A data-driven adaptive multivariate steady state detection strategy for the evaporation process of the sodium aluminate solution

The evaporation process of sodium aluminate solution is one of the main processes for alumina production. Due to uncertainty of production environment and existence of random measurement noise, process condition is changed frequently, which easily leads to unstable production. In this paper, an adaptive multivariate steady state detection strategy is proposed for the evaporation process of sodium aluminate solution. In the proposed strategy, steady state detection variables are selected firstly according to partial correlation analysis between the concentration of the outlet mother liquor and its major influencing variables. Then, an improved K-means clustering algorithm is presented to identify the outliers which would affect steady state detection results. By analyzing changes in velocity and acceleration of variables, a steady state evaluation index is defined with quantitative trend information extraction. The flexibility and effectiveness of the proposed steady state detection strategy are validated by industrial data of sodium aluminate solution evaporation process