In-line moisture measurement during granulation with a four-wavelength near-infrared sensor: an evaluation of process-related variables and a development of non-linear calibration model

The near-infrared set-up based on simultaneous detection of four wavelengths was applied for in-line moisture measurement during fluid bed granulation. In addition to the spectral response, several other process measurements describing the state of the granulation were evaluated. The near-infrared moisture measurement is disturbed by the variation in physical properties of the sample (e.g., temperature, particle size, bulk density). The factors explaining the non-linearity of spectral response during different phases of granulation could be extracted. Combining all this process information improved the prediction capability of the multivariate calibration models tested (partial least squares and artificial neural network (ANN)). The back-propagation neural network approach was found to have most predictive power with the independent test data.