Application of Multivariate Process Control on Particle Size Analyzer

Improving the Robustness of Particle Size Analysis by Multivariate Statistical Process Control

“…At this moment, the behavior of the transmission is rather random and depends on the manner which the free surface of the sample drops through the flow cell. Momentary particle size distributions are measured once per second, and the final result of the particle size measurement of one sample line is calculated as a weighted average over the whole measurement time for phases 1 to 4 [5,6].…”

“…The present measurement sequence control stops the integration of the particle size distribution 3 s before the dilution tank is detected as empty, based on the rate of change of the transmission in phase 4 and some heuristic rules. This method leads to rather good repeatability but there is a risk that the detection will fail if the parameters associated with the detection have not been set appropriately [1,3,5]. The aim of this research is to find a more accurate and reliable way to determine the end of the particle size integration period.…”

“…If the correlation matrix is used in these types of cases there is a risk that noise will be amplified in the measurements with low variance [16,21]. Comparison of the correlation and covariance matrix based PCA models in the application studied is reported in more detail in [5]. Different PCA models of the normal operation were created using RLS, CLS and PSD data for the three training data sets from each measurement sequence.…”