Isotherms and Mass Transfer Resistance by Neural Networks

Identification of Isotherms and Mass Transfer Resistance by Neural Networks 

Schlinge et al. (2011) presented an alternative method that requires a lot of preliminary computer simulations but estimates isotherm parameters as well as mass transfer coefficients on the basis of only a few experiments and very small amounts of samples in short time. The basic idea is to simulate HPLC chromatograms for given isotherm equations and to vary the isotherm parameters vdthin the range of practical applications. The calculated data are then used to train and validate neural 

After having developed the NN for certain types of isotherms the estimation of the isotherm data for a new chromatographic system is done in following steps  

It has to be kept in mind that the development of the NN and especially the number of computer simulations increase exponentially with the bandwidth and the number of parameters, but this costs computer time and not additional man power. For some parameters such as axial dispersion, particle diameter and porosity mean values should be chosen and kept constant. The resulting deviations should be acceptable if they are balanced with the low experimental effort and small amount of samples necessary. 

As an example for the efficiency of the NN method, Schlinge et al. (2011) published a comparison between simulated and measured chromatograms for a racemic mixture. The NN were validated for a column of 4 x 300 mm and a flow rate of 

---