This conference session demonstrates how machine learning and advanced computational methods are transforming engineering across multiple domains. In smart manufacturing, machine learning analyzes credential reliability in cyber-physical systems, identifying that word-plus-digit patterns dominate 33-38% of passwords, which poses security risks for industrial control systems. In power engineering, optimized neural networks (multi-layer perceptron with 76 and 38 neurons) achieve 4.74% prediction error for insulator leakage current under nonlinear environmental conditions. In biomedical engineering, powder metallurgy with controlled compaction pressure (200 MPa) produces titanium alloys with 25-30 GPa elastic modulus matching human cortical bone, eliminating stress shielding while maintaining bioactivity. These applications illustrate how AI and advanced materials science enable more efficient, reliable, and biocompatible engineering solutions.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
SIEMS-2026Added:
Good morning dear colleagues, participants and guests. We are pleased to welcome you to the international conference on smart innovation and energy and mechanical system. The grand opening begins with the performance of the national of Ukraine. I ask everyone to stand up please.
Sh.
Yeah. Cross Oh, the people show me the go Show me what despite the serious challenges our country is facing, we continue to move forward. This conference is taking place amidst the ongoing full scale scale war in Ukraine. However, this uh has not broken our spirit or our commitment to science and international cooperation.
Let us honor the memory of the following heroes of Ukraine who gave their lives of the independence, freedom, and future of our state. I ask everyone to honor their bright memory this minute of silence.
Oo.
Hey.
The conference is significant platform for exchanging had ideas and research in the field of energy and mechanical systems. The conference was organized by national universities of Ria polyeku in partnership with university of start dans university of technology ofenborg university of appliance science cardiff university international association of techn uh technological development and innovations GPMS international incororated joint stock company mci Scientific and production complex is scrap the Parisia chamber of commerce and industry and spring nation. The main goal of the conference is to bride together leading academic researchers, scientists and industry stakeholders to exchange knowledge, share best practice and engage in a discussion of both fundamental and applied research as well as uh industrial application in field of mechanical and energy systems. We hope that this conference will serve as a valuable source of knowledge for researchers, practitioners and educators and that it will be aspire further innovations in the field of energy and mechanical system. Due to the current situation, we are unable to hold the conference in the traditional format. This year, it will be held in a hybrid mode from May 11 to 13. Most presentation will be delivered online while uh some will take place here in the conference hall of national universities of politic. We are expanding participants from all across Ukraine and from nine other countries online translation available in zoom.
Now it's my honor to invite the director of national university polyank professor Victor to deliver his welcome address smart innovation in energy in mechanical system 2026.
Economic Institute.
Fore.
So following the opening remarks of our repa we would like uh to continue with the welcoming part of our program it's our great pleasure to invite our distinguished guests to address the participants of the conference with their greetings and welcome words and now I would like to invite our first speaker professor Iba Kalpas director of the institute of proas power semiconductor systems and since the beginning of this year also director of the inst institute for power electronics and electrical drives uh university of Germany uh to deliver some welcome remarks and I would like to add that it's clear that the leading of the two two institutions double workload double responsibility that's why we appreciate even more that uh you found time to join us today and we are truly pleased to have you uh with us at our conference so professor Kafas the floor is yours Thank you, Alina. Um, welcome everybody.
Thank you very much for giving me the opportunity to give a short welcome address to this year's edition of our conference.
It is uh both with a lot of pleasure but also with some sadness that I give this uh welcome address. Um or as we say in Germany, I have uh one eye which is uh sad and crying, but the other eye which is happy and uh laughing. And um the sad eye comes from the fact that war is still ongoing and that you still have to endure all um the atrocities of uh this ongoing war.
um the very and that we are still not able to meet in person fully and that the meeting has to be in hybrid form. So this is the sad eye, but my happy and laughing eye is that um you are still willing and able to organize this conference as a meeting of scientists and exchange of knowledge uh in our areas of research also for the um for for learning of our young scientists and for the new generation.
And another laughing eye and happy eye is because our collaborations are intensifying.
Um I am very much looking forward to everything that is planned between Saporicha Polytenic and my group here at the University of Stoodkat. uh one of the photos of our recent meetings is given here and I'm looking forward to our summer school that we will organize for your students and our students to come together at the University of Stoodgard in summer.
Most of all I wish to all of Ukraine and to Saburicha Polytenic courage, justice, peace and prosperity and also to all of us. I am looking forward to a successful conference and I'm I wish you a successful conference.
Thank you very much and looking forward to seeing you soon. Goodbye.
>> Thank you. Thank you very much Professor Kapas and thanks to entire team of University of Stogat for the warm welcome during our recent visit and for very productive and inspiring discussion we had there.
Uh so the next one uh we're pleased uh to introduce our long-term partner and good friend Professor Marana Dean of International for the Middle East School of Computer Science Informatics Cardiff University Wales United Kingdom. Uh Professor Rana has always been a strong supporter of our conferences and act actively engaged with our initiatives.
uh although uh he's unable to join us in person this year due to other academic uh commitments, we are truly grateful for his continued support. Uh we highly appreciate his interest in our work and his recognition of the efforts we are making to sustain and develop the uh conference even uh in challenging um in challenging times. So we warmly thank professor Rana and wish him all the best.
Uh let's move on and uh we would like uh additionally to acknowledge the message from professor Stefan Traes director of forenbuk university of applied science.
Uh again unfortunately to do a prayer commitments he wasn't able to join us this time for a welcome address. However he has sent his warm greetings and wishes for successful conference and fruitful discussion. But I'm I'm absolutely happy to welcome here with us uh professor Gilter wayel I would like to mention that this is the person thanks to which uh uh our students has had wonderful time during their uh visit of book university they just recently returned and I think that this uh group even with stronger impressions and motivation uh so it's especially valuable in today's conditions So uh professor uh so prof vice dean of research faculty of mechanical engineering and uh process engineering of book university of applied science uh you are welcome the floor is yours >> thank you for the introduction Alina I'm the next okay Um, thank you ladies and gentlemen, dear colleagues, dear students. Thank you for the opportunity to say a few words uh here today about our cooperation.
Two weeks ago, 10 students from Zabarisha visited us at Offenborg University already for the second time.
Last year, Titro joined at the exchange as a supervisor.
During their stay, we visited three company companies together and gave the students the opportunity to works hands-on in our lab.
This was something very special because in Chaparisa Saparishia due to the war uh studies currently take place mostly online that made it even more valuable for the students to gain practical hands-on experience in person at a cooperation between our universities already began last year. uh two colleagues of mine if Zikosski and Oxana Lozeno both born in Chaparisha came up with the idea for the exchange uh our recctors um Victor Gresa Ga sorry Ga and um Mr. um met met in often work to discuss possible collaborations. I'm very happy to see how these discussions are now becoming real cooperation. My sincor thanks go to everyone involved in in organ uh organization the exchange and supporting the students. I would especially like to thank Ditro who conducted the first visit with his students last year. I would also like thanks Daria Tak and Natalia Honcha who continued this excellent work this year as well as Alina for the organization and the support in the international office.
Besides the academic aspects, personal connections are especially important in difficult times like these. In only 10 days, new friendships were formed along with the first ideas for joint research projects. This shows how important international cooperation and personal exchange truly are. I wish you a productive conference and look forward to the future joint projects and I sincerely hope for peace in the Ukraine.
Thank you very much.
>> Thank you. Thank you very much for being with us and uh it's true. So our additional thanks to and huge thanks to Yuga for a warm welcome for arranging everything. So and without all of you it wouldn't be possible to um to send our student to your university and additionally I would also like that we sincerely appreciate the solidarity shown by open book university together with the university of life science amber following the recent drone attack on our university. So this uh support is deeply meaningful to us. Um and I already heard that uh this initiative with uh uh our student exchange will will uh have a continuation and the next year uh of book university will will be welcoming our students again.
>> Thank you. You're welcome.
>> Yes. Thank you very much. And uh let's move on. So the next we are ready to welcome uh professor Gonogi, doctor of science, head of institute of digital technologies design and transport national university Ukraine. The floor is yours.
Fore! Foreign! Foreign!
Conference man.
Fore! Foreign! Foreign!
Yeah.
Keep it.
Spring.
It's a conferences.
Conference So let us now welcome our next respect guest and speaker professor Vitali Iano.
of doctor of science head of the I'm sorry uh uh vice dean for international affairs of the faculty of technical systems and energy efficient technologies professor of the department of manufacturing engineering machine and tools uh sum state university Ukraine >> dear participants dear colleagues researchers and guests it's my pleasure to welcome you to the 2026 conference us as a professor representing Suma State University and the president of the International Association for Technological Development and Innovations. I'm honored to welcome experts, young researchers, industry representatives, and students gathered here to share their knowledge, ideas, and innovations. In my opinion, conferences uh such as SEMS are not only a platform for presenting scientific results, but also an important space for cooperation and creation of new international partnerships.
Today, science and engineering are developing faster than ever before. The challenges we face in industry, technology, sustainability and education require interdisciplinary thinking and strong connections between universities, research institutions, companies. I believe that this conference will inspire productive discussions and future joint projects. I would like to express my sincere appreciation to the organizers, partners, reviewers and all participants for their contribution and dedication to this event. As an organizer of scientific conference myself, I understand very well how much time, effort and dedication are required to prepare an event of this scale and quality. Thank you very much. I wish you all of you inspiring presentations, fruitful discussions, valuable networking and pleasant stay during since 2026 during these days. Thank you and I wish you a successful conference.
>> Thank you. Thank you very much and let's continue our program. And the next in our program uh I would like to pass the floor to Dr. Andrez, associate professor, vice president of the Visia Chamber of Commerce and Industry, Ukraine.
>> With the floor and would like to say that dear distinguished guest, dear participants of CMS 2026 conference. Uh I would like to thank the team of national universities poly techchnic for organizing such huge conference.
>> Sorry for interruption, but can you speak loudly or closer to the microphone because it's not so uh >> so loud? Yes.
>> Yeah.
>> Okay. Let me let me just I'll I'll make it in another way. Just >> now better much better. Thanks.
>> Better. Okay. uh dear distinguished guests, dear participant of SIMC conference 2026. I would like to say thank you to the organizational team of national university technic for giving me the floor today. It's a great honor to present some details of activity of the project chamber of commerce and industry as a business support organization and uh and a partner of uh project poly techchnic in the projects that we make uh together and in uh combining our efforts in this. So if I may I will make some a few slides let let me share my screen. Uh uh do you see it? Is it seen now?
>> Yes. Yes, I can see it. Uh so let me say that the British chamber of commerce and industry it's the organization that meet the requirements of ESO standard 901 and we provide services for our business and uh still and nevertheless the war time we make our efforts to give the effective functioning of our enterprises and our work is carried out by licensed experts in accordance with ational legislation. It's uh exhibition center was a palace that we used to to make before the war we had a huge number of conference forums and uh our partners from the poly techchnics also supported us in this activity but since 2022 it's humanitarian hub that is operating in the course we plan to resume conference and exhibition activities uh in the process of rebuilding of Ukraine.
I won't stop on the numbers of our various certificates that were issued in the in the time of war. I would say only that uh you can see that we uh every year we issue more than 2500 certificates and a huge percentage of leads it's for mojour and of course for our business now it's vital it's vital service that make them to be more effective and uh to not to be influenced by uh some uh some some aspects of uh of contract work or work in the term of uh national legislation uh training and consulting services.
