Professional Documents
Culture Documents
Manuscript Draft
Yechun Zhang, M. Tajammal Munir*, Isuru Udugama, Wei Yu, and Brent R. Young
Industrial Information and Control Centre (I2C2), The Department of Chemical and Materials Engineering, the
University of Auckland, New Zealand.
Highlights
1
*Manuscript
Click here to view linked References
4 Industrial Information and Control Centre (I2C2), The Department of Chemical and Materials
5 Engineering, the University of Auckland, New Zealand.
6 Abstract
7 In this study, two commonly used types of milk powder evaporators: a conventional
10 simulator. The objectives were to predict the process variable trends (temperature
11 profile along the tube side and the shell side of the evaporator) and to compare energy
13 integrated into the model so that energy consumption could be compared for the two
14 processes. The size of the evaporator required was also estimated so that so that
15 operating and capital costs could be assessed. The model prediction results were
16 validated using industrial data from a local milk powder plant. The developed model
19 with a known fact that a three-effect falling-film evaporator with MVR could
21 developed models in a commercial process simulator are novel in the dairy industry
22 and can be used for simulating heat-recovery integrated evaporation processes but
23 they still remain sufficiently generic and flexible to be applied to other production
24 processes.
1
25 Keywords: Simulation, falling film evaporator, mechanical vapour recompression.
26 1. Introduction
27 Evaporators in milk-powder production plants are widely used for concentrating milk
28 not only to achieve the desired viscosity for subsequent spray drying but also to
29 reduce the energy required for spray drying. During evaporation, sterilised milk is
30 concentrated under vacuum at temperatures between 40–70 °C. The total solids
32 Evaporation of milk takes place under vacuum to minimise the adverse impact of heat
35 Various types of evaporator are used in the dairy industry such as falling-film, plate,
37 used for milk powder production and are usually used with mechanical and thermal
39 shown in Figure 1. Milk circulates inside vertical tubes while steam provides direct
40 heating from the evaporator shell for evaporation. Falling-film evaporators are
41 preferred since they are easy to design, have a large heat-transfer area and produce a
42 short residence time. Although various types of evaporators differ in the details of
43 their design, different evaporators have the same working principle: the heat flow
45
2
46
48 Evaporation is the second most energy intensive process in milk powder production
49 after drying (Ruan et al., 2015). Therefore, energy consumption during evaporation
50 has a substantial impact on the cost of milk powder production. Energy saving or heat
51 recovery during milk powder production are desirable in terms of revenue savings and
53 greenhouse emissions. Several authors, such as Ramírez et al. (2006) and Atkins et al.
54 (2010), have reviewed energy use in the dairy industry, and have recommended
56 integration, and use of integrated solar thermal energy. Increasing the temperature of
57 milk evaporation may reduce costs by speeding up the process but relatively high
3
58 temperatures cause whey protein reactions and damage the functional properties of
59 milk powder (Pisecky et al., 2012; Winchester, 2000). However, Bylund (1995)
63 film evaporator. With MVR, vapours generated from the boiling chamber are
65 temperature of the evaporator and are recycled back to the evaporator. Mechanical
67 other types because of they require low amounts of energy (Bylund, 2003; Pisecky et
68 al., 2012).
70 which various transport, and heat and mass transfer phenomena occur. These
73 extend our knowledge of the field further by modelling these evaporators. Evaporator
74 process models can be used to predict process behaviour with reasonably acceptable
75 accuracy and can play a major role in optimisation and decision making without the
76 need for costly full-scale research trials that would disrupt normal plant operation.
78 powder applications. For example, Winchester and Marsh (1999) proposed a first-
80 More recently, Medhat et al. (2015) developed both “lumped” and “distributed”
4
82 wide variety of evaporator models can be found in the literature for other food
83 applications. For example, Russell et al. (2000) presented three types of evaporator
85 analytically-derived model; and a linear regression model, and Bhargava et al. (2008)
87 mill. Moreover, various studies have investigated: flow regimes (Silveira et al., 2015);
88 heat transfer coefficient distribution (Gong et al., 2015); pressure drop and friction
89 factor modelling (Mura and Gourdon, 2017); and modelling of heat and mass transfer
91 Most of the evaporator modelling related studies cited above have developed
100 build a process simulation of a falling-film evaporator is the objective of the current
101 study.
