Identificação de uma planta piloto de flotação em coluna

Abstract

The aim of the present work is to investigate the application of nonlinearidentification techniques to a pilot column flotation which belongs to Centro de Desenvolvimento da Tecnologia Nuclear (CDTN). The flotation column technique used by the pilot plant is commonly used in ore beneficiation both in Brazil and other countries. This technique is well known for its efficiency and versatility in produce clean concentrates. The model representation used is multivariable nonlinear autoregressive, moving average models with exogenous inputs (NARMAX) of the polynomial type. The present work can be divided in three parts. The first one, composed by chapters one to three, presents a brief survey on system modeling, column flotation and nonlinear system identification. The second part, composed by chapter four and appendix B, is devoted to identification of multiple input multiple output (MIMO) systems. The basis of identification of single input single output (SISO) systems described in chapter three is extended to MIMO systems in chapter four. The computational routines derived of that theory and used in this work are presented in appendix B. The last part, composed by chapters five, six and appendix A, presents the main results obtained by applying identification techniques to the pilot plant. In chapter five and appendix A the dynamical testing of the plant is described as well as the best identified models. Inputs are the signal that controls the tailing flow pumps, wash water flow and air flow. The model outputs are the pressure values at the measuring points in the collection zone. The main conclusions are presented in chapter six.