摘要
Ultrasonic attenuation spectroscopy （UAS） is an attractive process analytical technology （PAT） for on-line real-time characterisation of slurries for particle size distribution （PSD） estimation. It is however only applicable to relatively low solid concentrations since existing instrument process models still cannot fully take into account the phenomena of particle-particle interaction and multiple scattering, leading to errors in PSD estimation. This paper investigates an alternative use of the raw attenuation spectra for direct multivariate statistical process control （MSPC）. The UAS raw spectra were processed using principal component analysis. The selected principal components were used to derive two MSPC statistics, the Hotelling＇s T2 and square prediction error （SPE）. The method is illustrated and demonstrated by reference to a wet milling process for processinR nanoparticles.
Ultrasonic attenuation spectroscopy （UAS） is an attractive process analytical technology （PAT） for on-line real-time characterisation of slurries for particle size distribution （PSD） estimation. It is however only applicable to relatively low solid concentrations since existing instrument process models still cannot fully take into account the phenomena of particle-particle interaction and multiple scattering, leading to errors in PSD estimation. This paper investigates an alternative use of the raw attenuation spectra for direct multivariate statistical process control （MSPC）. The UAS raw spectra were processed using principal component analysis. The selected principal components were used to derive two MSPC statistics, the Hotelling＇s T2 and square prediction error （SPE）. The method is illustrated and demonstrated by reference to a wet milling process for processinR nanoparticles.
