BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Optimization of a Nuclear Reactor
DTSTART:20260521T123000Z
DTEND:20260521T133000Z
DTSTAMP:20260525T141700Z
UID:indico-event-8018@indico.ph.tum.de
DESCRIPTION:Speakers: Paul Weimer (TUM)\n\nThis talk presents a framework 
 that couples a custom GPU-accelerated Monte Carlo neutronics simulation to
 ol (GAINS) with a large language model (LLM) -based agentic optimizer (NOV
 A) to automate and accelerate the reactor design process. The simulation s
 ide achieves a 5–23x speedup over conventional Serpent 2 calculations wh
 ile maintaining geometric and physical consistency. On the optimization si
 de\, the LLM agent is equipped with a structured toolchain covering geomet
 ry generation\, case management\, simulation execution\, and results parsi
 ng\, while being deliberately constrained to mitigate hallucination and ov
 erconfidence.\nAs a first demonstration of the framework's capability\, th
 e system was able to independently construct a complex RBMK reactor model 
 from scratch. Building on this\, the talk then moves to the FRM II HEU-to-
 LEU fuel conversion challenge\, a design space involving parameters such a
 s number of fuel plates\, involute radius\, plate thickness\, and fuel len
 gth\, subject to criticality\, cycle length\, flux\, and safety constraint
 s. This talk will show the current state and capabilities but also highlig
 ht how LLMs like to “cheat the system” and pass of wrong or incomplete
  data.\n\nhttps://indico.ph.tum.de/event/8018/
LOCATION:E18/ENE Seminar Room 3268 (TUM PH)
URL:https://indico.ph.tum.de/event/8018/
END:VEVENT
END:VCALENDAR