USG Boral Launches a Trailblazing AI Safety System on AWS

USG Boral

From Detection to Prevention

USG Boral is a leading manufacturer and supplier of gypsum-based wall and ceiling systems in the Asia Pacific and Middle East regions. With its headquarters in Malaysia, the company operates across 13 countries and has a mission to deliver innovation that helps customers work smarter, do more, and build better. In 2018, a forklift collided with a pedestrian in a warehouse at USG Boral and this triggered a breakthrough safety innovation in the Asia Pacific region. Safety is a core organizational value, so when the incident occurred, the business immediately took measures to prevent a recurrence.

At the time of the incident, the company relied on CCTV footage to record activities on the floor of warehouses, and employees would manually review the footage stored locally following any incident. USG Boral wanted a more proactive and intuitive solution that would help prevent accidents in the first place.

“We make use of the analytics behind AWS to churn out the data without having to hire data scientists or spend time … [on] manual tasks.”

Calvin Ng, IT Director, Infrastructure and ANZ, USG Boral

  • About USG Boral
  • USG Boral is a leading manufacturer and supplier of gypsum-based wall and ceiling systems, with offices in 13 markets in Asia Pacific and the Middle East. It delivers innovation to help people work smarter, do more, and build better. Safety is a core value of the organization. 

  • Benefits
    • Processes 12 images per second for near-real-time analytics
    • Records less than 300 milliseconds of latency for vision analytics
    • Offers dashboard visibility and customized reports for management
    • Improves safety capabilities to help prevent accidents
    • Provides a flexible framework for expansion and integration of new ML services
  • AWS Services Used

Niche Warehouse Solution

The company approached, Bigmate, a Select Technology Partner in the Amazon Web Services (AWS) Partner Network (APN), to develop an intelligent warehouse safety system that would sound an alarm when objects or people came within the 3-meter safety range of forklifts. AWS was the platform of choice for the project for two main reasons. First, it is hardware-agnostic, so teams can continue to evolve vision processing as new technology becomes available. Second, it offers a flexible framework for expansion, which means that architecture can be modified quickly to tweak security, logging, or networking.

This would be USG Boral’s first artificial-intelligence (AI) venture—and likely a first for the manufacturing industry, at least in the Asia Pacific region. “We knew this would be a niche solution, and there was nothing existing in the market yet. We spent a lot of time consulting with AWS and Bigmate to develop the framework and looked at how it could achieve what we wanted from a business and safety perspective,” says Yeow Kok Weng, CIO at USG Boral. This included working closely with staff on the ground to ensure the approach would support better safety outcomes without obstructing work activities.

Actionable Insights Lead to Re-education

In addition to preventing accidents, a key performance objective for the project was the provision of data in a unified format on the number of accidents and alarms, as well as the number of “near misses” per work site. This would allow for quick managerial oversight and monitoring that were impossible with traditional CCTV systems. “We know that people can get complacent over time and need support through continuous safety education,” explains Calvin Ng, IT director of Infrastructure and ANZ at USG Boral. The AI-powered solution would facilitate the implementation of re-education programs based on actionable insights. The new solution, dubbed Warny™, was nine months in development.

According to Bigmate, Warny is one of the most advanced vision applications on the market. AWS’s Internet of Things (IoT) technology is the foundation of Warny, particularly AWS IoT Greengrass and AWS IoT Core. AWS IoT Greengrass seamlessly extends cloud capacity to the warehouses so that they can execute predictions and actions based on trained machine-learning (ML) models, even when not connected to the internet. Local light and siren alerts go off when potential incidents are detected.

With AWS Lambda, Warny can run Lambda@Edge functions so that code is automatically executed, enabling industrial sensor control that can eventually extend to the forklift itself. “The ability to configure Lambda@Edge functions in AWS IoT Greengrass allows USG Boral to carry out local processing and remotely manage updates to gateways as needed. This granular and flexible approach means that they can continually evolve even at an accelerated pace,” says Brett Orr, general manager at Bigmate. In addition, the company uses Amazon CloudWatch to monitor gateway and cloud resources.

Real-Time Dashboard Analytics

The ML model is still being honed, but Warny can already perform near-real-time analytics, processing at least 12 images per second. It must continually detect, track, and calculate distance and velocity between objects, which require superior vision analytics with a latency of less than 300 milliseconds to enable rapid alerts for potential incidents. If an object enters the 3-meter safety radius, an alarm sounds.

A major benefit of Warny is that USG Boral can assess why near misses are happening and improve safety on the ground through aggregation of data from its many sites for analysis on the cloud. “We can use reporting and dashboard analytics according to specified output parameters to look at how we are progressing with safety for our employees,” Calvin says. “We make use of the analytics behind AWS to churn out the data without having to hire data scientists or spend time extracting data or doing manual tasks.” When a near miss occurs, text messages and emails are immediately sent to managers. Separate reports go to C-level executives, compiling regional data on accidents and near misses.

IoT Inspires Innovation

To date, Warny has been tested at one building site in Australia and will be rolled out to 10 more sites in the country over the next few months. An extended rollout across all 13 countries where USG Boral operates has also been planned and will be facilitated by Bigmate’s extended partner network. USG Boral has relied on Bigmate not only for selecting the best-fit technology stack, but also for updated advice on evolving safety standards in countries in which it operates.

With AWS, teams have developed a roadmap for future AI and ML innovations built on Warny. In one example, USG Boral will use Amazon SageMaker and Amazon SageMaker Neo to detect and ensure that workers are wearing safety equipment such as protective helmets, safety glasses, and high-visibility clothing. “With all the solutions in the AWS portfolio, there’s a lot of opportunity for us to use technology to help improve safety in our workplace and other business initiatives,” says Calvin. “IoT is key to that digital journey, and we can achieve much more moving forward with AWS.”

Learn More

To learn more, visit Machine Learning on AWS