Neural net based knowledge extraction from the historical data of an industrial leaching process
Date
1996
Authors
Rademan J.A.M.
Moolman D.W.
Lorenzen L.
Van Deventer J.S.J.
Aldrich C.
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
A methodology is proposed to analyse and model an ill -defined and poorly understood leaching process from historical data by artificial neural networks (ANN) and statistical techniques. It was found that the most effective process models were produced by a learning vector quantization (LVQ) neural network and a back propagation neural network. Rather than attempting a quantitative prediction of the performance of the process, the outputs of the process are classified as either high or low in order to cope with uncertainty in the data. Moreover, since this approach is similar to the way in which plant operators interpret the data, it is anticipated that acceptance and implementation of these models on the plant will be facilitated significantly. For the case study described in this paper, the LVQ neural net models were able to classify correctly the outputs of the process with an accuracy of between 75% and 82%. The LVQ neural net models were used to perform sensitivity analyses on the different process variables to determine the relative importance of the input variables. The information from the sensitivity analyses, together with the statistical means of the input variables for each output class, confirm the relative importance of each variable and give an indication of the effect that a change in the variable will have on the process. By applying this method of data analysis and modelling to historical data of an ill -defined industrial leaching process, valuable knowledge about the process was obtained, which could be used by the plant personnel and operators to improve the control of the process.
Description
Keywords
Data acquisition, History, Knowledge based systems, Mathematical models, Neural networks, Performance, Process engineering, Statistical methods, Leaching
Citation
Hydrometallurgy
43
03-Jan
43
03-Jan