Prediction model of quality indices based-on RBF neural network in the raw slurry blending process

In the raw slurry blending process, red mud, alkali powder, blending ore and limestone are translated into ball mills to produce the raw slurry whose quality indices include calcium ratio, alkali ratio and water content. The operation control objective of the blending process is to control the quality indices into their targeted ranges. However, in the raw slurry blending process, quality indices can not be measured on-line using the instrument, and it is also difficult to obtain the accuracy model of quality indices. Therefore, the only way to obtain the actual quality indices is the manual chemical examination. For the disadvantage of manual method such as long cycle and low accuracy, it is difficult to realize the operation control objective. To solve this problem, with the integration of the subtractive clustering, RBF neural network and operator's experience, an intelligent prediction model of quality indices of raw slurry is proposed. The application results in some alumina factory have proven the effectiveness of the proposed method.