Improving Pulp Level Detection in a Flotation Column Using a Neural Network Algorithm

"Pulp-froth interface position is a key variable in column operation. In plant practice, it is commonly estimated through a system of three pressure-transducers which allows the calculation of the pulp height. An alternate method based on the measurement of the conductivity profile across the interface has been proposed and successfully tested. Both methods however, present some practical problems related to the assumptions made in the calculation of the pulp height in the former case, and to the response time and the precision of the reading in the latter case.In this paper, the use of the neural network technique for improving the estimation of the pulp-froth interface in flotation columns is discussed. The technique is used with both conductivity-based and three pressure-transducer methods. Results obtained in a 2"" laboratory column and a 12"" pilot scale column are presented. A major obstacle in applying this approach in practise is the training of the artificial neural network. This problem is analyzed here for the conductivity-based method and some solutions are proposed."