2. 2004 I Italian Near Infrared • Limo
•Fiber Optic on the Ham
Lodi Symphosium •Baby Food Plasmon (Heinz Group)
12th International
2005 Conference on Near-
• On-line Maillard reaction monitoring for food
additives production (caramel)
Auckland infrared Spectroscopy
2005 VII CISETA • Ice cream mixtures
Cernobbio
• Inorganic Integrators
II Italian Near • Pectin
2006 Infrared Symphosium
• SO2
• Proteolysis index in P.D.O. Cheeses
Ferrara
2007 13th International • Cous cous
Umeå Conference on Near- • Vanilline
• Wheat flour rheological paramethers
infrared Spectroscopy
• On-line polymerization process
2008 III Italian Near • Licopen content in Tomatoes
Lazise Infrared Symphosium • Barilla FT-NIR Network for Flour monitoring
NIR Seminar – Campden – October 14th 2009
3. ! " "#
$
Lab & R&D
• University
• State agencies for control
• Private Labs
INDUSTRY
TECHNOLOGY
PRODUCERS
NEEDS
• Raw material control • Analytical
• Monitoring production processes
• Fianl products quality control
systems
NIR Seminar – Campden – October 14th 2009
5. #
DIFFERENT RAW
DIFFERENT
MATERIALS
PRODUCTS
Wheat Flours
Quality 1
Industrial
BARILLA bread Quality 2
BAKERY
Quality 3
Cakes
Quality 4
Quality 5
Quality up to
Snacks 12
NIR Seminar – Campden – October 14th 2009
6. () * "% '
TARGET: Assess Quality of incoming Wheat Flour batches
• Identify Parameters to check
• Identify an Analytical System able to perform quickly such
checking
• Define sampling methods
• Collect samples and verify
12 different Wheat Flour Qualities considered
8 – 100 – 106 – 108 – 110 – 114 – 120 – 141 – 144 – 158 – 164 –
175…
• Samples collected starting from March 2007
• Samples collected from different suppliers all over Italy
NIR Seminar – Campden – October 14th 2009
7. '
CHEMICAL PARAMETERS REFERENCE METHODS
Moisture UNI reference
Protein Content Methods
Falling Number
Brabender
RHEOLOGICAL PARAMETERS Farinograph
Farinographic
Baking Absorption
Alveographic
W
P/L
Chopin Alveograph
NIR Seminar – Campden – October 14th 2009
8. "
Spectra acquisition by diffuse reflectance
using an FT-NIR Spectrometer
Wavelenght range 4000 – 10000 cm-1
• 3 Sub-samples for each incoming Flour Batch (3 spectra
measurment for each batch)
• Each spectra as average of 64 scans having a rotating petri
dish system
• Samples Temperature: 20 5 C
Data management
• NIRCal Chemometeric software to
Chemometric software develop quantitative calibration models
with Evaluation Set Tecnique
NIRCal 5.0
NIR Seminar – Campden – October 14th 2009
10. "
Checking incoming raw materials
Why NIR? chemical composition
Checking chemical composition of
finished formulations
NIR Seminar – Campden – October 14th 2009
11. "
Raw materials are paid Example: Soy Flour according
according to chemical to protein content
composition
Check every batch supplied
means to pay the correct price
To produce according to the
Finished products declared composition by having
an instant monitoring
To obtain the same chemical
Plus composition they could be used
different and new raw materials,
maybe cheaper
NIR Seminar – Campden – October 14th 2009
14. . .
