Estimation of component concentrations of sodium aluminate solution via PLS and Hammerstein recurrent neural networks

Abstract

In this paper, a new on-line soft sensing method is proposed for component concentrations of sodium aluminate solution. With this sensing strategy, real-time control and optimization can be realized in aluminate production plants. Several advance techniques are used, such as PLS (Partial Least Squares), Hammerstein model, recurrent neural networks and least square algorithm. Industrial experiment results show that the proposed soft sensing algorithm is effective.