McLean, David,Slocomb, Richard A. W.2009-03-252009-03-2519941994Source: Masters Abstracts International, Volume: 33-05, page: 1548.9780315959781http://hdl.handle.net/10393/9591http://dx.doi.org/10.20381/ruor-16403A unique problem in the continuous processing industries (CPI), arising from integration of statistical process control (SPC) with automatic process control (APC), is choosing the correct variable(s) for process monitoring. It was shown that the manipulated variable displayed superior process monitoring potential for first-order processes with moderate amounts of inertia, subject to an ARMA(1,1) disturbance when the autoregressive parameter, $\phi$, is larger than the moving average parameter, $\theta$. It was also shown for large amounts of process inertia that the reconstructed output (disturbance) was the most effective charting variable ($\phi > \theta$). However, when $\theta$ exceeds $\phi$ the residual chart performed the best. In some very severe cases, process observations are so highly correlated that they display a wandering mean. Such nonstationary disturbances are extremely problematic in the CPI. The n$\sp{\rm th}$-difference chart developed in this work, proved to be efficient at identifying the presence of assignable cause in first-order processes under Minimum Variance feedback control when subjected to IMA(1,1) disturbances displaying moderate degrees of nonstationarity ($\theta >$ 0.7). Furthermore, it was shown that the performance of these charts is significantly affected by the presence of the autoregressive parameter (ie., when $\phi >$ 0.2). A case study involving a C$\sb3$ splitter provided the basis for development of a six-step procedure for the integration of SPC with APC. While carrying out integration on the C$\sb3$ splitter it was determined that the cause of disturbance nonstationarity must be sought out and removed if successful integration is to be achieved. For most nonstationary disturbances the amount of natural drift is often greater than or equal to the size of the step shifts that one is trying to detect. Thus, one is not able to differentiate between the common cause and assignable cause disturbances. (Abstract shortened by UMI.)178 p.Engineering, Chemical.Thesis