Timescale Prediction for Stably Stratified Turbulence (SST)

Modeling Framework

Best-Case Results for Turbulence LSTM and NODE models

Parametric Plots for SST

(a) LSTM model (b) NODE model (c) Combined contour plot. The blue lines are the LSTM results, and the red are the NODE results. The highlighted regions represent the lowest error areas. The error metric used is the maximum normalized rms error of the state variables.

Concluding Remarks

The ratio of timescales identified by the ML models is an integer multiple of the Reynolds number of the turbulent flow, which is a ratio of the flow's timescales. Thus, the results demonstrate the capability of our framework to extract timescales of highly complex dynamical systems.