Hello, and welcome to the Cemtech Live webinar. Welcome back. It's been a couple of months since I was last here. Many thanks to Jim O'Brien for hosting last month's edition. And welcome. My name is Thomas Armstrong, Managing Editor of International Cement Review, and I'm very pleased to be here for the next hour, hour and a half. Here we are three-quarters of the way through our planned webinar series, now talking about low carbon cements and low carbon cement technologies. A little introduction before we get started. Cemtech Live is organized by International Cement Review, and we're a monthly publication.
It's something that's been running for now over 35 years, and the magazines are perfect for you as a visitor, as a listener here today, to this webinar. The technology that we are covering, from the quarry to the factory gate, regular technology updates and case studies, and all the information you need to understand how to run your cement plants as effectively and efficiently as possible. We've just updated our website. We now have something like 450 issues of International Cement Review stretching back into the late 1980s, believe it or not. So there's a wealth of information, and it's all retrievable, all readable on Cemnet. So that's our little plug for this session.
Please take a look at the website. If you subscribe, you will receive your free copy of the Operations Handbook. That's the only way you can get it. So it's well worth having, especially for anyone sat in a cement plant or indeed building one. Cement Plant Operations Handbook. However, if you're on the market side of things, or you're looking to enter new markets or to understand the global cement market comprehensively, then this is your first stop.
"The Global Cement Report," now in the 16th edition, the leading statistical reference for the worldwide cement sector, profiling around 170 countries, data, 10 year database, forecasts, as well as the country profiles, where you have a comprehensive online database of plants and all the data that you need to download and analyze for your understanding of the market. So do take a look again on Cemnet for more information on the Global Cement Report. A quick word about next week's event. We'll be traveling over to Paris next week for Cemtech Europe, and we're really looking forward to this event.
It's a destination that we've wanted to hold our event in for many years, and now we're doing it with the help of amazing sponsors, including Fives. And Loic is here with us today representing Fives. And as well as an excellent conference program touching on all the areas of cement manufacturing and of course carbon reduction, we have a large exhibition and there are some of the exhibitors who will be participating in one of our biggest events in Europe, actually. Around 350 delegates are signed up already. It's going to be a great meeting point to start the autumn conference series.
But today we've got an excellent and very concise overview of three different technologies relating to low carbon cement manufacture and construction. We're going to look into the whole value chain. And later we will hear about the role of demolition concrete and how that can be transformed into its separate constituents, aggregates, the sand, and the concrete fines, and how those can then be used again and recycled. So the circular economy in action. A really interesting presentation later from Fives. We have Online Control from Switzerland, and they'll be presenting their AI control for cement mills.
Again, AI at the center of the initiatives that most cement plants now are taking to really optimize their operations throughout the process. And here we have an excellent example of that technology. But to start us off today, I'm really pleased to welcome back Pedro Ladeira, who is Sustainability Director at SET combustion. I'm going to allow you to share your slides now, Pedro. Pedro brings more than 30 years experience across technical and managerial functions, working across Latin America and cement, industrial minerals, lime, fertilizer, and supplementary cementitious material sectors.
His expertise spans advanced pyroprocessing technologies, with a particular focus on clay calcination and sustainable construction materials. He has led innovative R&D and industrial scale developments, including patented fertilizer processing technologies developed in collaboration with the University of Cambridge. Today, he leads sustainability engineering, R&D, product development, and optimization initiatives with a strong focus on advancing clay calcination technologies and their industrial applications. So, I'm going to leave it to you now, Pedro, to continue with your presentation. Welcome to the webinar. Well, thank you, Thomas, again for the opportunity.
It's always a pleasure to be here and work among friends on this path towards a low carbon cement. This presentation is about specifically on the rotary kiln. We're not going to talk about the flash calciner, its advantages, and so on. There are just two things that we have to address here. Whether or not it's possible to make a rotary kiln that is quite as efficient as a flash calciner, and how to best control color for clay in a rotary kiln. So these are the two main topics. But before we move on, just a quick overview on the company. We are a company that has been established about 40 plus years ago.
We have a worldwide presence, and you can see we are spread all over the world, and we have more than 1,000 references in different countries, in different industries, and more than 300 only in cement. We have a lot of different products we work with, from calciners to dryers to multi-channel burners to grinding and so on. And a number of references, some of the most relevant players in the market have experienced our products and technologies so far. And yeah, now going through the presentation, we have this question that it's always asked for us from most of the clients.
Why should we go for a rotary kiln if the flash calciner can be more effective in terms of heat consumption and probably quality of the product and so on? So while the flash calciner, it's normally seen as most energy effective, rotary kilns remain a very proven and well-established technology in cement industry. So these matters. This presentation is really focused on optimizing the configuration of the plant for the clay manufacturing using a rotary kiln. So what can we do in a rotary kiln to make it more efficient and almost as efficient as a flash calciner?
And I believe those are the two topics that are probably most relevant for the industry today in terms of clay calcination, how to have a rotary kiln that's efficient, and how to have a very good quality control method. So going for the configuration of a rotary kiln. This is the typical configuration. You have the rotary kiln. You should probably have a rotary cooler, and have a burning system, and you have an exhaust system that goes to a filter, and you have a feed bin that can have a deagglomerator. But it's basically that.
