Martin Holmberg

Drift correction for gas sensors using multivariate methods

Drift is one of the most serious impairments afflicting gas sensors. It can be seen as a gradual change in the sensor response over a long period of time when the external conditions are constant. This paper presents a new simple drift counteraction method based on PCA and PLS. The basic idea is to remove the drift direction component from the measurements. The direction of the drift, p, is calculated from measurements for a reference gas. Projecting the sample gas measurements on this vector gives the score vector t. The drift component tp T can then be removed from the sample gas data, which we call component correction (CC). The method is tested on a data set based on a reduced factorial design with four gases and a concentration gradient of hydrogen. It is found that the method works efficiently for both cases. Copyright

Drift counteraction for an electronic nose

Abstract Most gas sensors do not give stable responses over a long period of time. This drift causes electronic noses to classify gases wrongly after some period of time. In this paper this problem has been studied, and two methods for counteracting drift are proposed. One is based on reference patterns that are updated when new measurements are made, and the other uses models known from system identification theory. The two methods are suited for different experimental situations, and both give a better result than a back-propagation neutral network classifier when they are used under the right conditions.

Bacteria classification based on feature extraction from sensor data

Data evaluation and classification have been made on measurements by an electronic nose on the headspace of samples of different types of bacteria growing on petri dishes. The chosen groups were: Escherichia coli, Enterococcus sp., Proteus mirabilis, Pseudomonas aeruginosa, and Staphylococcus saprophytica. An approximation of the response curve by time was made and the parameters in the curve fit were taken as important features of the data set. A classification tree was used to extract the most important features. These features were then used in an artificial neural network for classification. Using the ‘leave-one-out’ method for validating the model, a classification rate of 76% was obtained.

Wavelet transform of electronic tongue data

Abstract A measurement in a multi-sensor system is characterized by a large array of numbers (a vector or a matrix), sometimes several thousands. In order to increase the interpretability of the measurements, decrease the calculation demand on the computer, and/or to reduce noise, an alternative, more compact, representation of the measurement can be made which describes the important features of the measurement well but with a much smaller vector. The purpose of this paper is to show that for a particular wet-chemical sensor system (pulsed voltammetry, also called an electronic tongue) the data compression can be made using a wavelet transform together with different wavelet selection algorithms for different purposes. The resulting compressed data can also be used for easy interpretation of the measurements and to give hints for improvements or simplifications of the measurement procedure. Two different criteria for selection of wavelet coefficients have been used, variance and discriminance, in two different cases. The variance criterion was used when variations of any kind in the raw data was studied during monitoring of water in drinking water production plant. In this case, the number of variables was reduced with a factor of 18, without loosing relevant information. In the other case, the focus was to separate different microorganisms, therefore, the discriminance selection criterion was successfully used. The number of variables was reduced by a factor of 144, this smaller data set captured the important information for separating the microorganisms, which led to better classification of the test set.

Development of a ChemFET sensor with molecular films of porphyrins as sensitive layer

The interaction of chemical species with molecular films of porphyrins causes variations of the work function of the film itself, as it has been recently demonstrated by using the Kelvin probe technique. This characteristic makes porphyrins films suitable to be used as sensitive layers in ChemFET sensors. In this paper, we present a preliminary report about the fabrication and testing of such gas sensitive devices. The technological solutions towards an optimised device are also illustrated and discussed.

Variable reduction on electronic tongue data

Reduction of the number of variables in data from a so-called electronic tongue contributes to simpler model calculations and less storage requirements. In this study, we have developed a model for this purpose. This model describes the response from the electrodes in the electronic tongue with two exponential functions plus a constant term, i(t) = k + kf e-ta + kc e-ts, where t is the time. From the model, five parameters which describe the signal are extracted. These parameters can be used as inputs instead of the original signal to any multivariate algorithm. The results show that the variables obtained are at least as good as the original data to separate between different classes, even though the number of parameters has been reduced between 80 and 199 times.

A new method for the detection of hydrogen spillover

Abstract In this work, a new method is used to study the spillover of hydrogen from a catalytic metal onto an insulator. The method makes use of the so-called scanning light pulse technique (SLPT) and is applied to palladium gate silicon dioxide-silicon capacitors with the substrate being p-doped. It is concluded that there exist two types of spillover species. The spiltover hydrogen changes the conductance of the SiO2 surface, making it possible to charge the surface. The results for p-type silicon suggest that the induced surface charge must be positive. The results of SLPT and inversion capacitance measurements obtained on similar samples during hydrogen exposure are compared.

Evaluation of on-line flue gas measurements by MISiCFET and metal-oxide sensors in boilers

Metal insulator silicon carbide field-effect transistor sensors, metal-oxide sensors, and a linear Lambda sensor in an electronic nose was used to measure on-line in hot flue gases from a boiler. Flue gas from a 100-MW pellets-fuelled boiler has been used to feed the experimental setup. Several reference instruments, which measure the flue gases in parallel to the sensor array, are connected to the electronic nose. Data was collected during six weeks and then evaluated. Using principal component analysis as the data evaluation method, different operating modes for the boiler have been identified in the data set. The different modes could be described in terms of high or low O/sub 2/ and CO concentration. Furthermore, we have shown that it seems possible to use a sensor array to determine the operating mode of the boiler and, by partial least-squares models, measure the CO concentration when the boiler operates in its optimum mode.

Chaotic chemical sensing

Examined the behavior of a chaotic circuit where one of the components has been replaced by a chemical sensor. The response of the sensor is manifested as a change in the attractor of the circuit. Furthermore, with a proper operating point, a small response of the sensor causes a change in the attractor of the circuit within a time much shorter than the normally defined response time of the sensor. This technique unites sensors and data processing as one combined unit in a unique fashion.

The characteristics and utility of SiC-FE gas sensors for control of combustion in domestic heating systems [MISFET sensors]

The possible utility of MISiCFET gas sensors in the application of combustion control in small-scale boilers has been tested and compared to commercially available resistive-type MOS sensors. The results suggest that by using the signals from one or more MISiCFET sensors, together with the measured temperature of the furnace, it seems possible to provide a rough picture of the state of combustion applicable to a control scheme in order to reduce emissions and increase the power to fuel economy.