new sensors for food industry - université · pdf filenew sensors for food industry...
TRANSCRIPT
NEW SENSORS FOR FOOD INDUSTRY Corrado Di Natale1, Ed Jones2, Gustavo Alcuri3 1 University of Rome Tor Vergata, Italy 2 CTech; Chester, UK 3 Alctra; Montreuil, France
www.cafe-project.org
www.cafe-project.org
Outline
• Extending sensing capabilities in CAFÉ project • Approach to new sensors
• Ultrasounds • Electric impedance • Gas sensors
• Electronic nose
• Applications to case studies • Conclusions
2
www.cafe-project.org
Sensors available in case study plants
Case study 1: wine
Temperature sensors, flowmeter for injected O2, flowmeter for produced CO2, on-line GC
Case study 2: beer
Viscosity, pH, conductivity, VIS-NIR, particle size distribution, turbidity
Case study 3: freeze-drying
Temperature (product and refrigerating liquid), chamber pressure
Case study 4: Ice cream
Temperature, pressure, power consumption, dasher speed, Focused beam reflection, on-line viscometer
3
www.cafe-project.org
CAFÉ approach to extend sensing capabilities
• During the processes a number of physical and chemical properties are altered.
• Considering the complexity of food composition and the changes occurring during the process the use of non-specific broadly selective sensors was considered
• Physical properties • Ultrasound probes • Electric impedance
• Chemical properties • Gas sensor arrays
4
www.cafe-project.org
Ultrasonic probes • Ultrasonic sensors investigate the propagation of acoustic waves in a media • Waves are reflected, transmitted and absorbed. • The acoustic properties of materials depend generally on the density (amount of
material moved by the wave) and the mechanical resistances (resistance to movement).
• There are several examples of applications of ultrasounds to the characterization of foodstuff.
Sound velocity vs. concentrations of simple and complex food ingredients Data from Saggin and Coupland, J. Food Sci. 2001
5
www.cafe-project.org
Emission-receiver and electronic unit
piezocomposite transducer Emission – reception – treatment unit
delay time
level E R
d E/R
d
Measurement modes: transmission pulse/echo Measurement features
Delay: sound velocity Attenuation: impedance loss
6
www.cafe-project.org
Freeze-drying process
Ultrasonic unit A in emission / reception mode.
Ultrasonic unit B in echoes mode.
Emission transducer
Reception transducer
Emission / reception transducer
Ultrasound frequency 1.25 MHz
7
www.cafe-project.org
Ultrsonic monitoring of Freeze-drying
-30
-20
-10
0
10
20
30
0
5
10
15
20
25
30
35
40
10:3
110
:37
10:4
810
:53
11:0
411
:13
11:2
211
:30
11:3
411
:36:
…11
:38
11:4
511
:49
11:5
311
:57
12:0
412
:09
12:1
712
:21
12:2
512
:38
13:0
113
:35
13:4
114
:11
14:4
715
:06
15:2
415
:37
tem
pera
ture
(C°)
prop
agat
ion
time
(µs)
mode E/R mode echo temperature
Transition zone at T=-10.6°C Abrupt change in E/R mode. Slow transition of echo mode signal due a progressive stratification down to T=-14.8°C.
8
www.cafe-project.org
Electric impedance
• The dielectric properties of foods are studied because of their implications with microwave cooking and processing.
• However, the impedance even at lower frequency can provide interesting information about the content and the structure of the food.
• Dielectric properties depend on a wide range of physical and chemical properties of the material, including: • conductivity • viscosity • density • composition • temperature …
9
www.cafe-project.org
Impedance probe
• Simple probe design to allow impedance measurements in the presence of dielectrics under test
• Impedance measurements can be calibrated to give real and complex permittivities: ε�r(f) and ε�r(f).
• Two dielectric standards are required for calibration (usually air and water)
• For CAFÉ monitoring actual dielectric values not required, just signals that change with target processing properties
sensor
analyser
sample fringing field
1.234
conditioning unit
10
Optimized probe
Fringe field diffusion
www.cafe-project.org
Fingerprint method with the reflection coefficient
• 1) The frequency dependence of the reflection coefficient of the initial states is measured • The reflection coefficient is related to the
change of dielectric constant with respect to the external medium.
• 2) The frequency dependence of the reflection coefficient of the ongoing state is measured
• 3) The frequency dependence of the difference between the two reflection coefficients is produced
11
www.cafe-project.org
Beers mixtures
Flat Hoegarden beer added to flat Leffe beer Bubbly Hoegarden added to bubbly Leffe
Flat Hoegarden added to fizzy Leffe Fizzy Hoegarden added to fizzy Leffe
12
www.cafe-project.org
Physical and chemical changes heating stirring
Ethanol addition Water addition
13
www.cafe-project.org
Test in real conditions Pilot plant in Wageningen
14
Real permittivity at 100 MHz
• A difference between filtered and unfiltered sample is observed
• Sensitivity to temperature disturbance • Necessity of a better screening from
environmental noise.
www.cafe-project.org
Headspace composition
• In principle gas sensors can be used to monitor the evolution of headspace during the processing of foods.
