Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
E18: Talking to Deswik about Mining vs Construction
60 Plays2 years ago
In this episode, Jason & Carlos are joined by Pieter & Pat from Deswik and discuss all things mining, 'schedule optimised design', mass haul routes... and yoga?
Deswik frees up mining engineers' time and gives them the tools and resources to engineer better plans, examine more scenarios and consequently drive lower costs.
Check out Pieter's Linkedin: https://www.linkedin.com/in/pieterrautenbach/
Follow Pat on Linkedin: https://www.linkedin.com/in/pat-banks-73575628/
Recommended
Transcript
Introduction and Podcast Banter
00:00:00
Speaker
Hello and welcome to episode 18 of the Offsite Podcast, where we chat all things construction and technology. My name's Carlos Cabolo. And I'm Jason Wansoumi. Good day, Carlos. How you doing? We live with about this shine down the middle of my head. You can see the light, the lighting set up in this wee work office. There's going to be a bit of a, like, full egg above here in the rain, which is not cool. You've self-ball joked yourself, too. Yeah, I'd rather self-ball joked than you say. It makes me feel a bit better. Yeah, cool.
00:00:28
Speaker
How are you doing?
Unique Interview Questions
00:00:30
Speaker
Yeah, I'm good. I'm good. As you know, we've got a couple of new starters joining the team here. And one of the things that we do whenever we post a job listing is we have like a bunch of questions. And one of them is to tell us a one liner that we wouldn't have seen on like the first page of Google. And there's one that I saw the other day when we were reviewing candidates since I'd be like laughing whatever I think about it. So I'm actually going to turn up with a one liner. And we can drink 10 liters of petrol and live to tell the tale.
00:00:58
Speaker
jerry can yeah that's good that's pretty good that's me then the broad range of providers that we've heard over the years that's definitely one of the better ones yeah with half of them you couldn't yeah i you would get out you would get like an explicit warning on the uh on the on the podcast absolutely
00:01:17
Speaker
Cool.
AI in Construction with Deswick's Experts
00:01:18
Speaker
Today, we have a couple of guests joining us Carlos. And I guess the context in prior weeks, you know, we, we spoke about, I guess you thought like non-traditional planning or things that are like programmatically generating schedules or maybe different use cases for AI. And this was triggered off the back of, I think you saw a demo of Alice being used on a section of our speed two projects over there in the UK.
00:01:42
Speaker
So today, I guess, uh, Peter Rautenbach and Pat Banks, and they work for, let's say, uh, kind of skunk works within Deswick. So Deswick's like a very successful mining software company, and they've got a ton of smarts around things like optimized mine planning and masterful routes.
Peter's Professional Journey
00:02:00
Speaker
And so, um, I guess I thought I'd ask Pat and Peter to join us. Peter, Pat, hello. Welcome. Thanks for staying up. Thanks for inviting us.
00:02:13
Speaker
Peter, you're LinkedIn. I've never scrolled for so long on the LinkedIn page before. There's a lot going on. RSCS, South Africa Council of QS's, Engineers Without Borders. You speak four languages, major awards. You've got courses in yoga, in Python. There's everything on there. So it's like, oh shit, where do I start here? You started as a QS.
00:02:39
Speaker
You then had quite a big focus on BIM. How'd you tie those two together? What was the transition? Oh, so you just come thinking about the ODE, actually. But you didn't do that one another day. Oh, it actually starts with the ODE. I want to hear this now. I want to see proof as well.
00:03:08
Speaker
I was working here in Australia. I was working for a mining company and I got married and my wife and I were like, oh, let's go traveling for a year before, you know, life gets this year. So we went traveling for a year and we started with the yogurt course in Dresden, which I thought would be
00:03:26
Speaker
easy, but it ended up being really intense. It was this like master yoga teacher called Nicky Knopf, who just like, I've never done anything so hard in my life, but it was, it was quite good. That was the start of it. Anyway, after the yoga course, and the trapping, I came back to Australia and I didn't have a bow.
00:03:46
Speaker
didn't have work and I ended up getting a job with a contractor in central Queensland doing a 10 on four or five foot roster as the commercial manager and it was a really tough job and it was just planes every day and it was you know it was it was rougher's guts and it was war and I was putting in all these planes and often you you know have to share extension of time and your trains and you know
00:04:14
Speaker
causation and all of that. And they were just looking for fee set schedules and Excel and it's so worked, yeah, the colour of the way. And then one of the engineers on our site, it was pretty good.
