Dead-end ultrafiltration (UF) has been considered as a more energy efficient operation mode compared to cross-flow filtration for the production of drinking/potable water in large-scale water treatment systems. Conventional control systems utilize pre-determined set-points for filtration and backwash durations of the constant flux dead-end UF process. Commonly known potential membrane fouling parameters such as feed water solids concentrations and specific cake resistance during filtration were not taken into considerations in the conventional control systems. In this research, artificial neural networks (ANN) predictive model and controllers were utilized for the process control of the UF process. An UF experimental system has been developed to conduct experiments and compare efficiencies of both the conventional set-points and ANN control systems. The novelty of this study is to utilize commonly available on-line and simple laboratory analysis data to estimate potential membrane fouling parameters and subsequently utilize the ANN control system to reduce water losses. Reduction of water losses were achieved by prolonging filtration duration for feed water with low turbidity using the ANN control system. This advanced control system would be of interest to operators of industrial-scale UF membrane water treatment plants for the reduction of water losses with existing facilities.