the chemical basis for everything else of the polymer. And to give you an insight, what does this mean? What do we really measure? And what you can see here is the manufacturing of composite structures. So here you can see every single colored curve represents a manufactured component. And here you can see you have three process steps that are used in here. And let's start on the top right of this plot. Here you can see that all the curves come together and look like identical in process. Step three at the end, this means the manufacturing protocol in this manufacturing environment works very well. So every single component meets the requirements of degree of cure in this case. But what we can also see is that they are all coming together at a specific point in time, earlier than the process ends. And here, while having the insight and while even taking into account what happens here and process step one where we see a deviation, and here it's a huge deviation in terms of degree of degree of cure between best and worst part. In this case, we can see that while having an insight into the manufacturing, we can cut down cycle times. And here in this case for this entire cycle along process. Step 1 to 3. We were capable of reducing the cycle time by 30%, just while having a chance to look into the material behavior, and then to send the signal to the machine that the component meets the requirements. And yeah, this is a huge impact we have here. And that's our philosophy as well. So we have highly sophisticated manufacturing lines and machines. And I mean that is what especially European Europe stands for. Um, to create very great manufacturing machines. But now it's the point in time with all the regulations we even see in the European Union, with the EU green law and stuff, that we need to reduce the energy consumption and that we need to reduce the material usage. And therefore now it's the point where the material behaviour, the hidden potential that is within the material comes into the foreground, and that is where we can help our customers with and that we even help Duhok and Intesa and Airbus behind to optimise the manufacturing cycles. So now a very, very interesting topic. I think, um, for our audience. How has the Da principle being approved a test Dia. So what were the steps and which kind of measurements have been done. Because the standards is um, you usually determine degree of cure like written down in the standards with the differential scanning calorimeter And the transition temperature was a dynamic mechanical analyzer. So what have been the steps to verify that what the Da tells you in process meets these quality criteria you have set with an Airbus and you are using for many, many years now. Mhm. Well, uh, I will say that first we started with a huge series of measurements with simple geometries because at the beginning we just wanted to see how the this technology works and how can we measure the, the electric properties of the resins. And we started with just pure resin in an oven with no brakes. And we also tested some samples of these plates with DMA and DSC. So so what we did is we prepare some test, we measure the the electric properties with your technology in that in those samples made of resin. And we compare the results uh obtained during the monitoring with the results given with the final part, and that we obtained doing a DMA and DSC, DSC to evaluate the glass transition temperature and the degree of cure of these samples. Uh, this gave us the chance to correlate the measurements to the existing quality standards in this expert team. Uh, what they did is they build up a machine learning model, and that is in charge of calculating the glass transition temperature and the degree of cure in real time during manufacturing, and furthermore, can predict the time of the required end values during the current cycle cycle. I mean, uh, we can with this method, we could compare the results given by the machine learning from expert to the with the values obtained by the DMA and DSC of those of of of those samples that were cured, and the predictions of sex, uh, have been validated as well, uh, through another measurement series. And we found that we were able to evaluate the critical parameters, the glass transition temperature and the degree of cure, uh, with adequate, uh, accuracy during the manufacturing. So, um, what we did was to, uh, prepare some samples. And once we stop the cycle in different parts, we compare the results as well, uh, with, um, with those samples and the ones that are given by the machine learning program or your machine learning software, and not only with, um, uh, finished parts, with the whole curing process or the, or the curing process finished. But also we establish some stopping points, uh, to see what was the accuracy of your model during all the curing process. So we did like different tests Contests with all the huge or with all the process. Sorry. And, uh, stopping the process in intermediate points and measuring as well the curing degree and the, sorry, the degree of cure and the glass transition temperature on those samples. -Great. Thank you for this insight. So it was a really extensive measurement series to prove everything. And I think we can tell the audience that we are working together since 2016. And there have been some years in between now. Um, and so bringing the technology really into the manufacturing environment. Before I ask my second question. Um, there was a question