00:04:25
Speaker
3D models and Navisworks. And we started time-lapsing what had happened with access issues and stuff. And then suddenly, I started winning these plates, which was good, so we got some money. But then also, the fitters on site and the electricians, and they got really interested in these models. And they'd be sitting in my side office, and the morning's been like, hey,
00:04:44
Speaker
can we get actors there? And the technician would be like, nah, forget that. You're not going in there. We're in there today. And I was like, oh, this is magic. Somebody's actually listening to a plan for what's in it. And I was actually struggling with the plans, because we never did what we said we were going to do. So it's hard to justify a thing. I won't say which job it was, but in turn, it was great that we were actually following the plan. But it was all visualized. So I got really into plan.
00:05:12
Speaker
through that and thought, you know, like, you know, I really like how we can use models to plan our work and then learn a bit of 4D planning and got into that a bit.
Generative Design and Rule-Based Scheduling
00:05:24
Speaker
And then, yeah, and then I was at a conference and I saw this guy doing a talk, his name is John Richard, and he was talking while, and while he was talking, he was coding this bridge
00:05:37
Speaker
And I was sitting there and my mark just dropped because it was the first time I saw a generative design or, and he was using grasshopper and he would say, you know, make three columns and design this bridge and bridge was getting designed forever. So I was like, wow, this is really blowing my mind. So I went and had a chat with him afterwards and I was like,
00:05:58
Speaker
you reckon we could build a schedule using this? And then John and I spent quite a lot of time building a, it was a structural building and
00:06:08
Speaker
We designed it programmatically, and then we wrote a program to design the schedule, and then it got built through the schedule, and we visualized that it was really cool, but it was useless because it had no resources or anything like that. If you use AI, it's all set up.
00:06:28
Speaker
If you're using AI to design something with schedule in mind, is it like a schedule optimized design? Do you know what I mean? So it's, you can ask to make it the most efficient design based on program. Yeah. None of this was AI. This was all just right.
00:06:44
Speaker
coding, where you give a computer a parameter and say, I want you to build a building for me with so many stories, and then it decides it for you. But it's not fun. You don't actually have to draw a line. And the same way with the scheduling, you don't have to type a task. You just say, if the task is concrete, call it concrete. And if the cubic meters are 30 cubic meters, give it this duration. And if it's on level two, make sure everything on level one is done.
00:07:13
Speaker
And so you just give me the rules and it's set up on a schedule. So it's not AI, it's rule-based. Yeah. So that's how I learned to build. And then I make desrec. And I was like, what? You guys did this for real life? Yeah. That's how I ended up doing this. And you still do the yoga?
00:07:35
Speaker
I do. I'm not nearly as good as I used to be. And I only do like 15 minutes, but yeah, I do a little bit of Taffyn and I go for runs. So I said, can you use me? So yeah. Nice. I think the main takeaway from me there was like, I used to have a construction director on a project that I worked in London and sometimes we would have a little, uh, complain about the quantities of errors. And he always say like, where are we getting these quantities of errors? We're just getting them off the street.
00:08:03
Speaker
And now I know that we're just getting them up yoga retreats in Dresden and that's where they're coming from.
00:08:12
Speaker
We haven't done one episode over here. Jason hasn't paged fun at a QS. So I guess to kick off Peter and maybe to set the scene, like we've talked a number of times before about the idea of generating schedules and that for some types of projects, um, traditional planning methods of like building steps up and then linking them together, don't drive like the optimal results. And you know, we talked about the use case of breathworks where.
00:08:41
Speaker
in my experience, Math Hall plans were, you know, the development involved a supervisor getting in a car, driving around a site, smoking a few cigarettes and saying, okay, here's the Math Hall plan. So I guess to maybe start, set the scene and give you a way to get into the context of a use case.
00:09:04
Speaker
If we were to focus on Earthworks, tell us what you see at Deswick and how National Planning looks today, and then what it can be.
