Perception based sensor systems in environmental applications

This thesis describes the work on the development of a new electronic tongue for water quality assessment. The aim of this work is to build a complete system including a sensor and associated data analysis methods which is able to detect quality changes in drinking water and warn the user if the water is not suitable for drinking. Therefore, the main design issues were to build a portable, robust electronic tongue which provides a fast estimate of the water quality rather than identify the type of the contamination. In this sense, the system has a simple interface resembling traffic light signals. The system responds with green light for good water quality, red for bad quality and yellow if the result is uncertain. The electronic tongue sensor is based on voltammetry, which gives a relatively simple and robust system. It is suitable for the considered applications, since it is able to measure small changes in the chemical and micro-organic content in water. These properties are very important for our goal to develop an electronic tongue for water quality assessment. During the development process, the sensor system was modified in order to improve the system’s mobility and robustness. Moreover, the long-term stability and the extent of the sensor drift were evaluated for continuous use. The response time for the sensor is relatively short, but it generates large amounts of data. Therefore, there is a need for pre-processing and feature extraction methods to deal with the complexity of data. This is done with the help of multivariate analysis method. The quality assessment is done with different classification algorithms. We implemented fuzzy clustering and self-organizing maps. In this thesis, two real applications are also considered, where the measurement is taking place in streaming water. In the first case, the electronic tongue could be placed directly at a water tap where random measurements occur. In the second application continuous measurements are performed and information about the sensor drift is used in the data analysis. The evaluation confirmed that the improved sensor system fulfills the requirements for on-line measurements of water quality.