Uploaded by Pramod Borse


Odor, aroma and VOCs constitute a important aspect of characteristics of various materials, systems and
processes. Most of the materials, processes or systems have certain odor or emit VOCs that are unique to
them. This odor or VOCs can be used as a marker to identify and discriminate that material. A change in
the odor of the material may indicate that the material is not the normal condition. For example, there is a
change in metabolic process of patients with diseases like cancer, which leads to presence of certain VOCs
in the exhaled breath which may not be present in the breath of a healthy person. If we are able to
discriminate the breath of a affected person and a healthy person, the presence of diseases can be detected
in a simple manner. Thus, the ability to detect the odor of a system or a material, allows one to get
information about its state of condition. Techniques like GC-MS (Gas Chromatography – Mass
Spectrometry), Ion Mobility Spectrometry, Selected ion flow tube are being used to detect the odor in
various fields like food processing and storage, clinical diagnosis, beverages, pharmaceutical, chemical etc.
GC-MS offers superior accuracy in detection, however expensive instrumentation, complex diagnosis and
time consuming operation hampers it’s wide spread usage. Electronic-Nose (E-Nose) is the emerging
solution for the detection of odors and VOCs. E-Nose is a intelligent chemical array sensor system that
mimics the mammalian olfactory system, in-short as artificial nose. Electronic Nose is inspired by the
mechanism of human olfaction. E-Nose can detect and distinguish the profiles of different VOCs, odors
and aromas from various sources and pathways based on chemical sensing mechanism. Chemical sensors
used in E-nose include: metal-oxide sensors, conductive polymers, FET, Acoustic wave sensors and
electrochemical sensors. Chemiresistive sensors based on metal-oxide semiconductors are the widely
reported ones for the E-nose. Simple mechanism, ease of fabrication, cost-effectiveness and simple
circuitry makes them a preferred choice. Chemical sensors are often based on two types of approaches
namely, selective approach and cross-sensitive sensor array mechanism. Selective approach is based on a
specific target mechanism in which a single sensor-active element is designed to interact and detect the
presence of a single compound. In order to detect this specific target compound, the sensing layer should
be specifically tested and fabricated and which should not show appreciable cross-response to any other
unintentional target material. Though this approach is highly sensitive to the target material, the procedure
to synthesize unique sensing material to each and every different target is a herculean task. This approach
is widely used when detection time is very crucial and the sensing atmosphere has less number of
unintentional target compounds. The second approach based on cross-reactive sensor array is employed in
E-nose. In this cross-reactive sensor array approach, instead of single sensor element, an array of sensors
are used. Mammalian nose consists of a large number of olfactory receptors which detect and discriminate
thousands of odor compounds based on cross-responsive. Brain stores this pattern specific to each odor.
This odor is identified as a result of collective response of many receptors based on recognition of the
pattern stored by the brain. Similar to olfactory system, the e-nose consists of ‘n’ number of sensors, when
the array is exposed to the target gas, each sensor in the array may react differently with the gas. Overall
interaction of gas with the array generates a pattern which is unique to that set of target gases. Thus each
pattern is finger print to that specific gas or collection of gases. By the help of pattern recognition
techniques the generated pattern can be recorded and analyzed with data in the database and target gas can
be identified. Thus, based on the response pattern generated by the sensor array and extraction and analysis
of the pattern by certain pattern recognition techniques, the target gas can be discriminated.
Figure 1 Schematic showing a) Selective approach and
b) cross-reactive array approach
Figure 2 Human Olfactory system and E-Nose technique
Multiple Sensor Array
Sensor array converts the target gas analyte into a machine-readable signal. Sensors used in array include:
metal-oxide sensors, colorimetric, conductive polymers, FET, Acoustic wave sensors and electrochemical
sensors. In array approach, sensor array is chosen to respond to a number of different chemical analyte. In
the array, there is no requirement for all sensors to be reactive towards target analyte. Sensor array must be
diverse enough to respond to largest possible cross-section of analyte. The distinct pattern of response
generated by the sensor array to each target serves as a fingerprint . The clear advantage of cross-sensitive
array over selective approach is that multiple sensor array can detect variety of analyte, including those for
which the array is not originally designed. The number of sensors that must be present in array depends
upon the target analyte. When the target species is mixture of many gases, the number of sensors in the
array must be increased so as to decrease the ambiguity in the response. (Schematic of mechanism of a ENose based on Colorimetric sensor array is illustrated in Figure 3)
Figure 3 Schematic of E-Nose sensor array based on dye-based colometric sensors
Pattern Recognition Techniques
Once the target analyte react with the sensor array, data of specific response of each sensor to analyte is
acquired through a data acquisition software (e.g. VI- Virtual Instrumentation). The DAQ software must
be equipped with capacity to handle and store data from all sensors simultaneously. The purpose of data
acquisition (DAQ) system is to acquire the signal and transfer it to the data analysis or pattern recognition
system. Pattern recognition algorithm (PRA) does the classification, identification and if necessary the
quantification of the analyte based on the data stored in the database. The database is generated by
exposing the sensor to various target analyte and related compounds and recording their corresponding
pattern, which is uniquely attributed to that analyte. This PR algorithm process the raw data offered by the
DAq software and based on the correlation of information in the database, it identifies and relates the data
to that analyte. There are various pattern recognition techniques to facilitate this process, which include:
•principal component
analysis (PCA)
discriminate analysis
•cluster analysis (CA)
•artifcial neural
network (ANN)
•radial basis function
Figure 4 Various pattern recognition techniques
Figure 5 Applications of E-Nose in various Industries