The invisible world in my infrared eyes
MVDA aided (N)IR spectroscopy and imaging
Szilveszter GERGELY Department of Applied Biotechnology and Food Science
Budapest University of Technology and Economics
Guest lecture at PCCL from BUTE NIR Spectroscopy Group 13
thof September, 2018 ● Leoben, Austria
2
The main topic of this lecture:
The invisible world in my infrared eyes
3
Newton’s experiments (over the white light)
(1704)
It was known earlier, but he wrote
it down...
4
Herschel’s experiments (below the red light)
11
thof February, 1800
Discovery of Uranus 13thof March,
1781 from court of
his house.
5
Coblentz’s experiments (the IR fingerprint)
(1905)
Ethanol CH3–CH2–OH Dimethyl ether H3C–O–CH3
C2H6O, but IR sp. ≠!
6
IR – heat – energy
An AC-130U gunship fires flares to emit masses of infrared and confuse heat seeking missiles.
http://www.wired.com/dangerroom/2012/10/infrared-obscurant/
Using IR against the
fire-and- forget missiles
The small
„stars” are more warm, than the engines
7
Developing of instruments:
from the military usage...
http://www.wired.com/dangerroom/2013/04/darpa-infrared-cameras/
A short movie?
DARPA:
„Cheetah”
Aims today:
smaller instrument,
better resolution
http://www.newsweek.com/2015/01/16/making-blood-draws-easier-friendlier-needle-296317.html 8 A short move?
(IDRC 2002 Baltimore, hospital ship) NIR 2007 Umeå (S):
painting
„H2O” on white paper by water
Developing of instruments:
... to civil applications
9
The spectrum is the sum total of changes of states:
chemical AND physical „fingerprint”
bending (deformation) changes in angle of bonds
symmetric stretching changes in length of bonds
asymmetric stretching changes in length of bonds
Vibrations of molecules
Paradigm shift with changes of millennium
NIR:
molecule / vibrational spectro- scopy
That is why we can measure moisture content
10
Moisture content below 0.5% vs. NIR (artificial fertilizer, 1 st test)
0,15 0,2 0,25 0,3 0,35 0,4 0,45 0,5
2018.03.14 18:45 2018.03.15 12:00 2018.03.16 4:00 2018.03.16 20:00 2018.03.17 12:00 2018.03.18 4:00 2018.03.18 20:00 2018.03.20 2:00
KF víztartalom Perten víztartalom
Royal Society of Chemistry (RSC): Infrared Spectroscopy (IR). (2009) 11
http://www.rsc.org/learn-chemistry/resource/res00000283/spectroscopy-in-a-suitcase-ir-student-resources#!cmpid=CMP00001302
Vibrations of molecules
NIR:
molecule / vibrational spectro- scopy
In case of MIR:
more sharp peaks S/N: 100×
Murray I.: Scattered information: philosophy and practice of near infrared spectroscopy. 12 In Near Infrared Spectroscopy: Proceedings of the 11th International Conference, Ed by Davies A.M.C., Garrido-Varo A., NIR Publications, Chichester, pp. 1–12 (2004).
Bonds have a vibrational “answer”
All we need:
water, proteins, lipids, carbo-
hydrates
13
A quick example for NIR spectra
0 1 2 3
1100 1300 1500 1700 1900 2100 2300 2500 Wavelength (nm)
OD
0 1 2 3
1100 1300 1500 1700 1900 2100 2300 2500 Wavelength (nm)
OD
Water (H
2O)
Ethanol (CH
3–CH
2–OH)
0 1 2 3
1100 1300 1500 1700 1900 2100 2300 2500 Wavelength (nm)
OD
Wine
10 mm 4 mm 1 mm 10 mm4 mm 1 mm
10 mm 4 mm 1 mm
Gergely S., Farkas K., Forgács A., Salgó A.: Quantitative and qualitative differentiations of alcoholic beverages by near infrared spectroscopy.
In Near Infrared Spectroscopy: Proceedings of the 11th International Conference, Ed by Davies A.M.C., Garrido-Varo A., NIR Publications, Chichester, pp. 569–572 (2004).
Many variables,
small differences
► MVDA
14
»chemical and physical „fingerprint”«
Dahm D.J, Dahm K.D.: The Physics of Near-Infrared Scattering.
In Near-Infrared Technology in the Agriculture and Food Industries, Ed by Williams P., Norris K., Association of Cereal Chemists Inc., St. Paul, pp. 1–17 (2004).
15
»chemical and physical „fingerprint”«
Dahm D.J, Dahm K.D.: The Physics of Near-Infrared Scattering.
