NOTE
Donkey milk powder production and properties
compared to other milk powders
Giovanni Carlo Di Renzo & Giuseppe Altieri &
Francesco Genovese
Received: 14 September 2012 / Revised: 1 January 2013 / Accepted: 7 January 2013 /
Published online: 29 January 2013
# INRA and Springer-Verlag France 2013
Abstract In order to adapt the seasonal production of donkey milk to constant
market demand, this study was aimed to define the project parameters of a pilot
spray dryer for producing soluble milk powder from donkey milk concentrate. The
concentrate (23% mean dry matter (wb)) was spray-dried using three different inlet
air temperatures (120–150–185 °C). Both cow and goat milk were used as reference
in the trials, and ascorbic acid was used as a chemical marker to evaluate thermal
damage to the powder. The thermal damage index (IDT) and insolubility index (IINS)
were used to assess the quality of the powders produced. Prediction models were
developed for each kind of milk to correlate spray-drying operating temperatures to
the IINS and IDT. The results of experimental trials were used to determine optimal
processing temperatures (both inlet and outlet air temperature) in order to obtain an
“extra-grade” milk powder from donkey milk concentrate (the maximum allowed
inlet air temperature that resulted was 173.5 °C).
Keywords Spray-drying . Donkey milk . Powder. Thermal damage . Insolubility index
1 Introduction
In recent years, the demand for replacement milks has increased considerably, because
of intolerances and allergies to cow milk (Hill and Hosking 1996; Iacono et al. 1992;
Monti et al. 2007). This applies not only to goat and mare milks, but also especially to
donkey milk, whose composition is very close to that of human milk (Hill and Hosking
1996; Monti et al. 2007; Salimei et al. 2004).
Dairy Sci. & Technol. (2013) 93:551–564
DOI 10.1007/s13594-013-0108-7
G. C. Di Renzo (*) : G. Altieri : F. Genovese
Scuola di Scienze Agrarie, Forestali, Alimentari e Ambientali, Università degli Studi della Basilicata,
Viale dell’Ateneo Lucano 10, 85100 Potenza, Italy
Donkey milk production, with particular reference to southern Italy, is mostly
performed in small livestock farms with no more than 60 animals. The average milk
production of each farm ranges between 12 and 72 kg.day−1 considering the donkeyspecific
milk production amount of 0.3–1.2 kg.day−1 (depending on the season and
the physiological status). At present, there are 12 farms in southern Italy, and they are
distributed over a wide area, with distances between farms of 80–300 km, but the
interest toward this livestock activity has been growing in the latest years.
Moreover, donkey livestock is just one of the farm activities, but it is very
important because of the possibility for donkeys to pasture on barren marginal ground
and to improve the farmer’s income. Indeed, given that the fresh pasteurized milk
market price ranges between 12.00 €.L−1 and 16.00 €.L−1, the milk is generally sold
as fresh pasteurized milk, and only a limited amount, in relation to the seasonal
variation in market demand, is set aside for powder production.
Therefore, in southern Italy, such replacement milks currently account for a marginal
line of business, with small farms dispersed over extensive rural areas. Further, as for
mares and donkeys, milk production depends on calving seasonality; herd productivity,
in terms of milk yield, does not make it viable to build industrial plants to bottle
pasteurized raw milk or UHT: this is due to the small quantity of milk produced against
the requirements of an industrial milk plant (Doreau and Boulot 1989).
As the typical composition of milk makes it very perishable, it has to be processed
into a more stable product. Among the several technologies available, spray-drying
offers many benefits (Daemen 1981; Rysstad and Kolstad 2006; Schuck 2002).
Drying extends milk shelf life (whole milk powder has a maximum shelf life of
about 6 months) and reduces weight, volume and the consequent cost of transporting
and storing the product (Daemen 1981; Rysstad and Kolstad 2006). As regards the
process involved, the milk is first concentrated by evaporation and then dried in a
drying chamber. The concentrated milk is atomized through a nozzle into droplets
that are co-current dried with a flow of hot air. During the first stage of the drying
process (i.e. constant rate drying), excess water is evaporated; in the final stage (i.e.
falling rate drying), the water bound in the solid droplets is also finally evaporated
(Chen and Lin 2005). Product quality is dramatically affected by the thermal level of
the final drying stage: if contact time between milk droplets and hot air is prolonged,
the powder may contain traces of charred particles that lower its quality (Birchal et al.
2004; Chen and Lin 2005; Pérez-Correa and Farías 1995).
