(92f) Bio-Reactor Monitoring with Multiway-Pca and Model Based-Pca

With the rapid growth of biotechnology and the PAT (Process
Analytical Technology) initiative in the pharmaceutical industry, more
attention is being focused on monitoring bioreactor production to create a safe
production environment and obtain a high-quality product. However, a bioreactor
is difficult to monitor mainly due to the following reasons: 1) The process is
always batch or semi-batch rather than continuous. 2) The dynamic behavior is
highly nonlinear and rarely is a high fidelity model available to describe the
dynamic behavior of the process. 3) The micro-organisms can be affected when
operating conditions change unpredictably.

Typically, process monitoring methods can be divided into
data-driven and knowledge-driven techniques. multiway-PCA developed by Nomikos
and MacGregor [1-3] is the first and most widely used data-driven method in
batch process monitoring. The basic idea of MPCA is unfolding the three
dimensional batch data to two dimensions so as to perform PCA on the data
matrix. Based on this pioneering work, more efforts have been made to make the
technique more powerful and applicable including: 1) New data unfolding methods
(batch-wise[1]; variable-wise[4]; hybrid-wise[5]); and 2) Batch data
synchronization (variable indicator[2]; dynamic time warping[6]; correlation
optimized warping[7]). Besides data-driven methods, Model Based-PCA (MB-PCA)[8]
is based on the fundamental knowledge of process behavior and can be
successfully used in batch and continuous processes. If the process model is
accurate, then the data unfolding and synchronization steps can be avoided by
applying the MB-PCA method.

In this work, we focus on finding an efficient and effective
way to perform PCA on a penicillin fermenter simulation model. The detailed
fermenter model was developed by G. Birol et al.[9]. MB-PCA method is also
applied and compared with MPCA with DTW. The effect of the coupling of
manipulated and controlled variables on PCA-based fault detection is estimated.

References:

1.
Nomikos, P. and J.F. MacGregor, Multivariate SPC Charts for Monitoring Batch
Processes. Technometrics, 1995. 37: p. 41-59.

2.
Nomikos, P. and J.F. MacGregor, Monitoring of batch processes using
multi-way principal component analysis. AIChE Journal, 1994. 40: p.
1361-1375.

3.
Nomikos, P. and J.F. MacGregor, Multi-way partial least squares in
monitoring batch process. Chemometrics and Intelligent Laboratory Systems,
1995. 30: p. 97-108.

4.
Wold, S., et al., Modelling and diagnostics of batch processes and analogous
kinetic experiments. Chemometrics and Intelligent Laboratory Systems, 1998.
44: p. 331-340.

5. Lee, J.-m., C.K. Yoo, and I.-B. Lee, Enhanced process
monitoring of fed-batch penicillin cultivation using time-varying and
multivariate statistical analysis. Journal of Biotechnology, 2004. 110:
p. 119-136.

6.
Kassidas, A., J.F. MacGregor, and P.A. Taylor, Synchronization of Batch
Trajectories Using Dynamic Time Warping. AIChE Journal, 1998. 44(4):
p. 864-875.

7.
Pravdova, V., B. Walczak, and D.L. Massart, A comparison of two algorithms
for warping of analytical signals. Analytica Chimica Acta, 2002. 456:
p. 77-92.

8.
Wachs, A. and D.R. Lewin. Process Monitoring Using Model-based PCA. in Proc.
IFAC Symp. on Dynamics and Control of Process Systems. 1998. Corfu.

9. Birol G., C. Undey, and A. Cinar, A modular simulation
package for fed-batch fermentation: penicillin production. Computers and
Chemical Engineering, 2002, 26, p. 1553-1565.