An integrated model for caustic ratio prediction in the alumina digestion process

The caustic ratio of the product in the alumina digestion process is obtained by off-line analysis with a long time delay, which leads to delayed process control, and therefore, fluctuation of the caustic ratio. So, it is significant to predict the caustic ratio on-line. On the basis of the reaction mechanism in the digestion process, a mechanistic model based on material balance for the caustic ratio was built. The fuzzy expert rules for caustic ratio prediction were summarized according to the operating experiences by taking the production conditions as inputs. The mechanistic model and the expert model are integrated. A kernel extreme learning machine (KELM) model was then built to compensate the integrated model, which uses time series of the error of the integrated model as inputs. So, finally, an integrated prediction model for the caustic ratio was established by parallel connection of the corrected mechanism model with the compensation model. The validation results showed that the integrated prediction model can effectively solve the problem of on-line measurement, which gives a foundation for the optimal control in the alumina digestion process.