The paper proposes an artificial neural network (ANN) model of multilayers perceptron type (MLP3:10:1) adapted for mechanical-chemical treatment system of an industrial effluent (i.e. coagulation-flocculation - sedimentation applied for an industrial effluent produced in a manufacturing plant of bricks and other ceramic products). This model of multiple inputs-one single output considers three input variables (independent variables) like the temperature (z1), dose of polyelectrolyte (z2) and agitation time (z3) and one single output variable (dependent variable) as the removal of turbidity (Y1) or colour (Y2). Consequently, the proposed ANN model is optimized and also tested for some data from outside of the training experimental field. The optimal removal of turbidity (91.7%) is performed working at a temperature of 20°C, with a polyelectrolyte dose of 20 mg/L, for 30 min of agitation at 50 rpm, and in the case of optimal colour removal (92.2%) by working at a temperature of 26°C, with a polyelectrolyte dose of 15 mg/L, for no more than 30 min of agitation at 50 rpm, respectively.