It's the services that we provide and according to our according to our law and according to that experience that we have more than 20 years uh of such organization of such training courses online distant learning and workshops for our business and uh this um this slide is uh concerning our qualification center that is a structural unit of our chamber and it was accredited accredited by national veritation con agency qualifications agency and I can say that in some in some aspect it's the European experience of dual system education for example in Germany or in Austria where chambers of commerce and dusty are the basis for providing some aspects of dual dual training systems and uh since 2020 22. Our organization is an official partner of European Enterprise Europe network which includes 70 countries and thanks to this preparation we our companies of the region have access to the network's information resource B2B events with companies from European Union including representatives of metological machine building industries consulting companies all more than 250 companies were included in this ENN system in the time of the war. Uh we are the full member of association of management consultants and program partner in business with Germany. It's a very huge program that assisted us in some way uh to deliver uh European uh European approach to business because uh thanks to this program more than 30 representatives of the companies. This is SM sector almost participated in internship program in Germany and uh this in some aspect lead us to with uh national university poly technique to making the cluster engineering automation machinery cluster which is operating as a public union since 2020 and now it's uh it's an organization that is supported by different donors and we are proud Proud that with the assistance of our partners from university we have uh such a new form of supporting machine building and materological center in our region. Uh disk center is also the the center that that is operating from 2024 and deals with intellectual property and innovation. Uh the same disk center is also is functioning on the basis of by Poly Techchnic also.
And this is uh what we are proud that in 2024 we were invited to the consortium by uh with the lead of national nurses technic and we are the coorganizers of European digital innovation h cup h on the basis of the university and became one of 12 associations that won the European commission's competition and joined the network of European digital innovation hubs under the EU digital Europe program and this is the the experience and they showcase how the business and how the scientific sector can join their efforts in uh in making in solving new uh new challenges, new decisions in uh uh in innovation ecosystem.
This is the platform that we organized also in pro business where we show the companies and the economic sector of the region and we invite our dear guests from uh different countries to see the profiles of companies from different industries and services. Uh we consider this platform to be not just the showcase show platform of our businesses but rather uh to be the platform for the future mutual work and for making new cluster in new cluster organizations in our region in future.
So uh once again thank you for inviting uh for inviting us today and I would like to say greetings and warm wishes to all the participants of this conference and to share the experience of uh of all of all the participants in the name of further proh for the prospects and cooperation of Ukraine and our European colleagues. Thank you very much.
>> Thank you. Thank you very much for information that you provided for us and thank you for welcome words. Uh we are very pleased to continue our opening session with the welcome remarks from uh Yuri Torba, deputy director of the enterprise for scientific work, head of the experimental and testing complex joint stock company progress.
Foreign speech. Foreign speech. Foreign speech.
for conference.
Fore speech.
for it's now my pleasure to give the floor to our next honored guest, Mr. Head of the chief technologist department motor joint stock company Ukraine is a bit student.
Uh let's move on. It's uh my great pleasure to introduce our distinguished guest professor Christina Hamoka uh from the University of Technology and uh since the end of the previous year honorary professor of the national universities of Bishia politic professor gam always gladly support our and joins our academic events whether with the presentation welcome speeches or participation in scientific discussions. uh she has made a sign significant personal contribution to the development of cooperation between our universities and has cons uh continuously supported our initiatives including the night of science and international conferences organized by our university. Unfortunately, this time, Professor Gamula couldn't join us personally because uh of her schedule.
She has right now lectures and but uh she kindly send us um her greetings and wishes all participants a successful, productive and inspiring conference.
And uh according our uh program uh the next our guest will be Adam Baritzki, chief executive officer MS International Poland. Uh but again he is not with us online but he kindly sent us his video records with a welcome speech.
Foreign speech. Foreign speech. Foreign speech.
Fore!
Foreign! Foreign!
togo international.
Foreigne foreign.
Thank you.
forive concurrency.
Foreign for both international.
for Zatiach.
IMS International.
Efficiently manufacturing single parts automatically, reliably, and with the highest quality. Our software lives up to this standard every day.
We are a global technology leader in the CAD CAM and MEES markets. Our customers use Tebis software to design, plan and manufacture high-quality models, molds, and dyes as well as production parts.
Tbis is a privatelyowned company headquartered near Munich, Germany with locations in 17 countries.
Our 400 employees worldwide support customers from the automotive, aerospace, and machine manufacturing industries.
For the past 35 years, automotive manufacturers and their suppliers around the world have used TEBAS for their CAD CAM and MEES needs.
A Tebis CAD CAM installation provides a virtual representation of the user's manufacturing environment.
The use of digital twins makes operation of the machines cost-effective and reliable.
Our customers can also store their proven knowledge in manufacturing and planning processes in their teas and prolife system. Both are important measures for quality and on-time delivery.
But these efficient and reliable processes cannot be achieved through software alone.
Our combination of software and comprehensive services is what sets us apart from our competition. Our consulting and implementation specialists analyze corporate structure and procedures to develop strategies for management topics and manufacturing processes.
Additionally, we can implement these concepts with our customers making a valuable contribution.
Teis software packages are tailored to operational applications and technologies build on each other and may be expanded as necessary for special design and manufacturing tasks.
Tebis customers can combine their CAD, CAM, and viewer stations with MEES packages for planning and controlling manufacturing projects.
In order to provide competitive lead times and remain profitable, your material logistics and planning of machines and personnel must be in sync.
In training programs and webinars, we teach your employees how to use the software and we present the latest developments.
Our highly regarded help desk team can answer any questions you may have about the software.
Manufacturing processes include software and services. We provide both and prepare your company for the future.
Thanks for this informative video and let's move on uh to the end of the program. We have three more uh guests.
Uh and the next one uh it's my great pleasure to introduce our next distinguished guest uh professor Peter Aras uh faculty of engineering technology department of mechanical engineering Kovven Belgium and honorary professor of the national university of Parisia poly techchnical our cooperation and friendship with professor Iris has continued for more than 10 years during this time he has made a significant contribution uh to the development of international cooperation, academic mobility and joint educational scientific projects between our universities.
Professor Aris has continuously supported Ukrainian technical education, welcomed students and staff of our university at KU with the mobility programs and actively participated in international academic initiatives and conferences organized by our university.
We are sincerely grateful for his longstanding support and cooperation with our university. So, Professor Aris, the floor is yours. You're welcome.
>> Thank you very much for this nice welcome and uh introducing me. Um I will just want to welcome all the participants on the international conference on smart innovations in energy and mechanical systems and uh would like everybody to have a a good conference where you can meet your colleagues from the exciting fields of mechanical engineering and energy. I myself am mechanical engineer so from the machine building department in Kulvin in Belgium. Um the study fields our study fields of both mechanical engineering and energy is quite shaken let's say the last couple of years and if you start with energy we have the geopolitical events of which uh one of the tough points of course you very well know near Ukraine and uh but it's happening all over the world the geopolitical events and they're putting a lot of stress on the production and distrib contribution and reliability of energy resources and uh new challenges in using new innovations and artificial intelligence also rise the demand for electricity and other power resources and on an unprecedented scale in uh recent history. So we as scientists and engineers we need to find solutions for this. At this moment when we talk about artificial intelligence and smart innovations and we talk about the data centers uh they account now for 2 to 4% of country's electricity use and this is predicted to double by 2030. This is a lot of electricity which is not there yet and which is used for maybe not the uh most efficient energy use. So it will double to an astonishing more or less 1,000 terowatt hours. This means this is about 10 times the total electricity use of Belgium as a reference country for me of course.
So there is a lot of work and in mechanical engineering we also see a breakthrough in a lot of mechanical hardware and use of artificial intelligence to name a few. The advancements in electric mobility and humanized robots show that this that the field of electro of mechanical engineering is uh prone also to a lot of new innovations and also going very fast.
So in the next days on the conference there is a lot of things to talk about and see where contributions can be made and I'm sure that despite all the circumstances which we are suffering from that the CMS community uh can stay at the forefront and following new directions and contributing to science in these exciting fields. I wish you a lot of energy and inspiration for a fruitful conference the next days and a lot of opportunities for cooperation.
Thank you very much for your attention.
>> Thank you.
>> Yes, thank you. Thank you very much for joining for uh delivering your welcome speech and we're always happy to have you here with us. And I noticed that it's the same flag on your background that we saw the previous year. So it's very it's great pleasure to see uh in the family >> and is part of my life and my career. So I'm very supportive of anything which is happening there and of the country because it's uh it's a very difficult times. It's already lasting for some years. I hope peace will come soon and that I would be able to return to Zaparia to see the university and to see the people again in another way than just watching kind of a television show on the computer.
>> Thank you. Thank you for your words. We also really waiting for this for the time when we will be able to welcome here in the walls of the Zapia poly technique all our partners or our friends from the abroad from universities who right now kindly welcoming us because uh we have only one option but still we we we truly believe that this time will come and let's continue And now it's my great pleasure to introduce our next uh guest Mr. Robert Newman um managing director of car LTD Kenny uh United Kingdom. But I would like to underline that uh Mr. uh Newman joining to us from the USA. So it's you understand the time uh difference between Ukraine and USA. So yeah, thank you very much for uh joining to us. And Mr. Newman is a valued partner and supporter of our university within the framework of our cooperation with carved. He has contributed to the development of modern engineering education and the implementation of advanced karma technologies in the learning process of our students. We sincerely appreciate his support and engagement in strengthening international cooperation and innovation in engineering education at our university. So you are welcome.
The floor is yours.
>> Hello there. Hopefully everyone can hear me.
>> Yes.
>> So wonderful. Thank you. Thank you. Um as they mentioned, um I am in America at the moment. Uh I wasn't expecting to be here. Uh but I think I woke up at 4:00 a.m. and I think it's coming up to 5:00 a.m. now to u be part of this wonderful conference. Uh especially for this first bit. So, um, thank you very much for, um, inviting me along to this. So, my name is Robert Newman and I'm from KFCO.
Um, I am the managing director and I'm a co-founder of the company.
And for most of my career, I thought I knew what we did. I thought I understood that we built software. I thought I understood that we are a software development company. ones and zeros coding.
I thought we solved problems and like most creators and engineers, that's really how we measure our success.
Now, the software that we develop spans multiple creative and engineering industries. The product's been going for 31 years, so we're not new to software development or technology. And you could probably say that billions of products designing our software are out there in the world today. And as a company, that's quite incredible that most places in the world that you go to, our products are used by people. I'd say that probably at some point every single person in the world has touched one of our products. And I'm not talking about nuts and bolts. I'm talking about coins.
Um, all the coins in the world are designed in our software.
things like jewelry, confectionary, most commonly known are the Haribo gummy suites, the Golden Bear, but their whole range they're designed in our software.
So, we used in a variety of markets, which I thought I really understood and that as a company is quite exciting.
We take the problems manufacturing and design industries face and we develop technology that solves them. It's simple.
And if I take a look at some of those industries, if you were to look through our company website, you would see that it's mostly used for hobby woodworking. That's the part we show. The rest of it is what our software actually is used for.
and we work behind the scenes in lots of industries.
The most important market for us, I'd say, is engraving.
And that itself is a broad but technically cool and complex market. And most people never think about it. But if I said to you that in a number of cases, they own mines.
They take the raw material out of the ground. They process it, they smelt it, they roll it, they stamp it, they turn it into a product, they package it and sell it, ship it all within the one facility. That's kind of the definition of a complete manufacturing process.
There's a lot of processes to fully understand when you take a look at manufacturing and engineering.
In our world, this is the world of coin design. This is one of our more niche, larger markets, but it's a global market. Everybody in the world has coins.
So, it's a really interesting process.
But under this umbrella, we have what we call the engraving market. And that also covers greetings cards. So that's the Christmas cards, the sympathy cards, the birthday cards that people sell, but also bottle labels.