• But: • Headspace build-up may requires a large time with respect to the typical
time scale of the process. • Physical conditions of food during the process (e.g. temperature and
pressure) do not allow for the formation of a readable headspace. • Nonetheless, gas sensors can be used in selected processes and
they are valuable to determine the quality of the final product. • In CAFÉ the properties of sensors based on metalloporphyrins were
considered.
15
www.cafe-project.org
Richness and versatility of porphyrins chemistry
16
Hydrogen Bond Acceptor
Hydrogen bond acceptor Hydrogen bond donor
Hydrophobe
Aromatic Hydrophobe
Coordination
Porphyrins building blocks
www.cafe-project.org
Porphyrins coated Quartz microbalances
17
• Mass to electric frequency converters
• TSMR sensors are simple to be prepared and the signal read-out can be done with great accuracy.
• Since any interaction results in a variation of the mass, the mass transducer is in principle non-specific.
• The resolution is of the order of 1 ng • About 1012- 1014 molecules have to be
absorbed to have 1 Hz of frequency shift.
Brunink et al. Analytica Chimica Acta 1996
www.cafe-project.org
Some applications
19
Food Quality and Control Eifler et al. PLoS ONE, 2011 Santonico et al. Food Chem, 2010 Santonico et al. Sens Act. B 2008 Olafdsottir et al. Trends Food Sci Tech. 2004 Saevels et al. Posth. Sc. Techn. 2003 ••••••
Medical Diagnosis Santonico et al. Lung cancer, 2012 Montuschi et al. Chest, 2010 D’Amico et al. Skin Res. Techn. 2007 Di Natale et al. Bios. Bioelec. 2003 ••••••
Spacecraft air quality control Martinelli et al. Microgravity Sci Techn. 2008
Fortezza et al. Acta Astronautica, 2006 •••••• ISS 2005
ISS 2011
www.cafe-project.org
Application to ice-cream volatile compounds vs. temperature
SPME
-10 -5 0 5 10 150
0.5
1
1.5
2
2.5
3
3.5 x 106
Temperature [°C]
Ab
un
dan
ce [
a.u
.]
Heneicosane, 3-methyl- 1,3,8-p-Menthatriene D-Limonene Crithmene Nerol acetate
20
www.cafe-project.org
Monitoring of ice-cream making
• Lemon sorbet • 5 experimental conditions changing process
parameters : • Flow ice-cream machine (35 Kg/h) • Tevap refrigerant liquid (from -9°C to -16°C)
• aim: to predict from gas sensors signals the ice crystals size and the apparent viscosity • ice crystals size measured with Focused Beam
Reflection • viscosity was model estimated • Ice fraction
Crystal size 6.53 μm
Crystal size 6.63 μm
21
www.cafe-project.org
Gas sensors measurements
• A measurement chamber is fixed at the output of the ice-cream machine.
• The chamber holds a fixed amount of ice-cream into a closed volume whose headspace is delivered into the sensors cell.
22
www.cafe-project.org
Crystal Size Prediction
RMSETrain : 0.1133 µm RMSETEST : 0.1980 µm about 8% of the range of variability of crystals size.
• Feed Forward Neural Network. • Single hidden layer (with 3
neurons) • 2/3 of the dataset been used
for training and 1/3 for test.
23
www.cafe-project.org
Estimation of viscosity and Ice fraction
24
RMSETrain : 0.3920 RMSETEST : 0.9926
RMSETrain : 0.0211 RMSETEST :0.0326
www.cafe-project.org
Application to freeze-dryed bacteria quality
25
• Concentrates of lactic acid bacteria are used as starters of fermentation processes for the production of cheeses, yogurts,……
• Freeze-drying is the method of choice to improve the shelf-life of concentrates.
active ingredient (Bacteria)
excipients (polymers, sugars,…)
Freeze-drying
lyophilized product storage
Quality criteria • water content (<3%)
• activity • cells viability
www.cafe-project.org
Experimental
26
• six batches of lactobacillus delbrueckii sbsp. bulgaricus • samples were prepared with different salts to
favour the drying • different freeze-drying conditions were applied to
change the quality of the products. • Nucleation LyoGamma Telstar
• half of each batch was used to measure water activity, acidification activity, and cells viability
• half for headspace analysis: electronic nose and GC/MS
www.cafe-project.org
Measurement setup
27
• Tor Vergata Electronic nose equipped with eight porphyrins coated quartz microbalances
• room temperature measurements
www.cafe-project.org
Measured quality parameters
28
• PCA of parameters indicates two groups of samples separated by the acidification activity.