In Near-Infrared Technology in the Agriculture and Food Industries, Ed by Williams P., Norris K., Association of Cereal Chemists Inc., St. Paul, pp. 1–17 (2004).
This is only glass + air / water / oil.
What are inside a grain?
16
Particle size vs. NIR
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7
950 1000 1050 1100 1150 1200 1250 1300 1350 1400 1450 1500 1550 1600 1650
A= lg (1/R)
Hullámhossz (nm)
2,0 mm 1,4 mm 1,0 mm 0,5 mm 0,3 mm
„Mixed feed”
Alja
17
How we can measure? – 3-in-1
dispersive (DS)
diode array (DA)
Fourier trans- formed (FT)
„bench-top”:
pre-
„process”:
post- dispersive
Robust; prompt data in the whole range (moving smpl, ON-LINE !!!) In this case:
pre- dispersive arrangement
Always post- dispersive
Sumriddetchkajorn S., Chakkrit Kamtongdee C.: 18 Optical penetration-based silkworm pupa gender sensor structure.
Appl. Optics51(4), 408–412 (2012)
How we can measure? – NIR LEDs
Non- destructive sexing of silkworms within pupa
Sumriddetchkajorn S., Chakkrit Kamtongdee C.: 19 Optical penetration-based silkworm pupa gender sensor structure.
Appl. Optics51(4), 408–412 (2012)
How we can measure? – NIR LEDs
20
How we can measure? – NIR lasers (tunable)
Pál MAÁK / Dept. of Atomic Physics, BME Non-
destructive sorting of watermelon
21
How we can measure? – NIR lasers (tunable)
Pál MAÁK / Dept. of Atomic Physics, BME
22
Step by step – (IQ)Q = IQ 2
IQ ► e.g. pharma
Q ► e.g. agro-food (overlapping:
e.g. moisture)
Identification
Who are you?
searching for sharp spectral differences
correlation, Euclidean distances etc.
Qualification
Where do you belong to?
define subpopulation based on more fine differences
principal component analysis (PCA), Mahalanobis distance etc.
Quantification
How many?
calibrations with reference parameters
partial least squares (PLS) method etc.
Main elements:
well-defined sample sets
accurate reference methods
reliable spectroscopic tools
exact mathematics and chemometry
Overview of calibration and validation
23
Calibration
to find multivariate correlation between spectroscopic and reference data sets
Validation
to estimate the performance of the calibration model using independent sample set or cross-validation
Overview of calibration and validation
24
Xcal ycal
reference data for calibration spectroscopic
matrix for calibration spectrum selection spectrum transformation
representative sample set
scanning
statistical evaluation
calibration modell (MLR, PLS, ANN)
Calibration
reference measurements
25
independent sample set, cross-validation
spectroscopic matrix for validation
scanning reference
measurement
reference data (measured) reference
data (predicted)
y’val
y’val
yval yval
+ ++++ ++ ++ ++ ++++++++
++++ ++++
+
+
Xval
comparison (scatter plots)
Validation
26
27
Analysis of reference data
Box & Whisker Plot Median 25%-75% Min-Max Raw Data
2,00 2,20 2,40 2,60 2,80 3,00 3,20 3,40 3,60 3,80 4,00
2,03 2,29 2,40 2,63 2,83 3,06 3,27 3,48 3,88 4,13
Elméleti hatóanyagtartalom [%] Mért (HPLC) hatóanyagtartalom [%]
1,5 2,0 2,5 3,0 3,5 4,0 4,5
3,67
Box & Whisker Plot Mean Mean±SD Mean±1,96*SD
2,00 2,20 2,40 2,60 2,80 3,00 3,20 3,40 3,60 3,80 4,00
2,03 2,29 2,40 2,63 2,83 3,06 3,27 3,48 3,67 3,88 4,13
Elméleti hatóanyagtartalom [%] Mért (HPLC) hatóanyagtartalom [%]
1,5 2,0 2,5 3,0 3,5 4,0 4,5
Bar/Column Plot of multiple variables CLX gélek analitikai eredményei in CLX gélek analitikai eredményei 5v*11c Elméleti hatóanyagtartalom [%] Mért (HPLC) hatóanyagtartalom [%]
2,00 2,20
2,40 2,60
2,80 3,00
3,20 3,40
3,60 3,80
4,00
2,03 2,29
2,40 2,63
2,83 3,06
3,27 3,48
3,67 3,88
4,13
1 2 3 4 5 6 7 8 9 10 11
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5
Scatterplot: Elméleti hatóanyagtartalom [%] vs. Mért (HPLC) hatóanyagtartalom [%]
Mért (HPLC) hatóanyagtartalom [%] = -,0495 + 1,0368 * Elméleti hatóanyagtartalom [%]
Correlation: r = ,99925
1,8 2,02,22,4 2,62,8 3,03,2 3,43,63,8 4,04,2
Elméleti hatóanyagtartalom [%]
1,8 2,0 2,2 2,4 2,6 2,8 3,0 3,2 3,4 3,6 3,8 4,0 4,2
Mért (HPLC) hatóanyagtartalom [%]
0,95 Conf.Int.
