Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Contents lists available at ScienceDirect

Innovative Food Science and Emerging Technologies

journal homepage: www.elsevier.com/locate/ifset

Closed-loop spray drying solutions for energy efficient powder production T

⁎
S.N. Moejes, Q. Visser, J.H. Bitter, A.J.B. van Boxtel
Biobased Chemistry and Technology, Wageningen University, P.O. Box 17, 6700 AA Wageningen, The Netherlands

A R T I C L E I N F O A B S T R A C T

Keywords: This paper introduces a closed-loop dryer system to reduce the energy consumption for milk powder production.
Spray drying The system is based on a monodisperse droplet atomizer which reduces the amount of fines in the exhaust air,
Milk powder and allows dehumidification and recirculation of the air over the dryer. In this way the latent and sensible heat
Air dehumidification from the dryer exhaust are recovered. Two adsorbent systems for dehumidification are discussed; a membrane
Zeolite
contactor with a liquid desiccant, and a zeolite sorption wheel. Four configurations for closed-loop spray drying
Membrane contactor
Pinch analysis
are simulated and optimized. By heat integration of the adsorber-regenerator system with the dryer and pre-
ceding concentration step, the energy consumption is significantly reduced to 4.9 MJ heat per kg milk powder.
The final heat integration solutions were obtained by simultaneous optimization of the operational conditions
and the heat exchanger network based on pinch analysis.
Industrial relevance: Drying is an energy intensive operation in processing. To comply with the upcoming reg-
ulations that arise from the EU goals for sustainable development, the energy consumption of drying processes
should be reduced drastically. Emerging technologies are the key for the next step in energy efficiency im-
provement. A closed-loop spray drying system for milk powder production is simulated and optimized in this
work. The proposed technologies are: monodisperse droplet drying, membrane contactor and a zeolite wheel. By
applying air dehumidification and heat integration the latent and sensible heat are recovered from the exhaust
air. The energy consumption for milk concentration and spray drying has the potential to be lowered from 8.4 to
4.9 MJ heat per kg milk powder. Although milk powder has been considered, the proposed system is also ap-
plicable to other food products, as well as in the (bio)chemical, pharmaceutical and paper industry.

1. Introduction are needed (Moejes & van Boxtel, 2017). An important development to
meet the future requirements on the energy consumption in milk
Thermal processes are responsible for 29% of the total energy powder production is the energy efficient concentration of milk prior to
consumption in the food industry (Okos, Rao, Drecher, Rode, & Kozak, the spray dryer. Walmsley, Atkins, Walmsley, and Neale (2016) show
1998). Spray drying systems are the main energy consumer in powder that application of mechanical vapor recompression (MVR) with op-
production. The energy efficiency of spray drying has been improved timal heat integration can potentially reduce the energy consumption
over the last decades by the introduction of multi-stage drying with per kilogram of powder by 78%. It is also demonstrated that with
fluidized bed dryers, air pre-treatment, heat pumps, and the optimiza- current state of the art technology the energy consumption can be re-
tion of the processes and operational conditions to a full extend duced to 5.2 MJ per kg of powder, and by an ultimate process and
(Ramirez, Patel, & Blok, 2006; Walstra, Geurts, Noomen, Jellema, & utility integration even to 2.5 MJ per kg of powder (Walmsley, Atkins,
Boekel, 1999; Westergaard, 2004). Furthermore heat recovery and in- Walmsley, Philipp, & Peesel, 2017).
tegration of current spray drying processes, like the production of milk MVR has a high potential to reduce the energy consumption, but
powder, has been studied (Atkins, Walmsley, & Neale, 2012; Walmsley, concerns the milk concentration step. The potential for energy recovery
Walmsley, Atkins, Neale, & Tarighaleslami, 2015). However, to reach from the dryer exhaust air, which contains significant amounts of latent
the energy ambitions of the EU to reduce the energy consumption with and sensible heat, is not yet fully exploited. Sensible heat recovery from
27% in 2030, and even more in the following decades (European the exhaust has been proposed by Atkins, Walmsley, and Neale (2011)
Commission, 2011), incremental improvements in energy efficiency, by and Golman and Julklang (2014) for spray drying. The application of
additional optimization, do not satisfy this requirement. Large steps heat exchangers and proper heat integration resulted in a reduction of
forwards, which can be achieved by introducing emerging technologies, the hot utility up to 21% (Atkins et al., 2011). In practice this option is

⁎
Corresponding author.
E-mail address: [email protected] (A.J.B. van Boxtel).

https://doi.org/10.1016/j.ifset.2018.01.005
Received 16 August 2017; Received in revised form 12 January 2018; Accepted 15 January 2018
1466-8564/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

