Two methods of selecting smoothing splines applied to fermentation process data

Thornhill, Nina F.; Manela, Mauro; Campbell, J. A.; Stone, Karl M.

doi:10.1002/aic.690400414

Cited by 6 publications

(3 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All spectra were baseline corrected using a prescanned reference spectrum (distilled water). The transmission spectra were then smoothed by a spline routine (Thornhill et al, 1994). Because the spectral data were smoothed and baseline corrected, their reproducibility was high; therefore, errors in the spectral data were considered negligible.…”

Section: Multivariate Analysismentioning

confidence: 99%

Near-infrared spectroscopy for bioprocess monitoring and control

Yeung

Hoare

Thornhill

et al. 1999

Biotechnol. Bioeng.

Self Cite

View full text Add to dashboard Cite

This article describes the calibration of a spectroscopic scanning instrument for the measurement of selected contaminants in a complex biological process stream. Its use is for the monitoring of a process in which contaminants are to be removed selectively by flocculation from yeast cell homogenate. The main contaminants are cell debris, protein, and RNA. A low‐cost instrument has been developed for sensitivity in the region of the NIR spectrum (from 1900 to 2500 nm) where preliminary work found NIR signatures from cell debris, protein, and RNA. Calibration models have been derived using a multivariate method for concentrations of these contaminants, such as would be found after the flocculation process. Two strategies were compared for calibrating the NIR instrument. In one case, samples were prepared by adding materials representative of the contaminants to clarified yeast homogenate so the contaminant levels were well known but outside the range of interest. In the other case, where samples were like those from the process stream after flocculation and floc removal, there was uncertainty of analysis of contaminant level, but the calibration was in the range of interest. Calibration using process stream samples gave results close to those derived from traditional assays. When the calibration models were used to predict the contaminant concentrations in previously unseen samples, the correlation coefficients between measurements and predictions were above 90% in all cases but one. The prediction errors were similar to the errors in the traditional assays. © 1999 John Wiley & Sons, Inc. Biotechnol Bioeng 63: 684–693, 1999.

show abstract

Section: Multivariate Analysismentioning

confidence: 99%

Near-infrared spectroscopy for bioprocess monitoring and control

Yeung

Hoare

Thornhill

et al. 1999

Biotechnol. Bioeng.

Self Cite

View full text Add to dashboard Cite

show abstract

“…All spectra were baseline corrected using a pre-scanned reference spectrum (distilled water). The transmission spectra were then smoothed by a spline routine (Thornhill et al, 1994). Since the spectral data are smoothed and baseline corrected their reproducibility is high; errors in the spectral data were therefore considered negligible.…”

Section: Multivariate Analysismentioning

confidence: 99%

Near‐infrared spectroscopy for bioprocess monitoring and control

Yeung¹,

Hoare²,

Thornhill³

et al. 1999

Biotechnol. Bioeng.

Self Cite

View full text Add to dashboard Cite

This paper describes the calibration of a spectroscopic scanning instrument for the measurement of selected contaminants in a complex biological process stream. Its use is for the monitoring of a process in which contaminants are to be removed selectively by flocculation from yeast cell homogenate. The main contaminants are cell debris, protein and RNA. A low cost instrument has been developed for sensitivity in the region of the NIR spectrum (from 1900 to 2500 nm) where preliminary work found NIR signatures from cell debris, protein and RNA. Calibration models have been derived using a multivariate method for concentrations of these contaminants such as would be found after the flocculation process. Two strategies were compared for calibrating the NIR instrument. In one case samples were prepared by adding materials representative of the contaminants to clarified yeast homogenate so the contaminant levels were well known but outside the range of interest. In the other, where samples were like those from the process stream after flocculation and floc removal there was uncertainty of analysis of contaminant level but the calibration was in the range of interest. Calibration using process stream samples gave results close to those derived from traditional assays. When the calibration models were used to predict the contaminant concentrations in previously unseen samples the correlation coefficients between measurements and predictions were above 90% in all cases but one. The prediction errors were similar to the errors in the traditional assays. 2

show abstract

“…In bioprocess engineering, various kinds of kinetic models have been proposed to quantitatively describe the dynamic behaviors of biological reaction systems. Several conventional methods, such as the graphical method and the gradient-based nonlinear optimization methods, have been employed as estimation techniques to obtain the kinetic parameters. − The graphical method can be applied only to those problems that can be converted to linear regression problems. Basically, a gradient-based nonlinear optimization method does not have such a restriction, but it requires gradient information about the error function with respect to the parameter estimates and often gets stuck at local minima.…”

Section: Introductionmentioning

confidence: 99%

Hybrid Differential Evolution for Problems of Kinetic Parameter Estimation and Dynamic Optimization of an Ethanol Fermentation Process

Wang

Jang

2001

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Hybrid differential evolution (HDE) is applied to estimate the kinetic model parameters of batch fermentation for ethanol and glycerol production using Saccharomyces diastaticus LORRE 316. In this study, we considered two kinetic models for describing the dynamic behaviors of S. diastaticus LORRE 316. In the parameter estimation problem, we used the worst observed error for all experiments as an objective function so that the parameter estimation problem becomes the min-max estimation problem. Several numerical methods have been employed to solve the min-max estimation problem for comparison. HDE could use a small population size to obtain a more satisfied solution as compared from these computations. To validate the two kinetic models, we respectively used the two kinetic models to determine the optimal feed rates for the fedbatch optimization problem. Because the fedbatch optimization problem was a constrained dynamic optimization problem, we introduced the HDE with a multiplier updating method including adaptive penalty parameters to obtain the feasible feed rates. Such feed rates are then applied for fedbatch experiments in a 5 L fermenter for model validation. From the comparison of batch and fedbatch experiments, we observed that the proposed kinetic model was more adequate than Monod's model.

show abstract

Two methods of selecting smoothing splines applied to fermentation process data

Cited by 6 publications

References 15 publications

Near-infrared spectroscopy for bioprocess monitoring and control

Near-infrared spectroscopy for bioprocess monitoring and control

Near‐infrared spectroscopy for bioprocess monitoring and control

Hybrid Differential Evolution for Problems of Kinetic Parameter Estimation and Dynamic Optimization of an Ethanol Fermentation Process

Contact Info

Product

Resources

About