Improving Empirical Models with Machine Learning

Inaccuracies of sediment transport models largely originate from the difficulties to describe the process in precise mathematical terms. Machine learning (ML) is an alternative approach to reduce the inaccuracies of sedimentation models. It utilizes available domain knowledge for selecting the input and output variables for the ML models and uses non-linear regression techniques to fit the measured data. Two ML methods, artificial neural networks (ANN) and model trees, are adopted to model bed-load and total-load transport using the measured data. Special effort was made to ensure that the empirical formulations known from the sedimentation research would be taken into account. ML models were found to be more accurate than the existing ones.