102 In this work, a simulation approach was proposed to model both conventional falling-
103 film evaporators and those with mechanical vapour recompression. With similar
104 baseline conditions assumed, the modelling results were presented and the evaporator
105 temperatures were compared with those obtained from a typical industrial plant.
106 Detailed comparisons were made in terms of overall heat transfer and energy
5
107 consumption. Capital cost and operational cost were estimated based on the values
111 A milk stream with ‘pseudo’ milk components was built in a commerical process
112 simulator during the authors’ previous work e.g. Zhang et al. (2014) with some
113 advancements as described in Munir et al. (2016). In these previous studies, the key
114 components of milk such as fat, proteins, lactose, and minerals were hypothetically
116 and physical properties such as heat capacity, thermal conductivity, density, and
117 viscosity of ‘pseudo’ milk components were modelled using the process simulator and
118 were compared with experimental data. The simulation of milk as a collection of
119 ‘pseudo’ milk components in the simulator allowed us to simulate dairy processing
120 unit operations, and to predict milk process behaviour reasonably accurately relative
122 Table 1. The composition and key physical properties of the raw milk at 4 °C, 101
123 kPa. The components marked with an asterisk are hypothetical compounds.
125
6
126
127 The composition of raw milk used in the simulation conducted here was based on that
128 reported by Bylund (1995) and Bon et al. (2010). Milk composition and simulator
129 predicted thermodynamic properties of ‘pseudo’ milk at 4 °C and 101.325 kPa are
130 presented in Table 1. Further details about modelling of ‘pseudo’ milk are provided in
132
134 In this study, the software VMGSim v10.0 (Virtual Materials Group Inc., 2014) was
135 used as the process simulator. The reasons for choosing VMGSim were: 1) a user-
136 friendly interface; 2) up-to-date thermodynamic data; and 3) its ability to incorporate
137 customized calculations using external computer program routines. The simulator has
138 been widely used e.g. for improving process design (Lee et al., 2011), optimization
139 (Saber and Shaw, 2008), controller assessment (Munir et al., 2012a, b), and
141 thermodynamic model was chosen due to its wide application and good accuracy in
142 predicting physical properties of polar and hydrocarbon compounds (Munir et al.,
143 2016).
145 with MVR were modelled in the commercial process simulator for a typical milk
146 powder production process. As the key unit operation, a falling-film evaporator uses a
147 shell-and-tube arrangement for heat transfer between the tube wall and shell, Figure 1.
148 Therefore, a vertical shell-and-tube heat exchanger was used to represent the calandria
149 of a falling-film evaporator. It should be noted that a real calandria consists of a large
150 number of tubes while the shell-and-tube heat exchanger unit operation in the
7
151 simulator only has single tube. Therefore, the single-tube approach can be considered
152 as a lumped cluster of multiple, smaller tubes. The key design parameters include heat
153 transfer factor U·A, where U is the overall heat transfer co-efficient and A is the total
154 heat transfer area. The co-efficient U can be broken down into four parts as described
(1)
156 where, hs and ht accounts for shell and tube side film heat transfer coefficients, xw is
157 the thickness of the tube wall, kw is the heat transfer rate of the tube wall, and ft is the
159 The flow directions for the tube and shell sides were set to be co-current. Each
160 calandria was vertically divided into 50 sections in order to reach sufficient accuracy
161 and resolution. The energy balance in each section is given by Equation (2).
(2)
162 where, Wi is the heat transfer duty within section i and ΔTi is average temperature
163 difference in section i. The total heat transfer duty is given by Equation (3).