*
Original Spectra
NIRC al : C uscus _Se mo la to_ Farina_U m id ita 2311 06 26/04 /20 07 10.34.43 A dm inistra tor
All Spectra
Calibration Spectra
Validation Spectra
0.8
R e fle c ta n c e
0.6
0.4
0.2
10000 9000 8000 7000 6000 5000 4000
Wavelengths
R
Samples Method Range SEP SEC
C-set/V-set
Umidità 210 PLS 10.00 - 16.09 0.99 / 0.99 0.12 0.12
Proteine 144 PLS 11.02 - 14.03 0.97 / 0.97 0.16 0.15
Ceneri 210 PLS 0.69 - 1.35 0.94 / 0.93 0.20 0.22
NIR Seminar – Campden – October 14th 2009
15. / '
Original Property / Predicted Property
N I R C al : Lat t i er o casear i o onl i ne- aggi or nat o2. ni r Lact ose - i m pl em ent ed2 23/ 05/ 2005 9. 25. 39 cam g
All Spectra
Pr e d ic t e d Pr o p e r t y la c t o s e
5.5 P ro p e rty Ou t l i e r S p e c t ra
V a l i d a ti o n S p e c tra f (x )= 0 .6 5 2 3 x + 1 . 7 1 3 7 r= 0 .7 7 7 8 3 8
C a l i b ra t i o n S p e c t ra f (x )= 0 . 6 5 5 3 x + 1 .6 9 9 3 r= 0 . 8 0 9 5 2 3
5.0 Original Property / Predicted Property
N I R C al : Lat t i er o casear i o onl i ne- aggi or nat o2. ni r Fat A - i m pl em ent ed2 23/ 05/ 2005 8. 49. 25 cam g
All Spectra
P r e d ic t e d P r o p e r t y f a t A
V a l i d a t i o n S p e c t ra f(x )= 0 . 9 8 8 9 x + 0 . 0 2 4 3 r= 0 .9 9 3 5 8 8
Ca l i b ra t i o n S p e c t ra f (x )= 0 . 9 8 8 2 x + 0 .0 3 6 3 r= 0 . 9 9 4 1 0 7
4.5
6
4.0
4
3.5
3.0 3.5 4.0 4.5 5.0 5.5
True Property lactose
2
0
0 2 4 6
True Property fat A
C-Set V-Set
Property C-Set SEE V-Set SEE
Regression Regression
(%) (SEC) (SEP)
Coefficient Coefficient
Fat 0.17 0.17 0.99 0.99
Protein 0.17 0.16 0.85 0.86
Dry matter 0.31 0.31 0.98 0.98
Lactose 0.16 0.17 0.81 0.78
NIR Seminar – Campden – October 14th 2009
16. / $
Olive grinding Olive
paste
Solvent Gramolatura
Husk Extraction
Estrazione Water
Extraction
Husk Separation Water
Oil
Filtration
Extra-virgin
Olive oil
NIR Seminar – Campden – October 14th 2009
17. / $
Spectrometer FT-NIR NIRFlex N500
• 2 acquisition each sample
• Every acquisition is tha average of 64
scan with rotating petri dish (total time<
1min)
• Temperature 20° C
Samples from different
N IR C a l : c o p y o f M o is tu r e , 0 .8 0 5 0 , 1 - 6 ./6 , 4 6 0 0 - 1 0 0 0 0 . 2 4 /0 4 /2 0 0 8 1 4 . 3 0 .4 9 A d m in is tr a to r
geograpical regions 2007
P r e d ic t e d P r o p e r t y M o is tu r e
Predicted Property vs. Original Property
A Spectra
ll
C lib tio S ectra f(x)= 73 0.73 r= 9 r2= 73 S ev(x-y)= 54 B S
a ra n p 0.98 x+ 00 0.9 37 0.98 d 0.82 IA (x-y)= 0 ran e(x)= .8.. 7 n 34
g 34 1.73 = 2
V lid S ctraf(x)= .0 2x-0 18 r= 20 r2 0 84 S v(x-y)= .8 8 B S
a ation pe 1 03 .1 1 0.99 = .9 0 de 0 17 IA (x-y)= .0 7 ra ge 4 .8.. 70 7 n 1
-0 64 n (x)= 1 .9 = 33
70
60
50 Standard
PARAMETR Range samples
errorSEP [%]
40
Moisture 0.8 34.8 – 71.7 240
30
40 50 60 70 Fat 0.9 15.8 – 31.1 200
O in P perty M isture
rig al ro o
NIR Seminar – Campden – October 14th 2009
18. / $ , '
Diffuse reflectance
Samples from different geograpical
PredictedProperty vs. Original Property
regions