But what you can expect here is that you have wet material coming in, and you have the drying occurring in the first part of the kiln, and then on the second part of the kiln, you have the calcination. And you do both things in a single piece of equipment, and then you go into cooling. The problem starts here. When you have a very high moisture feed project, the kiln diameter and length have to be sizable. So start to get big, because you have to have, say, 80, probably 100-meter-long kiln just to cope with high moisture and dry a lot before you enter calcination. You typically have, of course, one source of heat.
So we have to do supply heat for calcination and drying, one source to do two things. So you may have a very important variation in the kiln production because of that. The same that happened here happened here on the RPM. You only have one control of kiln rotation, and you have only one exhaust system for gases to be treated. So all the gases going through the kiln have to be treated for all those elements. And because of the fact that you have to bring air from the cooler to the combustion in the kiln, you don't have enough air. So typically what you have to do, you have to supplement during the cooling with injection of water.
And water is something that we want to avoid using because of many reasons like the availability and the price and our environmental awareness. So this is a typical kiln with some issues that we just see here, and this is the configuration that FCT has suggested for our clients. So instead of having only one kiln and cooler, you add a dryer before the other kiln. So instead of feeding the kiln straight, you go to a dryer. And what is important here, you can see this line here. We may call it also tertiary air. You bring air from the cooler straight to the dryer. So we separate the air streams. Some go to the kiln, some go to the cooler-- to the dryer, I'm sorry.
And then you can attach a hot gas generator to add some more enthalpy to the process of drying, and then you pre-dry the material before entering the kiln. So what happens there? You have two sources of heat, which makes your life easier because you can control separately heat for the drying and heat for the calcination. It makes more sense in terms of process control and keeping the productivity high. You have two exhausts. Those gases are more nasty in terms of pollutants than the gases from the dryer. So you don't have to treat those gases for most of the NOx, SOx, and so on, while you can concentrate the PSocs, the NOxs, and VOC removal only for this stream of gas.
Well, and apart from that, and that's very relevant, the kiln, instead of being a giant piece of equipment, can be much shorter because you're replacing most of the drying to be done here by something to be done here. And you have independent RPM control of drying and calcination. So everything makes sense because you can have everything separated, and you can control easily step of the way. And because you're using drafting more air through here for the dry, you can draft more air through the cooler. And because of that, you have a more aerated, a better-ventilated cooler, and you might not have to use water to supplement.
So by separating clay drying from calcination using a dedicated dryer, the proposed configuration allows for a significantly smaller kiln while reducing specific heat consumption and volume of gases required for treatment. So as you might see here, and we put together a comparison between in the case of 1,000 tonnes a day kiln for 25% feed moisture of the clay. The kiln length in the case of the kiln alone is 98 meter long, while the kiln with the dryer, it just go down to 42 meter long for calcination. And diameter go from 5.3 down to 4. These are very relevant and important, really massive difference in all aspects, mechanically speaking, and stability and so on.
And the heat consumption improves, and the off-gas treatment is decreased because you're only treating gases from the kiln, and then you increase the aeration of the cooler. So the rotary kiln design can be optimized by a combination of separate drying, improved heat consumption, and a reduced gas volume to be treated. That was for the configuration of the kiln. Now we want to evaluate the collar control method based on the inorganic modifiers. We're going to talk about that now, and compare with the conventional methods for collar control.
So basically, what we normally see is some people opt for a collar control based on oxygen depletion, which means, for instance, to feed coal along with the feed of clay. So it can be relatively successful, but you expend a lot of energy to do that because most of the energy you are putting through the kiln feed is just lost in the exhaust of the gas, with the exhaust gas. There is a quenching. You try to, as the material reach the kiln discharge to the cooler, you're trying to quench it with water, oil, and so on. We know we don't have to describe much on that. You know that. And this is our method, which has nothing to do with that. We call it sometimes iron kidnapping.
But basically, what we have here, and I just want to concentrate in this method, is to convert magnetite into other species. We wanted to determine the best reactant to combine with the iron species. And we have a side effect, which is the improvement of the concrete properties. And you can compare that with the other solutions with pros and cons. But the important thing is try to understand that we rely on calcium. So if you provide calcium with the right chemistry, with the right fineness, and you mix it together with a raw clay, and you submit it to the right temperature in the kiln, you will form calcium ferrites. There are many forms of them, but it's basically those.
So you consume all the free iron, and then you remove the reddish hue. Actually, those ones here tend to be green or black. So green is the exact opposite of red. So you're killing the red, transforming into this grayish form because you are co-feeding calcium So just to show you how it works in practice, plus we've done with this in our toll processing unit, basically, we mix together calcium with clay. You see the calcium here, and the clay is really reddish one, and it goes through the toll facility. And as you can see here, it starts to granulate. I'm sorry. It starts to granulate. So this granulation, it's very important for us.
It's a sign that, it's a signal that the thing is starting to combine calcium with iron. And then what we got from that is this very nice grayish color. You can see here, here, and you can measure the color by the LED system, which have a very decent and gray color on cement, on the clay. This was done in the pilot plant. Now in the real life, we go for 1,000 tons a day kiln in Brazil where this was done. And there you have a very, very high iron on the clay, and you mix it together with a source of calcium. You just blend it in the correct way. You don't have to over grind it or to inter grind it. And what you get in the end is this. It looks like clinker.
