Accepted Manuscript

Simulation of a sugar beet factory using a chemical engineering software

®
(ProSimPlus ) to perform Pinch and exergy analysis

Charlène Lambert, Benjamin Laulan, Martine Decloux, Hédi Romdhana, Francis

Courtois
PII: S0260-8774(18)30010-4
DOI: 10.1016/j.jfoodeng.2018.01.004
Reference: JFOE 9137

To appear in: Journal of Food Engineering

Received Date: 13 June 2017

Revised Date: 24 November 2017
Accepted Date: 8 January 2018

Please cite this article as: Lambert, Charlè., Laulan, B., Decloux, M., Romdhana, Hé., Courtois, F.,
®
Simulation of a sugar beet factory using a chemical engineering software (ProSimPlus ) to perform
Pinch and exergy analysis, Journal of Food Engineering (2018), doi: 10.1016/j.jfoodeng.2018.01.004.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
1 Simulation of a sugar beet factory using a chemical engineering software
2 (ProSimPlus® ) to perform Pinch and exergy analysis

3 Charlène Lamberta,b , Benjamin Laulanc , Martine Declouxa,b , Hédi Romdhanaa,b , Francis Courtoisa,b,d,∗
4
a AgroParisTech,UMR1145 Ingénierie Procédés Aliments, F-91300 Massy, France
5
b INRA, UMR1145 Ingénierie Procédés Aliments, F-91300 Massy, France
6
c VERI (Veolia Environnement, Recherche et Innovation), F-78520 Limay, France

7
d UMR QualiSud, University of Montpellier, C.C. 023, Place E. Bataillon, 34095 Montpellier Cedex 5, France

PT
8 Abstract

RI
To reduce energy cost and comply with recent European regulations, there is a growing interest for generic
computer-aided tools to optimize energy consumption in food factories. There are several chemical

SC
engineering software but none dedicated to food processes. Prior to Pinch and exergy analysis, the
entire process needs to be modelled, with specific constraints related to exergy analysis and scientific
bottlenecks related to the estimation of thermophysical properties of food products and the modelling of

U
unit operations specific to food processes. In this study, a generic methodology to model food processes
AN
to in fine allow Pinch and exergy analyses is proposed. The methodology was applied to the sugar beet
process. The developed model allowed simulation of a real sugar beet factory, with deviation ranges
within the uncertainties on industrial data.
M

9 Keywords: Food process simulation, chemical process simulator, generic model.

1. Introduction
D

10

11 In Europe, energy efficiency directives require, for large companies, to carry out frequent energy audits
TE

12 (European Commission, 2013a) and ask for a 20 % decrease of energy consumption and greenhouse
13 gas emission and a 20 % increase of renewable energy use by 2030 (European Commission, 2013b).
Considering also the current economic, energy and environmental context, the food industry may be
EP

14

15 under pressure to comply with new environmental regulations, improve energy efficiency and increase
16 production rate. Therefore, improving energy efficiency in such industry is a major challenge, both
C

17 economic and environmental.

18 Chemical industry benefits from ’chemeng’ computer aided tools (e.g. ProSimPlus® , Aspen Plus® ,
AC

19 Aspen HYSIS® , PRO/II® , gPROMS® ) to simulate their whole industry plants. These simulators are
20 composed of a) a calculation database of thermophysical and equilibrium properties using thermody-
21 namic models, and b) a database of unit operations. They are dedicated to gas and liquid mixtures
22 of pure components. Furthermore, specific methods, such as Pinch and exergy methods, can be used
23 systematically to improve energy consumption in industry (Sikdar and El-Lalwagi, 2001). They are also
24 available in some chemical engineering software (e.g. Pinch and Exergy modules in ProSimPlus® , Aspen
25 Energy Analyzer® and ExerCom® in Aspen Plus® and PRO/II® ). Also, ExerCom® (CCS Energie
26 Advies , 2014) plug-type software allows calculating exergy of material streams. In addition to the exergy
27 calculation, the ProSimPlus® exergy module allows for the estimation of exergy destruction and exergy

∗ Francis Courtois
Preprint
Emailsubmitted to charlene.lambert@agroparistech.fr
addresses: Elsevier January 9, 2018
(Charlène Lambert), benjamin.laulan@veolia.com
(Benjamin Laulan), martine.decloux@agroparistech.fr (Martine Decloux), hedi.romdhana@agroparistech.fr (Hédi
Romdhana), francis.courtois@umontpellier.fr (Francis Courtois)
 ACCEPTED MANUSCRIPT
28 efficiency (Gourmelon et al., 2015). Generally, Pinch method is used to find the minimum potential con-
29 sumption of utilities and the maximum potential saving of thermal streams, while exergy analysis may
30 be applied to locate the inefficiencies in the process.
31 Due to the complex nature of food products, food industry can barely use such powerful tools to
32 simulate their specific processes to optimize their energy consumption. Indeed, many food products are
33 mixtures of gas, liquids and solid components for which almost no available equations can predict their
34 properties. Food process simulation software are generally restricted to a single specific unit operation.

