الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 222 |  |  |
| | | from | 222 | | |
Abstract
Filtration is a solid liquid separation in which water passes through a porous medium to remove suspended or colloidal impurities. Deep bed filter is one of the most important types of filtration process in which solids are removed within the granular medium. Further, it is commonly used in either conventional water treatment plants or direct filtration plants. In the beginning of filtration process, the head loss is small and it can be easily calculated by different empirical equations, but as the filter bed gets clogged, the head loss increases and it can’t be easily calculated. In addition, Turbidity removal through filter run is influenced by varying operational conditions. The main objective of this research was to study the performance of deep bed sand filter under various operational conditions and utilize the electrical properties of the sand particulars inside the filter to stop and washout processes. In addition, Improve the performance of sand filters by using fiber paper sheet capping was studied. Also this study develop simple predictive models for head loss and effluent turbidity through deep filters , using both statistical regression analysis and the artificial neural networks (ANNs). In this study, experimental pilot plant was constructed in sanitary engineering laboratory, Faculty of Engineering, Mansoura university. Sand was used as a filtration media under different filtration rates ranged from 4 m/hr to 8 m/hr. Down flow was applied to the filter through sand media with size of 0.7-1.0 mm, while sand depth varied from 80 cm to 140 cm. Aluminum sulfate (alum) was used as coagulat in different doses ranged from 20 to 40 mg/lit. The used synthetic turbid water was prepared in different turbidities varying from 10 to 30 NTU. Turbidity removal and head loss were investigated as functions of sand depth, filtration rate, influent turbidity, run time and alum dose. Based on the experimental data, ripening periods for deep bed sand filters were obtained with different operational conditions. In addition, the effective sand depth for various filtration rates was determined. Sand filters could be controlled automatically for ripening, filtration, and backwash via the electrical charges throughout its media. Fiber paper sheet capping improved the performance of filters. In addition, filtration of high turbidity water could be achieved by using fiber sheet capping. Further, head loss and effluent turbidity predictive models for deep bed sand filter as function of five important parameters was proposed. These parameters are (filter media depth, filtration rate, influent turbidity, run time, and alum dose).The simple models yield highly accurate predictions with coefficient of determination (R2 of 0.88 for effluent turbidity model and R2 of 0.75 for head loss model). The sensitivity analysis using the proposed model showed that the most significant parameters on predicted head loss were the run time, filtration rate and filter media depth. The most significant parameters on predicted effluent turbidity were the filter media depth, filtration rate and influent turbidity. The artificial neural network models using the tanch axon function with architecture (5-100-1) for both models, with the same inputs as in the regression analysis yielded higher R2 values (R2 = 0.99) compared to the regression models.