Transforming manufacturing with artificial intelligence

Share this post:

KIST Europe manufacturingArtificial intelligence (AI) is no longer just a field for academic researchers; machine learning and deep learning are becoming mainstream technologies that any organization can harness. This could have dramatic implications for many industries, including manufacturing.

The impact of AI on manufacturing is likely to usher in a whole new era of industrial development. The first three industrial revolutions were triggered by the introduction of mechanical, electrical and digital technologies, respectively. Now, AI will be the driver of a fourth revolution known as “Industrie 4.0.”

The mission of the Korea Institute of Science and Technology (KIST) in Europe is to contribute to the globalization of Korean research and innovation by building an open platform where prominent Korean and European institutions and industrial partners can collaborate. Since manufacturing is a vital contributor to both the Korean and European economies, Industrie 4.0 is one of our main focus areas.

Smarter manufacturing

KIST Europe and IBM are partners in SmartFactory-KL, an innovative manufacturing facility built with modular components that can be reconfigured for different manufacturing tasks. All the components are instrumented with sensors and connected through the Internet of Things (IoT) to a “digital twin,” a complete digital replica of the factory’s physical assets, processes and systems, running in the IBM Cloud.

To demonstrate the value of this technology to our industrial partners, we looked for a project that would use AI to solve a real-world manufacturing problem. We decided to look at weight measurement, which is a common part of many industrial quality management processes.

Weight measurement can be used as a proxy for quality because industrial machines are calibrated to very fine tolerances. If a product is substantially heavier or lighter than it should be, it indicates that something unexpected happened during the production process, and there is likely to be a fault.

However, the high-precision scales that manufacturers use to take weight measurements are extremely sensitive and can be disturbed by other activities in the factory. If a heavy forklift drives past the scale while it’s taking a measurement, for example, you may get a false reading.

As a result, you might discard an item that is not actually faulty; or, even worse, you might allow a faulty item to pass the quality control check. If faulty products reach the market, it can impact customer satisfaction and lead to expensive warranty claims.

But what if your scales were smart enough to know when their own readings were inaccurate, and automatically decide whether to take a second measurement? We wanted to show how AI can make this possible.

Harnessing machine learning

Data scientists from KIST Europe and IBM used IBM Watson Studio to design, train, test and deploy a machine learning model that can predict whether a given measurement is likely to be reliable. The ability to select and combine the wide range of modeling techniques that are available in Watson Studio was a major advantage. It would have taken weeks for us to develop the models by ourselves from scratch.

The deployed model was trained on a data set containing more than 1,000 real measurements captured from the smart factory IoT sensors. Each iteration of training took just three to five minutes, so we saw the results improve very quickly.

The model’s predictions are now correct in 98.1 percent of cases, and we’re experimenting with additive learning techniques to make it even more accurate in the future.

AI-powered quality management

The weight measurement project shows how manufacturers can use AI to solve real-world problems on the production line. It’s a very simple example, but it shows how the combination of instrumented machinery, IoT and AI can work together to enable a decentralized, self-optimizing manufacturing system in which the machines themselves can make correct decisions without requiring human intervention.

If these Industrie 4.0 principles are applied more widely, the result could be lower defect rates, fewer returns and warranty claims, and increased safety for employees and customers alike. Looking at the bigger picture, it will help to build a stronger future for Korean and European manufacturers as pioneers of the fourth industrial revolution.

Read the case study for more details.

Marco Hüster, Business Lead AI Implementation, KIST Europe

More Cognitive stories

Bringing AI and esports together with Watson on cloud

Whether players are physically on the field or controlling virtual avatars in digital settings, IBM Watson is helping sports fans catch the most breathtaking moments. At the 2019 Game Developers Conference this week, IBM showcased how it’s using artificial intelligence (AI) on the IBM Cloud to make esports fan experiences better and improve player performance. […]

Continue reading

Providing around-the-clock call center support using a cognitive virtual assistant

According to Investopedia, financial technology has grown explosively since the mobile internet and smartphone revolution. Originally referring to computer technology applied to the back office of banks or trading firms, “fintech” now describes a variety of financial activities, from depositing a check with a smartphone to managing investments, generally without the assistance of a person. […]

Continue reading

How to remove the top 3 barriers to AI adoption in business automation

Many businesses are looking to automation to increase productivity, save costs and improve customer and employee experiences. For example, imagine a large insurance company that processes millions of claims a year. Around 60 percent of its claims are automatically processed, but it’d like to automate 85 percent of claims and reduce error rates, reduce costs […]

Continue reading