René del Villar

Neural net-based softsensor for dynamic particle size estimation in grinding circuits

Abstract A softsensor was developed to estimate the hydrocyclone overflow dynamic particle size distribution of a grinding circuit using neural network models. Because of the inherent dynamics associated with the grinding process, the time histories of the four variables measured around the hydrocyclone along with the past value of the cyclone overflow particle size or the past percent passing 53 μm were utilized as the inputs to formulate the neural model for the estimation of the present value of the particle size. The fact that the neural net-based softsensor performs well in the particle size estimation suggests that the model developed has captured the essence of the process dynamics. To cope with the time-varying nature of the grinding circuit, on-line adaptation of the neural network was considered. A simplified neural net-based softsensor was thus developed by making use of the principal component analysis for the data reduction so as to simplify the neural model structure and to render it more suitable for on-line adaptation.

Estimation of Bias and Entrainment in Flotation Columns Using Conductivity Measurements

Abstract The cleaning action of the wash water in the flotation column froth is enhanced by a net downward flow of water, called the bias . The evaluation of this variable in industrial columns, through a water balance over the lower part of the collection zone, presents some serious deficiencies since it requires extensive instrumentation and the calculated value is usually unreliable, particularly in transient operation which renders the implementation of automatic control very difficult. In this work, the relationship between the bias and the conductivity profile across the interface has been modelled using a neural network algorithm, a non-linear model whose performance in interpreting complicated patterns has been widely demonstrated. A 5.8 cm diameter column equipped with a series of conductivity electrodes in its uppermost section was used in an initial series of experiments designed to train the neural network with two-phase data. The excellent results obtained led to new series of experiments in both the 5.8 cm column with a three-phase (slurry-air) system and in a 30.48 cm diameter column with a two-phase system.

Automatic Control of Flotation Columns

Since their first commercial application for mineral separation in the early 1980s, flotation columns have become a standard piece of equipment in mineral concentrators particularly for cleaning operations. This chapter presents and discusses the most recent advances in instrumentation and automatic control of flotation columns. It also examines how current industrial practice could benefit from recent academic developments in these areas. A particular emphasis is placed on the development of specific sensors for the continuous monitoring of process operations and their regulation.

Learning control of an autogenous grinding circuit

Abstract This paper presents the results of the application of a new process control technique, the Learning Control, to a mineral processing operation. A hierarchical system of learning automata is used as a model of the controller. An empirical simulator capable of reproducing the dynamic of the autogenous grinding process is considered as the random environment in which the hierarchical system of automata operates. A probability distribution is associated to the manipulated variable. This distribution is continuously adjusted by the learning system using a reinforcement scheme. Numerical results have demonstrated its control properties, transparent tuning and robustness, while requiring minimal computational load.

Multivariable predictive control of a pilot flotation column

The aim of this work is the control of hydrodynamic variables of a pilot flotation column working with a three-phase system (air-water-ore) in an industrial environment. Since hydrodynamic variables are closely related to metallurgical and economical performances of the unit, the implementation of such a control strategy is crucial to optimize its operation. The hydrodynamic variables here considered are the gas hold-up in the collection zone and the fraction of wash-water underneath the interface. They are controlled by manipulating gas flow rate and wash-water flow rate respectively. Hydrodynamic variables are controlled through a constrained model predictive control (MPC) strategy. This choice is dictated by the interdependency of these controlled variables and the capability of predictive controllers to handle process constraints. Identification tests lead to a representative model of the system under nominal operating conditions but indicate variations of process behavior with changing operating regimes. Results show good control performances, confirming the potential use of MPC to control hydrodynamic variables for real-time optimization of full-scale flotation columns.

Using Semi-Physical Models for Better Control. Part I: Modeling of A Pilot Flotation Column.

Abstract Two ways to model a process are physical modeling and empirical modeling. Physical models rely on basic laws of physics whereas empirical models are obtained by fitting an empirical equation to experimental data. Empirical models are less complex and easier to obtain but they are valid only for specific process operating conditions. Moreover, their parameters do not have any physical meaning and a priori information if often completely neglected. Part I of this paper presents two models of a pilot flotation column by mixing both physical and empirical approaches. The models are accurate but remain simple enough to be used for controller design. Part II of the paper details the backstepping control based on the first model and the model reference non linear control using the second model. Comparison is then made with a typical industrial PI control for the same process.

OPTIMIZATION-FREE CONSTRAINED NONLINEAR PREDICTIVE CONTROL – MINERAL PROCESSING APPLICATIONS

Abstract A difficulty with constrained nonlinear control is the minimization of the cost function. With complex system representations such as fundamental models, the required optimization algorithm may be complex to implement, setting its parameters may be difficult and the calculation time may be long. To overcome these problems, an innovative optimization-free predictive control scheme is proposed. The minimization of the cost function is replaced by a simple and easy-to-compute simulation. Two mineral processing applications illustrate the very good performances of this new algorithm.