from Nitin Gupta coming in. Um, I see it in front of me. Uh, coming back to the slide I showed before. Um, how did we manage to reduce the cycle time by 30%? Um, we really looked to the degree of cure because we found out if the degree of cure meets the requirements, we can stop the cycle, because then even the mechanical performance, um, meets the requirements of the components. So here in this case, we really look just at the degree of Q and glass transition temperature. So to both parameters. And they are within the specification of Airbus. And that is how we made it happen in real time to stop the cycle earlier and to save up to 30% of cycle time. That is what we did there. Then a very important question that some of our customers raised from time to time, especially in aviation and aerospace, but also in the defence industry. How did you manage to implement the sensor technology into the series manufacturing and to agree with all the regulations? So that is a question a lot of our customers raised, because they see the regulations they are sticking to and they need to fulfil. How did you manage to bring this technology with other regulations into the manufacturing environment? Well, yeah, I think that that is a very good question because it's a big a big part or is the most important part for the audience. Audience, I think, and is that we need to know how to go from the lab or the laboratory to the real workshop. And well, once the process was validated with those tests that I mentioned before, and we need to test the equipment in a real situation. So we look for a port manufactory in Airbus that was simple, with an accessible manufacturing process with no high production rates, because we didn't want to impact on the deliveries of of Airbus, of course. And and we need, of course, a part that was made of the same material that we validated in the prior test. So we decided to test it in the A350 section 19 beams. It's a part of manufacturing. These parts are made of M6, a very well known resin in nervous because we have been using it for decades now. And, and and of course this this part is manufactured using RTM process. That means that it's not a cure in the autoclave, but is it is done in a pressure, in a curing pressure or in a curing station. And this is important because, uh, we didn't want to test it in pre-press because the sensors are not 35 by the EASA as airworthy. And they create like a little bit, uh, prints in the part. So, uh, to avoid that, uh, we wanted to, to study it first in resin before, uh, before pre-press. So we installed two sensors into different tools, and we purchase ADR equipment to you to expert to create a temporal setup. And we obtained the Da values of the resin in a real manufacturing process, not only in the laboratory, but also in the workshop, and we sent those data to expert to improve the actual machine learning software. And we also did some DMA and DSC to the parts that were manufacturing in those in that Kiruna station, and we compare the results given by the expert machine learning with the with the ones that were given by the DMA and DSC. All of that was processed by M.P. Department from Airbus materials and processes from Airbus. And with that, we could assure that we are obtaining the degree of cure and the glass transition temperature that was necessary, uh, or that was described in the Airbus process in that curing station for that part in particular. And and with the collaboration of expert and in Mapa, another Spanish company and of course from Austria, we we could, uh, implement that, uh, that technology in the current situation, and we could use the current degree as a such a such a, um, a trigger signal trigger and to start and stop the process. So that was, uh, the process that we follow the, the main part for to avoid the is a part, we could say that we decided to use the um, the, the, the Da sensors, uh, just measuring the resin that is not in contact with the real part. I mean, we side flow. Something like a side flow. Yeah. -That's it. In a side flow, we. So we could measure the process in the resin but not in contact with the part. Just in the same tool, but not in the, in the part that is going to be on, on the aircraft in the future. -Yeah. Great. Thank you for this insight. I think that's that's a very interesting topic for the audience to hear. here. How did you manage to bring this in? You already mentioned machine learning. Um, this brings me to the next question. Um, so because how can we calculate the degree of cure, how can we manage, um, to, to translate the signals that we measure into the quality criteria? So now we have talked about the pure measurement technology, what the Da is capable of and how can we help you as testers and even your mother company Airbus, to improve efficiency in manufacturing. But coming from the pure Da technology and that is related to the naming of the webinar. So the technology that learned to control what was still missing, to bring the technology, even if you have validated that the sensors are capable of calculating the degree of cure in class translation temperature, um, adequate to your standards. So what was still missing to predict technology into the series manufacturing? Hmm. -Well, uh, I will say that after validating the capability of the de da, It still was a pure measurement technology, but we needed a solution that can be integrated into our manufacturing and that is able to communicate, to communicate with our existing