The success of a new spray-drying product is related to determination of the drying
kinetics and degradation kinetics for heat-sensitive constituents (Birchal et al. 2004; De
Ritter 1976; Indyk et al. 1996; Wijlhuizen et al. 1979). As the residence time of milk
droplets in the spray-drying chamber is very short (usually no more than 30 s), the
process requires proper determination of both drying kinetics and degradation reactions
(Oakley 2004; Piatkowski and Zbicinski 2007; Straatsma et al. 1999a, b; Verdurmen et
al. 2002, 2005).
The aim of this paper is to define the operating parameters and their effects on powder
quality of a drying process/plant that should have the following characteristics: simple
and easy to manage, possible to use on livestock farms in order to avoid long transportation,
small in terms of size and production capacity, production capacity (in terms of
raw milk) in the range of 50–100 L.day−1 and low cost, in order to permit the single
producer to purchase the plant. Therefore, experimental trials have been carried out using a
552 G.C. Di Renzo et al.
pilot spray dryer for producing soluble donkey milk powder, using both cow and goat milk
as reference. Quality indices (insolubility, thermal damage and protein denaturation) were
used to evaluate the correct processing of milk, and a chemical marker (ascorbic acid) was
used by the authors to evaluate overall thermal damage.
2 Materials and methods
Milk samples were provided by local suppliers and were collected from 10 donkeys,
10 goats and 10 cows living in the same area; the mean physical and chemical values
of the raw milk samples are reported in Table 1. After milking, the milk was collected
in 20-dm3 containers, frozen and stored at −20 °C for 2 months until experimentation.
Low-temperature concentration of milk was performed in a low-pressure evaporator
pilot plant available in the laboratory. Evaporation was carried out at 55 °C and led to a
concentrated milk temperature of 35 °C. Evaporation capacity was about 2.5 kg.h−1 and
final dry matter content about 20%. Five repetitions of about 15 L of raw milk were
carried out for cow milk, three repetitions for goat and donkey milk. The equipment used
for spray-drying was a laboratory-scale FT80 spray dryer from Armfield Limited with a
pressure atomizing nozzle. The spray tower was modified to allow the internal flow to be
axisymmetric by using an internal cylindrical diaphragm that shapes an axisymmetric
discharge duct: this was done to develop the full evaporative capacity of themachine and
to standardise and increase the particles’ residence time. The main operating parameter
ranges of the FT80 spray dryer are reported in Table 2.
Spray-drying of the concentrated milk was performed in a one-stage spray dryer
pilot plant using three inlet temperatures (120, 150, 185 °C). The feed flow rate of
concentrate was set at 0.5 L.h−1, with dry matter in the range of 15–30%, according to
the machine operating parameter range; the ambient air temperature was 25 °C and
RH% about 40%; the other operating values used for the trials are listed in Table 2.
Powder samples were analysed to evaluate both physical and chemical properties:
thermal damage index (IDT); insolubility index (IINS); loss of ascorbic acid on dry basis;
powder dry matter percent (wb) (ORS%); titratable acidity, as percentage of lactic acid
content (OTA%); outlet discharge temperature (OT) and the RH% of discharged air
(ORH%) were continuously recorded during the trials. The IDTwas obtained through the
assay of total soluble undenatured proteins (the Biuret method was applied to the filtered
sample at iso-electric pH 4.8 which precipitates the denatured proteins) related to the
overall protein content calculated using the persulfate digestion method (Koroleff 1983)
as an alternative to the more time-consuming Kjeldahl. The IINS was obtained by
measuring the insoluble matter (volume expressed in millilitre) after milk powder
Table 1 Mean physical and chemical values of raw milk samples
Kind of milk Fat
(g.100 mL-1)
Protein
(g.100 mL-1)
Lactose
(g.100 mL-1)
Conductivity
(mS.cm−1)
Dry matter
(DM%) (% wb)
Cow’s milk 3.52 3.22 4.62 4.1 12.5
Goat’s milk 4.36 3.28 4.50 6.7 14.6
Donkey’s milk 0.57 1.75 6.74 2.3 10.1
Donkey milk powder production 553
reconstitution in standard conditions (International Dairy Federation 2005). Ascorbic
acid content was determined by the Official Methods of Analysis (AOAC 2002), a
standard amount of ascorbic acid (to a maximum of 5 mg.L−1) being added to the raw
milk samples before processing.
