Abstract:
Irrigated agriculture is faced with challenges that include sediment loading in the river basins and dams. The management of sediments in river basins and waterways has been an important issue for water managers throughout history. Water managers are faced with similar challenges caused by siltation of water reservoirs and irrigation water conveyance systems. As a copping strategy to counter the low irrigation application efficiency, designs of settling tanks are typically oversized. This is aimed at having enough detention time for the sediment particles to settle. This process is costly and tedious hence this study aimed at optimizing the hydraulic design parameters for settling basins with an intention of addressing such challenges. Artificial Neural Networks in MATLAB was used to build ANN models for predicting sediment settlement and eventually optimizing the hydraulic parameters of settling basins. The ANN was used to develop the relation between sediment settlement and different inflows by changing the different number of neurons in hidden layer from two to ten. The optimum hydraulic design parameters for a settling basin were calibrated using a physical model prepared in the Civil Engineering laboratory at Jomo Kenyatta University of Agriculture and Technology. Extensive data was collected to evaluate the physical and hydraulic parameters needed to calibrate and validate the ANN model. ANN model was trained and validated using a dataset obtained from the physical model. The developed simulated results were evaluated using recommended quantitative statistical analysis that compared the measured and the predicted data. The average accuracy between the artificial neural networks prediction and the real data in all the cases was over 90 %. The turbidity drop in a settling basin increased with the increase in flowrate and gave a biquadratic relationship, applicable in calculating turbidity drop. In addition, an ANN architecture of 1-9-1 was the best suited for predicting turbidity in Kiriku-Kiende settling basin where a quadratic relationship existed for flow rate against the optimum surface area for the settling tank. For flow rates of 5.7, 8.7, 9.9, 10.5 and 11.1 m3/s, the critical settling velocity was found to be 0.0034, 0.0044, 0.0024, 0.0026 and 0.0034 m/s respectively. The optimum surface areas for 5.7, 8.7, 9.9, 10.5 and 11.1 m3/s flow rates was calculated as 2.42m2, 3.04 m2, 3.75 m2, 4.20 m2 and 4.71 m2 respectively.
Further, on flushing, the settling tanks with continuous sediment removal gave a higher sediment efficiency flushing of 65.5% against 24.4% without flushing. Finally, ANN can be used as a decision making tool for turbidity prediction.
