Abstract
Purpose
Reliable process monitoring in real-time remains a challenge for the pharmaceutical industry. Dealing with random and gross errors in the process measurements in a systematic way is a potential solution. In this paper, we present a process model-based framework, which for given sensor network and measurement uncertainties will predict the most likely state of the process. Thus, real-time process decisions, whether for process control or exceptional events management, can be based on the most reliable estimate of the process state.
Methods
Reliable process monitoring is achieved by using data reconciliation (DR) and gross error detection (GED) to mitigate the effects of random measurement errors and non-random sensor malfunctions. Steady-state data reconciliation (SSDR) is the simplest forms of DR but offers the benefits of short computational times. We also compare and contrast the model-based DR approach (SSDR-M) to the purely data-driven approach (SSDR-D) based on the use of principal component constructions.
Results
We report the results of studies on a pilot plant-scale continuous direct compression-based tableting line at steady-state in two subsystems. If the process is linear or mildly nonlinear, SSDR-M and SSDR-D give comparable results for the variables estimation and GED. SSDR-M also complies with mass balances and estimate unmeasured variables.
Conclusions
SSDR successfully estimates the true state of the process in presence of gross errors, as long as steady state is maintained and the redundancy requirement is met. Gross errors are also detected while using SSDR-M or SSDR-D. Process monitoring is more reliable while using the SSDR framework.
Similar content being viewed by others
References
U. S. D. of H. and H. S. FDA, Guidance for industry PAT—a framework for innovative pharmaceutical development, manufacturing, and quality assurance, no. September, p. 16; 2004.
Ierapetritou M, Muzzio F, Reklaitis GV. Perspectives on the continuous manufacturing of powder-based pharmaceutical processes. AICHE J. 2016;62(6):1846–62.
Lee SL, O’Connor TF, Yang X, Cruz CN, Chatterjee S, Madurawe RD, et al. Modernizing pharmaceutical manufacturing: from batch to continuous production. J Pharm Innov. 2015;10(3):191–9.
Su Q, Moreno M, Giridhar A, Reklaitis GV, Nagy ZK. A systematic framework for process control design and risk analysis in continuous pharmaceutical solid-dosage manufacturing. In: J. Pharm. Innov, vol. 12; 2017. p. 327–46.
Knopf C. Introduction to data reconciliation and gross error detection. In: Modeling, analysis and optimization of process and energy systems. Hoboken: Wiley; 2011.
Narasimhan S, Jordache C. Data reconciliation and gross error detection. Houston: Gulf Publishing Company; 2000.
Alhaj-Dibo M, Maquin D, Ragot J. Data reconciliation: a robust approach using a contaminated distribution. Control Eng Pract. 2008;16(2):159–70.
Tong H. Studies in data reconciliation using principal component analysis. Ph.D. Thesis. Hamilton: McMaster University; 1995.
Cencic O, Fruhwirth R. A general framework for data reconciliation—part I: linear constraints. Comput Chem Eng. 2015;75:196–208.
Narasimhan S, Bhatt N. Deconstructing principal component analysis using a data reconciliation perspective. Comput Chem Eng. 2015;77:74–84.
Narasimhan S, Shah SL. Model identification and error covariance matrix estimation from noisy data using PCA. Control Eng Pract. 2008;16(1):146–55.
Benqlilou C. Data reconciliation as a framework for chemical processes optimization and control. Doctoral thesis, Universitat Politècnica de Catalunya; 2004.
Diehl M, Bock HG, Schlöder JP, Findeisen R, Nagy Z, Allgöwer F. Real-time optimization and nonlinear model predictive control of processes governed by differential-algebraic equations. J Process Control. 2002;12:577–85.
Liu J, Su Q, Moreno M, Laird C, Nagy Z, Reklaitis G. Robust state estimation of feeding–blending systems in continuous pharmaceutical manufacturing. Chem Eng Res Des. 2018;134:140–53.
Bai S, Thibault J, McLean DD. Dynamic data reconciliation: alternative to Kalman filter. J Process Control. 2006;16(5):485–98.
Haseltine EL, Rawlings JB. Critical evaluation of extended Kalman filtering and moving-horizon estimation. Ind Eng Chem Res. 2005;44(8):2451–60.
Crowe CM, Campos YAG, Hrymak A. Reconciliation of process flow rates by matrix projection. Part I: linear case. AICHE J. 1983;29(6):881–8.
