Sequential analysis 770 INDEX

A sequential procedure was further developed by Tong and Crowe (1996) by applying sequential analysis of the principal component test using the sequential probability ratio test (SPRT). Dunia et al. (1996) also used PCA for sensor fault identification via reconstruction. In that paper it was assumed that one sensor had failed and the remaining sensors are used for reconstruction. Furthermore, the transient behavior of a number of sensor faults in various types of residuals is analyzed, and a sensor validity index is suggested, determining the status of each sensor. 

The concentration data obtained from each sample analysis were expressed as fractional parts and normalized to sum to 100. The normalized data were statistically analyzed, and three principal components (A=3, Equation [1]) were calculated. The PCB constituents (varlbles) are numbered sequentially and correspond to peak 1, peak 2,. .. to peak 105. The structure and retention index of each constituent in the mixture were reported by Schwartz et al. (9). The tabular listing of the data is availedile from the present authors. 

One or more detectors is attached to the output of the columns. For routine analysis of linear homopolymers, this is most often a Differential Refractive Index (DRI) or a UV detector. For branched or copolymers, however, it is necessary to have at least two sequential detectors to determine molecular weight accurately. Branched polymers can be analyzed using a DRI detector coupled with a "molecular weight sensitive" detector such as an on-line viscometer (VIS) or a low-angle laser light scattering (LALLS) detector. 

---