PT
35 For instance, for the simulation of drying, one can cite Symprosis® (Devahastin, 2006), DryPAK®
36 (Pakowski, 1999), DrySel® (Kemp, 1999), Dryer3000® (Lambert et al., 2016), and SP2P® (Schuck
37 et al., 2009). There are few applications of chemical engineering software to the food industry, mostly

RI
38 limited to specific workshops. One can cite works in the fields of dairy industry (i.e. milk concentration
39 by evaporation workshop (Bon et al., 2010; Madoumier et al., 2015)) and sugar industry (e.g. juice

SC
40 concentration by evaporation and ethanol generation workshops (Ensinas et al., 2007; Palacios-Bereche
41 et al., 2013; Ortiz and de Oliveira Jr., 2014)). These workshops are only composed of non-food specific unit
42 operations and of liquid-vapor mixtures for which thermophysical properties can be estimated with classic

U
43 thermodynamic models (e.g. UNIQUAC (Starzak and Mathlouthi, 2006)). Moreover, some specialized
AN
44 commercial software (e.g. Sugar® (Weiss, 1989)) or models developed in laboratories (e.g. using the
45 EES® (Ensinas et al., 2009; Pellegrini and de Oliveira, 2011), QBasic® (Tekin and Bayramoglu, 1998,
46 2001) or Matlab® (Peacock, 2002) languages) allowed for the whole factory simulation and sometimes
M

47 also for the estimation of the irreversibilities and thermodynamic inefficiencies in all unit operations of
48 this process. Similarly, a software dedicated to the petrochemical processes, Aspen Plus® , has also been
D

49 used to perform the simulation of all the unit operations of a sugar cane factory and its exergy analysis
50 (Albarelli et al., 2014). The unit operations specific to the sugar process were simulated with global
TE

51 heat and mass balances, as in the models developed in laboratories, while not allowing for a detailed
52 representation of the irreversibilities and thermodynamic inefficiencies (Gourmelon et al., 2015).
Process simulation combined with Pinch and exergy analysis makes an interesting and powerful ap-
EP

53

54 proach for systematic improvement of food processes. Prior to any energy optimization, as a bottleneck,
55 a model of the factory is required. The methodology is usually in 3 steps (see figure 1):
C

56 1. Perform a detailed flowsheet of the factory and gather all required data. The pro-
AC

57 cess flowsheet of the factory includes all existing unit operations and the heat exchanger network.
58 Required data includes flow-rates, compositions, temperatures and pressures of the streams and
59 operating parameters of each unit operation as well.
60 2. List constituents and find thermodynamic models to predict the thermophysical prop-
61 erties of the mixtures. A number of thermophysical properties are required for each constituent
62 (e.g. molecular weights, liquid and gas specific heats, vaporization enthalpies, liquid and gas thermal
63 conductivities, . . . ). Hence, thermodynamic models are required to estimate the thermophysical
64 properties of the mixtures. When unsure, thermophysical properties or models have to be modified.
65 3. Model the network of unit operations in the factory and validate simulation results
66 with industrial data. If the factory model is not validated, some correction is performed by

2
 ACCEPTED MANUSCRIPT
67 checking the validity of the assumptions and the consistency of the supplied industrial data (see
68 figure 1). It should be emphasized that the required level of details of the model depends on its
69 purpose as it is well described by (Gourmelon et al., 2015). For instance, the level of details needed
70 to compute the mass balance of a process is far lower than the one required for exergy analysis.
71 In order to perform an exergy analysis, a unit operation module should only simulate one energy
72 change (i.e. pressure, temperature or chemical composition) at a time. Consequently, each unit
73 operation should be splitted into a combination of several generic and basic modules to distinguish

PT
74 the thermodynamic irreversibilities of the process.

75 The scientific bottlenecks are, in the above list, the second and third steps (see figure 1). Models of food

RI
76 specific unit operations are mostly unavailable in chemical engineering software. Adapted combinations
77 of simple and generic unit operation modules may overcome this limitation. Nevertheless, the underlying
model may not represent accurately the specificities of food processes, it is hence quite common to

SC
78

79 combine them with specification modules (known as "SPEC" modules in ProSimPlus® ). The latter
80 allow to calculate unknown process parameters in order to ensure some specifications (e.g. composition)

U
81 on certain outputs. However, using this approach, the simulator partly looses, locally, its ability to
82 extrapolate. While still allowing energy optimization –main purpose of the present study– it could
AN
83 hinder the scaling-up of the process.
84 In this work, the above described methodology has been applied successfully to the modelling of
a sugar beet factory to further perform both Pinch and exergy analyses, using ProSimPlus® . This
M

85

86 software was chosen for two reasons. First, it allows for the estimation of the thermodynamic properties,
87 the calculation of heat and mass balances and the selection of a large variety of plant equipments (e.g.
D

88 heat exchangers, evaporators etc.). Second, it supplies energy post-processing tools (i.e. Pinch and
Exergy modules).
TE

89

90 In this study, a brief description of the sugar beet process is given. Afterwards, we introduce the
91 method to model all the unit operations (specific and non-specific) of a sugar beet process using the
EP

92 ProSimPlus® software. The factory model is then validated. To do so, mass and energy balances and
93 industrial data correction are performed.
C

94 2. Sugar beet process overview

AC

95 For confidentiality purposes, the name and precise characteristics of the studied French sugar beet
96 factory are not disclosed in this paper. The studied factory currently consumes about 20,000 t of sugar
97 beet per day and more than 3 GWh/d. It produces about 2,500 t/d of white sugar crystals, 1,300 t/d of
98 low purity1 syrup (about 80 %), 4200 t/d of pressed pulps (about 25 % of dry matter) and 770 t/d of
99 sugar limes. The production is continuous during about 100 days a year, 7 days a week and 24 hours a
100 day. About 360 MWh/d of electrical power are produced and totally consumed within the factory.
101 The sugar beet process involves several chemical engineering operations such as cleaning, slicing,
102 sugar extraction, juice clarification, juice concentration by evaporation, sugar crystallization, and drying-

1 Purity: percentage of sucrose in the soluble dry matter

3
 ACCEPTED MANUSCRIPT
103 storage-packaging. It is well known and described in numerous books (e.g. "Sugar Technology, Beet and
104 Cane Sugar Manufacture" (Van der Poel et al., 1998)). A general flowsheet is given in figure 2. Workshops
105 represented by dotted frames in figure 2 are modelled in ProSimPlus® and are briefly described below
106 (i.e. sugar extraction, juice clarification, concentration by evaporation and crystallization workshops).