still a challenge due to fine powder particles (fines) present in the ex- external energy requirements. Atuonwu, van Straten, van Deventer, and
haust air, which cause fouling in the heat exchangers used for heat van Boxtel (2011) applied a simultaneous approach based on the work
recovery. Filter systems are needed but result in additional energy loss. of Duran and Grossmann (1986), where pinch analysis and optimiza-
Monodisperse droplet atomizers combined with proper airflow patterns tion of operational conditions were combined in one step. By con-
and well-designed drying chambers, have the potential to operate sidering streams and temperatures as variables, the pinch point can be
without these fines. Both Deventer, Houben, and Koldeweij (2013) and shifted resulting in an additional heat recovery. For a low-temperature
Rogers, Fang, Qi Lin, Selomulya, and Dong Chen (2012) showed that a drying system with zeolites the simultaneous optimization resulted in a
spray drying system with monodisperse droplet atomizers based on 13% improvement in energy consumption compared to the results ob-
inkjet technology results in a very narrow particle size distribution after tained with a standard step-wise pinch analysis (Atuonwu et al., 2011).
drying. Monodisperse droplet drying is now applied at pilot plant scale In line with this, Walmsley, Walmsley, Atkins, and Neale (2013) found
(Debrauwer, 2016). With successful upscaling this technology reduces that applying variable temperatures in pinch analysis for spray drying
the current limitations of sensible heat recovery from the exhaust air systems specific heat recovery can be increased by 30%.
and, in combination with air dehumidification, it offers also the pos- Next to the application of the existing methods for the reduction of
sibility for closed–loop drying and latent heat recovery. Dehumidifi- energy, like multi-stage drying with fluidized bed, air pre-treatment,
cation in closed-loop spray drying, which can recover both sensible and heat pumps, and heat exchange between inlet and exhaust air, emer-
latent heat, is a new instrument for energy saving and will be assessed ging technologies are needed to further reduce the energy consumption
in this study. in spray drying processes. In this work we discuss the potential for
Two technologies are proposed for air dehumidification in combi- energy reduction by air dehumidification in closed-loop spray drying
nation with closed-loop drying i.e., 1) contact-sorption system with a and compare the results with the common practice in milk powder
solid adsorbent (Atuonwu, van Straten, van Deventer, & van Boxtel, production. By combining emerging technologies different new closed-
2012), and 2) membrane contactor with a liquid desiccant (brine) loop spray drying configurations are proposed to increase energy effi-
(Isetti, Nannei, & Magrini, 1997). Both systems are already proven in ciency. Simultaneous optimization of the operational conditions and
other fields, and have potential to be implemented in closed-loop spray the heat exchanger network is applied to find an optimal process de-
drying. Contact-sorption systems use solid adsorbents with a high af- sign.
finity for water. Zeolite and silica are the most used adsorbents for these
systems. Since the spray dryer exhaust air has a temperature in the 2. Process description
range of 60 to 90 °C, zeolites are expected to be more effective com-
pared to silica (Boxtel, Boon, Deventer, & Bussmann, 2012). Application Spray drying systems are intensively used in the dairy industry. Milk
of zeolites for air dehumidification in low-temperature dryers has been powder is, therefore, used as model product. The processing steps for
discussed before, and shows a significant potential for energy savings standardized milk powder production are heating, concentrating and
(Djaeni, van Straten, Bartels, Sanders, & van Boxtel, 2009; drying. The focus in this study is on the dryer section. In the closed-loop
Goldsworthy, Alessandrini, & White, 2015). Likewise, zeolites are used spray dryer system a surplus energy stream is created from the re-
for the pre-treatment (dehumidification) of ambient air prior to drying. generation of the adsorbent. To be energy efficient the surplus energy
Advantage of this pre-treatment is the increase in dryer capacity and has to be exploited elsewhere in the production process. The surplus of
improved controllability of the dryer conditions (Boxtel et al., 2012). steam is the best used multi-effect evaporators. Mechanical vapor re-
Membrane contactors are currently used for selective separation of compression (MVR) systems or combinations of thermal vapor re-
gasses (Li & Chen, 2005) and in air conditioning systems (Bergero & compression (TVR) with MVR use no or only a minor amount of steam.
Chiari, 2010; Kneifel et al., 2006). In air conditioning systems moist air Therefore a pre-heater and a multi-effect evaporator are included as a
is separated by a hydrophobic membrane from a saturated brine (i.e. heat sink for the surplus energy of the drying process in this work. In
lithium bromide, lithium chloride, magnesium chloride, calcium Fig. 1, the dryer is given with the loops for air dehumidification, while
chloride, or a combination) (Abdel-Salam, Ge, & Simonson, 2013). The the pre-heater and multi-effect evaporator are given as two unit op-
partial vapor pressure difference over the membrane is the driving force erations. The concentrated milk is atomized with a monodisperse nozzle
in these systems, and only water vapor passes through the membrane. and the spray dryer is operated in closed-loop with air dehumidifica-
Successful applications of membrane contactors for air dehumidifica- tion. The section for air dehumidification consists of an adsorber (either
tion at ambient temperatures are already reported (Isetti et al., 1997; membrane contactor or zeolite wheel), regenerator, and a cooling/
Jain, Tripathi, & Das, 2011; Kneifel et al., 2006), but the potential of heating unit. The dehumidified air is heated/cooled to the drying
these systems for air dehumidification at elevated temperatures has not temperature before reentering the spray dryer.
previously been quantified. The total system is split into subunits, as represented in Fig. 1, and
The potential of zeolites and membrane contactors for the dehu- for each subunit overall steady-state mass and energy balances are used:
midification of the recycled air in spray dryers is investigated in this
Hl / s = Fl / s ∙ (cp, l / s ∙xl / s + cp, w ∙x w ) Tl / s (1)
work. Both dehumidification systems have in common that heat is re-
leased when water vapor is adsorbed, and external energy is required Ha = Fa (cp, a + ya ∙cp, v ) Ta (2)
for the regeneration of the adsorbent. Air dehumidification is only ef-
−1
fective when the heat released at adsorption and the remaining heat where H is the enthalpy of the flows (kJ h ), F the mass flow of liquid
from the regeneration are used elsewhere in the system (Atuonwu et al., (l), solids (s) and air (a) (kg h−1), T the temperature of the flows (°C), xw
2012; Djaeni, Bartels, Sanders, van Straten, & van Boxtel, 2007). This the water and ya the vapor content of the flows (kg kg dry air−1), and cp
makes heat integration a prerequisite for these proposed configurations the heat capacities of water (w) and vapor (v) (kJ kg−1 °C−1).
to be energy efficient.
Pinch analysis a well-established method for heat integration and 2.1. Evaporator
the design of heat exchanger networks to minimize external utilities
(Kemp, 2007). The pinch approach is a step-wise procedure in which Milk is first heated up in a pre-heater, and subsequently con-
operational conditions, like flows and temperatures, are optimized first. centrated in a multi-effect evaporator. The energy requirements for
Subsequently, given those optimized conditions, a heat exchanger heating follow from Eq. (1), and the size of the heat exchanger (Ahex, in
network is defined according to the pinch rules. The drawback of this m2) is based on the following equation. In which Qhex is the amount of
approach is the optimized operational conditions are not necessarily the energy exchanged, U is the heat transfer coefficient, and ΔT the tem-
optimal conditions for the heat exchanger network with the minimal perature difference.

25
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Fig. 1. Schematic representation of a closed-loop dryer for milk powder production.

Qhex concentrate (Hcm), and the exhausted air (Ha, out) and milk powder (Hp)
Ahex =
U ΔThex (3) is given by:
The mass and component balances for the evaporator are: Ha, in + Hcm = Ha, out + Hp (8)
Fm = Fcm + Fv,1 + Fv,2 + …+Fv, n (4) The energy required for operating pumps, fans, nozzle, etc., and
heat losses are not taken into account in this study.
Fm ∙x m, in = Fcm ∙x m, out (5)
where Fm and Fcm are respectively the milk feed and concentrate flow 2.3. Membrane contactor
(kg h−1), xin and xoutthe solids concentration in the feed and con-
centrate (kg kg−1), and Fv, i the amount of evaporated water in each In the membrane contactor the air and brine streams are separated
effect (kg h−1). The number of effects (n) is set to 7, and milk is con- by a hydrophobic membrane, which is only permeable for water vapor.
centrated to 50% total solids. The difference in vapor pressure over the membrane is the driving force
The energy requirements for the evaporator result from the fol- for vapor transport through the membrane. The brine heats up due to
lowing equation (based on Eq. (1)): the released heat of condensation, and therefore a cooling system is
integrated in module as shown in Fig. 2. The brine and cooling water
Hm,i − 1 + Hvap,i − 1 = Hvap,i + Hm,i + Hcw,i (6)
flow are in co-current, while the brine and air are in counter current
where Hm, i−1 is the enthalpy flow of the milk feed to effect i, and Hm, i configuration. The applied mass and energy balances are:
the enthalpy flow of the concentrated milk flow leaving effect i, Hvap, iis Fa ∙ya, in + Fb ∙x w, in = Fa ∙ya, out + Fb ∙x w, out (9)
the enthalpy flow of the vapor (or steam in case of the first effect), and
Hcw, i is the enthalpy flow of the condensate stream. Ha, in + Hb, in + Hc, in = Ha, out + Hb, out + Hc, out (10)

Pa, in − Pb, out = Pa, out − Pb, in (11)

2.2. Monodisperse spray dryer
where Fa and Fb are the flows of air and brine in the adsorber (kg h−1),
Water and energy are exchanged in the dryer due to the contact of ya and xw the vapor and water concentrations (kg kg−1), Ha, Hb, and Hc
hot air with atomized droplets. The mass balance for water is: the enthalpy of air, brine and cooling water streams (kJ h−1), Pa and Pb
Fa ∙ya, in + Fcm ∙x w, in = Fa ∙ya, out + Fp ∙x w, out are the partial vapor pressures in the air and brine. The vapor pressure
(7)
is based on the Antoine's equation (Patil, Tripathi, Pathak, & Katti,
where Fa and Fcm are the flows of dry air and concentrated milk 1990).
(kg h−1), ya is the concentration of vapor in the air flow, and xw is the The membrane contactor uses lithium bromide as brine solution,
water content of the milk flow (both in kg kg−1). The moisture content and the physical properties are taken from Florides, Kalogirou, Tassou,
of the inlet air (ya, in) depends on the level of dehumidification achieved and Wrobel (2003), Iyoki, Yamanaka, and Uemura (1993), and Patil
in the adsorber. The moisture content of the exhaust air (ya, out) is re- et al. (1990). Heat transfer between air, membrane, and brine, are
lated to the relative humidity of the exhaust air, which is set at 10%. based on the Maxwell-Stefan equation (Krishna & Wesselingh, 1997),
The overall energy balance for the inlet air (Ha, in) and milk the membrane properties are taken from Zhang (2006). The membrane

Fig. 2. Schematic representation of a membrane contactor

module with cooling water channel.

