The building information modeling (BIM) industry was valued at over $5 billion in 2019, and is expected to grow to over $15 billion by 2027, according to Allied Market Research. That represents a compound annual growth rate of over 14%. Builders, architects and designers are realizing that advanced tech tools are a valuable and important part of their businesses.
The most innovative companies in the construction and engineering industries are far more likely to be using technology like BIM, drones, virtual and augmented reality, and AI or machine learning, according to KPMG’s 2019 Future Ready Index. However, KPMG found a gap between tech-savvy firms and others in adoption of less advanced technology like basic data and analytics, and mobile platforms.
Artificial intelligence and machine learning are unlocking opportunities for construction firms to “harness project data, organize it, interpret it and uncover patterns faster” in order to be more successful, according to Allison Scott, director of construction thought leadership and customer marketing at Autodesk. Scott spoke on a webinar in July explaining how these advanced technologies are becoming commonplace, and how construction companies can do more once they get a grip on data and how to use it.
Scott pointed out that data by itself is not that useful. To turn data into information, and ultimately, wisdom, users need to put it into context.
“You get information when you start to make data useful,” she explained. “Data becomes information when it’s processed, and context and interpretation is applied so we can derive relations to other things.”
The types of data that artificial intelligence and machine learning draw from go beyond materials costs, equipment lifespans or worker hours. Drawings, contracts, project photos and emails between workers all represent sources of data that can be useful to companies trying to identify opportunities or risks within their business. They just need software that can help them find and analyze relevant data.
“AI is making it easier to translate information into knowledge, and two-thirds of IT decision makers expect their budget for things like automation to grow in the year ahead,” Scott said.
Those decision makers are using automation not just for tracking construction progress, “but also for accelerating the move to more leading indicators for things like health, safety, quality and productivity KPIs.”
Making construction safer
Safety is a critical measure of project success, and data tools can help builders monitor compliance and prevent injuries on jobsites. Identifying the steps that led to an accident can be tricky for builders. Crew members may disagree about what led to an injury, and accident reports may have conflicting or missing information.
Trying to anticipate an injury, or the conditions that might lead to an injury, can be even more difficult. Despite construction’s earned reputation as a dangerous industry, individual workers, even experienced ones, may witness “hundreds of near misses and first aid injuries, dozens of medical cases and lost work time injuries, and, perhaps, a few permanent disablement injuries and fatalities,” according to a 2016 paper published in Automation in Construction. “Because of this limited experience with incidents, they may misdiagnose the risk of a given construction situation,” the authors wrote.
The study, “Application of machine learning to construction injury prediction,” found that machine learning algorithms can be trained to predict the type of injury, the body part affected and the energy source causing the injury. One area where machine learning did not accurately predict outcomes was in injury severity. The authors suggested that unlike the other outcomes, there’s an element of randomness to how severely a worker will be injured.
Even with that caveat, “underlying patterns and trends exist and can be uncovered and captured via statistical learning when applied to sufficiently large data sets,” according to the authors.
Because accurate predictions rely on large data sets, “it’s really impossible to detect any kind of pattern without machine learning classification,” Manu Venugopal, Autodesk’s head of construction data and analytics, said.
COVID-19 presents a clear opportunity to use data to keep workers safe.
Machine learning is “helping measure social distancing and PPE compliance also,” Venugopal said. “I can’t overstress the importance of creating standards and benchmarks, and the current COVID-19 situation surely has shown the importance of having data to benchmark and improve your performance.”
Scott added, “The pandemic has definitely upended business as usual, and to ensure the health and safety of people, many contractors have started to use technology and data to understand areas that may put their workforce at risk.”
Quality starts at the beginning
The data science team at Autodesk turned its attention to RFIs and what builders and contractors can do to limit how many of these expensive tasks they have to deal with. Pat Keaney, director of product management and intelligence at Autodesk, noted that on average, RFIs cost about $1,000.
In an analysis of customer data, Autodesk found projects that had more than 6% of their value in change orders suffered margin erosion, according to Keaney. The company also found that the most successful companies weren’t closing all RFIs quickly, just the most critical RFIs.
“The takeaway here is that you need to expect some change, but you also need to invest more time and resources into the up-front earlier phases of project planning to keep that change in the right zone,” she said.
RFIs have a significant impact on project costs and timelines. Autodesk used machine learning models to automatically tag RFI data with a root cause.
“More than 70% of RFIs could have been resolved in design,” Venugopal said. “A more robust design review process gives you the opportunity to identify and mitigate some of these problems early on, and then make sure that they don’t become much more costly.”
Human nature is biggest data obstacle
One of the biggest challenges that builders will face when implementing these high-tech solutions is decidedly low tech. Data tools need a large sample to draw conclusions from. If employees don’t see the value in a particular data point, or simply don’t have an answer, they might leave it blank.
Keaney put it bluntly: “People don’t like to fill out forms.”
Even if employees do buy in to the value of data science, there may be disagreements about qualitative data. For example, users may disagree about the root cause of an RFI or the incidents leading up to an accident.
Keaney warned that tracking a particular data point doesn’t make it useful.
“If you measure the wrong thing, it can actually inadvertently drive the wrong behavior,” she said.
What should builders do
One way builders can use data to improve the quality and safety on their jobsites is to standardize processes across all their projects.
“Standardization is the key to project success,” Venugopal said. He noted that Autodesk found “projects using standardized checklists for proactive quality management had a much higher rate of success compared to projects that are reacting to things and managing them as they come up.”
Creating standards and common nomenclature across all projects helps builders avoid time wasted on ad hoc issues, Venugopal said.
Virtual design and construction is a starting point for builders who want to incorporate data into their processes, Keaney said.
Related: The downside of VDC
“If you start that today,” she said, “I guarantee you, five years from now, all the data you create today will be valuable to you in ways in the future that you cannot even imagine.”
Venugopal encourages builders to start by looking at their workflows and creating a checklist that can be used across their all their projects.
“If each project wants to make slightly small tweaks, that’s fine, but don’t let them start from scratch,” he said. Then they need to implement some kind of oversight process to ensure crews are actually abiding by the checklist.
Each person on a project needs to be empowered to use the checklist and collect project data, Venugopal said. “If it’s not working, you might want to adjust and optimize your final checklist. That way you are creating programs that work for everyone.”