[00:00:00] Todd R. Zabelle: Building upon what was just said, we’re going to go ahead and talk about two of the pillars we’ve done industrially. We’ve talked about operation science. Now let’s talk about autonomous and digital. We’ve put together a panel of who I believe are some of the leading experts involved in this in a variety of ways.

[00:00:22] Todd R. Zabelle: Martin Fischer from Stanford, Ravi Roopreddy from Cloudleaf, and James E. Craig, aka Jim from Chevron, but before we jump in with the experts, I just want to do a little bit of framing that maybe is something that we talked about this morning, but again, the framework that we see for this, and thinking about autonomous, robotic and digital is this idea that regardless of what we do in construction, companies that are involved in manufacturing and in technology and logistics, if you will, are creating.

[00:00:59] Todd R. Zabelle: Smart assets, whether it is a truck crane, a robot for production or even a hand tool. So I know Stanley, Black & Decker and others are actually beginning to put IoT sensors or, as Ravi will call it, instrumenting hand tools. So there will be data that will be available from people doing work.

[00:01:24] Todd R. Zabelle: Okay. So again, this production layer, which is critical, is beginning to produce data that’s available. And that data, whether we know it or not, is flowing through a network of various devices, data centers, antenna, whatever the case may be, satellite dishes, and it’s starting to become put in a format that we can use.

[00:01:52] Todd R. Zabelle: Now, Ravi is going to talk about what that means. And just to go back for a sec, Martin’s going to talk to us about what’s happening in the world of robotics and autonomous, and then Ravi is going to talk a little bit about what’s happening and the movement of the data and the consumption of the data, if you will.

[00:02:10] Todd R. Zabelle: And then finally, we’re going to look at what we can do with the data. Now we had a hint about that with what Microsoft is doing with the Power BI. All right. And there’s going to be an interesting discussion about who even owns the data. We’ll get to that in a minute. So think about this as the frame.

[00:02:25] Todd R. Zabelle: We’re going to go up the stack, starting with Dr. Fischer, moving to Ravi, and then to Jim. I think what’s interesting is, you know, what are your data generators? And I find this, like I said, it’s very interesting to me where people are very much dedicated to their critical past schedules and all that other stuff they got.

[00:02:53] Todd R. Zabelle: And very rapidly here the advent of robotics and autonomous vehicles and all the stuff associated with even dumb vehicles that have a GPS device on them is going to be creating more data than we know what to do with, and it’s already occurring, and we’re going to have to figure out how do we deal with that amount of data to create value, right.

[00:03:14] Todd R. Zabelle: And what people are really trying to get to is actionable insight, whether it’s coming through data analysis or data science. Okay? So the garbage in, garbage out, or nothing in nothing out, is really what we’re talking about here. So we’re going to start by introducing Martin here as our first speaker about robotics.

[00:03:34] Todd R. Zabelle: But I’ve known Martin, he’s been a very close personal friend for a long time, and a lot of this whole journey is because I personally had a desire to figure out how to make more money in construction. I was very fortunate to meet Martin when he was doing his PhD at Stanford. And many of you know Martin. I don’t even know why I’m introducing him, but his area of focus early on was the application of 4D computer modeling to better visualize construction.

[00:04:06] Todd R. Zabelle: And he’s probably forgotten more than most of us have ever thought. He is on the faculty in the construction engineering management department at Stanford. He also leads the Center for Integrated Facility Engineering that, if you’re not involved in, I encourage you to get involved. He’s published over a hundred reference papers and several chapters and books and done over 50 keynote lectures on his research.

[00:04:35] Todd R. Zabelle: He’s worked all over the world. He lives all over the world and it’s always interesting to track Martin down to have a discussion. He holds a diploma from and civil engineering from the Swiss Federal Institute of Technology. As many of you know, that’s an impressive place, or so Martin tells me anyway.

[00:04:55] Todd R. Zabelle: And he has that MS in Industrial Engineering – Engineering Management and a PhD in Civil Engineering – Construction Engineering from Stanford. So Martin, I’m going to go ahead and hand this over to you.

[00:05:05] Martin Fischer, PhD: Thank you, Todd and colleagues. Well, thanks for the introduction and thanks for putting on this event. I would say the marriage of, yeah, the things you mentioned.

[00:05:22] Martin Fischer, PhD: Particularly, I would say operation science has a new framework and then the data we can get from different devices that you just mentioned will dramatically change beyond what we can, I think, even imagine how we plan, execute, renovate, et cetera, our build in. But I will focus mostly on robotics, but reflect a bit.

[00:05:50] Martin Fischer, PhD: And one question I often get is, “Now you do all this work in robots with robotics and, you know, aren’t robots taking my job away?” And as I thought about this, if the experience you have had so far is that they will, only the robots will not so much take your job away or about the application of robots by your competitor will take your job.

[00:06:16] Martin Fischer, PhD: So that’s hence the maybe somewhat cooperative title, but I think that’s really the decision you have to make, you know, will construction robots take your job or your competitor’s job? Meaning it’s really you that are still, for the time being, in the driver’s seat, that robots are not anywhere close to being spot enough or autonomous, that they will really take things over.

[00:06:38] Martin Fischer, PhD: And the teams that, with the construction teams that have embraced it as a new tool, they have really actually already made significant progress very quickly and those that have sort of rejected it and, as well the guys, you know, taking my job, those crews have of course struggled with bringing them in.

[00:06:56] Martin Fischer, PhD: But if, you know, you reflect on projects today, right? We see lots of tools being used. I mean but still an amazing amount of manual or semi-manual or, you know, traditional work. And the future looks, of course, much more connected, integrated with information flowing as thought alluded to between people, tools, devices, and the building parts themselves.

[00:07:35] Martin Fischer, PhD: Where are we in terms of robots helping us with this transition? I have to admit that a couple, a few years back, was quite skeptical when one of my PhD students, and I should have thanked her, Cynthia approached me and said, “You know, I would really like to study robots in construction.” Yeah, we’ve tried this before.

