Video Interview: The Role of AI and E-Nose Technology in Mitigating Risk in Industrial Settings
Seb (00:00.692)
Hi everybody, welcome to this episode of Insights by Safetytech Accelerator. My name is Seb Corby and I’m a principal consultant with SafetyTech Accelerator. In this episode, I’ll be interviewing Viktor Bezugly and Ortwin Muehr and we’ll be discussing the opportunities for AI and nanotechnology in risk reduction and early fire detection.
So, Ortwin starting with you, what was the original business problem that brought you to Safetytech Accelerator? And can you give us a background of why you were exploring new technology in the first place?
Ortwin Muehr (00:45.04)
Actually, the problem in shipping in the maritime industry is already very old to have precise alarms in regards to fire detection or detection of incidents or dangerous cargo spills or fuel oil spills on ships. The technology we use presently on board is, if you ask me, I would carefully say it’s outdated for many years. It’s an old system.
which relies on optical sensors. So I was for years already on the search to find new technologies which would be more precise. And this was actually successful when we had our first initial kickoff meeting in Singapore. It was the Safetytech Accelerator. I understood that the people meeting there had the same issues and were looking for new ways to detect such problems on board.
Seb (01:44.02)
Viktor, do you want to start by giving an introduction to your organisation and the journey that you’ve been on to get here today in terms of your technology development?
Viktor Bezugly (01:57.719)
Yeah, thank you, Seb. So we are a startup, so a relatively young company which transferred the knowledge which was collected during many years of research project at the Technical University of Dresden. Then in know -how project further transferred towards founding of the company and bringing the first products to the market. And the…
Since we were experimenting, so to say, since many years already with nanomaterials to create really a very, very sensitive platform, we finally were able to develop it towards the concept of electronic nodes. So it was basically inspired by the function of our nose, like we interact with reality. Just…
taking into account the environmental effects in the sense of smells and gas surrounding around us. Until now, it was not really well digitized technology. So we know already since modern century cameras, we know microphone for digitization of human seeing and human hearing. But concerning, so smelling, there was no digital.
or well -established digital technology. And I believe the most, the main hurdle in the past was really not suitable materials actually for detecting on a very efficient way, gas molecules or volatile molecules in the air. And on the other side, as our human nose, is a combination of…
hardware and software, so our brain analyzes the signals which we get from the nose. Then of course there was not so much developed yet in the AI side, which followed in the last decade probably. There was a really big jump in the AI development. And that all led us at the end to create really something like real electronic nose, so not just a…
Viktor Bezugly (04:26.007)
science fiction, but something which goes already towards real applications.
Seb (04:32.02)
So Talk me through kind of how is this different to gas sensing that has been on the market for, you know, 100 years. What are we talking about that’s really different and innovative here?
Viktor Bezugly (04:46.487)
I think the concept itself is quite innovative. It’s kind of orthogonal to gas sensors to some extent. So in the development of gas sensors, the point is really to detect specific gas at very high precision. So at best, very high sensitivity and minimal cross sensitivity to other gases. Because the point is through the detection of some particular component in the air, independent on what is going on around. This component is…
focus and should be reliably detected at best with already some rough knowledge of concentration. And then the other components are left back. In our case, it’s like if you have orchestra playing, you just focus on picking up one particular instrument playing, like drum. So we can hear this, but nothing else. And this is your point. You really want to pick up something single from this orchestra.
In the concept of electronic nodes, we are working with many, so, say, metrics or array of sensors. So, this is again mimicking our own nodes, like we have multiple of receptors in the node, the same in our case, there’s multiple of detectors, very sensitive detectors, but they react to everything. They react in a different way, a little bit.
And that’s why they all together create like really digitization of orchestra music. So where every instrument is heard and recorded. And then the AI actually analyzes that based on previous experiences like typical AI working. So you need to teach it first. And basically the same we do when we teach, say, a dog to find some smell, we also give…
some sample of this smell and then the brain works and can first memorize this fingerprint of particular smell and then helps to find the same fingerprint. And the same is the principle of electronic nose. You see this is very different from a gas sensor, where in a gas sensor you want to really have one component, in electronic nose you want to have all components, but you know…
Viktor Bezugly (07:11.191)
what are these components to which scenario they belong to.
Seb (07:15.764)
So the kind of metaphor that you used for an orchestra that what you’re saying is like a traditional gas sensor was looking for one specific gas. And if you wanted to detect multiple gases, you’d need multiple sensors where the carbon nanotubes in your chip effectively detect all of the molecules in the air in any given makeup. So that actually, rather than looking for one specific.