107 Sugar extraction

108 Sucrose is extracted from cossettes 2 by contacting with hot water composed of condensates and water
issued from the press (see below). Several solutes diffuse from the cossettes to the juice:

PT
109

110 • "sugar ": sucrose;

RI
111 • "non-sugars": mineral matters, mixed-salts, organic and non-organic matter.

112 Preheating of the cossettes allows for the thermal denaturation of the cell membranes and promotes solute

SC
113 migration. The raw juice (also known as green juice) and pulps (i.e. cellular residues of the cossettes) are
114 the two output streams of the diffuser. The raw juice is recycled in this workshop to heat the cossettes
115 (using scalder and prescalder) and is then purified in the next workshop. Pulps are pressed and are finally

U
116 dried (Van der Poel et al., 1998).
AN
117 Juice clarification

118 The raw juice is treated twice with lime and carbon dioxide gas and is finally filtered. The addition
of lime to the heated raw juice induces a pH increase allowing for the precipitation and the coagulation
M

119

120 of the colloids, precipitation of anions and degradation of invert sugar and raffinose. The lime-treated
121 juice is then saturated with carbon dioxide gas to precipitate lime and calcium carbonate (carbonation
D

122 reaction) and allows the settling down of precipitates.

It should be noted that carbonation sludges are composed of all the precipitates removed during the
TE

123

124 filtration step (Van der Poel et al., 1998).

125 Concentration by evaporation

EP

126 Evaporation increases the dry matter concentration in the juice (Brix3 value raised from 15 % to 70 %)
127 by removing water. In addition, it produces nearly all the hot streams required by the heat exchangers
C

128 of all the workshops (steams and condensates).

129 A multiple-effect evaporator is classically used in this workshop (5 to 7 effects). All streams of steam
AC

130 are condensed in each effect, and condensates are flashed using flash-tanks (Van der Poel et al., 1998).
131 Steam used for the heating of the first effect is produced from pressurized steam. More specifically, in the
132 factory studied in this work, pressurized steam is a) expanded into a turbine to produce electrical power,
133 b) supplied to a turbo-compressor to re-compress steam produced at the first effect, and c) expanded in
134 a pressure valve.

2 Cossettes: small strips of sugar beet

3 Brix: percentage of soluble dry matter in the juice

4
 ACCEPTED MANUSCRIPT
135 Crystallization

136 Crystallization is the only batch unit operation of the sugar beet process. Crystallization and cen-
137 trifugation steps are repeated twice in the studied factory. Sugar crystals, produced in the second crys-
138 tallization step, are dissolved into thick juice in a vacuum boiler. The obtained mixture is vacuum boiled
139 and supersaturated. Sugar crystals are then added to the mixture to initiate crystallization. Crystals are
140 then separated from the liquid using a centrifuge. The obtained "poor-syrup" liquid supplies the second
141 strike of crystallization. During centrifugation, crystals are washed out with condensates to remove the

PT
142 thin layer of liquid around them, and the obtained juice is recycled. The white sugar crystals are further
143 dried in the drying-storage-packaging workshop (Van der Poel et al., 1998).

RI
144 3. Model of the sugar beet process

145 3.1. Constituents and thermodynamic models

SC
146 Thermodynamic properties are calculated using the Simulis Thermodynamics® software. According
147 to the above description of the sugar beet process, the following constituents are required for the modelling
of the streams:

U
148

• sucrose, "non-sugars" and water (for all workshops);

AN
149

150 • cellular residues (specific to the sugar extraction workshop);

M

151 • dinitrogen, dioxygen, carbon dioxide, calcium oxide, calcium carbonate (specific to the juice clari-
152 fication workshop).
D

153 In this analysis, the UNIFAC-Larsen thermodynamic model was preferred for its reliable predictive
154 capabilities. All the retained constituents were found in the 2013 standard package of the Simulis
TE

155 Thermodynamics® database. Using this model, the UNIFAC-Larsen decomposition of each constituent
156 is required. However, since calcium oxide and calcium carbonate were unavailable in the database, they
157 were approximated by pure water. Estimated thermal properties (i.e. heat capacity and boiling-point
EP

158 elevation) were compared to those calculated using (Bubnik et al., 1995) correlations. For a juice pu-
159 rity above 70 % and a juice Brix range of 15 - 90 %, boiling-point elevation were under-estimated by a
C

160 maximum of 1 °C. These deviations are within the range of uncertainties on industrial data.
161 As explained in the previous section, "non-sugars" are composed of a lot of different molecules and
AC

162 their precise relative composition is unknown. Using (Bubnik et al., 1995) correlation, one may consider
163 that assimilating "non-sugars" to "sugar" affects very little the thermal properties of the juices (i.e.
164 heat capacity and boiling-point elevation) with respect to differentiating "non-sugars" from "sugar".
165 Deviations below 5 %, corresponding to an absolute error below 2.5 °C for the boiling-point elevation,
166 were observed for a juice Brix range of 15 - 95 % and a juice purity above 70 %. Hence, in this work,
167 "non-sugars" were assumed to behave as a single constituent with same properties as sucrose.
168 Cellular residues are at solid state. Only liquid-liquid and vapor-liquid heat exchangers exist in this
169 workshop, except for the scalder and prescalder. Hence, precise modelling of cellular residues is pointless
170 for a thermal representation of the factory. In this work, cellular residues were assumed to behave as one

5
 ACCEPTED MANUSCRIPT
171 of its major constituents: cellulose. In addition, its physical state was forced to be liquid by modifying
172 the correlation used for the vaporization pressure.
173 Finally, in the crystallization workshop, sugar crystals were assimilated to sucrose without distinction
174 of its physical state (dissolved or solid) to simplify the model.

