Development of Sediment Transport Models for Selected Rivers in Malaysia Using Regression Analysis and Artificial Neural Network
Problems associated with sediments have become major issue for years as it affects not only the morphology of the rivers but the cause of flooding in several areas. Knowledge on sediment prediction is important in solving river engineering problems. In lieu of this, a study on the transport of sediments in natural rivers was conducted to observe both the bed sediment load and suspended sediment load with flow. The objective is to develop a sediment transport equation using hydraulic and sediment data extracted from 12 rivers. A total of 346 data were used in the analysis and validation works. Initially an attempt was made to evaluate the nine most commonly used equations. However, these equations failed to predict sediment transport to a desired accuracy. The poor performance resulted from the difference in sediment availability from the source, composition of sediments and river configurations. Some of the equations were physically based and derived using flume data. Naturally the conditions in the laboratory are not the same as in natural rivers. Thus, a new sediment transport equation is necessary for use in Malaysian rivers. Three statistical techniques namely the multiple linear regression, robust regression and artificial neural network were used to determine the functional relationship between the sediment discharge variables. Subsequently a total of 71 equations were derived. Analysis showed that the proposed equation derived using multiple linear regression gave the best prediction with an overall accuracy of 67.35%. Thus the proposed regression equation for Malaysian rivers under the flow range and sediment properties can be defined as a function of relative roughness on the bed, ratio of shear velocity and fall velocity, Froude Number and the ratio of shear velocity and average velocity. Examples of stable channel design using the proposed equations are also given.