From dynamic response surface models to the identification of the reaction stoichiometry in a complex pharmaceutical case study. Issue 4 (12th January 2019)

• Record Type: Journal Article
• Authors: Santos‐Marques, Jacob; Georgakis, Christos; Mustakis, Jason; Hawkins, Joel M.
• Is Part Of: AIChE journal. Volume 65:Issue 4(2019)
• Page Start: 1173
• Page End: 1185
• Publication Date: 2019-01-12
• Subjects: design of experiments -- data‐driven modeling -- batch processes; Chemical engineering -- Periodicals; Génie chimique -- Périodiques; 660.28
• DOI: 10.1002/aic.16515
• Languages: English
• ISSNs: 0001-1541

Abstract

Abstract : The dynamic response surface methodology (DRSM) (Klebanov and Georgakis, Ind Eng Chem Res. 2016;55(14):4022–4034) was proposed as a generalization of the classical response surface methodology (RSM). The power of DRSM is evaluated here in a blind test against a complex pharmaceutical process containing 10 observed species involved in eight reactions. The obtained DRSM models, one for each species, accurately represented the time‐varying concentrations of 8 of the 10 species. The DRSMs for two intermediate species suffered a bit from oscillatory behavior toward the end of the batch, due to their small concentration values and the polynomial‐based model. These DRSM models greatly facilitated the application of target factor analysis (Bonvin and Rippin, Chem Eng Sci. 1990; 45(12): 3417–3426) correctly identifying six of the eight true reaction stoichiometries, whereas all six false reactions were rejected. In addition, the ability to distinguish true and false reaction stoichiometries was not affected by a less informative design of experiments (DoEs) than the original center composite design (CCD). © 2018 American Institute of Chemical Engineers AIChE J, 65: 1173–1185, 2019