Ortwin Muehr (07:24.688)
Thank you.
Seb (07:43.86)
element or molecule or gas you’re actually able to detect a signature I guess in the air, is that right?
Viktor Bezugly (07:51.767)
Yeah, that’s completely true. And this concept, just this concept, allows us really to go to more complex scenarios than just detecting some gas. Let’s say a typical application of a gas sensor in safety is detecting some harmful gases, some leakage, for instance, of some harmful gas. Then you want, of course, to be able to detect really in any environment that there is a leak of this gas.
and lowest concentration as possible. This is really true application and this is focus. In our case, we would like, let’s say, to detect the coffee smell. Let’s say, let’s start there. And coffee smell, it’s known to consist about 120 components. However, we don’t want to know about this component’s details. We just want to see if there is a coffee smell or no coffee smell. That’s all. This is a different approach.
Ortwin Muehr (08:44.944)
If I may make one comment to this, the interesting fact is that the last fires we encountered on our vessels were all detected by nose, by the nose of a duty officer or by the nose of a seaman making a round. They were not encountered by the existing fire detection system. So this proves how important this was, Viktor is saying now.
Seb (08:46.58)
Go on, go on, Alter, you can come in.
Viktor Bezugly (09:15.383)
Thank you, Orton. This is actually a very good comment because the current fire detection system are based on smoke detecting. And smoke are actually particles and smells are molecules. So this is two different worlds, so to say. And it’s matter of fact that molecules are emitted area than the particles. Particles are emitted when there is already fire. And molecules can be emitted when it’s still…
just overheating. And also molecules have longer penetration distance. So it’s also a known fact that you can sniff from longer distance that you can see that the smoke is appearing.
Seb (10:01.236)
Okay, so the nanotubes adjust the molecules from the air. Talk me through how, I guess the exponential rise in the power of AI, talk me through how that is able to kind of unlock the use of the nanotubes within something like early fire detection. Because did you say that you’re almost developing a signature, a kind of footprint of a certain event?
So could you talk us through kind of how you use AI to enable early fire detection?
Viktor Bezugly (10:36.887)
And we were actually looking into fire prevention application of our technology since already some time. Though we were focusing as people from land, not sea and air we were focusing mainly on electric devices application. So what is actually not under control is electric cabinets, so switching boxes. So there many fires started actually there.
and there is no control, no one puts any smoke detector for instance there. And our compact, what is also important for our electronic news application, we are very compact, so it can be put in any closed box, so to say, to monitor overheating even, so even not before smoke start. And this has already got quite a big resonance, I would say, so people are really interested in applying such technology for safety, so safety…
fire safety issues in industrial environment, also for houses, it’s also quite a clear application. And we, that time, we were not looking towards a maritime application, I would say. And only thing to this story of Safety Tech Accelerator, which as I know, Ortwin initiated in this direction that let’s look into the potential application for electronic noses.
the fire safety on ships. So only this actually kind of came from two different sides and joined under the roof of a Safety Tech Accelerator and where we could go together next steps and really first of all we met each other. This already a very big success. So I think many technologies didn’t appear just because people didn’t meet each other at some point. And here it happened.
Ortwin Muehr (12:23.984)
you
Viktor Bezugly (12:33.975)
And now thanks to Ortwin we have really a big insight into this area of application and the problems which are behind that we all know, I think, on the same page. This is really a project worth really development and next efforts, I would say.
Seb (12:56.116)
Ortwin going to you, obviously Viktor mentioned that smoke detection was, I think it was you who said smoke detection has been the traditional way of detecting fire and that’s something that everyone knows about. We’re also looking at temperature, so observing rises in temperature either from the air or through thermal imagery. Why is this such a game changer?
in terms of speed of detection and what you’re able to detect, why is this so important? Why is it such a game changer compared to traditional methods?
Ortwin Muehr (13:35.184)
I personally think from the practical side, there are two things which really speak for this technology. One thing is you’re not focused on incidents where you have smoke. As Viktor explained, we can map and detect smells. And so we could detect, for example,
The lithium ion battery which is packed inside the container which is suddenly starting to heat up and going through a thermal runaway. And we don’t need to wait until the fire is there. We can actually detect it earlier. And on the other hand, we have the possibility to check on operational processes on board which could also have dangers. For example, dangerous cargo spillages.
Fuel oil overflows if tank ceilings are suddenly giving up and fuel oil is going somewhere in the bilge or wherever. Just all could be detected by such a system and especially also in view of engine room fires, which mostly start actually through fuel oil being sprayed or pouring out by high pressure fuel oil pipes.