P re d ic te d P ro p e rty F a t
A Spectra
ll
C lib tio S e f(x)= 6 1 0 7 1 r= .98 9 r2 0 6 S e
a ra n p ctra 0.9 2 x+ .1 4 0 0 = .9 21 d v(x-y)= .3 3 B S
0 8 6 IA (x-y)= 0 ra g (x)= 1.. 1 .1 n 2 6
ne 2 =9
N IR C a l : S V _ G r a s si_ 2 3 0 4 0 8 2 3 / 0 4 /2 0 0 8 1 2 . 1 7 . 3 7 A d m in ist r a t o r
12 V lid tio S e f(x)= .9 6 x+ .3 3 r= .9 2 r2 0 4 6 S e
a a n p ctra 0 3 6 0 8 0 0 7 9 = .9 6 d v(x-y)= .3 7 B S -0 3 7 ra g (x)= 2.. 8.7 n 8
0 8 8 IA (x-y)= .09 9 n e =4
10
8
6
4 Standard error
PARAMETR Range Samples
2 SEP [%]
0 Moisture 1.8 24.7 – 70.7 170
0 2 4 6 8 10 12
Original PropertyFat
Fat 0.3 1.0 – 12.1 180
NIR Seminar – Campden – October 14th 2009
20. + ! &
$ '
General calibrations developed
thanks to refernce lab
Customization according to the
needs of a specific industry
Predicted Property vs. Original Property
All Spectra
User Spectra
Calibration Spectra f(x)=0.9709x+0.1267 r=0.9854 r2=0.9709 Sdev(x-y)=0.2279 BIAS(x-y)= 0 range(x)= 2 .. 8 n=181
Validation Spectra f(x)=0.9876x+0.0605 r=0.9915 r2=0.9831 Sdev(x-y)=0.1812 BIAS(x-y)=-0.006503 range(x)=2.49 .. 7.4 n=58
8
Predicted Property Grassi
NIRCal : Sanse vergini grassi <8% 121207 13/05/2008 13.56.38 Administrator
6 Standard error
PARAMETRO Range samples
SEP [%]
4
Moisture 1.3 41.8 – 67.7 120
2
Fat 0.18 2.00 – 8.00 120
2 4 6 8
Original Property Grassi
Higher measurement
accuracy
NIR Seminar – Campden – October 14th 2009
21. / $ 1 ' *
Transflettanza
Spettrometro FT-NIR NIRLab N200
Original Property / Predicted Property N IR C al : adriaoli - bas s e c onc entraz ion impurez z e.nir impurez z e bass e conc netraz ioni 0307 13/05/2008 14.25.56 c amg
All Spectra
P re d ic t e d P ro p e rt y Im p u re z ze
Property Outlier Spectra
Validation Spectra f(x)=0.9227x+0.0132 r=0.975797
Calibration Spectra f(x)=0.9487x+0.0229 r=0.974025
User Spectra
1.00
0.75
0.50
Standard error
0.25 PARAMETER Range Samples
SEP [%]
0.00
Solvents 0.06 0.02 – 1.03 105
0.00 0.25 0.50 0.75 1.00 1.25
True Property Impurezze Impurities 0.07 0.02 – 0.99 105
NIR Seminar – Campden – October 14th 2009
22. Istituto Zooprofilattico
2 2/ 2 ( ) $ Sperimentale della Lombardia
$ e dell'Emilia Romagna
Spettrometro FT-NIR
NIRFlex N-500 with
liquids cell
Parameters for olive oil
quality evaluation
Acidity Polifenol
Tocoferol
Perox.
K232 K
K270
NIR Seminar – Campden – October 14th 2009
23. Istituto Zooprofilattico
2 2/ 2 ( ) $ Sperimentale della Lombardia
$ e dell'Emilia Romagna
Predicted Property vs. Original Property
All Spectra
Calibration Spec tra f(x )=0.9959x +0.0021 r=0.9979 r2=0.9959 Sdev(x -y)=0.0400 BIAS(x-y)= 0 range(x)=0.01 .. 2.97 n=150
Validation Spectra f(x)=0.9887x+0.0130 r=0.9918 r2=0.9837 Sdev (x-y )=0.0452 BIAS(x -y )=-0.008318 range(x)=0.04 .. 1.795 n=75
1.4
Predicted Property Olio Acidità
1.2
1.0
0.8
NIRCal : Olio oliva acidità 14/05/2008 9.25.43 Administrator
0.6
0.4
0.2
0.0
-0.2
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Original Property Olio Acidità
Standard
Coeff. N. Spt.
