By Yukihiro Kimura, Chief Operating Officer, NEXT-SYSTEM
As AI is increasingly implemented in various aspects of our daily lives, it is also making its way into an industry that has an outsized impact on all of us, but is often forgotten due to its invisibility in day-to-day activities: the manufacturing industry. With broadening acceptance and understanding of digitization in the corporate sector, the market of digital modernization in factories is rapidly growing and is expected to reach $300 billion by 2023, changing and enhancing the way factories and the people in them work.
Future Factories and Their Need for AI
Worldwide crises have had a great impact on the manufacturing industry, including skilled labor shortages and transportation disruptions, accelerating the adoption of new technologies – including AI – into factories and logistics. According to a Google Cloud survey, 76% of manufacturing companies turned to data, analytics, cloud, and AI technologies due to the pandemic, paving the way for innovative workflows and improved quality control.
The implementation of AI technology in the manufacturing industry is highly anticipated and necessary: according to a 2021 survey from The Manufacturer, an annual growth of 57.2% over the next five years is expected for implementation of AI technology into warehouses and factories, including AI fields such as robotics, natural language processing, machine learning, computer vision, and many more.
The main focus of all these AI fields is to increase the efficiency, quality, and safety of day-to-day operations. Flagging safety hazards, detecting defects in products, preventing stops of the assembly line – these tasks have been overseen by human eyes for a long time. However, humans are prone to mistakes and oversights no matter how attentive they are. AI that can rationally and accurately identify defects or situations out of the ordinary would increase efficiency to prevent mistakes and accidents, and let humans focus on other tasks.
Possibilities of Computer Vision in Manufacturing
The type of AI technology to perform such detection is computer vision. Visual inspection systems use cameras or IoT sensors to identify defects automatically, preventing faulty products from being overlooked. AI is already in use for this type of quality control, identifying deviations from the standard in real-time during the production or for finished products. But this type of visual inspection can also be expanded to assembly inspection, process monitoring, or security monitoring, ensuring that all processes in the factory are running smoothly.
The Important Role of Behavior Analysis
Especially in the field of security, computer vision AI has the potential to bring immense change, not only through object detection, but through human detection as well. The manufacturing industry, especially on-site in factories, is known to be one of the highest-risk sectors to work in, and many accidents do not only result in injuries, but also fatalities. The implementation of AI-enhanced security systems can help reduce fatal accidents and increase overall safety for employees.
For this type of security monitoring, AI behavior detection technologies can play an important role in prevention and early detection measures. Whether someone is in an accident involving machinery or is collapsing due to a health issue – AI behavior analysis can help detect movements or activities out of the ordinary. AI can automatically and swiftly react by setting off alarms, notifying security personnel, or stopping machines. And thinking a little further, behavior analysis could also supervise the human workflow, making sure that processes such as assembly or packaging are performed in the right order, helping to spot inefficient workflows and enhance quality.
Special Demands of Factories for AI
The idea of AI revolutionizing the manufacturing industry is indeed promising, but when looking at the actual implementation, there is still a long way to go.
In manufacturing, every factory creates different products through different workflows. Thus, there is no standardized one-for-all AI solution that every factory can simply start using. Most of the time, factories implementing AI solutions require changes in systems, additional development, customization, and training using the factory’s unique data – which essentially means every factory needs a custom AI system.
And not only that: workflows in factories change, as do their products, so there is a need for further and ongoing adjustments after implementation of a system. Yet expecting manufacturers to perform additional development themselves is unrealistic. What is actually needed is not only the AI system itself, but tools to empower customers without developer skills to build, adjust, and use their own models with their own data: a user-trainable system.
User-Trainable AI Behavior Analysis
Such a user-trainable AI behavior analysis system would idealistically enable manufacturing companies to create their own customized AI system that fits their needs and the workflow in their factory, without any in-depth developer knowledge necessary.
The key to adopting AI into manufacturing lies in its ease of deployment, user-customization, and utilization, which can only be achieved when handing appropriate tools to manufacturers to adjust AI systems and make them their own.
Yukihiro Kimura joined NEXT-SYSTEM in 2009, first working as an engineer developing iPhone applications and tools, but eventually taking over product planning and sales, getting more involved in AR projects and developing not only systems and apps, but content as well. He became chief of the Tokyo office in 2013, and is currently in the position of Chief Operating Officer.