Could the key to transforming workplace safety lie within the realm of virtual reality? In today’s episode, we explore the exciting domain of digital twins—a technological marvel with the potential to overhaul how we approach operational safety across various industries.
Using the paper, “Digital Twins in Safety Analysis, Risk Assessment and Emergency Management.” by Zio and Miqueles, published in the technical safety journal, Reliability Engineering and System Safety, we examine intricate simulations that predict traffic flows to emergency management tools that plan safe evacuation routes, and we delve into how these virtual counterparts of physical systems are redefining risk assessments and scenario planning.
As we navigate the world of operational safety, we discuss the diverse array of models—from geometric to sophisticated hybrid simulations—and their groundbreaking applications in forecasting fire spread and optimizing evacuation procedures. These digital twins aren't just theoretical concepts; they're powerful, real-time lifesavers in emergency situations, emblematic of the future of safety science.
Discussion Points:
Quotes:
"Ideally, a digital twin is a complete virtual copy of a product or service that is an electronic simulation that is completely accurate compared to that real product or service.”- Drew
“One of the first documented digital twins was in the aerospace industry - NASA [used it] during the Apollo 13 program.” - David
“this idea of having a complete digital picture of the thing that you're building is becoming fairly common, so that lends itself very much towards using it for things like digital twins.” - Drew
“we may never quite know exactly how different the digital twin is from the physical object itself. That’s the challenge.” - David
Resources:
The Safety of Work on LinkedIn
You're listening to the Safety of Work podcast episode 117. Today we're asking the question, can digital twins help improve the safety of work? Let's get started.
Hey, everybody. My name is David Provan, and I'm here with Drew Rae. We're from the Safety Science Innovation Lab at Griffith University in Australia. Welcome to the Safety of Work podcast. In each episode, we ask an important question in relation to the safety of work, or the work of safety, and we examine the evidence surrounding it.
As always, when Drew and I are preparing the podcast, we go back and forward with paper ideas. It usually involves me suggesting something and Drew thinking that it's not quite good enough quality for the podcast. But Drew, you fired back this paper last week. Before we get stuck into it, what was interesting about this paper that you thought might be worth talking about on the podcast?
Drew: David, the main thing is just that I've been seeing this term digital twins cropping up. I wanted to make sure that I knew a bit more about it before either endorsing it or dismissing it. I don't know about you, but throughout my career, one of the holy grails that I keep running into is this idea of fully automated or real time risk assessment.
Instead of having humans trying to do these risk assessment practices that don't necessarily do anything that they're not already doing, this idea that we could have some system running in the background that measures the amount of risk and tells us objectively whether we're headed into danger and gives us warning to get back out again. I had this vague idea that digital twins were the latest iteration of this idea, but I didn't really know enough about it. I thought, okay, here's an opportunity to just get into it, find out what's going on, and find out what the current state of the technology is.
David: Right. I was really interested too. It's not an area where I'd naturally or normally go and read about, so it was fun to have a bit of a read-about about this world. Let's talk a little bit about digital twins and a bit of background to them before we maybe talk about this paper. Do you want to talk a little bit about what is a digital twin and how we might use them?
Drew: Okay. I think we should caveat here that there's a ideal version of what a digital twin is meant to be. As we'll get into the paper, not everything that's called a digital twin actually matches this idea. Ideally, a digital twin is a complete virtual copy of a product or service that is an electronic simulation that is completely accurate compared to that real product or service.
If you imagine, there's the real world and there's the matrix. The matrix is an exact copy of the real world running at the same time doing the same things. But the idea is that because you've got this virtual copy, you can do things to the virtual copy that you can't do to the real world. For example, you can set off an explosion and see what happens. More realistically, you can make changes to the virtual copy. You can run it a little bit faster and predict what's happening in the future, or you can adjust something and see if that makes things safer or less safe.
