SDS 252: Data Science In Construction

Podcast Guest: Kirill Eremenko

April 12, 2019

Welcome to the FiveMinuteFriday episode of the SuperDataScience Podcast!

Today, we’re continuing our discussing of data science in various industries, in today’s episode: construction.
First off: we define construction as an on-site creation for a known client (as opposed to manufacturing). It’s a massive industry, larger than even mining. The GDP across the board is 15%. It’s also one of the slowest industries to introduce the use of AI. Their use cases are simple but that makes it exiting. There’s a lot of open space to work with and a lot of creativity.
1 – Work Process Streamlining
Construction sites are complex. There’s lots of moving parts: people, materials, etc. People might not be using optimal movements or paths to get to where they need to go. Data science comes in as a way to streamline optimal routes, locate bottlenecks, and reduce the amount an average walker needs to walk. This can give an extra hour per worker per day.
2 – Reducing Waste
Construction projects result with almost 30% of their materials ending up as waste. In the UK, 50% of all landfill is construction waste. This can be as much as 25% of a building’s total cost. It’s massive numbers which means there’s a ton of opportunity to use data science to get the numbers down. VOCA utilizes their technology to track the delivery of materials in real time and determine their use case and apply that data to future projects.
3 – Augmented Reality
We think of this as a gaming technology. But AR has huge applications in construction. AR can help developers understand how a building will fit in its surroundings before construction starts. Smart Reality can allow users to upload their designs into a 3D map. Another example is using AR for disaster relief and understanding the damage.
4 – Smart Bidding
Construction projects typically work like this: the client needs something built, then there is a contractor who comes in and bids for the work with their cost and projected timetable. The contractor then subcontracts the work for various stages of the construction (cleaning, digging, bricklaying, etc.). So, bids could be off or all over the place because of all the parts that have to be taken into account. Data science can assist the estimation of costs for the bidding process. Software like e-Builder allows contractors to track the data across the life cycle of the project.
5 – Operation and Maintenance Optimization
After construction is over, the maintenance and operation of a building can cost 5x as much as building it. Accurately tracking traffic, stress points, use of heating, etc. can save thousands of dollars. Data science is the way to do these. One solution is offered by Leanheat to track use of heat in a building and optimize it.
ITEMS MENTIONED IN THIS PODCAST:
DID YOU ENJOY THE PODCAST?
  • What other problems in construction can you think of that data science could offer a solution for?
  • Download The Transcript
  • Music Credit: Our Lives Past by Phantom Sage feat. Emily Stiles [NCS Release]

Podcast Transcript

This is FiveMinuteFriday, episode number 252, Data Science in Construction.

