Everything You Need to Know About Machine Learning for Construction Safety | Tech News
Safety has always been a problem in the construction industry. More than 5,000 workers died on job sites in the most recently reported year, and many more were injured.
The good news? Thanks to advances in AI and machine learning for construction safety, all that’s about to change.
Why You Need to be Paying Attention to Machine Learning Right Now
Until recently, machine learning was a future technology that had yet to have a lot of practical application for the construction industry. Today, machine learning is being applied all over the world to substantially improve safety in construction.
Why the sudden emergency of practical machine learning applications for safety? Two primary reasons:
- Cloud and mobile technology adoption
- Massive explosion of data sources in construction
- Recent, rapid advances in machine learning and deep learning technology
When mobile technology became widespread on construction sites, along with it came an explosion of data. Field workers make daily safety inspections using checklists on mobile devices. They track safety observations and notes, snap photos, take video, and record audio all day long, while on some sites drones take additional footage. We are also starting to see the use of sensors/IOT and wearables being used on jobsites to continuously collect data. The average job site now generates 50GB of data.
Meanwhile, cloud technology has made it possible to aggregate all of this data into one location for analysis. Many leading contractors have been using BIM 360 as their construction management software platform and aggregating BIM, drawings, markups, issues, checklists, RFIs, Submittals, clashes and project and business profile and related project metadata over the years. More recently, BIM 360 is also supporting connections with multiple data sources (see BIM 360 integration with 3DR) in the industry such as drone captured images and wearable technology to provide a connected data ecosystem for construction.
Finally, in the past two years, researchers into machine learning, and especially the branch of deep learning techniques, have created rapid advances in what is possible with machine learning. High performing algorithms that can translate speech in real-time and algorithms that can beat us at complex games such as Go are just some examples.
That may seem hard to believe, but it’s true. Properly trained artificial intelligence can now analyze and categorize data not only faster than a similarly trained human, but with greater accuracy. This is a stunning advance, and makes it possible for machines to contribute substantially to job site safety.
Current Applications of Machine Learning for Safety
Forward-thinking construction companies like Layton Construction and Skanska are already using machine learning for safety on their job sites, and technologies like Smartvid.io and BIM 360 Project IQ are making it more powerful and accessible than ever before.
Current safety applications include:
- Analyzing and tagging visual and audio data for safety hazards and unsafe practices
- Analyzing and tagging visual and audio data to record and make searchable best practices
- Identifying potential safety risks and hazards by location, and other criteria based on inspection data, safety observations, issues, checklists, photos, etc.
- Identifying Subcontractors with unsafe practices based on jobsite data
- Identify the most recurring safety risks and hazards across your jobsite based on current and historical data
- Prioritizing targeted safety improvements by subcontractor, project or business unit level based on recurring hazards and risks from present and historical data
These applications are currently available through Smartvid.io and Project IQ. Let’s take a look.
How Smartvid.io Improves Construction Safety
The Smartvid.io team has been instrumental in making the analysis and prioritization of visual and audio data accessible for analysis by artificial intelligence and machine learning. They’ve also developed the machine learning platforms necessary to make that data actionable.
The Smartvid.io platform aggregates visual and audio data from multiple platforms and formats. It then uses SmartTags to pull out relevant safety-related data and analyze it.
It is capable of identifying hazardous conditions based on similarities with previously identified hazards, such as misplaced hole coverings, improperly used ladders, and incorrectly installed barriers
It can also identify individual people in photos and videos, and analyze the presence or absence of safety protocol, including whether they are wearing appropriate PPE. This information can then be tagged and prioritized to create notifications for safety officers, supers, and others who can act to correct the safety issue.
How BIM 360 Project IQ Improves Safety
Project IQ, a technology from Autodesk that is built into BIM 360, uses AI-based deep learning techniques to analyze past and current projects for safety and efficiency and provide targeted warning about delays and threats to workers’ safety. Project IQ sifts through millions of data points from your construction documents, issues, checklists based inspections, related metadata and historical data, and analyzes it to identify and prioritize safety risks on each job, including:
- Project level safety risk analysis and predictions
- Flag Subcontractors with higher safety risks and risky behaviors
- Specific hazards and risks on a project
- Best practices and targeted improvements for mitigating safety on the job site
Safety is also about building a positive culture. Project IQ helps in this aspect by highlighting positive safety work done by Subcontractors and creates a leaderboard from a Safety performance perspective. The Project IQ interface allows you to view safety risk priorities by project, or across business units or regions or by project types or even across your entire organization. You can easily see which of your projects is more prone to risk, and the best practice measures to take to mitigate those risks. With integrations with sensor data, schedule, jobsite photos from drones and Building information modeling information, smart systems such as Project IQ will be able to provide a holistic view of the Project risk. Project IQ is available as a pilot for existing BIM 360 customers today.
Watch: Machine Learning for Construction: A Project Manager’s Perspective
In this webinar put on by Autodesk University and Layton Construction, Cooper Darling shares his project manager experiences involved in helping identify and reduce safety risks on a construction job site with the aid of assistive technologies.
How You Can Leverage Machine Learning for Safety on Your Job Sites
Machine learning for safety is already here and already in use and the best part is technologies such as Project IQ requires no additional set-up or configurations. If you are already using BIM 360, Project IQ just sits on top of your data and works for you. To make the best use of machine learning on your projects, there are definitely some best practices.
Engage the highest levels of your organization
Implementing machine learning requires a cross-departmental effort to organize and coordinate all of your data and best practices, and to standardize across many projects. This requires the vision and support of the company’s leadership.
Digitize your workflows and start documenting ALL project data
If you are doing inspections and checklists on paper, and not digitizing your workflows then you are missing out on leveraging that critical information and making it available for machine learning algorithms to process and predict risk. Make it a best practice to start documenting all project data and start incentivizing project teams to do the same.
Establish standard processes for data capture and categorization
A smart machine is only as smart as its data and training. Establish standard processes and terminology for capturing data on the job site, and for categorizing safety hazards. For example, Layton Construction established a practice for the super to walk the site, and video record and narrate hazards as they encountered them. By capturing video and audio together, and then manually tagging them, they train their AI to recognize similar potential hazards in any video or audio data that is received.
Document best practices
In order to identify and prioritize safety risks, it’s important to understand your safety best practices. Once documented, your AI will be able to tag and notify you when those best practices are not being followed. If you recommend running a checklist based safety inspection every single day, then AI can alert you when you have not completed your safety inspection on time.
Ensure coverage of entire site and all stakeholders
In order to get the most out of the AI, the data must be complete. Put processes in place to ensure that the entire site is captured in the data on a regular basis. Also, ensure buy-in from all project stakeholders. If every team member is logging safety observations, then you have a rich data set for risk analysis.
Connect all the data
Machine learning is more effective the more data it has access to. Choose data platforms like BIM 360 that allow you to integrate and aggregate all of the text, document, model, visual, audio and sensor data coming out of all of your job sites, and make the data available to your machine learning platforms.
Conclusion
Machine learning for safety is already here, and this is only the beginning. We expect the field to explode with new applications over the next few years. The result will be a major transformation in safety in the industry. We look forward to a day when a 100% injury-free job site is the norm rather than the exception. Be ready to take advantage of machine learning applications in safety by investing in capturing and documenting jobsite data in a digital form.
Comments are closed.