This week’s show is all about the DIGITAL TWIN! I sat down with David Cameron, Coordinator at the Sirius Centre for Research-based Innovation, at the University of Oslo.
This research lab focuses on the digitalisation of oil and gas operations and is (in my opinion) one of the most advanced research programs on infrastructure digital twins - mainly because the oil and gas sector has been doing ‘digital-twin’-type applications for more than 20 years.
In this episode, we’ll be answering some big question like:
- What can we learn from the Oil and Gas sector who have been developing ‘digital twins’ for over 20 years?
- What are some of the current research challenges being tackled by the Sirius Centre?
- How is the centre collaborating with industrial partners to put this expertise into practice?
Links & more information:
- More About the Sirius Labs @ University of Oslo
- Sirius Labs Beacon Projects
- Sirius Labs Research Programs
- David Cameron (LinkedIn)
- Will Needham (LinkedIn)
We are joined by David Cameron. There is the coordinator here with the Sirius Lab in the University of Oslo. Thanks for joining me, David. How are you?
Thank you. Yeah, very well. Very well. Yeah.
Excellent. Thanks for hosting me here. Lovely.
There’s a better view outside if the weather’s better.
Absolutely. And to start with, perhaps you could give us a bit of a background on the Sirius Labs .The research that you are undertaking and a bit of an overview of what’s going on.
Yes. Okay. Sirius is a construction called a center for research-based innovation, which is something that a Norwegian Research Council has done. Actually based on a model from Canada and Australia to try to get research to link together with industrial and community players to develop things which could be related to innovation. This is now the third round of the sort of centers. And what you do is that academics get together with an industrial consortium to propose the center as part of a call. In fact the deadline for the fourth set of these is actually this Wednesday. So there’s a lot of people being very busy about that. Sirius came about in a call in 2013 and we started in 2015 which was an interesting time to start because what the center looks at is something called scalable daughter access, which is about how do you get access to data on the value proposition that most people who try to make decisions in industry spend about 80% of the time getting data, checking data, making true of the data. They want to make better decisions, okay. So that was a fairly compelling industrial problem. But we started in 2015 when you may recall that your price had just fallen from $100 a barrel to about 30 which meant that was challenging to get the oil and gas industry to being the center but we set up with the interesting group of partners led by Equinor. We have IBM, as a large IT company, we have a cluster of smaller Norwegian IT suppliers and hardware people, and we have a company called DMVGL, which is a big services company. Since then, we’ve been able to… what we discovered in 2016 was that what we actually were doing was a very large part of something which became very trendy called digitalization. So we were fortunately able to say that research we do is a necessary prerequisite to good digitalization in the industry. And that has enabled us to define up an interesting research program around different aspects of oil and gas digitalization. And it also meant that additional companies have joined the center, such as SAP and Technique FMC and Darker Solutions. So that’s what we do. Computer science research to solve problems in the oil and gas industry is basically what we’re trying to do.
Great. Thanks for that. Interesting overview, I think. Can you get a sense of the scale here? How many researchers?
Yeah, the core funding we have gives this the capacity to put through about 10 to 15 PhD’s, but what we do is we bring together postdocs and people in existing research groups and other projects. So, we’ve actually got a cent a population of around about 50. And what we’re also doing is bringing in additional projects, for example, with just one, another competitive research program which will bring in another four researchers. So we’re hoping that will be getting up to around about 60 to 70 people by the time that the center is finalized. Last.
Okay. Yeah. You mentioned Started this in 2015 and what you started in became trendy in those years following. Yeah. I think, yeah, that’s one of the reasons that drew me here is your approach and your stance on some of the research I think you seem to be leading. So if you can tell us a little bit more about some of the core areas.