[00:07:56] Martin Fischer, PhD: I was one of those old guys in denial. We have tried this before and it didn’t really work. and then she said, “Well, maybe we should take another look.” And so actually we looked at the precursor of the Hilti quite carefully on the application, on the project in Norway. And I have to, as you will see from the data, I had to sort of revise a bit of my thinking about robots and where they are.

[00:08:22] Martin Fischer, PhD: But I’ll come to that, the question for sure arises as you’re reading about a new robot almost every day. That supply for construction and investments in this field, well, you have to decide, you know, which of these robots works for you in the context of your schedules, your projects, your budgets, et cetera, your challenge.

[00:08:46] Martin Fischer, PhD: And this is a task that will be with us for some time because it will be a long time before it’s just for robots to build projects. But I think, if not already today, but very soon, you will need to figure out how to incorporate robots into your cruise or have them work with your crews or become a tool of your cruise.

[00:09:07] Martin Fischer, PhD: And for that we developed through many applications, a robotics evaluation framework so we don’t have to invent the wheel again and again. Because the question is basically is it robot feasible, which has to happen in the context of your product, organization process of your project, because they all shape what your project is.

[00:09:26] Martin Fischer, PhD: But then at the same time, we need to see the robot fit to the project and the task fit to the robot. That needs to happen in the context of the typical objectives that every project has: safety, quality, schedule, and cost, which of course are prioritized by the client and project objectives that you have to reach.

[00:09:46] Martin Fischer, PhD: And so based on that, basically thinking a systematic way of thinking through these questions, it forces you to collect data, and that supports the analysis and comparison between your traditional and the robotic supported way, and then ends up with recommendations. So we were able to, over the last couple years, study 14 robots with 13 contractors in eight countries.

[00:10:17] Martin Fischer, PhD: And I’ll just share a couple quick examples just to illustrate and the sort of summary impact that we have seen so far from these 14 robots in comparison with traditional ways of doing work. So one is to test drive all robots developed by a company called Canvas here in the Bay Area, and that has been, it’s been tested on quite a few projects.

[00:10:45] Martin Fischer, PhD: There’s some competitive products out there as well. But basically what we see here is, when we look at comparison, you get a drive level five finish at level forecast and many developments. If you’re a property developer, then you’ll know what this means, but basically you got the highest level finish at the cost of the one level below, which is typically what is specified because the highest level finish tends to be too expensive, and it does offer time reduction because it tackles the drying time.

[00:11:23] Martin Fischer, PhD: But it also requires, enables, depends how you want to look at it, a new business model, because the tasks get broken up differently and that’s something we see actually in quite a few robot applications. Another robot we studied was a robot hiring robot on a bridge project together with the trailer brothers.

[00:11:47] Martin Fischer, PhD: I was a Reva layer in college. That’s how I made my money. Now, I’ll probably be paid a little bit less in that sense, yes, it will take that part of my job away, but it does 1200 ties per hour for a bridge deck. What we saw is a 21% scale reduction, 25% cost reduction, but something we see again and again in most of our applications significant decreases in strenuousness.

[00:12:12] Martin Fischer, PhD: So significant health and safety benefits and reduction in rework and in waste. So, as I said, just two examples from the 14 robots we studied. Looking across the 14 and the comparisons we found is that we saw safety, especially on the strenuous work at 25% to sometimes a 100% improvement.

[00:12:33] Martin Fischer, PhD: Basically just that part of the work was eliminated. A significant health and safety benefit, which I think is really important in the context of the shortage of craft labor, is that they can stay in the job longer because they have, yeah, better [health]. We saw a significant improvement in accuracy and with that reduction in rework.

[00:13:00] Martin Fischer, PhD: And then on the scheduling cost side, it was very much a mixed bag. We saw some applications like, for example, layout, show incredible reduction in time and cost of 90%. And then we saw other situations, where there was an increase in duration and cost because of the robot, there were still typically the other benefits.

[00:13:21] Martin Fischer, PhD: We will need to study these robots in the context of our work for the foreseeable future, hence this framework that should help you make these managerial decisions. We’ve also started to look into, well, what should we do onsite? Offsite? That’s interesting how the robots are changing this question because all of a sudden some work onsite becomes more feasible that you would want to shift offsite.

[00:13:43] Martin Fischer, PhD: But of course there can be other reasons why you shifted offsite or onsite. So we haven’t done quite as many comparisons here. This one was also on the drive side, on framing and comparing traditional and onsite robotics with offsite robotic work. We saw this in this particular case study. Of course, your numbers could be different.

[00:14:09] Martin Fischer, PhD: Again, this significant increase in the safety of the health and a decrease in schedule, but also an increase in cost, which was obviously quite significant. So PAB made it really cost prohibitive, even though the other benefits would’ve been nice to have. But as you can see, the cost for this particular application on those projects were still a bit too high.

[00:14:34] Martin Fischer, PhD: But in any case, the main point here being we are really confronted with opportunities to make work that is safer, higher quality, and typically faster. The cost is sort of really the big variable that sometimes also moves in the good direction and sometimes not. So we may have to be creative there for a while, but then the same is true also with the mix of onsite-offsite work.

[00:15:05] Martin Fischer, PhD: So we have to do the process design of, you know, of course that relates very much to the product design as well. With more knowledge and more carefully, we are also getting more data. So what I’ve shown you is basically a way of, as you are developing the project, to think already about how you are going to bring robots to your work and what you’re going to do onsite and offsite.

