The application of multivariate statistical projection based techniques has been recognized as one approach to contributing to an increased understanding of process behaviour. The key methodologies have included multi-way principal component analysis (PCA), multi-way partial least squares (PLS) and batch observation level analysis. Batch processes typically exhibit nonlinear, time variant behaviour and these characteristics challenge the aforementioned techniques. To address these challenges, dynamic PLS has been proposed to capture the process dynamics. Likewise approaches to removing the process nonlinearities have included the removal of the mean trajectory and the application of nonlinear PLS. An alternative approach is described whereby the batch trajectories are sub-divided into operating regions with a linear/linear dynamic model being fitted to each region. These individual models are spliced together to provide an overall nonlinear global model. Such a structure provides the potential for an alternative approach to batch process performance monitoring. In the paper a number of techniques are considered for developing the local model, including multi-way PLS and dynamic multi-way PLS. Utilising the most promising set of results from a simulation study of a batch process, the local model comprising individual linear dynamic PLS models was benchmarked against global nonlinear dynamic PLS using data from an industrial batch fermentation process. In conclusion the results for the local operating region techniques were comparable to the global model in terms of the residual sum of squares but for the global model structure was evident in the residuals. Consequently, the local modelling approach is statistically more robust.
L'application de techniques basees sur la projection statistique multivariee est reconnue comme etant une approche qui contribue a une meilleure comprehension du comportement des procedes. Les methodologies cles incluent l'analyse des composantes principales (PCA) a plusieurs criteres de classification, les moindres carres partiels (PLS) a plusieurs criteres de classification et l'analyse des niveaux d'observation discontinus. Les procedes discontinus presentent typiquement un comportement non lineaire et variable dans le temps et ces caracteristiques mettent au defi les techniques mentionnees ci-dessus. Devant ces defis, la methode PLS dynamique est proposee pour saisir la dynamique des procedes. Des approches semblables pour supprimer la non linearite des procedes incluent le retrait de la trajectoire principale et l'application des PLS non lineaires. On decrit une autre approche ou les trajectoires discontinues sont subdivisees en regions operatoires avec un modele dynamique lineaire/lineaire adapte a chaque region. Ces modeles individuels sont raccordes pour obtenir un modele non lineaire global. Une telle structure presente un potentiel pour une approche differente du suivi des performances des procedes discontinus. Dans cet article, plusieurs techniques sont considerees pour la mise au point du modele local, incluant les PLS a plusieurs criteres de classification et les PLS a plusieurs criteres de classification dynamique. En utilisant la serie de resultats les plus prometteurs d'une etude de simulation d'un procede discontinu, le modele local comprenant les modeles de PLS dynamiques lineaires individuels a ete compare a la methode de PLS non lineaires dynamique globale utilisant des donnees d'un procede de fermentation discontinu industriel. En conclusion, les resultats pour les techniques des regions operatoires locales sont comparables au modele global en termes de somme des carres des residus mais pour le modele global, la presence d'une structure dans les residus est evidente. En consequence, l'approche de modelisation locale est statistiquement plus robuste.
[1]
S. Wold,et al.
Nonlinear PLS modeling
,
1989
.
[2]
Svante Wold,et al.
Modelling and diagnostics of batch processes and analogous kinetic experiments
,
1998
.
[3]
J. Macgregor,et al.
Monitoring batch processes using multiway principal component analysis
,
1994
.
[4]
W. Harmon Ray,et al.
Dynamic PLS modelling for process control
,
1993
.
[5]
John F. MacGregor,et al.
Multivariate SPC charts for monitoring batch processes
,
1995
.
[6]
B. Kowalski,et al.
Partial least-squares regression: a tutorial
,
1986
.
[7]
A. J. Morris,et al.
Non-linear dynamic projection to latent structures modelling
,
2000
.
[8]
Roderick Murray-Smith,et al.
Multiple Model Approaches to Modelling and Control
,
1997
.
[9]
E. Martin,et al.
Non-linear projection to latent structures revisited: the quadratic PLS algorithm
,
1999
.
[10]
S. Wold.
Cross-Validatory Estimation of the Number of Components in Factor and Principal Components Models
,
1978
.
[11]
Tor Arne Johansen,et al.
Non-linear predictive control using local models-applied to a batch fermentation process
,
1995
.
[12]
B Lennox,et al.
Process monitoring of an industrial fed-batch fermentation.
,
2001,
Biotechnology and bioengineering.
[13]
P. Ghosh,et al.
Kinetic analysis of gluconic acid production by Pseudomonas ovalis
,
2007
.