26
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

contactor is operated in a continuous mode with a moderate increment

of the water concentration in the brine between in- and outlet of the
module. Therefore, in the countercurrent operated membrane contactor
the vapor pressure difference is assumed to be equal at every place. The
brine temperature is related to the vapor pressure difference over the
membrane (Eq. (11)).
After passage through the membrane contactor the brine is re-
generated in a continuous operation. Two options for brine regenera-
tion are considered: 1) water evaporation in a two-effect evaporator, or
2) by superheated steam. The energy needed for brine regeneration by
the two-effect evaporator (Hst) is given by:
Fb Δx mc
Hst = ∙Hv1/ Fst
1 + Hv1/ Hv2 (12)
−1 Fig. 4. Superheated steam (SHS) cycle in adsorber/regenerator loop.
where Hv, i is the heat of evaporation (kJ h ) in effect 1 and 2, and
depends on the temperatures in the effects. Fst is the amount of steam
required for the first effect (kg h−1), Fb the brine flow (kg h−1), and Ha, in + Hz, in − Hdes = Ha, out + Hz, out (16)
Δxmc is the concentration difference between the brine in and out. −1
where Fa and Fz are the flows of air and zeolite (kg h ), ya and xw the
Boiling point elevation related to the lithium bromide concentration is
taken into account. vapor and water concentrations (kg kg−1), Ha and Hz and the enthalpy
of air and zeolite streams and Hdes the desorption enthalpy (kJ h−1).
For regeneration with superheated steam a second membrane con-
tactor unit is needed, which operates in the same way as the membrane The sorption isotherm for a commercial zeolite (CeCA, 4A) is used
(Boxtel et al., 2012).
contactor for dehumidification, only without a cooling water section.
Due to the low vapor pressure of the superheated steam, water vapor Hot air or superheated steam can be applied for zeolite regenera-
tion. The balances for regeneration of both regeneration systems are
passes from the brine to the superheated steam. The energy balance for
brine regeneration by superheated steam is: based on Eq. (16), and the balance for the cooling/heating is expressed
in Eq. (14). The superheated steam flow is recycled after usage in the
Hshs, in + Hb, in = Hshs, out + Hb, out (13) regenerator as shown in Fig. 4. Part of this flow is upgraded and reused,
the other part is a surplus flow and is exploited elsewhere in the system.
where the enthalpy of the superheated steam (Hrm) depends on its
temperature (Tshs) and pressure (Lachkov, Lysenkov, & Mamonov,
1999). After regeneration with superheated steam the temperature of 3. System optimization
the brine is adjusted to the optimal temperature for adsorption by either
heating or cooling. The energy balance for the cooling/heating system 3.1. Process configurations
is:
Four different configurations (see Table 1) are optimized, evaluated,
Hb, in + Qheat = Qcool + Hb, out (14) and compared to a conventional spray dryer system without air dehu-
midification and recirculation. The energy consumption per kilogram
where Qheat and Qcool is the required energy for heating or cooling
produced milk powder for these configurations are compared to that of
(kJ h−1).
a conventional milk powder production system. In this conventional
system a pre-heater, 7-stage evaporator and spray dryer are used and
2.4. Zeolite waste energy from the evaporator is recovered. The pre-heater and
multi-stage evaporator are the same for all configurations. General data
The zeolite sorption system consists of a wheel system with three for the system are given in Table 2.
separated sections (see Fig. 3): 1) adsorption, 2) regeneration, and 3)
cooling or heating. Each section is modeled as an individual unit. For
3.2. Operational conditions
the adsorption section the following balances apply:
Fa ∙ya, in + Fz ∙x w, in = Fa ∙ya, out + Fz ∙x w, out (15) Main objective for a closed-loop dryer is to minimize the external
energy input. The decision variables that affect the energy input for all
configurations are the operating temperatures of the evaporator (Tst, evap
and Tcm), the dryer inlet temperature (Ta, in), the moisture content of the
air at the dryer inlet (ya, in), and the temperatures of the regeneration
medium entering and exiting the regenerator (Trm, in, and Trm, out). The
concentration difference of the brine between the in- and outlet of the
membrane contactor (Δxb), and the temperature difference between the
brine and the air side (ΔTmc) are decision variables specific for the
membrane contactor. The inlet temperature of the zeolite for the

Table 1
Overview of the considered process configurations.

Configuration Adsorber Regeneration medium

Conventional None None

1 Membrane contactor Evaporator
2 Membrane contactor Superheated steam
3 Zeolite Hot air
Fig. 3. Schematic representation of a zeolite wheel with the adsorption, regeneration, and
4 Zeolite Superheated steam
heating/cooling section.

27
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Table 2
⎛ ⎞
Fixed process variables. Qex = min ⎜∑ QiHU ⎟
⎝ i ⎠
Variable Value s. t . mass and energy balances (Eq.1 − 15)
Fm (kg h−1) 100,000 lower bound < decision variable < upper bound (17)
Tm (°C) 10
xm, in (kg kg−1) 0.09 where Qex is the total amount of required external heating (MJ heat kg
xm, out, evap (kg kg−1) 0.5 milk powder−1), which is based on the sum of hot utilities (QiHU) of all
xw, out, dryer (kg kg−1) 0.035
the streams (i) and to total amount of product produced (Fp). The boiler
efficiency is not taken into account. Ambient air or ground water are
sufficient for cooling, active chilling is not necessary, therefore cooling
adsorption section (Tz, in) is of importance for the zeolite system. The
not included in the objective. Simulation and optimization were per-
upper and lower bounds for each decision variable are related to pro-
formed in MATLAB R2014b using the genetic algorithm solver.
duct and process constraints, and are listed in Table A.2.

3.3. Heat integration design 3.4. Economic evaluation

In pinch analysis hot and cold composite curves are identified from To address the industrial viability of the proposed technologies the
the defined target temperatures and flows, and subsequently the heat utility and investment cost are estimated for the energy optimal sce-
recovery is estimated for a given temperature difference at the pinch narios. Each configuration is evaluated on the total annualized costs
point. In the applied procedure the target temperatures and flows are (TAC), consisting of the equipment costs (Cequip) and utility costs (Cutil)
not fixed at beforehand, but estimated in a simultaneous optimization for every process j and yearly operating time (t).
of the operational conditions and the heat exchanger network
(Atuonwu et al., 2011). By altering the flows and temperatures in the
TAC = ∑ Ceq,j + ∑ Cut,j t
j j (18)
system during the optimization steps, the hot and cold composite curves
are adjusted to minimize the energy consumption.
i (1 + i)nlife
The hot and cold streams for all process configurations are given in Ceq = Cinv ∙LF∙
(1 + i)nlife − 1 (19)
Table 3. For each configuration a different number of hot and cold
streams is available. It should be noted that, depending on the opera- in which LF is the Lang factor to include the costs for building and
tional conditions in the system, some streams could then be either hot installing the equipment, i is the interest rate, nlife is the lifetime, and
or cold streams. For example, depending on the operational conditions Cinv is the equipment initial investment costs. The investment cost are
of the membrane contactor the brine needs cooling or heating after derived from standard engineering data in Table 4. For up- and down
regeneration. The switches between heating and cooling are included in scaling from the given equipment dimensions, the following equation is
the procedure to optimize the operational conditions. used:
Fig. 5 shows the generic heat exchanger network for the different
configurations, including the switch streams. The number of hot and neq
Aeq ⎞
cold streams affects the size of the heat exchanger network. For the Cinv = Cref ⎜⎛ ⎟
A
⎝ ref ⎠ (20)
pinch point an individual minimum temperature difference was used,
10 °C for air streams and 5 °C for liquid and steam streams (Kemp, with Cref the costs for a reference installation with dimension Aref as
2007). The optimization problem is defined as follows: given in Table 4, Aeq the actual equipment dimension in required in the
configurations, and neq the scaling factor. Membrane contactors operate
Table 3
on the same principle as membrane distillation units, therefore the
Overview of all feasible hot and cold streams for the four configurations and the con-
ventional configuration.
membrane costs are based on a membrane distillation unit. Industrially
used zeolite wheels have a diameter up to 4 m, this means multiple
1 2 3 4 Conventional wheels are placed if a larger area is required.
The utility costs consist of the heating and cooling cost required for
Hot streams
H1 Vapor from the last effect of the milk X X X X X
all processes:
evaporator
H2 Vapor from the last effect of evaporator X
Cut = Cheat ∙Qheat + Ccool ∙Qcool (21)
for brine regeneration
H3 Cooling water from the membrane X X in which Cheat is the heating cost, and Ccool is the cost for cooling.
contactor Electrical usage of pumps, fans etc. is not taken into account. Table 5
H4 Superheated steam surplus from X X lists the used economic data.
regenerator
H5 Hot air after regeneration X

