(661b) Multivariate Statistical Algorithm to Monitor and Control a Continuous Powder Mixing Process Using Online NIR Spectroscopy

NIR spectroscopic was used to get on-line spectra and obtain immediate results of API concentration at the output of a continuous powder mixer in order to assess the efficiency of the mixing. The results were compared with values obtained with ultraviolet spectrophotometer. Calibration curves were developed prior to continuous mixing experimentation for both UV and NIR methods. Considering that the mixing experimentation was performed between 0 and 100%w/w of Naproxen, it was created eight different Partial least squares (PLS) calibration models for NIR ranging from 0 to 100% instead of using only one calibration model so that a better prediction were obtained. The NIR precision and exactitude achieved were about 2-5% and 1-3% w/w respectively depending on the value of concentration studied. The calibration models were generated taking into consideration the variability of the process working at different RPMs of the mixer and at different feed rates of the feeders and considering as well the powders properties mainly the powder cohesion variation in the mentioned range of API concentration.During the process of mixing three different step changes were applied to the feeders to obtain three different API concentrations in the range 0-30% to characterize the dynamic response of the continuous mixer. The results obtained by UV and NIR showed stable mixing at the different API concentrations and achieved a reasonable mixing with values of RSD lower than 6%. The error obtained with the NIR and UV increases when increasing the API concentration and it was attained to the increase in powder cohesion which made more difficult the process of mixing.  The parameters of the dynamic response were similar to those obtained theoretically.