Computational Intelligence Techniques Applied to Coagulant Estimation Models in Water Purification Process

Four coagulant estimation models in water purification process are presented. For its developing computational intelligence techniques are used, which include neural networks, fuzzy logic (Sugeno and Mamdani type) and ANFIS structures. The methodology described is based on the operator's experience extraction from operational historical data (in neural network, fuzzy logic Sugeno type and ANFIS case), and the linguistic information given by an experimented plant operator (in fuzzy logic Mamdani type case). According to the reached results, by applying some of this models in a control system strategy, could be possible overcome some of the current limitations found in the most widely used techniques for coagulant dosing in water purification plants; jar tests and streaming current detector.