manufacturing lines towards process automated automation. Sorry. So I will say that a that is one of the main points. And the other one, uh, we will say that, uh, the capability of measure the, the degree of cure in big thicknesses or great thicknesses in imperfection, I will say, because we have a very good penetration. But when we are talking about such a very good or, sorry, very big, uh, aircraft parts, maybe we will need a little bit more of, uh, penetration in the signal of the Da, uh, sensors. So I will say that those are the two points. Uh, they they are a little bit of. I will not say weaknesses. I will say opportunities. Yeah for sure. -Yeah, yeah, I would say that for sure. And, um, we made some, some kind of project like this in the past together already with another, uh, uh, Spanish supplier of Airbus, where we created bigger sensors and we made it happen to look through 25mm of dry fiber mats, carbon fiber mats. Um, the sensors have been integrated into a carbon fiber tool, and, um, we could see the flow front arriving on the top layer. So the position that was most far away from the sensor, then we could see the, the set, um, uh, flow of the material through thickness or the impregnation. And then we could see when the material arrived at the sensor surface and then we could follow the curing. So there are, uh, technological possibilities to make that happen. Um, but always we always have to decide what is the best benefit for you with the lowest effort, um, that we can spend on there. So that's always the the balance we have to find. Great. Thank you. You're welcome then. Based on the measurement technology. So how are we really capable of implementing the setup? And how can we manage to communicate with the machine and to control the machine based on the material behavior? So what are we doing here? In this case, let's imagine we have an autoclave or maybe even a press or whatever. So let's stick with the the autoclave in here. Then with the autoclave we have our sensor expert device. The blue box is more or less the brain of our technology. So here we have the communication with the machine control system. We have the communication with the dielectric sensors. But we can even implement third party sensors like pressure temperature or whatever or even humidity if you have that within your manufacturing. So we can put all the the data into the blue box here. And on top we put in the kinetic behavior of a material. So even material science that means we have three data sources coming together. We have the machine parameters and the profiles, for example, the temperature profiles that are being used. We have the sensor information for the material insights. And we have the material science coming from the lab. So what you can find in some material data sheets where you see degree of cure or viscosity behavior based on the cure cycle you're using. So this is the chemical physical basis. But this chemical physical basis from the lab is not capable of taking into account all the variations that can occur on the material side because of storing, um, freezing the material, then taking it out from the fridge again, then freezing it in. Um, even if you, um, especially since George has mentioned the material item six, um, that has been used for decades. There is um, ATM six can be stored at room temperature for two weeks, and we will result in the same mechanical performance of ATM six. That's totally true. Um, but the way to get to the same degree of cure. If the material is taken out of the fridge and being used for production on day one after taking it out, or on day 14, that's different. So the way to get to the same degree of cure and the mechanical performance, this is different. So the behaviour is changing and we can make use of this deviation here. And we can cut down the cycle times while ensuring the component quality in this case. To do so our package is separated into two levels. One level is what is going on on the shop floor and the other one is even our cloud connection. What is happening there? I come to the point in a few sentences. What is happening on the shop floor is that we have this blue box installed at the machine, communicating with the sensors, having a materials database on this blue box, and communicating with the machine. This gives us the possibility for every single component to have a digital process map, and to follow the dependency between machine parameters, manufacturing cycle and the material behavior. And even if you have more information about the lifetime and the storage of the material, then we can put it in to our our machine learning models as well, to be more precise towards the current state of production. In addition, George has mentioned machine learning models already. What we are doing is we collect the three data sources machine parameters, material insights from the sensors and the material, database data. So the material science into machine learning models. We put everything in and the machine learning models learn the dependencies. Why are we doing so? Because we as humans we are restricted. We can deal maybe with one, 2 or 3 parameter variations, and we can understand what's the result in terms of mechanical performance, quality of the component. But if we talk about at the end 2025 parameter variations, even if we think then about aging of material and the influence of storage and material deviations. aviation's best batch, for example. Then it's out of our scope. We as humans