So we also have some bizarre markets as well like movie props, fashion brands, automotive and woodworking. So we span quite a lot. But as I said recently, I thought I knew what we did until we started thinking differently. And this was around our investigation into AI.
And because sometimes the work that you think simply is solving a problem ends up doing something far more profound.
It changes how someone experiences the world.
And that's where I really want to go with the rest of my talk is about how can we think differently with the technology we have or the technology we produce or the engineering the manufacturing processes to do things differently to make change.
And actually that change is what led me to being in a hotel room at this 5:00 a.m. in the morning doing this presentation is because this evening we're going to be collecting or I'm going to be collecting three awards for AI technology alongside Google Claude and Whimo.
So, it's not something I ever expected to do, but I think it's around the thought and the use of technology differently and it's exciting.
So, but it's not as exciting as what led us to the journey of how we got here.
And that's the journey I want to take you on now.
So, I want you to imagine that you walk into an art gallery. There's art all over the world, all over the walls. It's everywhere. It's beautiful to see.
People standing there, they're quiet.
They're taking in the art. Some people smile. Some people pause longer than others.
You step closer, you're noticing more detail.
Now imagine you're there but you can't see any of it. You hear movement, you feel the space, but the experience, the thing everyone else is having is just out of your reach.
And we've built a world that is deeply beautiful and visual, but not everybody gets to experience it equally.
Which brings us to the AI part. What if AI could be used for something genuinely positive?
Not just a buzzword or not just as a machine for generating endless content, but something that could be truly useful and transform people's experiences.
What if AI isn't about removing effort?
What if it's about creating an experience for somebody?
And this is what we did last year. We explored and put on an event we called Touch Beyond Vision. And I want to show you a short video about this event.
But I'm going to skip forwards to actually the blind people's interaction.
So what we did was we used AI and we took visual information. We took art, we took images, we took scenes and we translated them into 3D tactile art. Now we did this with CNC machines and 3D printers. Now the tactile art historically is known as bass reliefs.
It's the sort of beautiful art you'd see in many of your countries. Um, in Greece, in Italy, all of the old carved onto a surface artwork. It's something physical. It's something of real beauty, but something that you can explore with your hands, not just through sight. So, suddenly, an image is no longer something you look at. It becomes something you can feel.
So what we did was we developed within our own software platform AI technology capable of turning pictures into tactile touchable artwork and then we created an exhibition around it. We invited the visually impaired schools, charities, blind people basically to the exhibition.
But importantly with the whole event was we wanted to talk about democratizing it. We wanted to talk about how anybody could do this themselves. We wanted to see it in art galleries, which is why we put on an art gallery. But actually, we wanted people to be able to do this themselves. Using AI would take away the manual process, which honestly manual process to take anybody's face and turn it into the artwork would take many many years of um expertise in design. But actually it is a specialism and even if when you're an expert it's somewhere between 50 and 150 hours.
Well, we thought what if we could just take a picture and follow a step-by-step process and it instantly turns out into a 3-day bass relief and that's what we've done.
So, we put on this art gallery beside the river temps, but we didn't do it alone. And and this brings us to where we are here. It's about collaboration.
That sort of work that we did doesn't happen alone. It's collaborative. We collaborated with artists, technologists, charities, and most importantly, the people it was going to affect, the visually impaired community themselves.
And that's because I suppose real innovation doesn't happen by itself or as one company. It doesn't work when industries work in isolation.
It happens when different perspectives collide. And that's really why I'm here today because actually university collaboration matters. It keeps companies like mine on our toes.
And the world changes fast. Technology changes fast. Some of us we well all of us we we age and it's important to keep technology moving and forever evolving and the younger generation experts in other fields can move technology further further than working in isolation itself. We like to talk about it as breaking down the silos.
And while technology moves quickly, meaningful in innovation itself requires understanding people, understanding process, understanding problems worth solving.
And actually universities are one of the few places where experimentation, creativity, research and challenge all still exist because they have the time, they have the curiosity.
Universities bring curiosity, industry brings experience.
And when those two things meet, powerful things can happen.
What started for us as software development became accessibility.
What started as AI tooling became human experience.
And I think we forget a lot about human experience, human emotion when things are created.
And I suppose my segmentation into why we're here is that we work with our host university here who gratefully invited me along to this event. And that started a few years ago to supply software software that's used globally as you've heard some of the examples today. And the aim really is to provide a platform of curiosity and a platform of learning where on a positive side it gives skills for students to go out to the market.
But on a bigger picture, it gives students and staff the idea and and the researchers to think differently about what they create or what is the future that we could be creating because we should be thinking about the future.
And every year they run classes on making money. Not how to make real money and create fakes, but how to design money, a real world application using realworld tools in a niche market.
And I think that's the opportunity we all have right now. It's not just to automate.
It's not just to generate more content, but to create things that make the world more inclusive, more accessible, and ultimately more human. So, it's about taking the modern technology and AI is just the modern technology and it is it is a thing at the moment that's going faster than any of us can really keep up with.
But it's about using it for the right reason.
Creating loads of images and burning through, as we heard from our previous speaker, costs and burning through energy for the sake of it, which is what a lot of these images that people create are for, isn't changing the world. But actually using AI for the right reason, that's a responsibility I think we all have and that's where it can impact people the most.
As I'm here in New York, it was interesting. I got to go to Metal Labs and they have the glasses with the camera in and that can view the world that we live in.
Now, a I know a lot of blind people use these glasses to view and navigate the world. And the AR technology, they can ask them where they are. They can ask the glasses what they're doing, and they're acting as their eyes. This is good use of AI technology. The LinkedIn posts that people put up where it shows them in a different pose or them in a plastic wrapped vinyl um packaging.
That's okay and it's funny and it's slightly interesting, but it's not the real power of AI, which often is overlooked, especially by the media.
And really, what we should be doing as engineers, as educators, as creators, is think about the use of what we create, the use of the world around us. Now hopefully I'm going to be able to share this video. I will just play the first couple of minutes and then I'll skip to what people say. It's really interesting on the exhibition. Um but this is our use of AI technology.
So hopefully uh if I am just able to uh share my screen if that works.
Um, and hopefully I assume the sound comes through this as well.
Um, am I correct in saying that the sound comes through?
For some reason, I can't hit play.
Does the sound work?
Yes, we can see your video.
But there is no sound.
Just We can still do video without sound, but we can subtitles there soon. Maybe it should be.
So, I'm I'm not going to play the whole video because it is quite lengthy, but this is the art exhibition we put on. We used AI technology and this wall that sits at the back here, that was actually people submitting their own photographs and us turning their photographs instantly into 3D touchable tactile art in front of them.
And that's what we were able to do. We took a series of 3D printers and we were able to take their images And what I want to show you in the next part is this. This video is a wonderful emotional video, but I want to skip through to their words cuz the users words are the most important part of all of this exhibition and what they say.
So, I'm going to skip through to in their words because this hopefully it plays. cuz I am on a hotel a hotel Wi-Fi. So, in their words, so I'm going to skip forward to this bit and just play what the blind people say about the event or the artwork itself.
So, I'm going to stop that there and stop the share um and really talk about um to just close off this section here from myself, the positive use of AI, the use of AI in in the fact that we did it which was to really try to make a difference. And although our technology has been taking images, it has been taking vector outlines and it's been turning it into bass reliefs for 31 years.
This technology that we've now got was not possible with our AI at all. It wasn't something that we could create, democratize, consumeriize, uh, make accessible to absolutely anybody. So, you know, we're promoting AI for positive use. We're promoting collaboration in working with universities, with other businesses to think about the way that we as engineers um change the world and make things differently. So from myself, Robert Newman at KFCO in a hotel room in New York, thank you very much for your time and for listening to me.
>> Thank you. Thank you very much for such informative presentation and um we'll continue and I just want to say additionally that uh this software we are using not only for students in educational program but also for conducting some kind of master classes for school pupils and they quite engaging and excited about using this software. So, thank you for your support.
>> And let's move on. And last but not least, uh our guest uh we are pleased to welcome Dr. Eric Behopper, CEO and chief engineer of GPMS International incorporation, USA. Uh Dr. Big Hopper is a good friend and long stand longstanding supporter of our university who actively contributes to our educational scientific initiatives. uh he also initiated the international hackathon uh competition for students post-graduate students and young scientists and we sincerely appreciate his continued support of innovation and engineer education at our universities.
So Egg uh thank you that you are here with us and the floor is yours.
>> Well, good morning. Good morning. My name is as Elena said Eric Becker. I'm the founder, chief engineer, and CEO of GPMS International. I'm a company based in Vermont, which is a very small state just south of in the United States, but south of Montreal to give you context.
Uh, and we develop advanced health and usage monitoring system. Short name is HUMS for aircraft primarily for helicopters. My sort of background is I'm a former naval aviator, a senior member of the Institute of Electronic Engineers, a fellow of the Prognostics Health Management Society, and vice chair and fellow of the machinery fault prevention technologies. At GPMS, we build systems that combine sensors installed on the aircraft with embedded firmware and a ground station application to provide operators with actual information for making safety critical decisions.
HUMS supports routine maintenance activities such as rotor track and balance while also providing health on the aircraft drivetrain including the shafts, gears, bearings and the engines and then some aspects even the pilots how they fly the aircraft. We can give feedback on that. The result is improved safety and greater operational readiness by detecting problems they become that before they become failures. homes reduces unscheduled maintenance and increases aircraft availability.
This is especially important for commercial operators as it means improved safety and higher revenue because they don't have to take the aircraft down for maintenance. For military operator, it enhances sustainment, readiness and combat capabilities.
I am here because I'm a strong believer in the mission of the international conference on smart innovation in energy and mechanical systems. The challenges we face today are increasingly complex and solving them requires close collaboration among researchers, engineers, students and industry practitioners.
Conferences such as this are essential for advancing scientific knowledge and accelerating innovation.
In n in 2024, I had the privilege of supporting Zeparia polytenic national university and organizing the all Ukrainian scientific hackathon for young researchers in intelligence information technologies.
I provided the helicopter data set for a challenge focused on turbo shaft engine performance and sponsor the competition's prize.
Initiatives like this help inspire the next generation of engineers and scientists. This conference and the work that Zaparisia Poly Techchnic advances critical areas of including intelligent manufacturing, aerospace and energy systems, artificial intelligence and sustainability.
They are important and worthy goals and they deserve the full support of both academia and industry. Thank you very much. Just a short note. I just I'm just happy to be here. Thank you.
>> Thank you. Thank you very much for joining for finding the time and in spite of the time uh time difference you are here with us and additionally I want to thank you for the reviewing the papers and supporting our conference in the event. So now I think that we come to the end of the opening ceremony of our conference but it's not the end of the conference for today. Uh so I would like to once again um express our uh great uh gratitude to all our speakers to all our guests. Uh thank you that you find time and to join to our conference and support us in in in this initiative in this event. So it's very meaningful for us especially in this uh condition. So again just want to repeat that we hope and we believe that the very in nearest future we will be able to welcome you here in the university and conduct our uh conference offline.
So thanks to all uh to all a small announcement that we have a small break and we continue our conference in um half past uh 13 so 1:30 yes uh it will be the first session um devoted to artificial intelligence and digital technologies in engineering. So we'll be waiting you here in 1313.
Thanks to everyone.
Is it over? Is it real?
Dear uh participants of the course parents, let's wait for two or three minutes while everyone joining and then we will start our sessions.
So once again hello to everyone. Uh we will continue our conference uh smart innovation in energy and mechanical systems 2026.
uh thank you for joining uh and that you are here with us and we will start our first session which is devoted to artificial intelligence and digital technologies in engineering.