David: Drew, in thinking about this paper, I've read a lot in the last couple of years about traffic flow modeling as well, taking a highway or a section of road, creating a digital twin version, and looking at different vehicle volumes, speeds, and looking at the impact on traffic, traffic flow, and things like that. Is that one example? Are there other examples that our listeners might be familiar with about digital twins?
Drew: There are certain areas of engineering, where people have always used a lot more modeling. Traffic flow is one of them, fire engineering is another. In water infrastructure, they've for many decades, used finite element analysis. I guess the difference is that instead of trying to do engineering by calculation, you're trying to do engineering by accurate simulation.
I guess the step that they're doing with digital twins is having it running simultaneously in having this back and forth exchange of information between the real system and the twin. The twin is always matching the real system as opposed to just being a model that you use during design or a model that you use to predict the future generally.
David: Which is where it gets to that point you made earlier about potentially having this real time dynamic up to date view of the system and hopefully the risk of the system. We'll get to that in a moment, but I was also fascinated reading in the background or the introduction to this paper, one of the first documented digital twins was in the aerospace industry, so NASA during the Apollo 13 program.
Some of our listeners might have seen the Apollo 13 movie with Tom Hanks. They actually had simulators on the ground that mirrored the state of the crewed spacecraft, which actually played a significant role after that in-flight explosion that damaged the main engine. They were actually able to, on the ground, use this digital twin simulation to explore their contingency strategies that, ultimately, they were able to put into place on the physical vehicle itself and bring that crew back to Earth after not being able to land on the moon.
I thought that was a fascinating look-in, because some of the work that I've been lucky to be involved in at NASA is because of the necessity of the innovation that NASA needs to do to do what it does, it actually helps advance technology for the rest of the world. Things like laptops and actually were a result of payloads and computing power required for the space shuttles.
Drew: Yes. I guess that's part of what they can do with digital twins. One of them is this simulation aspect. On Apollo 13, they were only going to have one opportunity to repower up the module. But on the ground, they could try it 50 times until they got it right and then pass those instructions on to the astronauts.
The other thing is just routinely during space launches, they've got all of the things that are going on in the shuttle also going on in the ground, and they can very, very quickly spot when things start to deviate by this constant comparison between the simulation of how it's meant to work and the actual feedback from the instrument.
David: I think based on the conversation that we've already had so far, hopefully our listeners are forming a view of starting to actually think about and see how this might be the use of potentially useful applications for safety management. Before we go into that, there are two different models they talk about, a three dimensional model and a five dimensional model. The three dimensional model is that you've got this physical object. This is the actual physical system itself. It might be a plane, it might be a electricity network, it could be a highway.
You've actually got this physical twin, and then you create this digital object, which is in the virtual space, you create a digital version of this physical system. You've got the third dimension, which is the twinning process. How seamlessly you can integrate and flow information from the physical twin into the virtual twin. They call that the twinning process. Drew, is there anything else you want to say about how digital twins work before we introduce the paper?
Drew: No, I think that's pretty well for background.
David: Great, so are you happy for me to introduce the paper?
Drew: Yeah, please do.
David: Okay. It's a recent paper, it's a 2024 paper. The authors are Enrico Zio and Leonardo Miquelas. They're both connected to the energy department of Milan in Italy, but also to the Center for Research into Risk and Crisis Management at Mines Paris. That institution in France just seems to have a whole bunch of collaborative research centers and research partnerships. I think for a long time in the resilience engineering space, it has published some really useful safety research.
The title of the paper, Digital Twins in Safety Analysis, Risk Assessment and Emergency Management. It's being published in Reliability Engineering and System Safety, which is a very large technical safety journal. Drew, is there anything else you want to share about the background of the paper?
Drew: No, except perhaps just to mention that this is a systematic literature review. The idea is they're trying to be pretty comprehensive using a set of fixed search terms to discover all of the papers that match those search terms. Because digital twins are fairly recently, they've restricted it to just basically the last five years. They're looking for all papers that use the term digital twin paired with something relevant to safety, so risk assessment, safety analysis, accident, emergency, and things like that. They're basically creating a landscape saying, what generally is now been published and said about digital twins in this safety space?