are not capable of them. Bridge the gap in between all the different variations. But that is why we use machine learning models. Because machine learning models. So statistical methods from mathematics, they are capable of identifying these relationships. And they can easily tell us what's the dependency and how to interact if we see deviations there. In addition, the cloud environment, besides the storage capability and visualization for our customers whenever they want and wherever they are, the machine learning, retraining, AI core. That's a very important topic for us, because if we think about machine learning and what we do in in terms of quality control within the manufacturing, we cannot know every failure mode that can occur on the customer side. We can learn some of them in an early stage. If we produce the first components with our customers together for a specific time. But from time to time, there will be failure modes that are not known by our system and therefore the cloud environment is very important because from there the customers can give us the direct feedback if it was an okay pod or not okay pod, and if it was not okay pod even, and that would be very beneficial. But it's not always necessary to give us the information. What exactly was the failure? Because then our AI retraining core is starting to calculate new with the feedback from the customer. So the customer is really getting a part of our solution. And then we have a new machine learning model that knows the additional information and can be sent down to the shop floor solution to improve the efficiency of our solution within the manufacturing. And the benefit to show you what's then the benefit for testing are here, in this case, for the components we have manufactured. Then you can see this is an excerpt of a specific manufacturing cycle we have done together with test here in Spain and what our model is doing in the background. This is what you can see on the bottom right in this green and pink plot. Here what you can see and that comes there. The world comes together between what George has mentioned at the beginning, with all the trials that have been done to stop components in the intermediate curing phase and then to quantify degree of cure and glass transition temperature. This is how these ground truth curves have been created that you can see here. So the blue and the purple one, the dark purple one. This is the ground truth. These are the curves that have been collected together with test here. When we have stopped components at specific times, and then we have evaluated the degree of cure and glass transition temperature, then we have calculated the machine learning model. And from there, the machine learning models are even capable of predicting the future. So based on the data that are already coming in. So the green area you can see in the plot that it's moving, the more data we get, the more precise the model will be, but at a specific point in time. And here it's roughly between 60 to 80 minutes. You can see the precision is increasing dramatically and around 80 minutes plus, the model is already that precise that we can predict the degree of cure and glass transition of the cycle. So that also means we can, in the best case, even trigger additional logistics actions that help to increase the efficiency. But while having a look into the material behavior and calculating degree of cure and glass transition temperature in real time, we have the possibility to cut down the additional safety times. And this is what you can see on this histogram plot. On the top right here you can see the amount of components where we have specific, um, average um, cut down potentials. Uh, so saving times. And here you can see the histogram consists of 44 components that have been manufactured in a specific time frame together with test data. And we can see that coming from an initial cycle time of 117 minutes. For this cycle, we have an average cut down cycle time of 13.2%. And in this case, you can see there are components where we have no saving potential, but there are even components where we can save up to 25 minutes. This is just the hidden potential that is within the material. And that comes from the the polymer science that is natural. So not every batch of polymer is behaving the same. It's it's natural that some components are changing, that we have changes from summer to winter, even on the manufacturing side of a material supplier. But that's normal. But we have learned to take this into account and to deal with the material deviations and being able to help our customers to optimize the manufacturing cycles in this case. From here, we are at the end of Georges and my story, and I'm looking forward now to the question answer session, where he and I are keen on answering your questions that were coming up during this webinar. Yeah. Okay. Thank you very much for your insights. And thank you. Very much, Alex. And sorry, because I have I, I had some problems with my connection. So that's why my camera was turning on and turning it off constantly. And it had a little. But I expect that I have answered all your questions and I have been clear enough for the for the audience. If not, of course you can. I can repeat any other unsure or if you have any question, go ahead. And we are uh, we will be pleased to to answer you. Of course. Thank you so much, Alex and George. That was great to hear all of this, um, description and understanding of the technology. Um, I