(3)
164 All other heat exchangers for heat recovery purposes followed Equations (2) and (3),
165 except that they consisted of only 20 sections, and the flow direction was set to
167 The orifice and the distributor as shown in Figure 1 generate a pressure drop for the
168 milk stream before it enters the calandria tubes, and they were represented as a valve
8
170
171
174 evaporator with MRV used in an industrial milk powder production plant (Bylund,
175 2003) were modelled. The models for both the conventional falling-film evaporator
176 and the three-effect falling-film evaporator with MVR were built in both steady state
177 and dynamic modes with the common assumptions and baselines outlined below in
179 1. The milk feed has a temperature of 4 °C, pressure of 50 kPa and flow rate of
180 30,000 kg per hour, which is a typical processing rate for a dairy plant (Bylund,
181 1995).
182 2. The heat recovery and preheating process ensures the milk enters the first
183 calandria at 81.5 °C, and 50 kPa (at its boiling point).
184 3. Each calandria has no pressure drop on the tube side and 1 kPa pressure drop on
186 4. The condensate from each calandria is used for preheating the cold milk feed.
188 After building a model of a multi-effect falling-film evaporator with and without
189 MRV, the stream information, composition and key physical properties predicted by
190 the process simulation were compared with relevant industrial plant data. The
191 temperature profile along each calandria was also observed from the model. The
192 temperature profile is essential to study whether or not heat has been effectively
193 transferred and if the incoming temperature is outside of the acceptable range.
9
194 2.4. Energy consumption and preliminary cost analysis
195 The difference in energy consumption between the five-effect falling-film evaporator
196 without MRV and three-effect falling-film evaporator with MRV was compared in
197 this study using the process model. For the conventional five-effect evaporator, the
198 energy is mainly used for final stage preheating (“Preheater” in Figure 2) and
199 producing fresh steam (“Steam heater” in Figure 2) for calandria 1. For the three-
200 effect evaporator with MVR, the energy usage is in the two MVR compressors (CP1
201 and CP2 in Figure 3), the preheater (“Preheater” in Figure 3) and the steam heater that
202 provides fresh steam for thermo-compression (“Steam heater” in Figure 3).
203 In this study, the costs of operating the five-effect falling-film evaporator without
204 MRV and three-effect falling-film evaporator with MRV were compared using steam
205 cost. Steam used in dairy plants in New Zealand is mostly produced by burning
206 natural gas (Ministry of Business Innovation and Employment, 2017). Annual savings
207 of new installations were also considered. The cost calculations were based on a gas
208 price of 3USD/GJ, and the price was taken from the average market price for gas in
209 mid-2017. Furthermore, it was assumed that gas was burned to produce the heating
210 duty required and an efficiency of 70 % was achieved in the conversion process.
211 Natural gas was used in this instance as it is currently used by industrial dairy
212 manufacturers and provides a lower base cost for heating in comparison to the use of
217 2.
10
218
220 A total of five shell-and-tube heat exchangers (Effects 1-5) were used to represent
221 five calandria. Fresh steam at 81.5 °C, 50 kPa was fed into the first effect where the
222 external energy is provided to the evaporator system. The vapour separated from the
223 concentrated product of each calandria was fed into the shell side of the next calandria,
224 while the liquid phase was expelled through the orifice valve. The liquid expanded as
225 it was expelled and, therefore, its temperature dropped. The evaporation process was
226 then repeated in the next calandria. The vapour condensate from each calandria was
227 used to transfer heat in the preheating area, where most of the energy was recovered
228 through heat exchangers Hx 1-6. The final preheater raised the temperature to 81.5 °C,
229 and 50 kPa. The simulation calculated that for the conventional five-effect evaporator,
230 4230 kg/hr of steam is required for a 30, 000 kg/hr feed, which resulted in 0.19 kg of
231 steam required to evaporate 1 kg of water. This simulated result is slightly lower to
232 those reported in literature (0.20 kg of steam per 1 kg of water), presumably due to
11
234 The model for a three-effect falling-film evaporator with two-stage MVR was
235 developed. This is a generic example typical of the industry and a schematic is shown
236 in Figure 3.
