(697c) Data Accuracy for Smart Manufacturing in Continuous Pharmaceutical Systems | AIChE

(697c) Data Accuracy for Smart Manufacturing in Continuous Pharmaceutical Systems


Ganesh, S. - Presenter, Purdue University
Moreno, M., Purdue University
Su, Q., Purdue University
Shah, Y., Purdue University
Nagy, Z. K., Purdue University
Reklaitis, G. V. R., Purdue University
The principles of Quality by Design (QbD) and the innovative applications of Process Analytical Technology (PAT) tools have enabled the adoption of continuous processing paradigms to advance pharmaceutical manufacturing. As the community progresses towards strategies for real-time release testing and implementing advanced process control strategies, it is important to ensure accuracy in process data. Process and material quality measurements using PAT tools are highly valuable, however, present multiple challenges such as extensive calibration, data handling, sensor fouling, measurement bias, measurement drifts etc. The utility of such tools in real time hence can become limited unless properly managed. However, by leveraging the extensive real time generated by robust equipment sensors and field devices along with predictive capabilities offered by process models, the application of data reconciliation and gross error detection can enable enhanced real-time process data accuracy and validation [1]. Improved accuracy in raw measurements enhances plant-enterprise integration with increased reliability.

Data reconciliation (DR) and gross error detection (GED) are systems engineering tools for further improving the accuracy and consistency of process data through systematic data rectification and checking [2]. DR can be posed as a constrained optimization problem, which has at its objective providing the best estimate of process variables, which satisfy the process model, including material and energy balances. Under DR random errors and gross errors associated with measurements are rectified through a model-based approach providing the sensor network has the necessary redundancy. Reliable state estimates of process variables not only enhances process monitoring capabilities but also improves the performance of control systems.

DR is a powerful tool for continuous processes whose application in industries such as oil and gas is well known. It is a powerful tool for leveraging large datasets for improving monitoring consistency in the pharmaceutical industry. In this paper, we demonstrate improved accuracy in real-time monitoring of relevant process variables in a continuous dry granulation tableting line. The dry granulation process exhibits nonlinear response and fast dynamics and hence poses interesting challenges for monitoring and control. In this paper, an earlier simulation study highlighting the application of DR for a partial dry granulation line [3] is extended to investigate experimental performance using the entire tableting line. The DR is conducted using a steady-state model in order to achieve fast computational times. The DR framework is experimentally demonstrated using real time measurements of the relevant CQA's on the continuous tableting pilot plant at Purdue University.


[1] M. J. Bagajewicz, Smart Process Plants: Software and Hardware Solutions for Accurate Data and Profitable Operations, New York: McGraw-Hill, 2010

[2] S. Narasimhan and C. Jordache, “The Importance of Data Reconciliation and Gross Error

Detection,” in Data Reconciliation and Gross Error Detection, Elsevier, 1999, pp. 1–31.

[3] S. Ganesh, M. Moreno, J. Liu, M. Gonzalez, Z. Nagy, G. Reklaitis. Sensor Network for Continuous Tablet Manufacturing, Proceedings of 13th International Symposium on Process Systems Engineering, 2018 (accepted)