[00:15:33] Martin Fischer, PhD: But we’ve also seen that with the current version of robots while having already quite positive impacts. And so I wasn’t frankly a bit surprised personally by the many good positive impacts that robots already have, even though they’re still quite early. I would say that we also found that there’s very few tasks in construction

[00:16:01] Martin Fischer, PhD: that a robot can do by itself. So there’s always a human-robot collaboration and that’s an area where we are doing quite a bit of research because that’s something we need to improve, I think, significantly, so we can leverage what people are good at and what robots are good at, because even when you do offsite fabrication, then you still have, you know, quite a bit of work to do onsite.

[00:16:28] Martin Fischer, PhD: And quite a bit of it is repetitive. That could be done at least mostly by a robot, probably again, to higher accuracy and we need to find ways of teaching the robot. So we’ve explored the human-robot collaboration for drywalling, painting, bolting, welding and joint-sealing. And actually developing a model to teach a robot how to weld in sort of, you know, small places where otherwise you would have a lot of setup.

[00:17:07] Martin Fischer, PhD: But many projects have these kinds of situations where you need to weld hundreds of columns to base plates or things like that, or need to put the base plate to which you can then bolt a column or things like that. And the difficult thing here is to do this truly autonomously, which would require a lot of technology in terms of vision and ability to correct and ability to recognize clutter and all kinds of things.

[00:17:32] Martin Fischer, PhD: So things that are unrealistic for our project site. What strikes us as more realistic is to teach a robot the things that are really repetitive, which is putting the weld at the plate once the robot is at the plate, and so for that, we can, in a virtual event, first of all, we are also using a virtual environment to design this human-robot collaboration, and then we can use it to interact with the actual world, so we can have this basically augmented reality

[00:18:06] Martin Fischer, PhD: approach where a human is with a haptic interface. And if you haven’t experienced a haptic interface, you should find a way to do that because it’s remarkable how much feedback you can, what kind of feedback you can really get in terms of guiding a robot. And so the idea is that, for example, a human would guide the robot to the base plate, but then the robot would’ve learned, “Okay, yeah, now I’m at this corner.”

[00:18:33] Martin Fischer, PhD: “I’m going to weld myself.” So if you had several robots, you could see how you, how a human could guide many robots to many locations. And maybe in some areas the robot could eventually learn to go from base plate to the next base plate. That’s certainly within the realm of what looks feasible. But we’ve developed this simulation environment including haptics and basically an augmented reality approach to explore human-robot collaboration and then to design the particular robot capabilities and human capabilities so that they can be deployed safely and productively.

[00:19:11] Martin Fischer, PhD: And we’re in the middle of actually testing the first prototype, so I’ll report on that in a few months. But this is a very exciting area in terms of, I think, human-robot collaboration. Instead of going fully autonomous, basically start to teach the robot degrees of autonomy that work for the people you have on site and that work for the conditions that you have on site.

[00:19:40] Martin Fischer, PhD: And finally, then the last bit, you’re probably wondering is if the design is always fit for robotic construction, and that’s also something we’ve learned as we’ve worked on these case studies that you can often quite quickly see, well, if there was just a small change made, the robot would’ve been deployed much more beneficially like here on the first application of the predecessor of the Hilti.

[00:20:08] Martin Fischer, PhD: They had two diameters of holes and bolts and that created quite a bit of setup, cost and switching costs. And the reflection was afterwards that really, the sort of the smaller bolts really didn’t save hardly anything. It would’ve been much better to just have one diameter throughout as a simple example, but in a material handling robot that SHI is developing and deploying on its projects in,

[00:20:37] Martin Fischer, PhD: you also see issues of level changes, types of elevators, access issues that if you know them, you can probably incorporate them in the design and then that makes robotic construction much more feasible. If you don’t know about them, then well, you may have a design that either you have to fix or where you deploy the robots, but if you don’t deploy ’em, you lose out on the benefits you could have, or deploy ’em, but not as efficiently as you could.

[00:21:06] Martin Fischer, PhD: So that’s another project we are currently going to figure out, you know, how do we design for robotic assembly? To think about, of course, dimensions, obstacles, variability in part design, how the joints  formulated, tolerances and so on. So this is where, so what I would say in conclusion, there’s definitely robots worthy of your attention now if you are not already using robots on your projects.

[00:21:48] Martin Fischer, PhD: Definitely, I would say imperative to reflect on your offsite, onsite mix of use of robots and people.

[00:22:00] Martin Fischer, PhD: And what’s a little bit down the line emerging is the human-robot collaboration onsite. And the feedback loop with design. As you know, some of these take sometimes a bit longer, so if we put this together right, the robot evaluation framework helps us guide what robot is applicable and useful for us, and the on- and offsite question.

[00:22:28] Martin Fischer, PhD: And then the human-robot collaboration design for robotic assembly. This is connected with a lot of the outdoor information, data related and other technologies that we are seeing that have already been mentioned. IoT, OT, sensing, logistics management, digital twins, BIM, and so on. And generally speaking, it’s a question of management and design.

[00:22:53] Martin Fischer, PhD: Like everything, you will have to make your management decisions in terms of where you deploy, what robot, and mix of robots and people. And then we have to also think about the design of this human-robot environment. And that includes the product. But I alluded to the data side, and this is something that I haven’t heard mentioned as much, I think in the long run it will be a really big deal.

[00:23:28] Martin Fischer, PhD: I mean, I think the health and safety benefits, the quality benefits and schedule benefits we’re already seeing are quite a big deal. I think, I imagine the cost, as with all technologies, will get better over time. So I think this is very much a real opportunity. As I was thinking about, you know, here we are in five years, companies will have deployed lots of robots and I was wondering who will know more about construction at that point?

[00:23:58] Martin Fischer, PhD: The traditional construction company or the robot company? And it’s very clear the robot company will know so much more when we do these comparisons. It’s really hard to get data on the traditional way of doing work, and it’s really easy to get data on what the robot does, because you just get data as you go.

[00:24:18] Martin Fischer, PhD: So for example, here from this, an automated data collection system from Versatile, I think, may have existed before. It collects everything that Crane is doing. And for the first time in my career I was able to see for different design conditions, like three different beam conditions, the average installation time and precast beams and the variability that existed.