SEL= 0,017044 R2MAX
= 0,999385
Never forget:the reference laboratory has error, too : standard error of laboratory
Analysis of reference data
Differences between day-shift and
night shift:
controller is sleeping...
28 Box & Whisker Plot: nedvesség
Mean ±0,95 Conf. Interval ±SD
1 2
# 8,0 8,5 9,0 9,5 10,0 10,5 11,0 11,5 12,0
nedvesség
Histogram: nedvesség K-S d=,13839, p<,01 ; Lilliefors p<,01
Expected Normal
7,2 7,4
7,6 7,8
8,0 8,2
8,4 8,6
8,8 9,0
9,2 9,4
9,6 9,8
10,0 10,2
10,4 10,6
10,8 11,0
11,2 11,4
11,6 11,8
12,0 12,2 X <= Category Boundary 0
5 10 15 20 25 30
No. of obs.
Categ. Histogram: nedvesség
#: 1 nedvesség = 90*0,5*normal(x; 8,8744; 0,6454)
#: 2 nedvesség = 126*0,5*normal(x; 10,9651; 0,4847)
nedvesség
No of obs
#: 1 7,07,58,08,59,09,510,010,511,011,512,012,513,0 #: 2 0
10 20 30 40 50 60
Categ. Histogram: nedvesség
#: 1 nedvesség = 90*0,5*normal(x; 8,8744; 0,6454)
#: 2 nedvesség = 126*0,5*normal(x; 10,9651; 0,4847)
nedvesség
No of obs
#: 1 7,0
7,58,0 8,59,0
9,510,0 10,511,0
11,512,0 12,513,0 0
10 20 30 40 50 60
#: 2 7,0
7,58,0 8,59,0
9,510,0 10,511,0
11,512,0 12,513,0
Mathematical pretreatments
SNV / MSC;g-s / S-G D1 / D2
29
Mathematical pretreatments
Double- edged sword!
„trial and error”
(next slide)
30
Mathematical pretreatments
1,5 31 2 2,5 3 3,5 4
850 900 950 1000 1050
hullám hos s z [nm ]
A = log(1/T)
-0,2 0 0,2 0,4 0,6 0,8 1 1,2 1,4
1100 1300 1500 1700 1900 2100 2300 2500 hullám hoss z [nm ]
A = log(1/R)
d < 315 µm 315 µm < d< 710 µm
710 µm < d
transmission (diffuse) reflection
What isthe aim?
Determination of protein content or particle size?
||: Double- edged sword! :||
The reason is physical:
granulation (corn grits).
Mathematical pretreatments
32 The effect on
calibrations for glucan content of corn fiber
„trial and error”
33
c o mmo d it y G × E in s tr u me n t
c ro p y e a r
Filtering of spectra (outliers),
searching for patterns
34
Drugs – building spectrum library
65 products, 390 batches
10 / 10 vs.
100 / 10
35
65 products, 390 batches
API from China, India
(e.g.
tetracaine).
Drugs – searching for patterns (PCA)
36 liqueurs
vermouths alcohol content
extract content low
low high
high
Alcoholic/hard drinks – searching for
patterns (PQS)
37
From „off-line” to „in-/on-line”
„[...] These measurements can be:
at-line:
Measurement where the sample is removed, isolated from, and analyzed in close proximity to the process stream.
on-line:
Measurement where the sample is diverted from the manufacturing process, and may be returned to the process stream.
in-line:
Measurement where the sample is not removed from the process stream and can be invasive or noninvasive [...]”
Because NIR is quick
and non- destructive
38
The Process Analytical Technology (PAT) tools
& within it: research fields of our academic group
In order to implement a successful PAT project, a combination of three main PAT tools is essential
Multivariate data acquisition and data analysis (MVDA) tools
Usually advanced software packages which 1) aid in design of experiments (DoE), 2) collection of raw data and 3) statistically analyzing this data in order to determine what parameters are Critical Process Parameters (CPP).
Process analytical chemistry (PAC) tools
In-line and on-line analytical instruments used to measure those parameters that have been defined as CPP. These include mainly near infrared spec- troscopy (NIRS); but also include biosensors, Raman spectroscopy etc.
