Tuesday, March 4, 2008

Electronic Smell

** Niraj Kumar & I presented this paper at Felicity '04, held at IIIT, Hyderabad in February, 2004. We won the 3rd prize. **
**Moksh Walia & I presented this paper again at Troika '04, held at Delhi College of Engineering, New Delhi in March 2004. We won the 2nd prize. **
____________________________________________________________________
It is now possible to add a new sense of smell to the Internet. This is done through the virtual nose called the e-nose. An electronic nose is not a replacement for people, it is a supplement. The "electronic nose" is a relatively new tool that may be used for safety, quality, or process monitoring, accomplishing in a few minutes procedures that may presently require days to complete

The electronic nose consists of two components,
(1) an array of chemical sensors (usually gas sensors) and
(2) a pattern-recognition algorithm.
The sensor array "sniffs" the vapors from a sample and provides a set of measurements; the pattern-recognizer compares the pattern of the measurements to stored patterns for known materials. Gas sensors tend to have very broad selectivity, responding to many different substances. This is a disadvantage in most applications, but in the electronic nose, it is a definite advantage. Although every sensor in an array may respond to a given chemical, these responses will usually be different. In recent years, electronic noses have been sniffing out landmines, detecting contraband drugs, sensing for chemical and biological weapons, identifying batches of spoiled food, and even showing promise for aiding in the diagnosis of diseases like lung cancer and pneumonia. These interesting devices are designed to mimic the ability of the human nose to detect very small quantities of odorants. E-noses can also detect chemicals that have no odor, such as toxic carbon monoxide. Compared to the senses of sight, hearing, and touch, scientists know relatively little about how humans smell and taste. Designers of electronic noses have tried to mimic human noses by linking together sensors that detect a variety of volatile compounds.

The advantages of ANN-containing electronic noses over chemical sensors. The `nose' is trained on examples rather than rules, negating the need for expert description of the domain. The number of odours classified is greater than the number of sensors because the network can discriminate between patterns of activation across all the sensors. Fewer sensors are needed. Thus one can use less selective (and less expensive) sensors. Real-time odour identification. The time consuming part of the process is training of the network. Once trained the system's performance is governed by the speed of the chemical sensors. It processes new smells, despite never having having been trained on them.

Need of an electronic nose : Human noses have always been the best odor receptors distinguishing between very similar ones. Contrary to physical senses (dealing for instance with acoustic or optic mechanisms), some aspects of the human taste and olfaction physiological working principle are still unclear. Because of these intrinsic difficulties toward the understanding of the nature of these senses, only sporadic research on the possibility of designing artificial olfactory systems was performed until the end of the eighties. But as good as human noses are for chemical detection, they have drawbacks. They "fatigue" if subjected to repeated smelling tasks or strong scents, as anyone knows who has initially been shocked by an overpowering smell and then has become acclimated to the odor. The exquisite sensitivity of the nose can be defeated by a common cold, and for obvious reasons, human noses have limitations on sniffing out highly toxic compounds.The electronic noses are unbiased. They are not subject to interference by emotional states (e.g., tiredness, mood) or illness (e.g., allergies). They can be used in dangerous situations (e.g., contamination testing). They are not subject to habituation.


DIFFERENT TYPES OF SENSORS
1. The sensors used in an electronic nose can be mass transducers (such as Quartz microbalanz or QMB)

2. Chemo resistors (based on metal-oxides )
3. Chemo resistors (based on conducting polymers)
4.Some arrays comprise both types of sensors.
5. Nanomechanical Cantilevers

The artificial nose demonstrator is based on micro fabricated nanomechanical cantilever sensors - thin silicon beams - a few hundred micrometers long and one micrometer thick. Eight cantilever sensors, each is coated with a different sensor layer, are integrated in an exchangeable array. On exposure to an analyte, the analyte molecules adsorb on the cantilever’s surface. This leads to formation of interfacial stress between sensor and adsorbing layer. The bending pattern is characteristic for each analyte.


How smell sensors work : A smell sensor can be made from a quartz crystal with electrical connections and a special plastic coating. Quartz crystals are used in electronics because they can be made to vibrate at a precise frequency. A quartz crystal is what is used to control the speed of a processor in a PC. The frequency of vibration of the quartz crystal depends on its size, shape, stiffness and mass. The plastic coating on the crystal absorbs some chemicals so increases the crystals mass. The whole device is called a Quartz Crystal Microbalance (QCM) A quartz crystal can be thought of as mass on a spring.The frequency of oscillation of a mass on a spring is given by the formula: f = ½*PI*( (k/m)).Where k is the stiffness of the system in N/m, m is the mass of the system in Kg, f is the frequency of the system in Hertz.