175 3.2. Modelling of the flowsheet

176 Sugar extraction, juice clarification, concentration by evaporation and crystallization steps were mod-
elled with the ProSimPlus® chemical engineering software (version 3.5.3.1). Separate files for each

PT
177

178 workshop were used for improved convergence.

179 General hypotheses

RI
180 Following assumptions were considered to model the sugar beet process:

• in all workshops: negligible pressure and thermal losses within pipes and heat exchangers;

SC
181

182 • in the juice clarification workshop: dioxygen, dinitrogen, and carbon dioxide are not dissolved in
183 the juice during both carbonation reactions;

184

U
• in the concentration by evaporation and crystallization workshops:
AN
185 – Negligible thermal losses within the evaporators and pan boilers;

186 – Produced steam is pure water overheated to respect the boiling elevation point;
M

187 The last two assumptions are common in a first step of a factory modelling due to the lack of available
188 industrial data. On the contrary, assuming that a) produced steams are composed of pure water and
D

189 that b) dioxygen, dinitrogen and carbon dioxide are not dissolved in the juice during both carbonation
TE

190 reactions, is questionable. On the one hand, solute concentrations in the steam and concentration of
191 dissolved solutes in the juices are hardly ever measured in sugar beet factories. On the other hand, some
192 specific simulations have shown that their concentrations were negligible with respect to other constituent
EP

193 concentrations in the steam and juice. Hence, one could reasonably expect that these assumptions may
194 not alter the validity of the simulator.
C

195 Food specific unit operations

196 Usually, input variables (i.e. pressure, temperature, composition) of the inlet streams and operating
AC

197 parameters of unit operation modules (e.g. splitting ratio of a "stream splitter" module) are required to
198 simulate a process with a chemical engineering software. Indeed, output variables of outlet streams of
199 the unit operation modules are forward calculated (i.e. pressure, temperature and composition) using
200 the input variables and operating parameters.
201 The ProSimPlus® software has been initially developed for petrochemical industry. Hence no food
202 specific modules were included. To overcome this issue, in this study, unit operations specific to the sugar
203 process were modelled using a combination of generic and basic unit operation modules and "specification"
204 modules. Also known as "constraints and recycles" modules, they are depicted as "SPEC" in following
205 screenshots of ProSimPlus® (see figures 3 to 7). They allow optimizing the operating parameters of the

6
 ACCEPTED MANUSCRIPT
206 basic unit operations to respect a fixed outlet variable (i.e. composition or temperature). Consequently,
207 the more "SPEC" modules are included, the less predictive is the model unless used only for initialization
208 purpose. Moreover, as stated in the introduction, each unit operation was splitted into several modules
209 to distinguish the thermodynamic irreversibilities of the process.

210 Heat exchangers

211 A particularity of the sugar beet process is the fact that the hot stream production and use (in the heat
212 exchangers) are generally located in different workshops. Since each workshop was modelled separately,

PT
213 the following workaround was used for the heat exchangers. To estimate the –unknown– (except for
214 the prescalder) input mass flow-rate of the hot stream for all heat exchangers, a combination of two

RI
215 "cooler/heater", a "SPEC" modules and an "information stream" (see figure 3) was used:

216 • "Heater/cooler" modules to model the heating or cooling of respectively cold and hot streams.

SC
217 Outlet temperatures of the cold and hot streams were fixed as operating parameters. Power required
218 for the heating of the cold streams was provided by the cooling of the hot streams via an "information
219 stream";

U
220 • A "constraints and recycles" (SPEC) module to calculate the mass flow-rate of the hot stream to
AN
221 respect its fixed output temperature.

222 Sugar extraction

Diffusion, which is a unit operation specific to the sugar beet process, was modelled using a combi-
M

223

224 nation of a "mixer" and a "component splitter" modules (see below). "Sugar" and "non-sugars" losses
225 within wet pulps were fixed using a "constraints and recycles" (SPEC) module modifying the recovery
D

226 ratio of each constituent in the "component splitter". An "information stream handler" module allowed
calculating the mass flow-rate of the recycled raw juice to respect the fixed volume withdrawal4 (see
TE

227

228 figure 4).

229 Press modules are available in the ProSimPlus® database (i.e. "belt-filter" and "plate-frame filter"
EP

230 modules). Cake moisture content before filtering and solid fraction in the filtered juice are the required
231 operating parameters of these modules. In this study, the cake constituents were considered as solids by
232 the thermodynamic model. However, cellular residues were only considered in the liquid phase in our
C

233 model. Hence, the press was modelled using a combination of a "component splitter" and a "constraints
AC

234 and recycles" (SPEC) modules (i.e. generic and basic unit operation modules). "Sugar" and "non-sugars"
235 losses (output variables) in the pressed pulps were fixed by a "constraints and recycles" (SPEC) module
236 modifying the split ratio of the constituents (see figure 4).
237 The prescalder (i.e. a cossettes – raw juice heat exchanger) was modelled using a "generalized heat
238 exchanger" module in which output temperature of the raw juice was fixed.

239 Juice clarification

240 As stated earlier, to simplify the modelling of the clarification workshop, four reactions were neglected:
241 precipitation and coagulation of the colloids, precipitation of anions and reaction of degradation of invert

4 Volume withdrawal: ratio of the volumetric flow-rate of the outlet raw juice and cossettes

7
 ACCEPTED MANUSCRIPT
242 sugar and raffinose. Carbonation reactions were modelled using a combination of three modules (i.e.
243 "reaction tank", "component splitter" and "SPEC" modules), see below and the left panel of figure 5.
244 The reactor is composed of two inlet and two outlet streams. Saturated gas and limed juice enter the
245 reactor in which reaction 1 was modelled while reaction enthalpy was fixed using available data in the
246 literature (Van der Poel et al., 1998).

CaO + CO2
CaCO3 (1)

PT
247 Gas and liquid are separated at the outlet of the reactor. A "component splitter" module allowed
248 separating the dissolved molecules in the cloudy juice (dioxygen, dinitrogen, carbon dioxide) from the

RI
249 other ones. The dissolved molecules are then mixed in the gas at the outlet of the reactor. The required
250 mass flow-rate of carbon dioxide (an input variable) was calculated using a "constraints and recycles"
251 module (SPEC) to respect a fixed mass flow-rate of calcium oxide in the carbonation juice.

SC
252 Impurity removal by filtration was modelled with a combination of a "component splitter", two
253 "constraints and recycles", a "stream splitter", a "mixer" and two "information stream handler" modules

U
254 (see below). Calcium carbonate and a part of the non-sugar are washed out with the carbonation sludge.
255 Hence, the obtained liquid was divided into two streams: one represents a part of the carbonation sludge
AN
256 and the other the light juice. The two parts of the carbonation sludge are then mixed. Split ratio of
257 each constituent of "component splitter" and split ratios of the "stream splitter" were calculated using
258 "information stream handlers" and "constraints and recycles" (SPEC) modules (see below and the right
M

259 panel of figure 5). On the one hand, split ratios of each constituent were adjusted to respect a fixed purity
260 of the light juice. On the other hand, split ratios of the "stream splitter" were calculated to respect a
D

261 fixed efficiency for the filters (i.e. ratio of mass flow-rates of the light juice and carbonation sludge).
TE

262 Concentration by evaporation

263 All unit operations existing in the concentration by evaporation step are common in chemical engineer-
264 ing. Specific modules of evaporation or turbo-compression are available in the ProSimPlus® software.
EP

265 However each unit operation of this workshop was split into a combination of several modules to identify
266 the thermodynamic irreversibilities of the process. Model description of each unit operation existing in
267 this workshop is given below.
C

268 Every single effect evaporator was modelled as a combination of ten modules (see figure 6a):
AC

269 • an "expansion valve" module to decrease the pressure of the entering juice (except for the first
270 effect) where part of the water content flash-evaporates;

271 • an exchanger (i.e. a "cooler/heater" module) to heat the juice. The power required for the heating
272 of the juice is provided by the condensation of the steam (via an "information stream"), while the
273 mass flow-rate and the steam temperature are known;

274 • a "liquid-vapor separator" module to separate vapor and liquid phases. Vapor phase contains
275 "sugar" and "non-sugars" at infinite dilution concentration. These constituents are mixed in the
276 liquid phase using a combination of a "component splitter" and a "mixer" modules. The mixture
277 is both the concentrate of the current effect and the liquid to concentrate in the next effect;

8
 ACCEPTED MANUSCRIPT
278 • steam supplying the evaporator is condensed in a flash-tank, which was modelled with a combination
279 of an "expansion valve" and a "liquid-vapor separator" modules. The steam flash-produced is mixed
280 with the purified steam. This mixture is divided into two streams: the first stream supplies the
281 next single effect. The second stream was modelled as a process outlet. This outlet steam was
282 considered as a hot utility for all heat-exchangers of all the workshops and its mass flow-rate was
283 calculated as the sum of the mass flow-rates required for all exchangers (see sub-section 3.2).

284 As explained in the section 2, the steam supplying the first effect, is produced from high pressured steam.

PT
285 The latter is a) expanded into a turbine to produce electrical power, b) supplied to a turbo-compressor
286 to re-compress first-effect steam, and c) expanded using an expansion valve (see figure 6b).

RI
287 The mass flow-rate of high pressured steam supplying the turbine was manually adjusted to respect
288 the required production of electrical power.

SC
289 The turbo-compressor itself was modelled as a combination of a "turbine", a "compressor", and a
290 "constraints and recycles" (SPEC) modules and two "information streams" (see below). The mass flow-
291 rate of high pressured steam supplying the turbine was calculated using a "constraints and recycles"

U
292 module (SPEC) to respect the output pressure of the re-compressed steam. The power required for the
293 steam compression was provided by the steam expansion via the "information stream".
AN
294 Finally, the mass flow-rate of the compressed first-effect steam and total mass flow-rate of high pres-
295 sured steam were manually adjusted to respect a fixed Brix of the concentrated liquid and to minimize
M

296 the steam in excess in the fifth effect.

