Download files
Abstract
A paste thickener produces underflow with a high solids concentration by removing water from a dilute suspension. Paste thickening technology has been gradually gaining acceptance in the mining industry. Some of the main challenges in thickening are the complex dynamics due to severe interactions between operating parameters and the existence of a large range of process time constants. Additionally, being a downstream process, paste thickeners suffer from fluctuating feed conditions from upstream processes. Furthermore, the operation of paste thickeners involves a tight operating window as the underflow properties are sensitive to small changes in concentration.
The above challenges motivate this work to develop a systematic advanced control strategy, in particular, model predictive control (MPC) to effectively and efficiently regulate this complex, multivariable process. In this thesis, the sedimentation-consolidation model is adopted and analysed to identify important dynamical features and the key process parameter which will be considered in control system design. The model is then validated using industrial plant data. An extended Kalman filter is developed to estimate in real-time the key process parameter which affects thickening dynamics significantly. Through extensive simulation studies, it has been demonstrated that the proposed MPC approach can deliver a higher underflow solids concentration and a better regulated underflow removal rate.
In addition to the basic MPC, several important extensions are developed based on three practical considerations. Firstly, it is illustrated that incorporating ``future'' time-varying constraints in the MPC can help deal with the control difficulty arising from the co-disposal of underflow and coarse reject, leading to improved control performance. Secondly, an MPC with a non-uniformly spaced optimisation horizon is proposed to deal with the timescale multiplicity of thickening. The proposed approach provides similar levels of performance as compared to a conventional MPC with the advantage of reduced computational complexity. A stability ensuring condition for the proposed MPC is also developed. Thirdly, an MPC is developed to control both underflow solids concentration and rake torque. The nonlinear rake torque control problem is converted into a linear MPC problem, reducing the computational complexity of the optimisation problem. Simulation studies have shown that the proposed approach can effectively control the underflow solids concentration while minimising the rake torque, potentially preventing serious operational problems such as underflow blockage.