This option is normally set to Standard. It uses internally the standard parameters of the evolutionary algorithms: (1+7) evolution strategy with standard parameters. Set the option to Manual to edit these parameters.
Number of Parents
A parent represents a set of the optimization variables as nominal and stochastic design parameters. The number of parents is the number of the sets of design parameters at each optimization step a a generation. Parents are selected from the best children of the last generation. They will be used to generate the children in the next generation. For a smooth objective function, a parent is good enough. For a discontinuous and rough objective function, a great number of parents should be used.
Number of Children
A child represents a set of the optimization variables as nominal and stochastic design parameters. The number of children is the population size in a generation as optimization step. It characterizes the number of simulation runs in an optimization step. For each set of optimization variables, a simulation run must be executed. the number of children must be greater than the number of parents, because parents are selected from children. The relative of parents and children realizes the pressure of selection. If the pressure is low, the optimization will convergence slowly. If the pressure is high, the optimization will convergence high, but the optimization could not be able to leave a local optima. The relative of parents and children should be selected depending on the objective function.
Evolutionary Parameters
With the Standard option, all standard evolutionary parameters are used. Set the option to Expert to edit these options.
Number of Pareto Frontier
A multi-objective optimization uses not only the parents, also the other best solutions of the last generations to generate children for the next generation. The number of Pareto frontier is therefore the number of the used solutions generating children. The option is normally set to 20. For a great number of optimization criteria, it should be increased.
Kind of Recombination
It characterizes how a child can be generated from 2 parents. There are several kinds of recombination: Discrete, Mean Value, Normal Distributed, Uniform Distributed and Rampen Distributed. The option is normally set to Rampen Distributed.
The recombination improves the optimization process. It is normally used if there are at least 2 parents in a generation.
Kind of Selection
A kind of selection (Komma, Plus) causes, whether the parents survive on the next generation or not. If the option is set to Komma, parents die on the next generation. They are used only in a generation, children must be new generated. If the option is set to Plus, parents may survive on the next generation. If parents have better fitness then children, they will be adopted in the next generation.
Kind of Step Control
It characterizes the adaptation of the optimization step on the local topology of criteria properties and the improvement of the optimization process. Set the option to Multi Step to do it and to Single Step to no do it.
Select Step
It causes a accumulation of the optimization step from all parents steps which will be adopted in the next generation. with Best Value, the next step is the step of the best parent. Mean Value is the mean step of steps of all parents.
Random Generator
The option is similar to sampling methods settings.
User-Defined Parameter
If this option is selected, user can define and import the start parameter for the parent population.
Data
The option is visible if the "User Defined Parameter" is selected. Here, user can define and import the start parameter for the parent population.
Hypervolume computing
The option turns on or off the computing of hypervolume for multi-objective optimizazion.