 |
|
This scheduler automatically samples the design space where the interpolation is less accurate.
In many circumstances the designer is initially more interested in the exploration of the design space more than in the search for the optimum that is typically obtained after a fairly long decision process. This activity is usually done using DOE techniques. DOE algorithms are of two types: the one that makes no assumption on the behavior of the design space and the one that assumes polynomial behavior of the design space. Random and Sobol sequences belong to the first class while Reduced Factorial, Box-Benken, Taguchi, etc. belongs to the latter class. None of these algorithms have an iterative behavior while it would be desirable to sample the design space in order to maximize the extraction of information from the engineering model.
This scheduler fills this need:
Starting from an initial set of points each new experiment is placed in the design space region where the interpolation error of a Kriging model is larger for each of the responses being analyzed.
The user must specify:
Number of Generated Points: is number of designs to add to the design space.
Cross-Validation Using: points out which specific points are difficult to estimate from the observed data. It represents the type of error to consider during running. It is possible to choose between Absolute Error, Relative Error and Variance. View an MACK Example to understand the difference between these three cross-validation types.
K-nearest Designs: is the number of neighbor points to consider. The maximum number of neighbors is 100.
Random Generator Seed: an integer between 0 and 999 for repeatability. If two MACK Sequences are created with the same seed, they will generate and return identical sequences of numbers. If the seed value is 0, the sequence is automatically seeded to a value based on the current time.
Notes:
The number of concurrent designs evaluation can be set in the Run Project Dialog. The number of concurrent designs evaluation is limited by the number of output variables (e.g. if the project has 4 output variables, the maximum number of concurrent evaluations will be 4 even if there is a greater selection in the Run Project Dialog).
The first N number of individuals entries in the DOE table are used as the initial design space distribution.
The number of individuals in the DOE table must be equal or greater than the value edit in K-nearest designs textfield.
Have a look at an MACK Example to see how it works.