297 Crystallization
D

298 Vacuum pan boiling, in the crystallization step, is a classical unit operation commonly found in
299 chemical engineering. However, for exergy analysis, it has been modelled as a combination of seven
TE

300 modules to distinguish the thermodynamic irreversibilities of the process (see figure 7):

301 • An "expansion valve" module to decrease pressure of the entering juice while part of the water
EP

302 contained in the juice flash-evaporates;

303 • An exchanger (i.e. a "cooler/heater" module) to heat the juice. The power required for the heating
C

304 of the juice is provided by the condensation of the steam (via an "information stream");

• A "liquid-vapor separator" to separate vapor and liquid phases. Vapor phase contains "sugar" and
AC

305

306 "non-sugars" at infinite dilution concentration. These constituents are mixed to the liquid phase
307 using a combination of a "component splitter" and a "mixer" modules.

308 • Mass flow-rate of the steam supplying the under-vacuum boiler is calculated, using a "constraints
309 and recycles" module (SPEC), to respect a fixed Brix value of the mother liquor and sugar crystal
310 mixture.

311 A centrifuge module is available in the ProSimPlus® database (i.e. "clarifier" module) with solid fractions
312 in the sludge and in the outlet liquid as required operating parameters. The sludge constituents had to be
313 considered as solids in the thermodynamic model. However, sugar crystals were considered in the liquid

9
 ACCEPTED MANUSCRIPT
314 phase in our model. Hence, the centrifuge was modelled with a combination of generic and basic unit
315 operation modules. A combination of two "component splitter" modules was used for its simulation (see
316 figure 7). The first "component splitter" module allows separating sugar crystals from syrups (mixture
317 of rich and poor syrups) and the second one separating the different syrups. The sucrose ratio in the
318 first "component splitter" was calculated using a "constraints and recycles" (SPEC) module to respect
319 mass flow-rate of sugar crystals. Another SPEC module was used to calculate a) the split ratio of each
320 constituent in the second "component splitter" module and b) the mass flow-rate of the condensates

PT
321 washing the sugar crystals in order to respect the composition and mass flow-rate of the rich and poor
322 syrups.

RI
323 4. Validation on industrial data

324 Factory models are usually assumed to be valid if simulation data are within uncertainties on industrial

SC
325 data. Fluctuations in the sugar beet composition affects compositions, temperatures and mass flow-
326 rates for all streams of the process. Hence, data supplied by the factory were the mean of measures
recorded while the production was stabilized. Besides, some of the supplied data were measured daily

U
327

328 while others where calculated by performing mass balances within each workshop (workaround for non-
AN
329 measurable data). Indeed, mass flow-rates, temperatures and compositions of juices (at liquid state
330 only) and pressures, temperatures and mass flow-rates of steams were measured. Mass flow-rates and
331 compositions of all solid-liquid mixtures (e.g. pulps, carbonation sludge, limed juice, sugar limes) were
M

332 calculated.
333 Data were supplied only for validation purpose, no identification of variables or parameters were ever
D

334 performed in this study. However, their consistency was checked by performing global heat and mass
335 balances using a spreadsheet software prior to any process modelling in ProSimPlus® . Finally, due to
TE

336 sensor precision and fluctuations in the stream compositions, the following measurement uncertainties
337 were retained for the supplied data from the studied sugar beet factory:
EP

338 • Temperature: ± 3 °C;

339 • Mass flow-rate: ± 3 %;

C

340 • Brix: ± 2 %;
AC

341 • Purity: ± 2 %;

342

343 Each workshop model was validated against industrial data. Deviations between simulation and
344 industrial data were below 1 % for sugar extraction and crystallization workshops. Nonetheless, many
345 output variables of these workshops were forced by "constraints and recycles" (SPEC) modules. In
346 the sugar extraction workshop, this was due to the fact that a) sugar extraction is a unit operation
347 specific to the sugar beet process and b) cellular residues were considered at liquid state. Concerning
348 the crystallization workshop, since manufacturers can measure the "sugar" content of all streams and the
349 mass flow-rate of white sugar crystals, a simple mass balance allowed calculating all compositions and

10
 ACCEPTED MANUSCRIPT
350 mass flow-rates of this workshop. After fixing a majority of the model parameters (using SPEC modules),
351 the simulation of the industrial workshop gave, without surprise, a deviation from industrial data below
352 1 %.
353

354 For the juice clarification and the evaporation steps, some data correction have been performed prior
355 to modelling. For the juice clarification workshop, industrial data was parsimonious and entailed with
356 large uncertainties due to the difficult measurements of compositions and mass flow-rates of solid-liquid

PT
357 mixtures. Hence, the measured compositions and mass flow-rates of raw juice, thick juice, sugar limes
358 and coke were not consistent in regards to a mass balance. This was due to two reasons. First, measuring
359 mass flow-rate and composition of a solid phase in suspension in a liquid phase is difficult. Second,

RI
360 little fluctuations in the mass flow-rate or composition of the raw juice greatly affect mass flow-rate and
361 composition of the other streams. This is classically observed both in the factories and by simulation.

SC
362 Hence, a global mass balance of this workshop has been performed using a spreadsheet software.
363 Standard deviations on mass flow-rates and compositions of the streams of this workshop were quite
364 substantial (up to 10 %). However, mass flow-rates and compositions of the raw juice, lime (inlet streams)

U
365 and thin juice (outlet stream) were considered to be reliable since these streams are only in liquid state.
AN
366 Hence, these inlet and outlet variables were taken as given by the manufacturer. In addition, all fixed
367 output variables (i.e. purity of the light juice, filter efficiency5 , and sugar lime composition) were either
368 validated industrial data or recalculated values by a global mass balance of the workshop. Indeed,
M

369 recalculated values were in the range of classical values existing in sugar beet factories.
370
D

371 Concerning the model of the concentration by evaporation workshop, the deviations between simula-
372 tion and industrial data were below 1 %, except for the:
TE

373 • temperature of steam produced in each effect (deviation below 0.5 °C for the first three effects, and
374 maximum deviation of 3 °C for the fourth and fifth effects)
EP

375 • dry matter composition of liquid concentrated in each effect (deviation below 3 %);

376 • mass flow-rate of steam supplying the second to the fifth effect (maximum of 3.5 % of deviation);
C

377 • mass flow-rate of re-compressed first effect steam (deviation of 7 %).