PARAMETEr error Range
Reg. R (cp. X 3)
SEP
Acidity 0.04 0.99 0.01 – 2.97 228
Scores vs. Scores
All Spectra Peroxide
0.2 1.8 0.97 3.2 – 43.4 237
num.
0.1
K232 0.12 0.98 1.99 – 5.27 237
NIRCal : Olio oliva acidità 14/05/2008 9.26.57 A dministrator
PC 2
-0.0
k270 0.06 0.97 0.08 – 1.49 237
-0.1
0.0001 –
K 0.0002 0.95
0.0555
237
-0.2
-0.2 -0.1 -0.0 0.1 0.2 0.3 Polifenol 30 0.70 139 - 300 93
PC 1
Tocoferol 21 0.95 7 - 282 129
NIR Seminar – Campden – October 14th 2009
24. / $ Dipartimento di Chimica e tecnologie
) $ Farmaceutiche e Alimentari - Univ. Genova
Application of differenet multi-variate
analysis techiniques to identify the
geograpichal origin of olive oil
200 samples of olive oil
“Near infrared spectroscopy and class
modelling techniques for the
geographical authentication of Ligurian
extra virgin olive oil”
Journal of Near Infrared Spectroscopy,
November 2007
NIR Seminar – Campden – October 14th 2009
26. ! ! !
Predicted Property vs. Original Property
A Spectra
ll
Calibration Spectra f(x)=0.9887x+0.0421 r=0.994357 range(x)=1.32-6.32 Sdev(x-y)=0.1134 BIAS(x-y)=1.35324e-014 n=108
Validation Spectra f(x)=0.9756x+0.0837 r=0.994598 range(x)=1.49-6.03 Sdev(x-y)=0.1096 BIAS(x-y)=0.00778034 n=52
opy of O ogeniz ato gras i 040906 07/09/2006 17.23.04 buchi
6
P d te P p rty F t
re ic d ro e a
4
s
2
m z
0
N C : c
0 2 4 6
IR al
Original Property Fat
Fat
Parameter
[%]
Samples 80
Regres. C-set 0.99
Regres. V-set 0.99
Measurement SEE C-set 0.11
time = 15 sec. SEP V-set 0.11
Range 1.32 - 6.32
NIR Seminar – Campden – October 14th 2009
27. 6 ' + 7
PARAMETHERS
Protein
Total fat
Saturated Fatty Acid
Unsaturated Fatty Acid
Lactose
NIR Seminar – Campden – October 14th 2009
28. )
8 * 2 2/ 2
Predicted Property vs. Original Property SEC/SEP
Pr e dic te d P roper ty pr ote olis i TC A 1 2 %
User Spectra Parameter Samples Samples Range [%] R
User Spectra
[%]
NIR Cal : R agusano proteolis i TCA 12% 250606 13/ 05/ 2007 23.58. 49 Adm inis trat or
Calibration Spectra f(x)=0.9300x+0.5710 r=0.964346 range(x)=0.43-22.04 Sdev(x-y)=1.1547 BIAS(x-y)=8.82512e-015 n=1207
Validation Spectra f(x)=0.9472x+0.4541 r=0.963864 range(x)=0.63-20.02 Sdev(x-y)=1.1610 BIAS(x-y)=-0.0124169 n=597
Property Outlier Spectra
20
15 Soluble C-set 408 0.11 – 6.84 0.96 0.31
10 nitrogen
5
TCA 12% V-Set 197 0.12 – 15.67 0.96 0.30
0
0 5 10 15 20 25 Proteolysis C-set 408 0.43 – 22.04 0.96 1.15
Original Property proteolisi TCA 12%
index TCA
12% V-Set 197 0.63 – 20.02 0.96 1.16
NIR Seminar – Campden – October 14th 2009
29. -
Milk
Vanilla
DIFFERENT MIXTURES OF Creme
Yogurt
ICE-CREAMS Chcocolate
Only one calibration for each parameter
Fat Protein Dry matter
SEP = 0.4% SEP = 0.10% SEP = 0.41%
NIR Seminar – Campden – October 14th 2009
30. / !
