| 초록 |
Ethylene vinyl acetate (EVA) is produced by reacting ethylene with vinyl acetate (VA) in a high temperature and pressure. In the reaction, a decomposition reaction is also occurred, which is exothermic and faster than polymerization reaction. The temperature of reactor increases drastically causing a thermal runaway. To avoid the phenomenon, the autoclave reactor of the EVA production process has to be monitored using multivariable statistical monitoring method. Many kinds of multivariable statistical monitoring methods have been developed in past several decades. However, Principal component analysis (PCA) is still the most popular multivariable statistical monitoring method because of its simplicity, low computational intensity, and it has shown good detectability of a fault prognostics. So, in this research, the PCA is applied to the process. The PCA extracts the most dominant principal components (PCs) on the basis of the latent values. When the extracted PCs are below (or above) 3 sigma of the PCs variances, it is considered as a fault. The result shows that the applied method has a good detectability and false alarm rate. |