And in fact, it's the same thing that happens in clinker. You start to form some liquid phase, things bind together and generate different things from that. So what you have to do in terms of the plant configuration to have the inorganic modifiers, you have to add a bin, and probably depends on the granometry, some sizing or some crushing, but it's probably all you have to do to get this into your system. There's nothing really special about applying this technology.
So, if we check on the footprint, because this is important, the CO2 footprint, because people may say, "Okay, you are adding calcium." If calcium comes from calcium carbonate, so you're, instead of reducing CO2, you are increasing CO2. Yes, we are. But actually, if you see all the technologies of color control for rotary kilns, they will increase the CO2 footprint. Some will increase more and some will increase less. If you, for instance, use petcoke in the feed of the kiln, there's going to be a release of CO2, a release of CO. If you account for all the heat consumption increase and so on, you're going to have an increase in CO2.
And if you have, you're going to have an increase in heat consumption. So you can compare it, this is a comparison we've done for one client, that this technique of the inorganic modified with EOD with a lower increase of CO2 and a lower increase in heat consumption. So the message here is, there are several solutions to calcined clay to be evaluated case by case. Sometimes you have your own system, muffle kiln that you want to convert. It might be a small kiln. By adding a dryer, it will increase the potential capacity of this muffle kiln. So we are here to support you in every step of the way.
In terms of color control, we believe that, we call it inorganic modifier, this co-feeding of calcium to the system. That's very reliable. It will increase the initial strength of the concrete and the cement. It will decrease the variability, and it's a very good technology. And above the rotary kiln design, your kilns, this can be optimized, and it can be very competitive if compared with a flash calciner unit. With that, I finish my presentation, and I thank you for your time. Thank you very much. That's a great and very concise presentation there, but really neat. Taking those two different technologies. Firstly, with the rotary kiln and rotary cooler, and then the inorganic modifiers.
A couple of questions came up, and I think you've answered them very well already regarding the color control system especially. It's really interesting to hear how effective that is. What's the source of calcium? What kind of sources are you looking at? The cheapest you may find and the one that can be readily available for the combination. For instance, if you have gypsum, gypsum is a very strong binding between calcium and sulfate. It wouldn't work. If you have a calcium carbonate, it should work. If you have calcium oxide, it should work. The other thing is it has to be fine enough.
What we try and encourage people to do is not to do by themselves, not to do in the muffle, because you need a rotary kiln to do this. You should only stir the things up together and blend in the right temperature. And if you don't know the mineralogy well enough and understand the whole phenomenon, don't do it. We can definitely support you on this path. Mm-hmm. Because it tends to, we sometimes we may oversimplify it, but there's a whole science of how to bind them together. A related question, is a clay rich in calcite? Will that work? Yeah. I'll give you the worst answer possible. It depends. Depends on the mineralogy. So it tends to be, it makes sense.
You saw there is a calcium ferrite of different sources, one to one and one to two. So it will depend how it presents and how it forms up the layers in the microstructure. But yeah, it's a good starting point. And you're referring to the differences with flash calcination. You're saying that wouldn't be appropriate for this kind of color treatment because of the mixing required. Is that right? Sorry, say it again. So flash calcination is a different process and wouldn't lend itself to this kind of color system. No, it's impossible. Yeah. It works only because we start having molten phase, and you can have molten phase or liquid phase in a flash calcine. Yeah.
And just on the output from those two different systems, is there a noticeable difference and in quality or in characteristics because of the different processes? Some people advocate that flash calcine... There's not many flash calcine in operation. So people say, and we have our pilot plant in operation, and we know they tend to be more softly burnt, which tends to be nice for the flash calcine being better. But when you apply the inorganic modifier, the calcium, let's call it, there is so many advantages. One being you'd reduce dramatically the water demand, so concrete will not suffer that much.
You increase the early strength, which makes it easier for people to accept blended cement compared to regular cement. So I wouldn't say, and some people say they're just the same, so I cannot completely differentiate. Yeah. Lots of competing views. Yeah. Okay. Well, that's fantastic. And thank you very much, Pedro, for that presentation. The slides will be distributed to everyone on this webinar, so don't worry about that, along with the recording. But it was a great presentation. Thank you very much. Thank you. Very good. Let's move on to our next presenter. I'm very pleased to welcome Michael Amrhein, who's head of online control.
Pedro, if you could unshare your slides, we can get Michael's slides up. And he specialises in industrial process control, optimisation and automation. He holds a degree in process engineering from the University of Stuttgart in Germany and a PhD from EPFL Switzerland. His expertise spans process control, chemometrics, predictive AI, and data-driven modelling, with a focus on applying advanced technologies to optimise industrial processes and enhance operational performance. He's spoken here before, so it's great to have you back, Michael. And I can see your slide, so you're ready to start now. I look forward to what you have ready to tell us. Over to you. Yeah.
Thanks, Thomas, for the introduction and for this invitation. Yep. Now the fourth time here since the last time it was back in '23. Yeah. So, yeah, this time we would like to present our newest product, actually application, soft sensor or predictive AI for control to be integrated into our autopilot systems, OBLEND and OMILL. So in the first slide, just our company, and then we're going to focus then on this application for cement mills at this time. So yeah, we're around since 1998. We are in the Lausanne area. So yeah. So our focus is on... Main focus, we are coming from industrial automation, but then our main focus is then on process analysis.
