Manufacturing in the Time of AI – What to Expect With Data-centric AI Inspection

A recent exchange between leaders from FPT Corporation, FPT Software, Landing AI and Schaeffler explored the potential of applying artificial intelligence (AI) in various methods of inspection and defect detection, within the manufacturing industry.

Identified as the key for further enhancing competitiveness and workforce reinforcement, AI has the potential to be included in various operation processes. One such field, AI in visual inspection – with computer vision and machine learning on the rise – has been becoming more popular thanks to the engagement of top players across industries.

In the recent online sharing, experts from FPT Corporation, FPT Software, Landing AI, and Schaeffler have discussed their visions for the future of AI in Real-time Quality Inspection. All parties emphasised on the use of data-centric approach to shorten AI training duration in machine learning and addressed critical issues faced by brownfields factories.

Until just recently, factory owners were equipped with rule-based vision inspection, which required IT experts to write pages of rules for the algorithm to detect product defects. These systems are unable to “self-learn” or take on new status outside-of-the-book. Solely capable of detecting errors in their rule books, rule-based inspection might miss anomalies, requiring manual inspection by onsite staff.

“Rather than having a vision engineer write pages and pages of rules, you can now show some data to the AI system and build a system that works just as well, but much faster”, said Dr. Andrew Ng – Founder & CEO of Landing AI.

Data-centric AI continuously proved to be highly efficient, significantly reducing the time to bring trained AI into use. In a recent project between Landing AI and a Japanese manufacturer, for example, compared to rule-based inspection that took up to 6 months to build, data-centric AI only took 6 minutes to learn from the qualified data input and accurately identified product defects.

Recommended AI News: Husqvarna Affirms its Commitment to Autonomous Mowing and Leadership in Robotics as a Service (RaaS) with Expanded Investment in Robin Autopilot

Another challenge is multi-variations among different production lines, not to mention production plants. Each variation required dozens more rules for computer vision to abide by. Therefore, factories are constantly looking for new solutions, especially AI ones, integrating with high-resolution cameras and are capable of self-learning.

“Applying AI broadly in our manufacturing processes is a core element of Schaeffler’s digital strategy and will enable the next level of optimization within production”, Roberto Henkel – Senior Vice President Digitalization and Operations IT of Schaeffler.

Earlier this year, Accenture and Frontier Economics estimated that by 2035, AI applications might boost labour productivity by as much as 40% across 16 industries. AI would also bring $3.8 trillion (up 45%) in added value to the manufacturing industry, according to the report.

The technology was considered much more reliable in identifying microscopic defects without compromising the intense production paces. As such, product quality, labour productivity was also expected to rise while saving manpower, thereby reducing production costs.

In Vietnam, FPT Software has collaborated with Landing AI to bring the LandingLens – an MLOps visual inspection solutions – to hundreds of its customers, including Schaeffler.

Dr. Truong Gia Binh – Chairman of FPT Corporation asserted: “Data-centric AI can also work for sound inspection, electromagnetic inspection or light spectrum inspection.”

This potential direction quickly became an interesting topic. With excitement, the leaders shared a vision for data-centric AI to inspect the quality of material and internal error without physical interaction, as well as receiving sound cues to detect defects.

Recommended AI News: Crescent Bank Partners with Informed.IQ To Simplify the Lending Process for its Auto Dealers

However, all parties needed to overcome the challenges in change management and staff training before successfully applying cutting-edge technologies.

“We need easy-to-use platforms which bring high class technologies to the shop floor and make them (staff) capable to use it”, said Roberto Henkel. This was also the problem which Dr. Ng was striving to solve, making AI accessible for everyone. “Data-centric AI technology is the key that makes it feasible for people to engineer the data”, Dr. Ng shared.

Having accompanied more than 1,000 global partners to achieve digital transformation, FPT has helped transform both business structure and technology for companies in various industries, providing them with staff training and new technologies updates. “We work on the 3 C – community, communication and building shared content,” Dr. Binh said.

The FPT chairman emphasised: “The key point is change management and people management. We’re talking about business formations, technology transformations, and people’s transformation.” With an intensive background in the education field, Dr. Andrew Ng also expressed his enthusiasm for training staff on new technology, ensuring all key personnel learn the desired skill sets.

The high-quality workforce for high-tech industry was one of Vietnam’s current competitive advantages within the Southeast Asia region, according to a recent remark from the representative of Savills. This attracted a wave of investment from large tech companies from the US and around the world.

Recommended AI News: Leavened Announces New ConfigAI Feature

[To share your insights with us, please write to sghosh@martechseries.com]