Yeah. Our concern is that so much digitalization debate is about the application of really trivial techniques to, frankly, some trivial problems and what we discover in oil and gas in any large complex industry, actually, is that the data you have is probably not huge amounts of data like you get in Facebook or Google, where you’ve got very what I call shallow amounts of data for billions of people. What you have here is rather complex, very, very structured data, about 14, 15 works. But it still becomes incredibly challenging about a data is what I call middle size data. You know, 30,000 things, which it’s too hard for a person to get their brain around but is not the big data that the Microsoft’s and the Facebook’s and the Googles are talking about. And what we want to do is we’ve got six areas of computer science which we believe can be motivated to be better with true engagement to oil and gas problems. The first one of those is actually something called semantic technology, which is about how you get computers to understand what things they’re talking about. So instead of having a table, a database, you know that this table in this database represents this feature of a pump, or a tank or an oil platform, then we have some database technology of people who are capable databases which were able to run really fast and effective. You know, that sort of data? Then we have people who are good at simulating complex systems. They started off simulating actual computer system, high performance computers about a colleague in the next office. He’s still doing simulation and design of how do you make a high performance computer work better, but you could see know lots of other complex things like plans, like instrument features on supply boats. We’ve got those three areas, we’ve got people who are doing work in language technologies, especially around the area of how you get language technologies to work when you’ve got a very specialized vocabulary. The problem you have now is that all the Google translators’ great on daily conversation. The trouble is, if you’re talking oil and gas, so you’re talking computer science, you’re talking medicine, you’re not in that space and the terms that means something in oil and gas don’t mean that in the Google world. And the problem is you’ve got very, very… got much smaller corpuses to learn stuff from. So we’re doing work around that.
And what are you hoping to achieve?
We’re hoping to achieve that you could actually get some sort of alignment between what is written down and what is actually happening in facilities. And we can also simplify things like saying we’ve got 50 years of standards, now those standard to just piles of paper that you throw at each other and it costs lots of money. If we can digitalize those requirements, we can save lots of money in the engineering industry. However, you got these and that requires some way of rationally going through all documentation at extracting out that information in a digital form. We’ve also got a group of industrial sociologists, people with social sciences background who are looking at one of the biggest challenges of digitalization of that we’ve got technology running out areas. But the challenges of getting implemented are related to how we think and how we work and how we actually relate to the world around us. And so they’re looking at ways in which we can, well, we’re actually following our projects with a view of trying to sharpen up, use the satisfaction and also look it if we can get some best practices on how you actually introduce these new technologies.
Great. Yes, definitely. Really cool concern, it gets addressed a little in your research program which is good, your focus.
Yeah, it’s It’s good. It’s good to have it in. And it’s actually being an area where we’re seeing a lot of interests. And we’re actually adding value in the area around interaction between technology and social sciences, and I want to actually see a lot more of that.
Yeah, you also, maybe not in the oil and gas sector, but if you expand into more general infrastructure, where maybe you have a big impact on citizens.
I’ll just tell you a bit about my background, I’m not an academic. I did a PhD many moons ago, but there have been industry for about 25 years and came back in to academia into this sort of coordination role about three years ago. Yeah, my background’s process, industry process, automation. So, it’s always been about, you know, we have to have a clear online real time observation of what’s happening here, otherwise it will blow up, and everyone will be really, really unhappy. Now, what I see with people talking about buildings and construction and engineering is that they’re coming to that situation now where they’re saying, we suddenly got measurements all over the place, can we do something sensible about them? And so what you’re seeing is that you’ve got to migrate and control system engineering companies, the Schneider Electric, Honey Waltz have always been engineer’s area, so they’re providing lots, lots of information to people. And so you’re starting to see that things which were applied previously in oil and gas in the process industries are now applicable in buildings that, whether it’s cost just fine, is a good question. And indeed, even companies, even company’s new on the gas sector, there’s a company called you Mayford Shell. They have a really, really good concept for their facilities, which is called an appropriate level of smartness. So they decide, if I have gone through all the things I own in the world and said that this thing here, which is producing six barrels a day, doesn’t need to be smart, whereas this thing here which is producing this floating energy facility, first of its kind in the world sitting on the terrace of Western Australia? Yes, he really needs to be smart. So we’ll justify that. And, yeah, So the construction industry is an area where I think that there’s a great deal of experience and learning from oil and gas that could be taken. And you see that in Norway in that project manager for the last big rail tunnel project in Norway had a background from the oil and gas industry, and a man who’s responsible for running the main airport in Norway actually had an experience… his background was all process industry.
Really interesting, and I’ll be interested to learn more about the collaborations that you have, in that case with the oil and gas companies here and some of the projects that you’re working with industry partners.