[00:24:48] Martin Fischer, PhD: So you see, for example, that beam condition two has a higher average insulation time than beam condition one, and a really high variability and beam condition three has an even higher average insulation time, but less variability. So these kinds of insights we have been lacking so far, but this really

[00:25:11] Martin Fischer, PhD: provides us now the opportunity to work with all the five levels of optimization, right? I’ve already alluded to the product and process design interaction and now we are getting data on the variability, which helps us then, as you know from other sessions, understand the whole production system. And we also get data on CAPA, capacity utilization.

[00:25:34] Martin Fischer, PhD: And so now we can really understand what is the working process for a productive system and how all of these parts interact in terms of the product design, enabling or requiring a certain process that is enabled by certain types of robots that give you data and so on. So this is really an exciting world that I’m entering,

[00:25:55] Martin Fischer, PhD: that we are all entering, that I look forward to being part of it. In case you are a quick plug-in case and you want to do a traditional work comparison, let me know because we will do more of these case studies in the winter quarter as we do our construction robotics class. Okay.

[00:26:15] Todd R. Zabelle: Thank you very much,

[00:26:15] Todd R. Zabelle: Martin. I’m going to go ahead and move on here. Excellent presentation. From my perspective, a couple things that come to mind. A few things. Martin said, the robot won’t take your job, but your competitor with the robot might. And I would say not having an understanding of how to approach all this might just take, and you might put yourself out of business.

[00:26:52] Todd R. Zabelle: What Martin has just presented was all about production, about robotic production and understanding the implications of robotic production. And then of course, it automatically brings you to the product and process design. And for the life of me, I have no idea why people keep messing around when, if you’re going to use a robot, we need to come up with another type of penalized material that’s fabricated completely and the robot welds or fuses the corners to.

[00:27:19] Todd R. Zabelle: But you know we started the day off with the Drucker quote, and then finally this idea that Martin introduced, that Keith talked about, of how tremendous amounts of data, as we said earlier, are going to flow from and through that robotic production to informed design. All right, so, so moving on here because I’m concerned about time and the people upstairs are sending me nasty texts.

[00:27:47] Todd R. Zabelle: To get moving, we’re going to have Ravi from Cloudleaf. Talk a little bit about what’s going on this network layer and where all this data that Martin’s talking about might end up. But again, I think, you know, think about what Martin was talking about on the implications of what he just said, but specifically, Ravi’s a technical co-founder and oversees the cloud operations and software engineering for Cloudleaf.

[00:28:14] Todd R. Zabelle: He has 31 years of experience in software design. He holds a Bachelor’s degree in Mechanical Engineering from JNT University India. And he’s done some graduate level coursework at Arizona State University, Tempe. So with that, I’m going to hand it over to you, Ravi.

[00:28:30] Ravi Roopreddy: Okay. So it’s great to be on this panel.

[00:28:33] Ravi Roopreddy: Also last year we went through a next generation of supply chains kind of thing. We’ll touch a little bit on that, but, mostly there, there’s just a lot of progress that happened since last year to this year in terms of the IoT in general and then construction in a specific way. You know, how the things are applied to construction.

[00:28:53] Ravi Roopreddy: The picture, the way it’s evolving is, like Todd said, and Martin kind of said, data acquisition is becoming really commoditized. We used to talk about last year’s sensors and you know, the sensory kind of streams coming in. A lot of the shift, there’s multiple changes that’re happening, right?

[00:29:13] Ravi Roopreddy: You know, typically the whole digitization of the supply chains, meaning there’s a discreet view that was happening, both the inbound outbound logistics versus, you know, in the warehouse. It changed into a bit of a continuous modeling through digital trends. I mean, digital twin is typically talked about in a different context, but in this context, it’s like supplier networks, you know, your routes, your transportation modes, and all of these things are modeled into a world.

[00:29:39] Ravi Roopreddy: You see a connected view basically, right? So that’s just the modeling of the portion of it. We’ll talk a little bit about the AML at the end of it. The whole data acquisition, if you’re going to step back a little bit, these are all signals coming through the network. These signals are the sensors through cellular networks, through satellite networks, different back hauls, whether it’s flights, you know, whether it’s oceanic and vessels.

[00:30:04] Ravi Roopreddy: So signals of different kinds, devices, carriers, and different, other kinds of, you know, carriers. So there are multiple signal sources. Integration has never been a problem. Enterprise integration has been there. So getting these signals wired up to a bigger picture into feeding that into the digital trend became like the theme basically.

[00:30:25] Ravi Roopreddy: Right. You know, it kind of evolved to the next stage into the modern construction signal attribution. So typically when you look at certain things that are moving within this ecosystem, you can have a nesting of items. Ultimately, you get a view on a transportation system, like a ship or trucker or something.

[00:30:47] Ravi Roopreddy: A container, but not to the level that you go to the final asset itself, like a piper, you know, the boxes in a pallet and things like that. Our goal is not even a goal that is happening today. Go, you know, acquire the signals from a flight or wherever, right, a truck, but attribute those signals into the, into the digital item level so that you get visibility into this network.

[00:31:15] Ravi Roopreddy: We used to have continuous visibility in discrete silos, right? That is changing. That means the whole picture. These silos are completely, and I’ll talk about that a little bit in the context of the digital twin, these silos are completely getting stitched up. You know, if you have suppliers both on the supply side of the

[00:31:37] Ravi Roopreddy: manufacturing, on the outbound logistics side, they’re all getting connected basically to make it a one continuous, a stream of a model, basically. Right? If you look at the construction, what are the different use cases? You see, the whole data acquisition then, then the IoT kind of can help or transform, right?

