HIPPO: An Iterative Reparametrization Method for Identification and Calibration of Dynamic Bioreactor Models of Complex Processes

Unstructured, dynamic bioreactor models of complex processes usually possess many nonidentifiable, insensitive, or statistically nonsignificant parameters. However, an exhaustive search to find a reduced set of identifiable parameters is computationally demanding. We developed a heuristic iterative...

Full description

Saved in:
Bibliographic Details
Published in:Industrial & engineering chemistry research 2014-12, Vol.53 (48), p.18514-18525
Main Authors: Sánchez, Benjamín J, Soto, Daniela C, Jorquera, Héctor, Gelmi, Claudio A, Pérez-Correa, José R
Format: Article
Language:English
Subjects:
Bioreactors
Dynamics
Fittings
Heuristic
Iterative methods
Mathematical analysis
Mathematical models
Matlab
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Unstructured, dynamic bioreactor models of complex processes usually possess many nonidentifiable, insensitive, or statistically nonsignificant parameters. However, an exhaustive search to find a reduced set of identifiable parameters is computationally demanding. We developed a heuristic iterative procedure for parameter optimization (HIPPO), generic and free of symbolic mathematical manipulations, to help bioprocess engineers to find and estimate a set of significant parameters, thus obtaining reliable models. This methodology includes local sensitivity, significance and identifiability analysis, metaheuristic optimization, and fitting performance assessments. We tested this method with two bioreactor models: a microalgal fed-batch bioreactor (MFB) 12-parameter model and a solid substrate fermentation (SSF) 14-parameter model. Applying this procedure, we were able to find a MFB reparametrization with 4 fitting parameters after 479 iterations and a SSF reparametrization with 6 fitting parameters after 918 iterations. In both cases, all fitting parameters were statistically significant and accurately estimated and models had a better fit than the originally fitted 12- and 14-parameter models. The heuristic was programmed using MATLAB and is freely available to the research community (http://www.systemsbiology.cl/tools/).
ISSN:0888-5885
1520-5045
DOI:10.1021/ie501298b