Now everybody calls it AI, but it's more, we call it interpretable AI, using then multivariate statistics, but really also then all the machine learning toolbox, if you want, and to apply to industrial systems. Okay, advanced feedback control, data-driven optimization, and then advanced modeling. So that's our main focus, and that's to increase then product quality, if possible, maximize productivity, and reduce energy consumption. So for that, we deliver then industrial PCs to be implemented and commissioned in the plant. And therefore, cement industry, we have then the OMILL for cement mills grinding system, optimizing grinding system.
Then the OBLEND for feeders adjustments, automatic ones, and then the PCM Toolbox for estimation product quality estimation, and other tasks also, yeah. Okay. Just one slide, initial slide for OMILL, but we're coming back. So that's the OMILL. We're around since 25 years, working in this area. And we have this OMILL specifically developed then for ball mills and/or combined grinding with press, roller press. And that based on adjust, for example, the separator speed, fan speed, by controlling then on finest measurements and also then controlling the filling also. And that's a self-optimizing system to increase then productivity. And then we have the OBLEND.
So there we are active in various industries here. And it's all about adjusting the feeder proportions this time, based then on all sorts of chemical measurements, for example, XRF online measurements, or then nowadays even XRD measurements, mineralogical measurements. And then we have also, we can help the customer to also have even not only achieving the quality, reducing the variability of the quality, but also reducing costs. And that we have here for multi-objective optimization toolbox in order that helps us to reformulate the quality objectives. So not only working on targets, quality targets, but reformulating in order to consider then self-optimizing raw material cost reduction.
And we have installation worldwide with our partners. Yeah, our various partners. So the PCM Toolbox. PCM stands for Process Chemometric Toolbox. It's quite a wide technology myself. I did my PhD in this field very many years ago when nobody talked about AI. It was machine learning at that time, and working a lot with chemical industry, pharmaceutical industry in that time. And it was upcoming here, near-infrared, mid-infrared at my time. And there we developed many different methods to actually explorative data analysis to get to fit dynamic models on it, endpoint detection, and so on and so forth. So many different applications. And then clearly also this predictive part of the regression.
So that's the quantitative analysis, but also classification. And these tools, classification and regression, all the different toolbox we apply in this field for the PCM Toolbox strengths for prediction of compressive strength, so online prediction of compressive strength. So now let's go then into how this looks like here. So just... Okay. This is... Sorry, there is one... No. Okay. So here we go. No, that is something I don't want. Okay. Yeah, this one. Yeah. So just a quick overview of our OBLEND cement modules. So we are here from the quarry up to then to the cement output. Here we have two systems for this in the stockpile.
So one autopilot system, which is then adjusting the feeders and then an assistant systems, if there is only just the crusher, to optimize and to get the batch objectives of the quality for the stockpiles, yeah. Which can be longitudinal or circular. Then we have the system for the raw mill, either working directly with XRF measurements or with online measurements, and the combination, the compensator, which then corrects for bias between the reference XRF measurements and the online measurements. So that's something. And then more and more upcoming is also now the kiln applications in the context of alternative fuels, especially when they are changing ash chemistry.
This is becoming interesting in order to adjust then also that the chemistry to get a percentage with the raw kiln feed. And that's, again, based on XRF measurements and others, yeah, of the clinker, for example. And then our main focus is then on the cement mill here. So let's have that here. So we have a zoom-in of our cement mill here. Okay, so what we have here now, we have more and more than XRF also for the cement mills, and the particle size distribution analysis can be also plain or other measurements, but best is clearly then particle size distribution in the lab, automated lab, or manually.
And then there is a modeling part, and then the idea is then, or the goal is to estimate these compressive strengths, various types of the two-day, for example, or 28-day, and hand that over these predictions together with prediction intervals, alarms, and everything that can happen in the real life, a real plant, hand it over to a customer OPC UA and DA server. So we call it a standalone system. And then clearly we can then integrate that with our systems or OBLANT, which then actually where we can then not only control directly on compressive strengths set points, yeah. That's the idea, yeah. To close the loop.
Use these measurements that are available every hour, for example, every two hours. All the filtering is done, everything automated in real time, gets the prediction, and then we adjust the feeders to reduce then the variability of the compressive strengths and then clearly also the next steps or if possible, then reduce the clinical factor. So for example, increase the limestone percentage. And then, also what we know when we are on, for example, limits of the feeders, we would like to adjust the feeders only in a certain range. Then, whenever we are on a limit here, we can then increase or decrease the finers.
So there is this connection to our own mill system, which then adjusts the separator fan speeds and also the total throughput. So yeah, let's have a closer look at the PCM Toolbox strengths. So just we're going to see some slides or some data, real data from this cement type here. There is clearly a first step here, phase one. The offline is building the models. There are several models, a family of models. There is a lot of insight going on. We need to generate explorative data analysis before, and data cleaning and so on and so forth. So these are real life data.
And so we have different kinds of data here, XRF, XRD, particle size distribution, that we call it as the input data X, but which are available for hourly measurements are taken right after the cement mill, before the silo. And then we have then also cumulated samples, which then are used to formulate and cast the mortar, where then after two days and 28 days, for example, we get the compressive strengths. So we have two types, and every time these are different kinds of data, which are coming in real time whenever they are available. They are not available at the same time, so there is all this treating the readiness to be handled and so on and so forth.