[00:31:57] Ravi Roopreddy: One is the modeling itself, farming a bigger network, you and your suppliers, and then the, you know, networks, transportation networks, all of that. Just modeling that. Construction information modeling is a key thing so that you don’t have a siloed view. You have a complete view, right? Then when you’re dealing with the sites and the monitoring, where the majority of the work is happening, real time remote site maps, who, what, where, kind of a view, you know, people, equipment, tools, all of that view, right?

[00:32:27] Ravi Roopreddy: Sensors and trackers in the drones. A lot of these are like  signal sources that kind of give you that visibility. Including the machines, you know, just devices. All of these robots also, like Martin talked about in the session before this, they’re all signal generators and kind of helping this bigger picture when it comes to the people, you know, the wearable tech.

[00:32:47] Ravi Roopreddy: I know it’s happening in the consumer world, but it’s also seeping into the construction world basically and starting to utilize it, they’re getting sophisticated. Signals kind of are really helping to build a bigger picture of, you know, the efficiencies and resource management once construction happens, you know, the whole new thing about the green building, so intelligent building management, they’re all still going through the same sensory kind of, you know, signals coming in to manage it often.

[00:33:14] Ravi Roopreddy: Todd talked a little bit about resource management. Tool tracking is a kind of scenario, so I think we keep seeing it actually as another additional use case into the construction world, check in, check out, and you know, put ’em in, you know, renting them, all of those kinds of things. New use cases are coming on to utilize these kinds of advancements.

[00:33:36] Ravi Roopreddy: So this is a picture we had last year, a bit of a painting. This looks like a continuous spectrum, right? That’s what happened this year, right? In a way, this is how we looked at the thing in the last year, you know, which is, let me just go, this can, okay. It doesn’t move. Okay, so this is basically how we viewed it as silos, right?

[00:34:02] Ravi Roopreddy: You know, we viewed the entire efficiencies and the IoT supply chain network within four walls from the warehouses. We viewed this as inbound logistics, and then the outbound logistics basically that completed those lines are getting blurred. Meaning you have one continuous modeling for digital networks or digital twins, so

[00:34:26] Ravi Roopreddy: So if you look at the modeling, your value chain is a hundred percent. Now instead of silos, let’s say imagine you’re bringing a supplier on board. You know, the supplier has its own network and you’re connecting these two networks through a digital twin. It’s as simple as, you know, basically connecting these graphs of these networks and then it becomes a far bigger network.

[00:34:50] Ravi Roopreddy: I’ll talk a little bit about what this kind of modeling allows. Bringing onboard subnetworks and then connecting them together and then allowing data sharing, right? That’s what it kind of made possible from a modeling side of it. And the visibility side. We talked last year, you know, because that was the beginning of some of these things that are happening.

[00:35:11] Ravi Roopreddy: A lot of the visibility now, the big, the minute you kind of have a 100% modeling side of it, these gaps, the silos get connected. The visibility can now increase to 80% to 90% across the value chain. Ultimately, in the construction, the simple question that gets asked, is the material going to be there when I need it?

[00:35:30] Ravi Roopreddy: Right. Doesn’t matter where it comes from. That’s raw material or finished goods, you know, to the site. And this is these two, the whole digital twin based modeling plus the, you know, signal acquisition through various means is kind of providing an answer to that, basically with higher level tools on the top.

[00:35:49] Ravi Roopreddy: So this is a bit of an overview. If you see these things, what I’m highlighting here, these are all signal sources. These are not necessarily just devices anymore. Carrier signals, for example, you know, if you have certain goods getting transferred through the carriers or the supplier networks, those signals will come in along with the device based kind of signals.

[00:36:13] Ravi Roopreddy: If you have certain things that are transported through the oceanic kind of vehicle through satellite as a backhaul, those signals will come in. And ultimately you are ex— you are kind of attributing these signals all the way from the top level shell like. And if you take a truck, you know, transportation management to your containers, to all the way to the

[00:36:31] Ravi Roopreddy: end, you know, a kind of assets that go into pallets kind of thing, right? They’re going from the raw material on the inbound logistics side. Same signals are generated and the same tracking is happening. This is part of the network basically. And then, within the four wall warehouse,

[00:36:48] Ravi Roopreddy: you know, you could be using RFID systems, you could be using a BLE, you could be using satellite based, you could be using GPS based trackers, doesn’t matter in a transit kind of sense. They all generate signals and you get one continuous normalized view of where things are. You know how they’re moving and you get them in a time kind of view.

[00:37:12] Ravi Roopreddy: A little bit of breakdown. In the manufacturer side of it, the work in process and a factory tracking side, basically. Right. There’s a different kind of, you know, again, of signal generators, so we go back on the receiving side, raw material, receiving side, you get an acknowledgement saying that you got the right amount of the raw materials coming in and within the warehouse, basically you’re going through different cells for work and process.

[00:37:41] Ravi Roopreddy: You get a view into that. Also, you know, starting with the, you know, the raw material inputs to finish goods when it ends, you know, step one is to step in and the sensors are giving you a view of how they’re moving. What is the velocity? You know, and of my finished goods. So finished goods finally when I end up in the outbound dock.

[00:38:01] Ravi Roopreddy: So you get a complete view into this side of the picture, basically. Right? So that’s, again, throw the signal, generators, various kinds, you know, there is a signal that gets attached to the boxes, pallets, and then, you know, there’s a zone sensor, there’s a, you know, whatnot, any kind. And then there is, these days, the whole mobile model is in the picture, too.

[00:38:24] Ravi Roopreddy: You know, you have AR you know, like barcodes and RFIDs that are there, scan, throw that. You get a view from those signals also into your information system basically. Okay. So these are components in transit. Basically finished goods, you know, getting into the outbound, you know, into the job sites.

[00:38:46] Ravi Roopreddy: It’s the same thing there. Different kinds of signal generators, basically going through different kinds of back hauls, you know, satellites or ceiling. There’s GPS trackers, condition trackers like, you know, where sharks and other kinds of conditions are fairly common. If you’re transporting goods that are shock send still.

