Publications Details

Visualization of high dimensional turbulence simulation data using t-SNE

Wu, J.; Wang, J.; Xiao, H.; Ling, Julia L.

Computational mechanics simulations often output large, high dimensional data sets. Analyzing and understanding these data sets can prove challenging because of the difficulty associated with visualizing data in more than two dimensions. In this paper, the t-Distributed Stochastic Neighbor Embedding (t-SNE) methodology is used to reduce the dimensions of computational fluid dynamics data sets for improved visualization. This visualization technique enables easy comparisons between data sets. These comparisons are particularly useful in assessing the applicability of data-driven turbulence models, which are most accurate on flows that have similar characteristics to the flows on which the data-driven model was calibrated. The t-SNE technique is applied to a range of different flow configurations and the results and modeling implications are discussed.