ARTICLE

EVALUATION OF THE FOSS MILKOSCAN™ FT3 INFRARED ANALYSER

The MilkoScan™ FT3 is a FTIR mid infrared analyser manufactured by Foss Analytical A/S (Danmark) and
commercialised in France by Foss France SAS. It is used for the determination of the major and minor composition
components in liquid and semi-solid dairy products, such as milk, cream, whey, retentate, chocolated milk…
The instrument is computer controlled with the Nova™ software, which ensures the signal treatment. A mini screen
allows the start of a measurement without going through the computer. The control of the humidity in the optical
unit is achieved through a patented automatic drying system. The « zero » measurement and the standardisation
are simultaneously and automatically realised. The cleaning is defined according to the properties of the
programed matrixes. The fluidic system has been designed to handle the analysis of a full range of dairy products,
from milk to thick and viscous products like yogurt or chocolate milk with undissolved particles. However, this
means that for each product to measure, some settings are required by selecting one or more special features in
the software (sample dilution, viscosity, blend).
Within the context of the evaluation, no dilution analysis was performed.
The prediction models can be performed via the « adjustment » program which is designed as a calibration
assistant. The calculations are offered directly by the assistant.
The instrument used in this study was:
- MilkoScan™ FT3
- Serial number: 91840178
- Part: 60062098
- Environnement Windows: Windows 10 Pro
- Software: Nova™ MilkoScan™ FT3

Prediction models for many matrixes are available in the instrument.

The instrument was installed by Foss in a temperature controlled room (20-23 °C – air conditioning), without direct
sunlight.

A cleaning solution [Solution Msc W-960 Cleaning Agent (24 g Msc W-960 in 5 l of demineralised water ISO 3696
grade 3)] and a zero solution [Solution Msc Zero (5 ml Msc Zero in 5 l of demineralised water ISO 3696 grade 3]
are required. The instrument automatically controls the use of solutions: a message appears on the screen when
the contenairs are empty and have to be filled.

LES TESTS

The evaluation tests were performed in ACTALIA Cecalait’s physico-chemistry laboratory from May to September
2020. After preliminary tests of stability, the repeatability and accuracy on tank raw cow milk, tank raw ewe milk,
cream, whey and retentate for fat, true protein, dry matter and total nitrogen were evaluated.
The accuracy of the instrument was evaluated according to the following standardised methods:
- Fat in milk: Gerber method according to ISO 19662|IDF 238
- Fat in ewe milk: Acido-butyrometric method according to NF V04-155
- Fat in cream: Röse-Gottlieb method according to ISO 2450|IDF 16
- Fat in whey : Röse-Gottlieb method according to ISO 1211|IDF 1
- Protein in milk: Amido black method according to NF V04-216
- Total nitrogen: Kjeldahl method according to ISO 8968-1|IDF 20-1
- Dry matter: Oven method according to ISO 6731|IDF 21

___________________________________________________________________CECALAIT's Newsletter no. 114, 4th quarter 2020 7

 ARTICLE
1. EVALUATION OF THE SHORT-TERM REPRODUCTIBILITY

The short-term reproducibility was evaluated by analysing 4 samples of tank raw cow milk with preservative
(Bronopol 0.02 % final), with different concentration levels of fat and protein in triplicate, every 15-20 minutes to
obtain at least 20 sequences.
To evaluate the stability of the instrument, the repeatability and reproducibility were calculated by level.

Level 1 Level 2 Level 3 Level 4

Fat (g/l) 22 38 54 75
Protein (g/l) 24 35 39 55
Table 2: Content of the samples used for the short-term reproducibility evaluation

Le tableau suivant présente les résultats obtenus :

