Occupational injury accidents in Taiwan have been increasing in recent years. For instance, drug factory employees accidentally fell from a 3-meter tank, marble factory workers were crushed by a 500-kilogram slate, three workers were injured by a biogas power plant gas explosion, and steel plant technicians were accidentally pinned in the machine. These occupational accidents in various factories in Taiwan raised awareness of a stark fact—enterprises need more effective risk prevention and safety management solutions to reduce the risk of occupational injuries.

“Although many companies installed surveillance cameras onsite, the image data is not fully analyzed and utilized, and is difficult to detect potential occupational injuries in real time, ” explains Jackie Yang, Product Director at ASUS AICS. “In the past, our surveillance cameras could only be used in retrospective, but these accumulated images should be effectively used to proactively predict and prevent future risks.”

Therefore, AICS launched the EHS Management Service deployed on Microsoft Azure, with Azure Data Factory, Azure Databricks and Azure Machine Learning for data collection and model training. For example, in Personnel Tracking & Management, each personnel’s location and movement data will be sent to Azure Databricks for unauthorized entry analysis and cross-zone movement detection; In Man-Machine Collaboration Risk Identification, footages of complex operations are sent to Azure Databricks to determine risk scores according to pre-defined standard operating procedures in real time along with Azure Kubernetes Services (AKS) to continuously improve AI modeling accuracy. Also, Azure Machine Learning is deployed for MLOps to upload collected image data into containers in Azure Kubernetes Service.

Story image 3

With these solutions, it can actively identify and prevent potential risks through AI deep learning models. Furthermore, with flexible public cloud scalability, turn-key deployment, and elastic provisioning, the service can significantly reduce upfront cost of implementing AI solutions and dramatically improve development flexibility. As a result, this cloud-based, man–machine collaboration solutions help monitor, educate, and prevent occupational injury.

Two pain points in occupational injury risk management: Ineffective use of image data, increasing demand for customization

Yang indicates that occupational injury is a seriously underestimated issue worldwide. 88 percent of the causes of occupational injuries are related to human misconduct, 45 percent of which occurred when workers operated a machine. It follows that occupational injury can be effectively managed if the person’s behavior can be effectively managed.

However, behavior management is difficult and unpredictable. In the past, many cameras were set up at a factory site, and monitored by a security or factory process manager to ensure frontline workers were following the standard operating procedure (SOP). Real-time monitoring is labor-intensive, and it is a challenge to immediately detect non–SOP-compliant behavior with the human eye. Therefore, these surveillance cameras were unable to provide a real-time warning, although they could only provide historical footage to analyze the cause of the accident.

AI deep learning technology can fill the gaps between traditional technology. The AI deep learning model replaces the human eye to automatically predict and analyze image behavior, avoid human negligence, and quantify the concept of behavioral risk for further management. However, AI applications in occupational injury risk management require system integrators to provide customized services. The higher the degree of customization required, the more difficult it is to rapidly replicate those applications to large and multiple manufacturing fields.

Story image 4

Mitigating field risks: AI reduces factory occupational injury risks by 90 percent

Since the two major pain points of occupational injury risk management—ineffective use of image data and increasing demand for customization—the company sought to develop AI technology as a modularized platform, and it launched the EHS Management Service. Through streaming surveillance images, the AICS platform can label dangerous areas from the monitored images, and the AI deep learning model automatically learns and combines with online data feedback to continuously train and improve analytical accuracy.

According to Yang, “Anyone without AI background can deploy AICS EHS Management Service with deep learning models to meet their customized needs and correct AI logic at any time.” By deploying on the Azure cloud platform, these models can be flexibly updated without any vendors customizing services. Therefore, EHS Management Service reduces the implementation difficulty for AI solutions from a technical and cost perspective.

According to Andrew Lu, Azure Solution Sales Lead in Taiwan, “Microsoft Azure can elastically scale according to your data throughput. Besides, with [the] pay-as-you-go pricing model, organizations do not have to worry about infrastructure to store data, but can effectively reduce hardware cost.”

Story image 5

Bridgestone, a well-known tire manufacturer, implemented AICS EHS Management Service to manage safety risk. “Bridgestone not only reduced the risk in the field to less than one-tenth after six months of implementation, but [it] also used images on AICS as education and training material to raise internal security awareness,” says Yang.

’Empowering people with AI’: High brand trust and comprehensive support from Azure and ASUS together

When asked about the reasons of choosing Azure as the company’s cloud platform, Yang points out that the turnkeys are brand trust and professional support from Microsoft.

AI needs to process massive amounts of data—which involves the privacy of individuals and businesses—and is prone to ethical concerns. Therefore, the attitude and best practices on data protection of a cloud service provider become crucial. Microsoft’s data management, compliance, and security practices are known to be rigorous and highly reliable.

Story image 6

In addition, Microsoft not only provides the cloud platform, it also offers comprehensive and professional technical advisory support to assist AICS in solving problems in the product development process. According to Lu, Microsoft’s technical advisory team not only regards ASUS as a “customer”, but as a “partner.” The Azure team stands in the customer’s shoes to continuously optimize cost and performance of the customer’s solution, from validation in the early stages of product development, to solution architecture tailor-making fitting to AICS’s customer requirements. Not only that, the Azure team also regularly shares new features and services with AICS team after solution deployed.

Lu echoes the words of Tai-Yi Huang, CVP and CTO at ASUS. “If we want to measure the impact of your solution to the society, it will depend on how many people are actually using your service,” he explains. Working with the Azure team on the two key solutions of AICS—”Intelligent Industrial Safety” and “Intelligent Healthcare”—together they achieve the mission of field safety improvement, helping people across society to work in a better place.

In the future, the Azure team will continue to take advantage of innovative services to enhance AICS EHS Management Service solutions. The team will share the latest market trends and product updates with the ASUS AICS team, to help manufacturers to improve safety in the workplace and realize the vision of “empowering people with AI.”

Learn more about AICS EHS Management Service.

As featured on Microsoft’s Customer Stories. Original Chinese article by TechOrange.