David: They had five research questions that they were looking into. They wanted to understand the expected tasks or functions that digital twins can be applied to safety analysis, risk assessment, and emergency management. The second question was around the technique. What techniques and approaches are being applied for the implementation in these areas?
The third question was, which types of model representations are being applied? That's a little bit more the technical aspects of the different types of digital twins. Research question four was about the aspects of the twinning process that are being taken into consideration, and then research question five was about limitations and challenges.
They really wanted to say how and in what ways are digital twins potentially being used in the published literature? Like you said, Drew, they did a whole bunch of search strings. I was surprised at how many publications I found.
After they went through their selection and their evaluation criteria, which did rule out a whole lot of different potential studies, they still found 40 papers. In relation to digital twins and safety analysis, they found nine studies, digital twins and risk assessment, 24 studies, digital twins and emergency management, seven studies. Drew, I was surprised there was that many digital twin safety publications that met their criteria in the last five years. Did that surprise you?
Drew: I've only run into the term fairly recently, and I have a little bit of a suspicion that not everyone who does this thing calls it digital twinning. It's possible that there is other related work that's being missed just because it's being published under totally different labels. As far as the concept of digital twin itself goes, it's a bigger thing than I realized.
David: Yeah, it's a good point, Drew. You mentioned about the search string. When they were searching, unless the words digital twin appeared in the title, abstract, or key words of the publication, then it would never be included in the initial scan. That's a good point about language. Perhaps that is one limitation of this study. Maybe they need to think about other synonyms for digital twins that people might be using in the literature.
Drew: It's a trade-off because you try using a term like simulation or real time risk assessment. Suddenly you've got a thousand papers, and it's not at all specific in what you're looking at.
David: Let's talk about these findings then, Drew. Some of the expected tasks and functions, do you want to give an overview of what they found in relation to how digital twins are potentially being used?
Drew: Okay, this is just basically what people are trying to do with digital twins. The first thing they looked at is just what industries are using them. There was a pretty big range, but the main topics that came up were construction, naval engineering, and manufacturing, which I guess shouldn't be surprising because these are industries that already use a lot of digital models. When you're using a digital model already, then you start to think, okay, what else can I use this model for?
An example is in construction, building information systems have been a big thing. There's a few different terms there, not just building information systems. I think there's building integrated modeling. The general idea is this idea of having a complete digital picture of the thing that you're building is becoming fairly common, so that lends itself very much towards using it for things like digital twins.
Some of the typical things people are doing are things like there's a paper about creating a digital twin of a crane as it's being constructed, and then calculating the stability of the crane and providing warnings in real time as the digital model is built along with the real model of the train. Creating a digital twin of a construction site, and then having cameras around that site identifying mismatches between the site as it's being built, the site as it is in the integrated system, and then using augmented reality to then report those hazards back to the construction workers. Creating a digital twin of a production line, and then playing around with different changes to that production line to measure how it improves or whether it creates any safety or logistical difficulties. That's the broad types of things that they're doing.
They actually came up with a long list of different things. We probably won't read out the long list, but some examples are calculation of safety influencing factors, safety assessment, stability analysis, identification of hazards, real time visual warnings, risk identification, quantitative risk assessment, risk and reliability assessment, and pretty much anything you do in terms of safety analysis. There's probably someone who's trying to use digital twins for it.
David: I thought some of these things I was familiar with, but I guess many of these things wouldn't be immediately considered in a safety management system, let alone a risk management standard or risk analysis standard in the safety management system. Maybe that's coming for our safety management systems now and into the future.
You mentioned a lot of those bullet points. In this paper, there is quite a lot of detail. I like the way that they presented their findings in easy to understand large tables. The good thing is the paper is open access, so you'll be able to follow the link in the show notes and grab a copy if you want to look into all of these applications in detail.
