An artificial intelligence assisted engineering analysis at the Savannah River Site in Aiken, South Carolina, traced a persistent throughput shortfall to a part measured in fractions of an inch, and the correction pushed the facility past its original design specification.
The Salt Waste Processing Facility has operated since 2021, treating salt waste drawn from the site radioactive liquid waste tanks. It strips cesium and other radionuclides before the remaining material moves to the Defense Waste Processing Facility for vitrification. The plant was designed for 21.6 gallons per minute and had never sustained that rate.
The analysis identified the weir, a circular component inside the centrifugal contactors that governs liquid separation and outflow, as misaligned because its diameter ran slightly small. Engineers increased the weir diameter by one tenth of an inch. After the changeout, throughput reached 23 gallons per minute, which the American Nuclear Society reported as 106 percent of the original design specification.
"Prior to the weir adjustment, SWPF processed 17 million gallons of waste feed in almost six years of operation. We have already processed more than 3 million gallons in only five months after the changeout, showing how a small mechanical change can make a big difference," said SWPF Senior Director Ryan Lentsch.
Thomas Burns Jr., president and program manager at Savannah River Mission Completion, the Department of Energy contractor running the work, said that without artificial intelligence the change is something that would have taken five years to even discover, and it was put into place in the facility in mere months.
Source: ANS Nuclear Newswire - https://www.ans.org/news/2026-09-15/article-8389/ai-leads-to-improved-efficiency-at-srss-salt-waste-processing-facility/