Statistical analysis was carried out by non-parametric analysis of variance with
respect to the processing temperatures and kind of milk. This was followed up by a
multiple comparison test (MCT): the MCT results on the overall paired samples were
analysed by the Mann–Whitney U test, and the familywise error rate (FWER), set to
95%significance level, was controlled by Hommel’s method (Hochberg 1988; Hommel
1988, 1989), adjusting the p values of each comparison.
3 Results and discussion
Table 3 shows the results of statistical analysis: the data report, for each milk, the
value of the IDT, IINS, the constant rate of destruction of ascorbic acid (KVITC), the
OT, the ORS% and the ORH% of the outlet air vs. the drying air temperature (in
degree Celsius). The significance is indicated by different lowercase letters along the
columns for each temperature and kind of milk and by different uppercase letters
along the rows for each milk and temperature.
The percentage variation of KVITC as the temperature increased is the same for all
milk samples. This confirms that thermal treatment performed with spray-drying was the
same for all milk samples, thus allowing comparative evaluation of global product
resistance to thermal treatment.
Absolute values of IDT show that donkey milk has a very low resistance to thermal
treatment. Its low fat content makes proteins very susceptible to denaturation with
Table 2 FT80 spray dryer main operating parameter ranges and operating values used for the trials
Parameter Range of allowed values
Feed flow rate (dm3.h−1) 0.2–7.0
Evaporated water (dm3.h−1) 0.1–3.0
Air flow rate (m3.h−1) <60
Feed dry matter percent (wb) 10–60
Hot air temperature (°C) 50–250
Nozzles (n) 2
Mean residence time (s) 0.5–5.0
Sprayed diameter of solid particles (μm) 20–200
Operating values used for the trials
Feed flow rate (dm3.h−1) 0.5
Ambient air flow rate (m3.h−1) 40
Feed dry matter percent (wb) 23
Hot air temperature (°C) 120–150–185
Nozzle Co-current pressure nozzle
Sprayed diameter of solid particles (μm) (Sauter’s mean diameter from
Nukiyama and Tanasawa’s model; Marshall 1954)
194
554 G.C. Di Renzo et al.
exposure to high temperatures. However, the difference from other kinds of milk
diminishes as the treatment temperature increases. For all milk samples, IDT values
increase with the rise in spray-draying inlet temperatures as expected. However, cow
and donkey milk present lower percentage increments than goat milk with processing
temperature. Goat milk IDT increases about 1.7 times as the processing temperature
increases by 65 °C. By contrast, IDT of donkey and cow milk increases 0.6 and 0.2
times, respectively. The milk powder insolubility index is much lower for goat and
cow milk due to the lower content of lactose which is the main cause of insoluble
compound formation due to high temperature exposure. As regards the IINS, goat
milk presents the highest increase (1.3 times) due to temperature exposure in
Table 3 The thermal damage index (IDT), insolubility index (IINS), constant rate of destruction of
ascorbic acid (KVITC), outlet discharge temperature (OT), powder dry matter percent (ORS%) and
RH% of discharged air (ORH%) vs. the inlet drying air temperature; in parentheses are shown the
percentage variation with respect to the value obtained at 120 °C
Drying air temperature Donkey’s milk Goat’s milk Cow’s milk
IDT
120 °C 45.17 (+0.0%) a A 22.18 (+0.0%) a B 56.09 (+0.0%) a C
150 °C 51.80 (+14.7%) b A 36.54 (+64.8%) b B 54.85 (−2.2%) b C
185 °C 72.58 (+60.7%) c A 60.19 (+171.4%) c B 68.07 (+21.4%) c C
IINS
120 °C 1.00 (+0.0%) aa A 0.50 (+0.0%) a B 0.58 (+0.0%) a C
150 °C 1.07 (+6.7%) aa A 0.80 (+60.0%) b B 0.72 (+24.1%) b C
185 °C 1.30 (+30.0%) aa Ab 1.13 (+126.7%) c ABb 1.18 (+103.4%) c Bb
KVITC
120 °C 2.26 (+0.0%) a A 1.94 (+0.0%) a B 2.11 (+0.0%) a C
150 °C 2.39 (+5.5%) b A 2.05 (+5.6%) b B 2.23 (+5.6%) b C
185 °C 2.54 (+12.3%) c A 2.18 (+12.5%) c B 2.38 (+12.5%) c C
OT
120 °C 70.7 (+0.0%) a A 68.6 (+0.0%) a A 71.5 (+0.0%) a A
150 °C 93.3 (+31.9%) b A 95.4 (+39.1%) b A 92.9 (+30.0%) b A
185 °C 122.9 (+73.9%) c A 121.4 (+77.0%) c A 120.4 (+68.5%) c A
ORS%
120 °C 96.2 (+0.0%) a A 97.2 (+0.0%) a A 96.3 (+0.0%) a A
