physicalrobustness (also duringcommonperipheral services), no needfor
skilledpersonneland cost effectiveness(e.g.capital costs, low power consump-
tion). Several companies tried to fulfil these needs by the productionof
commercialsensor systems.Manyof these systems are like standard chemical-
analytical devices (e.g.for gas or liquidchromatography),whichcan be used
onlyin laboratories.Kress-Rogersand Brimelow(2001)give a generaloverview
of thesecommercialsystemsand recentpapersreviewcommercialsystems,e.g.
chemical sensors by Stetterand Penrose(2002), or biosensorsby Rodriguez-
Mozazet al.(2004).
Thissituation addsone of the most importantissues: the abilityto use sensor
devicesas an in- or on-linetool,i.e. directprocess monitoring, startingwiththe
incoming goods inspection. In a nextstep, process monitoring couldstart
directlyon farm-side,which wouldagainincreasethe sensor requirements,
especially robustnessand easeof handling.
The following section is dividedinto two parts.The first will concentrate on
devicesthat are capableof controlling by air sampling chemical sensors, while
the second one will concentrate on liquid detection, i.e. devices such as
electronictongue devices and biosensorsystems.
19.2 Sensortypes
19.2.1 Volatile detection± electronic nose
Sensors for volatile (vapour phaseanalysis) detectionare already used in
automobiles, medicine, process control and laboratory measurements. The
responseof thesesolid-stategas sensors is usuallynot veryspecific,for example
towards reducing(fuel)gases, oxidising gasesor ammonia and molecules with
ammonia structure. To overcomethe problemof non-specificdetectionor cross-
sensitivity,Persaud and Dodd proposed in 1982the concept of so-called
electronic noses', the detection of volatilesby sensor-arrays. Gardnerand Bartlett(1993)defined electronicnoseas an instrument,whichcomprises an
arrayof electronic chemical sensors (e.g.solidstategas sensors) withpartial
specificity and an appropriatepattern-recognition system, capable of recognising
simpleor complexodours'.
To ensure the sensors receivea largeamountof information,the systems are
equippedwithan arrayof sensorsof the sameworking principle or withsensors
of differentworking principles (so-called hybrid or multi-sensor systems). The
amountof raw data fromup to 30 single sensors requires complex statistical data
processing. Evaluationsoftwareof sensorsystemsoftenincludesfunctions for
thesesophisticated statistical approaches suchas principal componentanalysis
(PCA), partial least squares (PLS) or artificial neural network (ANN)
calculations.
Commercial instruments appeared on the market and in the scientific
laboratories,in the early1990sand sincethat timehundreds of publications
aboutpossible applicationshavebeenwritten.The advantages of thesedevices
Improvinghygieniccontrol by sensors 289