Cold streams 4. Results and discussion

C1 Milk feed X X X X X
C2 Steam to the 1st effect of the milk X X X X X
4.1. Conventional configuration
evaporator
C3 Steam to the 1st effect of the evaporator X
for brine regeneration The energy requirement for the optimized conventional steam
C4 Superheated steam reheating after X X driven configuration is 8.4 MJ heat per kg milk powder. For heat re-
regenerator covery, vapor from the last evaporator effect is used to pre-heat the
C5 Hot air for regenerator X
milk to the evaporator and to pre-heat the drying air (Fig. 6). The en-
C6 Ambient air to the dryer X
ergy consumption of the conventional configuration is in agreement
Hot/cold streams
with reported values (Depping, Grunow, van Middelaar, & Dumpler,
H6/C7 Air from adsorbent to the dryer X X X X
H7/C8 Adsorbent from regenerator X X X X 2017; Finnegan, Goggins, Clifford, & Zhan, 2017; Ramirez et al., 2006;
Yildirim & Genc, 2017).

28
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Fig. 5. Generic heat exchanger network for the different configurations. The number of hot (n) and cold (m) streams depend on the configuration.

4.2. Configuration 1 – membrane contactor with evaporator Table 5

Economic data.
In this optimized and heat integrated system a moderate energy
Item Value
increase in the air after dehumidification was achieved, and the de-
−1
humidified air needs further heating to the drying temperature. The Operational time (h year ) 8000
other energy input in this system is required for the regeneration loop of Interest rate (%) 0.6
Lang Factor (−) 3.5
the brine. The energy gain in this system is realized by recovery of the
Cold utility cost (€ kJ−1) 0.1 × 10−5
heat which is obtained at condensation of the water vapor in the brine Hot utility cost (€ kJ−1) 1.3 × 10−5
in the membrane contactor. The temperature of the internal cooling
water (H3), that is used to recover the heat, is raised to 92 °C. This flow
is used to pre-heat the milk feed before concentration (C1) and the 4.3. Configuration 2 – membrane contactor with superheated steam
brine after regeneration (C8). Fig. 7 shows the optimal heat exchanger
network for this configuration. The energy requirement for this opti- For this configuration the optimal conditions of the air loop over the
mized system is 7.3 MJ heat per kg of milk powder, which is a 14% dryer and the dehumidification unit are comparable to previous con-
improvement compared to the conventional configuration. The optimal figuration. Most important difference is the high temperature of the
values for the decision variables are listed in Table 6. superheated steam which changes the brine after regeneration into a
In this configuration the spray drying temperature (Ta, in) is the hot stream (H7). The surplus superheated steam (H4) acts as a high
same as in the conventional configuration. The hot streams have a quality energy flow, and is used efficiently in heat integration. Related
maximum temperature of 92 °C, (see Fig. 7) and cannot be used to heat to this difference, the cooling water flow (H3) is of a lower quality,
the air after dehumidification, the spray dryer is most energy efficient therefore, most of the latent heat released in the brine is transported to
at the highest temperature, i.e. 220 °C. The other decision variables are the regenerator via the brine flow, and a negligible amount to the
all on, or close to, the boundaries for this system. To transfer most heat cooling water system. This is achieved by a large brine flow and a very
from the brine to the cooling water and air, the temperature of the small cooling water flow, resulting from a small concentration differ-
evaporator (Trm, in), and the temperature- and concentration difference ence of the brine between in- and outlet of the membrane contactor
(ΔTmc and Δxmc) over the membrane contactor are set to their maximum (Δxmc). The heat exchanger network, as shown in Fig. 8, does not in-
values. clude a heat exchanger connected to the cooling water stream (H3), as
the energy content of this stream is so low it can be neglected (0.004 MJ

Table 4
Cost data for each unit operation.

Process Aref Cref neq Life time Reference

Heat exchanger 80 m 2
€ 32,800 0.68 20 years (Smith, 2005)
Evaporator 7700 kg water h−1 € 830,000 0.53 30 years (Seider, Seader, Lewin, & Widagdo, 2010)
Spray dryer 400 kg water h−1 € 600,000 0.29 30 years (APV Dryer Handbook, 2000; Garrett, 1989)
MC (membrane) 1 m2 € 50 – 4 (Elsayed, Barrufet, & El-Halwagi, 2014)
MC (equipment) 1 m2 € 200 – 10 (Elsayed et al., 2014)
Zeolite wheel 1400 kg water h−1 €250,000 0.78 5 (Voogt, 2016)

29
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Fig. 6. Heat exchanger network of the conventional configuration, including the hot and cold utilities (HU and CU respectively). The flow descriptions are given in Table A.3.

heat per kg milk powder). The energy consumption for this optimized Table 6
and heat integrated configuration is 4.9 MJ heat per kg milk powder, Optimal values for the decision variables for all configurations. Trm is the regeneration
which is a large improvement compared to the conventional config- medium temperature which can either be superheated steam, hot air or steam for the
evaporator, depending on the configuration.
uration (42%).
The relative high temperature of the superheated steam (over Decision variable 1 2 3 4 Conventional
187 °C) is a possible drawback for this system. No literature is published
on membrane contactors operating at these temperatures. Nevertheless, Ta, in [°C] 220 220 220 180 220
ya, in [kg kg dry air−1] 0.010 0.012 0.002 0.002 –
thermal stable membranes exist (Li & Chen, 2005), hence, this config-
Δxmc [kg kg−1] 0.099 0.014 – – –
uration could be feasible. As an alternative the temperature of the su- ΔTmc [°C] 1.2 1.0 – – –
perheated steam could be reduced. Decreasing the temperature of the Tz, in [°C] – – 125 125 –
superheated steam from 187 °C to 150 °C results in an increase of energy Trm, in [°C] 128 187 298 350 –
Trm, out [°C] 50 100 – 250 –
consumption of less than 1%.
Tsteam [°C] 83 88 87 91 75
Tcond [°C] 27 32 35 30 19
4.4. Configuration 3 – zeolite with hot air

(C1). Fig. 9 shows the optimal heat exchanger network for this con-
The energy consumption for this optimized and heat integrated
figuration. A drawback of using hot air as regeneration medium is the
configuration is 7.5 MJ heat per kg milk powder, which is 17% below
lower quality compared to steam, and the need for air–air and air-liquid
the energy consumption in the optimized conventional system. The
heat exchangers. This type of heat exchanger need larger heat ex-
energy reduction is comparable to configuration 1. The optimal values
changing surfaces, and have therefore higher investment costs, than
for the decision variables are given in Table 6.
steam-heat exchangers.
The dehumidified air (C6) exits the adsorber section of the zeolite
with a temperature of 163 °C, and to reach the drying temperature of
220 °C a low amount of external energy is needed. This difference 4.5. Configuration 4 – zeolite with superheated steam
compared to the conventional configuration is the main contributor for
the reduction of energy input. The configuration needs, however, a The energy consumption for this configuration is 5.2 MJ per kg milk
large amount of energy to heat the hot air for regeneration (from am- powder. This is a large improvement in energy efficiency compared to
bient temperature to 298 °C), which can only partially be supplied from configuration 3, and is similar to the energy consumption of config-
available hot streams. The regenerator exhaust air (H5) has a tem- uration 2. Like in configuration 2, the energy reduction is possible due
perature of 125 °C, and this stream can only be used to pre-heat the air to the high quality of the surplus superheated steam after regeneration.
for regeneration (C5), or to pre-heat the feed flow to the evaporator The optimal temperature of the surplus superheated steam in this

Fig. 7. Heat exchanger network design for configuration 1. HU: hot utilities, CU: cold utilities, the flow descriptions can be found in Table A.4.