And then once we build then here that with this reference data, historic data, we get the models here, different kinds of models, which then are then deployed in phase two as a monitoring the predictions and then update automatically the models. So that's also a very important part that we have an automatic model update here within. And then as a optional phase three is this integration what we see before with OBLANT and OMEAL. Yeah. So yeah, so data, real data, many things can happen. So just as an example here that can, that's a two-day and 28-day compressive strengths. The blue data is from '23, '24, and then the red one, '25.
And you clearly see with the eye, there is a shift to higher strengths in 2025. And that you clearly need to update, motivate, update your model because it won't, if you use the model here from '24, you won't be able to predict for '25 because all these models are, these are database models, machine learning models, they are good for interpolation. So that's very important also, yeah. And then, so what is also important to handle, so for example here, the chemical, the XRF data has then, for example, suddenly some bumps here in '24 and then again, and so there are regime changes in the data you can observe and this you need also to handle and then motivates for automatic model update. Yeah.
So we need to compare the different approaches, challenge them, and so on, and everything is happening in phase one. We simulate then also the real output, and there is a model design approach. Okay? There is loops, there are lots of optimizations and discrete optimizations and so on. I don't want to go into it. So we have here the input data and here the reference data, and there is a whole procedure we call also the causal validation, where we simulate also the future application or challenge these models already for future application deployment. Yeah. And, so there is at least different correlations or grouping and so on, outlier removal, very important outlier removal.
So it's not only in the XRF, XRD, and particle size distribution that you have outliers, but also in the reference, even the cast, the compressive strengths can have outliers. You need to remove them already during phase one, but more importantly is even then also done to the automatic model update that this is then automated, supervised automated. So there is a lot of effort here in this phase one. And then, okay, for example, what we do is also this causal validation strategy where we have here, for example, initial training data, and then we test the future data. So the training is done only here. The model building is with the past data.
Then we say here, for example, the gap, for example, the compressive strength two, there is two and a half day later, we only test the data. And so we test, we train and test this kind of increasing data, training data, so more and more coming in or rolling window strategies for training. There are different strategies that we need to test, and we test during phase one. And then we have here, for example, automatic update, which can be triggered on every two new compressive strength measurement, or weekly or every three days or on another kind of event-driven trigger can be done this update. And so then in phase two, the deployment is then here.
So for the normal prediction, we just need to input data, XRF, XRD And PSD, for example, data, and then it's in real time. They're coming in whenever they come in, become available. Outlier detection, all these things are done here in the soft sensor. And then finally, if everything is okay, we predict the compressive strengths, give the prediction interval, and so on and so forth. And then we have this automatic model updater, yeah, whenever there are new references available. But clearly, it's with this gap and so on. So this can be also the third part with the standalone version of this PCM toolbox.
So we have here typical situation is here that all the devices are, or the measurements are already available in a database, customer database, ABB knowledge manager, and so on. And this database, so then are automatically exported into files here whenever the measurements become available. And this is then transferred into a virtual machine which is running. Okay, we supply here the virtual machine, which is running on a customer PC. So customer PC, which is already validated, that's very important in this IT security. So we supply here the virtual machine, yeah, and commission that also from remote.
So this, and then the data, all these predictions, prediction interval alarms can be then either returned to the database or make available in a OPC UA DA server for further use. Here, for example, what is important is not only supplying, here we see the blue dots. These are hourly or two hourly predictions here for the CS2. And we give always here all the prediction intervals that we set here, the 95% prediction interval, but also other intervals. And we see that it's about reliability, how important, whether we can rely on that.
And then also what we always plot with it or can be plotted is the real measurements here, sorted with the variable importance and protection, for example, which have contribute most to these predictions, so to these compressive strengths, for example, here, it's not a surprise here that CC3 isoclinologic is on the top, and then others. So we have around 36 predictors which contribute to these measurements here. And then what's important for the control applications is not so much the absolute value of the compressive strength predicted, but it's the tendency. Do we capture increasing or decreasing tendencies?
And you see here the green one, that's again the today, it's decreasing, then increasing, decreasing, and so on. And we have here then the change between, and so this is then decreasing here below zero, and then this is increasing. And what we like to have is then the blue one. That's what we're going to control on to have it on the same side. Okay. It's almost always corresponds, but not always, yeah. But anyway, and this is very important because the bias, we're going to handle it differently in control loops. But what is very important also for all sorts of sensors, by the way, online sensors and so on, to capture, give the correct direction.
And you see that it's coming every-- and clearly, the green dots came much later into the picture. So you see it immediately every hour you get that one, and the green one, you get only two and a half days later. Yeah. So we have this anticipation, this main advantage of using these measurements and to be able to work already on that. So time is going fast. So yeah. And how to now put it into the whole picture here. You have here the cement group. We have here clinker supplementary cementitious material, gypsum, limestone, and maybe iron sulfide, and maybe contrast.
And so here these measurements go into the soft sensor, and we have this prediction, and then we compare it in this module with the quality set points or limits for the compressive strengths for the specific cement type. And then we adjust these set points, which then finally Oban takes and adjust these proportions of the clinker and the other features. And whenever they are on the limit, somehow we then even give you a delta fineness measurements, to then adjust the fineness. And yes, okay, we saw it before, the PCM toolbox as standalone, but this can be then easily connected in a third phase to this industrial PC.
