Partial least-squares regressive analysis and modeling for the yield of acrynolitrile reactor

Modeling the acrynolitrile reactor is the foundation for the optimizing and advanced control of the reactor. The algorithm of partial least - squares regression (PLSR), which is a new kind of statistical data-analysis method, is briefed. Because the reactor has these characteristic: long delay and low signal-noise ratio, and there is grievous multicollinearity in the operating variables. The model with the selective independent variables (ratio of raw material, temperature, etc.) and the yield of acrynolitrile as the dependent variable is built up by using PLSR approach. And then we analyze and evaluate the model. The results demonstrate that the variable screening is reasonable and the satisfied values of the yield can be obtained from the PLSR model.