DATADVANCE is adding new features and functionality to its pSeven design space exploration software platform at a lightning pace. pSeven 6.6 Release was unveiled April 5, followed by pSeven 6.7 three days ago. Highlights of the two releases include a new interactive Predictive Modeling Toolkit built on the company’s pSeven Core algorithms (formerly called MACROS), new ANSYS Workbench integration, and a rich array of feature updates and enhancements to further increase pSeven’s performance, stability and ease of use.
Highlights of pSeven 6.6 that make predictive modeling and data analysis easily available:
- Use any data source a user wants: bring in data from Excel using the updated and improved Excel import tool, read it from text files in virtually any tabular format, or load it from the pSeven project database.
- Train models with a click of a button: With the new SmartSelection technology implemented in the Predictive Modeling Toolkit, pSeven automatically selects the best training method and adjusts its parameters in order to train the most accurate model.
- State-of-the-art modeling algorithms let users add a priori knowledge and subjective information to their data. For example, users can set noise levels for data samples, or assign higher weights to important or more accurate sample points.
- Fine-tune model builder settings: Advanced users will benefit from the range of settings provided in the Model Builder. If a user wants to direct the modeling process in a given way, pSeven provides this steering capability. On the other hand, if a user is not sure where to go next, the SmartSelection technology will provide guidance.
- Validate models in a clean and intuitive way using the Model Validator tool to visualize model accuracy metrics.
- Evaluate a model that the user has trained and view the results immediately, thanks to the tight integration of the Predictive Modeling Toolkit with other pSeven data analysis tools. Or add the model to a pSeven workflow automating mass calculations.
- For multitasking, pSeven lets users train multiple models, validate and evaluate them in parallel, all in the multifunctional pSeven Analyze.
- pSeven models are available to use wherever the user wants: model code can be exported to MATLAB or C, or compiled into a DLL and called from Excel spreadsheets.
- Automate everything the user needs, even model training and export: pSeven supports approximation workflows through its std.ApproxBuilder block. In pSeven 6.6, std.ApproxBuilder also received significant updates and now supports all features available in the Predictive Modeling Toolkit, including support for incremental model training, data sample weighting and processing noisy data.
Like all previous releases, pSeven 6.6 aims to deliver greater flexibility and a wider feature set while staying backward-compatible in both program compatibility and continuing support for data analysis, optimization and process integration methods:
- pSeven 6.6 includes an updated robust optimization algorithm for problems that include computationally expensive objectives. Efficiency of the new method is based on more intensive use of modeling capabilities, which allows the optimizer to reduce the required number of objective function evaluations.
- Updated PTC Creo integration block provides full support for integration with PTC Creo 3.0.
- Updated SOLIDWORKS integration block with improved error handling and higher stability.
- Updated Python integration block with simplified import of third-party Python modules, further increasing pSeven’s capabilities for scientific computing.
Predictive Modeling Toolkit, the latest addition to pSeven’s wide range of data analysis and post-processing tools available in its Analyze mode, is a set of interactive tools that enables thorough data analysis, approximation model training, model validation and evaluation with high accuracy and less effort. The Predictive Modeling Toolkit lets users:
- Train predictive models on their data with Model Builder.
- Let SmartSelection automatically find the best modeling algorithm.
- Find the best model with Model Validator.
- Evaluate models in pSeven or export model code for external use.
- Predictive Modeling Toolkit: train predictive models on user data.
- Let SmartSelection find the best modeling algorithm and tune its parameters to improve model quality.
- Control every aspect of model training if desired.
- Store models in pSeven project database where they are always available for evaluation or export.
- Predictive Modeling Toolkit: find the best model.
- Analyze model quality using various error metrics.
- See how models compare in terms of accuracy and robustness on training and test data.
- Visually test quality of even complex high-dimensional models.
- Newest pSeven capabilities for surrogate-based robust design optimization.
- Integration with LMS Imagine.Lab Amesim in design optimization.
- Run-ready workflows—a diverse collection of complex workflows wrapped into an easy-to-use customized run interface, illustrating pSeven’s capabilities for producing platform-based applications.
Algorithmic core updates
- New surrogate-based optimization algorithm for robust design optimization problems with computationally expensive objective functions.
- Improved optimization results filtering, increasing overall solution quality.
- New approximation features: support for incremental model training, sample weighting and additional output noise data for more accurate model training.
- Full support for PTC Creo 3.0.
- Improved SOLIDWORKS integration with better stability and error handling.
- Advanced Python integration and the capability to import third-party modules in embedded scripts.
Usability and performance
- Less memory usage for long-running workflows.
- Better-looking plots in Analyze.
- Faster file browsing and cleaner graphical interface.
- Many additional smaller enhancements.
In the just-announced pSeven 6.7 Release, major updates include:
- New ANSYS Workbench integration block, std.ANSYSWorkbench, enables integrating ANSYS Workbench projects into pSeven workflows. This block maps a project’s input and output parameters to ports, allowing use of ANSYS Workbench simulations in pSeven to perform optimization and other design space exploration tasks. A new project showing std.ANSYSWorkbench usage is also added to the pSeven 6.7 examples package.
- The Predictive Modeling Toolkit now supports exporting approximation models to a special format that provides better compatibility with Excel. It exports model C code ready to be compiled into a DLL that can be imported in Excel without making changes in the code manually.
- SOLIDWORKS 2016 support added in std.SolidWorks.
- Decreased memory consumption when loading high-dimensional approximation models in the Predictive Modeling Toolkit and std.ApproxPlayer.
- Redesigned 2D plots with better rendering and more visualization settings.
This release also updates the std.NX integration block with better handling of errors from Siemens NX, and provides several other improvements in stability and performance.
ANSYS Workbench integration
- Run ANSYS Workbench simulations from pSeven workflows.
- Map ANSYS project parameters to ports so they can be analyzed and changed in the workflow.
- Usage example included in the pSeven examples package.
Updated Modeling Toolkit
- New special C code format with better Excel compatibility.
- Excel-compatible C code is ready to be compiled and used as a standalone DLL in VBA functions in spreadsheets.
- Less memory consumption when loading high-dimensional approximation models.
- SOLIDWORKS 2016 support in std.SolidWorks.
- Better error handling in the Siemens NX integration block, std.NX.
- Detailed logging in all block that support external error handling, in particular all integration type blocks.