부산대학교 도서관

ABSTRACTDuring high stages in rivers, water flanks its banks and inundates its adjoining floodplains. This creates difference in depth of flow between the main channel and floodplains making the flow behaviour complex, thus leading to inaccuracies in estimation of discharge. At this stage, the section of the stream is termed as compound channel. Assessment of actual discharge carrying capacity in river sections always has its special importance as it has direct impact on planning, operational and management problems. Due to lack of observed river data at these high stages, simulation is usually relied on the experimental outputs. Present study tries to simulate quantum of flow observed in the laboratory in compound sections for different roughness with actual river flow. Using soft computing techniques, predictions have been carried out successfully and the results are compared with various traditional approaches. In this study a common platform is used with low stage channel data to high stage river data with different hydraulic characteristics using 11 types of approaches for discharge prediction. Out of all the methods used, ANFIS model performs best with high degree of coefficient of correlation (R) as 0.99 with mean square error of 0.164