150 °C 97.3 (+1.1%) a A 97.7 (+0.4%) a A 97.3 (+1.0%) a A
185 °C 97.7 (+1.5%) a A 98.7 (+1.5%) a A 97.5 (+1.2%) a A
ORH%
120 °C 5.8 (+0.0%) a A 6.3 (+0.0%) a A 6.5 (+0.0%) a A
150 °C 2.9 (−50.0%) b A 2.2 (−65.4%) b A 2.4 (−63.9%) b A
185 °C 1.3 (−77.6%) c A 1.3 (−78.7%) c A 1.1 (−83.8%) c A
a This becomes “a ab b” at p value ≥0.06, “a b b” at p value ≥0.07 and “a b c” at p value ≥0.12
b This becomes “A B B” at p value ≥0.07 and “A B C” at p value ≥0.22
The family wise error rate is set at 0.05, and the significance is indicated by different lowercase letters along
the columns for each temperature and fixed kind of milk and by different uppercase letters along the rows
for each kind of milk and fixed temperature
Donkey milk powder production 555
comparison with cow milk (1.0) and donkey milk (0.3). On the basis of these results,
donkey milk appears to have lower resistance to thermal treatment than other milk
used, though increments in processing temperatures do not produce major declines in
quality parameters, as occurs in other milks.
The OT values increase significantly with the rise in inlet air temperature.
However, due to the high spread around the mean, they do not significantly differ
as the kind of milk changes. Hence, the average OT is 70.3, 93.9 and 121.6 °C for
120, 150 and 185 °C of inlet air temperature, respectively, representing an increase
over the value at 120 °C of 33.6% and 73.1%, respectively. Moreover, these changes
are only due to the different amount of evaporated water that is directly related to
ORS% and ORH% values at the three inlet air temperatures.
Further, the ORS% values, though slightly increasing as the inlet air temperature rises,
show a non-significant difference when compared to both the inlet temperature increase
and the milk change. Such values fall in the range 96.2–98.7%, with an average of 97.3%.
As the processing temperature increases by 65 °C, the increase in ORS averages 1.4%.
In addition, ORH% values decrease significantly as the inlet air temperature
increases, but they do not significantly differ with respect to the kind of milk
involved. Hence, the average ORH is 6.2%, 2.5% and 1.2%, respectively, for inlet
air temperatures of 120, 150 and 185 °C. As the processing temperature increases by
65 °C, there is an average decrease in ORH of 80%.
The changes in OT, ORS% and ORH% are due primarily to different inlet temperatures
and subsequently to the final stage of drying when the water bound in the solid
droplets has evaporated. What determines the quality of the powder produced is the
exposure of the droplets going through the machine in terms of both heat level and
exposure time.
Moreover, although the thermal level, which severely affects milk powder quality, can
be optimized by lowering both the inlet temperature and the outlet air temperature, the last
one through the variation of the feed dry matter content, the exposure time to the thermal
level, in terms of the residence time of the droplet crossing the machine, depends both on
the machine fluid dynamic (on machine geometry, process air flow rate and temperature)
and droplet diameter (nozzle operating parameters and dry matter content) which are
difficult parameters to control. Therefore, the parameter of choice to achieve powder
quality control remains the inlet process air temperature because the outlet air temperature
depends directly on the feed flow rate of dry matter (held constant during the trials).
In Table 4 are shown the data for the linear regression between the natural
logarithm of KVITC vs. the inverse of the inlet absolute air temperature (1/Ti) and
vs. the inverse of the outlet absolute air temperature (1/To) for each milk treated. An
Arrhenius law relationship is considered between KVITC and Ti and/or To, and
therefore, a linear relationship exists between 1/Ti and 1/To. The energy of activation
(Ea) and the constant rate (ko) were estimated from the experimental data, and the
root mean square error (RMSE) of prediction is based on the “leave one out cross
validation” (LOOCV) algorithm (Picard and Cook 1984). The low RMSE value
shows that the Arrhenius law relationship fits the experimental data remarkably well.
The RMSE of the donkey milk ascorbic acid destruction rate is the same for all the
milk treated and confirms the uniformity of thermal treatment utilized for the
experimental trials. Ea proves quite similar for all the milk treated and ten times
lower than the values commonly found in the literature for milk. This result is due to
556 G.C. Di Renzo et al.
the air pressure nozzle used for spray-drying, which enhances the oxygen oxidizing
action.
Moreover, Table 5 shows the linear regression data between the natural logarithm
of the IDT vs. the inverse of the inlet absolute air temperature (1/Ti) and vs. the
inverse of the outlet absolute air temperature (1/To) for each milk. The RMSE value
shows that the Arrhenius law relationship fits the experimental data well. The RMSE
of the donkey milk thermal damage index is very low for all the milk treated and
confirms the different behaviour of the milk types to thermal treatment. The high Ea
level for goat milk confirms its very high resistance to thermal treatment.
Furthermore, Table 6 shows the linear regression data between the natural
logarithm of the IINS vs. the inverse of the inlet absolute air temperature (1/Ti)
and vs. the inverse of the outlet absolute air temperature (1/To) for each milk.
Despite the higher level of RMSE compared with the previous regression, the
Arrhenius law relationship fits the experimental data, as resulting from Adj R2.
Table 4 Linear regression between the natural logarithm of the ascorbic acid loss ratio referred to the solid
content (KVITC) vs. the inverse of the inlet absolute air temperature (1/Ti) and the inverse of the outlet
absolute air temperature (1/To) for each milk
Milk Intercept Slope Adjusted
R2
Average RMSE
% (calibration)
Average RMSE
% (prediction)
Estimated
Ea (kJ.mol−1)
Estimated
ko (s−1)
Ln(KVITC) vs. (1/Ti) (inlet drying air temperature)
Donkey 1.6345 −322.0262 0.9957 0.90 1.03 2.68 5.13
Goat 1.4889 −325.2391 0.9972 0.89 0.99 2.70 4.43
Cow 1.5790 −326.9099 0.9942 1.02 1.10 2.72 4.85
Ln(KVITC) vs. (1/To) (outlet air temperature)
Donkey 1.6872 −299.0477 0.9826 1.80 2.18 2.49 5.40
Goat 1.5230 −294.6502 0.9750 2.65 3.01 2.45 4.59
Cow 1.6877 −323.4086 0.9812 1.84 2.05 2.69 5.41
Table 5 Linear regression between the natural logarithm of the thermal damage index (IDT) vs. the inverse
of the inlet absolute air temperature (1/Ti) and the inverse of the outlet absolute air temperature (1/To) for
each milk
Milk Intercept Slope Adjusted
R2
Average RMSE
% (calibration)
Average RMSE
% (prediction)
Estimated
Ea (kJ.mol−1)
Estimated
ko (s−1)
Ln(IDT) vs. (1/Ti) (inlet drying air temperature)
Donkey 7.1209 −1,314.6777 0.9333 3.23 3.59 10.93 1,237.57
Goat 10.1371 −2,767.2580 0.9897 2.94 3.07 23.01 25,262.46
Cow 5.3523 −536.7618 0.6255 3.28 3.51 4.46 211.10
Ln(IDT) vs. (1/To) (outlet air temperature)
Donkey 7.3472 −1,225.0271 0.9282 3.35 3.89 10.19 1,551.81
Goat 10.4010 −2,497.0149 0.9588 5.87 6.47 20.76 32,892.82
Cow 5.5471 −536.9917 0.6327 3.25 3.49 4.46 256.50
Donkey milk powder production 557
However, the high RMSE values shown in the table are related to an analytical
measuring method that does not permit high measuring precision, and a very
high prediction error results.
Therefore, due to the choice of the inlet process air temperature as the quality
controlling parameter, holding the dry matter feed flow rate constant, and with the
aim of simplifying the quality evaluation of powder samples, the KVITC index was
correlated with IDT (see Table 7) and IINS (see Table 8), in order to predict the
processing temperature in relation to expected quality using the values arising from
Tables 4, 5 and 6.