And at the beginning, there is no fire. There is just fuel oil coming through a crack of these pipes or through a broken ceiling. And this could be detected at a very early stage. And this is a true game changer, yes.
Seb (15:14.068)
So you’re now at the point where you’re not even detecting fire, you’re actually looking at molecular signatures in the air that represent events that may occur that would precede a fire but are still very risky. Do you want to talk about the work that you’re doing on ship and in the Cargo Fires and Loss program to start to profile these events?
Ortwin Muehr (15:28.784)
Exactly.
Ortwin Muehr (15:37.104)
Yeah, well, what we did, we put several sensors on board of a container vessel and we were simulating some incidents. For example, we put a sensor in the compartment of the vessel where the fuel oil is cleaned and we simulated an incident, a fuel oil incident, so we poured some fuel oil in a bucket and we were looking thereafter, you know, if the sensor were detected.
or we generated some smoke on deck of the vessel at certain locations and saw how the sensor was reacting on it, how much smell of smoke do you need. And I mean, this is just the beginning, but even at the beginning, it already shows very high potential. You know, it already shows from the first trial on, it’s better than the traditional fire detecting method. I mean, just from the beginning on, it’s better. And…
It’s hard to imagine what could be possible maybe in a couple of years with this technology.
Seb (16:42.74)
Yeah, the… Go ahead, Nick.
Viktor Bezugly (16:43.191)
If I may add to this What Ortwin already said, why smell detecting is really a breakthrough, because actually smell also has a good penetrability. So it penetrates everywhere. So imagine like Ortwin said something in the container, it’s kind of sealed container, but still, so you cannot see from camera, from infrared camera or something like that from outside or a temperature. Temperature doesn’t change, but the…
the molecules will already go out and already warn us that something inside the container container container is maybe burning or goes toward the scenarios it starts to burn. And you cannot prevent molecules to escape. It’s really a great potential, just a new level of interaction with reality. Of course, cameras or temperature sensors, this was already known technologists for a long time. But this is really new.
It may be even not completely substitutes the other technologies, but will be absolutely very valuable at on anyhow to the other.
Seb (17:54.356)
So talk me through kind of the, how does the work that you’re doing in the Cargo Fire and Loss Program possibly vary compared to other engagements that you have? Obviously, I know from personal experience, it’s very difficult to, I mean, once you have a great piece of technology, it’s very difficult to get it into industry quickly, prove that it works, start to produce it at scale.
Can you talk through kind of the where you are at in terms of your own technology development, the work that you’re doing in the program and how does your relationship with Ortwin kind of overcome some of those barriers?
Viktor Bezugly (18:36.055)
Yeah, thank you for this question. Of course, as you pointed out, this is a novel technology and it has some particular entrance barriers as any novel technology. So at the end, later on, everybody thinks, wow, good technology, we’re happy to have it, but not when you enter. So it’s really, you need to get some acceptance from the potential users that this technology really delivers what you declare.
Ortwin Muehr (18:58.256)
Mm -hmm.
Viktor Bezugly (19:03.927)
And people also have sometimes wrong ideas what they would like that this technology will deliver. It’s not probably in the current stage, it’s not mature enough. And all this we should really demonstrate in some short pilot projects, so to say, when we really come together with a potential user or customer to that place, how it will be used, how it will be then…
working in reality. So no one is interested in theory. In theory, everything is fine. But the question is always, please show me it works. And that’s exactly what helped us, where we got the help from SafetyTech Accelerator. So we could demonstrate our technology to really a big number of potential users. So that we do this step by step.
We started with some very local experiments and some fire safety test facility where we could burn some material which is relevant to safety on the ships and demonstrate that our technology outperform other technologies and really can give some response or warning in real time. So not waiting for hours or at least minutes. So it’s quite…
quite early detection, which really gives you a possibility to win a very valuable time when you’re fighting against a fire or preventing the fire. This first step we did, we demonstrated really within this round about performance of our technology. And then of course, another step was really thanks to Ortwin it’s really for us, absolutely not possible alone.
So really visiting the real container, huge container ship, at least to my understanding, it’s huge. Looking into really construction, into the scenario of application. So looking where this sensor make sense to place. So it’s really huge. So one can easily make any mistake without understanding the situation there.
Viktor Bezugly (21:25.399)
And I think with Ortwin it was really extremely helpful. And of course, this always, due to SafetyTech Accelerator also, because of this program, we were able to meet Ortwin I should also mention another partner, which was not yet mentioned. This is Duotech, our partner company, who helped us also with installations in this environment, which is not just indoor.
room installation. This is already many things which should be already thought about and done reliably. But we did this. I’m really proud that now the installations are there, they are running. We collected the measurement data, we collected background smells, we collected event smells. Everything is perfect and we are really hoping to proceed with next steps.