Data Processing Methods: The signals generated by an array of odour sensors need to be processed in a sophisticated manner. The electronic nose research group has obtained considerable experience in the use of various parametric and non-parametric pattern analysis techniques. These include the use of linear and non-linear techniques, such as discriminant function analysis, cluster analysis, multi-layer perceptions, genetic algorithms, fuzzy logic, and adaptive models.


Pattern Recognition : A sensor comprises a material whose physical properties vary according to the concentration of some chemical species. These changes are then translated into an electrical or optical signal which is recorded by a device. The sensors are non-selective A chemical compound is identified by a pattern of the outputs given by the different sensors, thanks to pattern recognition methods. There is an exhaustive database which contains the information about patterns of different chemicals. The pattern now generated by the sensors and the data processor is compared with every entity of the database. If a match occurs then the chemical is recognized by the system.


Other techniques of operation: Electronic odour sensing devices have arrays of sensors that detect the presence of vapors. In this way they act as volatile chemical detectors. The sensors respond by producing electrical signals that are passed on to an artificial intelligence system programmed to interpret them.


APPLICATIONS:
Environmental Monitoring:
· Monitoring of factory emissions, air quality and household odours.
· Detection of oil leaks.
· Analysis of toxic wastes and fuel mixtures.


Medicine:
. Breath odours. The Highland Psychiatric Research Group is pioneering a
breath odour analyser for the prediction of acute schizophrenic illness in vulnerable patients; normally an extremely complicated procedure.


Body fluids. The smell of urine and blood can help in the diagnosis of liver and bladder problems.
Wounds: Smell can be an important indicator that the operation is not going well and so a remote electronic nose coupled with a local odour generator would help in the
transmission of olfactory information for medicine


Food industry applications.
- Inspection of food to test for ripening/rotting.
- Testing of packaging materials for odour containment.
- Microwave oven cooking control.
- Verifying if orange juice is natural.
- Grading whiskey and controlling fermentation.
E-nose could sniff out time of death. It could detect the time of death of a corpse by identifying odors.


Army applications. Inspecting the presence of landmines, detecting contraband drugs, sensing for chemical and biological weapons.


Thus we see that no one expects e-noses to duplicate all the capabilities of the human nose anytime soon. But they can deliver substantial benefits in situations where, given the choice, we'd prefer not to use our own sniffers. No instrument is complete without its shortcomings and an electronic nose is no exception.

Its drawbacks include:
- Difficulty in maintaining an exhaustive database of different fingerprints of chemicals.
- It may be very difficult to analyze a complex mixture of different chemicals.
- The precision of the device while analyzing similar smell is controversial.


There are numerous potential applications of electronic noses from the product and process control through to the environmental monitoring of pollutants and diagnosis of medical complaints. However, this requires the developments of application-specific electronic nose technology that is electronic noses that have been designed for a particular application. This usually involves the selection of the appropriate active material, sensor type and pattern recognition scheme. The work of the group has led to several commercial instruments, one employing commercial tin oxide sensors (Fox 2000, Alpha MOS, France) and another employing conducting polymer sensors (NOSE, Marconi Applied Technology, UK). Collaborations also exist with Osmetech (UK) and Cyrano Sciences (USA) Future developments in the use of hybrid micro sensor arrays and the development of adaptive artificial neural networking techniques will lead to superior electronic noses.

____________________________________________________________

References :
REFERENCES

1 http://www.inapg.inra.fr/ens_rech/siab/asteq/elba/sommelen.htm

2 http://www.cogs.sussex.ac.uk/lab/nlp/gazdar/teach/atc/1998/web/sloss/index.html

3 Danny Kingsley – ABC Science Online

4 Gardner J W and Bartlett P N 1999 Electronic Noses ( OUP Press, Oxford); Gardner J W and Bartlett P N (Eds) 1992 Sensors & Sensory Systems for an Electronic Nose (Dordrecht: Kluwer Academic Publishers) NATO ASI Series:
Applied Science Vol. 212 pp.327

5 Gardner J W and Bartlett P N 1994 Sensors and Actuators B 18 211-220 "A
brief history of electronic noses"'

No comments: