Michael Shellenberger smiling in a blue suit at Atomic Insights event.

3 Comments

  1. I suspect Constellation is getting little out of Blue Wave AI… The solution biases in power distribution/margin and reactivity in these core simulator ‘codes’ are manifestations of modeling/method limitations made for run time (runs fast). The BWR operators are generally locked into using the fuel vendors’ licensed, design of record, core simulator to do the design and monitoring work. The dominant vendor’s core simulator uses very old diffusion methods that ‘run fast’ and have been ‘good enough’ for at least two decades. Biases could be reduced with better methods like those developed by Studsvik, but generally are accommodated by additional design margin fat [biasing low] with the ‘bad code’. So, Blue Wave seems to be making a science project analyzing/reconciling relatively ‘bad answers’ with instrumentation that until recently was considered ‘calibrated’ when within 10% of a reference power distribution reconstructed from periodic (40 day) movable detector traces using the ‘bad code’. Like you said it’s heady, technical stuff and way more down in the weeds than typical for your blog. FWIW, the fuel vendor has for decades been using an ‘AI’ written by our BWXT colleague William E. Russell (Prometheus) to balance all the different design constraints with a best solution. Still, designers find the program hard to constrain and generally just do the design work themselves. My coworkers and I earn our livings this way.

  2. The New York State new build article would have had a lot of comments… My ask there would have been why build FOAK? I’d love to see little BWRs built, but think the ABWR or AP1000 would be better choices.

  3. Revised:

    WRT Blue Wave, the solution biases in core simulator ‘codes’ are manifestations of modeling/method limitations/compromises made for run time. Most BWR fleet operators are locked into using the fuel vendors’ licensed, design of record, core simulator to do the design, monitoring, and maneuvering work. The dominant vendor’s core simulator uses old diffusion methods that ‘run fast’ and have been ‘good enough’ for at least two decades. Biases could be reduced with better methods (COTS and in development), but are accommodated by additional design margin fat [biasing low]. Reload batch size is going to be what it’s going to be for a given cycle duration +/- 4 assemblies (~1%), with or without talent or AI; room for improvement is incremental at best.

    Blue Wave seems to be analyzing/reconciling relatively ‘bad answers’ with instrumentation. The instrumentation is considered calibrated/operable when readings are within 10% of a reference power distribution, reconstructed (splined) from periodic (40 day) movable detector traces, *using* the ‘bad code’. As you wrote, it’s heady, technical stuff and way more down in the weeds than typical for your blog. FWIW, the fuel vendor has for decades been using an ‘AI’ written by a former colleague (Prometheus) to balance all the different design constraints with a best solution. Still, designers find the program hard to constrain and generally just do the design work themselves.

Comments are closed.

Similar Posts

  • Atomic Show #173 – Nuclear communicators bid goodnight to Hurricane Irene

    Hurricane Irene took a path up the east coast of the United States that put a large population at risk from heavy rains, storm surges, and high winds. Before the storm there were stories about the fact that among all of the other parts of the infrastructure that might be affected, there were a number…

  • Atomic Show #303 – Bret Kugelmass, CEO Last Energy

    Last Energy is an innovative new company governed by a philosophy of avoiding the invention of anything that has not been done before. They have created a business that is laser focused on building, owning and operating small (20 MWe), modular pressurized water reactors and selling the electricity they produce under long term power purchase…

  • Risk below 100 mSv is so low you cannot measure it

    One of my favorite jokes about the difference between scientists and engineers is the one in which a scientist and an engineer are both put into a room with a pot of gold on the other side. They are given the rules of the challenge – the gold will be given to the person who…

  • B&W mPower™ Reactor Control Room Simulator Begins Operations

    (CHARLOTTE, N.C. – December 4, 2012) – The Babcock & Wilcox Company (B&W) (NYSE: BWC) is pleased to announce that the production-standard control room prototype for its B&W mPower™ small modular reactor (SMR) is now operational. This engineering simulator is a key milestone in the B&W mPower development program. The B&W mPower control room prototype,…

  • Atomic Show #333 – Kurt Terrani, CEO Standard Nuclear

    Standard Nuclear emerged from the start-up stealth mode in early June 2025 with the announcement of successfully raising $42 million from a group of venture capitalist led by  Decisive Point with participation from Andreessen Horowitz, Washington Harbour Partners, Welara, Fundomo and Crucible Capital. Though Standard Nuclear is young enough to have a single page web site, it owns and operates the largest…

  • Atomic Show #258 – Energy and the Modern State With Professor George Gonzales

    Dr. George Gonzales is an energy-focused political scientist and professor at the University of Miami. He has written a number of books about the relationships between energy, the environment, international relations, and political decision making. His upcoming work is titled Energy and the Modern State; it is scheduled to be published in 2018. On this…