Drew: Yup. David, I guess if there's a consistent theme here, these are all things that you would not be surprised to see appear in a system safety or safety engineering analysis of a new design. But through the digital twinning, they're being pushed towards operational safety. Not just being done at the desk before the design, but being done during the construction or during the operation of the tunnel. That seems to be the general aim here is to use these more engineering based analyses, but make them available to much more operational scenarios.
David: I think in these operational scenarios, I'm very familiar with industries like utilities that are establishing full digital twins of their network, mainly from an asset management perspective to understand capability, capacity, and how to best optimize demands on different parts of the network and capacity, and things like that. You may be in an organization as a safety professional, and you actually may have your asset management team or your engineering team actually investing very heavily in digital twinning for non safety application optimization purposes that might give you a perfect launchpad to introduce the idea of actually using the digital twin potentially for safety analysis.
Drew, research question two was about the main different types of models. You've listed a couple here. Do you want to talk briefly about the different types of models that are being applied to these different applications?
David: Okay. This is the heart of, how realistic are these digital twins? They mention a few different types of models that people use. The most simple one is a geometric model. A geometric model is like an exact physical match, but it doesn't have the dynamic aspects to it. You could imagine it as being a really, really sophisticated diagram of the physical system.
The next layer is the idea of a physics model. If you've ever played a computer game like Angry Birds, that's a really simple example of a physics model. It starts to incorporate things like gravity and collisions so that the system behaves like a real world model. The more laws of physics you build into your model, the more it behaves exactly like the real system. That's a clear sign for the limitations because there's a heck of a lot of physics, and typically people are very selective about which physics they include or don't include.
In fact, most of these models are not full physics models. Mostly what they use is geometric models or physical models for certain well-defined things. I guess in engineering, I'd consider them like traditional models, things like state based models or machine learning that are really just calculating or interpreting what's going on rather than creating this full fidelity twin. In the case of machine learning, you wouldn't be able to step inside the matrix and see what's going on. It's just drawing conclusions, which is really quite different from this idea of a full twin.
David: I personally haven't played Angry Birds. I have played Mario Brothers though, so might be the same. I remember when doing some work in offshore oil and gas a long time ago, nearly 20 years ago, we were doing pushover analysis on offshore platforms to do with force of waves, wave height, frequency, and things like that. Is that what you were referring to in terms of a physics model, where you actually create a model of the engineered system, and then you can apply different parameters to it and see what may happen to the physical system?
Drew: Yes, I guess that's a good indication of the difference. In order to calculate whether something's going to fall over, you could just do your turning moment calculations. There's no real twin there, there's just this calculation. You could try to create this exact electronic copy of it, and apply the laws of physics to each part of that model and see what happens, and be able to even tell if it's split in half. You'd still be able to keep modeling and see what happened. The reality is that most of these digital twins are a mix between the two. They're sometimes using one, sometimes using the other depending on exactly what they're interested in.
David: Okay, great. After the physics model, do you want to talk about any of the other types of models?
Drew: I guess most of these fall into the category of what you call a hybrid model, which is to say they're using some mix of calculations and the realistic simulation.
David: Yeah. I've seen some interesting ones in the emergency management space looking at emergency evacuation of buildings in these digital twin models and actually simulating the fire propagation, how the smoke will flow based on the ventilation system, where the people are going to be located, how the people are moving and the time it's going to take to move people. Some of that is quite advanced now and I'm assuming in the emergency management papers that this study referenced some of those types of case studies.
Drew: Yeah. Emergency management has always been one of those spaces, where they've tended towards this modeling of all of the individual people moving around. You can see on their diagrams little dots of people moving through the building, how they might react in terms of crowd behavior, where the smoke would go, and so forth.
It's particularly important in emergency management to be able to have this realistic simulation because you need to be able to say, okay, what if part of the building falls down? How does that change the way everything behaves? In your simple calculation based model, you can't do that. The goal of some of these things is to say, okay, could we dynamically change the lighting in the building and change the directions we're giving to people depending on what's happening to the building in terms of fire and collapse?
