A Machine Learning Approach to Qualitatively Evaluate Different Granulation Phases by Acoustic Emissions

Wet granulation is a frequent process in the pharmaceutical industry. As a starting point for numerous dosage forms, the quality of the granulation not only affects subsequent production steps but also impacts the quality of the final product. It is thus crucial and economical to monitor this operation thoroughly. Here, we report on identifying different phases of a granulation process using a machine learning approach. The phases reflect the water content which in turn influences the processability and quality of the granule mass. We used two kinds of microphones and an acceleration sensor to capture acoustic emissions and vibrations. We trained convolutional neural networks (CNNs) to classify the different phases using transformed sound recordings as the input. We achieved a classification accuracy of up to 90 % using vibrational data and an accuracy of up to 97 % using the audible microphone data. Our results indicate the suitability of using audible sound and machine learning to monitor pharmaceutical processes. Moreover, since recording acoustic emissions is contactless, it readily complies with legal regulations and present Good Manufacturing Practices.

2.2 Granulation Process

We used a binary placebo formulation of pharmaceutical powder materials for the granulation process. In particular, we used Lactose Monohydrate (GranuLac® 200, Meggle, Wasserburg, Germany) and Microcrystalline Cellulose (VIVAPUR® 101, JRS Pharma, Rosenberg, Germany). For all experiments, the ratio of Lactose Monohydrate to Microcrystalline Cellulose was 80-to-20. The granulation process was performed with a total dry mass of 150 g as described in Reference [1]. Water was added with 2 mL min−1.

Each of the 10 granulation runs took 75 minutes to complete and was divided into three phases. These three phases were defined on the basis of granule properties, related to their moisture content. We have recently shown that, for the given binary placebo formulation, a moisture content on dry basis of 33 percent (25 percent on wet basis) gives optimum material properties for processing [1]. According to that, we defined the granulation phase containing the optimum amount of water as opt. The granulation phase containing less water than the optimum is called dry. The granulation phase containing more water than the optimum is called wet. In the first phase (dry), water was continuously added for the first 25 minutes until the moisture content on dry basis reached 33 percent, i.e. until 50 g of water has been added. In the second phase (opt), mixing was carried out for 25 minutes without adding water. In the third phase (wet), water was added for 3.5 minutes to achieve a moisture content on dry basis of 38 percent, i.e. 7 g of water was added. After that, mixing was continued for another 25 minutes. We chose time intervals of equal length both to get sufficient data and to ensure that the data sets are evenly distributed throughout the phases. Table 1 lists all phases, the corresponding timing information, and the details about water addition during each phase.

Download the pre-print as PDF here: A Machine Learning Approach to Qualitatively Evaluate Different Granulation Phases by Acoustic Emissions – pre-print

or read it here

Fulek, R.; Ramm, S.; Kiera, C.; Pein-Hackelbusch, M.; Odefey, U. A Machine Learning Approach to Qualitatively Evaluate Different Granulation Phases by Acoustic Emissions. Preprints.org 2023, 2023070857. https://doi.org/10.20944/preprints202307.0857.v1

Read more articles about “Machine Learning” here:

• Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

• Optimizing environmental sustainability in pharmaceutical 3D printing through machine learning

• Revolutionizing Drug Formulation Development: The Increasing Impact of Machine Learning