Abstract:
In this paper soft sensing on coking models of reactor in the process of continuous catalytic reforming is studied. At first according to the analysis of technics principles several major parameters affecting coking are found out and used as assitant variables. BP algorithm is used for training neural networks to establish coking model. In order to modify the model structure PLS BP algorithm is also used for training neural networks again to establish another coking model. At last the second model is applied in practive after revised. The result of measurement is satisfactory.