AC

378 Observed deviations on the temperatures of the produced steams were due to the selected thermody-
379 namic model determining thermal properties of mixtures (i.e. UNIFAC-Larsen model). Moreover, a
380 little modification of the mass flow-rate or composition of the thin juice is known to greatly affect mass
381 flow-rates and compositions of the streams of the workshop. For this reason, contrary to evaporator
382 pressure, compositions and mass flow-rates of produced steam and concentrated juice were provided by
383 manufacturers with larger uncertainties.
384

5 Filter efficiency: ratio of the mass flow-rate of juice carbonation and carbonation sludge

11
 ACCEPTED MANUSCRIPT
385 5. Conclusion

386 Therefore, the presented model allowed for the simulation (after industrial data correction) of all
387 workshops of the sugar beet process of a typical french factory, with a low deviation range except for
388 some compositions or mass flow-rates which are difficult to measure. The observed discrepancies were,
389 anyway, within the experimental uncertainties. Consequently, energy analysis and optimization of the
390 sugar beet factory can be performed with the current model at the factory scale. Indeed, data required
for the Pinch and Exergy analyses are automatically extracted from the modules in ProSimPlus® and

PT
391

392 all hypotheses of energy optimization can be checked with the model. Furthermore, the lack of some
393 modules, specific to the sugar beet process, made it necessary to use many "SPEC" modules tuned to

RI
394 fit one specific operating point, limiting the ability to predict other operating conditions. Hence, to
395 perform a sensibility analysis on the operating conditions or to scale-up each unit operation (except for

SC
396 the evaporation workshop and heat-exchanger network), three methods may be considered. One may
397 either:

398 • run once then remove all the "SPEC" modules with caution. In this case, the analysis should be

U
399 performed using inlet variable values closed to the identified ones by the "SPEC" modules;
AN
400 • replace fixed mass flow-rates and compositions of output streams by the ratio of input/output
401 variables;
M

402 • develop modules for the food specific unit operations using "script modules" in ProSimPlus® . This
403 third method is more time consuming (at least seven man-days for each unit operation).
D

404 Besides, the "SPEC" modules can also be kept, in the model, to identify the operating parameters
required for the simulation of another sugar beet factory (using other industrial data).
TE

405

406 Current work in progress is focused on optimizing the energy consumption of this factory by combining
407 exergy and thermal Pinch analysis.
EP

408 6. Acknowledgments

409 The authors wish to thank the French Agence Nationale de la Recherche (ANR) for their funding, and
C

410 the partners of the project COOPERE2 led by VERI (Veolia Research & Innovation), in collaboration
AC

411 with ProSim and LGC (Laboratoire de Génie Chimique, ENSIACET).

12
 ACCEPTED MANUSCRIPT
412 Albarelli, J. Q., Ensinas, A. V., Silva, M. A., 2014. Product diversification to enhance economic viability
413 of second generation ethanol production in brazil: The case of the sugar and ethanol joint production.
414 Chemical Engineering Research and Design 92 (8), 1470–1481.

415 Bon, J., Clemente, G., Vaquiro, H., Mulet, A., 2010. Simulation and optimization of milk pasteurization
416 processes using a general process simulator (ProSimPlus) . Computers & Chemical Engineering 34 (3),
417 414–420.

PT
418 Bubnik, Z., Kadlec, P., Urban, D., Bruhns, M., 1995. Sugar technologists manual, Chemical and physical
419 data for sugar manufacturers and users. Bartens, p. 417.

RI
420 CCS Energie Advies , 2014. What is exercom? http://www.cocos.nl/en/548/ccs/Wat_is_ExerCom.html.

421 Devahastin, S., 2006. Software for drying/evaporation simulations: Simprosys. Drying Technology 24 (11),

SC
422 1533–1534.

423 Ensinas, A., Modesto, M., Nebra, S., Serra, L., 2009. Reduction of irreversibility generation in sugar and
424 ethanol production from sugarcane. Energy 34 (5), 680–688.

425

U
Ensinas, A. V., Nebra, S. A., Lozano, M. A., Serra, L. M., 2007. Analysis of process steam demand
AN
426 reduction and electricity generation in sugar and ethanol production from sugarcane. Energy Conversion
427 and Management 48 (11), 2978–2987.
M

428 European Commission, 2013a. Energy Efficiency Directive. Article 8: Energy audits and energy manage-
429 ment systems.
D

430 European Commission, 2013b. Green paper, a 2030 framework for climate and energy policies.
431 URL http://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX:52013DC0169
TE

432 Gourmelon, S., Thery-Hetreux, R., Floquet, P., Baudouin, O., Baudet, P., Campagnolo, L., 2015. Exergy
433 analysis in prosimplus® simulation software: A focus on exergy efficiency evaluation. Computers &
EP

434 Chemical Engineering 79, 91–112.

435 Kemp, I. C., 1999. Progress in dryer selection techniques. Drying Technology 17 (7-8), 1667–1680.
C

436 Lambert, C., Goujot, D., Romdhana, H., Courtois, F., 2016. Toward a generic approach to build-up air
AC

437 drying models. Drying Technology 34 (3).