Continuous improvement and/or knowledge management tools
Software packages which accumulate Quality Control data acquired over time for specific processes with the aim of defining process weaknesses and implementing and monitoring process improvement initiatives.
39
ICH Q10 approach of PAT
& within it: research fields of our academic group
Is it any faster than the light?
No. So the NIR light is good enough for us,
multivariate data analysis (MVDA) [calculating, solving]
near-infrared spectroscopy (NIRS) [off-line testing]
design of experiments
(DoE) [planning, projecting]
Summary:
time is money, so we need PAT.
40
The tool of in-/on-line – the fiber probe
Yes, it is breakable...
:(
41
How we can collect spectra?
The theory ...
diffuse reflectance (R)
trasmittance (T) transflectance (TR)
attentuated total reflectance (ATR)
42
How we can collect spectra?
... and the practice (with probes)
diffuse reflectance (R)
trasmittance (T) transflectance (TR)
attentuated total reflectance (ATR)
43
An example: from cuvette to tank
scaling-up from 1 mL to 5000 L
0 2 4 6 8 10
0 3 6 9 12 15 18 21
Time (hour) Gly (g L-1)
0 2 4 6 8 10 12
0 3 6 9 12 15 18 21
Time (hour) Ace (mmol L-1)
Black dots:
reference data by hours
Colour lines:
NIR based predicted data
by 4 mins!!!
44
An example: from cuvette to tank
scaling-up from 1 mL to 5000 L
From „off-line” to „in-/on-line”
off-line / at-line on-line / in-line
the same optics = quick calibration transfer
Is it any moving part?
No:
fix grating.
Automatic change of source = continous operating digital VIS camera, IP65, OPC, modbus ASCII etc.
46 Bühler AG: NIR Multi Online Analyzer MYRG.
http://www.buhlergroup.com/global/en/products/nir-multi-online-analyzer-myrg.htm#.V34_q_NPqWg
From „off-line” to „in-/on-line”
47
Industrie 4.0
You can find this sign on your bank card...
Budapesti Kereskedelmi és Iparkamara (BKIK): Ipar 4.0 (szakértői tanulmány).
http://bkik.hu/iparitagozat/ipari-tagozat/osztalyok/bkik-vii-hirkozles-informatika-osztaly/vii-osztaly-hirei/ipar-4-0/.
Király Olívia: Ipar 4.0 avagy beléptünk a jövőbe – 5 fogalom, ami segít az eligazodásban.
http://konzervtelefon.blog.hu/2017/07/12/ipar_4_0_avagy_beleptunk_a_jovobe_5_fogalom_ami_segit_az_eligazodasban
48 Lengyel Adrienn: A Győri Gyufagyár (Várostörténeti puzzle – 8. rész)
http://gyoriszalon.hu/index.php?mact=News,cntnt01,detail,0&cntnt01articleid=2375
Industrie 1.0 – Győr / Raab / Arrabona
Match factory, smoking chimney
49 id. Konok Tamás (1951, képszám: 43348)
http://www.fortepan.hu
Industrie 2.0 – Győr / Raab / Arrabona
Textile factory
50 Bauer Sándor (1975, képszám: 109843)
http://www.fortepan.hu
Industrie 3.0 – Győr / Raab / Arrabona
Vending machine
51 AUDI AG (2015, képazonosító: 146eaa22)
https://images.audi.hu/gallery/tn2/146eaa22.jpg
Industrie 4.0 – Győr / Raab / Arrabona
It is not a vending machine
52
Industrie 4.0 – here we are
Perten Instruments: DA 7440 On-line NIR.
https://www.perten.com/Products/DA-7440-On-line-NIR/
||: = IoT internet of things :||
We need more and more dedicated solutions
53
Industrie 4.0 – resulting...
Király Olívia: Ipar 4.0 avagy beléptünk a jövőbe – 5 fogalom, ami segít az eligazodásban.
http://konzervtelefon.blog.hu/2017/07/12/ipar_4_0_avagy_beleptunk_a_jovobe_5_fogalom_ami_segit_az_eligazodasban
Macro/micro imaging
54 T (trans- mittance)
Genuie steroid tablet Fake steroid tablet
Tablets – NIR microscopy
PCA Abs.
55
Package materials – „as is”
Péter GORDON / BME EFI-Labs
56
Package materials – light microscopy
metal layer
57 Péter GORDON / BME EFI-Labs
58
Package materials – IR microscopy
59
Package materials – identification
PET
60
Package materials – identification
PA
61
Package materials – identification
PE
Biodegradable polymers (TPS/PLA)
62
Biodegradable polymers (TPS/PLA)
63