30
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Fig. 8. Heat exchanger network design for configuration 2. HU: hot utilities, CU: cold utilities, the flow descriptions can be found in Table A.5.

configuration has a temperature of 250 °C, and can be applied for many and 4 is a high quality energy stream and is used efficiently; in con-
purposes. In the optimized system the superheated steam (H4) is used figuration 4 even to heat the dehumidified air to the dryer temperature.
to heat the dehumidified air (C7) to the drying temperature which is, in In configuration 2 more energy is used to operate the adsorber and
contrast to all other configurations, 180 °C instead of 220 °C (Fig. 10). regeneration section than in configuration 4, but in configuration 2
The zeolite requires cooling after regeneration (H7). The cold streams more energy is recovered by heat integration. There are two main
in the system, milk feed flow (C1) and evaporator steam flow (C2), are causes for this effect. First, zeolite in configuration 4 has to be cooled
not suitable for direct cooling, hence the cooling energy from the zeolite after regeneration but the energy cannot be recovered, while the energy
cannot be recovered by heat integration. Cooling of the zeolites with in the brine solution of configuration 2 is recovered by a cold stream.
ambient air is most advisable. Secondly, the heat of adsorption in configuration 2 is released in the
brine, whereas in configuration 4 the majority it is transferred directly
to the dehumidified air. As a result the dehumidified air (C7) in con-
4.6. Comparison of the different configurations figuration 2 has a temperature of 95 °C versus 142 °C in configuration 4.
Hence, the dehumidified air flow in configuration 2 needs more
The energy requirements for the four configurations and the con- heating.
ventional configuration are summarized in Fig. 11. The total energy
savings range between 11 and 42% compared to the conventional
configuration. Atkins et al. (2011) applied heat integration whereby the 4.7. Sensitivity analysis
dryer exhaust air is used to pre-heat the inlet air and reached a re-
duction in energy consumption up to 20%. The configurations in this A sensitivity analysis on variations over the operational window, set
work with dehumidification realize higher energy savings. by the bounds of the decision variables, showed for all systems that the
The energy consumption of configurations 1 and 3 is higher than in energy consumption is almost constant. Examples of variations are
configurations 2 and 4. The quality of the hot streams in configurations given in Fig. 12. Only at the edges of the operational window is the
1 and 3 is not sufficient to heat the dehumidified air to the optimal sensitivity meaningful. Variations in air temperature and humidity at
dryer temperature. The surplus of superheated steam in configuration 2 the dryer inlet has a maximal effect of 4% on the total energy efficiency

Fig. 9. Heat exchanger network design for configuration 3. HU: hot utilities, CU: cold utilities, the flow descriptions can be found in Table A.6.

31
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Fig. 10. Heat exchanger network design for configuration 4. HU: hot utilities, CU: cold utilities, the flow descriptions can be found in Table A.7.

steam for regeneration and results in high quality heat, which is

exploited effectively in the current system.

4.8. Used optimization method

Simultaneous optimization of operational conditions and heat in-

tegration resulted in an additional energy reduction varying between 1
and 11% for the different configurations. For configuration 1, 3, and 4
the additional energy reduction is minor (1 to 4%). These heat in-
tegrated systems already proved to be energy efficient and during op-
timization the gain achieved for one stream is counteracted by extra
energy needs for another. Configuration 2, however, has an additional
energy reduction of 11%, and has the most advantage of the simulta-
neous approach. The gain for this configuration is the utilization of the
surplus superheated steam after regeneration, which has a high energy
content to be reused for heating of other streams in the system.
Atuonwu et al. (2011) achieved for low-temperature drying an im-
Fig. 11. External energy requirements with and without heat integration for configura-
provement of 13% in energy reduction for the simultaneous optimiza-
tions 1 to 4 and the conventional configuration.
tion, which is in line with the improvement found for configuration 2.
The closed-loop spray dryer considered in this work operates at higher
for all configurations (see Fig. 12a and c for configuration 2 and 4). The temperatures, which makes drying more energy efficient than low
dryer can, therefore, be operated at a broad range of operational con- temperature drying (50 °C). Although the simultaneous optimization
ditions, without a large influence on the total energy efficiency. This approach did not result in large energy reduction for all configurations,
low sensitivity allows tuning of the operational conditions to the pro- the approach allows an open search for solutions without an a priori
duct specifications rather than energy requirements. On the other hand choice of heat integration possibilities. In this work this approach
Fig. 12d illustrates that some combinations of superheated steam tem- proved to be powerful for systems with streams which, depending on
perature for regeneration (Trm, in) and zeolite temperature (Tz, in), in the chosen operational conditions, can be used as either hot or cold
configuration 4 will increase the energy consumption significantly, streams.
however, there is still a large operational window where energy con-
sumption is low. In contrast, for configuration 2 (Fig. 12b), the energy
consumption is hardly influenced a change in temperature of the su- 4.9. Economic evaluation
perheated steam used for regeneration (Trm, in).
In the presented results the inlet concentration of the milk for the The total annual costs in k€ year−1 (TAC) of the different config-
dryer was set to 50% total solids, which is common in milk powder urations are summarized in Table 7. The costs of the pre-heater, eva-
production. Increasing the inlet concentration reduces the total energy porator, and spray dryer are the same for all configurations. The TAC
consumption (Walmsley, Atkins, et al., 2017). However, with increasing for configurations 3 and 4, which use the zeolite systems, have com-
solids concentration the viscosity also increases, which may lead to parable and significant lower costs than the conventional system.
clogging of the spray dryer nozzle. Advantage of the monodisperse Configuration 1 and 2 need a high membrane surface which result in
atomization nozzle is the better handling of high viscosity products high costs for the membrane contactor system. These high costs are
(Deventer et al., 2013), which makes the increase of concentration result of the applied objective function that minimizes the energy
possible. Fig. 13 illustrates the effect of variations in dryer feed con- consumption. Optimizing configuration 2 with respect to the TAC, in-
centration on the energy consumption. All configurations show a de- stead of energy consumption, results in a 50% cost decrease, while the
creasing energy consumption with increasing feed concentration. energy consumption is only increased by 4%. The resulting costs,
However, for configuration 2 the energy consumption is almost con- 12.9 M€ year−1, are still higher compared to the configurations with
stant for feed concentrations above 45% total solids and for config- zeolites. However, with the development of new membranes the flux
uration 4 above 30% total solids. These configurations use superheated can be increased, which will decrease the costs drastically.

32
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Fig. 12. Energy consumption variation with different operational conditions for configuration 2 (a and b) and 4 (c and d).

Table 7
Cost overview (k€ year−1) for all different configurations.

