Predicting the particle size distribution in twin screw granulation through acoustic emissions

Abstract

A non-destructive process analytical technology for monitoring the complex particle size distributions inherent to twin screw granulation (TSG) was presented, based on acoustic emissions (AE). AE spectra were collected during the wet granulation of lactose monohydrate at different liquid to solid ratios from 8 to 14% and correlated with the particle size distributions (PSD) to train a neural network model. Predicted PSD for particle sizes from 44 to 7000 μm based on the AE spectra showed the largest root mean squared error of 4.25 wt% at 2230 μm. After transforming the AE data with a newly created digital filter based on particle impact mechanics to address auditory masking, the maximum error for predicting fractions was reduced to below 1 wt%. This technology shows great promise in predicting the complex size distributions present in TSG in real time.