ArticlePDF Available

New Transient Feature for Metal Oxide Gas Sensor Response Processing

December 2012
Procedia Engineering 47:52–55

DOI:10.1016/j.proeng.2012.09.082

License
CC BY-NC-ND 3.0

Authors:

Maryam Siadat

University of Lorraine

Martine Lumbreras

University of Lorraine

This paper presents the performance of metal oxide gas sensor response processing for the concentration detection of an analyte diluted in a neutral atmosphere. In the field of electronic nose, two applications are generally studied: identification of a gaseous atmosphere from other atmospheres, or the determination of the concentration of one gaseous atmosphere. This second application needs more accuracy either in the measurement set-up or in the response analysis. We propose in this study the performance comparison between two traditional features extracted from the sensor response and a new feature corresponding to the maximum (Peak) of the derivative curve of the time sensor response. The performance of this feature to obtain fast odor concentration identifications is discussed and compared to other traditional features.

(a) time-responses of all the sensors to a fixed concentration; (b) responses of one senor to all the used oil concentrations

…

(a) Example of a derivative curve without filtering; (b) Derivative curve of SP-MW0 for all the used pine oil concentrations

…

Evolution of two selected features along with the pine oil concentrations: (a) Vs-V0; (b) Peak

…

No caption available

…

Figures - available via license: Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported

Content may be subject to copyright.

Available via license: CC BY-NC-ND 3.0

Content may be subject to copyright.

Procedia Engineering 47 ( 2012 ) 52 – 55

Sp. z.o.o.

doi: 10.1016/j.proeng.2012.09.082

Proc. Eurosensors XXVI, September 9-12, 2012, Kraków, Poland

New transient feature for metal oxide gas sensor response

processing

M. Siadat*, H. Sambemana, M. Lumbreras

LASC, University of Lorraine,Metz, France

Abstract

This paper presents the performance of metal oxide gas sensor response processing for the concentration

detection of an analyte diluted in a neutral atmosphere. In the field of electronic nose, two applications are

generally studied: identification of a gaseous atmosphere from other atmospheres, or the determination of

the concentration of one gaseous atmosphere. This second application needs more accuracy either in the

measurement set-up or in the response analysis. We propose in this study the performance comparison

between two traditional features extracted from the sensor response and a new feature corresponding to

the maximum (Peak) of the derivative curve of the time sensor response. The performance of this feature

to obtain fast odor concentration identifications is discussed and compared to other traditional features.

Keywords: gas sensor, signal processing; transient feature; concentration detection,; rapid identification

1. Introduction

Electronic noses are intelligent systems that play a constant growing role as general detectors of vapors

in many applications. In these devices, the sensor array plays a major role. Several types of sensors can be

employed, semiconductor metal oxide based sensors (MOX) are often used, due to their qualities: robust,

cheap, and able to react in presence of many organic or inorganic gases. Unfortunately, they are not

selective, but a strategic choice of several non selective sensors can improve the selectivity of the system,

to obtain a good discrimination of gaseous substances. In this way, an appropriate pattern recognition

method must be selected, based on an accurate treatment of the response of all the sensors. The key of a

* Corresponding author. Tel.: +33-387547626; fax: +33-387315232.

E-mail address: siadat@univ-metz.fr.

Available online at www.sciencedirect.com

Sp. z.o.o.

M. Siadat et al. / Procedia Engineering 47 ( 2012 ) 52 – 55

fast and successful recognition is the feature extraction method, which needs to extract robust information

[1-2] of the sensor response curves. This fast recognition can be obtained from the transient part of the

sensor response. In this paper, we propose the performance comparison between the traditional features

and a new feature corresponding to the maximum of the derivative curve, obtained in the first minutes of

the gas exposure and deduced from the time sensor response, after an adequate filtering.

2. Characterization measurement

An array of seven MOX sensors (from Figaro and FIS) was characterized under different gaseous

dilutions from liquid pine essential oil (EO) in synthetic air by using a dynamic flow measurement [3]. To

generate different concentrations of pine EO, a constant inflow of synthetic air was bubbling into the

liquid oil. Then the outlet flow, containing evaporated EO substances, was combined with a flow of pure

synthetic air to have a total constant flow rate (100 ml/min). So the EO concentrations are defined in

terms of percent ratio between the EO outlet flow and the total flow. The resistance variation of each

sensor was collected in terms of a voltage magnitude using a fast sampling rate (2 samples/sec). The

measurement protocol was adjusted and a cycle of 10 minutes exposition of EO volatiles molecules

followed by 20 minutes of synthetic air for recovery process was adopted. This procedure allowed us to

obtain a good stabilization and recovery of all the sensors after each concentration exposure. More than

20 experiments were made for each studied concentration to constitute a consequent data basis.