On the basis of collected data, first- and second-order polynomial multilinear
regression, between the natural logarithm of the IDT vs. the natural logarithm of
KVITC, was calculated. For each milk, the RMSE of prediction is based on the
LOOCV algorithm. The results demonstrate that the second-order polynomial
Table 6 Linear regression between the natural logarithm of the insolubility index (IINS) vs. the inverse of
the inlet absolute air temperature (1/Ti) and the inverse of the outlet absolute air temperature (1/To) for each
milk
Milk Intercept Slope Adjusted
R2
Average RMSE
% (calibration)
Average RMSE
% (prediction)
Estimated
Ea (kJ.mol−1)
Estimated
ko (s−1)
1+Ln(IINS) vs. (1/Ti) (inlet drying air temperature)
Donkey 2.8269 −727.1098 0.8535 10.04 10.82 6.05 6.21
Goat 6.0890 −2,265.2037 0.9858 18.19 21.36 18.83 162.23
Cow 5.4268 −1,973.7515 0.9134 30.91 33.59 16.41 83.67
1+Ln(IINS) vs. (1/To) (outlet air temperature)
Donkey 2.9474 −675.8126 0.8438 10.37 11.28 5.62 7.01
Goat 6.3408 −2,057.1053 0.9692 26.80 31.95 17.10 208.69
Cow 6.1357 −1,971.8529 0.9207 29.58 32.60 16.39 169.98
Table 7 The first- and second-order polynomial multilinear regressions between the natural logarithm of
the thermal damage index (IDT) vs. the natural logarithm of the ascorbic acid loss ratio referred to the solid
content (KVITC) are reported for each milk
Ln(IDT) vs. Ln(KVITC)
Milk Polynomial
degree
Polynomial coefficients
(a+b×X+c×X2)
Adjusted R2 Average RMSE %
(calibration)
Average RMSE %
(prediction)
Donkey 1 0.4250, 4.1087 0.9512 2.76 3.07
2 18.8782, −38.1606, 24.1343 0.9925 1.00 1.38
Goat 1 −2.5233, 8.4994 0.9901 2.88 3.06
2 −8.2437, 24.4065, −11.0096 0.9910 2.55 3.48
Cow 1 2.7165, 1.6956 0.6762 3.05 3.26
2 23.0453, −48.7644, 31.2010 0.9693 0.90 1.10
558 G.C. Di Renzo et al.
equation could be preferred due to a lower average RMSE% prediction (except for
goat milk) (Table 7).
Table 8 shows data for the model based on a first- and second-order polynomial
multilinear regression between the natural logarithm of the IINS vs. the
natural logarithm of KVITC. The prediction RMSE is based on the LOOCV
algorithm; data show that the second-order polynomial equation could be used in
most cases, due to lower average RMSE% prediction errors, except for cow
milk. Polynomial coefficients in Table 7 and 8 can then be used to determine the
optimal processing temperature (both Ti and To) having set the IDT (≤80) and
IINS (≤1.2 mL) values, in order to obtain an “extra-grade” milk powder as
shown in Figs. 1, 2 and 3. Moreover, with regard to the upper graph within
Figs. 1, 2 and 3, the relationship existing between KVITC and 1/Ti and/or 1/To
is the same apart from a scale factor and offset with respect to the inverse of
absolute temperature which depends on the kind of milk and thermal treatment
(i.e. 1/Ti and 1/To are linearly correlated as previously asserted). Therefore, once
the kind of milk is fixed and the Ti value identified from IINS and IDT by
means of the polynomial model, the To value is found by the linear relationship
existing between 1/Ti and 1/To as 1/To=M×(1/Ti)+Q where coefficients M and
Q depend only on the thermal treatment for each kind of milk.
Application of the polynomial models to donkey, goat and cow milks to
estimate the optimal processing temperature is shown in Figs. 1, 2 and 3.
Starting from the constraints on both IDT and IINS values, two temperatures
are predicted for Ti and two temperatures for To (one Ti and one To for the IDT
target and one Ti and one To for the IINS target): the optimal Ti and To
temperatures (allowing both IDT and IINS values according to the constraints)
Table 8 The first- and 2nd-order polynomial multilinear regression between the natural logarithm of the
insolubility index (IINS) vs. the natural logarithm of the ascorbic acid loss ratio referred to the solid content
(KVITC) for each milk
(1+Ln(IINS))1/4 vs. Ln(KVITC)
Milk Polynomial
degree
Polynomial coefficients
(a+b×X+c×X2)
Adjusted R2 Average RMSE %
(calibration)
Average RMSE %
(prediction)
Donkey 1 0.5709, 0.5199 0.8739 2.28 2.43
2 3.3825, −5.9203,
3.6772
0.9195 1.69 2.30
Goat 1 −0.8255, 2.4019 0.9261 9.15 10.25
2 −10.3920, 29.0042,
−18.4119
0.9967 1.78 2.42
Cow 1 −0.5830, 1.8642 0.9109 6.89 7.64
2 2.4198, −5.5894,
4.6088
0.9111 6.61 7.90
Donkey milk powder production 559
will be the lowest among those found, in order to minimize the IINS. Starting
from Figs. 1, 2 and 3, the optimal operating parameters of the powder-producing
spray dryer were estimated (Table 9), starting from the three types of milk, after
a concentration with a mean dry matter percentage of about 23%. For the donkey
Fig. 1 The relationships between IDT vs. KVITC and IINS vs. KVITC for donkey’s milk are used to select
the optimal temperature setting, both for inlet and outlet air temperature, for spray-drying of donkey milk
