Applying Mechanistic Models in Bioprocess Development - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 151

Fig. 4 Comparison of model predictions versus experimental data collected for cultivation 1
(black line model prediction, black circles experimental data) and cultivation 2 (blue dashed line
model prediction, blue stars, experimental data)
(HPLC), it is significantly higher for dry weight measurements, which are less
reliable, especially for low biomass concentrations (too large sample volumes
would be required for increasing accuracy). Additionally, at the end of the fer-
mentation, the biomass dry weight may include a fraction of nonviable and/or
dormant cells.
2.3.1 Confidence Intervals for Estimated Parameters
The estimated parameter values as such only have limited value if they are not
presented in combination with a measure of the degree of confidence that one can
have in them. Therefore, the confidence intervals for each of the parameters are
defined based on the covariance matrix and Student t-probability distribution. The
covariance matrix is calculated using the residuals between model predictions and
the standard deviations of the experimental measurements (further details are
provided by Sin et al. [ 14 ]). An experimental error of 5 % was assumed for
glucose and ethanol measurement by high-performance liquid chromatography
(HPLC), as well as for the oxygen measurements using a gas analyzer for deter-
mining the composition of the exhaust gas, and a 20 % error for the determination
of the cell dry weight. The confidence intervals at (1 - a) confidence level were