[00:39:03] Ravi Roopreddy: Your windows and so on and so forth. Attach it to them. You get a view in real time. You know, if some kind of damage happens, this is all real time as it’s happening through the network, right. You know, the things are loaded into the trucks. And then the gateways that are living in the outbound lock, you know, the dock kind of, you know, says that, yep.

[00:39:22] Ravi Roopreddy: It’s living at the manufacturer’s site now, and then it’s in transit, and then where exactly it is, what is the ETA? All of those get collected and then, it kind of, you know, you start to get a view of where the progress of this transportation model is, and now on the delivery side of it where the assembly is happening.

[00:39:44] Ravi Roopreddy: The finished components are delivered to the job sites, and you get the same thing. It’s auto acknowledgement that’s happening for these sensors, saying that, yep, received the goods. Again, throw signals are kind of giving you a view, but ultimately there is a higher level software, a higher level. A functional view is being built, which is, you know, there is a geo offense, alerts are happening, or it’s the ETTS of saying, “Hey, it’s a 30 minute ETA to the job site so you can line up your labor and other things.”

[00:40:10] Ravi Roopreddy: So you know, when the exceptions are happening, you get those notifications. All of that is happening in real time in a truthful way. Right? When something, the system says it’s 30 minutes away, it is 30 minutes away in a way, unless there’s a, you know, some risk or some, you know, accidents, other things involved there, right?

[00:40:27] Ravi Roopreddy: A lot of this is logic that gets delivered through these notifications and the exceptions, basically.

[00:40:37] Ravi Roopreddy: So here is the fun part, right? You got a lot of data, you acquired it, you kind of used all these.

[00:40:42] Todd R. Zabelle: Ravi, can we just go another 30 seconds? Is that all right?

[00:40:46] Ravi Roopreddy: Sorry. Okay. Yeah, I’ll try to wrap it up by then. Okay. Okay, so this is the, this is the control watch tower. You’re trying to build a higher level kind of metrics on the views basically of these signals coming in, right?

[00:40:56] Ravi Roopreddy: So, so you know, where and what and all that thing. It’s the same thing here. You know, you can start to build a lot of fun analytics here on time delivery cycle times, you know, the web turns and so on and so forth. Using that data pretty much you’re working  data that’s giving you a real picture there.

[00:41:14] Ravi Roopreddy: Same thing with the in-transit logistics metrics, transit, time on-time deliveries, and so on and so forth. The minute you have a signal AC acquisition and you transform the data, you can start to formulate a lot of these, you know, very useful analytics. So I’ll take like around 30 seconds on this one. We started with raw data acquisition.

[00:41:35] Ravi Roopreddy: We built some analytics, we acquired the data built analytics. What is happening is basically the next level, the baby steps are happening around the whole data science. Imagine millions of parts are getting transported, huge digital networks. It’s very hard to handle, deal with, you know, like a general business rule centric approach.

[00:41:55] Ravi Roopreddy: So a lot of this is now getting through the data science. You feed a lot of this data, which is the real data, and the cleaned up and quality oriented data starts to build machine learning models on it. Now, e-predictions are a very hot commodity now on top of the data. Anomaly detections exceptions.

[00:42:12] Ravi Roopreddy: Basically you don’t want to deal with the normal on these millions of parts. You kind of want to deal with the exceptions. So those kinds of things are already happening for this thing in the IoT space with the large amounts of data access in the networks. So I’ll kind of pause there, hopefully wrapped up within time and then gave some insight

[00:42:33] Ravi Roopreddy: into what’s happening in the realtime world of IoT and construction.

[00:42:37] Todd R. Zabelle: Thank you very much, Ravi. Just to keep things moving along we’ll send any questions that we have to you. Okay. So again, I want to show a quick graphic here for no other reason than to say we’ve begun to develop a model for doing analysis of what

[00:43:00] Todd R. Zabelle: data are you capturing? And if this is something that’s of interest to you, we’ll make it available. I think we’re working on a paper on this for those of you that are trying to figure out what your strategy is for data. But you can see that the data and the digital are very interconnected with the production.

[00:43:18] Todd R. Zabelle: Okay. So we’ve talked about the robots and autonomous production of this data along with others. We’ve talked about how this data’s moving around, and spending time with Ravi will be fascinating. For those that are interested. Now going to introduce Jim Craig to talk about the top layer. And one thing I want you to think about is, as Jim’s presenting, James talked about the process mapper and what we’re doing with the production system and what’s happening out there.

[00:43:47] Todd R. Zabelle: The big secret here is to take the IoT sensors and use them on the digital twin upon which you applied the operation science. All right, so with that, I’m going to introduce Jim with Chevron’s Project Resource Company based in Houston. He’s responsible for project production management across the corporations worldwide.

[00:44:07] Todd R. Zabelle: Major capital portfolio, along with being involved in the innovation team. He joined Chevron in 2004. He’s been on a lot of projects all over the world in nasty places. He has a degree from Texas A&M in mechanical engineering. He also has an MBA and he is, I believe, a licensed engineer. So with that, Jim, I’m going to hand it over to you.

[00:44:30] Todd R. Zabelle: Thanks.

[00:44:30] James E. Craig, PE: Todd. And thanks for having me. So, what I’m going to talk about, and there’s been some really good topics today, you know, and a lot of things we can learn from each other, as the things I’ve listened to, especially by Keith, you’ve heard his presentation, that we can definitely learn from him as well as Microsoft.

[00:44:48] James E. Craig, PE: Right? And how they do their Power BI, which, to me, was really fascinating. But some of the stuff we’re doing at Chevron in this space is, you know, we’ve been on this journey with operations science for a long time, it feels like, you know, with seriously since about 2017 and dabbled a little bit earlier.