So and the first uh presentation uh have the next title. So enhancing cyber physical system resilience machine learning analyszis of credential relability in smart manufacturing of the authors Andread Andreko Julia and Allahanka.
So is anybody here of the presenters?
>> Yes. Hello Andrea here.
>> Okay. So as we did the previous Yes. So we have the video record. So for your presentation so we will start this video and after the presentation the auditori have an opportunity to set some questions if they have.
Yeah.
>> Okay. Yeah.
My name is Andri Kada from Adessa Poly Techchnic University. I'll present our work on how machine learning can analyze password reliability and why it matters for cyber physical system resilience in smart manufacturing.
This is a joint work with my colleagues Writes and Alropeno.
Smart manufacturing brings together two worlds. Enterprise information technology and operational technology.
The systems that run physical equipment.
When they merge, a password on a human machine interface is no longer just a security artifact. It becomes part of the reliability chain like any reliability critical component. And by credential in this work we mean specifically the password used to authenticate into industrial systems like operators HMI scala consoles engineering workstation etc. And the scale is significant across the major password bridge data sets in our study more than 40% of passwords low complexity. So we asked what can this data set tell engineers about credential reliability in industrial systems. From this question our goal followed naturally. We organized our work around three concrete tasks. First build a unified clean corpus from several bridge data sets. Second apply a diverse set of analytical techniques. And third translate the findings into recommendation for industrial environments. This diagram shows the full pipeline at a glance. We start with three bridge data sets. Rocku, linkadin and have I been pawn. They pass through preprocessing, then statistical and probabilistic modeling, then machine learning. In parallel, we run cross data set comparison.
The outputs at the bottom are what we offer back to engineering password families, risk categories, anomaly detection and recommendation for operational tech.
A few words about the data sets themselves. Ro is plain text with informal vocabulary. Linkerin is hash recovered with many professional names and have I been pawn is aggregated and continuously updated with the highest diversity.
These are the established benchmarks in the field. We choose them because they let us compare against decades of prior work and recent independent analysis of Roio 2024 confirms the same structural patterns that persist in modern data.
After preprocessing, we have 10 million records around 1 million unique passwords. Reprocessing is critical here because raw bridge dumps are noisy. We applied four steps encoding repair length filtering datation and uniform integration on ethics only public data sets processed offline only aggregate patterns reported with clean data. We begin with classical statistics. Three methods each measuring how predictable a password is. Frequency analysis tell us how often each password appears. Marco modeling captures local sequence regularities given one character what is likely to follow and probabilistic contextf free grammarss or pcfgs the compos passwords into structural pieces and words stem digits a symbol so we can ask how likely each structure is building on this baseline we apply four machine learning techniques described by what they do not by algorithm names clustering finds families of similar passwords automatically. Neural embeddings map each password into a vector. So, similar passwords end up close together. This is what let us detect mutations as one family. Uh, deep classifiers assign passwords to risk categories and anomalia detection isolates strings that look different from typical human chosen passwords, usually machine generated secrets. Our final methodological step is to compare the three data sets against each other.
If they showed very different patterns, our conclusions would be platform specific accidents rather than general behavior.
How much frequency distribution differ?
Uh structural categories overlap and embedding clusters look the same. This protects the industrial transfer claim.
Now to the results. Chart on the right.
Horizontal axis rank. Passwords sorted from most common to least common and vertical axis how often each appears on logarithmic scale. The solid line is what we observed and the dash line is the zip reference of our log curve seen across human language. Our line tracks the reference almost perfectly. Uh the zip exponent is 0.85.
It's a textbook signature of human choice. The top thousand passwords account for 7 to 12% of all occurrences and typical length is uh 8 to 10 characters. For industrial systems, this means a small set of patterns dominates operator credentials and that elevates operational availability risk. We can characterize these dominant patterns more precisely. PCG analysis reveals that word plus digits is the most prevalent template 33 to 38% of all passwords. The dominant families are dictionary words capitalized varants like password one short numeric suffixes and keyboard works. In operational technology contexts time pressured operators on HMIs or scala consoles naturally choose exactly this. This table puts numbers on those families.
Word plus digits is 36% overall.
Combined with pure dictionary words.
It's 59% of all passwords fall into just two categories and symbol containing passwords are only 4%. And these proportions are remarkably stable across all three data sets. The core vulnerabilities persist regardless of platform. Our clustering and embedded analysis supports this picture.
Clustering produced clean groups.
Dictionary passwords, names, numeric strings, simple combinations, and neural embeddings refined this by exposing semantic families. Dragon, Dragon One, or Dragons clustered together because the model learned the mutation. And finally, about 0.15% of all passwords, roughly 1,500 strings were flagged as outliners. These look different from human chosen passwords and most likely machine generated secrets.
Bringing this together as risk classes, our classifiers reach 83 to 88% accuracy and more than half of passwords fall into high risk. About 40% medium risk and fewer than 10% are high entropy or policy generated. And when we say high-risk swords have already been compromised, we mean that they are the kind that get compromised similar to known weak families very likely to be guessed in a dictionary attack.
In engineering terms, these are credential reliability classes.
It's like failure mode categories in failure mode and effect analysis.
Are these patterns universal? Comparing the three data sets, we observe moderate variability.
Jensen Shannon divergence between 0.09 to 0.23.
Rocku is informal, link professional and HIPP is most diverse.
An honest limitation shown at the bottom. These are not operational technology data. We treat our results as a behavioral baseline for prioritization, not as ground truth for a specific. We can translate the findings into practical actions for smart manufacturing. The basic idea is a diagnostic instrument for credentials.
Similar to predictive maintenance, instead of treating all passwords equally, the model identifies which ones are structurally likely to fail, likely to be guessed, reused, or based on a weak template.
The slide shows four directions for ICS SCADA and HMI access. PCFG detection rejects high-risisk passwords at creation. For MES terminals, embeddings detect reuse and trigger stepup authentication. For PLC and maintenance accounts, anomaly detection separates machine secrets from guessable patterns.
And for industrial IoT on boarding, pattern clusters reveal default structures. So this does not replace existing security systems. It adds a lightweight scoring layer at IT boundary and showing engineers where credential risk is concentrated.
To summarize, users mostly choose simple and guessable patterns. Machine learning gives richer insight than horistic rules and cross data set comparison shows platform specific drift. And for operational technologies, the tech weak credentials is a reliability control for cyber physical system resilience. It's measurable, automatable, aligned with engineering objectives.
We would like to thank the conference organizers, the research community for maintaining public benchmark data sets and springer for publishing this proceedings.
Thank you for your attention. I would be happy to take your questions.
So thank you very much for this uh presentation and um if anyone anyone would like to set a question you are welcome to unmute yourself and set question if you Yes, we can hear you. Do you have question?
>> Uh could you please uh precise what kind of passwords are mostly frequently used in the industrial area?
Um okay. Um first important uh caveat that our data is not collected collected directly from the industrial plants but uh uh we treat large public data sets or bridge data sets as a proxy like for for human password behavior >> and uh uh this uh because these failure modes are transferable. It's user under pressure choose memorable or repeatable passwords and it is regardless of context. And we see that the dominant pattern is a word plus digit or it's like a dictionary word play plus some uh numbers and as as it was shown on the slides it's about 33 or 38% of the overall data set password and plus uh just dictionary words it's another 23% so only two categories and more and roughly 60% of all passwords fall into these categories.
>> Thank you.
>> Thanks for the question.
>> Does anybody else have a question? If no, so we have one more question if you don't mind. So how can data analyze help uh help improve the security of operators passwords?
Again question about the password.
Um okay so data analysis uh help us like by shifting the password security from like static checklist to evidence-based risk assessment and instead of just checking the passwords um that password meets some uh character rules or minimum length we analyze data sets to identify the structures of these passwords uh not their uh length.
So uh we identify which are commonly and which are predictable and therefore easier to guess and we use and we use the statistical models like and clustering and animal detection algorithm to identify these weak patterns. Uh and as we uh as we shown we identified these patterns like dictionary words, words plus digits keyword walks and so on. And this weak passwords may influence manufacturing continuity not just like IT security and uh using this scoring mechanism organization can then decide either to reject the weak passwords or rotate high-risk credentials or trigger step authentication and so on.
So in in in short it's like risk adaptive uh access control instead of treating all available passwords equally.
>> Okay. Thank you very much for your response.
So >> yeah if there is no question yeah okay we can continue >> and uh we would like to invite next presenters presentation on topic modeling and optimal neural network parameters selection of predicting leakage current of 35 kilohz in insulators and nonlinear environmental effects and kesh Katina Petrova was uh uh Is anybody here?
>> Yes.
the questions if if they will be we're now sharing presentation here.
Good day honorable chair, distinguished colleagues and participants of the CMS 2026 conference. I would like to present the results of our comprehensive study titled modeling and optimization of neural network parameters for predicting 35 kilows post insulator leakage current under nonlinear environmental influences.
This research addresses one of the critical challenges in power engineering. Improving the diagnosis accuracy of high voltage insulation systems.
Reliable operation of modern power systems depends directly on the technical condition of insulation.
leage insulation health in real time allowing for the early detention of fleshover risks caused by moisture and contamination.
However, the formation of rage current is an exceptionally complex process driven by the interaction of nonlinear environmental factors.
Existing evaluation approaches based on classical polomial regression models have significant limitations.
They are highly sensitive to measurement noise and failed to adequately describe sudden changes under critical weather conditions.
Our research focuses on developing and verifying an optimized multair perceptron model. Our primary objectives included optimizing the network architecture, selecting drawing parameters for cyber convergence, and verifying the models robustness again against measurement errors to enhance overall diagnosis reliability.
Main aim is to develop and verify and optimized neural network model multi-layer perceptron.
for predicting the leakage current of 35 kilovolts post insulator under nonlinear and environmental influences.
The key research tasks first architecture optimization to substantiate the optimal neural network configuration number of hidden layers and neurons for the accurate approximation of complex nonlinear relationships.
Second, training parameter selection to determine the most effective combination of the training algorithm and activation function to ensure stable model convergence.
Third, the robustness verification to verify the model's resilience to measurement errors, filtering input noise levels of up to 5% and to minimize the prediction errors.
And force the reliability improvement to provide higher accuracy and reliability in insulation condition diagnosis compared to conventional regression based approaches experimental setup and data acquisition.
Since neural networks require high quality data for training, we focused extensively on primary data collection.
Research was conducted in specialized automated climatic chamber figure one which simulated a wide spectrum of environmental conditions. The setup allowed us to simulate environmental impacts by concurrently varing six key factors temperature gradient absolute and relative humidity waiting duration airflow velocity and surface deposit density.
Figure two on slide illustrates the schematic circuit used for leakage current measurement to enable the model to understand the underlying physics. We write six interconnected environmental factors difference between the insulator surface and ambient air. Absolute humidity, relative humidity, waiting durations, equivalent salt deposit density using nitro chlorium, airflow velocity.
Data collection was performed using Unity precision measurement equipment especially UT 33TH temperature humidity loggers UT61E plus digital multimeters and UT 321 high precision thermometers.
The use of such instrumentation ensured that the neural network was trained on physically meaningful relationships rather than instrumental artifacts or errors.
The proposed neural network architecture.
We selected the multi-layer perceptron architecture as a universal approximator of nonlineer dependencies. Our input layer consists of six neurons corresponding to the environmental factors. The output layer contains a single neuron with a line activation function to generate the continuous target value. The predicted leakage current.
The primary challenge was determine the optimal number of hidden layers and neurons. Insufficient depth leads to underfeitting while excessive complexity cause overfitting.
where the model loses its generational capability.
After 20 modeling iterations, we identified the most effective structure two hidden layers containing 76 and 38 neurons respectively.