David: Like real time evacuation plan A, B, C, D, E, and real time changes to that evacuation plan based on what's the specific context of the emergency situation itself rather than the very static, I guess, signage, directions, and drills.
Drew: Exactly, and it can work in both ways. You could have a customizable emergency management plan, but you could also run 50 simulations of 50 scenarios, and go back and redesign the building and redesign your way of approaching emergencies to be able to handle that flexibility.
David: Drew, we've talked about the applications and the types of digital twins. Research question three was about the specific techniques and approaches. Do you want to talk a little bit about some of these more specific techniques and approaches to digital twinning for these safety applications?
Drew: Okay, research question three is basically about, how do you create the digital twin? Research question four is about, how do you maintain that digital twin as an accurate twin? They've got a massive list of different types of techniques that people use about how the twin is created. Broadly speaking though, they fall into the categories of using models that would otherwise be available. This is where people are using models already as part of their engineering and design, copying those across, using things like machine vision, putting up cameras or drones with cameras, basically taking a direct picture of the physical layout, and using that to create the digital twin.
Most of them fall into some form of machine learning. There's obviously a lot of interpretation of reality that is going into creating this twin. It's not the case of you feed your building in and get an exact copy back out. You feed pictures of your building in, and it goes through a lot of analysis in order to produce the digital twin.
In terms of maintaining it, they're talking about using things like wireless sensors, Internet of Things enabled devices, or again using cameras to keep a picture of the thing. It's here where there's a lot of challenge and difficulty because obviously, you're not going to be able to put sensors on every single component of your real system in order to translate that data across. There's always going to be a lot of interpretation between how you're monitoring your physical system and how you're communicating that to keep the digital copy up to date.
You can imagine how hard that is in an emergency. What happens if one of your walls catches on fire? It's not like the sensors in that wall are going to still be there accurately reporting the state of that wall so that you can keep the digital system up to date.
David: Yeah. I think also, Drew, the assumptions. If we think about engineered systems, the assumptions that you'd have to make about degradation rates and other impacts over time on components of that system. We'll talk about limitations and challenges shortly, but I think what you're alluding to there is that we may never quite know exactly how different the digital twin is from the physical object itself. That’s the challenge.
Drew: Yes. This is where the advantage is supposed to come from. The moment we start making lots of assumptions and lots of interpretation, then we lose the advantage of having the digital twin in the first place, and we may as well just do it as a set of calculations rather than try to do it as a twinning process.
David: Drew, are we ready to talk about some limitations and some challenges?
Drew: Yes.
David: Okay. The paper lists out six different limitations and challenges. How do you want to talk to these, Drew? Do you want to group them together or just go through each of them?
Drew: I think we could just run through them quickly. They're all things that we have touched on so far.
David: Let's just summarize these limitations and challenges, and then we can talk about some practical takeaways. The first limitation or challenge they talk about is uncertainty quantification. These digital twins require this stream of data that's necessary to, I guess, remove the uncertainties that the digital twin may not be a good representation or accurate, complete representation of the physical object. I guess the challenge is really about just how much uncertainty or variability is there in the actual functional output of the digital twin compared with the real physical object itself.
Drew: Yeah. The old risk assessment question is your risk assessment gives you a result, but what's the risk that the risk assessment is wrong? Digital twins don't have a magic solution to that problem that we always face in risk assessment.
David: The second challenge they talk to is about consistency and persistence. This is this digital twin consistency. You mentioned it just before that it's critical for risk based decision making. You need some sort of guarantee that the information inside mirrors the current state. Just how real time and how complete is it? It's a little bit like the first challenge there, but I think this is more about just how you keep the thing up to date.
Drew: Yeah, and I guess this second one is more of a technical problem that is getting better with technology. In particular, as we just expand the number of sensors we can put on things, we get better and better. I think my favorite example here is the smart paint that they put on aircraft that can detect the amount that metal has flexed to update the metal fatigue. We're always getting better at being able to measure the current state of things. That's why things like digital twins are now so much better than the types of modeling that we had 20 or 30 years ago.