[00:45:05] James E. Craig, PE: So we have a lot of production data out there that we’ve captured through all our different projects. And so what did we start doing with that data? Well, we started using it to our advantage, right? Because you know, data’s important as you’ve heard. You know, making sure you get it in and out of the cloud.

[00:45:18] James E. Craig, PE: And you know, also standardization of data to make sure you understand what you have and you can manipulate to be able to have the best utilization so you can increase your project performance. Right. So, you know, one of our projects, you know, we capture a lot of variability through Chevron PM which I’ll talk about a little bit later on.

[00:45:36] James E. Craig, PE: And we’re able to use that variability to start looking at when we started going into some production planning of, okay, what’s some big variability impacts out here that are keeping us from getting work done? Right? And one of, we learned it, one of our sides was weather impacts, right? These were always causing us delays, right?

[00:45:51] James E. Craig, PE: People weren’t thinking about, oh, the wind’s going to keep us from a heavy lift or something like that. And so we started taking the data we were getting from the Chevron production manager, and integrating it with, you know, just some standard data feeds of what the weather is going to be like tomorrow.

[00:46:07] James E. Craig, PE: And that helped us a little bit in our production planning to make sure that we weren’t planning work that we couldn’t get done. So it helped us be forward-looking in this space, which was very powerful in helping us, you know, kind of eliminate that variability. Right. We also looked at things like task completion, right?

[00:46:23] James E. Craig, PE: We started getting a lot of data in this, and we looked at some algorithms and started looking at, okay, we know this task isn’t going to happen. Right? We noticed that 80% of the time it doesn’t happen. So what can we do to prevent that? This allowed us to be very forward-looking in our planning, right?

[00:46:38] James E. Craig, PE: Which again, helps us reduce variability and helps us reduce the cycle time out there at the site when we’re doing our construction activities. And then, you know, I think Ravi talked about how data can come from anywhere, right? We do a lot of motor vehicle transportations in some of our operations and this one project we had a lot of vehicles moving around.

[00:46:57] James E. Craig, PE: So as part of that, you know, we keep people safe because people, NBCs, as we call ’em, is one of the number one injury makers on our sites and within the corporation. So we put some instrumentation out there to make sure drivers are staying alert and to make sure they’re staying safe. As a result of that, we started collecting data on truck routings and how they were going around the site.

[00:47:19] James E. Craig, PE: Right? And this allowed us to use this to help optimize those truck routings to one, reduce cycle time, you know, and really be able to get some benefits of how those trucks were moving around, right? Especially in our dewatering operation, because the site had a lot of dewatering going on and it really helped us optimize that so we could remove that variability in that system.

[00:47:40] James E. Craig, PE: Which again, you know, you never know where you’re going to get your data from. You never know how you’re going to use it, but once you put it in the power of the people actually using it, you can actually give some really good benefits from it, which is a really nice benefit. [00:47:56] James E. Craig, PE: Next slide. You know, there’s lots of stuff out there. You know, we’re in this world. And we’re thinking about how we can do it all the time. We’re always looking at how can we even see here, right? How can we do better on our projects? You know, making sure we have competitive, predictable projects.

[00:48:13] James E. Craig, PE: Chevron’s went through a big transformation recently, where we realized that data is really important and how can we make things recycle time with it, how can we make our efficiency and operations— how can we use that? Right? So we see a lot of things in here around project management, because we do have, Chevron came out there and we do get a lot of data from it, but we are doing it on the engineering side too, right?

[00:48:35] James E. Craig, PE: We’re looking at operational data and how can we pull that back into our designs? How can we get design standardization? How can we get this data moving between our digital twin, you know, starting our digital early when we identify concepts and take that all the way through and handle that operation.

[00:48:53] James E. Craig, PE: So there’s a lot of things we’re working on, right? And it’s an emerging market and we, and we’re trying to team with various aspects in the industry, and so we can get there because we know that data is extremely important, right? And right now our performance data, I mean, we are sucking it into our data like four times daily right now.

[00:49:12] James E. Craig, PE: Right. Which again, gives you an immense amount of data that you can do this. And again, it enables our projects to be able to be more predictive rather than be more backwards looking, because we have real time data we can analyze and it also, you know, allows us to really take that data and as Mark mentioned earlier, a lot about modeling, looking at variability, looking at WIP, you know, looking at where’s the optimum place to be?

[00:49:40] James E. Craig, PE: And we take, I’ve taken that data and actually have modeled portions of our project right from a complete project. You know, from all the way through to very small pieces of engineering, if we need to, you know, we can take it and say, okay, how’s our ISO production going in engineering? How’s that being effective?

[00:49:55] James E. Craig, PE: Fabrication? We’ve modeled that stuff, right? ‘Cause we have the data. We can look at it, we can take instant data, you know, from our production planning session, which I’ll talk about a little bit later, and we can do that. We’ve become very good at it and very predictive and we can say, “Hey, we need to add people.”

[00:50:11] James E. Craig, PE: We can tell you which people we need to add. So we’ve matured a lot in this journey that we’ve gone through, which has been, it’s been quite eye-opening for us. So how do we collect this data again? You know, we use a tool similar to what you were talking about earlier, and teamed up with SPS and we have, we’ve kind of rebranded Chevron Production Manager.

[00:50:29] James E. Craig, PE: That tool is very powerful because that allows us to work with the people doing the work. Identify the processes, map out the processes which James showed on the PPI website. We take those processes, right? And depending on where we are on the project, we work with the people doing the work and we generate, and projection schedules are usually about two weeks in advance looking at it.

[00:50:50] James E. Craig, PE: And we do that. And then we work with the production team and we get the production planning. We go out to execute the work. What are you going to do? What do you not do? And that allows us to capture variability, allows us to add resourcing and be able to make the right decisions as we go through our project execution, which again, you know, we’ve been on this journey for a while and we’ve learned a lot and we still have a lot more to learn as we go through this.

