(603a) Big Data-Based Fault Detection with Advanced Analytics in the Pharma Industry

Motivated by market pressure to reduce product costs, regulatory incentive to modernise manufacturing and a growing environmental concern to reduce waste has led the pharmaceutical industry to improve its manufacturing process. The impetus to adapt and modernise the manufacturing process can largely be attributed to the promotion of the Quality by Design paradigm (QbD) and guidance on Process Analytical Technology (PAT) by the FDA in 2004. The promotion of these methods coincided with a time where the cost of computing and data storage had declined, but also with developments in non-invasive online sensor technology, such as Raman and Near-Infrared (NIR) spectroscopy. This culminated into an increase intro the application of multivariate analysis (MVA) and specifically Multivariate Statistical Process Control (MSPC).

MSPC exploits the Big Data generated from a complex chemical process such as pharmaceutical production and can be used for process understanding, troubleshooting and on-line fault detection. MSPC tools include multivariant process charts and soft sensors and can be combined with APC methods to improve overall plant efficiency. MSPC aims to use big data to better understand a process, to develop a definition of normal operation and then use that to detect faults on-line. There are three main categories to classify as SPC methods (1) model-based (2) knowledge-based (3) data-based. Data-based methods are often the preferred method for pharmaceutical manufacturing. The advantages of data-driven methods are that they do not require first principals understanding of the process or a unit as they construct a model using input and output data. The most popular data-driven methods are multivariate latent/MVA methods (which include PCA, PLS and their adapted versions). These methods project data into the latent space and work as a feature extraction method and reduce the dataset into a smaller or manageable set that can be used in further applications of fault detection.

This paper presents an industrial case study which focused on fault detection towards operational improvements for a key API production, pursued in collaboration with a major multinational UK pharmaceutical company (GSK). The aim of this project is to exploit data availability towards constructing and validating a data-driven model of a specific API production unit of the GSK plant (Montrose, Scotland, UK). The project involves using large data sets of process data (temperatures, partial pressures, flow rates, and stream compositions, from an on-site GCMS analysis station), and implementing advanced MVA methods in MATLAB, in order to diagnose operational patterns and recommend feasible improvements towards reliable API production intensification.

LITERATURE REFERENCES

[1] E. Tomba et al., “General procedure to aid the development of continuous pharmaceutical processes using multivariate statistical modeling-An industrial case study,” Int. J. Pharm., vol. 444, no. 1–2, pp. 25–39, 2013.

[2] T. Kourti, “Application of latent variable methods to process control and multivariate statistical process control in industry,” Int. J. Adapt. Control Signal Process., vol. 19, no. 4, pp. 213–246, 2005.

[3] J. F. MacGregor and T. Kourti, “Statistical process control of multivariate processes,” Control Eng. Pract., vol. 3, no. 3, pp. 403–414, 1995.

[4] T. Kourti, “Multivariate dynamic data modeling for analysis and statistical process control of batch processes, start-ups and grade transitions,” J. Chemom., vol. 17, no. 1, pp. 93–109, 2003.

[5] J. F. MacGregor, P. Nomikos, and T. Kourti, “Multivariate Statistical Process Control of Batch Processes Using PCA and PLS,” IFAC Proc. Vol., vol. 27, no. 2, pp. 523–528, 1994.

[6] S. Garcia-Munoz and D. Settell, “Application of multivariate latent variable modeling to pilot-scale spray drying monitoring and fault detection: Monitoring with fundamental knowledge,” Comput. Chem. Eng., vol. 33, no. 12, pp. 2106–2110, 2009.

[7] M. Boiret, D. N. Rutledge, N. Gorretta, Y. M. Ginot, and J. M. Roger, “Application of independent component analysis on Raman images of a pharmaceutical drug product: Pure spectra determination and spatial distribution of constituents,” J. Pharm. Biomed. Anal., vol. 90, pp. 78–84, 2014.