438 Madoumier, M., Azzaro-Pantel, C., Tanguy, G., Gésan-Guiziou, G., 2015. Modelling the properties of
439 liquid foods for use of process flowsheeting simulators: Application to milk concentration. Journal of
440 Food Engineering 164, 70–89.

441 Ortiz, P. S., de Oliveira Jr., S., 2014. Exergy analysis of pretreatment processes of bioethanol production
442 based on sugarcane bagasse. Energy 76, 130–138.

443 Pakowski, Z., 1999. Simulation of the process of convective drying: Identification of generic computation
444 routines and their implementation in a computer code drypak. Computers and Chemical Engineering
445 23, Supplement (0), S719–S722.

13
 ACCEPTED MANUSCRIPT
446 Palacios-Bereche, R., Mosqueira-Salazar, K. J., Modesto, M., Ensinas, A. V., Nebra, S. A., Serra, L. M.,
447 Lozano, M.-A., 2013. Exergetic analysis of the integrated first- and second-generation ethanol produc-
448 tion from sugarcane. Energy 62, 46–61.

449 Peacock, S. D., 2002. The use of simulink for process modelling in the sugar industry. In: Proceedings of
450 South African Sugar Technologists’ Association. Vol. 76. pp. 444–455.

451 Pellegrini, L. F., de Oliveira, S., 2011. Combined production of sugar, ethanol and electricity: Thermoe-

PT
452 conomic and environmental analysis and optimization. Energy 36 (6), 3704–3715.

453 Schuck, P., Dolivet, A., Méjean, S., Zhu, P., Blanchard, E., Jeantet, R., 2009. Drying by desorption: A

RI
454 tool to determine spray drying parameters. Journal of Food Engineering 94 (2), 199–204, food Powder
455 Technology.

SC
456 Sikdar, S. K., El-Lalwagi, M. (Eds.), 2001. Process design tools for the environment. Taylor & Francis.

457 Starzak, M., Mathlouthi, M., 2006. Temperature dependence of water activity in aqueous solutions of

U
458 sucrose. Food Chemistry 96 (3), 346–370.

Tekin, T., Bayramoglu, M., 1998. Exergy analysis of the sugar production process from sugar beets.
AN
459

460 International Journal of Energy Research 22 (7), 591–601.

461 Tekin, T., Bayramoglu, M., 2001. Exergy and structural analysis of raw juice production and steam-power
M

462 units of a sugar production plant. Energy 26 (3), 287–297.

463 Van der Poel, P. W., Schiweck, H., Schwartz, T., 1998. Sugar Technology, Beet and Cane Sugar Manu-
D

464 facture. Verlag Dr. Albert Bartens KG, p. 1118p.

TE

465 Weiss, L. W., March 1989. Sugar factory process optimization using the sugars computer program. Tech.
466 rep., American society of sugar beet technologists, Englewood, Colorado 80110 USA.
C EP
AC

14
 ACCEPTED MANUSCRIPT

PT
RI
Step 1 : Carrying out a precise flowsheet of the factory and gathering of all required data

SC
Step 2 : Choice of constituents and thermodynamic model

UNo
AN
Step 2' : Validation of the prediction of thermodynamic model
Yes
M

Step 3 : Modeling of the unit operation network of the factory

D

No Yes
Step 3' : Validation of the factory model
TE

Industrial data correction - Systematic energy analysis of the factory

EP

- Scaling up of the workshops

- Updating the equipments ....
C

Figure 1: Modelling steps of a food factory using a chemical engineering software

AC

15
 ACCEPTED MANUSCRIPT

Process inlets
Process outlets
Sugar beet

Stones, sand,
Cleaning organic waste

PT
Slicing,
Condensates Press Pressed pulps
sugar extraction

RI
Raw juice
Coke

Coke CaO Sugar limes

Juice clarification Carbonation Vapor
oven

SC
CO2

Limestone
Thin juice

U
Electricity
Steam Condensates
AN
(40 bar) Concentration Vapor
(multiple effect evaporators (5 effects)) Thick juice
storage

Thick juice
Steam
M

Low purity syrup

Crystallization, centrifugation
Vapor
D

(two crystallizers and centrifuges)

Syrup, sugar cristals Wet sugar cristals

TE

Drying, storage, White sugar crystals

packaging
EP

Figure 2: Sugar beet process diagram

C
AC

Figure 3: Flowsheet of the model for the heat exchanger (screenshot from ProSimPlus® software)

16
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Figure 4: Flowsheet of the model for the sugar extraction workshop (screenshot from ProSimPlus® software). Note: ’green
M

juice’ refers to raw juice.

D
TE
C EP
AC

Figure 5: Flowsheet of the carbonation and filtration unit operations (screenshot from the ProSimPlus® software)

17
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
(a) Flowsheet of the model for a single effect evaporator
M
D
TE
C EP
AC

(b) Flowsheet of the model for the production of steam supplied to the first effect

Figure 6: Flowsheets of the model for the ’concentration by evaporation’ workshop (screenshot from ProSimPlus® software)

18
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP

Figure 7: Flowsheet of the model for the vacuum boiler and the centrifuge (screenshot from ProSimPlus® software)
AC

19
 ACCEPTED MANUSCRIPT
• innovative use of a chemical engineering software for the simulation of an entire plant in the
sugar industry, confronted with actual data from industry
• provides a methodology to simulate food operations by assuming food is equivalent to
mixtures of simple chemicals
• provides and discusses solutions for each workshop of the plant
• provides a solution to unlock further energy optimization at the plant level
• discusses the limits of the proposed methodology for advanced optimization

PT
RI
U SC
AN
M
D
TE
C EP
AC