Previous Topic Back Forward Next Topic
Print Page Frank Dieterle
Ph. D. ThesisPh. D. Thesis 3. Theory  Quantification of the Refrigerants R22 and R134a: Part I3. Theory Quantification of the Refrigerants R22 and R134a: Part I 3.6. Conclusions3.6. Conclusions
About Me
Ph. D. Thesis
  Table of Contents
  1. Introduction
  2. Theory Fundamentals of the Multivariate Data Analysis
  3. Theory Quantification of the Refrigerants R22 and R134a: Part I
    3.1. Experimental
    3.2. Single Analytes
    3.3. Sensitivities
    3.4. Calibrations of the Mixtures
    3.5. Variable Selection by Brute Force
    3.6. Conclusions
  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
Downloads and Links
Site Map
Print this Page Print this Page

3.6.   Conclusions

In this example, 6 polymers were investigated for application in a sensor system for the quantification of the refrigerants R22 and R134a in mixtures. By the use of all polymers an accurate quantification of both analytes could be performed. Additionally, 2 polymers were selected on the basis of the sensitivity patterns for the application in a small low-cost 4l RIfS setup. By the use of these 2 polymers, the 4l RIfS setup could quantify R22 in the presence of R134a quite well but not vice versa.

It was shown that the best selection of 2 sensors is the combination of 1 microporous polymer with 1 polar polymer enabling a discrimination on the basis of two interaction principles. The two interaction principles show the most different sensitivity patterns for the two analytes, which is the common selection criterion of sensor coatings for an analytical problem.

The polymers used in this example show two different types of sorption, the specific Langmuir sorption and the unspecific Henry sorption. For the classical feature extraction, which uses the height of the raw signal after a definite time of exposure to analyte, the unspecific Henry sorption is advantageous as the immediate sensor responses allow very fast measurements. Yet, a drawback of this common feature extraction is the extraction of only one single variable per sensor. This limits the number of analytes to be quantified to the number of sensors in the ideal case. This means that a 2-sensor setup can be calibrated for 2 analytes or contaminants and the 6 sensor array setup can be calibrated for up to 6 analytes and contaminants. Contaminants in the samples, which do not sorb into the sensitive layers, do not interfere the determination of the concentration of both analytes and can be ignored during the calibration. Yet, contaminants, which sorb into the polymer layers, bias the predictions of the analyte concentrations unless they can be considered during the calibration process as additional analytes in combination with additional sensors.

In chapter 5, the principle of a time-resolved feature extraction will be introduced. Thereby the kinetics of sorption and desorption of analytes is exploited allowing the extraction of a virtually unlimited number of variables per sensor. Thus, the number of analytes, which can be quantified per sensor, is limited only by similarities of kinetics and is not fixed by the device. The time-resolved approach removes the limitation of the common sensor array approach of being able to quantify simultaneously only a maximal theoretical number of analytes. The time-resolved approach also changes dramatically the search and the rating of polymers, which might be suitable for a specific analytical problem. The static measurements with a single feature extraction need sensitivity patterns as different as possible, whereas the kinetic feature extraction needs different shapes of the sensor responses during sorption or desorption.

The selection of the best combination of polymers based on the sensitivity pattern in this example could be verified by a brute force variable selection approach. Yet, the time-resolved measurements introduced in this work generate many variables putting new challenges to calibration methods and variable selection techniques, as a brute force variable selection is rendered impossible. Therefore, the introduction of new calibration techniques combined with variable selection methods are one of the focuses of this work.

Page 65 © Frank Dieterle, 03.03.2019 Navigation