Previous Topic Back Forward Next Topic
Print Page Dr. Frank Dieterle
Ph. D. ThesisPh. D. Thesis AbstractAbstract
About Me
Ph. D. Thesis
    Abstract in German (Zusammenfassung)
  Table of Contents
  1. Introduction
  2. Theory Fundamentals of the Multivariate Data Analysis
  3. Theory Quantification of the Refrigerants R22 and R134a: Part I
  4. Experiments, Setups and Data Sets
  5. Results Kinetic Measurements
  6. Results Multivariate Calibrations
  7. Results Genetic Algorithm Framework
  8. Results Growing Neural Network Framework
  9. Results All Data Sets
  10. Results Various Aspects of the Frameworks and Measurements
  11. Summary and Outlook
  12. References
  13. Acknowledgements
Research Tutorials
Site Map
Print this Page Print this Page


During the last decade, the application of sensors for the detection and determination of various substances has gained an increasing popularity not only in the field of analytical chemistry but also in our daily life. Most sensor systems such as exhaust gas sensors for automobiles are based on single sensors, which are as selective as possible for the analyte of interest. The problems of interfering cross-reactive analytes and the lack of specific sensors for many analytes have ended up in the development of so-called sensor-arrays. Thereby, several analytes can be simultaneously quantified by the multivariate data analysis of the signal patterns of several cross-reactive sensors. Yet, this approach is also limited since the number of sensors in the array has to exceed the number of cross-reacting analytes.

In this work, a new approach is presented, which allows multi-analyte quantifications on the basis of single-sensor systems. Thereby, differences of interaction kinetics of the analytes and sensor are exploited using time-resolved measurements and time-resolved data analyses. This time-resolved evaluation of sensor signals together with suitable sensor materials combines the sensory principle with the chromatographic principle of separating analytes in space or time. The main objectives of this work can be subsumed into two focuses concerning the measurement principle and the data analysis.

The first focus is the introduction of time-resolved measurements in the field of chemical sensing. In this work the time-resolved measurements are based on the microporous polymer Makrolon as sensitive sensor coating, which allows a kinetic separation of the analytes during the sorption and desorption on the basis of the size of analytes. Multi-analyte determinations using single sensors are successfully performed for three different setups and for many multicomponent mixtures of the low alcohols and the refrigerants R22 and R134a.

The second focus concerns the multivariate data analysis of the data. It is demonstrated that a highest possible scanning rate of the time-resolved sensor responses is desirable resulting in a high number of variables. It is shown that wide-spread data analysis methods cannot cope with the amount of variables and with the nonlinear relationship between the sensor responses and the concentrations of the analytes. Thus, three different algorithms are innovated and optimized in this study to find a calibration with the highest possible generalization ability. These algorithms perform a simultaneous calibration and variable selection exploiting a data set limited in size to a maximum extend. One algorithm is based on many parallel runs of genetic algorithms combined with neural networks, one algorithm bases on many parallel runs of growing neural networks and the third algorithm uses several runs of the growing neural networks in a loop. All three algorithms show by far better calibrations than all common methods of multivariate calibration and than simple non-optimized neural networks for all data sets investigated. Additionally, the variable selection of these algorithms allows an insight into the relationship between the time-resolved sensor responses and the concentrations of the analytes. The variable selections also suggest optimizations in terms of shorter measurements for several data sets. All three algorithms successfully solve the problems of too many variables for too few samples and the problems caused by the nonlinearities present in the data with practically no input needed by the analyst.

Together, both main focuses of this work impressively demonstrate how the combination of an advanced measurement principle and of an intelligent data analysis can improve the results of measurements at reduced hardware costs. Thereby the principle of single-sensor setups or few-sensor setups is not only limited to a size-selective recognition but can be extended to many analyte discriminating principles such as temperature-resolved measurements leaving room for many further investigations.

Page 7 © Dr. Frank Dieterle, 14.08.2006 Navigation