pubs.acs.

org/IECR Article

Modeling and Simulation of an Industrial Formaldehyde Absorption

System
Catarina G. Braz, Nicole Lutters, Jorge Rocha, Ricardo Alvim, Eugeny Y. Kenig, and Henrique A. Matos*

Cite This: Ind. Eng. Chem. Res. 2020, 59, 5996−6006 Read Online

ACCESS Metrics & More Article Recommendations

See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
Downloaded via KAROLINSKA INST on July 17, 2020 at 23:39:46 (UTC).

ABSTRACT: Although the absorption of formaldehyde is a crucial stage within its production process, the number of models
available in the literature is limited. The complexity of the absorption of formaldehyde in water, in which chemical reactions in the
liquid phase are combined with multicomponent mass transfer, makes the development of an adequate process model a challenging
task. In this work, a rate-based model of an industrial absorption system for tray and packed columns was developed. The
multicomponent mass transfer was described based on the two-film theory, and the model was implemented in the simulation tool
gPROMS ModelBuilder 5.1.1 and validated against data from the design flowsheet provided by the industrial unit licensor and
typical operation values measured in an industrial formaldehyde production unit. The simulation results were in good agreement
with the expected data. With the validated model, the influence of some operating parameters on the column behavior was studied.

1. INTRODUCTION (CH 3 OCH 2 OH) and poly(oxymethylene) hemiformals

Formaldehyde is produced in large amounts due to its broad [CH3O(CH2O)mH].1
application as a raw material in industrial and end-use products It is possible to obtain concentrated formaldehyde up to
in more than 50 industrial segments. Production of form- 95 wt % through evaporation, but temperatures of about 120 °C
aldehyde by partial oxidation of methanol on an iron− are necessary to prevent its solidification. Alternatively, it is also
molybdenum catalyst, commonly referred to as the Formox marketed in solid form as its cyclic trimer, trioxane [(CH2O)3]
process shown in Figure 1, is one of the main formaldehyde or as its polymer, paraformaldehyde [HO(CH2O)mH] with m =
manufacturing processes. In this process, methanol vaporizes,
8−100 monomers of formaldehyde.2
mixes with fresh air and recycled gas from an absorption system
and reacts in a multitubular packed bed reactor. The resulting The development of a reliable model describing absorption
product stream consists of nitrogen (75 vol %), water (10 vol %), columns is very important for the design, control, and
formaldehyde (7.0 vol %), oxygen (6.0 vol %), and traces of optimization of the formaldehyde absorption process. A higher
methanol, carbon monoxide, and dimethyl ether. Formaldehyde concentration of formaldehyde in the liquid stream reduces the
is highly unstable in its pure gaseous form. Therefore, the column performance. Moreover, it also increases the extent of
product stream leaving the reactor must be sent to a system of the polymerization of methylene glycol and, consequently, the
absorption columns, in which formaldehyde is collected as an
aqueous solution, normally with concentration 30−55 wt %,
commonly referred to as formalin. In such solutions, form- Received: December 23, 2019
aldehyde reacts with water to form methylene glycol Revised: February 19, 2020
[CH2(OH)2] that polymerizes to higher poly(oxymethylene) Accepted: February 28, 2020
glycols [HO(CH2O)mH]. Methanol is added to these mixtures, Published: February 28, 2020
in order to prevent the formation of long-chained polymers, by
reacting with formaldehyde to yield hemiformal

© 2020 American Chemical Society https://dx.doi.org/10.1021/acs.iecr.9b07013

5996 Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 1. General scheme of the Formox process.1

Figure 2. Scheme of the industrial formaldehyde absorption system.

risk of formation of paraformaldehyde, which solidifies at low absorption system presented in Figure 2. The lack of published
temperatures and damages the columns. models capable of capturing the complexity of formaldehyde
The number of models available to describe formaldehyde absorption columns was the main motivation of our work. The
absorption columns is scarce. Winkelman3 developed two above mentioned study by Winkelman3 was concerned with a
nonequilibrium stage models of an absorption column with two completely different system than that considered in our
packed sections with external cooling, one considering only the investigation, regarding both the applied catalyst and absorption
chemical reactions of formaldehyde with water and another one column configuration. Winkelman3 investigated a single
additionally taking the methanol reactions into account. absorption column, whereas our system consists of two columns,
Acceleration of mass transfer due to the chemical reactions a first one with two sections with random packing and a second
was accounted for via enhancement factors. Results obtained by one with trays. In his modeling, Winkelman3 used overall mass-
the author indicated that approximate methods for the transfer coefficients with enhancement factors. In our treatment,
estimation of mass-transfer enhancement factors might fail
we do not use such simplifications.
when the chemical reaction is accompanied by simultaneous
Our work suggests an extension of a previously developed
desorption of the product to the gas phase. Winkelman3 also
model,5 in which the same column configuration (cf. Figure 2)
studied the influence of process parameters on the performance
of the absorption column and found that lower liquid was modeled, but under simplified assumptions of only gas-to-
temperatures and an increase in the recirculation fraction and liquid mass transfer of water and formaldehyde and only the
the water make-up would enhance absorption efficiency. reaction of these two components to form methylene glycol and
Another study of a formaldehyde absorption column was its polymers. In the present study, the transfer of methanol to the
presented by Brásio et al.4 The authors used a simple liquid phase, its reaction with formaldehyde in the liquid phase,
equilibrium stage concept, but, considering the complex and and the following polymerization reactions, as well as the
kinetics-dominated nature of the system studied, the selected evaporation of methylene glycol and hemiformal are taken into
modeling basis is certainly too simplified to describe the account. The model is validated using industrial data, and a
occurring phenomena accurately. sensitivity analysis is performed to find the operating parameters
The aim of this work was to create a comprehensive model having the highest impact on the performance of the absorption
and simulate the behavior of the industrial formaldehyde columns.
5997 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

