Materials Design: Computational Alloy Design by Combinatorial Optimization

An overview is made of methods allowing to design new alloys with optimal characteristics, within very large compositional domains. Multi-objective optimization by stochastic metaheuristics like genetic algorithms can be used, while relying on fast predictive models such as the CALculation of PHAse Diagrams (CALPHAD), and on tools from the domain of artificial intelligence (data mining/machine learning) such as neural networks or Gaussian processes, performing flexible regressions of properties as a function of composition. The performance of the method is illustrated by examples of alloys designed using such techniques.