Presentation by: Raf Roelant (Laboratory for Chemical Technology, Ghent University)
This presentation demonstrates that standard NLSQ regression is inadequate to fit kinetic parameters from transient experiments such as experiments carried out in a batch-reactor and pulse-response experiments (e.g. in a TAP reactor) since there is strong correlation between the subsequent data points. This problem can be overcome using second-order statistical regression (SOSR). After integration this requires PCA (principal component analysis) to remove the correlation between the data, followed by a suitable rescaling to achieve proper heteroskedasticness. The improvement obtained with the SOSR approach is demonstrated using real experimental TAP data.
- Download the presentation: Second-order statistical regression of kinetic time series - Downloaded 342 times