Identification and control of wet grinding
processes - Application to the Kolwezi
concentrator
December 2013
Ecole Polytechnique de Bruxelles
Thèse présentée par
Moïse MUKEPE KAHILU
en vue de l’obtention du grade
de Docteur en Sciences de
l’Ingénieur
Promoteur :
Prof. Michel KINNAERT
Co-promoteurs :
Enhancing mineral processing techniques is a permanent challenge in the mineral and metal industry. Indeed to satisfy the requirements on the final product (metal) set by the consuming market, control is often applied on the mineral processing whose product, the ore concentrate, constitutes the input material of the extractive metallurgy. Therefore much attention is paid on mineral processing units and especially on concentration plants. As the ore size reduction procedure is the critical step of a concentrator, it turns out that controlling a grinding circuit is crucial since this stage accounts for almost 50 % of the total expenditure of the concentrator plant. Moreover, the product particle size from grinding stage influences the recovery rate of the valuable minerals as well as the volume of tailing discharge in the subsequent process.
The present thesis focuses on an industrial application, namely the Kolwezi concentrator (KZC) double closed-loop wet grinding circuit. As any industrial wet grinding process, this process offers complex and challenging control problems due to its configuration and to the requirements on the product characteristics. In particular, we are interested in the modelling of the process and in proposing a control strategy to maximize the product flow rate while meeting requirements on the product fineness and density.
A mathematical model of each component of the circuit is derived. Globally, the KZC grinding process is described by a dynamic nonlinear distributed parameter model. Within this model, we propose a mathematical description to exhibit the increase of the breakage efficiency in wet operating condition. In addition, a relationship is proposed to link the convection velocity to the feed ore rate for material transport within the mills.
All the individual models are identified from measurements taken under normal process operation or from data obtained through new specific experiments, notably using the G41 foaming as a tracer to determine material transport dynamics within the mills. This technique provides satisfactory results compared to previous studies.
Based on the modelling and the circuit configuration, both steady-state and dynamic simulators are developed. The simulation results are found to be in agreement with the experimental data. These simulation tools should allow operator training and they are used to analyse the system and to design the suitable control strategy.
As the KZC wet grinding process is a Multi-Input Multi-Output (MIMO) system, we propose a decentralized control scheme for its simplicity of implementation. To overcome all the control issues, a Double Internal Model Control (DIMC) scheme is proposed. This strategy is a feedforward-feedback structure based on the use of both a modified Disturbance Observer (DOB) and a Proportional-Integral Smith-Predictor (PI-SP). A duality between the DOB and PI-SP is demonstrated in design method. The latter is exploited to significantly simplify the design procedure. The designed decentralized controllers are validated in simulation on the process linearized model. A progressive implementation of the control strategy is proposed in the context of the KZC grinding circuit where instrumentation might not be obvious to acquire.