Editor’s note: Welcome to So That Happened, our editors’ takes on things going on in the manufacturing world that deserve some extra attention. This will appear regularly in the Member’s Only section of the site.
Humans Limitations Teach Vision Systems to Think Less and Embrace Lower Precision
We’ve heard it for generations – computers are better at bulk data processing than people, computers can interpret vast amounts of data quickly and digital cameras are incredibly precise compared to our flawed jelly-filled eyeballs.
To which researchers at Purdue University in Indiana say, “Yeah, so what!?”
Jianguo Mei, the Richard and Judith Wien Professor of Chemistry in Purdue’s College of Science, says our silly, fallible human eyes and our ability to ignore useless data make humans better at reacting to things going on in the world around us, like moving objects. And we can use our frailties to improve computer vision systems.
“Our eye and brain aren’t as high resolution as silicon computing, but the way we process the data makes our eye better than most of the imaging systems we have right now when it comes to dealing with data,” Mei said. “Computer vision systems deal with a humongous amount of data because the digital camera doesn’t differentiate between what is static and what is dynamic, it just captures everything.”
Camera-based quality systems have become extremely common on manufacturing plant floors, capturing terabytes of data to pinpoint minor flaws in finished goods. That flood of data and the expense of such systems can be a barrier to adoption. Purdue’s more-human computer eyes could reduce that load, potentially lowering the amount of data needed to make such systems work, which would cut costs for data processing and data storage.
Instead of immediately converting light into a digital signal, as most digital cameras do, Mei’s devices convert the light into a flow of ions, similar to how human retinal cells transmit light input to our brains. Those ions flow through an electrolytic gel, and repeated exposure to light increases the charge imbalance in the gel, a feature that can differentiate between the consistent light of a static scene and the dynamic light of a changing scene.
At Purdue, Mei and graduate student Ke Chen are joined in the research by Hang Hu, Inho Song, Won-June Lee and Ashkan Abtahi as well as researchers at the University of Texas at San Antonio. "Artificial Retina Based on Photon-Modulated Electrochemical Doping” was published in Nature Photonics with the support of Ambilight Inc.
—Robert Schoenberger
Repurpose, Repair, Recycle
The DOE has launched the Re-X Before Recycling Prize, a sustainability prize that will award a total of $4.5 million in cash prizes and $1.1 million in national laboratory analysis consultation and technical assistance. This three-phase prize promotes clean energy and circular supply chains through identifying and developing ways to reuse, repair, refurbish, remanufacture or repurpose products or parts before recycling to extend their lifetime. The three phases are Identify, Prepare and Develop.
“While recycling is an important part of the circular economy, finding ways to extend the usage of a product or part can often use less energy than recycling. That’s what we’re hoping to explore with this prize: how to keep products in use before they need to be recycled, and how to do so in a way that will engage communities,” said Jeff Marootian, principal deputy assistant secretary for Energy Efficiency and Renewable Energy.
In phase one, applicants will identify and create a plan to add an innovation to a new or expanded Re-X supply chain. Phase one winners will receive a cash prize of $50,000 and up to $5,000 in consulting support. Applications for phase one are due on March 12, 2024. Learn more about the prize here.
—Anna Smith
Product Recalls by the Numbers
Earlier this week, Stellantis issued a recall impacting more than 32,000 vehicles in the U.S. market. Cantaloupes recently were recalled due to a Salmonella outbreak linked to these popular melons. Warnings of defective foods, consumer products, automobiles and other items seem to march steadily across the newswire. How do those numbers add up?
The good news: In the third quarter of 2023, U.S. product recall activity recorded its largest quarterly decline in almost three years, dropping to 740 product recalls from 856 recorded in the second quarter. That's a 13.6% decline. Moreover, the number of impacted units declined as well, and much more dramatically, dropping 61.9% to 89.15 million units.
These numbers come courtesy of Sedgwick, a provider of technology-driven risk and benefit solutions. Its quarterly U.S. Recall Index report tracks recall data across six industries: automotive, consumer product, food and drink, medical devices, and pharmaceuticals. The company analyzes data from the U.S. Consumer Product Safety Commission, the U.S. National Highway Traffic Safety Administration and other agencies to compile its reports. It released the Q3 index this month.
No obvious factors explain the quarterly decrease, the report said, "especially since there is evidence of increased monitoring and enforcement actions among the regulatory agencies."
The latest data provide more good news, comparatively speaking. The year-to-date number of recalled units totals 528.7 million, making it unlikely that 2023 will be the third consecutive year with more than 1 billion units recalled. Unlikely, but not impossible, the company notes. The last three years each have had a quarter with more than 400 million units recalled. Let’s hope Q4 avoids that fate.
All that said, 2023 likely will end on at least one low note, despite the improved Q3 numbers, Sedgwick points out. The year is on track to hit a five-year high for total recall events.
—Jill Jusko
Everybody into the U.S. Steel Pool
Nearly four months after U.S. Steel directors hung up a ‘For Sale’ sign, who is still in the picture as a possible buyer? The better question might be to ask who’s not.
CNBC’s David Faber reported Nov. 28 that final bids for Pittsburgh-based U.S. Steel are due this week and that the company’s goal is to finalize a deal by year’s end. And while Cleveland-Cliffs head man Lourenco Goncalves—who this summer catalyzed the sale process by offering to pay $7.3 billion for U.S. Steel—early on tried to clear the field of competitors by gaining the full-throated endorsement of the United Steelworkers union, Faber and other recent reports say he has plenty of company in the bidding.
Reported to be joining Cliffs in the auction are ArcelorMittal (which just three years ago sold its U.S. operations to Goncalves) as well as Nucor, Canada’s Stelco, Steel Dynamics and Argentina’s Techint, which owns NYSE-listed Ternium. Various reports have speculated that only Cliffs and ArcelorMittal executives are prepared to buy all of U.S. Steel; others in the mix are supposedly likely to partner up and divide the company’s assets.
This high level of interest is a welcome turn for U.S. Steel shareholders, whose investments were trailing the S&P 500 benchmark by more than 20 percentage points year to date before Cliffs went public with its bid in early August. Since then, U.S. Steel shares have climbed more than 50% and the auction appears to be promising a cherry on top of that cake.
—Geert De Lombaerde
