DRINKING WATER TREATMENT: A NEURAL NETWORK MODEL FOR COAGULATION DOSING B. Lamrini*٫ A. Benhammou*, A. Karama* and M-V. Le Lann** * Laboratoire d’Automatique et d’Etude des Procédés, Faculty of Sciences Semlalia, PoBox:2390‚ 40000-Marrakech, Morocco E-mail :{lamrin.b, benhammou, karama@ucam.ac.ma} ** LAAS/CNRS‚ 7, Avenue du Colonel Roche‚ 31077-Toulouse cedex 4, France and INSA, DGEI, 135,Avenue de Rangueil, 31077-Toulouse cedex 4, France E-mail: mvlelann@laas.fr All correspondence should be addressed to: benhammou@ucam.ac.ma Abstract: The aim of this paper is to present the development and validation of a neural network model for on-line prediction of coagulant dosage from raw water characteristics. The main parameters influencing the coagulant dosage are firstly determined via a PCA. A brief description of the methodology used for the synthesis of neural models is given and experimental results are included. The training of the neural network is performed using the Weight Decay regularization in combination with Levenberg-Marquardt method. The simulation results of neural model compared to a linear regression model are illustrated with real data. Copyright © 2005 IFAC Keywords: Coagulation process; drinking water treatment; neural networks; weight decay regularization. 1. INTRODUCTION The use of artificial neural networks for process modelling and control in the drinking water treatment is currently on the rise and is considered to be a key area of research. The coagulation process which requires the addition of chemical coagulant is the critical process in the drinking water treatment. The control of a good coagulation is essential for maintenance of satisfactory treated water quality and economic plant operation. Basically, coagulant dosage is chosen empirically by operators based on their past experience, laboratory jar-testing and various information on water quality parameters. The jar-test apparatus simulates mixing, flocculation, setting, and a single test may take about one hour to be performed. Disadvantages associated with jar- testing are that regular samples have to be taken requiring manual intervention and operators can make manually in raw water quality. There is no mechanistic model describing the coagulant dosage related to the different variables affecting the process. Consequently, there is a need for a fast and reliable method for determining the required coagulant rate which can be used instead of the jar- test analysis. The purpose of this paper is to highlight the utility of artificial neural networks in drinking water treatment in particular coagulation modelling and control. Process data can be used directly to represent input-output process relationships. Neural networks proved to be extremely flexible in representing complex non-linear relationships between many different process variables (Cybenko, 1989). They do not require any a priori precise knowledge on the relationships of the process variables. Various applications of these models have been recently