Inverse Analysis of Experimental Scale Turbidity Currents Using Deep Learning Neural Networks

Abstract Despite the importance of turbidity currents in environmental and resource geology, their flow conditions mechanisms are not well understood. This study proposes verifies a novel method for inverse analysis using deep learning neural network (DNN) with numerical flume experiment data sets. Numerical sets turbidites were generated forward model. Then, DNN model was trained to find functional relationship between by processing The performance evaluated 2,000 test five Inverse results on indicated that can be reconstructed from depositional characteristics turbidites. For experimental turbidites, spatial distributions grain size thickness consistent sample values. Concerning hydraulic conditions, depth, layer‐averaged velocity, duration certain level deviation. depth had percent errors less than 36.0% except one experiment, which an error 193% duration. velocity 2.38%–73.7%. Greater discrepancies measured values concentration (1.79%–300%) observed relative former three parameters, may attributed difficulties measuring during experiments. Although did provide perfect reconstruction, it proved significant advance currents.