What's New

Altair HyperStudy 2022.3 Release Notes

New Features

New Low-Discrepancy Quasi-Random Sampling Method: A Sobol Sequence based sampling method has been added in both DOE and Sampling Fit approaches. It is extensible and supports design variable constraints.
On-demand Model-Independent Responses: Internal metadata-type or user-defined responses can be optionally added as channels in the Post Processing step.
New Data Source Tool for Extracting Data of Text Files: A new extraction tool has been added to read data from XML files.
Parameterization of Additional Resource Files: Files added in Model Resource can be parameterized independent of the main resource file.
Understanding The Role of Each Design Variable Constraint in Generating Datasets: Each Design Variable Constraint plays a role in redefining the initial design space and it is important to identify their individual contributions. Hence, individual and combined pass rates have been added.

Enhancements

Improving Efficiency and Accuracy of Predictive Models: Zero gradient information, when available, is consumed by fit models to boost efficiency and accuracy.
Multiple Responses from a Single Text File: File Assistant now supports extracting multiple outputs from the same text file.
Confidence Interval Plots for Non-LSR Fit Models: Previously, confidence intervals were available only in LSR. Now, it is extended to MLSM and RBF.
Auto-Detecting Dependencies in RADIOSS Connection: Radioss integration identifies dependencies such as engine, include, and initial state files and auto-include them in the Model Resources area.
Customizable Model Connections: A model-specific settings option has been introduced to run integrated processes in a particular manner. For example, Excel connection can be set to save a copy of the updated session in each run directory.

Altair HyperStudy 2022.2 Release Notes

Note: In version 2023, Altair will deprecate HyperWorks Desktop. Hence, HyperMesh Desktop Model will be retired and MotionView Model will be updated to use hstp connection scheme.

New Features

Auto-Create a New DOE Approach Based on Pareto Ranking: Input variables are ranked based on the percentage of their global contribution. Using the ranking, a new DOE approach can be created with only the most influential ones being active.
Visualize Range of Estimates for Each Prediction Point: For both Scalar and Vector Fit, an option in Trade-Off has been added to display confidence intervals.
Pedigree Information for Each Run: The origin of run number and source approach is now reported for each evaluated process.
Dedicated Post-Processing Tab for Curve Matching: In the Post-Processing step, curve-matching performance for each evaluated point can now be displayed.
New Data Source Tool for Extracting Data of Text Files: In addition to the ASCII Extract option, Regex Extract tool has been added to extract data via regular expressions.
Mobility for Vector Predictions: Vector Fit models can now be exported in FMU format.

Enhancements

Excel Connection: Option to Save as Copy: A copy of an excel session can optionally be saved in each run directory.
Building Vector Fits with Efficient Sampling: Sampling Fit approach has been extended to support vector fit models.

Resolved Issues

Correction to numerical discrepancies in HS-1610.
Undo action now works for removed models.

Altair HyperStudy 2022.1 Release Notes

Note: In version 2023, Altair will deprecate HyperWorks Desktop. Hence, HyperMesh Desktop Model will be retired and MotionView Model will be updated to use hstp connection scheme.

New Features

Predict Vector-Based Responses: Functionality has been added to predict series of data points indexed in order.
New Report Option for Fit Models: Regression models can now be exported in Functional Mock-up Unit (FMU) format. This will allow the use of HyperStudy regression models to be deployed in any platform that supports Functional Mock-up Interface.
New Software Integration: Altair Flow Simulator Model has been added to support parametric Thermo-fluid System Design studies.

Enhancements

Improved Accuracy and Efficiency of Mathematical Methods via Special Gradient Definition: In a multi-model study environment, a response may be completely independent of some design variables. That is, a response of model A and design variables of model B. Introducing zero correlation between a response and design variables help reduce dimensionality, which improves efficiency and accuracy of fit and optimization methods. Add Always Zero button and Always Zero option have been added in Gradient Tab to automatically detect and define such relationships.
Support for Additional Variable Types in B PreProcessor Model: The connection has been updated to support both string and integer variable types.
Additional Convergence Criterion for GRSM: A threshold option has been added to track improvement of objective value over the span of user-defined number of iterations. Once the threshold is reached, GRSM terminates.
Vector Response Data via Custom Reports: New report API has been introduced to export data sources through custom python scripts.

Resolved Issues

Pyfit now handles inactive fits correctly.
HyperStudy now writes the actual values of upper and lower bounds for linked variables in hst_input.hstp files. B PreProcessor connection requires bounds explicitly specified.
You can now remove a model without removing up all dependencies first.
Area Curve in Area Tool now scales correctly.

Altair HyperStudy 2022 Release Notes

New Features

Review Images of Design Iterations: Like Data Sources, you can now define Media Sources. Use this new feature to associate images of simulation results with each run to later enhance the postprocessing experience in the MediaView.
New Report Generation Options: Altair Binary Format (.abf) report generator is now available.
New File Export Option: Altair Binary Format (.abf) can be exported with data from all Approaches.

Enhancements

Visualize Near-Best Design Points: A tolerance option has been added to visualize alternative optimum design points on top of the best one. The user-specified tolerance is the relative difference between objective value of the best point and each alternative design.
Import Variables per Model: Import Variables task can be performed individually per model.
Constraint Markers in Parallel Coordinates: Responses with constraint definition can be optionally annotated in Parallel Coordinate Plots.
Image Support by HyperView Model: To complement the new MediaView capability, HyperView model connection has been updated to auto generate images for each run.
User-Defined DOE Algorithms: The use of external sampling method is now supported in DOE approach.
Leveraging HyperStudy algorithms: The internal methods of HyperStudy are now allowed to be used in external python processes.
New Tutorial: The HS-1670 tutorial has been added to show how to set up a Radioss model.

Resolved Issues

Order of constraints influenced GRSM results.
Goal columns of inclusion matrix were not being populated which caused data being omitted in Optima tab.
Progress bar showed negative remaining time.
Filtering outliers per response caused discrepancy between Internal and Excel fit.

Announcements

Support for studies earlier than v14.0 has been deprecated, please contact support if you need to open an older version.
Run Before option in Model Conditions has been retired.
Inspire Studio Model will be deprecated in the future.