The characterization measurements show a good sensibility of all the sensors (Figure 1a) to pine oil

vapour even at the lowest concentration. We note a very stable and rapid response of the SP-MW0 FIS

sensor. The other sensors have more or less sensitivity and reactivity (corresponding to the slope of the

transient part). The analysis of the same sensor responses under different concentrations of pine oil,

showed a good evolution of the dynamic and stabilized part of the signal along with the concentration.

The Figure 1b shows this observation for the SP-AQ1 sensor, where we can observe the evolution of the

stabilized value (Vs) but also the evolution of the transient part of the signal response.

Fig. 1. (a) time-responses of all the sensors to a fixed concentration; (b) responses of one senor to all the used oil concentrations

3. Signal processing and results

In electronic nose application usually a set of selected features, which depend to the measurement

protocol, are studied in term of their performance to classify the used substances. To obtain an accurate

classification, the selection of the features and their processing are very important. In this study, we have

selected principally two traditional features from the time-dependent signal, as the response amplitude

named Vs-V

(where Vs is the final value of the sensor response and V

its initial value), and the slope of

the dynamic phase calculated during the first three minutes of the gas exposition (named Slope). We have

observed that the transient part of the sensor response presents significant evolution in the two cases

54 M. Siadat et al. / Procedia Engineering 47 ( 2012 ) 52 – 55

described on the figure 1a (signal response of all sensors at a fixed atmosphere) and 1b (signal response

of one sensor to all the studied concentrations). So, we have selected one new feature corresponding to

the maximum value (called Peak) of the signal response derivative curve. This maximum appears for all

the sensors in the first minutes of the gas exposition.

For the calculation of this Peak feature, the time-response was first filtered to eliminate signal noises,

and then derivated. Several digital filters were tested to optimize the type and the characteristic

parameters of the filter to be unique for each sensor and all the exposures. Butterworth type filtering has

given the best results. For each sensor, the Peak values show an appreciable variation along with the all

oil concentrations (Figure 2).

Fig. 2. (a) Example of a derivative curve without filtering; (b) Derivative curve of SP-MW0 for all the used pine oil concentrations

The three selected features are tested for their capacity to differentiate the oil concentrations. In Figure

3 we present the evolution of the Vs-V

and the Peak features along with the oil concentrations and for all

the sensors. So, to take into account all these informations, the values of each extracted feature must be

assembled in a data basis and then treated by classification methods.

Fig. 3. Evolution of two selected features along with the pine oil concentrations: (a) Vs-V

; (b) Peak

The capacity of each feature to classify the pine EO concentrations was studied using first the Principal

Component Analysis (PCA). This non-supervised statistical method permits us to represent all the

observations of a multidimensional data basis in a reduced dimension. Figure 4 presents the PCA results

obtained with each selected feature taken separately. In the case of the Slope values the 3, 4, 5 and 6%

concentration groups are overlapped and not differentiable. With the Vs-V

values the group of 4, 5 and

6% concentrations are very closed and difficult to be distinguished. By using the Peak feature, we obtain

M. Siadat et al. / Procedia Engineering 47 ( 2012 ) 52 – 55

a better classification and only the group 4 and 5% are closed. We can then conclude that the new feature

Peak is more accurate than the classical features for our application. In addition the Peak is less dependent

on the sensor drift (derivation) and it is obtained faster than the other ones. To confirm this finding, the

discrimination power of each feature is tested using the supervised Linear Discriminant Analysis. The

concentration identification is validated by cross validation technique: a success rate of 95.2% is obtained

with the Slope, 99.2% with the Vs-V

and 100% when using the Peak feature.

Fig. 4. PCA diagrams obtained with each feature taken separately: (a) Slope; (b) Vs-Vo; (c) Peak

4. Conclusion

We have defined a new feature and highlighted its capacity for a rapid and accurate identification of

odor concentrations, comparing to the results obtained from the traditional features used in metal oxide

senor response processing. However, the determination of this parameter, deduced from the derivative

curve of the time response of the sensors, needs accurate measurement system and adequate signal

filtering. This data processing can be easily implemented to the intelligent system like electronic nose.