Welcome back to the SuperDataScience Podcast ladies and gentlemen, super excited to have you back here on the show. Today we’re continuing our series of episodes about data science in different industries. And the one for today that we have is constructions. Very cool, very exciting industry to be in. Let’s see how data science is applied.
First off, construction is the process of constructing a building or infrastructure. It’s different from manufacturing in that it typically takes place on a location for a known client or as in manufacturing, we just create a lot of stuff and then somehow we sell it off. The revenue for construction in 2015 was $8.5 trillion and is expected to go up to 15.5 trillion. That’s trillion with a T by 2030. The percentage of total GDP for construction worldwide is about 15%. So as you can imagine, that’s a massive industry, a way bigger than, for instance, even mining, one of the largest industry in the world actually. 
And in developed countries, the percentage of total GDP as you can expect is a bit less, actually quite a bit less since the countries already developed. It’s only 6-9% in developed countries. One thing, one caveat before we jump into the examples of data science in construction is that according to McKinsey, 2018 McKinsey report construction is one of the slowest industries to start using AI, which means that the current use cases are going to be quite simple. But on the other hand, that also means it’s a very exciting place to be because there’s a lot of untouched territory and a lot of places where you as a data scientist, as a machine learning engineer, as an AI expert, where you can add value. So listen up, let’s see, what’s going on in the world of construction today.
Example, or use case of data science number one is work process streamlining. So construction projects are very complex. There’s a lot of materials moving around, equipment, people, workers, and this can be very inefficient. At times materials might go in one direction when they might be needed to go into another direction. People might not be using optimal routes for their movements and how they approach their tasks and things like that. So by using wearables and computer vision algorithms, we can pinpoint under-utilized resources to find bottlenecks, optimize routes and so on. The example here is that analyzing this type of data allowed construction companies, Skanska USA to reduce the amount an average worker needs to walk in a day from six miles to four and in turn this boosted productivity by roughly an hour per worker per day. As you can imagine for a massive construction project that can be a lot of time saved and a lot of increased efficiency.
And by the way, as usual, you’ll be able to get all the links for these case studies in the show notes at www.www.superdatascience.com/252.
Example or use case of data science in construction number two is reducing waste. So when I read these figures, I was actually quite shocked. This is pretty shocking stuff here. So construction projects actually have a lot of waste. I never knew. I never thought about this way. In fact, estimates say that as much as up to 30% of the material delivered to a building site ends up as waste. That’s quite a large number, but that’s not all yet. As an example, in the UK, 50% of all landfill is actually construction waste. How crazy is that? 50% of all landfill in the whole of the country, or the Kingdom is actually from construction sites.
And according to one expert, this can represent as much as 25% of a building’s cost. Now you don’t need to take those numbers with a grain of salt. And even you go ahead and half them except for the 50% one because that one came from an official UK government document. Even if you half the other ones, it’s still massive number. So as you can imagine, there’s a lot of opportunity to use machine learning, data science, and other technology to reduce this waste and create a lot of saving opportunities. So the example here is that computer vision like that developed by a company VOCA, V O C A can be used to track the arrival of materials on a site and compare them to a project schedule in real time, to signal under-deliveries and to prevent delays. There we go. So that was example number two or use case number two.
Use case number three, augmented reality. So we usually think of augmented reality or AR as applied to gaming. A lot of examples we’ve heard of is focused on entertainment and games and so on. For instance, Pokémon GO, which was quite, quite popular back in 2015 and still is around. But AR actually has huge applications in construction. For instance, when a building has been constructed or before has been constructed, it’s very important to understand how it will fit within its surroundings, both aesthetically and structurally. And can the ground actually hold that much, how will the whole setup work with old existing buildings and all the existing roads and things like that. And that’s what augmented reality can be used for. We’ve got two examples here. So example one is apps such as Smart Reality allow users to upload their 3D models and superimpose them directly into the environment or on a 2D map.
I had a look at the video, they have a new chip just briefly, it looks quite exciting. Example number two is in 2011 after an earthquake hit New Zealand and augmented reality model of Christchurch before the disaster was used by city planners and engineers to visualize what used to be there, as well as what the actual scope of the destruction was. So as you can see having this information, in augmented reality tools can actually be very helpful in situations like this even.
Example or use case number four, use case of data science in construction number four, smart bidding. So this is an interesting one. My brother actually works in construction. He’s a, or one of my brothers is a civil engineer, so I’ve heard quite a few stories about this. So construction projects typically work like this. So there’s a client, some company or maybe the government needs something built like a bridge or a skyscraper or something else.
A warehouse, maybe a dam, whatever it could be. And then there’s a contractor, so the company that actually comes in and bids for the work, they say, all right, this project, based on your description, based on your requirements, based on all the things that you put out in the documents is going to cost as this much, is going to take me this much time. Well that’s not the end of it. The contractor doesn’t actually build the object, complete the project on their own. The contractor actually subcontracts the work to different, a variety to hundreds if not thousands of different subcontractors. So some subcontractors might go and clean the land and other subcontractor might do the digging. And other subcontractor might bring in the cement and other subcontractor might do the brick laying, somebody might do the seals, another subcontractor will go and do the telecommunications and then subcontractors some will do the plumbing and then the finishing touches.
So construction projects are actually very involved in terms of the amount of parties and different, basically different parties that are working on them. So there’s not just one contractor, especially for the bigger projects, there’s tons and tons of subcontractors and then subcontractors might subcontract that and so on. So estimating the value of these projects is very difficult. Not in just in terms of materials, but also in terms of the actual labour that’s required and how subcontractors might engage. So for instance, a client or an engineer working for the client, for the company that actually needs the building constructed or for the government, they might estimate the price or project to be like $2 million or something. But then at bid might come in at one point $1.5 million, so much lower than the estimate or $2.5 million much higher because it’s hard to keep all those things, to take hold of those things into account.
Or a company might be expecting a massive profits for instance. They’re building a residential building and they’re expecting that once they build and sell off they’ll have massive, massive profit. But in reality, because they didn’t calculate everything appropriately, they might be barely covering the costs or running at a loss. So that’s where data science, artificial intelligence, machine learning can come in to assist the estimates of such projects and their whole bidding process on both sides from the clients, from the contractors and from the subcontractors. And just to reduce that uncertainty and make things clear on where the costs are coming from. And the example here is e-Builder, which is a cloud based construction management solution that gives users performance data across the project life cycle. They’re claiming that users of their software report savings of over 4%. It might not sound like a lot, but imagine a $10 million construction project that’s $400,000 in savings right there.
And as we discussed at the start, the industry is still in its starting stages. If you think you can do better than 4% then you’ve got the green light to go ahead and create something even cooler than that for smart bidding.
And use case number five of data science in construction, operation and maintenance optimization. So once a building is complete, the construction work is over, but the building stays and actually requires maintenance and a good rule of thumb is that it costs about five times as much to maintain and operate an office building during its lifetime as it does actually building it. That’s quite a staggering number, five times as much to maintain a building over its useful life. And what we’re talking about here is for instance, properly heating and cooling buildings or tracking traffic stress levels on bridges. Things like that can dramatically reduce the operating and maintenance costs. And as you can imagine, with the proliferation of data, these are things that are very straight forward, very straight forward.
We think about these that we can use data science for a heating, cooling buildings, calculating when to heat or cool it. Tracking stress level on bridges. That’s also a typical data science, something that we could probably solve it there. The example here is Leanheat is an internet of things solution that uses AI to optimally control and monitor centrally heated buildings. Using this system reduces energy consumption by 6% and peak power demand by 17%.
So there you have it. That’s data science in construction and five use cases with examples. Of course there are more examples and use cases and there are probably even more opportunities because as we discussed at the start, construction is one of the slowest industries to take up AI. Therefore, if you’re interested in construction, well, hopefully some of these examples gave you some great ideas and I truly hope you will revolutionize the space if you get into it.
And on that notes, you can get the sources for this research at www.superdatascience.com/252 if you’d like to dig in further. And I look forward to seeing you back here on the podcast next time. Until then, happy analyzing.
Show All

Share on

Related Podcasts