This paper studies the effluent turbidity control of a deep bed rapid sand filter run by direct filtration method. In spite of the highly nonlinear filter dynamics, the maximum filter run time and optimum chemical usage are achieved. The disturbances such as the filter flow rate and the influent (raw water) turbidity that directly affect the effluent turbidity are compensated successfully. The behavior of the filter is determined by on-line measuring of the influent and effluent turbidities, the bed pressure drops, the flow rate, and the temperature of the influent. The required alum is directly dosed into the filter by a dosage pump controlled by the main computer. The extraction of a mathematical model of a sand and polymer based filter is very difficult and many times impossible because of the complex nature of the system. Furthermore the filter dynamics is slightly time-dependeni and quite nonlinear, making the use of linear control ineffective. Fuzzy control is proposed as most effective for this application, because of its ability to finely tape the control actions nonlinearly across different turbidity error ranges. An integral controller is added in parallel to the fuzzy controller to enhance the control performance for steady state errors.