1 2 3 4 Conventional

Equipment costs 7573 18,414 1636 1658 1135

Evaporator 739 739 739 739 739
Spray dryer 372 372 372 372 372
Zeolite 493 493
MC 6014 4882
MC regeneration 420 9174
Heat exchangers 27 92 31 53 23
Utility costs 7287 4926 7397 5182 8286
TAC 14,860 23,340 9033 6841 9422

per kg milk powder for the best practice of milk powder plants based on
a combination of MVR&TVR evaporation (Walmsley, Atkins, et al.,
2017). These MVR&TVR systems still need about 1.0 MJ steam per kg
powder, which can be provided by the surplus energy form the dehu-
midification adsorbent regeneration system. Then, the energy use of the
MVR&TVR system decreases to 4.3 MJ per kg milk powder. For the
Fig. 13. The effect of level of concentration before spray drying on the energy con- remaining amount an additional heat sink has to be found. Potential
sumption is shown for the different configurations.
sinks are the boiler house, CIP unit etc. Walmsley, Atkins, et al. (2017)
also performed projections for the energy consumption with other
4.10. Discussion emerging technologies and it would be interesting to assess in future
work how this new instrument for energy saving fits in the total range
Atkins et al. (2011) presented the potential of sensible heat recovery of possibilities.
from the dryer exhaust air. This concept is also used in a recent study of Heat integration is essential for all configurations to realize the re-
Walmsley, Atkins, et al. (2017) on process and utility integration. In duction in energy consumption. The realization depends on the feasi-
this work we have shown that the recovery of latent heat in closed-loop bility of the proposed heat exchanger network in industrial applica-
drying is a new instrument in the spectrum for energy savings in milk tions. Plant layout, possibilities for energy transport between units, and
powder production. The best solutions for the systems in this work, interference with other operations may limit the application of heat
with a steam driven evaporator, resulted in an energy consumption of integration. The usage of heat pumps was not investigated in this study,
4.9–5.1 MJ heat per kg milk powder, which is around the 5.3 MJ fuel but has the potential to further increase the recovery of the efficiency of

33
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

heat recovery in spray drying systems (Jensen, Markussen, Reinholdt, & Nomenclature
Elmegaard, 2015; Walmsley, Klemeš, Walmsley, Atkins, & Varbanov,
2017). Variables
The absence of fines in the dryer exhaust, realized by the mono-
disperse nozzle, is a main requirement for air dehumidification in the C Costs (€)
closed loop. The negative influence of fouling of the membranes or cp Specific heat capacity (kJ kg−1 °C−1)
particle deposition in the zeolite, on the process performance should be F Flow (kg h−1)
avoided. Monodisperse droplet drying is proven at pilot scale H Enthalpy flow (kJ h−1)
(Debrauwer, 2016) and is under development for larger production i Interest rate (%)
capacities. The shown potential for energy reduction motivates further LF Lang factor (−)
development of the monodisperse atomizers and the investigation of P Vapor pressure (Pa)
monodisperse powder properties. Monodisperse drying has also other Q Energy requirement (kJ kg milk powder−1)
energy efficiency advantages. When all particles are similar in size no RH Relative humidity(−)
overheating of smaller particles occurs, which leads to a better con- T Temperature (°C)
trollability of the dryer and a further reduction in energy consumption t Operating time (h year−1)
(Atuonwu & Stapley, 2017). TAC Total annual costs (€ year−1)
The application of membrane contactors is new in this sector, and U Heat transfer coefficient (W m−2 °C−1)
dehumidification at elevated temperatures is still at technology devel- x Component content of stream (kg kg−1)
opment level, further research is required to prove the effectiveness and y Moisture content of air stream (kg kg dry air−1)
compatibility of this technology. Furthermore, although the brine is
used in a closed loop and is isolated from the drying air by a membrane, Subscripts
the usage of a strong brine can be a hurdle for the acceptance of
membrane contactors in the food industry. Zeolite dehumidification a Air
systems, on the other hand, are already tested in the food industry for abs Absolute
the pre-treatment of the air prior to drying (Boxtel et al., 2012). ads Adsorber
In 2015 the EU produced 2.9 million tons of skimmed and whole b Brine
milk powder (Eurostat, 2016), which requires around 26 PJ for pro- c Cooling water
duction (based on 8.4 MJ heat per kg milk powder). Implementation of cm Milk concentrate
the proposed closed-loop dryer system, with savings between 11 and cw Condensate
42% in energy consumption, results in an energy reduction of 4.3 to cool Cooling requirement
11.6 PJ per year. The discussed emerging technologies are essential and d Dryer
the optimal systems surmount the 2030 EU goals on energy reduction. des Desorption
Although this study focusses on the production of milk powder, the eq Equipment
proposed configurations are applicable to any spray drying process. evap Evaporator
Other industries like the pharmaceutical and chemical industry already ex External heating
use closed-loop spray drying in order to recover solvents and gasses heat Heating requirement
(GEA Niro Pharmaceutical, n.d.). These industries, and also others, can hex Heat exchanger
now benefit from the discussed solutions to improve their energy effi- hu Hot utility
ciency. in/out Inlet or outlet stream
inv Investment
5. Conclusion l/s Liquid or solid
life Lifetime
Energy reduction by closed-loop drying is not possible in current m Milk
spray drying systems. Energy from the dryer exhaust air cannot be re- p Milk powder
covered due to the fines present in the air. Application of monodisperse reg Regenerator
droplet atomizers result in exhaust air without fines, and hence closed- rm Regeneration medium
loop drying with energy recovery can be applied. Heat integration is st Steam
then essential to make the system highly energy efficient. Four different shs Superheated steam
configurations optimized in this work resulted in a potential energy ut Utility
savings of 11 to 42% energy compared to current practice of milk v Vapor
powder production using steam driven evaporators. The configuration w Water
with the lowest external energy requirements (4.9 MJ heat per kg z Zeolite
powder, consists of the membrane contactor with superheated steam as
regeneration medium (configuration 2)). However, for air dehumidifi- Acknowledgements
cation at elevated temperatures zeolite adsorbent wheels are closer to
commercial implementation compared to membrane contactors. A The research leading to these results has received funding from the
zeolite system with superheated steam as regeneration medium (con- European Unions's Seventh Framework Programme for research, tech-
figuration 4) results in an energy requirement of 5.1 MJ of heat per kg nological development and demonstration under grant agreement nr.
powder. Furthermore, in this work simultaneous optimization of the 613732 (Enthalpy). We also thank partners from the ENTHALPY pro-
operational conditions and heat exchanger network proved to be an ject: P. Debrauwer, J. van Medevoort and J. Voogt (TNO), J.H.
effective approach for additional energy reduction. Hanemaaijer (TNO, I3), and R. Houben (Bodec) for the provided data.

34
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Appendix

Table A.1
Parameters used for optimization.
(Doran, 1995; Kreith, Manglik, & Bohn, 2012; Perry & Green, 1997).

Parameter Value

Pabs (Pa) 101,325

cp, a (kJ kg−1 °C−1) 1
cp, m (kJ kg−1 °C−1) 1.54
cp, z (kJ kg−1 °C−1) 0.84
cp, w (kJ kg−1 °C−1) 4.18
cp, v (kJ kg−1 °C−1) 1.93
Hv (kJ kg−1) 2500
Ua−l (W m−2 °C−1) 170
Ua−a (W m−2 °C−1) 35
Ua−st (W m−2 °C−1) 200
Ul−l (W m−2 °C−1) 1200
Ul−st (W m−2 °C−1) 2200
Ushs (W m−2 °C−1) 100

Table A.2
Lower and upper boundary values for each decision variables. The boundaries for Trm, in depend on the used regeneration medium; [1] is for air and
superheated steam, and [2] is for steam for the brine evaporator.

Variable Lower bound Upper bound

Ta, in (°C) 180 220

ya, in (kg dry air kg−1) 0.002 0.1
Tst, evap (°C) 75 100
Tcw (°C) 12 50
Δxmc (kg kg−1) 0.01 0.1
ΔTmc (°C) 1 20
Tz, in (°C) 2020 125
Trm, in (°C) [1] 200 350
Trm, in (°C) [2] 100 150
Trm, out (°C) 100 250

Table A.3
Stream data for the conventional configuration.

Stream name ID F (kg h−1) Ts (°C) Tt (°C) Q (MJ h−1)

Milk feed C1 100,000 10 65 2.2 × 105

Steam to the 1st effect of the milk evaporator C2 11,968 65 75 2.8 × 105
Ambient air to the dryer C6 211,149 20 220 4.3 × 105
Vapor from the last effect of the milk evaporator H1 11,543 40 39 2.8 × 105

Table A.4
Stream data for configuration 1.