M Sr Sr(%) SR SR(%) r R
Level 1 22.121 0.0507 0.229% 0.0693 0.313% 0.141 0.192
Level 2 38.079 0.0565 0.148% 0.0849 0.223% 0.157 0.235
Fat (g/l)
Level 3 53.858 0.0611 0.114% 0.0970 0.180% 0.169 0.269
Level 4 77.526 0.0703 0.091% 0.1039 0.134% 0.195 0.288
Level 1 25.548 0.0665 0.271% 0.0792 0.323% 0.184 0.219
True protein Level 2 31.693 0.0410 0.129% 0.0546 0.172% 0.114 0.151
(g/l) Level 3 39.061 0.0657 0.168% 0.0861 0.221% 0.182 0.239
Level 4 54.343 0.0843 0.155% 0.1019 0.188% 0.233 0.282
Level 1 10.319 0.0171 0.166% 0.0193 0.187% 0.047 0.054
Level 2 12.490 0.0071 0.057% 0.0092 0.074% 0.020 0.025
Dry matter (g/l)
Level 3 14.640 0.0101 0.069% 0.0133 0.091% 0.028 0.037
Level 4 18.252 0.0114 0.062% 0.0177 0.097% 0.032 0.049
4
Table 2: MilkoScan™ FT3 stability criteria for fat, true protein and dry matter

The results for levels from 1 to 3 indicate that the standard deviation of repeatability for fat and protein are below
the limits required in ISO 8196-3|IDF 128-3 standard for milk with an average fat and protein content (0.28 g/l). For
milk with a high fat and protein content (level 4), the results indicate that the standard deviation of repeatability for
fat and protein are below the limits required in ISO 8196-3|IDF 128-3 standard for milk with a high fat and protein
content (0.56 g/l).
As no standardised value exists for dry matter, it can be noted that the reproducibility of the instrument is lower
than the reproducibility of the reference value (0.20 g/100 g).

2. EVALUATION OF THE REPEATABILITY

The repeatability of the instrument was performed by analysing:

 for tank raw cow milk: 39 samples of tank raw milk from a French plant (West of France).
 for ewe milk: 33 samples of ewe milk from a French plant (South-West of France).
 for whey: 24 samples of skimmed whey from a French plant (East-Center of France).
 for cream: 24 samples of cream from a French plant (West of France).
 for retentate: 20 samples of proteic milk retentate. 5 samples of retentate from a French plant (West of France)
are reconstituted with skimmed milk to obtain a range of 10 to 15 g of dry matter/100 g.
Bronopol was added to the samples to give a final concentration of 0.02 %. They were analysed (after heating at
40 ± 2 °C and 37 ± 2 °C for cream) in non consecutive duplicate according to the following sequence: Set 1 rep 1 –
Set 2 rep 2 - … - Set n rep n.
The repeatability was calculted using results in duplicate obtained with all the data and for the criteria:
- Fat for all the matrixes except retentate,
- Dry matter for all the matrixes,
- Protein for tank raw milk and ewe milk,
- Total nitrogen for whey and retentate.

4
M: mean; Sr and SR (Sr% andt SR%): absolute (and relative) standard deviation of repeatability and reproductibility; r and R: maximum
deviation of repeatability and reproductibility in 95 % of cases.
___________________________________________________________________CECALAIT's Newsletter no. 114, 4th quarter 2020 8
 ARTICLE

The following table presents the results obtained:

n min max M Sr Sr% r
Fat (g/l) 39 37.43 41.39 39.85 0.064 0.16% 0.177
Tank raw cow
Protein (g/l) 39 31.18 33.46 32.25 0.051 0.16% 0.141
milk
Dry matter (g/100g) 39 12.46 12.91 12.66 0.009 0.07% 0.026
Fat (g/l) 33 72.15 85.60 80.57 0.084 0.10% 0.232
Ewe milk Protein (g/l) 33 55.94 64.99 61.03 0.083 0.14% 0.230
Dry matter (g/100g) 33 18.19 19.77 19.25 0.014 0.07% 0.039
Fat (g/100g) 24 0.034 0.055 0.040 0.001 3.09% 0.003
Whey Total nitrogen (g/100g) 24 0.60 0.92 0.87 0.004 0.50% 0.012
Dry matter (g/100g) 24 5.01 6.56 6.26 0.006 0.10% 0.018
Fat (g/100 g) 24 40.81 41.76 41.42 0.072 0.17% 0.200
Cream
Dry matter (g/100g) 24 46.08 46.85 46.61 0.067 0.14% 0.186
Total nitrogen (g/100 g) 20 4.60 9.01 7.03 0.009 0.13% 0.025
Retentate
Matière sèche (g/100g) 20 10.60 14.45 12.71 0.009 0.07% 0.025
Table 3: MilkoScan™ FT3 repeatability criteria for fat, true protein, dry matter and total nitrogen
5
in tank raw cow milk, ewe milk, whey, cream and retentate