[00:51:13] James E. Craig, PE: So we have a huge library of standard processes now, right. And we did some cross sharing with Ford and we learned that, you know, when they design a vehicle, they now have a bill of material, but they have a bill of process, a BoP, as they call it, and they integrate those into engineering.

[00:51:29] James E. Craig, PE: And so that’s something that we’ve been looking at quite intently to see how can we merge those going forward and really get that into the design early when we’re doing that design, which we have done a little bit of so from there we have, from that,

[00:51:49] James E. Craig, PE: standard processes. A person could go to standard processes and here’s the library of it. I can tell you right now that this library has not only engineering processes in phase three and phase four, which is feed engineering, as well as detailed engineering. We also have procurement processes in there.

[00:52:05] James E. Craig, PE: We have construction processes in there as well as some fabrication. And one thing you have to remember is that these processes, no matter where you are in the world— Right. You do the engineering the same, you do a PID the same way you do it in Houston, as you do it in London. You put a piece of pipe together the same way you do in Kazakhstan, the same way you do in Australia, the same way you do it here.

[00:52:24] James E. Craig, PE: So these things are very transposable, which is nice because you can use this, like I said, anywhere, right? So, you know, we have a library of standard processes you can pull that process from and from there you can say, okay, if I’m in construction, how can I make sure that I’m doing the right work in the right sequence?

[00:52:41] James E. Craig, PE: And if you’re an engineer you can get into a different type of design process. We talked earlier, which is earlier around concurrent design. And so with that, you know, you can build a process, which I think was marked earlier that says you almost have a 3D twin of your, of how you’re going to build it.

[00:52:59] James E. Craig, PE: And you take that with the design and you integrate that. Right? And we did this on a project where we knew we were doing some electrical cutovers. We already had all the work processes mapped. We had the design, we did some 3D laser scanning, and we merged those and what would be called FFL two or phase.

[00:53:16] James E. Craig, PE: To really try to reduce that cycle time of that engineering and advance the maturity of engineering further than we would before we went into what we call feed and then detailed design. So it’s still working through it, but it’s been a good success so far, which is nice. Again, there’s a lot of things you can do with data once you get it out there.

[00:53:36] James E. Craig, PE: Once you understand what you have and where you can go. So, again, this is just a high level, you know, everyone has a cloud. I heard earlier understand your cloud, understand where earlier your data is going in your cloud. You know, and this is kind of high level how it flows into our data warehouse as well as our data lakes.

[00:53:53] James E. Craig, PE: You can see we have a lot of data around production management. We are utilizing Power BI to go out there  and grab data outta that, and be able to manipulate it. So it’ll help us be more predictive in our projects rather than backwards looking, which again is a huge enabler to reduce cycle time and have predictable projects, which is a huge focus of Chevron right now.

[00:54:14] James E. Craig, PE: I can’t tell you how big it is. And of course we also have, you know, the same type of situations for designs. Again, we are looking to take operational data, you know, how’s that injecting decisions and equipment. How are we doing our 3D digital twin and taking that all the way through the design life?

[00:54:28] James E. Craig, PE: How are we getting the— how the work processes into that digital twin so we know what we’re building before we go build it, so then we can pretty much simulate everything building it. Operating it before we even get out there. Again, taking pieces from the automotive and manufacturing world, we know it’s not exact exchange, but taking those elements where we can, learning from them and seeing what we can do to better our projects and better our processes.

[00:54:54] James E. Craig, PE: So like always, data’s huge, right? And the lake is going to keep growing as it always will. You need to make sure you’re smart about it, you know what you’re going to do with it, make sure you get it in a standard format. We’re doing a lot of that standard formatting and our cost data and project control so we can be able to compare projects globally.

[00:55:13] James E. Craig, PE: Engineering’s the same way. We have different assets, different codes. We’re trying to get that standardized so we can use it and be able to use it to our advantage. And at the end of the day, this has become a bigger elevation as we own that data. So we know that this is going to change the way we work internally, as we know, we know it’s going to be able to change the way we work with our strategic suppliers because it is a different business model, different

[00:55:36] James E. Craig, PE: business proposition that we’ll have to work through and don’t know the answers yet, but we know we have to work through it. But Chevron has realized there’s a lot of value in this data, and we’re going to own it. We’re going to use it. Thank you.

[00:55:50] Todd R. Zabelle: Couple things. Just to wrap it up really quick on this particular session, some things that I thought were amazing is what Martin is talking about. What is going on with robotics is a complete game changer, all production.

[00:56:04] Todd R. Zabelle: We were texting while Martin was speaking. He wasn’t talking about flying drones around and running robots to measure progress. He was talking about doing actual production work, right? We heard about how things are starting to move around the data layer through the networks from Ravi, so I would tune into what Cloudleaf’s doing.

[00:56:22] Todd R. Zabelle: I took a little snippet here from Jim and from the Microsoft presentation from Efron. And you know, this is a really big deal right here in Microsoft’s getting it every two hours. Chevron’s getting it four times daily. This data’s going into a central data lake. And what’s going to happen, if you put all this together, is we’re automatically through this

[00:56:47] Todd R. Zabelle: production data flowing through these networks, we’re going to move to what we call intelligent production, where the digital twin is going to create itself. All right? So something to think about. The digital twin’s going to create itself, and as the digital twin creates itself, right, we’re able to model, simulate, analyze, and optimize based on what Mark was talking about, what Mark was showing, right?

[00:57:09] Todd R. Zabelle: And then obviously there’s the element of control that both Microsoft, well Mark, Petros and Chevron talked about. Okay, so control really is going to start to, at some point, move from human and management to more control of the elements that are creating the production. Again, whether it’s vehicles that are moving or it is robots that are doing work.

[00:57:34] Todd R. Zabelle: Okay, so I’m going to go ahead and hand this back to first take over at this point. Thank you, Todd. Great.