Arora N, Biegler LT. Redescending estimators for data reconciliation and parameter estimation. Comput Chem Eng. 2001;25(11):1585–99.
Almaya A, De Belder L, Meyer R, Nagapudi K, Lin H-RH, Leavesley I, et al. Control strategies for drug product continuous direct compression—state of control, product collection strategies, and startup/shutdown operations for the production of clinical trial materials and commercial products. J Pharm Sci. 2017;106(4):930–43.
Jackson JE. A user’s guide to principal components. New York: Wiley; 1991.
Tong H, Crowe CM. Detection of gross erros in data reconciliation by principal component analysis. AICHE J. 1995;41(7):1712–22.
MacGregor J, Cinar A. Monitoring, fault diagnosis, fault-tolerant control and optimization: data driven methods. Comput Chem Eng. 2012;47:111–20.
Wold S, Sjöström M, Eriksson L. PLS-regression: a basic tool of chemometrics. Chemom Intell Lab Syst. 2001;58(2):109–30.
Kourti T, MacGregor JFJF. Process analysis, monitoring and diagnosis, using multivariate projection methods. Chemom Intell Lab Syst. 1995;28(1):3–21.
Imtiaz SA, Shah SL, Narasimhan S. Missing data treatment using iterative PCA and data reconciliation. IFAC Proc Vol. 2004;37(9):197–202.
Nelson PRC, Taylor PA, MacGregor JF. Missing data methods in PCA and PLS: score calculations with incomplete observations. Chemom Intell Lab Syst. 1996;35(1):45–65.
Folch-Fortuny A, Arteaga F, Ferrer A. PCA model building with missing data: new proposals and a comparative study. Chemom Intell Lab Syst. 2015;146:77–88.
Kim I-W, Kang MS, Park S, Edgar TF. Robust data reconciliation and gross error detection: the modified MIMT using NLP. Comput Chem Eng. 1997;21(7):775–82.
Gupta A, Giridhar A, Venkatasubramanian V, Reklaitis GV. Intelligent alarm management applied to continuous pharmaceutical tablet manufacturing: an integrated approach. Ind Eng Chem Res. 2013;52(35):12357–68.
MacGregor JF, Yu H, García Muñoz S, Flores-Cerrillo J. Data-based latent variable methods for process analysis, monitoring and control. Comput Chem Eng. 2005;29(6):1217–23.
Hotelling H. The generalization of Student’s ratio. Ann Math Stat. 1931;2(3):360–78.
Miller P, Swanson RE, Heckler CE. Contribution plots: a missing link in multivariate quality control. Appl Math Comput Sci. 1998;8:775–92.
Le Roux GAC, Santoro BF, Sotelo FF, Teissier M, Joulia X. Improving steady-state identification. Comput Aided Chem Eng. 2008;25:459–64.
Bagajewicz MJ, Chmielewski DJ, Tanth DN. Smart process plants: software and hardware solutions for accurate data and profitable operations, 1st edn. New York: McGraw-Hill Education; 2010.
Cao S, Rhinehart RR. An efficient method for on-line identification of steady state. J Process Control. 1995;5(6):363–74.
Nomikos P, MacGregor JF. Multivariate SPC charts for monitoring batch processes. Technometrics. 1995;37(1):41–59.
Parikh DM. Handbook of pharmaceutical granulation technology, 3rd edn. In: Drugs and the pharmaceutical sciences, volume 198. New York: Informa Healthcare USA; 2010.
Austin J, Gupta A, McDonnell R, Reklaitis GV, Harris MT. A novel microwave sensor to determine particulate blend composition on-line. Anal Chim Acta. 2014;819:82–93.
Vanarase AU, Alcalà M, Jerez Rozo JI, Muzzio FJ, Romañach RJ. Real-time monitoring of drug concentration in a continuous powder mixing process using NIR spectroscopy. Chem Eng Sci. 2010;65(21):5728–33.
García-Muñoz S. Phi MATLAB toolbox. Personal communication; 2015.
Ganesh S, Troscinski R, Schmall N, Lim J, Nagy Z, Reklaitis G. Application of X-ray sensors for in-line and noninvasive monitoring of mass flow rate in continuous tablet manufacturing. J Pharm Sci. 2017;106:3591–603.
