Eigen Innovations’ AI-driven systems optimize quality control for automotive, paper manufacturers.

Since the advent of manufacturing automation, machine operators and plant managers have dreamed of pinpointing the optimal settings that consistently produced quality end products at the fastest rate.

Although better ways of examining end product quality have come along over the years, it’s been a long and frustrating shot in the dark, for the most part.

That’s partly due to the increasing complexity of manufacturing processes. Additional factors aff ecting product end quality have come into play; subtle differences in raw materials, more sophisticated levels of automation and even ambient plant temperature can all throw quality off optimum. But it’s mostly because there’s been no way to collectively analyze all the factors that aff ect quality and then tie them meaningfully back to machine settings.

According to Scott Everett, CEO of Eigen Innovations, those days of shooting in the dark are over. His New Brunswick-based automation fi rm specializes in designing AI-driven quality control systems for two key sectors: Paper manufacturing and injection molding for Tier 1 automotive parts manufacturers.

Eigen also serves customers in plastics joining and welding, die casting, sheet metal production and adhesive dispensing.

While all very diff erent sectors, each of those customers’ manufacturing processes can be optimized through a combination of high-resolution imaging, thermographic data and AI algorithms to reduce waste, increase cycle speeds and magnify efficiencies.

“We adapt our platform, working closely with each customer, so that it is confi gured to enable continuous improvement,” Everett explains. “Th e fi rst step is to identify all the data that aff ects end quality. We then put Eigen hardware in place for continuous realtime data collection and ongoing analysis of the manufacturing process, which always uncovers brand new insights.

“Our AI provides machine operators with setting recommendations to achieve process optimization, and also detects quality issues in real-time, allowing for immediate fixes,” he adds. “Over time, the AI begins to predict outcomes before they happen.”

The Road to Here
Started in 2012 as a spinoff from the University of New Brunswick’s (UNB) Intelligent Controls and Advanced Manufacturing Lab, Eigen Innovations was founded by Dr. Rickey Dubay (still a UNB professor) and CEO Scott Everett, who was Dubay’s graduate student.

Today, its customers include four of the top 10 Tier 1 automotive parts suppliers and one of the world’s largest producers of specialty paper products. The firm has about 25 employees and has secured more than $8 million in funding from angel investors and provincial agencies. The company has also won several awards, including the Cisco Innovation Grand Challenge, Dell ‘Connect What Matters’ and a Kira18.

However, when Everett and Dubay started presenting their solution to manufacturers eight years ago, artificial intelligence was a hard sell.

“AI is becoming common now, but no one had heard of it in 2012,” Everett remembers. “We had to explain how all this data, generated from end product quality analysis, generated by a given machine, could be analyzed by our software to create machine setting recommendations so that target quality levels are reached all of the time. To many early customers, it seemed fantastical.”

Of course, achieving those results depends on capturing good raw data. First, AI need to discern the ins and outs of a particular manufacturing process. Only then can it then analyze the captured data and turn it into information operators can act upon.

Early on, Everett says getting good raw data to the AI was the primary challenge. The team first had to find the right camera combination to capture ultra-resolution images as products came off the line.

As it turned out, this real-time analysis was key to the process. Typically, the quality of injection-molded parts is analyzed hours after parts come off the machine, when the material has already set. In die casting and sheet metal manufacture as well, defects are typically only visible later. Eigen’s high-res imaging, however, makes it possible to analyze quality differences right away.

“What makes it complex too is that each setting usually affects the others,” Everette explains. “For example, injection molding operators can change the rates at which the plastic is heated and squeezed. That means our algorithms need to be able to tease out what mix of settings is best – taking into account different batches of materials, the moisture content of raw material, the ambient plant temperature – to try to hit the ideal speeds at various parts of the process.”

Detecting quality issues is also hard in many manufacturing situations, Everett explains, because a finished part often goes through multiple steps before it’s integrated into a final product. Quality problems sometimes only emerge when these parts are being installed. And because parts don’t usually have a barcode or any other way to trace them, it’s difficult to know what manufacturing parameters were out of whack.

While Everett reports that parts traceability is evolving, he says that, “when our system is in place, there are no more quality issues to detect later on because we’ve identified them at the machine level and fixed them at that point.”

He adds that once quality issues have been tackled, manufacturers can then increase cycle time as much as it can be increased without crossing the line to where quality drops.

Looking Forward
Today, Eigen’s challenges are more like opportunities to take advantage of technological advancements, allowing their customers, in turn, to realize bigger benefits and faster ROI.

“AI computing power continues to increase and imaging also, as more computing power allows for faster and better analysis of the high-res video,” Everett explains. “The ‘edge computing landscape’ has also evolved so that very deep AI platforms can be run at any plant.

“We continually improve the AI’s programming so that it refines its definition of target quality and what process parameters produce that quality,” he adds. “Again, that allows us to push cycle time and make the cost return shorter.”

Everett says costs for customers are also kept to a minimum through Eigen’s standardized solutions. “Each plant is different but we’ve focused on the deep vertical applications that are very scalable across the industry,” he says, “and that helps drive adoption.”

In 2020, it’s now common for companies to ‘orient themselves to data,’ he says, hiring leaders who understand the power of leveraging real-time feedback.

“It’s a big shift and we’ve had to figure out how to implement our system so that it scales with it’s a core technology,” he says. “But the use of high-resolution imagery is growing across many sectors and the number of true AI applications is going up. It’s creating a critical mass.”
https://eigen.io