This configuration provided the lowest error and most stable convergence.
The training process involved minimizing the mean squared errors by iteratively adjusting weights and biases using the back prop back propagation algorithm.
The slide illustrates a cyclic training process. It consists of three stages.
First uh forward propagation. The network generates a predicted leakage current based on current weights. Uh second is the loss calculation. We compare the prediction with experimental data using MSE and back propagation. The algorithm itally adjust the weights and biases to minimize the loss in the next cycle. Our goal was to find the optimal combination of network size optimization algorithm and the learning rate to avoid both underfitting and overfitting.
The choice of algorithm and activation function is critical for stability. We performed a comparative analysis of four optimizers.
Aen, SGD, RMS prop, NAM and three activation functions 10h real ELU. As shown in table one, the performance will rise significantly. The best results were achieved using the nadam optimizer with the eu activation function row 11.
Uh this combination yielded the lowest error matrix n of 0.139 and the mean absolute error of 0.02652.
The scientific advantages of ELU over the standard VLU is the ability to handle uh handle negative input values ensuring smooth gradients and leading to faster more stable convergence.
Next, we analyze the impact of the learning rate which determines the step size the algorithm takes towards the errors minimum.
uh as seen in table two uh and the MSE versus aograph an excessively high rate causes the model to overshoot the optimum while a right to low results installed learning of trampen a local minimum.
We tested values from uh 0.1 to uh 0.001.
A rate of 0.1 red line showed ultra fast convergence uh 8.24 seconds but suffered from gradient instability.
The optimal value was uh 0.001 001 the orange line uh providing the best compromise maximum accuracy with a reasonable training time of under 43 seconds for a full cogance over uh 270 epochs.
In the final stage, we verified the model using the new experimental data set of 1,000 data points. A scatter plot shows the correlation between measured and predicted leakage current.
Uh the point a cluster is tightly and along the identity time with a mean relative prediction errors is only 4.7400s%.
Furthermore, the MO provided robust uh even when the hardware sensor inaccurates the deviation did not exceed uh 0.02 0 3 milliampers.
The model effectively filters stoastic noise significantly outperforming traditional regression models.
An optimized neural network model for predicting the leakage current of 35 kilov volts post insulator was developed and validated. The main conclusions are summarized.
First, an optimal multi-layer perceptum architecture with two hidden layers containing 76 and 38 neurons respectively trained using the NAM optimization algorithm with ELU activation provides minimal prediction errors while avoiding overfitting.
Second, uh the verification of the proposed model using a data set of 1,00,000 experimental data points demonstrated a mean relative prediction errors of four 4.74% confirming the high adequacy and robustness of the neural network for leakage current prediction in the presence of stoastic measurement noise.
And third uh the obtained results indicate that the proposed approach is suitable for accurate regression based diagnosis of high voltage isolenti conditions outperforming traditional regression models in terms of prediction stability and the noise immunity.
And fourth, the developed model can be integrated into smart grid based conditioning monitoring systems uh enabling in transition from uh schedule maintenance to condition based maintenance strategies. Uh future work will focus on extending the input parameter set to include ultraviolet radiation intensity and adapting the model for long-term diagnostics of polytric insulation.
This research was made possible through the collaboration between the Central Ukrainian National Technical University and the National University of Life and Environmental Sciences of Ukraine and the key National University of Technologies and Design.
Thank you for your attention. We would be happy to answer any questions.
Thank you for your presentation and uh we have a question in the audience please.
Uh hello. Um I I have um one question.
Maybe it's not uh very professional one uh but like a a major question because I'm not a specialist in um artificial intelligence. But uh could you show me your uh first uh slide if it's possible?
No.
uh where video video and uh present one slide.
>> Okay, there are different u indices like uh humidity and uh different ones. Uh how all these different data could be connected in one uh complex index.
Uh sorry can you please uh repeat this question because I not can't clearly understand.
>> Yes here uh environmental stress level and you take into account humidity pollution voltage stress etc. and uh how all these different data could be combined in uh one uh index of environmental stress.
I'm sorry how old you as >> how they could be combined in one index because they have a different physical nature, humidity, pollution, voltage >> and here I see one index >> uh >> okay I understand uh we combined all these factors uh because they uh influence on this uh one very this one important uh complex uh uh complex structure uh of insulators because we uh we have if you can show the second slide the chamber laboratory chamber when where we u where we uh When we made the experiment with all these uh types of environmental influences and um we we taking all the data uh from uh different states and um put the data onto our uh neural network to um to organ organize these environmental parameters and uh select the optimal configuration for learning and training on this data. And uh on the output we had the model that can uh predict some uh leakage of on this insulators.
Uh I can see maybe you want some other answer.
>> Uh you see they all have different uh dimensions humidity, pollution, voltage, stress. How they could how can you make uh from these three absolutely different uh indexes uh one complex index that has a uh influence on uh current leakage.
uh okay their influence uh can be combined you know as any uh data we uh used uh sensors we collect the data I don't see any complexity we we collect this data we combine this uh you know the simple numbers the percentage uh you know degrees it all combines as the simple parameters and we use them in a model.
M sorry I don't understand how it's possible to uh plus humidity plus uh volt pro uh percent and then uh voltage >> from the from the side input parameters uh from from the input the parameters were not the volts were the environmental factors uh the humidity humidity the other temperature parameters and >> okay humidity and temperature they have different dimension >> but they have scale we can use the scale okay >> okay thank you okay and uh from this from the results of learning. Could you tell which factor has the greatest influence on leakage current?
>> Uh I'm sorry we did not select the most influential factor but the combination of factors was uh described and uh investigated.
>> Thank you. Thank you.
So anybody who wants to ask presenters, we haven't seen any questions. Uh so thank you very much for your presentation and we continue with the next present presenters.
Yes, once again thank you very much for your presentation informative um uh speech. So the next one uh the next topic is research of the smart greenhouse automatic control system of the authors which we quite well know. It's Salena and Dina Kazra. So we see that is here with us. So the floor is yours. Oh just a second. No, firstly we will start the video records and then uh we'll have an a question uh we will have the questions from the auditorians.
Yes.
Our research on the smart greenhouse automatic control system. Our team represents the Polish poly techchnic national university.
Agriculture is very important for country's development. It provides food for people, create jobs, supports local communities and helps export products.
This makes the agricultural sector important for a stable and strong economy. Today, smart technologies are used in many areas, especially in agriculture. In STEM education, special laboratory equipment based on micro:bit and in sensor helps students learn and develop research skills.
The relevance is twofold. First, it's about smart farming technologies to improve productivity and save energy.
Second, it's about using modern tools like micro:bit in education to help students learn automation and programming through practice.
The goal of the research is to improve the efficiency of a smart greenhouse.
This will be done by studying and optimizing an automatic microclimate control system. The system should create the best conditions for plant growth, use less energy, and reduce environment impact. Research tasks. One, heating and ventilation systems to keep a stable temperature considering plant tape, grow stage, and weather conditions. Two, study and control air and soil humidity and develop better automatic watering systems. Three, study and improve the lighting system based on the light needs of different plants.
This is our laboratory stand. It use the BBC micro:bit microcontroller. The microbit is small and has built-in sensors. It is easy to program. Because of this, it is good for making simple smart device prototypes.
External sensors and relay are connected to it. Based on these components, a closed loop control system is built.
The placement of the elements of the smart house and the connection between them are shown in slide five. The central control unit includes a micro bit processor module and expansion board and a liquid crystal display. The central module receives power from the power supply unit. It is connected to the elements of the greenhouse body through the loop of 14 conductors. The greenhouse has an LED lighting module and relay unit, temperature and humidity sensors, a temperature controller, a heater, and 12 to 24 voltage converter.
A thermal drive of the roof controls the ventilation hatch. A pumping station with a voltage tank is used for irrigation. All components are connected through a specially designed crossboard.
It provides separate connection points and switching. The smart greenhouse microclimate control algorithm works as follow. The heater is switched on or off when the temperature change by plus or minus 0.2° C from the set value. The ventilation hatch open when the temperature is 1° C above the set value and close when the deviation is 0.2° C. The irrigation pump works for 10 seconds when humidity is below 75%.
The internal lighting is switched on or off when external light is below 200.
This diagram illustrates four key functional blocks of the system. First, we have the lighting control circuit which includes the lamp and a light sensor. Next is the temperature regulation circuit combining a heater, a temperature sensor and a ventilation hatch. The third component is the humidity control circuit consisting of a pump and a humidity sensor. Finally, everything is managed through the human machine interface.
During the physical experiment on current temperature inside the greenhouse of 21° C, the task for the temperature regulator was set to 28° C.
The values of the set and current temperature expressed in IDC units were transmitted from the central model to the computer via the USB interface and recorded using the port monitor tool in the make code environment.
The result of experiments with the wiring heater power are shown in the slice 7 for supply voltage 24 30 and 36 volt respectively.
If the heater power is too low, the heater stay on all the time. In this case, the temperature does not reach to the set value. Instead, it's stabilized at certain level.
This level depends on how much head is lost through the greenhouse walls. When the heater is closer to the heat losses of the greenhouse, the controller works with small temperature changes. Figure B. However, when the heater power is much higher than the heat losses, the relay switches on and off too often.
This cause temperature oscillations around the set point. Figure C. Such changes can negatively affect plant growth. Also, walking at full power creates peak loads on the power grid.
This increases the risk of overloads especially when there is limited power supply.
The problems can be reduced by using heaters with power control or more advanced control methods such as speed controllers. They provide smoother temperature regulation, reduce fluctuations, create better conditions for plant growth. Therefore, pitbased control is more suitable in practice.
To determine the time constants, a system equations is created. The analytical solutions of which is quite complex.
Therefore, a graphical method is employed. Slide eight shows the transient responses of the original system one and the system approximated using the audible storious method two.
From slide eight, it can be seen that the approximated response is slightly different from the real system. At the beginning, it is deleted because of the system block. Later it becomes faster than the real system. However, the identified parameters can still be used for controller tuning.
The main disadvantage of this method is that is difficult to draw an accurate tangent as the inflection point of the curve.
In addition, the order tutorials method has limitations.
If the ratio A to B is less than 0.736, the straight line does not intersect the normal curve and the method can not be used. In such cases, a higher order transfer function must be used instead of a second order one. This makes the parameter identification process more complex.
The control objects parameters were identified using the parameter estimated tool in MATLAB simul. For this purpose, the acceleration data were imported from a CSV file.
The control object was modeled in Simolink as a second order initial system with unknown parameters defined in the workspace.
The results show the parameter estimator provides more accurate identification compared to other methods. In addition, the obainse second order initial model without delay is convenient for control tuning.
Another advantage of these tool is that it can also identify high order initial models. The oven satur methods can be used as an alternative when this tool is not available.
The dis disadvantage of the methods include the influence of the human factor which affects the accuracy of graing the tangent.
Another drawback is that it is produce a second order initial model with delay which require special methods for control tuning. There are also limitation in its application as mentioned earlier.
In conclusion, we created a working smart greenhouse prototype. Our research showed that relay control cause unwanted oscillations. So, speed control is better. Our standing models will be used for student training and further research. Thank you for your attention.
>> Thank you very much for this presentation and your audience if you have uh audit radio auditor if you have any questions so you are welcome to ask.
>> Yeah I would like to ask a question. My question is um to what extent does the use of the parameter estimator tool increase the accuracy and speed of identification of in national model parameters compared to classical methods in particular in the pres presents of delight and non loner ities uh of the control object.
>> Thank you very much for your questions.