Seb (22:19.828)
So what’s next for the two of you in terms of this kind of ongoing partnership? What does the next year hold in terms of the types of tests that you’re doing, the things that you want to find out about the technology?
Viktor Bezugly (22:34.231)
Actually, coming back to your additional question or previous question about marked entry. So here, two activities are running in parallel. Of course, first, technical visibility demonstrations, but also awareness of the community. So I think the next step, I believe Ortwin will support me in that we need to demonstrate to the community where we are now that what we have done is really what we expected and what will really bring.
the favour to everybody. And then after the community is convinced, yeah, and I hope that we reach this stage, then we can do the next technological plan or the going towards prototyping. And yeah, but also we need, of course, feedback from the community, not just as acceptance, but also some.
used scenarios, some technical specifications. That’s everything which we’ll need for the next step.
Seb (23:38.932)
So I know we originally came together to look at early fire detection, but I guess the nature of the technology means that it’s got quite wide implications for risk, I guess, full stop in terms of given scenarios, generating molecular signatures that you can detect. Ortwin are you looking at this technology for a kind of…
a wider scale. For example, things like EV battery fires and lithium -ion breakdown you mentioned. And there are wider risks alongside more kind of traditional fire events that you’re looking at with the technology.
Ortwin Muehr (24:19.12)
Yeah, absolutely right. We are looking on a very wide range. It’s not only about detecting fires, as I initially said, there are so many other use cases we could think of to monitor operational processes by smell. I mean, just to say it in easy words, if you, for example, are toasting some bread, there will be a certain smell when the bread is ready. And…
So as we can detect smells, we can detect any production case or any procedure where the smell is changing at a certain stage. And if we can detect this, we have a new way of sensing. Bringing the nose into it is completely new. And there are so many use cases.
We cannot even think about now, but I’m sure it’s a key technology.
Seb (25:23.988)
Okay, both thank you very much. Any final comments or questions that you have for me?
Ortwin Muehr (25:31.216)
Well, how is your judgment or what is your opinion how this technology will go ahead? What is your impression so far? I mean, you’re dealing with different technologies. You have the broad overview of what is going on at the moment on the market and within our working groups. So what is your opinion, please?
Seb (25:54.356)
Yeah, I think from our perspective, we think this is extremely promising. It kind of represents the next step beyond traditional, even digital approaches. So we’re looking at two very significant new technologies in nanotech and AI. I think the analog nature of it is what distinguishes it. So we’ve worked in gas sensing.
a lot and it has its its found its limitations in its current format and I think this is demonstrating a next generation or two beyond the approaches that exist. So I think we think it’s very very promising. I guess the key is and this is where we want to help you is how do we demonstrate its performance, generate trust and and
create that transition from a piece of cutting edge technology that’s really interesting and can demonstrate high performance in and of itself. But how do we transition that into a solution that people are able to buy, that they trust and is robust, generating insights and affecting decisions within an industrial environment. And that’s the interesting next step is.
really solutionizing what’s clearly a fantastic piece of technology.
Ortwin Muehr (27:23.824)
Thank you.
Seb (27:25.428)
All right.
Viktor Bezugly (27:25.623)
Maybe from my side also questions of when we are talking about all stakeholders in this.
So one important party, I think, is also insurance companies. And I think they should also be addressed quite soon to show the potential advantage of use of such technology, which is, as I mentioned, not tend to replace some other important safety technologies, but really add on. And maybe you can already express that. Maybe.
have already a feeling how the insurance companies will look like on the introduction of such new technology.
Seb (28:13.268)
Yeah, I think the issue around insurance is a very interesting one and it’s becoming very prevalent because as we kind of move to a space in society where we do have technology that can start to demonstrate a credible reduction in risk, people are starting to question the concept of insurance and premiums. I think we’ve still got a little way to go before we can…
we can demonstrate that risk reduction and we need a lot of data and we need evidence. But it’s up to individual insurers to be first movers in the space of kind of, I guess, reviewing the technology that people have and recognizing the risk reduction. But I think it will be a conversation that continues and increases in frequency that…
better the technology becomes and every year we’re seeing new opportunities to reduce risk. So at some point they’re going to intersect properly.
Viktor Bezugly (29:24.183)
Thank you.
Seb (29:25.524)
Okay, both. Thank you very much for your time. And thanks to those who’ve watched this episode of Insights, and we look forward to seeing you soon.