What's New
View new features for HyperStudy 2024.1.
Altair HyperStudy 2024.1 Release Notes
New Features
- Training PhysicsAI Models within HyperStudy
- PhysicsAI ribbon-style workflow has been added in the GUI to enable training models.
- Defining Precision for Numeric Design Variables
- A new tool has been introduced to specify the format for numeric input variables. Format definition will work in conjunction with some mathematical models to produce meaningful values such as tolerance requirement in the manufacturing process.
- Variable Table for Templatized Text Files
- TPL Editor tool now has a dedicated table to conveniently edit formats and navigate to the variable location in the file.
Enhancements
- Auto-Capture Variable Formats in TPL files
- When Parameterized File model type is used, HyperStudy will automatically detect the user-specified formats in .tpl files and store them in the Data Format Builder settings.
- Support for Dynamic Cases in FMU
- FMI model type now supports both static and dynamic simulation modes. In dynamic cases, HyperStudy will generate Data Sources for both simulation data and timestep component.
- Building romAI models on Vector Responses
- Previously romAI (hosted library in Twin Activate and Compose) fit model was available for only scalar responses and now data sources are supported as well.
- Hyperparameter Support in Custom Reports
- Defining user inputs for custom reports via GUI is now available.
- Post-process Reliabilities with Constraints
- Reliability plot in Stochastic Approach now displays constraints if applied.
Resolved Issues
- Timeout option for child processes was not supported on Linux.
- In a multi-model environment, Spreadsheet model failed when multi-execution used.
- Special characters in Windows username were not handled correctly.
- HyperView model failed at importing variables.
Announcements
- In optimization approach, hyperopt and status files are no longer generated. Please use deprecate flag with 8157 to restore the legacy behavior.
- Xopt is no longer supported.
- Next release, Data Source Tool and Hstp Reader will be deprecated and will be entirely removed in 2025.0 release.
- Testing object,
getAlgorithmUsable
, has been introduced to external algorithm API to check and ensure the usability of the fit model on the chosen response. The object is derived from the newFitExtensionBase/OptExtensionBase/FillExtensionBase
class which is a replacement ofExtension
.#old class FitRandomForestExtension(hstfitext.Extension): def isAlgorithmUsable(self, setup, algorithm): if algorithm == 'RF': return "" #new class FitRandomForestExtension(hstfitext.FitExtensionBase): def getAlgorithmUsable(self, setup, algorithm): import alt.hst.algobase.hstalgocommon as hstalgocommon useInfo = hstalgocommon.CommonUsabilityInformation() if algorithm == 'RF': # error & block useInfo.addError('Error reason') # warning but allow useInfo.addWarning('Warning reason') return useInfo
Altair HyperStudy 2024 Release Notes
New Features
- Make FMI Models part of design exploration
- A new model type to leverage machine learning models standardized with Functional Mock-up Interface.
- Leveraging romAI for predictions
- Altair romAI is now integrated as one of the out-of-the-box fit models.
Resolved Issues
- Preview in Media Sources did not refresh when result file was changed.
- Media Sources tools did not track original files of inclusion data.
- Genetic Algorithm with Hooke-Jeeves method went into an infinite loop.
- In tutorial HS-1615, .cfx file required to be updated.
- Model resources were omitted when archives were generated.
- OptiStruct models with PSHELL thickness parameters did not properly update the fiber distances for stress calculations.
Announcements
- As of this release, Xopt is deprecated.
Known Issues
The following known issues will be addressed in a future release as we continuously
improve performance of the software:
- The drag-and-drop option for loading archive files (.hstx) is not functional. Please load the files via browsing.