Vol. 10, Special Issue 12 (2021)

Author(s):
Tushar Rathod, Vikram Singh, Devi Singh, Mukesh Kumar and Alex Thomas

Abstract:
For river structures, a sediment prediction is required. The quantity of suspended sediments has been estimated using station observations and machine learning modelling approaches. The sediment concentration was measured during this research using hydrometeorlogical data such as river flow, sediment and precipitation observed between 2006 and 2015 at the Hoshangabad station in Madhya Pradesh. The generalized feed forward (GFF) approach with various learning rules and activation functions has been used to estimate the sediment load. The correlation coefficient (r), mean square error (MSE), normalized mean square error (NMSE), efficiency (CE), and coefficient of determination (R2) were used to compare these models. When comparing the observation and model results, the GFF models provided consistent results in estimating the silt content of rivers. Even so, the overall performance of LinerAxon – Conjugate showed slightly better correlations and lower error performance than the others.