Reduced-order model

A reduced-order model (ROM) is an accurate surrogate of a high fidelity model. Based on intelligent sampling and robust data fitting, the resulting ROM enables accurate near-real time simulations. 

Our process

Based on high-resolution Thermal Desktop models, Veritrek captures the relationship between input factors (e.g. Thermal Desktop symbols) and output responses (e.g. node temperatures). Our process is simple:

  • First, a clear understanding of your ROM requirements is quickly developed. 
  • Next, highly trained engineers at LoadPath (our parent company) convert your Thermal Desktop model into a ROM.
  • A thorough evaluation of the ROM is provided which includes detailed test metrics that quantify its accuracy.
  • Finally, your ROM is delivered along with our Exploration Tool.

This process depends on the complexity of your model and ROM, but typically takes only a few days.

LHS.png

Sampling

Sampling is based on optimized Latin Hypercube methods that efficiently fill a design space. This approach is a widely-used approach for interrogating computer models. The result is a set of sampling points that provide the basis for generating training data.

Data Fitting

Data fitting is based on Gaussian Process (GP) methods. Robust in nature, GP data fitting can accommodate a wide-range of data without prior knowledge of its shape. In addition, it fits training data exactly which is advantageous for computer experiments.