(410d) Applications of Soft Sensors in Pharmaceutical Manufacturing Processes | AIChE

(410d) Applications of Soft Sensors in Pharmaceutical Manufacturing Processes

Authors 

Wang, Z. - Presenter, Pfizer Inc.
Kamyar, R., Pfizer Inc.
Mehdizadeh, H., Illinois Institute of Technology
Drying is an essential processing step for residual solvent reduction in pharmaceutical processes. For the manufacture of Active Pharmaceutical Ingredient (API), drying is mainly used to produce crystals with the right form and/or particle size distribution that are suitable for down-stream processing. As an energy-intensive and time-consuming process, drying is usually a bottleneck of pharmaceutical manufacturing processes. Therefore, it is critical to well understand the drying process and improve the process efficiency. For the experimental characterization of drying progression, loss on drying (LOD) is a commonly used off-line measuring approach. LOD requires to take samples periodically and to weigh the loss in mass when the sample is continuously being dried until the weight becomes constant. However, LOD is a time-consuming and intrusive method that can disrupt the drying process and increase the cycle time. It is thus desired to develop on-line measurements to monitor the drying process.

Process Analytical Technology (PAT) has been developed as an in-line or on-line monitoring approach to provide real-time data about the process status and critical process attributes (CQA), which can facilitate the process control and optimization. Spectroscopy methods, such as near infrared (NIR) and Raman, have been widely adopted to enable the implementation of PAT in monitoring the removal of solvents in the drying process [1, 2]. However, such methods require the development of chemometric models, which typically requires large amounts of experimental data for calibration and validation. Additionally, mass spectrometry has been developed to predict the drying end point without the need of chemometric models [3]. The common feature of the mentioned PAT methods is that they usually require the installation of spectroscopic devices which can be prone to fouling and introduce additional capital costs [4].

Recently, soft sensors have emerged as a powerful tool to provide real-time process monitoring capabilities in many industry fields. Soft sensors are mathematical models that use easily accessible on-line data to estimate other variables of interest that can only be determined at low sampling rates or through off-line analysis [5]. For pharmaceutical processes, soft sensors have great potential to either replace the existing hardware sensor or work in parallel to provide redundancy and verify whether the hardware sensor is working properly [6]. However, applications of soft sensors in the pharmaceutical industry are still limited, partly due to the challenge of validation by regulatory agencies [7].

In this work, we proposed a soft sensor based on a hybrid modeling approach to estimate the product LOD in real-time during the drying process. The model combines first-principle relationships with empirical correlation models derived from historical data, which is more robust than data-driven models in handling variabilities in raw material and process conditions. The developed soft sensor has been successfully calibrated, validated with experimental data, and implemented in an API full scale commercial process that follows good manufacturing practices (GMP) procedures.

  1. Hamilton, P., et al., Studies of particle drying using non-invasive Raman spectrometry and particle size analysis. 2011. 136(10): p. 2168-2174.
  2. Morris, K.R., et al., Accelerated fluid bed drying using NIR monitoring and phenomenological modeling. 2000. 26(9): p. 985-988.
  3. Dodda, A.G., et al., Using online mass spectrometry to predict the end point during drying of pharmaceutical products. 2014. 19(1): p. 122-131.
  4. Pla, D.L., et al., Moisture soft sensor for batch fluid bed dryers: A practical approach. 2018. 326: p. 69-77.
  5. Luttmann, R., et al., Soft sensors in bioprocessing: a status report and recommendations. 2012. 7(8): p. 1040-1048.
  6. Kourti, T.J.A. and b. chemistry, The process analytical technology initiative and multivariate process analysis, monitoring and control. 2006. 384(5): p. 1043-1048.
  7. Gernaey, K.V., et al., A perspective on PSE in pharmaceutical process development and innovation. 2012. 42: p. 15-29.