*
Spectrometer FT-NIR Buchi Nirflex N-419
Original Property / Predicted Property
Predicted Property Assorbimento a 610
All Spectra
Validation Spectra f(x)=0.9389x+0.0292 r=0.968015
Calibration Spectra f(x)=0.9497x+0.0220 r=0.974547
0.50
N IRCal : BS111.nir ASB 610 - new 0.92* 11/03/2005 14.48.06 fer g
0.45
0.40
0.35
0.30
0.25
0.30 0.35 0.40 0.45 0.50
True Property Assorbimento a 610
Original Spectra
All Spectra
0.8
0.6
NIRCal : B S111.nir A SB 610 - ne w 0.92* 11 /03/2005 14.51.44 ferg
Transmittance
0.4
Parameter Samples Range R C-Set/ V-Set SEC/ SEP
0.2
550mn 125 0.325-1.133 0.98/0.97 0.04/0.04 0.0
5000 6000 7000 8000 9000
1/cm
610nm 65 0.275-0.527 0.97/0.96 0.009/0.010
Reading at a 610nm
NIR Seminar – Campden – October 14th 2009
31. Parameter NaHCO3 CaHPO4 CaCO3 MgO
State University of Parma
Camp. 107 107 64 64
R C-Set 0.99 0.99 0.99 0.99
+ R V-Set 0.99 0.98 0.99 0.99
San Marco Plant
SEC 1.1 1.7 2.0 1.8
+
Büchi SEP 1.0 1.8 2.0 1.8
NaHCo3 CaHPO4
NIR Seminar – Campden – October 14th 2009
32. CONSTANT MONITORING
OF PRODUCTION
PARAMETERS PROCESS
Moisture Analisi dei campioni tal
Esterification quali in uscita dalla
ratio produzione
OPTIMIZATION OF
Galacturonic
PRODUCTION
Acid content PROCESS
Una sola scansione
tutti i parametri
contemporanemante
PRODUCT WITH
HIGHER QUALITY
NIR Seminar – Campden – October 14th 2009
34. */ 9
:
Regressione con set di validazione.
NIR Seminar – Campden – October 14th 2009
35. ! - $
<< ; %
P r e d ic t e d P r o p e r t y v s . O r ig in a l P r o p e r t y
Al l S p e c tr a
C a l i b ra t i o n S p e c t ra f (x )= 0 . 9 5 5 6 x + 0 . 0 7 6 5 r= 0 . 9 7 7 5 3 5 ra n g e (x )= 0 .6 4 2 -3 . 4 9 9 S d e v(x -y )= 0 . 1 1 7 7 B IA S (x -y )= -1 . 6 3 5 7 3 e -0 1 5 n = 6 0
V a l i d a t i o n S p e c t ra f (x )= 0 . 9 5 2 7 x + 0 . 0 9 3 1 r= 0 . 9 5 8 9 9 1 ra n g e (x)= 0 . 7 7 6 -2 . 3 4 5 S d e v(x -y )= 0 . 1 2 1 1 B I A S (x -y)= -0 .0 1 3 4 0 9 5 n = 2 8 Range SEC/SEP
3 Parameter Set Spectra [%] R [%]
IRCal : Vanillina quantitativ 140507 14/05/2007 22.46.48Administrator
Predicted Property Vanillina
2
C-set 60 0.64 – 3.50 0.97 0.12
1
Vanillin
a
1 2 3 V-Set 28 0.77 – 2.34 0.96 0.12
O r ig in a l P r o p e r ty V a n illin a
N
NIR Seminar – Campden – October 14th 2009