CHARACTERIZATION OF FLOTATION PROCESSES WITH SELF-ORGANIZING NEURAL NETS

Flotation processes are difficult to describe fundamentally, owing to the stochastic nature of the froth structures and the ill-defined chemorheology of the froth. Considerable information on the process is reflected by the structure of the froth. In previous work it has been shown that structural features extracted from flotation froths can be related to the behavior of flotation processes in a qualitative way through the identification of certain behavioral regimes or classes by using a supervised neural net as classifier. Although useful as an aid to control decisions, this method is less suitable for quantitative or dynamic analysis of the behavior of flotation plants. In this paper a new method for the analysis of flotation plants is consequently proposed, based on the use of order preserving maps of features extracted from digitized images of the froth phase. The construction of these maps by means of a self-organizing neural net is demonstrated by way of examples concerning the analysis of industrial copper and platinum flotation plants.     

NEURAL NETWORK-BASED PREDICTIVE CONTROL FOR MULTIVARIABLE PROCESSES

In this paper, the systematic derivations of setting up a nonlinear model predictive control based on the neural network are presented. This extends our previous work (Chen, 1998) into a multivariable system to explore the characteristics of the design. There are two stages for the development of nonlinear neural network predictive control: a neural network model and a control design. In the neural network model design, a parallel multiple-input, single-output neural network autoregressive with a model of exogenous inputs (NNARX) is proposed for multistep ahead predictions. In control design, the controller with extended control horizon is developed. The Levenberg-Marquardt algorithm is applied to training the NNARX model. Also, the sequential quadratic programming is used to search for the optimal manipulated inputs. The gradient of the objective function and constraints that require computation of Jacobian matrices are completely derived for optimization calculation. To demonstrate the control ability of MIMO cases, the proposed method is applied through two nonlinear simulation problems.     

NEURAL NETWORK FOR CLASSIFYING FLOW IN A TANK

A neural network trained with responses to a step input in concentration for a vessel with various percentages of perfect mixing, short circuiting, and distance-velocity lag classifies accurately flow patterns that the program has not encountered previously. Flow responses digitized at a selected sequence of times were the inputs and the percentages of three modes of flow were the outputs for a commercial program for neural networking. Default values of the program worked quite well for the layout of the network and for the convergence error for training, and tweaking these values had little effect on training time or performance.