encourage our attendees to ask questions. There are no dumb questions. Um, and you can do that by entering them in the Q&A pane. Um, so one of the questions. Um, how many sensors were you actually using in the mold during, um, during this testing and in and also in the RTM part? Yeah. -I don't know. If you want, I can answer that question. Uh, we are talking about a part that is around three metres, and we just used one sensor. Uh, we saw in, uh, other tests that the difference, the temperature difference in the, in the part was not very big. So we said that if we can control the most critical part, that is always the coolest one. We could have a very good measure of the evolution of the curing of all the all the parts. And of course, we were very close to the resin part that was tested in the future to calculate or that was used by a MMP. Materials and processes. Interface to calculate necessary to measure the degree of cure and the glass transition temperature of that part. So we were very close to that part that was used to measure that. And also we were in the coolest zone of the of the beam crate. -There was there was a question quickly showing up, um, from Doctor Parrish, um, how do we evaluate the the right spot of a sensor? You mentioned that already. Um, for, for the coolest part, in this case, the coolest part of the the components doctor even ask thickest or thinnest part. I think that's also related to that because usually in the in the thickest section, um, the, the heat transfer is a little bit different to the thinnest one. So even there, I'm always a fan of not overwhelming the mold with sensors, but to place the sensors always at the spots where we know, um, they are critical ones. So the weakest spots in the mold at the component, because if they are cured and meet the requirements, then we can be sure that the rest does as well. Mhm. Yeah. Someone has done a puppy. -Yeah. Yeah I think it's my background. Um we have a young one since Sunday. Well there is, there is another question. Um, Nitin Gupta again, uh, thank you for the presentation. If I use a wireless sensor like the one you used for carbon fibers. Um, I just need to ensure we don't use a wireless sensor. It was still wired, but it wasn't in my old sensor. That can be reused several times. So the sensor remains a part of the mold. And therefore we can use it for for several thousands of components. Or if we use, for example, Idex It's sensors. So the disposable ones. Um, do you think there will be a difference in the measurements? Definitely, yes. Um, because it's sensors, you have the the finger structure, it's a printed finger structure. And they have a very, very limited penetration depth. So you scratch just on the surface. That's the big difference between the kind of sensors we have used together with Tessier. Um, where with our installed sensors we can look through a specific penetration depth. And even in the best case, look through the entire components and the sensors that are just scratching on the surface. And in addition, I think you can agree on that if we use it sensors, you always have this manual effort you need to bring in the ID sensors. Um, they can be used just one time. You don't need to have some kind of disposable sensor then remaining in your component. So there are a lot of I think a lot of um, disadvantages when exactly disadvantages when using disposable sensors wear, because that is why I'm always a fan of, and I'm always suggesting to use in mold sensors that can be used, um, several times a year. Yeah, I will say as well that maybe them I'm not sure about that because you have more knowledge about about your sensors, of course. But, uh, even the, the glass fiber that is used in the, in the, it's ones in the disposable sensors, uh, do you think that that that can affect to the measurement or not really just the structure of the of the sensor? Yeah, they will definitely affect, um, the measurement because, um, when you're using. Exactly. That's another point that you mentioned because with the enameled sensors, the permanent ones that are remaining in the mold, we have the coating on top. So, um, we don't need a glass fiber patch on top. For the high tech sensors, you need a glass fiber patch on top to separate the sensor surface from carbon fibers. So just that the resin can flow through. And that will definitely change even the heat transfer through the material, because then you don't measure the real component. You just measure the resin that was flowing through the glass fiber patch. And even this makes a difference. That's like when you when you try to analyze pure resin in the lab, and you want to translate this into a composite component with glass or carbon fibers in process, you will see the deviations because the heat transfer is simply different. Yeah. Yeah. -So are there any shortcomings and restrictions for the sensor? Can it be used with all materials? Um, all different kinds of resins and fibers. And what about different kinds of tooling? Um, metal tooling and composite tooling? Yeah. Um, both is possible. Usually the, the composite tooling are much thinner from the structure. So, um, to implement the sensors is a little bit more tricky, because then on the back side of the mold, you need to put in an additional thread or whatever that the sensor can be mounted there. But we are doing that with some customers already, so it's possible. Definitely. Um, yeah. I mean, that's it. Um, the big advantage of metal molds