Challenges in Mass Haul Planning
00:09:19
Speaker
realizing what a problem of asshole was until we got to psych with our clients, to be honest. It wasn't the first thing that we approached. And because it's a little bit different in construction to the lining as well. It's quite intricate in construction compared to mining. I guess that can jump in there. But what we found is that we often find that there's a problem with quantification, quantifying the job as a stock. A lot of our clients
00:09:45
Speaker
the supervisors and engineers on site on to where what their current points are. So we helped them achieve that first on site by actually taking the design model
00:09:57
Speaker
using the drone scan of the site, turning the drone scanner to a surface. And then we could chop up where the work was done to and where we were supposed to get to. And the difference would be your remaining work and the opposite, obviously, with kind of build. So we're able to build these pretty cool reports effectively. And it's basic stuff. The engineer would be like,
00:10:20
Speaker
Thank you. I finally got some quantities. And then those quantities were kind of by a different material types. So you had a better idea of what you could do. But then very quickly, the mess hall came after that. Because what we're finding is that it's done in Excel. It's really slow. People sit there with this sheet. And if it's a big job, hundreds of
00:10:46
Speaker
Hundreds of rows and hundreds of connoisseurs and they're trying to make it go from top, next to bottom, right? Because it should be the least tall, but then they've got to get stuck because I've already felt that film with other dirt and they're like, oh, damn, where am I going to find?
00:11:02
Speaker
like some kind of spill all the opposite and then you got to start again it was quite a painful process and yeah I guess what we found was that they often messled us and you've done a lot on projects by advance and it kind of ends up being the supervisor driving around me at
00:11:19
Speaker
on the back of a cigarette box or PowerPoint. So that was kind of how we got there and how we solved it. And in minding, we do solve similar problems, but this one was a little bit unique. And I can probably talk about that because he wrote a lot of the code.
00:11:37
Speaker
Yeah, well, basically what we do in mining is it's a slightly different problem because you've got a lot of materials in a tight space that are very constrained by what you call your cut-shed where you've got to dig it up.
00:11:56
Speaker
because you're digging up something that's valuable and then you've got to put it somewhere because you're dumping your waste somewhere and you might have to be careful about your waste materials because they might be acid forming, need capping or something like that.
00:12:12
Speaker
But it's very constrained on the cut schedule, which is in mining, the mining schedule where you're targeting valuable material to market on a ship. But when we get to the mass wall problem in civil, it's not driven so much by that, but more by practicality. And what we were able to do is take some optimization knowledge we have on a differently framed scheduling problem and then pull it back out.
00:12:41
Speaker
take the time element out of it and more optimise where material goes in a global scale and then start scheduling at that point, because that more soon turn a surface-based, infrastructure-based problem.
00:12:56
Speaker
It's super interesting. I've not really thought of it like that. So I guess if I was to try and stay back to you, what you're saying is like in a, in a, in a mining scenario, you've got this, uh, can't schedule where you're constrained by what you've got, like more fixed constraints on this material ends up here, this material ends up here, this material ends up here. But in the construction scenario, you've got a, it's coming from here. It needs to go here, but, but you're almost like balancing. You've got more unknowns maybe. Is that a,
00:13:24
Speaker
Yeah, well, the additional, the constraints in mining, a lot of vertical stacking constraints where you're digging, say, a medium-sized open pit might be a couple of hundred meters deep, and you've got 10-meter benches. So you've got a couple of dozen benches that you have to basically take one at a time with maybe a one or two benches. You might have three open benches at once.
00:13:49
Speaker
But when you're looking at a big surface infrastructure problem, you've only got one or two materials, maybe three materials ever open at once. And you can do this over the whole area of the project, whatever timing suits your plan.
00:14:07
Speaker
you know, your optimal distribution of resources, you're not as constrained as mining. So we took, we reversed the mining problem and we said, well, what would we do if we could get access to all these surface areas at once and try and distribute materially more cost-effectively?
00:14:23
Speaker
Yeah. So that like lack of those constraints so that you could almost start anywhere and it's not stuff stuck beneath other stuff means not as much, but the problem is almost like harder to solve for a human because there's like infinite number of ways you could try and like solve the Rubik's cube.
00:14:41
Speaker
And we've actually got an algorithm that what it's built to solve and in fact just jump in the others you've done a lot more of this than I have but you know you cut the schedule and then it says where's the best place to place the stuff that I've just cut?
00:14:58
Speaker
Whereas that's speaking to us for a couple of years, but it's more about scheduling. Yeah, it's a greedy algorithm is a technical term. It's something that just solves. That's what I call costs right in front of it.
00:15:13
Speaker
But then we actually funded that was very similar to the given behavior that we were seeing outside as well. Exactly. Supervisors are driving around in jutes doing the same thing. It's like, how can I get out of this shift and just get to the next one?