Uh using the parameter estimator significantly increases identification accuracy because this tool minimize the influence of the human factor such as errors when drawing a tangent line in the audible storius method. Unlike classical graphical methods which have limitation uh for instance they do not work in the graphical ratio a to b less than uh 0.736 ml always for the construction of higher order models.
The resulting model is more adequate for the real object and more convenient for controller synthesis because it does not contain a poor delay link.
>> Thank you. Yes, thanks uh for your answer. If there are any other questions, so you are welcome. If no, then we can continue.
No, as I see there is no question uh from the auditory. So thank you very much uh for your presentation and we will move on to the next uh presentation and it will be the last before the break and after the break we will continue but uh the next presentation is uh local microclimate modeling based on safe echobot data for PV uh sighting and urban energy hub management.
and >> yes, I'm here.
>> Okay, we start share your uh presentation video and uh after that we're uh waiting for questions for you.
Good day colleagues. I am presenting our research titled local microclimate modeling based on save ecobot data for PV sighting and urban energy hub management conducted at the national technical university khark polytechnic institute. As urban environments transition toward decentralized energy systems, the ability to model hyperlocal climate conditions becomes essential for optimizing solar generation and managing proumer behavior. This study focuses on constructing a minimalist yet realistic microclimate model using crowdsourced data to optimize solar power plant SP.
sighting. We examined a triangular urban area in Kark defined by three Save Ecobot stations. By integrating temperature, humidity, and pressure data with a realworld PV generation log, we evaluated spatial interpolation methods to refine generation forecasts and manage demand within an energy hub framework. Our methodology relies on three specific sensors. Metropodski Lane, Valentineka Street and Kirpikova Street. These stations provide data at 3inut intervals. We also utilized a PV installation log from the Kirpikova site with a 5-minute resolution to correlate weather patterns with actual energy output. All coordinates were converted to a local cartisian system to facilitate precise spatial interpolation.
To ensure data integrity, records were converted into a wide format and physical outliers such as impossible humidity or temperature readings were filtered out. We aligned all time series to a regular 10-minute step using mean values. This resulted in a longitudinal data set spanning between 1,141 and 1,84 days across the three sensor locations, providing a robust statistical foundation for our analysis. As shown on this slide, our analysis of daily average temperatures revealed significant spatial variations.
While we observed a very strong correlation or approximately 0.97 between the Metro Budski and Valentinoka sensors, there was almost no linear correlation or approximately 0.02 between the Kurpikova sensor and the others. This discrepancy indicates a distinct physical microclimate at Kirpikova, likely a heat island effect or localized shading, emphasizing the need for sight specific calibration in urban energy planning.
To determine the most accurate way to reconstruct temperature fields, we performed leave one out cross validation LOCV comparing a planer model to inverse distance waiting IDW over a 46day concurrent period. IDW demonstrated superior accuracy with a mean absolute error M AE of 1.74° C nearly half that of the planer models 3.18° C during ostable spring and early summer periods IDW errors dropped as low as 1.2° 2° C confirming its effectiveness for limited point urban networks. We introduced an empirical transparency index KT a star derived from the ratio of real energy generation to nominal capacity. The index exhibits clear seasonal dynamics peaking in June at approximately 5.95 hours and reaching minimums in December at 0.15 hours. Correlation or approximately 0.48 48 between temperature and KTS star confirms that this index is a viable proxy for assessing illumination conditions and PV potential without complex radiative modeling. The implications for solar power plant sighting are three-fold. First, a simple three sensor IDW model provides sufficient accuracy for preliminary thermal screening. Second, data quality in crowdsourced networks is a more critical constraint than the interpolation algorithm itself. Finally, this low-level microclimate module can be integrated into fullcale AI tools supplemented by ERRA5 land data and satellite estimates to optimize SP placement at the scale.
Beyond the sighting, we model these microclimate results within a smart grid context. The study area is treated as a distributed generation cell where meteorological accuracy directly influences economic performance. By applying our IDW model, we can more precisely calculate the thermal derating of PV panels, which significantly impacts generation during the summer months.
Our analysis from July 2025 demonstrates that the local temperature deviations from city average forecasts can reach 3 to 4° C for a standard 1,000 square meter industrial rooftop installation.
This leads to a generation forecast error of 1.2% to 1.6%.
In the context of a balancing electricity market, this level of precision is vital for proumers to avoid financial penalties associated with generation imbalances.
This slide classifies the participants within our virtual energy hub. Hub A, the university campus, is an institutional proumer with a distinct microclimate and a daytime consumption peak that aligns with PV generation. Its strategy focuses on maximizing self- consumption. Conversely, hub B is a residential proumer with evening load peaks. For habi, the microclimate model is used primarily to forecast air conditioning loads which are often desynchronized from solar peaks.
The KTSR index serves as a demand response signal. Values above 0.7 trigger energyintensive processes like EV charging while values below 0.2 signal a shift to economy mode or battery storage. Furthermore, integrating these temperature maps into Escada systems allows distribution system operators to identify blind spots and dynamically assess line overload capacity, which is critical during hot days when PV generation and overheating risks both peak. To conclude, we have successfully implemented a minimalist microclimate model that leverages open data to improve urban energy management.
We proved that IDW interpolation provides a significant accuracy advantage for PV thermal calculations.
Future work will focus on introducing geostatistical tools like creing and integrating ER5 reanalysis to create an end toend AI system for automated SP sighting and grid optimization.
Thank you for your time. This research represents a step toward more resilient and datadriven urban energy ecosystems.
I am now open to any questions you may have.
>> Okay. Thank you for your presentation.
Is anybody wants to make questions and also uh I have a question uh for presenters.
How will the integration of geostatistical methods and satellite data affect the accuracy of local microclimate modeling and evaluation of photo volte system efficiency?
>> Thank you for your question. It's like our future strategy for this work when we want to uh transition into geostatic and satellite data as era 5 GHI. It allows to us transform local our model into a precision system for forecasting.
So um it mean that in our work AI um we use not like uh instrument but like our uh final goal to create uh new instrument for forecasting and provide um uh more optimal sighting of PV uh plants into urban cities. um and so on. And uh when we use this geostatic data, we uh improve our accuracy uh like uh when we start to use uh methods of uh crean coing uh it allows to us um um use new uh more wide data such as um train belief, land cover and so on and so forth. Um it give us uh opportunity to provide u boundary and atmospheric condition. Um it gives uh to us chance to improve our economic valability. So our research indicates that a temperature forecasting uh discrepancy of only as uh from 3 to four uh degrees of calcium can lead to a generation forecast error of 1.2 uh to 1.6% 6% and uh as it was uh told into our presentation uh it gives a very big opportunity uh to avoid financial penalties when we want to build some industrial PP plant um and so on. So it will be a future uh our work.
>> Thank you. And as as uh connecting about first question, the second question is how effective is the use of the empirical atmospheric transparency index as a control parameter for balancing energy consumption between uh in institutional and uh residential consump consumers within a local energy hub.
>> Thank you for your question. Um I can say that this uh empirical index is u such like effective and more cheaper uh like u signal for management. Um by this index we can balance our uh urban uh micro grids uh our consumers and so on.
Um for instance if uh for instance if we want to um talk about demand response uh in this case this index will uh be like some trigger for automatization.
Um if our index will be more than 0.7 it will be like a signal to activate some um heavy load like EV charging uh and so on. But if we uh have index lower than one um 0.2 to uh it gives signal to us that we need to transition into economic regime uh into regime of using our battery energy storage system and so on.
Uh and as it was described for different profiles like uh energy hub A and energy hub B, we can balance other profiles. Uh in case of hub A uh we synchronize our day pics for hub B we synchronize and forecasting our evening picss and uh so on and this provide like a final step um our like um system utility uh we um um we can provide this uh stability uh we can uh start to be like exporter uh of energy when we have a proficit uh from April to September uh and in different period of time we can optimize our a lot.
>> Okay, thank you for your answer. Uh is anybody wants to ask We haven't seen any uh uh questions. So, thank you for your presentation and answers and now we have a short break before uh 3 p.m. And uh after that we continue our section. Uh please uh don't connect uh don't disconnect with us and continue our section uh before the break.
So yeah, see you after the small Break.
So we hope that everyone uh have time to rest a bit to uh to have a cup of coffee or tea or something else. So now we can continue our conference the first session which is devoted to artificial intelligence and digital technologies in engineering.
So let's continue >> and yes so we um would to ask to present uh their research love Nicolin and Petroak hourly based for a casting for photo voltech power plants and its impact on low voltage distribution.
networks.
Uh we start uh sharing your video presentation and after that we want to ask you some question.
Dear colleagues and friends of the conference, let me present our research hourly based forecasting of photoaltic power plants and its impact on low voltage distribution networks uh made by Yaslo Basala Noland and Petro Kurlak.
Accurate prediction of photoaltic power generation is a key requirement for efficient integration of renewable energy sources into modern electrical power systems and low voltage distribution networks. In Ukraine and many other countries of the world, the number of photovoltaic plants with micro and mini capacitors that are connected to low voltage networks is increasing.
The activation is carried out according to simplified requirements without conducting the necessary analysis which can cause problems with un generation voltage dismatch and other energy parameters. A detailed analysis of the difference in the requirements for connected PV sources to national networks has shown the importance of monitoring the electricity quality indicators as well. Besides for restoration of Ukrainian energy system with photovoltaic station, it is necessary to do research on the problems of integration of renewable energy sources and connecting them to the electrical grids for betteranding the impact of PV uh station on the parameters and operating modes of the local grid as well as to plan the operation of hybrid system with storage physically interpret ed forecasting models are required. The calculation of the power of PV systems can be performed using daily or monthly generalized indicators such as daylight hours or average solar radiation. Although these methods are highly effective, they provide limited understanding of shial for low voltage distribution networks and smart grid applications.
First of all, we need to improve the methods of forecasting the level of generation of photovoltaic power plants and optimal management of power grid modes taking into account the dependence of the power of PV station on solar insulation, air temperature, cloudy weather like clear sky, cloudy day.
Our previous studies have shown different ways to represent the performance of photovoltaic plants. For example, it is possible to approximate the power curves of piv plants using regression or harmonic models related to the length of daylight and selected methologological factors. However, daily averaging mask the intraday variability caused by cloudiness, temporal weather radiation. As a result, such models have high errors due to the risk of inaccurate weather forecast and may under underestimate power fluctation and they impact on the voltage leverage level and energy balance in distributions networks. As you can see on the slide number five, we use statistical data on the amount of electricity produced at PV plants.
Metrological data from the metal blue server and the wetter radar information site is also used. This allows the Excel interface to extract data for the project and use it in individual devices. Most of these platforms provide data on generation for different days, months as well as changes in solar insulation, voltage and electricity consumption for photovoltaic station.
It is quite difficult to find up-to-date data on the forecast of changes in solar isolation during the day for a selected location for free and most modern models based on neural networks requires significant funds which owners of small PV plants and researchers are not always ready to spend. So our idea was to analyze the possibility of creating simplified models that would take into account available online data.
Therefore, we combined our previous models with the factor named dynamic cloud factor. We take weather focus from the weather website. The application is convenient with relatively accurate cloud movement and information about the number of sun hours during the day.
Determine the cloud factor is possible through the image J program for image processing which is free also using artificial intelligent chart and visual focus was also made which is the fastest.
We made several variants of the model taking into account only the cloudiness coefficient only the number of sunny hours during the day taking into account temperature. Some results are displayed on sliders.
The comparative analysis of PV power generation forecast demonstrates the strengths and limitations of different cloudiness estimation approaches including expert-based hybrid and pixelbased methods.
To demonstrate the approach, the model was validated using data from photovoltaic system located in different climate regions. For example, Ivana Francis, Ukraine, Poland, Munich, Germany, and Prague, Czech Republic.