is that the entire mold is getting a part of the sensor. So with the metal molds, our sensor is designed in a way that every mold that is in connection with our sensor is getting a part of one of the measurement electrodes, and therefore, the electric field we are using for measurements, um, is screening a higher volume of material for composite molds? It's not the case there. Um, we have this penetration of the sensor, the specific penetration of a specific sensor depending on, um, the dimension of the sensor. So the, um, the thickness and the diameter of the sensor and, um, yeah, they would just use the penetration field inside for the composite nails. -Did you see the question on the Idex sensors? Alex? -Uh, yes. Gupta I use High-Tech sensors as the sensors are small, so we generally place it on the surface of the part and not in the middle of a part. Otherwise, the contact pads are destroyed due to the pressure. Yes. Is it okay to just have information on the surface or you have any suggestions? Um, maybe you can you can add 1 or 2 sentences to my answer. I'm when it comes to the real process optimization, to have to have information from the surface is better than having no information, that's for sure. But if you really want to get a representative information of your components, you need to ensure that you can look into the material as much as possible. Therefore, placing sensors just on top gives you the information of the surface and not of what's going on through thickness. And that is always the case. If you have, for example, a one sided tooling. So you have a tooling on the bottom, then you have a dry fiber mat on top. You do in vacuum infusion setup. Then you have the vacuum setup vacuum back on top and you measure on the top side. Then you will measure a different behavior of the material between top and bottom, because on the molding side you have a different heat transfer then on the vacuum back side. So they are just measuring on the surface is not really representative for what is happening through thickness of the components. So that is why I'm always a fan of using In-wall sensors that can look deeper into the material. Maybe you have additional suggestions to that. Yeah, a well, I just will add that, yeah, it's a very good explanation. The, the, the fact that the Idex is not measuring the evolution of the material around, along the thickness, just in the surface or with very little thickness. It's important to know how how the, the the the difference between the temperature difference between the tool and the autoclave is affecting to your part. So we when we start, uh, when we started the validation test or other kind of test, uh, we usually used two sensors, one in the tool in the tool surface and another one in the back surface in the vacuum back surface. And if the if the product was very thick, we usually put one of the sensors just in the middle of the even the of the laminate to measure the, the Dia values and the temperature just in the middle of the of the laminate. In the middle of the layer. Yeah. -For the for the tri series, for the validation. Yeah, yeah. But you but, uh, you cannot do that uh, in the series manufacturing. No, not at all. At least that you install it in a part that is going to be removed in the, you know, in the tree. In a tree. Mean or in. -Yeah. Huh? Yeah. So yeah, it's. Yes. That that is the case. For example, we used the the reusable sensors are very good, but they make a little fingerprint in the part in the fabric. So we decided to put it in a place where the carbon fiber is going to be removed after the, after the manufacturing process. I don't know the name in English. A is just a part that is going to be removed, uh, when the autoclave phase ends during the three min, uh, stage of the part. Yeah. -Yeah, yeah. So so some kind of overflow area. Yes. It's. Yeah. Or even it could be a part that's going to be drilled out later for an access hole or whatever. I mean, there are a lot of opportunities. Exactly, exactly. Yeah, exactly. Um, there is another question of Doctor Polish. Um, is there any for solution, uh, to, to select the right spot. Um. some of our customers definitely do an FDA simulation to see the temperature variation along the entire component in the mold. So, um, especially in but that's not related to the aviation industry. That's usually then related to injection molding, where we see that a lot that there are consultants, um, that perform a lot of temperature simulations of the entire process and the mold and the heating plates of the machine, where you can then really see, okay, where is the weakest temperature spot. And this is then usually the place where we would like to place the sensors for the composite industry. Um, here we are always discussing with our customers, and we even did together with Airbus to learn from their experience. So to learn from the experience of our customers, because they usually have experience, um, depending on their manufacturing environment and their component geometry to know, okay, usually here and here, these are the weakest spots. And that is then usually the place where we would like to place the sensors. So I guess the question to both of you, but I'd like to hear George's response. How easy has it been to integrate this solution into that RTM part, and how easy do you think it will be to integrate the solution into production? -Well, I will say that it's a little bit tricky because of course, in the RTM and we know we all