It can be noted that:

 for tank raw cow milk: for fat and true protein content, the standard deviations of repeatability are lower than the
requirements of the ISO 8196-3|IDF 128-3 and ISO 9622|IDF 141 standards (Sr < 0.14 g/l). For dry matter, the
standard deviation of repeatability is in the same order that the results for the other components and lower than the
repeatability standard deviation of the reference method (Sr = 0.036 g/100 g).
 for ewe milk: for fat and true protein content, the standard deviations of repeatability are lower than the
requirements of the ISO 8196-3|IDF 128-3 standard (Sr < 0.28 g/l). For dry matter, the standard deviation of
repeatability is in the same order that the results for the other components and lower than the repeatability
standard deviation of the reference method (Sr = 0.036 g/100 g).
 for whey: the standard deviation of repeatability is equivalent for all the criteria. As no standardised value exists
for whey, it can be noted that the standard deviations of repeatability obtained with the instrument are lower or near
of those obtained with the reference methods when they exist (Sr = 0.001 g/100g vs 0.013 g/100g for fat; 0.006
g/100g vs 0.036 g/100g for dry matter).
 for cream: the standard deviation of repeatability is equivalent for all the criteria. For fat, the relative standard
deviation (Sr%) is in accordance with the recommendations of the ISO 8196-3|IDF 128-3 standard for milk with
high content (Sr% < 0.35 %).
As no standardised value exists for whey, it can be noted that the standard deviation of repeatability obtained with
the instrument is lower or near of those obtained with the reference methods (Sr = 0.067 g/100 g vs 0.072 g/100 g).
 for retentate: the standard deviation of repeatability is equivalent for all the criteria. For total nitrogen, the relative
standard deviation (Sr%) is in accordance with the recommendations of the ISO 8196-3|IDF 128-3 standard for milk
with high content (Sr% < 0,40% for protein).
As no standardised value exists for dry matter, it can be noted that the standard deviations of repeatability obtained
with the instrument are lower or near of those obtained with the reference methods (Sr = 0.025 g/100g vs
0.108 g/100g).

3. EVALUATION OF THE ACCURACY

The accuracy of the instrument was evaluated by using the same samples than those used for the repeatablity
evaluation. The mean of the duplicates of the results obtained in the repeatability evaluation was used for the
calculation of the results. Outliers samples (samples whose regression residues are greater than 2 times the
standard deviation of deviations: P at 5 %) have been discarded.
The following table presents the results obtained:

5
N: number of the results; min and max: minimum and maximum values; M : mean of the results; Sr (Sr%): absolute (and relative) standard
deviation; r: maximum deviation of repeatability in 95 % of cases
___________________________________________________________________CECALAIT's Newsletter no. 114, 4th quarter 2020 9
 ARTICLE
n min max X Sx Sy,x Sy,x% b a
Fat (g/l) 36 37.47 41.32 39.79 0.83 0.304 0.77% 0.963 0.987
Tank raw
True protein (g/l) 37 31.24 33.45 32.26 0.46 0.092 0.28% 0.903 3.166
cow milk
Dry matter (g/100 g) 37 12.46 12.90 12.66 0.12 0.031 0.24% 0.855 1.808
Fat (g/l) 33 72.21 85.57 80.57 3.36 0.463 0.57% 1.017 -1.231
Ewe milk True protein (g/l) 30 56.01 64.95 61.24 2.09 0.207 0.34% 1.091 -6.294
Dry matter (g/100 g) 31 18.20 19.76 19.25 0.43 0.033 0.17% 1.019 -0.406
Fat (g/100g) 21 0.036 0.054 0.040 0.005 0.006 16.78% 0.466 0.015
Whey Total nitrogen (g/100g) 22 0.61 0.92 0.87 0.08 0.033 3.91% 0.879 0.087
Dry matter (g/100 g) 22 5.02 6.56 6.26 0.36 0.023 0.36% 0.922 0.455
Fat (g/100 g) 22 40.99 41.69 41.41 0.16 0.183 0.44% 0.884 0.986
Cream
Dry matter (g/100 g) 23 46.33 46.84 46.62 0.13 0.104 0.22% 0.986 1.147
Total nitrogen (g/100g) 19 4.60 9.01 7.00 1.44 0.015 0.20% 1.064 0.205
Retentate
Dry matter (g/100 g) 18 11.00 14.44 12.85 1.18 0.039 0.28% 1.159 -1.040
Table 4: MilkoScan™ FT3 accuracy criteria for fat, true protein, dry matter and total nitrogen
6
in tank raw cow milk, ewe milk, whey , cream and retentate