Okay. It’s very easy to do a research center if you just say we’re going to do some PhDs. What we wanted to do was actually try to capture the things we did. So we had a very, very you know, a two year process of was actually very frustrating, confusing sometimes because what would happen is that the industry people would say we want computers to do something for us, but something magic has to happen to that to happen and we were going and my computer science colleagues were going, well, we’ve got the semantic technology and there was a lot of time was spent trying to find where we could actually be useful for each other. Yeah, so in the end, we came out and define these research programs I was talking about, where each of those research programs produces some concrete themes you can use for something. You know, a simulator, a database to a set of semantic models of the world. But then we have discussed this forward to what we call beacon projects, which are fairly large areas where we believe what we could do to make the world a better place. And there are six of them. One of them we’re trying to computerize, provide computer I support for geologists in their work. Another one which is looking at all the subsurface drilling and prospect arc that you have any industry and how we can get expected access to that quicker, the third one which is looking at how we can improve data access around the actual management of how you produce an oil and gas facility, we’ve got a project around better planning using our analytical simulators to visualize and simulate plans so that people who are planning, for example, shipping of equipment to all regions of sea, get a better control on visualization of what’s happening and a better ability to respond to unexpected changes. We’ve also got a project around digitalizing standards. So much of engineering is about a company says “I want this to be built like this, here is a couple of 100 pages of detailed instructions on how it should be done” that is in printed out, sent off as a pdf to an engineering company who then, so it’s incredible amount of money and work and labor being used in doing something which is incredibly unproductive. And also, we reckon it can probably cut engineering costs by about 5 to 10% if you moved to a better way of sending requirements backwards and forwards. So we’re working with the industry around that and then I guess we come to last, which is the last speaking that I’d like to talk better, I think, which is the one we’re going to talk about today, which is about digital twins, which is about saying that everyone’s trying to build the digital twin now. Our argument is that where you needed help if you could actually try to structure in the dark in your digital twin in a sensible way so that you can expand it so that you don’t have to reinvent the wheel that we try to build.
Great, yeah, this leads us nicely into the digital twin localization and I’d like to, maybe get it going with, what are some of the areas of research in oil and gas that you work in this digital twin space? Is that mainly focused on optimizing the operations through simulation?
Well, it’s very interesting because you may have heard something called Gamma Heart Curve, a very well paid consultancy bureau, which says that these things are at the absolute top of heart, which gets most people in their nothing into talking about. And last year there was two things of talk for heart curve, digital twins and deep learning. So if you have a deep learning based digital twin, you’re absolutely at the top of these. Your head’s exploding. Yeah, now the oil and gas industry has actually had digital twins for a long time, and most of the process industries have had some sort of vision about digital twins ever since computers have been around. I used to work in the steel industry and my first job when I graduated was that I was made apprentice to the man who had the model off the plant, which kept the steel industry in Australia working. And his ambition in 1987 was say “Good. I’ve got this model. I want to put it online.” In 1997 you couldn’t You didn’t have the computer user interface. Mac had just been invented, and so we were still using VAX machines with foster screens. You connected up to control system. And yes, we were out to get real time data occasionally, and it was coming at 48 beats a second. So you had this vision of saying we’ve got simulation wouldn’t be good if we could actually connect it to what’s happening so we can actually analyze what’s happening because what was being done was that process would be down in people’s heads, took the data, and you ran the model, you said, how do I leak this model to make it fit to what’s happening and you’re doing analysis but it was all in people’s heads. It was only in the 1990s that control system has got to the point where they’re actually capable of putting data into databases, and it was impossible to take simulators and leave them up to facilities. And so the process industry has had real time optimization systems, which take city state models and analyze the performance of plants. And you’ve got more recommendations since the 1990s. The oil and gas industry has had dynamic models that are used to get with control systems to do very, very important jobs like control all the model phase, gas pipelines coming from an North sea that’s being available since around 2000 and that is one of the may technical for all the subsea developments in the Gulf of Mexico or North coast in Norway, you may have heard of your men longer field, which supplies some percentage of a gas in Britain that would not have been possible without a digital twin. We took a dynamic model of the 100 kilometers of huge pipeline, which went from the production of oil to land and we’re able to actually simulate alongside the process avoiding operational problems. So, in some ways, beyond gas industry, process industry, full of people who are like me 55 years old, blue shirts and bit too much of the gut saying, digital twins? We’ve had him for about 20 years. Well, okay, but it’s not really true because what you’ve had is specific applications, which would be incredibly laborious to build and which do specific, very specific things. Now we’re in a situation where you’ve got a lot of information out there that you’ve seen, that you’ve got access to the three things you need for digital twin. You start to get access to good design information and building, mean in engineering, it’s your design database. You’re starting to get access to some sort of measurement data. Now, given the quality of most things, measurements, most that out is going to be rubbish, in fact, even in the oil and gas industry, where you’ve got senses for 50 £60,000. But pop, you’re still looking at about only the ones which are being used to actually control plan to really very accurate. So you’ve got this set of sense information you got to try to fix, and then you’ve got a simulator or analysis tool that may be a physical model, maybe deep learning. So you need all these three things to fit together to give you the best estimate of what the thing you’re interested in is doing.