References

[1] T. Eklov, P. Materson, I. Lundstrom. Selection of variables for interpreting multivariate gas sensor data. Analytica Chimica

Acta; 381(1999) 221-232

[2] O. Gualdron, E. Llobet, J. Brezmes, X. Vilanova, X. Correig. Fast variable selection for gas sensing applications IEEE

sensors; (2004) 892-895

[3] H. Sambemana, M. Siadat, M. Lumbreras. Gas sensing evaluation for the quantification of natural oil diffusion. Chemical

Engineering Transaction;, 23 (2010) 177-183

A New combined transient extraction method coupled with WO 3 gas sensors for polluting gases classification

Article

Sep 2021
SENSOR REV

Purpose – The purpose of this paper is to deal with the classification improvement of pollutant using WO3 gases sensors. To evaluate the discrimination capacity, some experiments were achieved using three gases: ozone, ethanol, acetone and a mixture of ozone and ethanol via four WO3 sensors. Design/methodology/approach – To improve the classification accuracy and enhance selectivity, some combined features that were configured through the principal component analysis were used. First, evaluate the discrimination capacity; some experiments were performed using three gases: ozone, ethanol, acetone and a mixture of ozone and ethanol, via four WO3 sensors. To this end, three features that are derivate, integral and the time corresponding to the peak derivate have been extracted from each transient sensor response according to four WO3 gas sensors used. Then these extracted parameters were used in a combined array. Findings – The results show that the proposed feature extraction method could extract robust information. The Extreme Learning Machine (ELM) was used to identify the studied gases. In addition, ELM was compared with the Support Vector Machine (SVM). The experimental results prove the superiority of the combined features method in our E-nose application, as this method achieves the highest classification rate of 90% using the ELM and 93.03% using the SVM based on Radial Basis Kernel Function SVM-RBF. Originality/value – Combined features have been configured from transient response to improve the classification accuracy. The achieved results show that the proposed feature extraction method could extract robust information. The ELM and SVM were used to identify the studied gases.

Reduction of the Measurement Time by the Prediction of the Steady-State Response for Quartz Crystal Microbalance Gas Sensors

Article

Full-text available

Jul 2018
SENSORS-BASEL

This paper presents a new approach to reduce the measurement time by the prediction of the steady-state using the transient response to ethanol for quartz crystal microbalance gas sensors coated with ethyl cellulose. The experimentally measured response curves were successively fitted using a mathematical model based on the sum of two exponentials with different time constants. The parameters of the model were determined, and the time constants and the magnitude of the steady-state response were analyzed. Even though the time constants did not stabilize well, the parameter corresponding to the magnitude of the steady-state response quickly converged and stabilized after 37 s. Moreover, this calculated parameter was highly correlated with the measured values of the steady-state response, which was measured at five times the longest time constant (83 s) of the model. Therefore, the steady-state response could be predicted with a 55% reduction in the measurement (detection) time.

Classification of Data from Electronic Nose Using Gradient Tree Boosting Algorithm

Article

Full-text available

Oct 2017
SENSORS-BASEL

In this paper, an approach that can fast classify the data from the electronic nose is presented. In this approach the gradient tree boosting algorithm is used to classify the gas data and the experiment results show that the proposed gradient tree boosting algorithm achieved high performance on this classification problem, outperforming other algorithms as comparison. In addition, electronic nose we used only requires a few seconds of data after the gas reaction begins. Therefore, the proposed approach can realize a fast recognition of gas, as it does not need to wait for the gas reaction to reach steady state.

A Transient Signal Extraction Method of WO3 Gas Sensors Array to Identify Polluant Gases

Article

Full-text available

May 2016
IEEE SENS J

Electronic nose is a system, which can determine the fingerprint of gas sample by a sensor array coupled to pattern recognition system. In this paper, a sensor array based on WO3 gas sensor has been described, and a feature extraction technique, including integral and primary derivative, is reported, which leads to higher classification performance as compared to the classical features: fractional resistance change Delta R-s and Delta R-f. The sensor array has been exposed to ozone, ethanol, acetone, and a mixture of ozone and ethanol while keeping the temperature constant. The data array has been normalized and auto scaled then analyzed with the principal component analysis. Results indicate that successful classifications have been gotten in the discrimination of three kinds of oxidizing and reducing gas with the proposed feature extraction method using support vector machine which shows that 97.5% were correctly discriminated with integral and primary derivate comparing the classical variable with 85%. The results of data analysis implied that coupling the extracted features in the same database would be interesting and more robust then the standards features.

The design and validation of a fast and low-cost multi-purpose electronic nose for rapid gas identification