Raw milk - Fat Raw milk - Protein

42.00 34.00
y = 0,9627x + 0,9866 y = 0,9031x + 3,1663
41.00 33.50
Reference (g/l)

Reference (g/l)

R² = 0,8779 R² = 0,9551
40.00 Sy,x = 0,304 33.00 Sy,x = 0,092
39.00 32.50
38.00 32.00
37.00 31.50
36.00 31.00
36.50 37.50 38.50 39.50 40.50 41.50 31.00 31.50 32.00 32.50 33.00 33.50 34.00
MilkoScanTM FT3 (g/l) MilkoScanTM FT3 (g/l)

Raw milk - Dry matter

12.90 y = 0,8553x + 1,8082

Reference (g/100g)

12.80 R² = 0,9165
Sy,x = 0,031
12.70
12.60
12.50
12.40
12.40 12.50 12.60 12.70 12.80 12.90 13.00
MilkoScanTM FT3 (g/100g)

Figure 1: Relation between MilkoScan™ FT3 and reference results for fat, true protein and dry matter in tank raw cow milk

Ewe milk - Fat Ewe milk - Protein

90.00 66.00
y = 1,0172x - 1,2312 y = 1,0908x - 6,2944
Reference (g/l)

Reference (g/l)

R² = 0,9825 64.00
85.00 R² = 0,9922
Sy,x = 0,463 62.00 Sy,x = 0,207
80.00 60.00
58.00
75.00
56.00
70.00 54.00
70.00 75.00 80.00 85.00 90.00 54.00 56.00 58.00 60.00 62.00 64.00 66.00
MilkoScanTM FT3 (g/l) MilkoScanTM FT3 (g/l)

6
n, min, max: number of results, minimum and maximum values; Y: mean of the results using the reference method; Sy: standard deviation of
the results from the reference method; d, Sd: mean and standard deviation of deviations; Sy,x (Sy,x%): absolute (and relative) residual standard
deviation; b,a: slope and intercept of the linear regression.
___________________________________________________________________CECALAIT's Newsletter no. 114, 4th quarter 2020 10
 ARTICLE

Ewe milk - Dry matter

20.00
y = 1,0188x - 0,4055

Reference (g/100g)
19.50 R² = 0,9945
Sy,x = 0,033
19.00

18.50

18.00
18.00 18.50 19.00 19.50 20.00
MilkoScanTM FT3 (g/100g)

Figure 2: Relation between MilkoScan™ FT3 and reference results for fat, true protein and dry matter in ewe milk

Whey - Fat Whey - Total nitrogen

0.060 y = 0,4662x + 0,0152 1.00
y = 0,8599x + 0,0997
Reference (g/100g)

Reference (g/100g)
R² = 0,133 0.90 R² = 0,7635
0.050 Sy,x = 0.006 0.80 Sy,x = 0,039
0.040 0.70
0.60
0.030
0.50
0.020 0.40
0.020 0.030 0.040 0.050 0.060 0.40 0.50 0.60 0.70 0.80 0.90 1.00
MilkoScanTM FT3 (g/100g) MilkoScanTM FT3 (g/100g)

Whey - Dry matter

7.00 y = 0,9221x + 0,4565
Reference (%)

R² = 0,9926
6.00
Sy,x = 0,028
5.00

4.00

3.00
3.00 4.00 5.00 6.00 7.00
MilkoScanTM FT3 (%)

Figure 3: Relation between MilkoScan™ FT3 and reference results for fat, total nitrogen and dry matter in whey

Cream - Fat Cream - Dry matter

42.60 47.70
Reference (g/100g)

Reference (g/100g)

