Hybrid Intelligent Optimising Control for High-Intensity Magnetic Separating Process of Hematite Ore

I n spite of its abundant deposits in China, hematite ore has few uses due to its low grade, complex components, and tight adnation. Therefore, the ball mill is used to grind the ore so as to obtain the granularity-eligible ore pulp. The ore pulp is separated into the grade-eligible concentrate pulp and tailing pulp using a high intensity magnetic separator. The concentrated pulp is then refined and turned into concentrate powder by the thickener and the drier. Through adjusting the magnetic field intensity of the high intensity magnetic separator, the rinsing water flow of the separator and the pulp density of the feed system, the concentration grade can be increased as high as possible and the tailing grade can be largely decreased. As a result the metal recovery and the concentration grade can be improved. For these reasons, the concentration grade and the tailing grade are crucial technical indices that reflect product quality and efficiency during the high intensity magnetic separation process (HIMSP). However, it is difficult to measure the concentration grade and tailing grade online continuously. These two grade indices have strong nonlinearity and uncertainty with respect to the rinsing water flow, the exciting current, and the feed density. They are so complicated that it is hard to describe them with mathematical models and complete feedback control by general control method. The technical engineer is needed to determine the set points of the rinsing water flow, the exciting current, and the feed density by experience. Furthermore, these three parameters should be controlled respectively to track set points, and the actual values of the concentration grade and the tailing grade within their target scopes. Unfortunately, the manual operation cannot accurately define these set points in time when the boundary conditions change. As such, the manual operation cannot ensure that the actual technical indices are inside their desired ranges of target values. In view of industrial processes where the dynamic characteristics between the technical indices and the output of the control loop can be represented by linear models, the model prediction control technology has been favourable in recent years in producing the optimal set points for the control loops. This can ensure that the outputs of the control loops track the set points so that the technical indices are controlled within their desired target ranges[1] [2] . However, if dynamic characteristics between technical indices and control loop outputs cannot be described by certain mathematical models (e.g., the step-heated furnace and laminar cooling systems), the hierarchy control method which generally includes a supervisory control layer and a loop control layer is preferred to control the technical indices within their desired target ranges [5] . There has not been any relevant report on its application in the HIMSP until now.