(259g) Development of Data-Driven and Hybrid Models for Continuous Pharmaceutical Manufacturing Lines Under Industry 4.0 Framework

Continuous manufacturing has been a critical research area in the pharmaceutical industry for the past decade¹. With intensive collaboration among regulatory agencies, leading industrial players and academia, a large number of data-driven and first principle models have been developed to guide the design and simulation process². Data-driven models attempt to capture the system behavior using only input and output data from the manufacturing line³. Even though they are accurate for specific datasets, the models are highly dependent on the choice of training data, and their description of process physical phenomena is not explicit⁴. First principle based models simulate the process with known mechanistic equations, but it is hard to model the complex behavior of the line completely⁵. Recent development in Industry 4.0 has led to a focus in cyber-physical system, internet of things, and smart manufacturing^{6, 7}, and combined with the quality by design initiative from the FDA, the volume of data generated for the entire manufacturing cycle increases dramatically, allowing for a better application of the generated data. As a data science tool, machine learning can be used for analyzing the large data set, and the resulting data-driven models can be used to enhance first principle models for the development of a hybrid model. This similar approach has been done in other areas like catalytic reactor modeling^{8, 9} and HVAC system analysis¹⁰.

As one of the three main routes of continuous pharmaceutical manufacturing, the continuous direct compaction (DC) line has mechanistic and semi-empirical models developed^{2, 11}. In this study, an integrated data collection and data analysis framework is constructed for the DC line using Industry 4.0 concepts to facilitate the development of data-driven and hybrid models. After the implementation of the Industry 4.0 framework, both process and analytical data of the DC line can be centralized into a cloud-based platform, where users can retrieve and visualize all relevant manufacturing information in the system. With a large amount of data collected, various machine learning techniques like Support Vector Machine (SVM) and Artificial Neural Network (ANN) are applied to infer data-driven models of each unit operation and the entire flowsheet. The data-driven models developed can then be used as a gap-assessment tool to improve the previously developed mechanistic and semi-empirical models, leading to hybrid models. The wealth of data in the system are also used to test the applicability of both types of models. The standalone data-driven models can provide quick input-output correspondence in the absence of mechanistic models, but on top of it, the hybrid models can offer a layer of system understanding. The highlight of this work is to apply a data-driven approach to supplement the mechanistic models and to form hybrid models, which can be used to improve the predictability of the continuous pharmaceutical manufacturing system.

1. Ierapetritou, M.; Muzzio, F.; Reklaitis, G., Perspectives on the continuous manufacturing of powder-based pharmaceutical processes. AIChE Journal 2016, 62 (6), 1846-1862.
2. Wang, Z.; Escotet-Espinoza, M. S.; Ierapetritou, M., Process analysis and optimization of continuous pharmaceutical manufacturing using flowsheet models. Computers & Chemical Engineering 2017, 107, 77-91.
3. Boukouvala, F.; Muzzio, F. J.; Ierapetritou, M. G., Dynamic Data-Driven Modeling of Pharmaceutical Processes. Industrial & Engineering Chemistry Research 2011, 50 (11), 6743-6754.
4. Rogers, A.; Hashemi, A.; Ierapetritou, M., Modeling of Particulate Processes for the Continuous Manufacture of Solid-Based Pharmaceutical Dosage Forms. Processes 2013, 1 (2), 67-127.
5. Boukouvala, F.; Muzzio, F. J.; Ierapetritou, M. G., Predictive modeling of pharmaceutical processes with missing and noisy data. AIChE Journal 2010, 56 (11), 2860-2872.
6. Liao, Y.; Deschamps, F.; Loures, E. d. F. R.; Ramos, L. F. P., Past, present and future of Industry 4.0 - a systematic literature review and research agenda proposal. International Journal of Production Research 2017, 55 (12), 3609-3629.
7. Zhong, R. Y.; Xu, X.; Klotz, E.; Newman, S. T., Intelligent Manufacturing in the Context of Industry 4.0: A Review. Engineering 2017, 3 (5), 616-630.
8. Azarpour, A.; N.G. Borhani, T.; R. Wan Alwi, S.; A. Manan, Z.; I. Abdul Mutalib, M., A generic hybrid model development for process analysis of industrial fixed-bed catalytic reactors. Chemical Engineering Research and Design 2017, 117, 149-167.
9. Luo, N.; Du, W.; Ye, Z.; Qian, F., Development of a Hybrid Model for Industrial Ethylene Oxide Reactor. Industrial & Engineering Chemistry Research 2012, 51 (19), 6926-6932.
10. Park, S.; Ahn, K. U.; Hwang, S.; Choi, S.; Park, C. S., Machine learning vs. hybrid machine learning model for optimal operation of a chiller. Science and Technology for the Built Environment 2018, 1-12.
11. Boukouvala, F.; Niotis, V.; Ramachandran, R.; Muzzio, F. J.; Ierapetritou, M. G., An integrated approach for dynamic flowsheet modeling and sensitivity analysis of a continuous tablet manufacturing process. Computers & Chemical Engineering 2012, 42, 30-47.