concentrate in order to obtain extra-grade milk powder
560 G.C. Di Renzo et al.
milk concentrate, the estimated maximum allowed inlet air temperature is 173.4 °C
and the estimated maximum allowed outlet air temperature is 114.3 °C. Further,
the polynomial models will be used for the subsequent monitoring of the milk
powder IDT and IINS values during the plant production.
Fig. 2 The relationships between IDT vs. KVITC and IINS vs. KVITC for goat milk are used to select the
optimal temperature setting, both for inlet and outlet air temperature, for spray-drying of goat milk
concentrate in order to obtain extra-grade milk powder
Donkey milk powder production 561
4 Conclusions
The results of the trials carried out on the laboratory-scale spray dryer on concentrated
donkey, goat and cow milk permit evaluation of the IDT, IINS and KVITC
Fig. 3 The relationships between IDT vs. KVITC and IINS vs. KVITC for cow milk are used to select the
optimal temperature setting, both for inlet and outlet air temperature, for spray-drying of cow milk
concentrate in order to obtain extra-grade milk powder
562 G.C. Di Renzo et al.
measured against the inlet drying air temperature (Ti) and outlet air temperature (To).
Comparative evaluation of global product resistance to thermal treatment shows that
donkey milk has lower resistance to thermal treatment than other milk used, though
an increase in processing temperatures did not produce major reductions in quality
parameters, as found for other milks.
Moreover, in order to simplify the quality assessment of powder samples, the
KVITC index was correlated with the IDT and IINS so as to predict the processing
temperature in relation to expected quality. On the basis of the collected data, firstand
second-order polynomial multilinear regressions between the natural logarithm of
the IDT and IINS vs. the natural logarithm of KVITC were found. The polynomial
coefficients in question were used to determine the optimal processing temperature
(both inlet air temperature and outlet air temperature), having set the IDT (≤80) and
IINS (≤1.2 mL) values in order to obtain an “extra-grade” milk powder. Furthermore,
during pilot plant operations, with the aim to control and manage the powder quality,
the found polynomial models could be used to monitor in real time the milk powder
IDT and IINS parameters.
References
AOAC (2002) Official methods of analysis. Association of Official Analytical Chemists.
Birchal VS, Passos ML, Wildhagen GRS, Mujumdar AS (2004) The influence of spray dryer operation
variables on milk powder quality. In: Proceedings of the 14th International Drying Symposium (IDS
2004), vol. A, São Paulo, Brazil, pp 389–396.
Chen XD, Lin SXQ (2005) Air drying of milk droplet under constant and time-dependent conditions.
