New block: Model builder
Available in: 
pSeven Enterprise v2025.06
The Model builder block enables predictive modeling in pSeven Enterprise. Data collected from simulations and physical tests can be transformed into an executable model that quickly computes results for any new inputs, accelerating complex simulations by orders of magnitude. For example, such models can be used to run full-factorial experiments and select candidate optimal design parameters, which are then verified using more precise methods.
The process of creating a model is called training, while using a model to compute results for new inputs is called prediction. The overall approach is known as predictive modeling, also referred to as metamodeling, surrogate modeling, response surface modeling (RSM), reduced-order modeling (ROM), machine learning (ML), etc.
Creating a model with sufficiently accurate predictions (good predictive power) requires selecting the most suitable modeling technique for the specific problem and dataset, plus extensive parameter tuning - often through trial and error. In pSeven products, a special technique called SmartSelection automates this entire process without requiring additional configuration or expertise from the user. It automatically searches for an optimal technique and parameter settings to balance computational efficiency and model accuracy based on the training data characteristics.
The Model builder block primarily serves as a SmartSelection launcher within pSeven Enterprise:
- Provide a CSV file path or Table value to the Training sample input.
- Optionally pass an existing model to the Initial model input as a baseline.
- Control the process by specifying high-level settings for training and validation, known sample properties and additional model requirements - needing only general methodological knowledge, no deep expertise in mathematics, modeling algorithms or specialized domains.
- Save trained models in native format or export them to source formats, including C source code and MATLAB-compatible - models are not locked into pSeven Enterprise.
Model builder offers quick setup through direct access to all settings in the Block properties pane, requiring no separate UI configuration. The block also features seamless integration with the Design space exploration block, allowing generated DoE to pass directly from the Result designs output to the Training sample input. This integration enables automatic detection of inputs/outputs, names, and properties without requiring additional setup.
For more experienced users, Model builder provides access to the most granular training settings through pSeven Core options. This enables full control over technique selection and parameter configuration. Experts can use Model builder for conventional training workflows while maintaining complete control over the process.