U of I team’s AI-driven solution hits the sweet spot

Hunter Hawkins ’23 traveled a long and winding academic road before he found success as a computer science doctoral student at University of Idaho Coeur d’Alene.

It took him far less time to help one of the largest manufacturers in Idaho keep their employees out of harm’s way while maintaining production.

Hawkins and John Shovic, director of the Center for Intelligent Industrial Robotics (CIIR), have been working with the operations team at Amalgamated Sugar in Nampa for over a year on a system to prevent plugging in the sugar beet processing company’s steam dryer. Using two different AI algorithms, Hawkins and Shovic created a way for Amalgamated to monitor the dryer’s operation and shut it down before a problem arises.

When the dryer becomes plugged, Amalgamated employees need to climb into it — and its dangerous combination of heat, pressure and moisture — to clean it out and get it back online.

“Working with Hunter and the U of I team over the last 12 months has been great,” said Scott Hyer ’10, operations manager at Amalgamated Sugar. “They’ve created an AI-generated control system that will monitor our process and let us know if there’s a problem coming up before it happens.” 

All about timing

Hawkins’ initial interest in computers came from playing video games on an older computer he got from his dad. He and his friends played Space Pinball around the clock, testing the limitations of the software to see if they could break it.

“That’s where my curiosity about computers came from,” Hawkins said. "It sparked that flame.”

But despite his interest in computers at a relatively young age, he wasn’t as interested in other subjects in school. He graduated high school in Coeur d’Alene with a subpar GPA and was not accepted by U of I when he first applied.

Undeterred, he enrolled in North Idaho College and improved his grades. It was also at NIC where he took his first computer science class.  

Now, in his doctoral program, Hawkins specializes in industrial automation. He is part of a Coeur d’Alene-area company, Wapiti Consulting, that focuses on creating AI-based solutions for rural water and wastewater districts. So when the possibility of working with Amalgamated came up, he jumped at the chance.

“They were having issues with their steam dryer getting plugged,” he said. "They had some educated guesses as to why it was plugging but they also realized there were different reasons it was plugging. Because their steam dryer has so many data points, we were able to use AI to monitor different areas of the dryer — like the temperature, steam pressure and flow rate of the pulp into the dryer — to find potential problems before it became plugged.”

Dirty, dull and dangerous

Having worked on AI-driven solutions for a long time, Hawkins knew he had some choices to make about what kind of algorithm to use on the Amalgamated project. He knew from experience that a long short-term memory (LSTM) model could gather all the data from the dryer’s monitors so they could predict when a problem might occur.

But he also wanted the information to be displayed in a manner where the Amalgamated employees watching the production line could monitor the system and know when, and where, they needed to act.

What Hawkins suggested to Shovic was something neither had heard of — a hybrid system that combines the data prediction capabilities of an LSTM with a predictive model that includes a monitoring system. The monitoring system would display a green light when everything is fine or a red light when a problem is detected.

“I thought ‘wow, that’s incredible and absolutely innovative’,” Shovic said. “It’s a great idea and will be applicable to other manufacturers.”

By reading the dryer’s data points and using AI to analyze the information every five seconds, the model developed by Hawkins will alert Amalgamated to a potential issue. When the monitor’s warning system flashes red, the Amalgamated team will have around 15 minutes to shut down the dryer and find where the potential problem is.  

As Hawkins and Shovic spend time with the Amalgamated team preparing for this system to go live, Hyer said the company is interested in testing out additional AI-driven solutions that will help them, especially in the areas of employee safety and production.

“We’ve already identified multiple areas where we think application of AI modeling can be very helpful,” he said. “It’s really important to us, as a factory, knowing that AI is always watching and will help us be successful.”

For Hawkins and Shovic, success means showing Idaho companies how CIIR can help them automate what Shovic calls “dirty, dull and dangerous jobs.”

“It really feels good to help Amalgamated Sugar — this is going to help a large Idaho company avoid shutdowns and avoid employees getting hurt,” he said. “This also shows our strength of helping in the manufacturing area. They had a problem that was perfect for AI, and we had the right technology and the knowledge to create a solution.”