Starting version 3.0, LS-OPT supports the definition of discrete variables. Discrete type of design parameters are helpful when there is a need to find the optimum values from a “set” of values. A good example would be the choice of element formulation. The possible values for LS-DYNA for a quad can only be a subset of formulations listed under ELFORM in SECTION_SHELL. Any value chosen by LS-OPT that does not uniquely match the subset will result in non-physical or error terminations by the solver such as LS-DYNA.
To enable discrete variables, the user has to mark the variable as ‘Discrete’ and provide a sequence of values separated by either a comma or a space. Discrete sampling of this definition is then performed by LS-OPT in two ways as defined below.
1. Continuous Discrete Sampling.
This sampling approach uses ALL possible values between the lower and the upper bound of the discrete set. For example if the discrete set consists of “2,10,16”, then any value that lies within 2 and 16 is considered as a valid sampling point.
2. Discrete Sampling
In this sampling, LS-OPT will only use values from the set with no interpolation. For the set 2,10,16, only values from the three will be used. Currently, this approach works for D-OPT and SPACE-FILLING sampling schemes only.
Thanks to Dr. Nielen Stander for sharing this information.