Stream name ID F (kg h−1) Ts (°C) Tt (°C) Q (MJ h−1)

Milk feed C1 100,000 10 65 2.2 × 105

Steam to the 1st effect of the milk evaporator C2 12,069 65 83 2.8 × 105
Steam to the 1st effect of the evaporator for brine regeneration C3 4692 92 128 1.1 × 105
Air from adsorbent to the dryer C7 210,205 95 220 2.7 × 105
Adsorbent from regenerator C8 37,987 50 92 4.4 × 105
Vapor from the last effect of the milk evaporator H1 11,543 40 39 2.8 × 105
Vapor from the last effect of evaporator for brine regeneration H2 4223 50 49 1.0 × 105
Cooling water from the membrane contactor H3 66,113 92 40 1.4 × 105

35
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Table A.5
Stream data for configuration 2.

Stream name ID F (kg h−1) Ts (°C) Tt (°C) Q (MJ h−1)

Milk feed C1 100,000 10 65 2.2 × 105

Steam to the 1st effect of the milk evaporator C2 12,202 65 99 2.8 × 105
Superheated steam reheating after regenerator C4 126,131 100 187 2.2 × 105
Air from adsorbent to the dryer C7 210,219 95 220 2.7 × 105
Vapor from the last effect of the milk evaporator H1 11,543 40 35 2.8 × 105
Cooling water from the membrane contactor H3 108 99 99 –
Superheated steam surplus from regenerator H4 8674 100 99 2.3 × 105
Adsorbent from regenerator H7 333,252 119 92 2.2 × 105

Table A.6
Stream data for configuration 3.

Stream name ID F (kg h−1) Ts (°C) Tt (°C) Q (MJ h−1)

Milk feed C1 100,000 10 65 2.2 × 105

Steam to the 1st effect of the milk evaporator C2 12,132 65 87 2.8 × 105
Hot air for regenerator C5 90,595 20 298 2.6 × 105
Air from adsorbent to the dryer C7 202,750 163 220 1.2 × 105
Vapor from the last effect of the milk evaporator H1 11,543 40 35 2.8 × 105
Hot air after regeneration H5 90,595 125 45 0.9 × 105
Adsorbent from regenerator H7 57,317 125 125 –

Table A.7
Stream data for configuration 4. The Zeolite stream after regeneration (H7) is cooled with ambient air, and is therefore not included as cold utility
requirement.

Stream name ID F (kg h−1) Ts (°C) Tt (°C) Q (MJ h−1)

Milk feed C1 100,000 10 65 2.2 × 105

Steam to the 1st effect of the milk evaporator C2 12,175 65 91 2.8 × 105
Superheated steam reheating after regenerator C5 113,000 250 350 2.2 × 105
Air from adsorbent to the dryer C7 275,000 142 180 1.1 × 105
Vapor from the last effect of the milk evaporator H1 11,500 40 30 2.8 × 105
Superheated steam surplus from regenerator H4 8720 250 65 3.3 × 105
Adsorbent from regenerator H7 64,800 166 125 –

References Debrauwer, P. (2016). Innovative technologies reducing energy & water consumption in
milk powder production - Executive summary. Retrieved from http://www.enthalpy-
fp7.eu/wp-content/uploads/ENTHALPY-EXECUTIVE-SUMMARY.pdf.
Abdel-Salam, A. H., Ge, G., & Simonson, C. J. (2013). Performance analysis of a mem- Depping, V., Grunow, M., van Middelaar, C., & Dumpler, J. (2017). Integrating en-
brane liquid desiccant air-conditioning system. Energy and Buildings, 62, 559–569. vironmental impact assessment into new product development and processing-tech-
http://dx.doi.org/10.1016/j.enbuild.2013.03.028. nology selection: Milk concentrates as substitutes for milk powders. Journal of Cleaner
APV dryer handbook. Rosemount. Production, 149, 1–10. http://dx.doi.org/10.1016/j.jclepro.2017.02.070.
Atkins, M. J., Walmsley, M. R. W., & Neale, J. R. (2011). Integrating heat recovery from van Deventer, H., Houben, R., & Koldeweij, R. (2013). New atomization nozzle for spray
milk powder spray dryer exhausts in the dairy industry. Applied Thermal Engineering, drying. Drying Technology, 31(8), 891–897. http://dx.doi.org/10.1080/07373937.
31(13), 2101–2106. http://dx.doi.org/10.1016/j.applthermaleng.2011.03.006. 2012.735734.
Atkins, M. J., Walmsley, M. R. W., & Neale, J. R. (2012). Process integration between Djaeni, M., Bartels, P., Sanders, J., van Straten, G., & van Boxtel, A. J. B. (2007).
individual plants at a large dairy factory by the application of heat recovery loops and Multistage zeolite drying for energy-efficient drying. Drying Technology, 25(6),
transient stream analysis. Journal of Cleaner Production, 34, 21–28. http://dx.doi.org/ 1053–1067. http://dx.doi.org/10.1080/07373930701396535.
10.1016/j.jclepro.2012.01.026. Djaeni, M., van Straten, G., Bartels, P. V., Sanders, J. P. M., & van Boxtel, A. J. B. (2009).
Atuonwu, J. C., & Stapley, A. G. F. (2017). Reducing energy consumption in spray drying Energy efficiency of multi-stage adsorption drying for low-temperature drying. Drying
by monodisperse droplet generation: Modelling and simulation. Energy Procedia, 123, Technology, 27(4), 555–564. http://dx.doi.org/10.1080/07373930802715682.
235–242. http://dx.doi.org/10.1016/j.egypro.2017.07.251. Doran, P. M. (1995). Bioprocess engineering principles. Academic Press.
Atuonwu, J. C., van Straten, G., van Deventer, H. C., & van Boxtel, A. J. B. (2011). Duran, M. A., & Grossmann, I. E. (1986). Simultaneous optimization and heat integration
Improving adsorption dryer energy efficiency by simultaneous optimization and heat of chemical processes. AICHE Journal, 32(1), 123–138. http://dx.doi.org/10.1002/
integration. Drying Technology, 29(12), 1459–1471. http://dx.doi.org/10.1080/ aic.690320114.
07373937.2011.591516. Elsayed, N. A., Barrufet, M. A., & El-Halwagi, M. M. (2014). Integration of thermal
Atuonwu, J. C., van Straten, G., van Deventer, H. C., & van Boxtel, A. J. B. (2012). A membrane distillation networks with processing facilities. Industrial & Engineering
mixed integer formulation for energy-efficient multistage adsorption dryer design. Chemistry Research, 53(13), 5284–5298. http://dx.doi.org/10.1021/ie402315z.
Drying Technology, 30(8), 873–883. http://dx.doi.org/10.1080/07373937.2012. European Commission (2011). Energy roadmap 2050. Brussels. Retrieved from https://
674996. ec.europa.eu/energy/en/topics/energy-strategy-and-energy-union/2050-energy-
Bergero, S., & Chiari, A. (2010). Performance analysis of a liquid desiccant and membrane strategy.
contactor hybrid air-conditioning system. Energy and Buildings, 42(11), 1976–1986. Eurostat (2016). Milk and milk product statistics - Statistics explained. Retrieved August
http://dx.doi.org/10.1016/j.enbuild.2010.06.003. 11, 2017, from http://ec.europa.eu/eurostat/statistics-explained/index.php/Milk_
van Boxtel, A. J. B., Boon, M. A., van Deventer, H. C., & Bussmann, P. J. T. (2012). and_milk_product_statistics.
Zeolites for reducing drying energy usage. In E. Tsotsas, & A. S. Mujumdar (Eds.). Finnegan, W., Goggins, J., Clifford, E., & Zhan, X. (2017). Environmental impacts of milk
Modern drying technology volume 4 (pp. 163–197). Weinheim: Wiley-VCH. powder and butter manufactured in the Republic of Ireland. Science of the Total