We deliver here a Siemens IPC, industrial PC, with a Siemens WinCC on it, software controller, and we do the connection via OPC UA. And then this can be very easily connected to customer Siemens PLCs, but directly using S7 connection or then again, via OPC server. So that's not a problem here. So just about then the grinding, the ohmmio system here. So we saw that already, and we see then from this study in phase one of PCM toolbox that actually these intervals here from 10 to 23 and 23 to 32, so both from 10 to 32, they are very much correlated to the CS2. So that gives us a very good also selection. So we would like then to control actually on this fineness interval. So it's an interval.
And that leads us here to this scheme here. And so by ohmmio then adjusts the separator speed, the fan speed in order to get these fineness intervals. And then if the fineness is okay, we then increase the total fresh feed flow rate. It's a self-optimizing system, and then also monitoring the anti-block. Yeah. So clearly, we use an electrical ear or if this is not working okay, then the minimal motor power and other quantities. And yes, for that system, for example, in Cimont Vigei, Swiss cement plant, we achieved such performance, okay, by controlling with respect to already an automation system, was able to reduce the 28 compressive strength variability by 0.7 megapascal.
Okay, sorry for the P here. And then also what is important on an annual basis was we're able to increase the production by 15%. So it's a lot, but it's always the question where we start from. But okay. But these are annual numbers. And then at the same time reducing the specific energy consumption. And that was then implemented for various other cement types and for second cement mill. And then let's see how we're going to extend that then to integrate, we saw that already before, with the PCM toolbox strengths. So we get here the particle size distribution, XRFF, XD, which then predicts the compressive strengths.
And then finally, we can then adjust these feeders with the OBLEND and also, for example, modify in order to give a set point, somebody gives the set point for this cement fineness. So this module then is there to automate that part. To add here delta SP, modify the set point for the fineness. So yep, I'm coming to the end. So let's go back here. So if there are any questions, thank you. Thank you very much, Michael. A very detailed presentation on your systems, the OBLEND and the OMILL. Really fascinating to see the predictions giving results two days in advance. I guess you can see that's a really powerful illustration of what your technology can do.
Really, a couple of questions that have come in, and they are, I guess, asking about how the system copes with variability. One of them is when incorporating highly variable cementitious materials, supplementary cementitious materials like calcined clay into the mix, how does the OBLEND predictive soft sensor adapt its algorithms to the shifting correlation between the raw mix chemical composition and the downstream compressive strength? Right. Okay. We saw already that previous plots that there are regime changes. We are detecting them. And then, either we adjust if these are then slow moves, then by model update, we can then track these changes.
If they are then coming back, if they are already done, I would say then we can then switch between models. So that's available existing models. So between model ensembles. But that clearly always, since these are machine learning and models, we need to have data. So if there is a completely new situation, we're going to see that. That's already the first step that we see that as outliers or saying, "Hey, there is a regime change. Hey, we don't give a prediction here." Yeah. So the first thing is to avoid false action. Yeah. And then once we build up then these new databases, we can then model and then do then the final prediction.
But first step is then to detect these unknown regime changes. Yeah. Yeah. So if it's an unknown regime, then the model has to be taught effectively to the new environment. It's like for any analyzer also, we have been working with so many different analyzers, online real PG&E, PF&A, and near-infrared and others, yeah. And we see that so much then if there are, for example, already stock buy changes, there is a huge change, I can tell you, and the current calibration, they don't handle that. Yeah. With real sensors, it's the same thing. Yeah. And you can imagine the soft sensors even, yeah, more.
But with experience of handling different materials, then perhaps you then have a model that you can draw on which applies to that. If it's a repeating regime, then okay. Yeah. But we need to have this somewhere in the database. And just related, I guess it's the same regarding moisture changes. If moisture is high in the raw materials and facing frequent inconsistent material flow from the feeders, again, it's a question of degree. If it's gradual, it's one thing, but- Okay. So you talk about then plugging or just the feeders that they don't follow the set points or? Yeah. Yeah. Okay. That's something else.
That doesn't concern then the prediction, but it's more about then the control and that we handle actually very nicely with our OBLEND system. That actually what it will do is then to increase to there is an offset between the set point and the real flow rate, and that can be then handled by the control loop. In order to increase, it will then even increase further the set point for this feeder, and at some point, and then it's going to pay back. It will get back the quality into target. But clearly it's a decrease in performance. Quality is adverse. Clearly, if the feeders are working not as nicely as the 1% or so, you lose performance, control, quality performance. Yeah. Very good.
Well, thank you, Michael. Michael's details there. And again, slides will be emailed out after the webinar ends. But for now, thank you very much, Michael. Yep. Thank you. So, let's move on now and across the value chain a bit further down, and we're going to have Loic Bottier from Fives. Loic brings more than 30 years of experience in engineering, project delivery, operations, and sales across the automotive, oil and gas, and cement industries. Sorry. Since joining Fives SEB in 2006, he's held senior commercial roles across the Middle East and sub-Saharan Africa, supporting major cement capacity expansion projects.
And since 2020, he's been leading the European market with a strong focus on decarbonization solutions, including energy efficient grinding systems, innovative calcination technologies, and construction materials recycling processes. So Loic, you're going to tell us about the RODAX now. If you'd like to share your slides, and we can get started. I can see them arriving. There we are. We've got them. Thank you. Over to you, Loic. All right. Okay. Thank you Thomas for the introduction and the invitation. Definitely yes, today we're going to talk about the downstream flow in the construction and dealing with low carbon cement.