237
238 Figure 3. Process flow schematic of a three-effect falling-film evaporator with MVR.
239 Similar to the model of the conventional evaporator, three shell-and-tube heat
240 exchangers (Effects 1-3) were used to represent the calandria. For Effects 1 & 2, the
241 vapour stream separated from the concentrated product of each calandria was sent to
242 its respective compressor in order to raise its energy. Therefore, the steam was
243 circulated in the calandria rather than introducing fresh steam to provide energy. For
244 process control purposes, the vapour was split into two streams, so that a degree of
245 freedom was added to Effects 1 & 2. In practice, flow controllers are in place to
246 control the amount of vapour to be circulated. For Effect 3, additional high pressure
247 steam (at 10 bar) (thermo-compression) was supplied to boost the vapour pressure
12
248 instead of using mechanical compressors. The reason for doing this is that the amount
249 of steam produced from Effect 3 is usually not sufficient to be compressed and sent
250 back to the calandria shell again. Similar to conventional evaporators, the condensate
251 streams were used for heat recovery purposes through heat exchangers Hx1-4.
252
253 3.2. Simulated stream information and comparison with plant data
254 One of the outputs from the model was that it can present stream information, such as
255 composition and key physical properties. The simulated stream information is
256 especially important for evaporating milk since some physical properties are used as
257 an indication of total solid content so they are a major index of the product. For
258 example, the viscosity of milk concentrate is used to control the atomisation process
259 in the spray dryer (Pisecky et al., 2012). This information can also be used for process
260 control.
261
262
263
264
265
266
267
268
269
13
270 Table 2. Compositions and key physical properties of the concentrate from each effect
271 for a conventional five-effect evaporator.
272 The compositions and key physical properties of the concentrated product are shown
273 in Table 2 for the conventional five-effect evaporator and in Table 3 for three-effect
274 evaporator with MVR. The temperatures of the concentrated product after each effect
275 with respect to the boiling point curve of simulated milk are shown in Figure 4.
276
277
278
14
279 Table 3. Compositions and key physical properties of the concentrate from each effect
280 for a three-effect evaporator with MVR.
282
283
284
15
90
80
1
Boiling point (°C)
70 2
3
1
60 4 2
5
50 3
40
0 10 20 30 40 50 60
Pressure (kPa)
285
286 Figure 4. The temperatures of the concentrated product after each effect for both
287 conventional five-effect evaporator (Red squares) and three-effect evaporator with
288 MVR (Blue circles)
289 Both types of evaporators demonstrated similar trends across effects, due to the nature
290 of evaporation. The pressure decreased with each effect and the product temperature
291 dropped accordingly, from an initial value of around 70 °C to 50 °C. The water
292 content decreased accordingly and finally reached 50 wt. % for both types of
293 evaporators. As a result, the key physical properties showed a clear trend whereby
294 liquid density (ρ) and viscosity (μ) increased while specific heat (Cp) and heat
295 conductivity (k) decreased, which agrees with the data reported by Zhang et al. (2014),
296 Madoumier et al. (2015), and Munir et al. (2016). However, it can be seen from
297 Figure 4 that a wider pressure range is required for the conventional five-effect
298 evaporator than the three-effect evaporator with MVR as a certain amount of pressure
299 drop should be guaranteed in each effect to create a temperature difference for heat
300 transfer inside the calandria. The pressure drop, temperature drop and amount of
16
302 Table 4. Pressure drop, temperature drop and amount of water vaporized across each
303 calandria for conventional five-effect evaporator and three-effect evaporator with
304 MVR.
306 The amount of water evaporated in each effect appears to be evenly distributed for the
307 conventional five-effect evaporator but the majority of the water was evaporated in
308 the first-effect in a three-effect evaporator with MVR. This difference can be
309 explained by the mechanism of MVR: the evaporated vapour from the first-effect was
310 compressed and sent back to the shell side of calandria 1. Due to the large flow rate of
311 the vapour, it can release a significant amount of energy as well, with the compression
312 ratio for the compressor CP1 only as low as 1.4. However, the MVR for the second
313 effect does not have a vapour flow rate as large as the first calandria; therefore it can
314 evaporate much less water even though the compression ratio for the compressor CP2
316 Since the three-effect evaporator with MVR was developed based on a real production
317 plant, a comparison of the stream temperature was made between the simulation
318 results and real plant data. It should be noted that the plant data is in real time but the
319 time axis has been rescaled for confidentiality reasons. The sensor locations are
320 shown in Figure 3 and the comparison results are shown in Figure 5.