David: The third is around cyber security and threat identification because there's so much data. For those who work in some of the industries we've spoken about, these economically critical infrastructure like utility networks and other public infrastructure, this idea that you're now putting a whole bunch of data about how your system functions and I guess how to mess with that system online. Clearly, the security of that digital twin becomes a foremost consideration.
Drew: Potentially, we have created a perfect copy of our building, everyone inside it, and what they're currently doing. A few minor privacy concerns there, particularly since a lot of what we're using to get this data with is cameras. We actually have a really quite serious attack surface that we've created by this data that could be used for purposes that are a lot more nefarious than what we're trying to do with the safety assessments.
David: Challenge number four was this virtual to physical twinning process. We specifically talked about risk assessment and emergency management applications. These are communication protocols and seamless connection between the two. There's just some limitations that they call out there. I don't know if you want to say any more about that, Drew.
Drew: I think probably all three of these next things fall under that general category. There's the twinning process, there's standardization, and there's interoperability. These are just your ongoing technical challenges. If you have to create a bespoke infrastructure just for your safety, then probably it's either never going to happen, or it's going to happen as this shadow of your other systems. The more you can have seamless integration between the things that you're using for your safety analysis and the things that you're using for your day to day operations, then the more accurate the safety model will be and the less overhead safety provides.
The ideal version is just whatever we're currently using to monitor and manage our systems, we use that to do the safety analysis on. There's no need for this extra system and extra protocols.
David: I guess we've been through the applications of digital twins to safety management activities, the different types of digital twins, how the twinning process is managed, and some of these limitations and challenges. Drew, is there anything else you want to say? The paper goes into two illustrative examples. I know that you love diving into the rabbit warrens provided in these podcast papers. Is there anything else you want to add before we go into practical takeaways?
Drew: Let's not take a deep dive into the examples. I might just mention one of them just as an idea of the promises of where this might get. Imagine that you are a building constructor. You have a building that you're putting together out of prefabricated components. You have a plan for how it's going to work, how each step will be done, and there is a system in the background which is completely shadowing that building as it's built.
The moment someone makes a mistake and they don't put the bolts in correctly, or they position a component not where it's supposed to be, that system in the background monitors how much you should be worried. Is this affecting the stability? Is this affecting the future stability of the system? It creates an alert and warns you of the danger.
How many people have been killed putting buildings up when concrete panels have collapsed or fallen over or when something that looked stable wasn't stable? That's the idea. In real time it tells you something that is going wrong, points this out to you, and we prevent the accident by being able to fast enough, do things that normally require an engineer sitting at their desk doing calculations as opposed to just constantly monitoring. That's the idea we're trying to work towards here. It definitely does have promise to be able to do not everything in that space, but some of those things do appear to be realistic ambitions.
David: All right. Let's talk a little bit about some practical takeaways. The paper didn't really offer too many. It wasn't that type of a paper, but I've got a few points here and then love to get your thoughts thrown in as we go. I guess the first takeaway is that this digital training process provides a real opportunity to maybe strengthen our safety management applications like safety risk assessment and emergency management by being able to create a simulation environment of our system, which should give us an opportunity to run different scenarios and perhaps maybe make our particular risk assessment practices a little bit more of a closer representation to the physical system that we're assessing.
Drew: I would say that there's no strong evidence offered here that at the design side, this is doing any better than we are already doing through our design safety. The promise is that we could shift some of those heavy duty risk assessment things and apply them into new areas where we're currently not doing very robust risk assessment.
David: The second one I came out with is this real time support for emergency management. You mentioned the design pieces. We can see the opportunity to do point in time assessments and design emergency management plans based on a better understanding of the system itself, but this real time support for emergency management, I think, is quite a promising area to think about because emergencies are unique, they're dynamic, they're not something that we can predict how they might escalate and go. If we can get real time decision support for emergency management, I'm sure that in wildfire management and other applications, even major natural disasters, that understanding, in real time, what's going on would be a good thing to be able to do.