42.10
47.20
41.60 y = 0,9862x + 1,1473
y = 0,8838x + 5,4021 R² = 0,619
R² = 0,3888 46.70
41.10 Sy,x = 0,101
Sy,x = 0,183
40.60 46.20
40.60 41.10 41.60 42.10 42.60 46.20 46.70 47.20
MilkoScanTM FT3 (g/100g) MilkoScanTM FT3 (g/100g)

Figure 4: Relation between MilkoScan™ FT3 and reference results for fat and dry matter in cream

___________________________________________________________________CECALAIT's Newsletter no. 114, 4th quarter 2020 11

 ARTICLE

Retentate - Total nitrogen Retentate - Dry matter

16.00 y = 1,1588x - 1,0403
y = 1,0636x + 0,205
Reference (g/100g)

Reference (g/100g)
10.00 15.00 R² = 0,9992
R² = 0,9999
Sy,x = 0,015 14.00 Sy,x = 0,039
8.00
13.00
6.00 12.00
11.00
4.00 10.00
4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00
MilkoScanTM FT3 (g/100g) MilkoScanTM FT3 (g/100g)

Figure 5: Relation between MilkoScan™ FT3 and reference results for total nitrogen and dry matter in retentate

Concerning the relation between MilkoScan™ FT3 and the reference results, it can be noted that:
 for tank raw cownmilk:
The residual standard deviations of linear regression obtained are equal to 0.304 g of fat/l, 0.092 g of protein/l, and
0.031 g of dry matter/100 g. The estimation precision of the instrument is therefore ± 0.61 g/l (± 2 x 0.304 g/l) for
fat, ± 0.18 g/l (± 2 x 0.092 g/l) for true protein and ± 0.062 g/100 g (± 2 x 0.031 g/100 g) for dry matter.
 for ewe milk:
The residual standard deviations of linear regression obtained are equal to 0.463 g of fat/l, 0.207 g of protein/l, and
0.033 g of dry matter/100 g. The estimation precision of the instrument is therefore ± 0.93 g/l (± 2 x 0.463 g/l) for
fat, ± 0.41 g/l (± 2 x 0.207 g/l) for true protein and ± 0.066 g/100 g (± 2 x 0.033 g/100 g) for dry matter.
 for whey:
The residual standard deviations of linear regression obtained are equal to 0.006 g of fat/100 g, 0.033 g of total
nitrogen/100 g, and 0.023 g of dry matter/100 g. The estimation precision of the instrument is therefore ± 0.012
g/100 g (± 2 x 0.006 g/100 g) for fat, ± 0.066 g/100 g (± 2 x 0.033 g/100 g) for total nitrogen and ± 0.046 g/100 g
(± 2 x 0.023 g/100g) for dry matter.
 for cream:
The residual standard deviations of linear regression obtained are equal to 0.183 g of fat/100 g and 0.104 g of dry
matter/100 g. The estimation precision of the instrument is therefore ± 0.37 g/100 g (± 2 x 0,183 g/100 g) for fat
and ± 0.21 g/100 g (± 2 x 0.104 g/100g) for dry matter.
 for retentate:
The residual standard deviations of linear regression obtained are equal to 0.015 g of total nitrogen/100 g and
0,039 g of dry matter/100 g. The estimation precision of the instrument is therefore ± 0.030 g/100 g (± 2 x 0.015
g/100 g) for total nitrogen and ± 0.078 g/100 g (± 2 x 0.039 g/100g) for dry matter.

4. CONCLUSION

We can conclude that the short-term stability of the instrument is in accordance with the requirements of the ISO
8196-3|IDF 128-3 standard.
Concerning the repeatability of the instrument, the results for fat and true protein in milk are in conformity with limits
of the ISO 9622IDF 141 standard. For the other products and criteria, the results obtained are in accordance with
the recommendations of the ISO 8196-3|IDF 128-3 standard, or lower or near to the repeatability limits of the
corresponding reference method.
Concerning the precision, no standardised requirements exist for the products tested during this evaluation (tank
raw milk, ewe milk, whey and cream).

According to the evaluation report of the MilkoScan™ FT3 – M. ESTEVES, A. OUDOTTE and Ph. TROSSAT – May-September
2020

___________________________________________________________________CECALAIT's Newsletter no. 114, 4th quarter 2020 12