17
321
322 Figure 5. Comparison of temperature (T) sensor readings and simulation results; data
323 shown for 24 hours.
324
326 As one of the key outputs of the model, the temperature profile along each calandria
327 can be viewed directly from the model. The temperature profile is essential to study
328 whether or not heat has been effectively transferred and if the approaching
329 temperature is outside of the acceptable range. The temperature profiles of calandria 1,
330 3 and 5 for the conventional five-effect evaporator are shown in Figure 6.
18
331
335 For falling-film type evaporators, the heat is transferred by condensing steam on the
336 tube side and evaporating of liquid product on the shell side so the latent heat of
337 vaporization (or condensation) is the main energy source rather than that stored by
338 heat capacity. Therefore, the temperature on both sides of the tube walls should
339 remain stable, which was accurately predicted by the model. The reason that the shell
340 side temperature showed a minor drop is due to the assumption of 1 kPa pressure drop
342 The temperature profiles of calandria 1, 2 and 3 for the three-effect evaporator with
19
344
345 Figure 7. Temperature profiles of the 3 calandria for a three-effect evaporator with
346 MVR. “T” stands for tube side temperature and “S” stands for shell side temperature.
347 Some interesting results can be observed from Figure 7. First of all, the shell side
348 temperature appears to be above the condensing temperature at the first three sections
349 for calandria 1 and 2, indicating the steam is superheated as a result of compression.
350 In real applications of MVR, desuperheaters are used to improve heat transfer, and the
351 excess heat is usually recovered. In the model, the desuperheater was simplified and
352 integrated into the calandria while the recovered heat was considered to be released to
353 the calandria tubes. Otherwise the temperature profile behaved similarly to that of a
355 The overall load Q and the heat transfer factor, U·A can be obtained from the model
356 directly. By assuming a constant U of 2200 W·m-2·K-1, the total heat transfer area can
357 be estimated therefore the size of the calandria can be indicated. The results from
20
359 Table 5. Heat transfer factor and estimated heat transfer area of each calandria for the
360 conventional five-effect evaporator and the three-effect evaporator with MVR,
361 assuming U = 2200 W·m-2·K-1
363 It can be observed that with the same total heat transfer load, the three-effect
364 evaporator with MVR required a 35 % lower heat transfer factor (and therefore heat
365 transfer area) than the conventional five-effect evaporator to achieve the same degree
366 of evaporation. However, it should be noted that the first effect of the MVR appears
369 The heating loads for the five-effect evaporator without MRV and the three-effect
370 evaporator with MRV are summarized in Table 6. The three-effect evaporator with
371 MVR used one third of the energy required for the five-effect evaporator. This result
372 is consistent with values reported in the literature (Bylund, 1995; Pisecky et al., 2012).
373
374
375
21
376 Table 6. Summary of unit operation loads for the conventional five-effect evaporator
377 and the three-effect evaporator with MVR.
