Sampling Methods Settings

For design of experiment, there are general options of all methods being setup. For this method, some additional options are required further:

• Parameter

  This option is normally set to "Latin Hypercube". There are several sampling techniques as "Latin Hypercube", "Monte Carlo" and "Sobol"

   

• Sample Size

  It is the number of model calculations being sampled.

   

• Random Generator

  It represents the way of the random number generator for sampling. It is normally set to "Init". It is possible to select a other options:

• Init

  The random number generator will always initialized for each optimization step (a tolerance simulation run). The seed of generator is set to 1. This option causes the same tolerance analysis results of the same Tolerance parameters. It is the reproduction of tolerance analysis. With this option, tolerance optimization will convergence, because the results of tolerance analysis will be approximated.

   

• No Init

  The random number generator will be initialized once at the start of OptiY. the seed of generator for the tolerance analysis depends on the last tolerance simulation run. Therefore, there is no reproduction of a tolerance analysis results for the same tolerance parameters.

   

• Init Time Dependent

  The random number generator will always be initialized for each optimization step. The seed of generator is set to the time constant of the computer. With this option, the results of tolerance analysis will differ from simulation to simulation, because the tolerances are generated with really random numbers.

• Additional Data

  User can import additional data as measuremetn data or old simulation data to design of experiment data. This data will be used as additional data for building the meta-model. It is useful for adaptive Gaussian process. User can import measurement data for building the first meta-model, then uses adaptive sampling to generate more data from simulation to build the better meta-model from both data.