I will focus on the concrete recycling and adding value to this kind of material. So I will start first with a few word about the Fives Group, which is a engineering group, strong from more than 8,050 people, 200 years of presence in the industry and evolution through the different industrial revolutions. And which give us also a strong backup from the financing point of view because the turnover and the order intake of the group is over two billion euros. And between the hundred location and 25 countries, our focus is to have long-term commitment with our customer and going all through the value chain, including the service life afterwards. So FCB is dedicated to the cement and mineral.
All our activity rely on process equipment and then comprehensive system installation integration. We are a member of some famous association in Europe and in the US mainly. And what we like to say in the cement industry, as today the focus is to decarbonize and we speak a lot about the price of cement, and I think in the background of your introduction, Thomas, you were mentioning the carbon capture, which make it very expensive. The position we have at the moment is to say that the cheapest CO2 is the one you don't emit.
And then this is an example of what you can achieve from existing cement plant or typical cement plant in Europe, which emission is a bit less below 700 kilogram of CO2 per ton of cement. With the existing technologies that we have in our portfolio, we can reduce it by almost 50%, driving on the grinding efficiency, on the alternative fuel, on the calcination. And the topic that we talk today is the demolition concrete recycling, because it also is a lever to decarbonize the cement and reduce the carbon footprint. This is encouraged by, let's say, the legal or the normative framework, both because we have, you see here on the right, some obligations in the construction industry.
This is for France. That's likely the same in all Europe. We have to reduce the carbon footprint of the construction. And so this goes through the construction materials. And we have, by chance also, the norms that have evolved with the cement norm EN 197-6, which allows the introduction of demolition concrete fines in the cement, and also the concrete standard that also accommodates with the recycled material. So either with a standard level of substitution, but also thanks to the introduction of the performance approach to have a bigger amount of aggregates or recycled aggregates provided that their quality is good enough.
And that's the lever that we want to work with and the topic of my presentation. So I'm going to talk about this machine, the Rodax, or also known today the Rodax 4D because there are four motors on it, which is a kind of vibrating crusher or grinder. And the idea is using this Rodax together with the high efficiency classifier TSW, is to have a selective grinding of the concrete waste to have high quality aggregates and high quality gravels, high quality sands, and highly concentrated cement paste in the fines. So the goal is to use this recycled aggregate in concrete, not as underlay or low-grade concrete, but high quality concrete. Same for the sand.
And the sand can be also recarbonate because it's still bear a little portion of cement paste. And the fines that are recovered can be used either as a low carbon SCM directly in cement, or it can be used as well as a raw material because then the calcium has already been decarbonized in the previous life cycle of this concrete. So it is a carbon-free material to produce clinker. You see here also some figures about the potential of CO2 capture. So there are different researches on the topic, but this figure are organized of 50 kilo of CO2 per ton of fines or 32 kilogram of CO2 per ton of sand, which remains an interesting level for certain of our partners.
So the idea is to put the materials through the Rodax, then the output of the Rodax will go in the screen, out of which we will recover the gravel fraction. And the fine fraction will then go through the classifier, from which we will recover in one side as a reject the sand and the fines as a finished product of the classifier. There are recirculation within the circuit, so they are not detailed here, but there are different way we can recirculate the material to obtain various performances or quality of the material. And this is a summary of an extensive research program we have made together with Heidelberg Material.
Where if we consider a feed composed of 45% of gravel, 20%, 21% of cement paste, and 34% of sand. When you go through the Rodax, then you have a 416 millimeters fraction, which is rich of 80% of original gravel, only 8% of cement paste and 12% of sand. So it means that it's a pretty high quality of gravels to be reused in concrete. And the other fractions of zero to two millimeters when it is classified, you have a sand with, let's say, 55% original sand in this material, 31% of gravels that you recover by size reduction, and only 14% of cement paste, which is the potential recarbonation.
And the fines, so zero to 125 micron fraction, which is rich of 65% of original cement paste, meaning that basically recover 60% of the original cement in the fines, which give to this material a very high potential, either as SCM or as raw material for clinker production. And during this campaign, we have tested the different possibilities of recirculation to assess the effect on the throughput and the quality of the different fraction we could obtain. Then the idea is to recycle 100% of the concrete waste, and then it will preserve natural resources because all this gravel and sand that recover will not be extracted for new concrete production.
You reduce the energy consumption and emission related to the extraction of this material as well. And you have ready-to-use material, which are equivalent to natural aggregates. And when we say that, we have made the test and compare the quality of this material to the standards. So in this table you have, let's say, the standard value for concrete, so in gray color. And you have in the green column the figures we obtain with the Rodax, and you see that the quality of the material is fully in line with the requirement of these norms to produce concrete, which make them usable for, let's say, high value applications.
And then we have put two picture of the two fraction of gravel that we have obtained in Heidelberg first plant. And we see from the picture that the gravels are really clean of cement paste. And if we look at the sand, we have also the same comparison. What is the requirement from natural material when you want to use it in concrete. And in the green column, you have the value of the corresponding sand that we obtain with the Rodax, which also fully meets with the requirement of the norm to produce concrete.
And on the picture here, it's an example of the sand recover with a very small amount of dust, meaning that the classifier has properly done its job to remove the very fine fraction from the sand. So we use the cement paste either as a low carbon raw material to produce clinker or to use as SCM. And as SCM, there are also different programs in FCB and outside to look the best way to activate it or to bind again some CO2 within this material. Which make it very interesting, but it is more interesting because it is very pure in cement paste.
