Modelling of ultrafiltration fouling by neural network

摘要: Optimisation of ultrafiltration pilot plants requires a better knowledge membrane fouling. In the field drinking water production, phenomena involved in fouling are very complex and interdependent because numerous compounds contained raw waters. As no model is available for this application, statistical modelling tool called neural network used paper to predict total hydraulic resistance at end filtration cycle after next backwash, using some parameters concerning quality (turbidity temperature) operating conditions, given experimental site. Different structures have been evaluated, information current previous cycle. Some them allow prediction with good accuracy. They take into account as inlets permeate flow rate, pressure turbidity, able effects reversible on resistance.