Drew: Yeah, that's definitely an error of promise.
David: I had in here about informing decision making particularly around operations, maintenance, asset management as well, and just the safety and integrity of the system itself. I know that's quite advanced in some industries like oil and gas, and you mentioned aviation as well, but I think that just continues to advance our ability to understand the current state of our systems and make decisions around how we manage those systems.
Drew: Even though they restricted this literature review specifically to safety, even within this review, it offered hints that this isn't really just a safety thing. This is more of a general approach to managing operational decision making. That really starts to blur the lines between what is safety and what is just normal operability and maintenance.
David: I guess I'm happy for those lines to be blurred and gray because safety isn't a sideline activity, it's a core operational activity. Drew, I wouldn't mind. This practical takeaway, even though we call them twins, they're not identical twins. To use that analogy, there is a difference or I guess a gap between the physical system and the virtual system.
The more we rely on these virtual systems to inform decision making about the physical system, what do you see as the opportunity for those designed in differences and assumptions that we make a decision based on the output of a virtual system that actually doesn't quite work or do what we want in the physical system?
Drew: This is the trap with all safety analysis, it’s that we risk getting a sense of assurance because we think we've done a complete analysis, and we think we really understand something about the system. Whereas in fact, what we understand is a simplified, distanced, or imagined idea of safety. I think there is an open question here whether the actual benefit of these digital twins in closing the gap is more or less than the apparent or imaginary sense that we have a really good understanding of what's going on.
So long as we are always realistic, we are always a bit skeptical, and we don't believe that these twins are telling us really what's happening, then we'll be able to leverage the advantages. But so long as we get sucked into just how realistic they appear to be, then there's the risk that we'll overestimate them and then not actually understand where the real threats to safety are coming from.
David: Great. Drew, anything you want to add before we wrap?
Drew: I guess the final thing is just the question that I came into this with, which is a sense of what the technology readiness is here. David, I'd be interested in your opinion, but the sense I get is that almost all of the papers we're talking about here are realistic, but they are engineering school type papers. They are not industry ready applications.
They're an indication of what might be mainstream engineering in maybe 10 or 15 years, but they're not something which is directly changing the way we work in safety today. Is that the sense you've got as well?
David: Yeah, I would definitely have a sense that it isn't directly changing the way we work in safety today, but definitely changing the way we work on system design and manufacturing construction. If we look at manufacturing, if we look at construction, if we look at some other industries like utilities, the digital twinning process is quite mature, a lot of advances in the last few years or last five years on that process.
I think it could be one of these areas that it's not really influencing the way we manage safety today, but it may be something that is actually available in many industries today to start to leverage into the safety management space.
Drew: Yup. Let me just retreat back a little bit from what I said, which is that I think there's strong potential. This is the type of area where the academic publications may be lagging the industry practice. Even though this is very much academic projects that they're talking about, it may be the case that there's actually even better stuff that is more industry ready just not being published in the academic journals.
David: I think if we think about Industry 4.0 that's been a decade long push now, if you did a literature review on digital twins in the engineering and other journals like that, obviously, this literature review didn't include anything that didn't mention safety, but I think you could do another review that looked at the broader digital twin literature and looked at potential safety applications that it wasn't the purpose of the paper, but you would see an application is being used for some other operational decision making purpose and think, oh, actually, we could use that exact approach for a safety application. I think you'd see a very mature field of engineering.
Drew: Okay, so maybe more ready than I thought.
David: I'd like to think so. The question we asked this week, Drew, was can digital twins help improve the safety of work?
Drew: I think the flat answer is yes, we have something here that has potential and is worth paying attention to and keeping an eye on with some concerns that there may be assumptions and hidden limitations, but that's true of any form of safety analysis. Don't be hyped, but do be aware and look at this as a potential tool.
David: Great. That's it for this week. We hope you found this episode thought provoking and optimally useful in shaping the safety of work in your own organization. Send any comments, questions, or ideas for future episodes to feedback@safetyofwork.com.