00:15:30
Speaker
So that wasn't actually solving the problem. So we almost had to build a middle stage where the engineer can say, firstly, I just want to cut and fill between this in this area because this is a separable portion and I need to finish this first.
00:15:46
Speaker
and this type of material can go there and then we let the algorithm solve it. So we've always had to bring a little bit more view intervention into it. Yeah, so you can apply some more like upfront constraints that sort of reduce the problem set a little bit or like apply more rules to it. Yeah.
00:16:06
Speaker
So Peter, right at the start, you mentioned that like this mass hall planning, like given it's done in a spreadsheet, it's really time consuming and difficult for an engineer to do. They have to go through iterative process in a spreadsheet. Therefore, I imagine with my engineer hat on that I probably want to do that the least number of times possible.
00:16:24
Speaker
As the project starts, as construction starts, dirt starts getting moved around, things happen, delays happen, we find rough where we didn't expect rock, we find issues where we didn't expect issues. So many times on previous projects, similar to the original Math Hall Plan, the supervisor drives around a bit and goes, here's the new Math Hall Plan to work around this problem. Is what you're doing with clients helping them
00:16:52
Speaker
typically at the tender stage, at like the pre-construction stage, or is it throughout the construction phase? Because it would definitely be something valuable then as well.
Mass Haul Tools in Practice
00:17:02
Speaker
Yeah, we work with different clients, different phases. We're working with an owner at the moment, pre-engineering even, on one of our projects.
00:17:14
Speaker
And in mining, traditionally, we stopped quite early. But because in construction, we don't really have a name, and nobody has money. The pre-construction place, we've kind of had to start at the back, and our keep, so to speak. So we started over right at the back when they were like, oh, we don't know what's happening. It was Germany, Balkan, Egypt, up the schizophrenia.
00:17:40
Speaker
We started with doing the drone flights. Our first job was effectively that, so that they could get paid. This is how much time this dirt Jeep cut this month, put that into a progress plan. And then we were like, hey, do a little bit of planning, we'll put that. That's kind of where we started. We're trying to get through the life cycle now, because the idea that you can really make those decisions that can never begin to work.
00:18:07
Speaker
I'd imagine like at that tender phase for sure. And, and, and immediately following you can have a massive impact the right that's all fine. Yeah. But, um, yeah, so then, you know, uh, but to ask you a question, yes. So, and that's not our technology. We use other tools, um, that do the flyover and processing of the dry technology. We just import that.
00:18:28
Speaker
usually as a last file and then bring it into our software. But that's really enabled something very powerful for us because you can effectively do a full update of your schedule at 10 meter changes to the quite accurate and do a progress update. But then you can plan off the remaining works with the same algorithm you previously had very quickly in minutes.
00:18:53
Speaker
Yeah, so you build a plan that will be able to progress it, I guess, really quickly and then adjust the remaining plan based on where progress is at. And most of the plans aren't quite there, but if you are tracking your productivity of your fleets and your vehicles, you can adjust your racial sync productivity rates as well. So you can get a more accurate view of what you are going to do in the next few weeks.
00:19:21
Speaker
Yeah, yeah. Cool. Sorry. I've hugged all the questions. Carlos, do you have anything you'd like to ask the jump? I'm going to caveat this with, um, I've never experienced a mining project.
Simplifying Mining for Beginners
00:19:32
Speaker
So apologies if my questions are like oversimplified. I think you're going to, I think you're going to caveat it with I'm a quantity surveyor, but I think by now that sufficiently, that's not a warning. That's like, uh, what's the opposite of a warning? Um,
00:19:46
Speaker
I want to focus on, obviously, the whole rich aspect. You mentioned optimization. With my curious hat on, obviously, you focused on the optimization being a vocation-based mission. How much, in terms of if you think about what is optimal, do you think about cost, and then maybe think about carbon based on, I don't know, distances that vehicles are moving and things like that?
00:20:09
Speaker
Yeah.
Sustainability and Electrification in Construction
00:20:10
Speaker
Excellent question. If I can take this one, Pete. I was just, I'll let you take the question, but I just want to say something as well, which is very important. You can be my shelter again. I'm also quite a bit severe. I know. My next question is about respect still there.
00:20:29
Speaker
I wasn't gonna say anything. I did my research before I'm coming back to my pizza.
