Physics Colloquium: Design of Alloys Resistant to Molten Salt Corrosion via Machine Learning and Optimization Algorithms

Physics Colloquium
Design Of Alloys Resistant To Molten Salt Corrosion Via Machine Learning And Optimization Algorithms

Dr. Rafael Herschberg
Postdoctoral Researcher, RIT BS Physics Alum, 2012, Institut des Máteriaux de Nantes

Abstract: 
Molten Salt Reactors (MSR) are a type of Generation IV nuclear fission reactors in which the primary coolant is a molten salt mixture. However, the extreme conditions in which their structural parts are being subjected makes it imperative to carefully choose alloys with outstanding characteristics, such as a stable microstructure and an excellent resistance to corrosion and to irradiation. In addition, molten salt corrosion is a highly sensitive phenomenon to its environment due to the influence of several factors – the temperature, the salt composition and the time of exposure, among others. In other words, it remains a challenge to compare the performance of different alloys in absolute terms.
The present strategy tackles this obstacle and accelerates the discovery of promising alloys by means of artificial intelligence approaches. Firstly, a database is built upon an extensively literature survey where groups of candidates have been previously tested at identical experimental conditions. Secondly, their overall performance is ranked with a pairwise comparison algorithm, where each alloy constitutes a node in a directed network. Then, the assigned score is fitted as a function of chemical content by a Gaussian process regression. Lastly, a multi-objective optimization algorithm is applied to obtain the best compromise between molten salt corrosion resistance and microstructural constitution (evaluated by computational thermodynamics).
  
Speaker Bio: 
Rafael Herschberg received his BS degree in Physics with a dual minor in Mathematics and Economics in 2012 from Rochester Institute of Technology (RIT). He later gained his Masters in 2015 in Materials Science and Technology from the Universidad Nacional de San Martin (UNSAM) under the supervision of Dr. Gastien and Dr. Zelaya where he experimentally studied the effects of isothermal treatments on the super-elastic properties of Cu-Al-Ni monocrystalline alloys. He then he earned his PhD degree in 2018 from the Université Paris – Saclay where he modeled the diffusion of interstitial atoms in bcc alloys using Atomistic Kinetic Monte Carlo simulations under the supervision of Dr. Soisson and Dr. Nastar. 
After obtaining his PhD degree he worked as a postdoctoral researcher for 1 year at the Université de Rouen – Normandie with Dr. Zapolsky where he modeled the Cottrell atmosphere of C atoms surrounding edge dislocations in pure bcc Fe using phase field crystal methods. At the present time, Rafael Herschberg is working at the Institut des Máteriaux de Nantes under the supervision of Dr. Tancret where he is designing alloys resistant to molten salt corrosion with artificial intelligence algorithms. 

Intended Audience: 
Students with interest in the topic

To request an interpreter, please visit myaccess.rit.edu