The results indicate that no single method consistently outperforms the other across all time intervals. The expert-based forecast shows the highest accuracy during morning hours where atmospheric conditions are relatively stable and cloud presents in minimal in this period. Human interpretation of sky provides a reliable estimation of solar radiation and correspond output.
Program image J helps to define coefficient of cloud news as well during peak generation hours. The adjusted model show the best overall performance. This approach benefits from moderated cloudiness value avoiding overestimation typical of pure optimistic assumption while still capturing the high irradiation levels associated with near clear sky conditions. In constant uh the pixelbased method demonstrates superior capability in representing intraday variability particularly under partially cloudy conditions in the afternoon. However, it tends to systematically underestimate absolute generation level likely due to the overclassification of thin clouds or atmospheric haze in satellite im uh images. The further analysis reveals that difference between forecasted and measured power in the afternoon period are caused not only by cloud cover but also operational constraints of the PV systems.
Forecasting with hourly resolution allows for a better assessment of the operation of PV plants in low voltage distribution networks, including the identification of peak generation periods that can cause voltage increases, reverse power flows, or increased load on power electronic interfaces. As a result, the model supports a more accurate assessment of the network operation conditions and facilitates the planning of voltage regulation and reactive power control strategies.
The study contributes to a deeper analysis of electromagnetic compatibility in distribution networks, which is relevant for networks with a high level of PV installations where electromagnetic interference and power quality problems are becoming increasing critically.
Thank you for your attention.
>> Thank you for your interesting presentation.
Um if anybody wants to ask uh presenters about uh this research.
Uh so we already have a few questions.
Um the first one is uh what are the primary materological variables such as solar radiance, temperature, cloud cover that contribute must to the error rate in hourly photovoltaic power prediction models.
>> Uh thank you for your question. So according to our search um uh the cloudiness is one uh gives uh the greatest impact and the greatest error and um actually we didn't study uh solar insulation deeply in this case I mean because it is very difficult to find a proper data um uh from forecasting uh resources. So u uh cloudiness contributes the most.
>> Okay, thank you. And the next one question is uh how does the study differenti between critical and non-critical energy use and what mathematical framework is proposed to quantify the efficiency of the critical energy components?
>> Uh thanks. Um I must admit that first of all uh we use forecasting uh models to uh determine network parameters in the context of electricity uh quality. uh but u u we take the initial assumption that critical nodes are powered uh in any case because uh they can't be uh turned off or they can't be uh postponed without risking system uh security or stability.
>> Thank you Rihanna. Uh so we continue.
So if there is no any other questions from the auditory so let's continue and the next uh speakers are Victor Hashtropeno Yuri and Mikail Liverski with the presentation which is entitled as modification of the elastic modules of titanium alloys for cervical interverral implants uh innovation approach Thank you. Uh dear colleagues, it's my pleasure to address this international audience on behalf of the Zaporasia Polych National University and the research team of our doctoral program in material science. uh the work I'm presenting today belongs to doctoral research dedicated to the development of the uh scientific foundation of low modulus biomedical alloys a field practical embodiment of which is the medical implant implants are no longer auxiliary devices uh they have become structural parts of the human body they share its lord dictate the kinetics of its haling and very often determine whether a patient returns to a productive life or to a chronic clinical pathway.
uh and uh from cervical cage and spinal fusion devices to loadbearing oropedic reconstruction uh the modern implant must do something that until very recently no manufactured material called do. uh it must behave mechanically like uh bone uh integrate biologically like living tissue and uh remain manufacturable at a cost compatible with mass clinical use. Uh this challenge has acquired particular urgency in the Ukrainian context uh where war related muscularkeeletal trauma and aging population and a growing burden of degenerative cervical pathology occured. Uh today we are uh facing a generation of uh patients uh uh for whom implant uh performance is not a matter of comfort but of national rehabilitation capacity.
It is against this background that our work targets the cervical intervertebral cage, a small but mechanically uncompromising device that sits at the very intersection of biomechanics, material science and clinical engineering.
Uh the uh central scientific questions of this uh presentation is can uh we engineer uh titanium uh the gold standard biioaterials to also acquire the elastic modules of cortical bone and can we do so using manufacturing road that is uh uh scalable?
Our answer uh developed over the next 12 slides is that powder methodology provides exactly such a route.
Uh on the left you see the post operative reality of cervical fusion surgery. Uh small interbody cage installed between the uh C6 and C7 vertebra. uh to restore disk uh height and uh stabilize the spinal segment. Uh the clinical efficiency of such devices depends above all on the lower vein shared between the implant and the bone rather than monopolized by the implant. On the right uh we see why this is so difficult. Uh traditional compact titanium has an uh elastic modulus of around 100 gigapascal. Uh whereas human cortical bone lies between 20 and 30 gap pascal. uh mismatch that drivers the well-known phenomenon of stress shielding leading to progressive degradation of the uh adjacent vertebrae.
The uh group of materials uh currently dominate the cervical interbody market and each comes with a structural uh compromise.
uh solid titanium offer excellence uh bio compatibility and strength uh but as I have just uh shown its uh stiffness drives stress shielding and subsidance uh p matches the modulus of bond beautifully but it's bionert surface results in slow and often incomplete uh also integration. Uh 3D printed titanium latises um can be uh tuned in modulus but their cost and production uh cycles remain a barrier to routine clinical adoption.
Uh uh our objective therefore is clear uh to combine the bioactivity of titanium, the modulus of peak and the cost effectiveness required for genuine scalability.
The uh proposed solution is simple but technologically powerful.
uh instead of uh casting solid titanium, we replace the methological root uh internally uh starting with PT5 titanium powder with a particle size fraction of uh 0.5 to plus 0.16 micrometers.
We applied cold pressing at room temperature followed by vacuum centerin and at 1,200 50° uh for u 5 hours. Uh this uh Swiss step processes transform titanium from uh dense monolithic materials into a non-compact cellular medium that mimics the trabacular architecture of bone. uh eliminated the need for toxic alloing uh addition and uh creating uh and uh interconnected poor network whose mechanical response can be precisely tuned by uh compaction uh pressure.
Uh this graph contains the most important engineering finding of our study. uh by varying the compaction pressure from 100 to 600 meapascal we obtain a continuous uh monotonic relationship uh uh porocity decrease from 35% uh to uh 13% uh while the elastic modulus rises from uh 18 to uh 60 gap pascal.
uh the optimum window of the so call it compaction sweet spot at approximately uh 200 meap pascal deals 27 to 30% porocity and the elastic modulus of 25 to 30 gap pascal uh which is near perfect mechanical match for human cortical All uh metallographic analysis at 100 and 200 magnifications confirms the formation of stable uh non-compact matrix in which the dense solid state structure of conventional implants is completely absent.
uh the pawn network is uh interconnected with characteristic dimension in the 100 to u 500 micrometer range. Uh exactly the geometric scale identified in the literature on poros titanium as fable for osteoblast migration vascularization and durable ointigration.
Uh in other words, the same uh processing step that lowers the elastic modulus also produces a biological scaffold that addresses the principal weakness of pig.
Uh determine the elastic properties of a non-compact metal is not a trivial task because uh classical static test overestimate or uh underestimate stiffness depending on porosity.
There overcome this we developed a hybrid res resonance based meth methodology. Uh in the first step, the natural vibration frequencies of prismatic prismatic samples uh captured experimentally using uh acoustic resonance in the uh 50 to 20 kilohz range.
In the second step, those frequencies are matched against an anis model analysis of the same uh geometry until the calculated and experimental frequencies agree to within five%.
This uh uh computational experimental coupling enabled use to determine the uh true dynamic elastic modulus and poison radio with verified uh accuracy providing a solid foundation for the cervical spine simulation uh that follow.
Uh having validated the material we built a fed element model of a functional uh spinal unit comprising the uh uh C6 and C7 vertebra with the implant in place. The lower segment of C7 was rigidly fixed physiological compressive load of 73.6 6 Newtons was applied actually through the upper C6 vertebra and dynamic moments of plus - 108 Newton m uh superimposed across multiple axis to reproduce flexion extension and lateral bending.
This uh boundary configuration faithfully reproduced the in vivo loading regime and implant experiences and is consistent with the established literature on cervical biomchanics.
uh to benchmark uh benchmark our material we pitted two uh uh architectures against each other. Uh on the left uh is the control uh 3D printed cage that achieves low elastic modulus through complex internal latice uh uh high performance but also high in manufacturing cost and uh production time. Uh on the right is our challenger homogeneous solid non-compact titanium cage uh with no internal latice at all where the entire mechanical adaptation is carried by micro level porosity engineered during the powder methology route.
Uh the finite element analysis shows uh that the non compact powder methology age distribute distributes equivalent stress uh evenly across its entire volume under both compressive and bending loss. This uh architecture eliminates the uh localized stress concentration that drive adjacent uh bone degradation and catastrophic implant subsidance. Uh there two fail mechanism that have haunted titanium cage for decades.
The clinical takeaway is that uh we are not merely reducing the modules. We are restoring genuine structural harmony with the vertebral in place.
This uh slide uh presents uh the studies most important comparative result across compression, tension and lateral bending. uh the stress and strain topologies of the non-compact powder metalology cage are virtually cannot uh be distinguished from those of the additively manufactured lettuce cage.
This uh big differences within uh 10 15% across all simulated physiological laws.
Uh in biomechanical terms uh this is a functional uh equivalence uh meaning that a much simpler uh cheaper manufacturing route can uh deliver the same clinical performance as the most advanced 3D printed devices currently available.
Uh this synthesis table summaries the entire research across uh four dimensions.
Modulus matching also integration stress shielding risk and manufacturing cost.
Uh solid titanium fails on stress shielding. Uh P fails on also integration.
uh 3D printed titanium fails on cost. Uh only the non-compact powder metal titanium satisfies all four criteria simultaneously.
Uh the final conclusion is that powder pathology successfully bridge the biomechanical gap delivering an implant that behaves mechanically like bone.
integrates biologically like titanium and scales economically for global clinical adoption.
Uh and uh conclusion uh we started from the proposition that the modern medical implants is no longer a passive insert but an active mechanical and biological part of the human body. Um the uh data I have shown today demonstrate uh that conventional powder methology is fully uh capable of producing such a partner at a compaction pressure of approximately uh 200 meapascal.
Uh our non-compact titanium alloys exhibit an elastic modulus of 25 to 30 gigap pascal uh and 27 to 30 porocity% porocity uh closely matching human cortical bone element analysis confirms that their stress strain behavior is biomechanically equivalent and to that of additively manufactured porous titanium with uh discrepancies that do not exceed uh uh 10 to 14%.
Uh from the broader perspective uh of implant engineering this works contribution is therefore three-fold. Uh first it uh removes the long-standing trade of between bone uh matching mechanics and biological uh compatibility.
For the first time a single materials processes both the modules of bone and the bioactive surface of titanium.
Second, it restores manufacturability to the uh equation. Uh uh unlike additive technologies which remain uh capital intensive and slow, powder methology is a major scalable industrial process available across the global manufacturing base.
And uh third it lays the uh methodological groundwork through the resonance based dynamic modulus determination and the validated framework uh for an entire family of law of modulus biomedical alloys which is the central scientific uh uh objective of of the doctoral dissertation that frames this work.
Uh and of course I like uh good uh engineering the work is not complete. Uh our future research priorities are clearly um defined cyclic fatigue testing under uh physiological loading perm77 in vitro bio compatibility and cell proliferation studies static compression testing to consolidate the uh static dynamic modulus.
relationship and in vivo animal model evolution of o integration kinetics.
