Fluoride-salt-cooled high-temperature reactor generative design optimization with evolutionary algorithms

Chee, Gwendolyn Jin Yi

Permalink

https://hdl.handle.net/2142/117715 Copy

Description

Title

Fluoride-salt-cooled high-temperature reactor generative design optimization with evolutionary algorithms

Author(s)

Chee, Gwendolyn Jin Yi

Issue Date

2022-09-29

Director of Research (if dissertation) or Advisor (if thesis)

Munk, Madicken

Doctoral Committee Chair(s)

Munk, Madicken

Committee Member(s)

• Kozlowski, Tomasz
• Stubbins, James F
• Tran, Huy Trong

Department of Study

Nuclear, Plasma, & Rad Engr

Discipline

Nuclear, Plasma, Radiolgc Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Fluoride-Salt-Cooled High-Temperature Reactor (FHR)
• Advanced High-Temperature Reactor (AHTR)
• Generative Design
• Generative Reactor Design
• Evolutionary Algorithm Optimization
• Additive Manufacturing
• OpenMC Monte Carlo Code
• Moltres Temperature Modeling
• Reactor evOLutionary aLgorithm Optimizer (ROLLO)

Abstract

Additive manufacturing of reactor core components removes the geometric constraints required by conventional manufacturing, such as slabs as fuel planks and cylinders as fuel rods, enabling further optimization and improvement of core geometries. Wide-spread adoption of additive manufacturing methods in the nuclear industry could drastically decrease reactor fabrication costs, reduce deployment timelines, and improve reactor safety. Due to the expansion of the potential design space facilitated through additive manufacturing, reactor designers need to find methods, such as generative design, to explore the design space efficiently. Generative design is an iterative design exploration process; designers define design goals and constraints in a generative design software and the software explores all the possible permutations of a solution, quickly generating design alternatives. Fully benefiting from the new ability to 3D print reactor components requires further research into generative reactor design optimization. Generative reactor design optimization for arbitrary geometries enabled by additive manufacturing is a new concept, and few research demonstrations have been done to explore the large new design space. In this dissertation, I apply evolutionary algorithms to conduct generative Advanced High-Temperature Reactor (AHTR) design optimization. First, I participated in the Organisation for Economic Co-operation and Development (OECD) Nuclear Energy Agency (NEA) Fluoride-Salt-Cooled High Temperature Reactor (FHR) benchmark to further our understanding of the AHTR design's complexities and to gain an intuition for the unique physics of the system. Next, I created the Reactor evOLutionary aLgorithm Optimizer (ROLLO) Python package tool that enables generative reactor design optimization with evolutionary algorithms for non-conventional reactor geometries and fuel distributions. I then applied ROLLO to conduct generative reactor design optimization for AHTR plank and one-third assembly models. ROLLO generated AHTR designs with varying fuel amounts, fuel distributions, and coolant channel shapes that optimize for three key reactor performance metrics: minimize the total fuel amount (PFtotal), minimize maximum temperature (Tmax), and minimize power peaking in the fuel (PPFfuel). I reported the FHR benchmark Phase I-A and I-B results, demonstrating the AHTR's passive safety behavior with negative temperature coefficients. A comparison of keff results between the reference case and the AHTR configuration with high heavy metal loading demonstrated that increased fuel packing does not always correspond with increased keff due to self-shielding effects. In addition, temperature modeling of the AHTR full assembly showed temperature peaks in the fuel stripes near the spacers, highlighting the impact of spacer material and location. Next, I reported and discussed the AHTR plank and one-third assembly optimization results. I characterized each objective's driving factors and relationship with each input parameter from the results. The final and largest optimization problem is the one-third assembly multi-objective optimization that optimized for all three objectives (PFtotal, Tmax, and PPFfuel) while varying all the input parameters (PFtotal, TRISO distribution, and coolant channel shape), also known as simulation a-3b. Simulation a-3b ran for 1800 node-hours on the Theta supercomputer with 6 generations and 128 reactor models per generation. It demonstrated 12 one-third assembly reactor models on its Pareto front that met all three objectives. The reactor models on the Pareto Front have different PFtotal, TRISO distributions, and coolant channel shapes, depending on the extent each objective is minimized due to the nature of multi-objective optimization that results in a tradeoff between objectives. The results demonstrated ROLLO's success in conducting multiobjective generative reactor design optimization, and underscored the challenges of multi-objective design optimization. In conclusion, through participation in the FHR benchmark, I contribute to deepening our understanding of the promising AHTR technology. By designing the ROLLO tool and demonstrating ROLLO's success in optimization of the AHTR beyond classical input parameters, I contribute to optimization tool development for reactors of the future. As additive manufacturing technology advances and the Transformational Challenge Reactor (TCR) program demonstrates the first 3D printed operational reactor, more reactor designers will begin to explore the vast design space enabled by 3D printing. ROLLO can be utilized to optimize other reactor types for arbitrary geometries and parameters, enabling further optimization and improvement of reactor geometries.

Graduation Semester

2022-12

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Gwendolyn Jin Yi Chee

Owning Collections