379 Conventional five-effect evaporators are less efficient because the residual energy in
380 low pressure, low temperature streams cannot be effectively recovered, for example
381 the “steam residual” and “milk steam 5” streams in Figure 3. In contrast, these low
382 pressure, low temperature streams were compressed in evaporators with MVR and
383 reused either by compressors or steam jets, so that the overall efficiency was much
384 higher. However, the installation of large-scale compressor and steam jets requires
388 The major economic variables considered in this study and a preliminary operating
389 cost analysis are presented in Table 7. The proposed three-effect evaporator with
391 compared with a conventional five-effect evaporator, since the proposed system uses
393
22
394 Table 7. Major economic variables and preliminary cost analysis
396 As illustrated in Table 7, installing a MVR system at an industrial plant similar to the
397 one considered here can save ~400,000 USD per annum in operating expenses. This
398 saving can be considered significant for a single plant. Some plants used other types
399 of power such as electricity. The cost of power is a substantial component of the
400 overall cost so higher priced power is likely to produce an even greater financial
403 Unlike the operating expense calculation, calculating the capital expenditure for the
404 two systems is significantly more difficult. To get a reasonable estimate, one really
405 needs a detailed design and vendor quotes; not feasible for a study such as this.
406 However, upon closer inspection we can see the following differences in plant design
407 should be factored in when calculating the relative capital expenditure of the two
408 systems.
409 The proposed system uses only three evaporators and heat exchangers as
23
411 The area required for heat exchange (indicator of cost of production) for the
414 From a capital expenditure point of view, the proposed MVR structure will be cheaper
415 to build only if the cost of installing an MVR is lower than building two extra
416 separators and heat exchangers. However, with significant savings in operating
417 expenses, the higher costs of installing an MVR system may be offset by the potential
419 4. Conclusions
420 In this paper, process models were built using a commercial process simulator for two
422 effect evaporator and a three-effect evaporator with mechanical vapour recompression.
423 These models were successfully used to do cost analysis of both types of evaporators
424 and to validate established facts such as MVR evaporator consumed only one third of
425 the energy compared with a conventional evaporator, given sufficient energy
426 recoveries. Furthermore, developed models were able to predict the process variables
427 trends. It was found that for the conventional evaporator, the evaporation rate is more
428 evenly distributed among calandria while the majority of the water was evaporated in
429 the first effect in the MVR evaporator. The reason was that the energy in the low
430 pressure, low temperature streams of the conventional evaporator could not be
432 Acknowledgments
433 The authors thank and acknowledge Fonterra for providing information and advice.
24
434 References
435 Andika, R., Lee, M., (2017). Techno-economic study of enhanced absorber–
436 regenerator configurations for improving an industrial Sulfinol-M-based acid gas
437 removal processes. Journal of Industrial and Engineering Chemistry.
438 Atkins, M.J., Walmsley, M.R.W., Morrison, A.S., (2010). Integration of solar thermal
439 for improved energy efficiency in low-temperature-pinch industrial processes. Energy
440 35(5), 1867-1873.
441 Bhargava, R., Khanam, S., Mohanty, B., Ray, A.K., (2008). Simulation of flat falling
442 film evaporator system for concentration of black liquor. Computers & Chemical
443 Engineering 32(12), 3213-3223.
444 Bon, J., Clemente, G., Vaquiro, H., Mulet, A., (2010). Simulation and optimization of
445 milk pasteurization processes using a general process simulator (ProSimPlus).
446 Computers & Chemical Engineering 34(3), 414-420.
447 Bourouni, K., Martin, R., Tadrist, L., Tadrist, H., (1998). Modelling of heat and mass
448 transfer in a horizontal-tube falling-film evaporator for water desalination.
449 Desalination 116(2), 165-183.
450 Bylund, G., (1995). Diary Processing Handbook. Tetra Pak Processing Systems AB,
451 S-221 86 Lund, Sweden.
452 Bylund, G., (2003). Dairy processing handbook. Tetra Pak Processing Systems AB.
453 Gong, L., Shen, S., Liu, H., Mu, X., Chen, X., (2015). Three-dimensional heat
454 transfer coefficient distributions in a large horizontal-tube falling film evaporator.
455 Desalination 357, 104-116.
456 Hall, C.W., Farrow, A.W., Rippen, A.L., (1986). Encyclopedia of Food Engineering,
457 2nd Edition. AVI Publishing Co., Westoirtm, Connecticut, USA.
458 Lee, S., Posarac, D., Ellis, N., (2011). Process simulation and economic analysis of
459 biodiesel production processes using fresh and waste vegetable oil and supercritical
460 methanol. Chemical Engineering Research and Design 89(12), 2626-2642.