Yeah, you. Some of your search relies on physics models as well in some…
Not us, not us we’re dependent on people who actually know something about the world out there. But my personal background is that I’m not a computer scientist. I’m actually… my background was building the physical models. So I’m sort of now on the other side of it saying, How can we make sure that we supply the data to for people who do the simulations in the best possible way, which means we need to work together with chemical engineers or reservoir engineers, the people who actually own the analysis. We’re saying, look, you can do your analysis much faster and much better if you’ve got good IT plumbing in the bottom of a regular work.
Can you maybe give us an example of an oil and gas research?
Okay, within digital twins, we’ve actually just got ourselves a project that’s not going to start until next year, but it’s about the best practice for digital twins, and what’s exciting about it is it’s actually funded by the Norwegian government and Brazilian government as a joint collaborative research program. So we have got Equinor and Petro gas on both sides for table and what we’ll be looking at there is that all your company’s going out building digital twins. If you look at your company website, they’re saying, yes, we’re going to do the twins. That is going to be a good thing. They’ll go out and buy them using commercially available products. Now what we’re trying to do in this research program is to say good, but what you’ll get is something that’s fairly disintegrated. It’s going to have lots of lots of point applications. Can we go in and find ways by building a common conceptual model of an oil and gas production facility, which fortunately is much the same in Brazil as it is in Norway? But everything’s got different names. Everything’s got different ways of describing it. So what we’re going to do in that project is build up a common conceptual model, a sort of semantic backbone which could be demonstrated to be applicable both in the North Sea and in Brazil. And together with our Brazilian research partners, that project’s going to have six PhDs, 4 or 5 postdocs working on problems around different aspects of this that had been linked models to it effectively about how you can also track the history of a facility along the lifestyle. Because he important thing about both the plant we’re looking at is that they’re still under construction, a bit of it there that is going to be another five or six bits of it put in the next 10 years.
And would you simulate the effect of bringing those online? Is that something that you should do?
Well, we would enable them to simulate the effect. You know, they’re going to have to do that to do the design. What we want to do is we want to provide the best practices and the computer glue to, um, able to do it twice as effectively, as they currently do it.
So you mentioned this semantic backbone, is this a concept that you worked on? (Yeah, exactly. Exactly, what you see is that people have been trying to build up an unambiguous description of the oil and gas reality to structure information for about 20 years. There’s something called Arta 50926 which at the moment is a very large dictionary. We’ve been working for a new version, Arta 50926 which enables computers to actually do reasoning calculations around it to work out about linkages between things so that it becomes computer graph, not just a dictionary. Whatever challenges we’ve had with the sort of semantic integration projects is that there’s being a problem with lack of quality of some of the software and system supporting it, and has also been a bit of an approach that everyone wants to produce a perfect model. Whereas what you would need is will produce a model which demands to be function. So we’re going to try to be instead of coming top down and saying, we come with the perfect description of the perfect oil and gas platform. For one you have in heaven will come and say, Look, okay, you’ve got the systems, let’s find a pragmatic model that we can demonstrate works from both your systems and the other systems. Now, having been in the industry for 15, 20 years, I’ve got a reasonable hope that, you know, everyone’s using vaguely the same sort of tools describing the same sort of equipment so that we can come to some sort of consensus reasonably quickly.
Very interesting. You mentioned there about the integration. Are you talking about the integration in smaller system?
So, your model that small level subsystem, maybe, then aggregate those systems on and do some sort of reasoning.