36
S.N. Moejes et al. Innovative Food Science and Emerging Technologies 47 (2018) 24–37

Environment, 579, 159–168. http://dx.doi.org/10.1016/j.scitotenv.2016.10.237. Moejes, S. N., & van Boxtel, A. J. B. (2017). Energy saving potential of emerging tech-
Florides, G. A., Kalogirou, S. A., Tassou, S. A., & Wrobel, L. C. (2003). Design and con- nologies in milk powder production. Trends in Food Science & Technology, 60, 31–42.
struction of a LiBr–water absorption machine. Energy Conversion and Management, http://dx.doi.org/10.1016/j.tifs.2016.10.023.
44(15), 2483–2508. http://dx.doi.org/10.1016/S0196-8904(03)00006-2. Okos, M., Rao, N., Drecher, S., Rode, M., & Kozak, J. (1998). Energy usage in the food
Garrett, D. E. (1989). Chemical engineering economics. Dordrecht: Springer industry: A study. (Washington D.C).
Netherlandshttp://dx.doi.org/10.1007/978-94-011-6544-0. Patil, K. R., Tripathi, A. D., Pathak, G., & Katti, S. S. (1990). Thermodynamic properties of
GEA Niro Pharmaceutical. (n.d.). Pharmaceutical GMP spray drying facility: Spray drying aqueous electrolyte solutions. 1. Vapor pressure of aqueous solutions of lithium
process development and contract manufacturing. Soeborg, DK. Retrieved from chloride, lithium bromide, and lithium iodide. Journal of Chemical & Engineering Data,
http://www.gea.com/en/binaries/pharmaceutical-GMP-spray-drying-facility- 35(2), 166–168. http://dx.doi.org/10.1021/je00060a020.
brochure-EN_tcm11-24405.pdf. Perry, R. H., & Green, D. W. (1997). Perry's chemical engineers' handbook (Seventh ed.).
Goldsworthy, M. J., Alessandrini, S., & White, S. D. (2015). Superheated steam re- McGraw-Hill.
generation of a desiccant wheel—Experimental results and comparison with air re- Ramirez, C., Patel, M., & Blok, K. (2006). From fluid milk to milk powder: Energy use and
generation. Drying Technology, 33(4), 471–478. http://dx.doi.org/10.1080/ energy efficiency in the European dairy industry. Energy, 31(12), 1984–2004. http://
07373937.2014.956118. dx.doi.org/10.1016/j.energy.2005.10.014.
Golman, B., & Julklang, W. (2014). Analysis of heat recovery from a spray dryer by re- Rogers, S., Fang, Y., Qi Lin, S. X., Selomulya, C., & Dong Chen, X. (2012). A monodisperse
circulation of exhaust air. Energy Conversion and Management, 88, 641–649. http://dx. spray dryer for milk powder: Modelling the formation of insoluble material. Chemical
doi.org/10.1016/j.enconman.2014.09.012. Engineering Science, 71, 75–84. http://dx.doi.org/10.1016/j.ces.2011.11.041.
Isetti, C., Nannei, E., & Magrini, A. (1997). On the application of a membrane air—liquid Seider, W. D., Seader, J. D., Lewin, D. R., & Widagdo, S. (2010). In W. D. Seider (Ed.).
contactor for air dehumidification. Energy and Buildings, 25(3), 185–193. http://dx. Product and process design principles: Synthesis, analysis and design, 3rd edition(3rd ed.).
doi.org/10.1016/S0378-7788(96)00993-0. John Wiley & Sons.
Iyoki, S., Yamanaka, R., & Uemura, T. (1993). Physical and thermal properties of the Smith, R. (2005). Chemical process design and integration. John Wiley Sons Ltd.
water-lithium bromide-lithium nitrate system. International Journal of Refrigeration, Voogt, J. A. (2016). Memo: Gegevens t.b.v. CAPEX/OPEX Sorptie-unit. Zeist: TNO.
16(3), 191–200. http://dx.doi.org/10.1016/0140-7007(93)90048-D. Walmsley, T. G., Atkins, M. J., Walmsley, M. R. W., & Neale, J. R. (2016). Appropriate
Jain, S., Tripathi, S., & Das, R. S. (2011). Experimental performance of a liquid desiccant placement of vapour recompression in ultra-low energy industrial milk evaporation
dehumidification system under tropical climates. Energy Conversion and Management, systems using Pinch Analysis. Energy, 116, 1269–1281. http://dx.doi.org/10.1016/j.
52(6), 2461–2466. http://dx.doi.org/10.1016/j.enconman.2010.12.052. energy.2016.04.026.
Jensen, J. K., Markussen, W. B., Reinholdt, L., & Elmegaard, B. (2015). Exergoeconomic Walmsley, T. G., Atkins, M. J., Walmsley, M. R. W., Philipp, M., & Peesel, R.-H. (2017).
optimization of an ammonia–water hybrid absorption–compression heat pump for Process and utility systems integration and optimisation for ultra-low energy milk
heat supply in a spray-drying facility. International Journal of Energy and powder production. Energy. http://dx.doi.org/10.1016/j.energy.2017.04.142.
Environmental Engineering, 6, 195–211. http://dx.doi.org/10.1007/s40095-015- Walmsley, T. G., Klemeš, J. J., Walmsley, M. R. W., Atkins, M. J., & Varbanov, P. S.
0166-0. (2017). Innovative hybrid heat pump for dryer process integration. Chemical
Kemp, I. (2007). In I. Kemp (Ed.). Pinch analysis and process integration: A user guide on Engineering Transactions, 57. http://dx.doi.org/10.3303/CET1757174.
process integration for the efficient use of energy(Second edi.). Oxford: Butterworth- Walmsley, T. G., Walmsley, M. R. W., Atkins, M. J., & Neale, J. R. (2013). Improving
Heinemann. http://dx.doi.org/10.1016/B978-075068260-2.50003-1. energy recovery in milk powder production through soft data optimisation. Applied
Kneifel, K., Nowak, S., Albrecht, W., Hilke, R., Just, R., & Peinemann, K. (2006). Hollow Thermal Engineering, 61(1), 80–87. http://dx.doi.org/10.1016/j.applthermaleng.
fiber membrane contactor for air humidity control: Modules and membranes. Journal 2013.01.051.
of Membrane Science, 276(1–2), 241–251. http://dx.doi.org/10.1016/j.memsci.2005. Walmsley, T. G., Walmsley, M. R. W., Atkins, M. J., Neale, J. R., & Tarighaleslami, A. H.
09.052. (2015). Thermo-economic optimisation of industrial milk spray dryer exhaust to inlet
Kreith, F., Manglik, R. M., & Bohn, M. S. (2012). Principles of heat transfer. Cengage air heat recovery. Energy, 90, 95–104. http://dx.doi.org/10.1016/j.energy.2015.03.
learning. 102.
Krishna, R., & Wesselingh, J. A. (1997). The Maxwell-Stefan approach to mass transfer. Walstra, P., Geurts, T. J., Noomen, A., Jellema, A., & van Boekel, M. A. J. S. (1999). Dairy
Chemical Engineering Science, 52(6), 861–911. http://dx.doi.org/10.1016/S0009- technology: Principles of milk. Properties and processes. New York: Marcel Dekker Inc.
2509(96)00458-7. Westergaard, V. (2004). Milk powder technology - evaporation and spray drying.
Lachkov, V. I., Lysenkov, A. I., & Mamonov, Y. V. (1999). Expressions for thermophysical Yildirim, N., & Genc, S. (2017). Energy and exergy analysis of a milk powder production
properties of superheated steam. Measurement Techniques, 42(1), 58–60. http://dx. system. Energy Conversion and Management, 149, 698–705. http://dx.doi.org/10.
doi.org/10.1007/BF02504202. 1016/j.enconman.2017.01.064.
Li, J.-L., & Chen, B.-H. (2005). Review of CO2 absorption using chemical solvents in Zhang, L. (2006). Mass diffusion in a hydrophobic membrane humidification/dehumi-
hollow fiber membrane contactors. Separation and Purification Technology, 41(2), dification process: The effects of membrane characteristics. Separation Science and
109–122. http://dx.doi.org/10.1016/j.seppur.2004.09.008. Technology, 41(8), 1565–1582. http://dx.doi.org/10.1080/01496390600634723.

37