Uh this step will move the developed ls from a validate material concept to a fully certified clinical implant.
Uh to sum up I would like uh to uh face uh that the implant of the future will not be uh defined by a single property uh not by stiffness alone uh not uh by biompatibility alone and not by manufacturing uh iligence alone. uh it will be uh defined by the uh simulation mastery of of all three. Uh today presentation has been uh an attempt to uh demonstrate that such a simulation simultaneous mastery is uh achievable and that is achievable in Ukraine at our university through classical powder methodology.
On behalf of my authors and our research team, I thank you for your attention and uh look forward to your questions and productive scientific discussion. Thank you.
>> Thank you very much for your presentation and for the quite important topic that was arised and we see that we have a question in the auditory. So hello for you're welcome.
Uh hello um I'd like to precise uh one point of your uh research. Uh could you please u give some details if there was found uh any statistical statistically significant correlation uh between compaction pressure and pon ratio?
>> Mhm. Mhm.
>> No, the variation of poison ratio with compaction pressure was not analyzed as any potential change is uh expected to occur within a very narrow range. Uh the primary focus of the study was on the evaluation of the elastic modulus.
>> Mhm. Thank you. Thank you very much. Any other questions?
>> Can I ask something? Uh, >> you're welcome.
>> Uh, please considering your future work regarding fatigue resistance, fatigue life. uh uh what uh do you think uh whether some surface modification is needed for your implants uh to increase fatigue resistance or not?
Thank you.
It is it is well known that surface modification can uh form compressive stresses that will increase fatigue. big resistance is it uh important for your implants? What do you think?
>> Uh to my mind I agree with you. Uh um this uh materials implant u must must modif modification. Yes, I agree with you.
>> Thank you.
Thank you very much. Uh if there are no more question from the auditory. So thank you very much for this presentation and we can continue and the last and not least uh uh presentation. It's a lost presentation.
Last >> uh yes the topic of the last presentation is application of fractile geometry methods for pro processing and analyzing thermoggrams in termral imaging inspection of non-metallic heterogenous materials.
The authors are Marina Yuri.
Is anybody?
>> Yes.
>> Yeah. Yeah.
>> I see.
>> So we um uh uh start sharing your video presentation and after that we uh want to ask some questions.
Dear colleagues, your attention is presented to the work called application of fractal geometry methods for processing and analyzing thermoggrams in thermal imaging inspectation of nonmetallic heterogeneous materials.
Nonmetallic heterogeneous materials are complex system made of several components. a polymer matrix uh with which are fillers and aggregates of various natures for example fibers fine particles crystals etc. Compared with traditional structural materials using of materials of this class allows to reduce the mass of the structure increase the service lives uh produce materials consumption increase the reliability and power of mechanism.
It creates the possibility of developing fundamentally new structures and their operation uh in extreme conditions. In addition, the advantage of this class of materials is the reduction of time uh for manufacturing parts and simplification of the technological process of their manufacturer. Act thermal methods of testing products made on made of non-metallic terriinous materials are quite promising today.
Here thermal energy distributed in the objects of study is used as an intensive parameter. Today the registration and visualization of the thermal pattern of the surface of control objects is carried out using infrared devices that implemented a non-cont temperature measurement method. This approach is characterized by high informativeness to negative impact of the objects of diagnosis, environmental front lines, mobility, the possibility of uh implementing control and measurements during the operation of the objects and therefore reducing cost.
The purpose of the work is to develop scientific approaches that will be the basis for the implementation and autom uh automated processing and analyzing of the thermographic images in the system for detecting defects in nonmetallic heterogeneous materials by the thermal imaging method to increase the reliability of products made from them.
uh as you can see research tasks are the analyszis of the thermal uh anomalis of the product surfaces was carried out using the fractal approach which is invariant with respect or to the measuring scale. An analyszis uh of existing image processing methods was conducted.
It has been shown that the method based on the using of fractal geometry tools is promising from the point of view in application in uh automated system of active thermal imaging control is a productive made of nonmetallicinous materials and control of technological processes of their manufacturer.
An analysis of existing image uh processing methods was conducted. Table shows the characteristics of method uh that can be used for processing and analyzing thermograms. After analyzing the possibilities of thermographic image processing method, it was found that the method based on the using of fractal geometry tools isn't promising in terms of application in automated system of active thermal imaging control of products made of nonmetallicous materials and control of technological processes of their manufacturer.
If we present a graph of temperature changes along a line on the thermogram, we obtained a signal by which we can determine the presence of or abs uh absence of a defect in the structure of nonmetallic heterogeneous material. Such a series consist of sequential temperature values and in the absence of a defect will characterize random changes in this intensive parameter.
This method is used to anal assessing the complexity of temp and allows us to measure how row or smooth the temperature series is. A high fractal dimension may indicate greater variability and complexity of oscillations.
Uh this is an integral characteristic that contains information about the set of random geometric configurations on the thermogram that uh continuously change from point to point. In the same case when our certain conditional line intersects a temperature anomaly it may indicate the presence of structural defect. A certain trend will be observed on the graph uh of temperature changes along this line.
Such a series of uh sequential data will have the characteristic or uh invariance with respect to scale transformations.
In this case, statistical methods of analysis are uh ineffective.
Therefore, the issue of analysis and processing of thermographic images can be translated into the study of the fractal set. that describes the series of sequential values.
On this slide, you can see the fractal dimension calculation procedure. Fractal dimension can be used as the generalized characteristics of a fractal diagram regardless of the special scale and resolution of the measuring instruments.
Today it is an important metric that is based on the signal and image processing algorithm. A measure of control rawness which is often used for the crow classification and prediction. This method is used to analyze temperature changes when assessing uh the complexity of temperature series and allows us to um measure how raw or smooth the temperature series is. This is an integral characteristic that contain information about the set of random geometric configurations on the thermogram that continuously change from point to point.
The first step in the study by the box uh counting method is to determine the cell dimension in the temperature change graph along the line depicted on the thermograph. We will consider the specified graph as an informationational signal. The later taking into account the probability characteristics can be stationary or non-stationary. Fractal dimension indicates the degree of judginess of the time series. Speaking in a broad sense, the conditions for stationerity are limited by the requirements for the independence of the mathematical expectation and variance from the parameter of absisa a axis and the dependence of the correlation function only on this parameter. uh with the help of this correlation function that a random stationary signal can be sufficiently fully characterized.
The structure of random signal can also be established by a known distribution density. If we apply such classification fter on the types of informative signals when studying the series of uh sequential data that describe the temperature gradient along the line depicted on a thermographic image during thermographic inspectation of products made of nonmetallicinous materials then the studies grabs can be attributed to non-stationary thermogram signals. in the case when for example uh there are trends on them. In order of to test the proposed approach to persistent thermal images, an experiment was conducted to determine the fractal dimension uh of informationational signal with a known host coefficient. This specified coefficient uh is an measure used to time series analyzis. On this slide you can see algorithm for determining the ho coefficient.
To analyze the operation of the box counting method on informationational signals from thermogram.
100 fractal signals were generated using the method of sequential random summation. The samples for uh each of these signals contained 3 uh,000 observation results. The first coefficient of these of the first 10 studied signals is 0.1. The second 10 signals are 0.2. The third 10 signals are 0.3 and so on up to one. Each of these signals was analyzed using the uh cellular fractal dimension method. The results of the calculation are shown in the left figure. As we can see uh from the graphs above, the value of the fractal dimension responds uh probability uh proportionally inversely to changes in the value of the ho cofficient. Uh instability of is observed only at small values of the fractal dimension. To access the accuracy of finding a random variable, the values of expanded uncertainty with a significance level of 0.05 were obtained. The values of the uncertainty in relative form of uh establishing the fractal dimension for each consider considerate value of the hurst coefficient are shown in the uh right figure. The largest value of measurement uncertment was obtained for the hurst coefficient 0.5 and doesn't exceed 10 persons.
The dependence uh of the value of the fractal dimension on the type of distribution law of random stationary signal was studied. For this purpose, 10 signals with a sample size of three uh,000 values each were generated for the following distribution laws.
uniform, arc sign, triangular, relian exponential, lelas, benom, b model and kashi. Left figure shows the results of measuring the fractal dimension by the box counting method uh with an indication of the expan uh expanded measurement uncertainty with a significance level of 0.05.
The uncertainty values in relative form for establishing the fractal dimension for each distribution low considered are shown in right figure.
Thus summing up the above we can state a strong correlation between the value of fractal dimension determined by the box counting method and the horse index.
This indicates good identification properties for the specified method for determining the fractal dimension that is can be used to determine the fractal dimension of informationational signals which are also graphs of temperature changes obtained during thermal imaging inspectation of products made of nonmetallic heterogeneous materials. The expanded measurement uncertainty in relative form doesn't exceed 9.5%.
It is shown that the value of the fractal dimension of a random stationary signals is also determined by the types of its distribution which make it possible to classify and identify them by the values of fractal dimension.
Moreover, the expanded uncertainty in the worst case for the Kashi distribution didn't exceed seven persons in relative form.
Thank you for your attention.
>> Thank you very much for your presentation.
Is anybody wants to ask questions to the author?
>> No.
>> No. So we have a few question for you.
Uh the first one is how applicable is the proposed fractile analysis method to different types of non-metallic heterogeneous materials?
Okay, thank you for your question. Uh this proposed method of product control is universal for non- metallic hoginous materials. But as for products made of traditional metal structural materials using of the proposed approach faces difficults. First of all, it's due to the rapidity of thermal processes in metals due to high thermal conductivity and accordingly very high requirements for the dynamic properties of the measuring equipments used. I mean the infrared um equipment.
>> Okay, thank you. And the next one question is what is the accuracy of determining the fractile dimension of theoggrams?
>> Oh, thank you. Okay, I understand your question. In this in this work, uncertainty of change was used as the characteristic of the accuracy of the method under study. The largest value of measurement uncertainty does not exceed 9 uh.5%.
>> Okay. Thank you.
>> Thank you.
>> Thank you very much for your presentation for your answers. If there are any questions from the auditori so we can finalize uh the the first day of our conference and the first session.
So thank you very much to all our esteemed ga guests to all our speakers.
Thank you very much for being with us during this day and uh we hope that everyone enjoyed and uh we have a huge variety of different top of different topics for today and tomorrow we will continue with the session number two.
uh but for today I wish to all of us a calm and peaceful evening. Hope that today will be the same but we know that we can expect some surprises. Uh so thank you to thanks to everyone.
>> See you tomorrow.
>> Yeah, stay safe. See you tomorrow.
>> Thank you.
>> See you tomorrow. Thank you very much.
>> Thank you.
>> Goodbye.
>> Goodbye.
Related Videos
OpenHuman VS Hermes AI: Who Wins?
JulianGoldieSEO
285 views•2026-05-29
Long-Running Agents — Build an Agent That Never Forgets with Google ADK
suryakunju
142 views•2026-05-30
This computer is made from real human brain cells. And you can buy it.
Talktmsmedia
3K views•2026-05-28
BREAKING: Microsoft’s New Image Generating Model Beat Out GPT 1.5 and Nano Banana 2
aimmediahouse
122 views•2026-06-03
I Made the Same Anime Fight Scene in Every AI Video Generator
NobleGooseAnime
295 views•2026-05-30
Nvidia Bets Big On AI PCs | New Chip To Power Windows Laptops | Technology | AI Updates | N18S
cnnnews18
3K views•2026-06-01
I Tested NEW Opus 4.8 on Four Projects (Updated LLM Leaderboard)
AICodingDaily
298 views•2026-05-29
3D Platformer Update - NO CAPES
SolarLune
294 views•2026-05-30