• As expected, acidification activity is anticorrelated to the cells viability
cells viability
water activity
Tmax
www.cafe-project.org
Electronic nose data
29
• The PCA of electronic nose data is largely distorted with respect to the quality parameters.
• Nonetheless, the two groups of samples found in the PCA of quality parameters and GC abundances are visible also in electronic nose data.
www.cafe-project.org
PLS prediction of quality parameters
30
AW viability Tmax
• PLS was used to estimate the three quality parameters • leave-one-out cross-validation was used for latent variables choice • 4 latent variables minimize the RMSECV
www.cafe-project.org
compounds identification
32
1 2 3 4 5 6 7 8 9 10 11 12 13 1,5-Hexadiene, 3,3,4,4-tetrafluoro- 53535 19979 45573 0 0 0 42421 0 0 0 0 0 0 2-Butanone, 3-hydroxy- 205615 982019 0 0 0 0 0 0 0 0 0 0 2,3-Butanediol, [S-(R@,R@)]- 665136 280857 1161775 0 0 0 0 0 0 0 0 0 0 Cyclopenta[e]tetrahydropyran-2-one, 5-(5-hydroxyindan-1-yl)-4a-methyl- 95293 114202 0 0 0 0 0 0 0 0 0 0 0 2-Propanol, 2-methyl- 698935 782189 227943 676121 556372 83147 0 146754 155908 196459 127855 0 222490 Oxime-, methoxy-phenyl-_ 197155 0 38095 0 0 0 0 0 0 0 0 158988 0 di-tert-Butyl dicarbonate 116648 119728 28742 132645 93394 12427 0 30913 33652 16980 13465 0 18794 1,1-Hexylenedioxybutane 295974 333641 132045 167240 125220 55304 0 40502 27932 81308 80091 20121 81936 Undecane, 2,2-dimethyl- 0 0 0 0 0 0 0 0 0 0 0 0 0 3-Hexen-2-one 128945 91092 60622 39834 8773 0 0 0 0 39564 25263 0 16199 1-Hexanol, 2-ethyl- 129539 0 0 0 0 0 0 0 0 0 0 39113 0 2,6-Heptadienal, 2,4-dimethyl- 7091 8144 0 8208 6528 0 0 0 0 0 0 0 0 3-Methyl-4-oximido-2-pyrazolin-5-one 0 3219 0 0 0 0 0 0 0 0 0 0 0 2-Propanol, 2-methyl- 0 5133 0 0 0 0 0 0 0 0 0 0 0 Di(1,2,5-oxadiazolo)[3,4-b;3,4-E]pyrazine, 4,8-diacetyl- 5707 0 0 0 0 0 0 0 0 0 0 0 0 2-Butanone, 3-methoxy-3-methyl- 11075 0 0 0 0 0 0 0 0 0 0 0 0 Fenproporex 13720 22489 12173 15452 13682 0 0 0 0 19893 19334 11620 14978 2,4,4-Trimethyl-1-pentyl methylphosphonofluoridate 81356 100433 63158 76281 74496 0 0 14381 8990 89896 103058 64750 86772 3-Butene-1,2-diol, 1-(2-furanyl)- 23937 29765 26056 18556 21781 10271 7458 7682 5516 14569 16054 14813 12565 3,5-Dimethyl-4-octanone $$ 42970 0 0 0 0 0 0 0 0 0 0 11694 0 Dibutyl phthalate 220206 0 28636 0 0 36198 35805 0 39562 0 167130 0
3-hexen-2-one butane-diol 3-hexen-2-one • The chemical functional descriptors
of VOCs more correlated with the acidification activity suggest a good detectability with porphyrins coated QMB.
www.cafe-project.org
Detection of filtration stages in beer production
• three formulations from a local brewery • Rajah: indian pale ale with large hop
content • Kerala: indian pale ale spiced and
falvored with orange, apple and cinnamon
• Spiga: weiss beer (31% filtered in the final product)
• three samples for each formulation
• unfiltered (as is from brewery) • paper filtered • siringe filtered
1mm sample of unfiltered
33
www.cafe-project.org
Principal component analysis of ENose data
non filtered
paper filtered
syringe filtered
• Filtration affects the ENose signal
• The removal of particulate increases the differences between samples
• The discrimination among beers largely increases after filtration.
• Results shows that the QMB
electronic nose is sensitive to filtration
• On line application in filtration unit is hindered by low-temperatures and high pressure.
34
www.cafe-project.org
Conclusions
• Broadly-selective sensors for physical and chemical quantities can provide useful information to monitor the evolution of processes in food industry.
• Electric and acoustic impedance can be used to estimate composition, density and texture properties
• The headspace composition can reveal changes in activity in the liquid material (e.g. due to crystalization) and to signal the presence of compounds affecting the quality.
• Eventually, these sensors can complement the sensors aimed at controlling the process parameters with feedback information about the quality of the processed product.
35