The application of knowledge-based systems to the simulation of gold extraction processes
Most unit operations and processes of gold extraction are ill-defined and cannot be simulated adequately by conventional mathematical models. This is caused mainly by the complex chemistry and mineralogy of most ore slurries. An object-oriented knowledge-based system (KBS) is proposed here for the dynamic simulation and fault-diagnosis of various sub-processes in the extraction of gold. Two approaches are used, viz.: (1) where the parameters of simple empirical expressions are related to a data base via an evolution of operating conditions, and (2) where the dynamic behaviour of a system is finger-printed by the gradient of change of a state variable. The first approach is applied to the simulation of gold leaching in batch reactors, cascades of continuous reactors, and countercurrent Kamyr towers. The depletion of cyanide and/or oxygen is also taken into account. The second approach is applied to CIP and CIL cascades. The centre (deep knowledge) of this model is a database containing concentration-time data and an accompanying generalized kinetic model. The knowledge-base is defined by various facts, objects, rules and functions, which capture both deep as well as shallow knowledge regarding the process. Furthermore, it is explained how KBS simulation can be used in fault-diagnostics and the identification and characterisation of ores, slurries and adsorbents. The proposed KBS produces realistic simulations of experimental and published data. It is evident that the accuracy of prediction is entirely dependent on the accuracy and population density of the data base. In most cases, dynamic simulation by KBS has reduced CPU time drastically in comparison with the numerical solution of more phenomenological models. © 1990.