know that there is a the resin is injected with pressure and you have to, um, be sure that you are going to install it in a place that is completely tight, that they are not going to be any leakage. And we didn't have any problem with that because our mold has several injection parts that were not used. So we decided to use those valves or do the that parts. We created um, a tool like an adapter to put the sensor inside the mold and, and, but in general terms, it was very easy because we finished, uh, at the beginning of this year, the implementation of the device in the Korean station and in we integrated that in the, in the control room, of course, as well. And, and we could uh, and we did it in just 2 or 3 days, a three days, I think with the collaboration of, of, of uh, the implementation was managed by NCA and in the electronic part was made by, by another company. And in just that part was very easy and, and the tricky part was that we started implementing that, uh, during the Covid. I don't know if I'll remember that. So, you know that the manufacturing or the production was already a stop. So we, uh, it took, uh, 1 to 2 years to collect all the data to be sure that we have a very good, or that we have the perfect setup to measure the degree of cure and the glass transition temperature in parts. So that took like two years. But once we have a very good production rate, it was very quickly and we can do it. I think that if you are going to implement it in your production, um, in your production process, the tricky part is to have a very good point to install the sensor in the mold without producing, uh, leakage or, and assuring the tightness of the of the model, of course. Yeah, yeah, yeah. And as you mentioned, the, the 2 to 3 days, that really included the entire communication between the machine control our system, then placing new cables around all the kind of things. Yeah, yeah. There is another question of Nitin Gupta. Um, if we have any setup photos, how the sensor are being implemented. I would like to come back to you, Nitin, after the webinar so that we get a direct exchange, because we definitely have some pictures that we can share where you can see sensors implemented in a mode. Do you know how this looks like? -One of the things you mentioned in the beginning, George, was that, um, you like the fact that since expert had already implemented this technology in other companies and industries? Um, Alex, can you give some more detail about that? What some of the other companies and industries are that you've already proven this technology with? Yeah. So it's not just composites. Um, yeah. We even, uh, servicing customers in the automotive industry in electronics encapsulation. Um, even in the direction of wind blades where we talk about composites again, um, so totally widespread. Um, we always like to improve composites manufacturing because the benefit is huge. So every company that uses composites in manufacturing. If they can make use of the material behavior, it's a very huge potential. Um, a very big, big amount of energy that can be saved and time. But um, depending on electronics, electronics, for example, here we are talking about insulating materials or materials that are resistive, uh, against um, humidity or acids. Um, so here in this case, um, it's even an encapsulation of electronics to shield it against all the influences. Then the technology is even being used in injection molding, um, for thermal sets, and even now for thermoplastics where even with the material deviations that are coming from the European Union pressure to use more recycled content on the thermoplastic field, we see higher deviations than we are used to when using virtual material. So even that is the case, um, when we use our technology. And then, um, there is even a nice use case, um, overseas, um, in South America, where we help a company, um, to control the manufacturing with rubber elastomers as an insulating material against high voltages. And here, um, we created two benefits. The initial target of this customer was if we can see that the component will fail the quality criteria or the quality assurance methodology they have implemented since years. Yes, we can do so. We see that in process. But we could do more. We could really save for elastomers. We could save them, um, in average 44% of the cycle time of curing of this rubber. So while making use of the material, it's a huge potential over widespread industries. And I even see and that's um, a topic where we are not in today, but I can even see a huge potential in the marine or boat manufacturing because they are they are producing big, big components and they have safety times on top like hell. So that is even an industry where we are not in, but where I see a huge potential in using such kind of technology in addition. -Did you see the question that just came up, Alex? -Yes, yes. Did I understand correct that it took two years to collect the data for the model? How many measurements are needed? So I think we need to separate this question. Yes. Yeah, definitely. Because maybe I create a misunderstanding. -Because one, one topic is so one part of this question is we needed to qualify that the technology is being capable of verifying or being verified, that it creates the same results like the existing quality standards. Yes. This process to verify this for the aviation industry, this took us some years together. So here we made a lot of trials. From the point we started. But