Yeah. I don’t think any digital twin will ever have the same implementation structure. What we need is some sort of layer of the top which ties together the bits you choose to make it relevant because there’s also a danger that you build a digital twin, just the pick of having a digital twin. That’s point, the manager should buy that sort of system. What I need to do is actually try to have a strategy at the back of the head. Yes. We want to move in the direction so that we in the end have this sort of level playing field of access to data. You know, we don’t want to continue to build lots and lots of little side applications, but you’re not going to get the money to do your projects unless you say I want to do something specific with this particular digital twin application, I want to decrease the time to first oil by 3% through getting better access to this information. So you’re necessarily going to have to build up these sort of mushroom projects. But if you have some sort of common semantic backbone, those mushrooms have a place to grow Rather than just being things which have grow in their own corners and paddock, it sort of makes sense.
Yeah, Yeah, absolutely. Like since you published many papers about this decision-making.
Yeah. We have worked with a company called Iver on reasoning around piping. You know, if you build on our platform and requires many, many tons of piping on, and probably about 15, 20% of a couple of cost off, your platform is still piping, so it’s not very sexy, but it’s an incredibly important area for saving money. Now, Iver said, okay, where we need to have a description of what piping is. Now, that description is currently in stacks of paper and the standards, we’ll digitalize their standards. Well, then link that up to our databases around what we can by purchasing. It’s a very narrow but actually by building a circled ontology, a semantic model of world of piping. They’re now able to have an automatic application which enables in to make sure that the piping they have in this part would meet the requirements of customer, that they ordered the right amount of material, that they avoid ordering the wrong type of material, that they can, in effect, make their engineers 10 to 15% more productive. Now that approach can be taken into our areas of engineering. And what we have there is a demonstration that it’s actually is worthwhile to build a semantic model and link it to your databases. It’s a very good project because it shows a clear operational implementation of semantic technology. And what were good I think was that it was aimed solving a very, very pressing business problem. It wasn’t phrased as “let’s try to get all our dock together that we can see what could happen,”
I noticed in the same report is the section about how University of Oslo and Sirius Labs working towards the UN SDGs maybe in a factor kind of way but working on the gas projects, remember?
Yes, I wrote that because it’s very clearly sure before we got a very clear message from University administration and also we ought to really start to, when you’re not aligned to the UN SDGs, you have to forget it. But in Norway, the oil and gas industry has been, for better or worse, the driver of technological development in the country last 30 years, the people you’ve got into the oil and gas industry are the brightest, the smartest in engineering, the technology world. I was talking to a man who was senior executive in Total on the project engineering site director of a field development, and his field actually was in health informatics. But it was 1985, he finished his PhD, where do we go? Yep, oil industry. Now that is competitive advantage, and what we’re doing is completely relevant to offshore, to large capital works to the tunnels, bridges, dykes, things you have to build as part of the infrastructure that you’re going to need to create to adapt to climate change. We’re going to continue to have to make the products, we’re going to have to continue to make smart manufacturing in Europe. So what we’re doing here is by making the Norwegian engineering and energy industries more effective. We’re supporting your sustainability goals. It was very easy to go through steadily goals and say that, yes, what we’re doing concretely supports these goals. It is reducing the costs of things, you’re getting better handle on what’s happening in the world, we’re collaborating internationally, we’re working towards the computer science that you need to create the most.
Like you mentioned, the technology is hopefully general enough to be used, and it’s related to different types of information.
Because computer science is full of computer programs which could be used to solve all the world’s problems. If you use a six years of consult and effort to build the solution. Computer science is basically a raw material that you use to trudge used to provide solutions. So what we’re trying to do here is to try to bridge that gap so that we can use some solutions to make sure that computers to drive research in computer science so that we can actually capture some hard problems. So often you see that a computer science paper will talk about some sort of toy problem. Samast Technologies is professors at universities and pizzas. Now it’s always frustrated people in the industry that you don’t I think it’s also frustrating academics that they’re not hitting real problems. Sometimes the real problems are so messy that you wind up not getting to the research problems. But at least we’ve got the privilege here. If we’ve got 8 years to actually build up some mechanisms to demonstrate that engaging with real industry problems in these six areas is actually I able to drive some good, fundamental research.
