Among the many surprises at PCB East this spring was the appearance of a pair of scientists from a semi-obscure (to we laypeople) government contractor called, obliquely, JLab.

JLab is shorthand for Jefferson Lab, or its official name, the Thomas Jefferson National Accelerator Facility (TJNAF). The facility is operated on behalf of the US Department of Energy, which has a budget larger than Jabil or Flex, and oversees, among other things, the US nuclear arsenal.

Now, in the event you haven’t been paying attention, the US government has been in the media kind of often of late, for reasons too numerous for this page to detail. But one big newsworthy item has been the administration’s efforts to change the federal government’s budget priorities.

Certainly, most readers are familiar by now with the Chips Act, the overarching legislation passed in fall 2022 that authorized more than $50 billion in direct spending and tax incentives to rally North American semiconductor production. It was perhaps the most significant government-led mandate since the European Union ratified the Restriction of Hazardous Substances Directive (RoHS).

And most are also aware of a similar, albeit smaller, bill to bolster domestic printed circuit board and IC substrate production that has been proposed in Congress but has yet to make it out of committee. (The latest version, called the Protecting Circuit Boards and Substrate Act, was introduced in late May.)

Far beyond the headlines, the team representing Jefferson Lab is working on an AI solution for the manufacturing side of the PCB industry. Which is how we came to find team members Dr. Thomas Britton and Dr. Nataliia Matsiuk walking the show floor at PCB East, talking with PCB manufacturers as part of a DoE academic program that aims to get scientists out of the labs to learn what problems the industry faces.

As Britton explains on a recent podcast we did, he and Matsiuk are technologists and problem-solvers out to provide tailored solutions applicable to the industry at large. What they need in return is for manufacturers to explain what their problems are.

“A lot of value is sitting above the manufacturing process,” Britton says, referring both to inaccessible collected data and missed opportunities to collect other useful data. “Can we make something adaptive and reactive utilizing that data coming from the manufacturing line?”

“There’s a lot of data being produced,” he elaborates. “It’s very complicated, lots of steps.” The DoE program seeks to take the data from those steps, aggregate it and use it to better assess the manufacturing process holistically, he says, getting in front of potential variances as opposed to, “ ‘Oh, what, there’s a problem here?’ ”

Clearly there is (or was) government momentum to support critical industries. While the Chips Act and Boards Bill are still trying to execute on their lofty goals, the DoE has been honing its craft for years.

The subtle gem of the DoE is it is already taxpayer funded. The crucial lever, then, isn't money – but communication.

“We’re doing things to help our science. And we’re funded from taxpayer dollars through the DoE to do the work we do,” he says.

One program, for instance, was for machine vision, which sought to replace shift workers with AI solutions.

“We’re looking at deep learning and those kinds of solutions,” says Britton. “One thing that I’ve heard of that’s a roadblock for a lot of factories, at least in the US, to develop this smart factory, is to be able to communicate with the AI systems. [Manufacturers] don’t have equipment that’s capable of reporting and communicating this [data].

“What we’ve found is Asia is very well-instrumented. Around [the US], legacy machinery maybe isn’t collecting the data. They didn’t know it would be valuable. We see PCB as a beachhead to prove out the technology because you have similar challenges, especially with substrate-like manufacturing techniques that are coming up.”

As part of their commercialization strategy, the DoE is considering an open AI model in which its researchers partner with a private manufacturer, and the improvements in the line outputs make up the proof of concept.

“I think it’s going to take a lot of partnerships between the lab, the researchers and industry partners to really dig into the data that is held so close to the chest for a variety of reasons.

“You take the knowledge you gain at one company and to go to the next one. And you are transferring that knowledge – just like an employee working at one firm and then moving to a different firm – without exposing the secrets of any party involved to enhance the manufacturing across the line,” Britton says.

“I think, right now, our top hypothesis is that you would do it through embedding with viable partners willing to give you the data, work with you, devote the time and then hopefully it would be made commercially available.”

Fabricators have often privately groused that the US government can be a roadblock to success. Will they take the opportunity to clear that path now that they know it’s available?

Mike Buetow is president of PCEA (pcea.net); mike@pcea.net.