to showcase it now with our customers today, it's like producing 20 to 50 components in a serious manufacturing environment that gives us the information about the material behavior. And from there we can already create the machine learning models. So that means if I translate this into composite manufacturing, maybe producing 50 components, maybe this takes two months or maybe a little bit more. Um, then we have enough data to create a machine learning model to optimize the manufacturing. If we talk about another industry, like injection molding, for example, then we have within one day we have enough data to create a machine learning model. So it's always depending on the industry. Yeah. And the period of time was higher because of the Covid 19. And as you may know, the aircraft industry was very affected by that. And the production rate was decreased. So the number of parts that were manufactured per month were very low. So that is why it took more time to to verify and to qualify that process and to be sure that the quality standards are the same that we had before in that, in that part or in that that stage that Alex mentioned. But that would change because if you were, say, press loading something in the cycle time was already a bit different, then it would be different, like you said, with injection molding. Yeah. Yeah. -Um, and then probably one last question before we wrap up for the day. Can you just I mean, we just started talking about the machine learning part. Um, I guess just to go back, would you, you would need to establish that machine learning based on the material system. Right. And the part I mean, that's the whole idea. Exactly, exactly. That's a very, very good topic and a good point changer because every manufacturing environment, even if two customers of us use the same resin system, the manufacturing environment is different. They use different machines, different mold geometries. Um, so every manufacturing environment of our customers is unique. Yes. We can translate a little bit of the material science that is there from the laboratory. But then to create the machine learning models is always depending on the manufacturing environment of our customers. So we train always with a new customer, a new machine learning model. So it's really unique for a specific manufacturing environment. Well, and that's the whole purpose to do it because like you said, what you discovered from years ago in niche is that even if you're doing the DSC and DMA, the resin isn't necessarily doing that when it's in the part. Exactly, exactly. Yeah. Yeah. It's just a picture of the momentum, what you can create in the lab. It's a picture of the momentum of this batch of material you're characterizing and you cannot take into account in the lab what's going on with the material during shipping, storage, whatever can happen, even in manufacturing, you can have a failure in the machine and the material sees 80 degrees. For instead of seeing it 30 minutes before injecting, maybe it sees the temperature for 1.5 hours. So the material is changing, but it doesn't mean that the material needs to be wasted. And this is where we can help to change the manufacturing capabilities. -It's a good to mention that, because that is another thing, that this has a very good potential, and is when some kind of vacuum liquid is detected in the back, for example, for bricks. I am talking now, not not RTM process is good to know. How was the process of the degree of cure and the glass transition temperature? To have a decision not to take a decision about if is necessary to keep on with the process, or is necessary to stop it, to avoid it and to see if we can relaunch another another cycle. Uh, repairing or fixing the vacuum back. And that is a pretty good thing, uh, that we can implement or that we can, um, uh, or that can be very useful for us to have that information and to take a decision based on the data that came from the part, not the external parameters like temperature or vacuum or pressure, because we have, uh, information directly from the part. -You know, one last thing I would like to bring up is that we're we're entering this new age where we want to use more bio based content and more recycled content. I would imagine that that could be really useful if you have this kind of system, because if you have your part, but yet now you have the opportunity to use, say, an epoxy that has more bio content. Well, it would be really nice to know how it's going to operate in your part. I mean, it's all great to have the lab data, but like you said, when you start changing over and to look at these new systems, it could be really nice to have this kind of data to say yes or no. Mm. Yeah. Yeah, definitely. -All the bio based materials, um, the water content is deviating from time to time. And that's a, that's a big challenge for our customers as well. Okay. -All right. Well, we're at the top of the hour. Um, I want to thank again Doctor Alexander Kaluga from science expert and Jorge Blanco Fernandez from NCA. For this webinar. We invite you to reach out to them to learn more about what they've done and where they see this technology going in the future. Um, and we also want to thank you, all of our attendees. We really appreciate that you are here with us, and we hope to see you again soon at the next Composites World Webinar. Have a great day, everybody. -Thank you. Thank you George, and thank you to all the attendees. Thank you very much and have a nice day.