Dow uses vision AI at the edge to boost employee safety and security with Azure

A center for digital innovation

Dow Inc. established its Digital Operations Center as a place where employees explore and evaluate emerging technologies to determine how the company can use them to boost employee safety, generate value, and improve operational efficiencies. 

“We bring together experts in a variety of areas, including analytics, AI, robotics, mobility, sensors, and digital thread,” explains Michael Dessauer, Senior Solutions Manager at the Dow Digital Operations Center. “Together, we decide how to integrate new technologies into existing systems and how to support and manage them in the long term.”

Dow is a leader in materials science, focused on delivering innovative and sustainable customer-centric solutions. The company has a global presence, with manufacturing facilities in more than 30 countries. The technology innovation specialists at the Digital Operations Center—part of a larger corporate digital manufacturing acceleration program—are using AI to develop new products and improve existing assets and processes, currently an important focus of the Center’s work.

Dow has a strategic relationship with Microsoft and has chosen Microsoft Azure as its trusted cloud platform. Even so, the Digital Operations Center evaluated a wide range of technologies before choosing Azure AI services. “We concluded that Azure offers the flexibility and scalability we need, with innovative services that facilitate cutting-edge AI development,” says Mayur Andulkar, Analytics & AI Specialist at the Dow Digital Operations Center. “And with this platform, we can easily design custom solutions or tie in existing third-party products.”

Because of the sensitive nature of its work, the Digital Operations Center made security a key factor in its decision. Once again, it determined that Azure had the built-in capabilities that it needed. “We have a cloud team that’s dedicated to hardening our services and storage, and Azure provides the capabilities to maintain strong security for our machine vision solution data in the cloud,” says Dessauer.

An AI solution for leak detection

The Digital Operations Center has explored a variety of possible uses for AI, including the use of video analysis for detecting containment leaks within its operations environments. To supplement existing safety programs, Dow wanted a robust, real-time solution and found it by combining AI and Internet of Things (IoT) technology with Microsoft Power Apps.

The Digital Operations Center designed an intelligent solution that monitors facilities around the clock and raises an alert if it detects a leak. “We used IoT-enabled cameras to build a system that looks for possible leaks and Power Apps to quickly build a notification app,” says Andulkar. “If the system detects an anomaly, it notifies the operator via voice, text, email, and the app, which runs on mobile devices and desktops. The operator can then take corrective action immediately.”

An essential element of the monitoring solution is Azure Video Analyzer, which extracts metadata from video feeds and provides real-time analytics. Importantly, Video Analyzer can run on edge devices, which reduces latency—a critical benefit in a leak scenario, when time is of the essence. Processing video at the edge also cuts down on bandwidth and cloud resource utilization, thus lowering costs.

“Because of its many built-in capabilities, we saved a lot of time using Video Analyzer instead of doing custom development,” says Andulkar. “This brought us to a solution much faster and freed us up to focus on the business logic and AI models that sit on the edge device and manage the detection and alerts.”

Dessauer adds, “In this situation, we were looking at more than just the speed at which we could put together a solution for monitoring video at the edge—there are open-source tools we could have used to cobble something together. But we need to scale our solution to thousands of deployments, and we couldn’t have done that so quickly without Azure Video Analyzer and the other Azure services we’ve used.”

The complete edge vision solution incorporates a variety of services, all of which the Center uses Azure DevOps to coordinate. The solution streams the video input through Video Analyzer for inference using a trained AI model deployed on edge devices. When a possible leak is detected, the Azure IoT Edge device sends an alert to Azure IoT Hub. The solution then forwards the alert to a custom-developed app with video footage of the situation. The Center uses Azure Machine Learning for model generation and retraining, which combines data from the cameras with operators’ knowledge to generate the intelligence needed for determining a possible leak. Operators evaluate the alerts generated by the solution, and as they flag any as valid or erroneous, the model becomes more intelligent by learning what is and isn’t an actual leak. The solution uses Azure Data Factory to orchestrate the data pipeline by pulling data from the edge devices through IoT Hub and connecting to Azure Machine Learning storage.

Each operator can customize the solution’s parameters to suit their needs using Azure Logic Apps. “We built the solution so that an individual operator can specify what sorts of alerts they receive and when they receive them,” says Dessauer. 

A flexible foundation for safety and security solutions

The Digital Operations Center focused on leak detection with its initial development, but it can use its edge vision pipeline with other types of detection or sensor devices by changing the AI model and business logic. For example, the Center used its AI framework with a model for personal protective equipment (PPE) detection to ensure that all personnel are appropriately protected in industrial settings. Other potential uses include monitoring unstaffed security gates and securing facility perimeters to prevent unauthorized access.

“We’ve identified many exciting possibilities for combining AI with IoT,” says Dessauer. “We can put sensors in places that are difficult or dangerous or cost-prohibitive for operators to go. And with advancements in AI models, we have numerous opportunities to make productive use of all the data we’re now able to collect.”

Dow ultimately wants zero safety-related events at its facilities, and the Digital Operations Center’s work is helping the company get there. “In addition to raising our already high safety standards, our video monitoring solution benefits our customers, helping them get their product on time, without the delays a leak could cause,” says Andulkar. “The use of Azure services—especially Video Analyzer—will help us reach our safety goals.”

Find out more about Dow on Twitter, Facebook, and LinkedIn.