Main Functionalities
- Several improvements for the Smart Superelement:
- Possibility to add parameters in MINN
- Possibility to enforce 0 eigenvalues in MINN
- Possibility to include files/mesh files
- Autofill creation form
- Improvement of learning speed for MINN
- Optimization: SQP algorithm bug fix
Known Bug
- Kriging interpolation (direct or with POD, FFT, Clustering) with cubic basis function is returning the error
“LU decomposition failed. at “
ODYSSEE Lunar – GUI
- Possibility to add parameters in MINN. It is now possible for users to add parameters such material properties, thickness and so on inside the MINN.
- Possibility to enforce 0 eigenvalues in MINN for rigid body modes.
- Possibility to include files/mesh files inside Smart Superelement.
- Autofill creation form for exporting Smart Superelement.
- Learning speed for Neural Networks (MINN and MLP) has been improved.
- The names of the parameters are now written in the <optimization.cmd> file.
- In the Lunar Help document, a note has been added for Hammersley and Halton DOE algorithms in the Appendix 2 concerning the number of parameters recommended for these DOE algorithms.
- The following error message is now displayed when FMU runs out of tokens:
“no license found for Quasar”
- If the Y curves are 2,000 points-long or more, the tested curves, validation curves, and imported curves are by default set to be displayed with a smaller thickness to avoid slow GUI response. For the tested curves and validation curves, the thinner style is applied to all datasets, while for the imported curves, it is possible to adjust the thickness from one dataset to another independently.
- Bug Fixes
- When creating a new project thanks to the import of datasets from hdf5 files, it was not possible to change the solver directly in the project configuration window during project creation, an error message appeared.
- In Sensitivity, with a project containing 2 datasets or more, when the user selects a curve to highlight in one dataset and switches to another dataset, the curve highlighted was the one of another case.
- In the Lunar Help document:
- Page 133, even though second order (quadratic) regression was presented, title was referring to first order (linear) regression, which could be confusing. Slide title has been corrected.
- The link page 161 was broken because cadlm.com domain does not exist anymore. The document the link was pointing on has thus been uploaded on the Product Documentation Center and the link has been corrected accordingly.
ODYSSEE Pulsar - Parser
- An error message now appears if the user tries to save his parsing project script under “pulsar.qsr”, “Pulsar.qsr”, “PULSAR.qsr” or any combination of upper and lower cases forming the word “Pulsar” such as “PuLSaR.qsr”. This allows to prevent a bug occurrence when importing the Pulsar library.
- The Pulsar Help document has been updated.
ODYSSEE Quasar – Solver
- Learning speed for Neural Networks (MINN and MLP) has been improved.
- New function read_OUTPUT() added in the <Useful_Functions.qsr> allowing to display in the command prompt the content of the OUTPUT file generated from external function execution.
- The error message of MLP function that appears when normalization is not possible has been updated.
- Bug Fix
- With the MLP solver, if the data contained columns full of zeros, the normalization was not possible, which blocked interpolation. This is fixed, if there are columns full of zeros, these are skipped, and normalization is not applied.
ODYSSEE Nova – Optimizer
- Bug Fix
- The SQP algorithm was diverging since 2023.1 version.