LS-OPT 6.0

The following features have been added:

1. Classifiers have been introduced to provide a new constraint handling approach for optimization or probabilistic analysis. The basic idea is to construct a decision boundary in the design space that predicts whether a given design is feasible or not. The support vector classification algorithm, a pattern recognition and machine learning method, is now available to approximate the boundaries. It focuses only on the decision boundary based on the pre-defined feasibility criteria and not on the prediction of response values. This approach is especially attractive for discontinuous or binary responses, and in the case of multi-disciplinary failure criteria.
2. Similarity measures. The Discrete Fréchet distance and Dynamic Time Warping (DTW) methods are introduced for computing the distance between two curves represented by respective point sets. DTW is highly effective for addressing noisy histories.
3. Multi-point histories have been introduced as a means to accommodate the full-field response of shell structures. The multi-point histories are defined at a user-specified coordinate set. Element or nodal values can be extracted from the LS-DYNA d3plot database. These can be used for parameter identification using Digital Image Correlation. A truncation function is available for filtering the multi-point histories.
4. Similarity measures for curve comparison. Based on multi-point histories, general expressions including Mean Squared Error, Discrete Fréchet and Dynamic Time Warping similarity measures as well as cross-plotting functions have been added. Multi-point cross-plots can be constructed to incorporate both histories and multi-point histories. The similarity measures can be used to compute the distance between any two curves or multi-point curves. All expressions apply to curves and multi-point curves imported from files as well.
5. Digital Image Correlation. Parameter identification can be done using full-field experimental data. Generic file-based as well as standard interfaces have been added for extracting full-field data. GenEx has been extended to accommodate multi-point histories. The gom/ARAMIS DIC system is supported.
   Postprocessing tools using LS-PrePost contour display of DIC and FE fields as well as field matching error and curve distance are available.
6. Binary databases have been introduced as a replacement of critical older XML databases to speed up interactivity during post-processing and to accommodate the potential growth in data volume brought about by the introduction of full-field data. Temporary databases (such as the JobResults and multihistory databases) have been converted to binary, and are also deleted after use.
7. Taguchi Method is now available as a classical parametric study method. It uses orthogonal arrays to study the effects of variables on the responses. These arrays enable the calculation of main and some interaction effect using a relatively small number of samples, but do not calculate higher order effects.
8. Efficient Global Optimization (EGO) has been added as an optimization strategy.
9. Stage library. To allow standardization of a design problem setup, solver stages or processes (stage groups) can be imported and exported to and from user-specified locations.
10. Interactive tables. Previously available simple (dumb) tables for design data have been enhanced to assume a more spreadsheet-like behavior. Multiple tables can be generated. Tables now allow new design point generation in a selected region of interest as well as the simulation of newly generated points. Using an existing metamodel, responses can be interpolated using user-specified parameter values. The highlighting of infeasibility has been enhanced and table rows can be re-ordered according to a selected column. As in the past, tables also interact with plots. The point categories feature has been enhanced. The current development progression is intended as a precursor to an independent graphical tool including many functions currently only available in the pre-processor GUI.
11. LS-TaSC has been added as a solver.
12. Training examples are packaged together with the LS-OPT installation and can be run from the starting menu. (Windows)