A dual surrogate memetic framework for single/multi-objective evolutionary optimization of computatoinally expensive problems

摘要

A method for the optimization of real-world objects in terms of physical parameters comprises the steps:- Creating an offspring population (P_o) by applying an evolutionary operator to a parental population (P_c), wherein the populations (P₀, P₁) comprise individuals representing a real-world object to be optimized in terms of its physical parameters;- Performing local refinements on at least one individual of the offspring population (P_o);- Selecting a new parental population (P_C) from a selection pool (P_S) by using a fitness function, the fitness function using physical parameters for assessing the fitness;- Repeating the above steps until a predetermined termination criterion is met; and- Outputting the optimized individual as a representation of an optimized real-world object.;The method may be applied in all forms of design or search problems that contain computationally expensive functions, simulator or analysis code, such as drug design, material design, rainfall prediction, aerospace design, aircraft design, aerodynamic design, structural design, electromagnetic design, physics-based modelling and etc.