Article

Sep 2022
SENSOR REV

Purpose The purpose of this paper is to design and validate an electronic nose (E-nose) prototype using commercially available metal oxide gas sensors (MOX). This prototype has a sensor array board that integrates eight different MOX gas sensors to handle multi-purpose applications. The number of sensors can be adapted to match different requirements and classification cases. The paper presents the validation of this E-nose prototype when used to identify three gas samples, namely, alcohol, butane and cigarette smoke. At the same time, it discusses the discriminative abilities of the prototype for the identification of alcohol, acetone and a mixture of them. In this respect, the selection of the appropriate type and number of gas sensors, as well as obtaining excellent discriminative abilities with a miniaturized design and minimal computation time, are all drivers for such implementation. Design/methodology/approach The suggested prototype contains two main parts: hardware (low-cost components) and software (Machine Learning). An interconnection printed circuit board, a Raspberry Pi and a sensor chamber with the sensor array board make up the first part. Eight sensors were put to the test to see how effective and feasible they were for the classification task at hand, and then the bare minimum of sensors was chosen. The second part consists of machine learning algorithms designed to ensure data acquisition and processing. These algorithms include feature extraction, dimensionality reduction and classification. To perform the classification task, two features taken from the sensors’ transient response were used. Findings Results reveal that the system presents high discriminative ability. The K-nearest neighbor (KNN) and support vector machine radial basis function based (SVM-RBF) classifiers both achieved 97.81% and 98.44% mean accuracy, respectively. These results were obtained after data dimensionality reduction using linear discriminant analysis, which is more effective in terms of discrimination power than principal component analysis. A repeated stratified K-cross validation was used to train and test five different machine learning classifiers. The classifiers were each tested on sets of data to determine their accuracy. The SVM-RBF model had high, stable and consistent accuracy over many repeats and different data splits. The total execution time for detection and identification is about 10 s. Originality/value Using information extracted from transient response of the sensors, the system proved to be able to accurately classify the gas types only in three out of the eight MQ-X gas sensors. The training and validation results of the SVM-RBF classifier show a good bias-variance trade-off. This proves that the two transient features are sufficiently efficient for this classification purpose. Moreover, all data processing tasks are performed by the Raspberry Pi, which shows real-time data processing with miniaturized architecture and low prices.

Implementation of Air Pollution Supervision System Using LabVIEW

Conference Paper

Mar 2018

WO3 sensors array coupled with pattern recognition method for gases identification

Conference Paper

Full-text available

Mar 2016

This paper presents the performance of gas sensors as electronic nose coupled with pattern recognition method for gases identification. In fact, the implementation of the electronic nose in a characterization process is based on two fundamental phases: a learning phase and a phase of identification. That is why we need an accurate extraction method in order to obtain performant classification. In this study, we propose to extract transient parameters in a dynamic mode: derivate and integral. The performance of these features is validated by the analysis method: principal component analysis (PCA) and K nearest neighbors (KNN), which present 98, 74% rate classification.

Gas sensing evaluation for the quantification of natural oil diffusion

Conference Paper

Jan 2010

Recent clinical studies show the bebeficial effects due to the inhalation of appropriate concentrations of certain natural oils on mental diseases like phobia, stress or Alzheimer. To demonstrate these effects, medical staff needs a system which permits an automatic control of the essential oil diffusion in a closed chamber. For this purpose, out work demonstrates the ability of chosen chemical gas sensors to identify the concentration of an essential oil evaporated in a neutral atmosphere. Sensor responses are studied using a selection of parameters which are tested for their performance to detect the oil concentrations and minimizing the drift effect of the sensors. Out system can detect various quantity of evaporated oil (pin and lavender) in the range of pleasant concentrations for humans.

Fast variable selection for gas sensing applications

Conference Paper

Jan 2004

We introduce a new variable selection approach, which converges much faster to the optimal set of variables for a given application. The new procedure runs in two steps. First, a coarse and very fast variable selection procedure is applied: A figure of merit is defined and computed for every variable, a threshold value set and only the variables whose figure of merit is higher than the threshold are retained for further selection. Then, a fine-tuning selection based either on deterministic or stochastic methods is conducted on the variable subset that resulted from the first step. The method is demonstrated using a database consisting of vapors of ethanol, acetone and toluene and their binary mixtures (120 variables/measurement). Vapors can be simultaneously identified and quantified with a 92.7% success rate and the time needed for variable selection is reduced at least by a factor of 4.

Selection of variables for interpreting multivariate gas sensor data

Article

Feb 1999
ANAL CHIM ACTA

In this work, methods to select relevant variables from a large set of available gas sensor parameters were examined. Two data sets, containing a large number of variables, were studied. The objective was to find the best descriptors, which could predict interesting properties of the measurements. Using a forward selection procedure, applying the root mean square error from a multilinear regression model as the selection criterion, it was possible to get good prediction accuracy from a back-propagation neural network (ANN). This procedure was fast, compared to the usual trial and error variable selection, objective, and can be made fully automated. In addition, the use of principal component analysis (PCA) and partial least squares (PLS) score vectors used as descriptors were examined. The ANN models constructed with either the PCA or the PLS score vectors as input gave for the first, rather smooth, data set, errors of the same size or smaller than for the forward selected parameters. For the second data set, containing many noisy variables, the forward selected parameters outperformed the other two data reduction methods.

Lumbreras Gas sensing evaluation for the quantification of natural oil diffusion