Produce prediction modeling of Industrial production processes using the improved PLS-CM

In the industrial production process, the configuration of raw materials plays an important role in predicting products and improving the efficiency of the production process. Therefore, this paper establishes the produce prediction model of industrial production processes based on the partial least squares(PLS) method integrating the correlation matrix(PLS-CM).The correlation matrix is used to evaluate the correlation among variables, and eliminate the weak correlation variables in ethylene production. Then the PLS method based on other strongly correlated variables is established to predict the ethylene products. Because the initial independent variables have serious correlation and redundant variables, the PLS-CM can simplify the variable system and remove the invalid variable to realize the regression modeling of small samples. Finally, the PLS-CM is used in the produce prediction model of the ethylene industry. The proposed model predicts the ethylene products based on the lightdoil, naphtha, raffinate, hydrogoil, lhydr, c345 in crude oil. Compared with the traditional neural network and the PLS, the PLS-CM achieves the best result which indicates the PLS-CM has more advantages in the model prediction.