Putting the intelligence in AI for the plastics industry

Artificial intelligence has the potential to enhance precision and productivity, increase uptime and bridge the skills gap.

The transition to AI will benefit plastics processors by enabling predictive maintenance, reducing scrap and energy usage, and shortening time to market, according to Samantha Peterson in her e-book, The Benefits of AI in Injection Molding.

From monitoring the behavior of raw materials to detecting process variations to real-time assistance for problems, AI offers manufacturers tight control over the entire production process, said Peterson, content marketing and events manager at Traverse City, Mich.-based RJG Inc.

Although she focused on injection molding, AI is playing a bigger role in extrusion, blow molding, thermoforming, tool design, part design and more.

"AI's capacity to analyze datasets and derive actionable insights enables you to optimize production processes and maximize output. By leveraging AI, you can fine-tune production parameters in real time, optimizing resource allocation and enhancing overall productivity," Peterson writes.

AI gives employees tools to fix a process no matter their skill level, which helps build morale and pride, Peterson says.

"And it does this all while freeing up engineering time so they can focus their time and energy on larger, more impactful projects," she says.

But first that artificial intelligence has to be gleaned from a variety of sources. Think training data, work orders, PowerPoint presentations, tool setup sheets and other related information such as text, images, videos, audio recordings or even scribbled notes on stained paper.

So far, most AI in the plastics industry has been limited to machine learning, which uses data to make predictions, like a vibration sensor detecting a motor failure. However, another subset of AI technology called generative AI is creating new content based on information that's analyzed. Generative AI can use the patterns and structures of all the inputs to generate new data with similar characteristics when asked about measurements, methods, malfunctions and more.

With this capability, plastics processors will be able to store the knowledge of retiring or relocating senior engineers and technicians, bring new hires up to speed more quickly, and access a secure global database related to machine operations, according to Derek Moeller, founder and CEO of CognitionWorks. Moeller also is the president of Surain Industries, a family operation founded in 1964 that does business as McConkey Co. and manufactures horticultural goods in Sumner, Wash.

Founded in March 2023, Seattle-based CognitionWorks specializes in generative intelligence for the plastics industry. The firm trains a processor's AI with manufacturing documents the client already has.

At McConkey, AI is used to interact with everything from work orders to manuals, procedure books and other documentation. The goal is to replace institutional knowledge at risk of going into retirements.

A lot of the early interest in generative AI is related to the shortage of skilled workers, Moeller said.

And while AI may help fill the gap in workplace shortages, Michael Cicco, CEO of robotics supplier Fanuc America Corp., said during a keynote presentation at NPE2024 that it can also help recruit young workers by presenting them with a new idea of what manufacturing can be.

The placement of automation equipment such as robots in educational settings is helping spur more interest among students in the manufacturing sector, Cicco said.

"We spend a lot of time sitting here talking about 'WTF?'" Cicco said. "And this re- ally means, 'What's the future?' We think about that quite a bit. ...We, as manufacturers, need to keep hiring people, and we can't because of the fact that there's not enough people out there.

"So jobs keep opening up, there's not enough people to fill them."

Tightening labor market

"In the United States, for the overall economy, unemployment is 3.8 percent, but in the plastics industry specifically, it's only 2.6 percent," Moeller said in an email. "That's like a third lower than the broader economy, and that introduces a lot of challenges. One is that pay is going up."

In response to the tighter labor market, plastics processors have been raising hourly rates and salaries to retain talent. However, even that's becoming a shorter-term solution.

In 2012, plastics industry tenure averaged 6.1 years. In 2022, it was 4.7 years, a 23 percent decline.

The outlook continues to be grim.

"Over the next 10 years, 3.8 million jobs are going to need to be filled specifically in manufacturing, and a good portion of those in the plastics industry," Moeller said.

About 2.8 million of those openings will be the result of retirements, he added.

"The people who have been at your company the longest and have the most industry knowledge are leaving. There's going to be a mass exodus of this skilled technical knowledge from manufacturing broadly and plastics processing in particular," Moeller said.

Retaining knowledge

Most companies have easy access to their explicit information, like manuals, schematics and documents. These materials usually can be easily duplicated for generative AI in a week or two. Even an engineer's note scrawled on a piece of paper can be processed.

"AI has an ability to take very messy information — like handwriting on paper stained with hydraulic oil — and extract the information to give us perfect answers to questions and show us the actual citation," Moeller said.

He described explicit knowledge as durable.

"Explicit knowledge does not walk out the door every day. It is part of your company, and it stays there permanently. The disadvantage is that it's kind of hard to find. It might be in a processing guide that came from a conference or an engineer's note on page 75 in the fifth binder from the left. What's the chance it will be found when it's needed?"

The chances are very good with AI, he added.

However, tacit information, which is out on production floor in the form of expertise and experience of the mold setters and technicians, is harder to obtain. "This informal knowledge about how to do things is just stored in their heads. It's incredibly valuable, and we cannot discount what they can contribute," Moeller said.

Tacit knowledge needs to be converted to explicit knowledge by writings and recordings. Something as informal as a cell phone recording of a company expert's rambling, stream of conscious can be processed. "AI is really good at taking this kind of informal knowledge and making it structured," he said.

He points to studies that indicate about 80 percent of a company's knowledge walks in and out the door every day as employees come and go.

"We're going to lose a lot of that knowledge over the next 10 years unless we do something about it," Moeller warned.

Sharing knowledge

AI has the possibility to let plastics processors tap into the knowledge of everyone across the world who has used a particular machine.

"There is no reason that the knowledge that you have within your plant about a particular piece of equipment should be the only knowledge that you get access to," Moeller said.

"Dozens and dozens of other plants have the same equipment, and maybe hundreds of technicians have used it. Why not be able to tap into that knowledge so you know every single problem that could happen with that machine? Probably somebody has come across it before."

It's time for the plastics processors to use AI to share their machine knowledge, Moeller said.

"We believe that our industry is at a disservice by not having access to that community common knowledge," he said. "That's something we would like to build for tomorrow and would love to have conversations with people about how to do it."

Security is paramount, Moeller added, because AI will have access to company knowledge and proprietary processes. That's why the information is encrypted and compartmentalized.

"The cognition engine doesn't actually know your knowledge. That's in a knowledge vault," Moeller said. "You never have to worry about the AI remembering anything that you don't want to remember because it's stored separately."

Putting AI to work

In the plastics industry, generative text and voice are helping processors the most. For example, suppliers to the automotive market can use generative AI for failure modes and effects analysis (FMEA) reports, which can take a long time to write. "With examples, AI can write 70-80 percent of an FMEA with a human doing the final review and polish," Moeller said.

The technology also can write summaries from bodies of knowledge, in one case condensing 30,000 pages of information into two paragraphs and summarizing the information needed.

Generative AI also is good at taking informal discussion and rewriting it to be formal documents.

With speech synthesis, Moeller said, "In the fast-arriving future, we will have AI colleagues."

Both machine learning and generative AI are useful to the plastics industry, but Moeller expects to see mostly machine learning AI applications at NPE simply because it has been in the market longer.

"Generative AI is so new and changing so fast that we are very early in the cycle. The first high-cognition model was only released 13 months ago," Moeller said.

* source : https://www.plasticsnews.com/news/putting-intelligence-ai