Vanarase AU, Muzzio FJ. Effect of operating conditions and design parameters in a continuous powder mixer. Powder Technol. 2011;208(1):26–36.
Acknowledgments
The authors will like to thank, Benjamin Rentz for his collaboration in the experimental procedure, Joonyoung Yu and Dr. Marcial Gonzalez for the design of the CDI-NIR holder, Dr. Salvador Garcia-Munoz for providing the “phi” latent variable modeling toolbox (version 1.7), Sudarshan Ganesh for his feedback in the design of experiments, and Yash D. Shah and James Wiesler for their continuous support in the project.
Funding
Funding for this project was made possible, in part, by the Food and Drug Administration through grant U01FD005535; views expressed by authors do not necessarily reflect the official policies of the Department of Health and Human Services nor does any mention of trade names, commercial practices, or organization imply endorsement by the US Government. This work was also supported in part by the National Science Foundation under grant EEC-0540855 through the Engineering Research Center for Structure Organic Particulate Systems.
Author information
Authors and Affiliations
Corresponding authors
Appendix
Appendix
Experimental Procedure and Information Flow
The experimental procedures for executing runs for cases E1 and E2 are quite straightforward. A run is initiated by turning on feeders 1 and 2, which take 1–2 min to arrive at steady state as defined by the conditions shown in Table 1. With the blender operating at 200 rpm, an additional period 2–3 min was allowed from the moment the feeders reached steady state [42] before reconciliation was initiated, corresponding to reaching a Q at or below the values in Table 1. In case E1, the process set points are as shown in Table 1. The measurements are collected and sent to MATLAB-2016a-64bit for execution, as shown in the system configuration (Appendix Fig. 10). During the runs, all variables are saved in the Emerson DeltaV v.13.3 process historian, operating using two Dell Precision T1700 64-bit computers with Windows 2008 RL2, configured as five virtual computers. The first server houses App Station (AS), ProPlus (PP), and Terminal Server 1 (TS1). The second server houses Engineering Station (ENG) and Terminal Server 2 (TS2). As configured, each virtual computer has a specific function: the ENG station is used for adding equipment and troubleshooting issues. Computer AS is used when other applications, such as MATLAB, need to communicate with the DeltaV system. The PP computer is used to add any figures in the main template for the operation of the pilot plant, process historian view diagrams, and to add or modify the control modules in DeltaV. Finally, TS1 and TS2 are used to operate the process.
The codes for the SSDR-M and the SSDR-D frameworks are implemented in MATLAB 2015b 32-bit and installed in the AS computer. Since MATLAB serves as a client, by using the OPC toolbox, the code reads the measured flows and composition from the DeltaV historian, solves the optimization problem, and writes the information directly into the DeltaV control module “SSDR.” All control modules and process historian view diagrams were created on the PP computer. For these experiments, the system was operated from the TS1 computer.
Another, physical computer: Dell Inspiron Windows 10 64-bit laptop using Intel Core™ i7 and 8 GB RAM called “PAT,” is used to connect to the necessary PAT tools, such as the CDI-NIR or the Mettler Toledo balance, and to link the required information to the other computers. All composition and flow measurements are in the DeltaV historian. However, it has to be noted, that of these four subsystem measurements, only the feeder flow values are directly sent to the DeltaV system. For the composition measurements, the CDI-NIR is connected via USB port to the “PAT” computer. By using the CDI-NIR software, called Spec32, the spectra measurement is collected as a “.spc file.” This file is generated every second and read by the multivariate statistical methods code. The code, which is in MATLAB 2016a 64-bit, executes the PLS model, created in “phi” software package and loaded as a “.mat” file. This software was acquired through personal communication with Dr. Garcia-Munoz. The results are sent to KEPServerEX v6. This server, provided by Kepware, sends the information to LinkMaster v3, which is connected to the DeltaV system. In parallel, the mass flow MATLAB code is executed using the “parpool” function. This code calls the information from the APW-Link™ software, which is connected to the Mettler Toledo balance through a COM port. It is important to mention that the acetaminophen composition and mass flow measurements are computed using an averaging time window of 10 s.
Rights and permissions
About this article
Cite this article
Moreno, M., Liu, J., Su, Q. et al. Steady-State Data Reconciliation Framework for a Direct Continuous Tableting Line. J Pharm Innov 14, 221–238 (2019). https://doi.org/10.1007/s12247-018-9354-9
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12247-018-9354-9