Optimised recrystallisation model using multiobjective evolutionary and genetic algorithms and k-optimality approach

Abstract

The meta-models are constructed for static recrystallisation of dual phase steels using evolutionary neural nets (EvoNN). Four mutually conflicting objectives—(i) overall kinetics, (ii) grain size, (iii) the amount of strain and (iv) the precipitate volume fraction—are optimised simultaneously using an emerging k-optimal approach incorporated in the EvoNN, using a predator–prey genetic algorithm. The first objective involved minimisation of error with respect to experimental observation. The grain size and the amount of strain were minimised, whereas the precipitate volume fraction was maximised. The aim is to control the recrystallisation process in order to achieve desired material properties of dual phase steel during the final stages of heat treatment.

Keywords:

• Multiobjective optimisation
• Recrystallisation
• k-optimality
• Genetic and evolutionary algorithms