Implementation of an artificial neural network as a PAT tool for the prediction of temperature distribution within a pharmaceutical fluidized bed granulator.

In this study, a novel in-line measurement technique of the air temperature distribution during a granulation process using a conical fluidized bed was designed and built for the purpose of measuring the temperature under the Process Analytical Technology (PAT) and introduced to predict the establishment of temperature profiles. Three sets of thermocouples were used, placed at different positions covering the whole operating range, connected to data acquisition measurement hardware, allowing an in-line acquisition and recording of temperatures every second. The measurements throughout the fluidized bed were performed in a steady state by spraying a solution of PVP onto a lactose monohydrate powder bed in order to make predictions of the temperature distribution and the hydrodynamics of the bed during the granulation process using Artificial Neural Networks (ANNs) and to establish the different temperature profiles for different process conditions through the precise predicted information by the constructed, trained, validated and tested neural network. The model's testing results showed a strong prediction capacity of the effects of process variables. Indeed, the predicted temperature values obtained with the ANN model were in good agreement with the values measured with in-line reference method and hence the method can have an application as a predictive control tool.