Skip to Main Content
During the operation of the industrial process, one of the optimal control objectives is to control some technique indices that represent the quality, efficiency and consumption of the product processing into their targeted ranges. So, it is important that technique indices can be obtained accurately and opportunely. However, in some industrial processes, technique indices can not be measured on-line using instruments, and there are complex natures between technique indices and the key variables that affect the technique indices, such as strong nonlinearity, heavy coupling and difficulty of description by the accurate model. It is difficult to obtain the technique indices accurately and opportunely in these industrial processes. To solve this problem, integrating the subtractive clustering, RBF neural network and operator's experience, a general prediction model of technique indices, which is suitable for many industrial processes, is proposed. Based-on the past and current process data, the prediction model, which is comprised of 7 modules, can predict the values and trends of technical indices on-line with high accuracy. An application case study is given to illustrate the method being applied to the raw slurry blending process in an alumina factory, and the application results have proven the effectiveness of the proposed method.