Data-driven-based self-healing control of abnormal feeding conditions in thickening–dewatering process

Abstract

The issue of self-healing control of abnormal feeding conditions in the thickening–dewatering process is considered herein. This process is widely used in the mineral processing industry to separate ores from the slurry. Owing to the instability of the upstream flotation process, the flow rate of feeding slurry generally fluctuates substantially. In this scenario, decision-making with regard to the economical operation of the thickening–dewatering process is challenging. Moreover, it results in the pressure rake of the thickener or ore leakage of the filter press. To address these abnormalities, a data-driven-based self-healing control scheme is proposed in this study. The contributions can be summarized as follows. (1) Based on the information provided by the pressure sensors inside the thickener, a dynamic data-driven model structure is established to predict the behaviors of the thickener. The underflow concentration is described by a nonlinear KPRM model, and the future trends of the pressure sensors are calibrated by several dynamic ARX models. (2) The FDA classifier is used to identify the abnormal feeding conditions, and the average future trajectories of the input variables are estimated; (3) The data-driven-based multitiered dynamic optimization problem is formulated to address abnormalities. The optimization results are obtained by solving these problems sequentially, with the former tiers providing the information to the latter ones. Experiments have been carried out in a self-developed simulation platform. The proposed self-healing control scheme can maintain the safety of the thickening–dewatering process as well as consider the energy consumption and operator convenience.