461 Madoumier, M., Azzaro-Pantel, C., Tanguy, G., Gésan-Guiziou, G., (2015).
462 Modelling the properties of liquid foods for use of process flowsheeting simulators:
463 Application to milk concentration. Journal of Food Engineering 164, 70-89.
464 Medhat, B., Fateme, Fanaei, M.A., Zohreie, H., (2015). Mathematical modelling and
465 dynamic simulation of multi-effect falling-film evaporator for milk powder
466 production. Mathematical and Computer Modelling of Dynamical Systems 21(4),
467 336-358.
468 Ministry of Business Innovation and Employment, (2017). Energy in New Zealand,
469 Energy and Building Trends. Ministry of Business, Innovation and Employment,, pp.
470 http://www.mbie.govt.nz/info-services/sectors-industries/energy/energy-data-
471 modelling/publications/energy-in-new-zealand/documents-images/energy-in-nz-
472 2017.pdf.
473 Munir, M.T., Yu, W., Young, B.R., (2012a). Recycle effect on the relative exergy
474 array. Chemical Engineering Research and Design 90(1), 110-118.
475 Munir, M.T., Yu, W., Young, B.R., (2012b). A software algorithm/package for
476 control loop configuration and eco-efficiency. ISA Transactions 51, 827-833.
477 Munir, M.T., Zhang, Y., Wilson, D.I., Yu, W., Young, B.R., (2014). Modelling of a
478 falling film evaporator for dairy processes.
479 Munir, M.T., Zhang, Y., Yu, W., Wilson, D.I., Young, B.R., (2016). Virtual milk for
480 modelling and simulation of dairy processes. Journal of Dairy Science 99(5), 3380-
481 3395.
25
482 Mura, E., Gourdon, M., (2017). Pressure drop in dairy evaporators: Experimental
483 study and friction factor modelling. Journal of Food Engineering 195, 128-136.
484 Pisecky, J., Westergaard, V., Refstrup, E., (2012). Handbook of milk powder
485 manufacture. GEA Process Engineering A/S Copenhagen.
486 Ramírez, C.A., Patel, M., Blok, K., (2006). From fluid milk to milk powder: Energy
487 use and energy efficiency in the European dairy industry. Energy 31(12), 1984-2004.
488 Ruan, Q., Jiang, H., Nian, M., Yan, Z., (2015). Mathematical modeling and
489 simulation of countercurrent multiple effect evaporation for fruit juice concentration.
490 Journal of Food Engineering 146, 243-251.
491 Russell, N.T., Bakker, H.H.C., Chaplin, R.I., (2000). A Comparison of Dynamic
492 Models for an Evaporation Process. Chemical Engineering Research and Design
493 78(8), 1120-1128.
494 Saber, N., Shaw, J.M., (2008). Rapid and robust phase behaviour stability analysis
495 using global optimization. Fluid Phase Equilibria 264(1-2), 137-146.
496 Silveira, A.C.P., Tanguy, G., Perrone, Í.T., Jeantet, R., Ducept, F., de Carvalho, A.F.,
497 Schuck, P., (2015). Flow regime assessment in falling film evaporators using
498 residence time distribution functions. Journal of Food Engineering 160, 65-76.
499 Virtual Materials Group Inc., (2014). VMGSim V. 8.0 User's Manual.
500 Winchester, J., (2000). Model based analysis of the operation and control of falling
501 film evaporators, School of Engineering & Advanced Technology Massey University,
502 Palmerston North, New Zealand.
503 Winchester, J.A., Marsh, C., (1999). Dynamics and Control of Falling Film
504 Evaporators with Mechanical Vapour Recompression. Chemical Engineering
505 Research and Design 77(5), 357-371.
506 Zhang, Y., Munir, M.T., Yu, W., Young, B.R., (2014). Development of hypothetical
507 components for milk process simulation using a commercial process simulator.
508 Journal of Food Engineering 121, 87-93.
509
26