Multivariate statistical process control (MSPC) forms the basis of process performance monitoring and detection of process malfunctions. The cornerstones of MSPC are the projection methods of principle component analysis (PCA) and projection to latent structures (PLS). These methods assume that process measurements are serially independent

which unfortunately is often invalid in real situations. For coping with serially correlated observations

A dynamic PCA (DPCA) based monitoring procedure is presented in this paper.By implementing PCA to observed disturbances other than process measurements

the new procedure eliminates the effects of process dynamics and meanwhile retains the ones of driving forces. Simulations carride out to a 6-dimesional linear system indicate that the DPCA can be viewed as a procedure of extracting driving forces

and the dynamic principle components obtained are essentially an estimation of them . A CSTR process model is finally utilized to verify the effectiveness of the DPCA-based statistical process monitoring.