Really interesting before we move on to something else, do you have examples of these avenues into other infrastructure projects? What were your maybe one of projects or worked on some research that the office steps into?
Well, at the moment we’ve got two beacons which are called cross domain where we’re working with other research groups and one of them is in health, where we see that the approach that you’re using to project execution, information management in the oil and gas industry are actually pretty adaptable to the issues required around hospital management, management of information in hospitals. And were involved in a beacon projects, lighthouse projects from the Norwegian government called Big Med. Another area is in the environment, so monitoring and observation where we’re collaborating with another center for research-based innovation called CFR for which is about article for observation. A common thing there is that Equinor is in both centers. And so we’ve got jointly supervised PhD students working around that semantic problems related to taking data from large satellites and tie it together with memory in observation data. So those big concrete areas. In the long term, we’re also working at the European level oil process industry. And it’s also been reasonably clear to me for a long time that the construction industry has basically the same sort of problems and the same sort of challenges and issues of integration of complicated business systems. You’ve got a okay somatic description of reality for the bit that sort of serves that purpose and it does need to be structured a bit more. But then you’ve got the challenge that everyone’s now selling smart bits with dodgy security and doing their own semantic description of reality. So it’s going to be a lot of challenges related to trying to link that together. And they I set a meeting on Oslo Innovation Week had a kick off of it yesterday. People were talking about building a digital twin of a city. It’s going to be very interesting to see how sustainable that sort of model will be. And the other thing you gotta remember about digitalization is it could be as digital as you want, but if your bus breaks down like mine did on the way into work today, you’re still stuffed.
Very interesting. So what are your reservations about the city digital twin?
It’s getting to be, too, to be grandiose. Let’s start with some building digital twins, we’ll find out that it’s going to be useful. At the moment it seems to be so much about violent society and people are talking about surveillance scenarios without having any ethical… seem to have any ethical qualms about it. And, uh, yeah, there are very obvious areas where a good control monitoring simulation system is going to do really good things very well and we’re getting there along. If you talk to the public transport providers or the people who are responsible for running large systems, that’s a place we can do stuff. The biggest challenge lies in all of these areas, is that you’ve got 30, 40 year old computer technology, and you’re still going to have 30 to 40 year old computer technology in 20, 30 years. So it’s going to have to be something that is capable of revitalizing this sort of legacy systems.
Great. Very interesting. Okay, we’re slowly running out of time. I have to finish with one last question. You mentioned about your 8-year program? 2015 to 2023. What are the goals, what do you want to achieve by year?
By 2023 I want to have had a successfully run or well underway large demonstration project in each of our beacons that, you know each, uh, beacons is either showing to the industry, yes, we’ve got this collaborative project working, which is producing good results for you, and that’s going to require external funding to get the number of people to work on these things. And so we can, in 2023, turn around and say, “look, here’s some good stuff that we’ve done that demonstrated that you could do things differently in the oil and gas and other industries and has advanced the core of computer science of it.” What we then also want to do is… we’ve just had a midterm review. We were a bit early to have midterm review, but just being given our authorization to continue for the full eight years, and everyone seems to be very pleased with what we were doing. I believe there’s an area which is very important, which I would call industrial computer science industrial, informatics. It sort of seats between computer science and automation, and it’s about how you make large, complex industrial systems work well and it’s not just the automation guys, it’s not just the engineers, it’s not just a computer scientists, it’s not just sociologists. You need to have centers who bring together these things, which cultivated discipline of making computers science relevant to the industry and I hope that it continues to be an industrial power in the next 15, 20 years. So, yeah, that’s my sort of dreams. The stuff about this sort of industrial dramatic center are not official. That’s sort of my thoughts. But the ambition to have six large demonstrated projects is actually a success criteria for us.
Thank you very much. Very interesting times ahead of you.
Yeah, I hope so. Yeah. Thank you so much. Okay, I’d like to also acknowledge that I’m just the industry guy and so all these wouldn’t work without the director Arild Waaler and his deputy, Ingrid Chieh Yu, Martin Geiser, Guy Horn and Thomas Estili…
Editor's note: the intro music for this podcast is kindly provided by Vidian. Also some links may be affiliate links, which means if buy something through that link it will cost you no extra, but I receive a small percentage from Amazon. It helps to support the show, so thanks!