Reduced-order model

A reduced-order model (ROM) is an accurate surrogate of a high fidelity model. Based on intelligent design space sampling and robust data fitting, the ROM acts as a statistical emulator constructed from high resolution simulation and provides accurate thermal analysis results in near real-time. 

Veritrek Creation Tool

The Veritrek Creation Tool uses a statistical scheme based on sampling and data fitting an underlying Thermal Desktop model to create a ROM. This approach is considerably different than nodal reduction methods in that it relies on a set of high-fidelity simulations (i.e. training data) to generate the ROM. In doing this, the reduced-order model approach is robust and can be easily applied to other problem classes, model types, and software packages.

Updated ROM process.png
LHS.png

Sampling

The first step in developing a ROM is carefully selecting sampling points. The Veritrek Creation Tool utilizes Latin Hypercube Sampling (LHS) space-filling design to efficiently identify and evaluate interior points. Veritrek's LHS sampling algorithm was developed based on concepts of the Maximin Method. This approach is widely used for interrogating computer models. The result is a set of sampling points that provide the basis for generating training data. 

Data Fitting

The Veritrek Creation Tool achieves data fitting using Gaussian process (GP) regression methods. This approach provides an exact fit to the training data, does not impose a specific model structure, and incorporates knowledge about the underlying function in the data. As such, GP modeling is a non-parametric modeling technique, where the training data is used to discover the model properties in a supervised manner.