engleski

Strategy for Developing and Optimizing Ion Chromatography Methods by Using Artificial Neural Networks

Optimization in ion chromatography involves the selection of experimental condition for adequate separation and acceptable retention time for each individually samples. But, in chemical laboratories obtaining a balance between resolution and analysis time is not always easy. An efficient optimization method should be employed during the method development process in order to deal with these optimization problems and to reduce method development time. Computers assisted optimization procedures have to be used as an efficient aid for ion chromatography method development. Empirical modeling is a useful technique whereby a global optimum can be located. The use of chemometric protocols like experimental design in combination with artificial neural networks can be extremely beneficial in the systematic construction and optimization of response surfaces. For most simple optimization applications, traditional techniques, such as simplex method, are capable of locating a global optimum efficiently. However, for nonlinear behaviors, like those of ion chromatography, these methods are either time consuming or become trapped in local optima. Another factor which also has to be considered in computer assisted optimization procedures is the speed of the computer algorithm used. It is crucial to produce results (optimal conditions) within reasonable time. The speed of the computer algorithm depends on many variables and has to be tested for each and every particular case. In this work artificial neural networks were used for optimization of ion chromatographic separation. It is shown that selectivity of ion chromatographic methods strongly depends on applied ion chromatographic conditions. The developed optimization model allows manipulating with appearance of the particular peak on the chromatogram and allows improvement of selectivity between particular ions. It is shown that optimized artificial neural networks are very accurate and fast optimizing tool with small amount of experimental data needed to model varied inherent non-linear relationships of retention behaviour with respect to the mobile phase parameters. By using artificial neural network optimization model it is possible both to improve performance characteristics of applied method and to speed up the new method development by reducing unnecessary experimentation.

ion chromatography ; optimization ; artificial neural networks

nije evidentirano