It's not only particle size reduction, which would produce some existing material that's used only for underlay or just some, let's say, additive to cement, which would be rather close to a limestone than to a reactive material. So if we go more in the technique, the Rodax is particular because it is working as in-bed compressive grinding principle. So if you compare to conventional approaching with impact, you would break the material and the particles, so it would break the gravel into smaller gravels. It would break the material in a, let's say, a non-cubic form, which is also something that normally you look for when you make concrete.
Instead of that, by using the in-bed compressive strength, you break the material along the joints, which helps to liberate the gravels and to liberate the sand and keep the cubic form of the different components of the concrete. And if you look at compressive grinding, it is also different from a cone crusher because a cone crusher has a constant stroke, whereas with the Rodax, we have a constant force. And due to this constant force instead of constant stroke, for a given power, there is no mechanical risk on the equipment itself and no breaking of the gravels in smaller pieces. The Rodax, in fact, is not a new equipment. You see here the reference list.
We have quite a long track record, which was mainly used in the mineral grinding industries to grind very hard and very abrasive material, which make for this machine, the demolition concrete, not a particular difficulty. The innovation was in the setting up the proper circuit to properly separate the gravel, sand, and the cement paste. And here in the four top lines, we have the latest references. So one is for Heidelberg Material in Poland, where they have set up a complete recycling platform to recycle concrete from this area near Gorzow.
And then this investment has been done after a thorough program of tests to define the best parameters to operate and to recover the product that the customer wanted to obtain. And the last three references are from the Taiwan Cement Corporation group, including OLYCK and Cimpor, where they will install in their respective countries, in Taiwan, Turkey, and Portugal, three Rodax for concrete recycling. And also, this decision has been made after a testing program that we have done in our testing facility to test concrete from the different origins and then different setups. And they are very happy in the perspective to recycle concrete that way.
So my conclusion here is to present the Rodax as the solution to recycle 100% of demolition concrete and not as just grain size reduction, but selective grinding to obtain high-quality material that will allow to reduce CO2 emissions in cement and concrete production and encourage to develop the circular economy and possibly promote urban mining, because as you can produce gravels where you have demolished some buildings, you already have the raw material on site and no need to make a long-distance transportation from the mines to the ready-mix plants and their site for the waste. That's it, Thomas. I think it was quite concise. Thank you, Loic.
Yes, fantastic presentation, and really, it's such a great vision of recycling and how we can create that holy grail of the circular economy. And I've seen for myself at the R&D facility at Fives in Lille just how amazing this product, the reclaimed aggregates and sand, really do look as you show them in the pictures. In terms of the actual available market, I think from my reading, the Netherlands is more advanced in terms of concrete recycling than other parts of Europe, but it's really only just beginning now. Is that right, in your- Yes... knowledge of the market? Yeah.
It's beginning, and it's always one equation between the, let's say, the different parameters and the return on investment is always a different case. Sometimes it relies on the transportation cost, sometimes it relies on the disposal cost of the concrete waste. So each case is particular and generally is different one to the other. Yeah. I wonder if it'll be similar to alternative fuels, maybe with landfill costs and disposal costs rising, then it creates the economic driver for these kind of technologies. Exactly, yeah.
And the location of this kind of installation is also something that has to be discussed and looked carefully because it can be either close to a cement plant because you want to minimize the transportation of this fraction of the material or to put it rather at a recycling facility and then transport the fine. Each case is particular and depends also on the driver from the customer. Sure. Yeah. Lots to consider there. In terms of the recycling of the concrete fines, how will the decision be made between whether to use it as an SCM or as an input for the clinker manufacturer? I guess that's a series of trade-offs as well. Yes, exactly.
It also depends on the interest of the customer, if they want to decarbonize their clinker or if they want to increase or, let's say, decrease the clinker content in the cement. Mm-hmm. It's always depending on their, let's say, pathway to decarbonize. Yeah. For sure, you have some technical aspects because the quality of the fine that recover depends a little bit on the quality of the gravel and the sand. So definitely you could have limitation due to the silica content if you put it in the raw mix. So it's also always a balance which rely or depends on the components and the quality of your unshot material. Yeah.
But I guess there's an application here for concrete producers to use it and create their aggregate sand and SCMs. Yes. Right. And maybe completely separate- We can address both. We can address cement producer who want to add, let's say, some activity in their portfolio who already work with concrete, and they will develop the recycling. And on the other hand, you can have direct ready mix producer who will use the material as their own raw material, as the gravel. So yeah, different profile of customer. Very fascinating, and we look forward to hearing more about it and seeing how this evolves over time. An exceptional technology.
Thank you, Loic, for presenting that and to all our speakers, Pedro from FCT, Michael from Online Control, and Loic from Fives FCB, who you'll be able to meet if you come to Paris. FCB are sponsoring Cembrit Europe in Paris. We're delighted to have them on board for that event. Otherwise, we'll see you online, fourth of November for our next webinar, when we'll be focusing on energy efficiency strategies for cement plants with Optimative, CTP, Christiaan Pfeiffer, and Klüber Lubrication. Another interesting collection of presentations guaranteed for you in November. But for now, thank you to our three presenters. Thank you to everyone for logging in today.
Looking forward to seeing you in real life, IRL, as they say, or at the next webinar. Thank you very much. Okay. Thank you. Thank you. Cheers. Thank you. Bye. Bye-bye.
