SUBJECT: Ph.D. Proposal Presentation
BY: John Stooksbury
TIME: Tuesday, June 23, 2020, 1:00 p.m.
TITLE: An Evolutionary Approach to Radiation Shielding Design and Optimization
COMMITTEE: Dr. Nolan Hertel, Chair (NRE)
Dr. Anna Erickson (NRE)
Dr. Steven Biegalski (NRE)
Dr. Stephen Croft (ORNL)
Dr. Walter Mcneil (K-State)


The time and cost of the development of radiation shielding systems can be greatly reduced by an automated tool which designs these systems by itself with minimal human input or interaction. Such an automated tool can be created by using a Genetic Algorithm (GA) coupled with a Response Matrix Method (RMM) transport solver. Radiation shielding problems typically have simple geometries concerning the shielding itself. Here, a 1-dimensional slab geometry RMM implementation is used to rapidly evaluate properties of radiation transmitted from some source through a mass of shielding material(s).

Like many engineering problems before it, radiation shielding design problems can be formulated for optimization by a GA. A fitness function is defined to quantify how well a shield satisfies some desired criteria. Then the properties of potential shields are evolved over hundreds of generations in order to arrive at an optimized arrangement of materials of various thicknesses. The nonlinearity of radiation transport and large solution space of possible shielding designs are well-suited to be explored using a GA. Although the principal application considered is shielding design, these methods and implementations can be used to solve any fixed-source problem with a sufficiently simple geometry.