00:20:42
Speaker
OK, in terms of optimization, we've got a lot of history optimizing things. And we've had plenty of discussions about what the word optimal even means. But in a mathematical term, it means you cannot do better. And what we're doing at the moment, mathematically, in this skunk works, as we've mentioned before, is getting the minimum possible combination of distance moved by volume.
00:21:10
Speaker
So at the moment, that's what we're optimizing. But what some of our other tools in mining that work in a different space in in the mining scheduling so that the stuff that is more based on the schedule and the excavation of benches and time periods that looks at a few more parameters. So instead of multiplying.
00:21:30
Speaker
distance by volume. They are to multiply fuel burn, cycle tolling, anything that contributes to your cost. So we're at the point of bringing those parameters into this approach to it at the moment. So we're going to be doing the same thing. So we can calculate the fuel burn of a particular route based on its change in gradient loss distance
00:21:58
Speaker
the ring pull of the truck, the fuel consumption of the truck. And we can calculate that for any given horn. So if we plug that in the same algorithm, we can then optimize for the minimum total fuel burn for any given scenario. And that's what we're working towards. We don't have it yet, but it's not far off.
00:22:20
Speaker
To double-click in, sorry to jump in on top of your colors, but it's a super interesting pad. So what does that actually look like in terms of the setup? So in the mining scenario, do you define the fleet going in, or is the fleet sort of an emergent thing? So you need to know the trucks that you're using, and you need to know the fuel consumption, their capacity.
00:22:40
Speaker
and all of those written pieces that tell you how much fuel you're burning to move a cubic meter of material from point A to point B. Okay, and we get that from the manufacturers, we get that from site records, calibration, GPS tracking. If we can clean that all up and say, okay, we can now confidently say that moving here to here along this route will burn this much fuel, then we can plug that in and say, well, this is how to burn the least amount of fuel possible.
00:23:08
Speaker
Yeah, that's super cool. And does that work across like a mix that might have like scrapers and stuff in it as well?
00:23:15
Speaker
Yeah. We usually just start with the cat ad book because it's quite readily available. But then... Pat doesn't believe the numbers in the cat handbook is what we thought. I don't believe the numbers in the cat handbook. I think they're a bit optimistic. So that's why we then go back to site records and we calibrate on what actually happens.
00:23:42
Speaker
But I was also going to say our mining business is looking very heavily at electrification, both electrified trucks or trotting, assisted trucks. So they're spending a lot of time on understanding that and developing technology to plan that better as well. There's even some battery trolley charging, so charge on the fly trolley trucks being worked on at the moment.
00:24:10
Speaker
Peter, do you see in any of the construction clients that you talk to any real interest movement in any electrification of fleets being on the biggest of projects? I haven't seen anything on electrification of fleets on the infrastructure projects, to be honest.
00:24:28
Speaker
I have had questions about carbon. I'm not sure if, you know, we're just missing it every year, but generally what I've seen is people just like reconciling fuel slips to try to report on carbon, but I haven't actually seen much carbon landing in the simple space, to be honest. And I've heard people talk about
00:24:48
Speaker
then using your cubic meters of dirt to do your carbon planning, which I guess is a start. But ultimately, I think your mass will drive it, because it's the distance and the volumes of the struts will move. And the gradients, if they move them up, that's going to drive your carbon on a civil job.
00:25:04
Speaker
I know we've spoken about this before and I think you were going to ask a question about carbon as well because there's a lot of emphasis over there in the UK about planning for carbon and considering carbon decisions about projects.
00:25:21
Speaker
And I actually had a conversation with a guy who got funding for a company that claims they can retrofit batteries to these large haulage mining trucks. So he claimed it was the next world thing. He added to that soon. But that completely changes your drivers then, right? Because if suddenly fuel isn't the issue, you can completely spin the other two parameters, which is cost of program, and drive it in a completely different direction if doing double the distance in the truck no longer has a negative effect.
00:25:48
Speaker
to an extent, you can then completely optimise it in a different way. Yeah, definitely. Just on that point that you said, I said that I haven't seen much real electrification stuff on the infrastructure side, but I have seen a lot of that online. Yeah, there seems to be, particularly on the major schemes here, there's quite the actual use of like electrically and everything like that, which is good.
00:26:12
Speaker
And conscious of time. So thank you very much guys for joining us and thank you everyone for listening.