Domain Aware

Machine Learning

Surrogate Modeling

Altaventus has experience in use of several types of surrogate models which are computationally cheaper versions of a full-fidelity model, which realizes comprehensive optimization and uncertainty quantification studies. Few examples of our past and current activities are shared below.

  • Data Fitting – Kriging, Radial Basis Functions, Regression

  • Machine Learning– Supervised/Unsupervised Learning

  • Multi-Fidelity– Computation-Experimentation Hybrid Database

  • Surrogate-based Custom Design Cycles

Machine Learning

Machine Learning is proving itself to be a very effective technology that has found quite a lot of usage in many fields. As Scientific Computing is used for design and operation phases of many engineering problems, large volumes of structured and unstructured data are being generated. This has led many researchers and engineers to adopt Machine Learning techniques into their workflow in the form of a surrogate model.

Domain Aware Machine Learning

As the problems of interest in our work have deterministic and/or statistically meaningful stochastic nature, the governing laws of nature can be used to leverage the information latent in the data. This form of an approach with the cutting edge research in the globe comes in significant advantages;

  • High Accuracy – Unphysical results in the solution space is avoided. 

  • Data EfficiencyThe available data is fused with the already know encoded physical laws of nature

  • High Performance and Robustness – The end result is superior platform augmented with both simulations and data