AICHE J 51(6):1790–1799
Table 9 Optimization of the processing
temperature for each kind
of milk concentrate to obtain an
extra-grade milk powder
Kind of milk Parameter Value
Common
parameters
Process air flow rate (m3.h−1) 40
Ambient air temperature (°C) 25
Ambient RH% 40
Average concentrate dry matter
percent (wb)
23
Cow’s milk Maximum inlet air temperature (°C) 187.6
Maximum outlet air temperature
(°C)
121.7
Concentrate feed flow rate (L.h−1) 0.5
Goat’s milk Maximum inlet air temperature (°C) 183.9
Maximum outlet air temperature
(°C)
121.8
Concentrate feed flow rate (L.h−1) 0.5
Donkey’s milk Maximum inlet air temperature (°C) 173.5
Maximum outlet air temperature
(°C)
114.3
Concentrate feed flow rate (L.h−1) 0.5
Donkey milk powder production 563
Daemen ALH (1981) The destruction of enzymes and bacteria during the spray-drying of milk and
whey. 1. The thermoresistance of some enzymes and bacteria in milk and whey. Neth Milk Dairy
J 35:133–145
De Ritter E (1976) Stability characteristics of vitamins in processed foods. Food Technol 30(1):48–51, 54
Doreau M, Boulot S (1989) Recent knowledge on mare milk production: a review. Livest Prod Sci 22(3–
4):213–235. doi:10.1016/0301-6226(89)90057-2
Hill DJ, Hosking CS (1996) Cow milk allergy in infancy and early childhood. Clin Exp Allergy 26(3):243–
246
Hochberg Y (1988) A sharper Bonferroni procedure for multiple tests of significance. Biometrika 75
(4):800–802
Hommel G (1988) A stagewise rejective multiple test procedure on a modified Bonferroni test. Biometrika
75(2):383–386
Hommel G (1989) A comparison of two modified Bonferroni procedures. Biometrika 76(3):624–625
Iacono G, Carroccio A, Cavataio F, Montalto G, Soresi M, Balsamo V (1992) Use of ass’ milk in multiple
food allergy. J Pediatr Gastroenterol Nutr 14(2):177–181
International Dairy Federation (2005) IDF Standard 129A. Dried milk and dried milk products. Determination
of insolubility index. International Dairy Federation, Brussels
Indyk H, Littlejohn V, Woollard DC (1996) Stability of vitamin D3 during spray-drying of milk. Food
Chem 57(2):283–286. doi:10.1016/0308-8146(95)00225-1
Koroleff F (1983) Simultaneous oxidation of nitrogen and phosphorus compounds by persulfate. In:
Grasshoff K, Eberhardt M, Kremling K (eds) Methods of seawater analysis, 2nd edn. Verlag Chemie,
Weinheimer, pp 168–169
Marshall WR (1954) Atomization and spray drying. Chemical engineering progress monograph series, vol.
50, no. 2. American Institute of Chemical Engineers, New York.
Monti G, Bertino E, Muratore MC, Coscia A, Cresi F, Silvestro L, Fabris C, Fortunato D, Giuffrida MG,
Conti A (2007) Efficacy of donkey’s milk in treating highly problematic cow’s milk allergic children:
an in vivo and in vitro study. Pediatr Allergy Immunol 18(3):258–264. doi:10.1111/j.1399-
3038.2007.00521.x
Oakley DE (2004) Spray dryer modeling in theory and practice. Drying Technol 22(6):1371–1402.
doi:10.1081/DRT-120038734
Pérez-Correa JR, Farías F (1995) Modelling and control of a spray dryer: a simulation study. Food Control
6(4):219–227. doi:10.1016/0956-7135(95)00009-G
Piatkowski M, Zbicinski I (2007) Analysis of the mechanism of counter-current spray drying. Transp
Porous Med 66(1–2):89–101. doi:10.1007/s11242-006-9024-0
Picard RR, Cook RD (1984) Cross-validation of regression models. J Am Statist Assoc 79(387):575–583.
doi:10.1080/01621459.1984.10478083
Rysstad G, Kolstad J (2006) Extended shelf life milk—advances in technology. Int J Dairy Technol 59
(2):85–96
Salimei E, Fantuz F, Coppola R, Chiofalo B, Polidori P, Varisco G (2004) Composition and characteristics
of ass’s milk. Anim Res 53(1):67–78
Schuck P (2002) Spray drying of dairy products: state of the art. Lait 82(4):375–382
Straatsma J, Van Houwelingen G, Steenbergen AE, De Jong P (1999a) Spray drying of food products: 1.
Simulation model. J Food Eng 42(2):67–72. doi:10.1016/S0260-8774(99)00107-7
Straatsma J, van Houwelingen G, Steenbergen AE, De Jong P (1999b) Spray drying of food products: 2.
Prediction of insolubility index. J Food Eng 42(2):73–77. doi:10.1016/S0260-8774(99)00108-9
Verdurmen REM, Straatsma H, Verschueren M, van Haren JJ, Smit E, Bargeman G, De Jong P (2002)
Modelling spray drying processes for dairy products. Lait 82(4):453–463
Verdurmen REM, Verschueren M, Gunsing M, Straatsma H, Blei S, Sommerfeld M (2005) Simulation of
agglomeration in spray dryers: the EDECAD project. Lait 85(4–5):343–351
Wijlhuizen AE, Kerkhof PJAM, Bruin S (1979) Theoretical study of the inactivation of phosphatase during
spray drying of skim-milk. Chem Eng Sci 34(5):651–660. doi:10.1016/0009-2509(79)85110-6
