Improving monitoring performance of on-line process based on PCA method

Principal component analysis (PCA) is very suitable for complex process monitoring and diagnosis, but it suffers many limitations such as great calculation load, poor real-time performance and lacking of on-line monitoring. Here, this paper presents a new method for multi-variable statistical process monitoring. Based on this new method, the principal component monitoring model can be generated in the principal component subspace, and the error monitoring model can be set up in the residual subspace. The method provides a human-machine monitoring interface and related fault-diagnosis interface for integrating Principal/Error/Multi-variable. This will change the real-time data of the multi-variable into the monitoring information of an integrated process, and present them effectively to the operators. With this method, on-line monitoring system was designed for the distillation process as an example, and the effectiveness of this method was illustrated.