2. STUDIED INDUSTRIAL ABSORPTION SYSTEM Instead of reacting with water, formaldehyde can react with
The main streams of the absorption system under study are methanol (M) in solution to form hemiformals (HF) through
schematized in Figure 2. The gas coming from the oxidation reaction 5, which, in turn, polymerize to poly(oxymethylene)
reactor enters the bottom of the first column, T-1, and passes hemiformals (HFm) by means of reaction 6.6
through two randomly packed sections filled with β-rings and kHF
separated by two valve trays. The heat of reaction in the liquid CH 2O + CH3OH HooI CH3OCH 2OH
*
kHF (5)
phase is removed by external cooling in the heat exchangers E-1
and E-2. The gas is further directed to the second column, T-2, kHFm
with bubble-cap trays and internal cooling coils. Cooling water is CH3O(CH 2O)m − 1H + CH3OCH 2OH HoooI CH3O
*
kHF
used to maintain the column T-2 outlet gas temperature below m

35 °C. Higher temperatures promote the vaporization of water, (CH 2O)m H + CH3OH; 2 ≤ m ≤ mmax (6)
which competes with methanol later in the adsorption in the
oxidation reactor catalyst. Two trays below the upper tray in with reaction rates
column T-2, a caustic solution is added to promote absorption, rHF = ctkHF(x FxM − x HF/KHF) (7)
especially at this stage, where the concentration of formaldehyde
in the gas is very low. rHFm = ctkHFm(x HFx HFm−1 − x HFmxM /KHFm) (8)
In the lower packing segment in column T-1, the liquid outlet
stream is cooled down, and a fraction is collected with a
formaldehyde concentration of ca. 55 wt % (F50), whereas the 4. VAPOR−LIQUID EQUILIBRIUM
rest is recycled to the top of this section. The liquid flowing As previously stated, multicomponent formaldehyde solutions
through the upper packing of column T-1 is entirely directed to are complex mixtures due to the tendency of formaldehyde to
the bottom of column T-2, where it is mixed with the liquid react with water and alcohols forming long-chained molecules.
flowing from this column. Formalin with a concentration of ca. This influences the phase equilibrium significantly. Further-
35 wt % (F30) can be collected here, depending on the market more, formaldehyde has more affinity for methanol than for
demands, and the remaining column T-2 outlet stream is cooled water; the result is that the influence of methanol in aqueous
down and introduced at the top of column T-1. formaldehyde solutions cannot be neglected even for very small
The tail gas leaving column T-2 is subdivided into two concentrations.7
streams, one is mixed with fresh air and sent to the methanol Maurer8 and Brandani et al.9 have developed physicochemical
oxidation reactor, whereas the other is directed to an emission models that describe the phase behavior in the binary systems,
control system for the complete oxidation of the hydrocarbons formaldehyde−water and formaldehyde−methanol and ex-
to CO2. The heat of reaction is used as a heat source for the tended them later to ternary systems with chemically inert
creation of superheated steam that goes through a steam turbine (e.g., trioxane)10,11 or reactive (e.g., methanol)10,12 compo-
to produce electricity. nents. For the ternary system, formaldehyde−water−methanol,
the physicochemical model can be schematized as presented in
3. REACTION KINETICS OF AQUEOUS Figure 3.
FORMALDEHYDE SOLUTIONS WITH METHANOL
The design and optimization of formaldehyde absorbers are
complicated due to several factors, above all, the reactions in the
liquid phase and the process exothermicity. Formaldehyde (F)
reacts with water (W) to form methylene glycol (MG) by
reaction 1 and higher poly(oxymethylene) glycols (MGm) by a
series of reversible reactions, according to reaction 2.6 Like most
gases, formaldehyde itself is only mildly soluble in water. The
good apparent solubility of formaldehyde in water is, in fact, the
good solubility of methylene glycol, and the capacity of the
solution to accommodate poly(oxymethylene) glycols.
kMG
CH 2O + H 2O HoooI CH 2(OH)2
*
kMG (1)
Figure 3. Scheme of the vapor−liquid phase and chemical equilibrium
kMGm for the ternary system formaldehyde−water−methanol.13
CH 2(OH)2 + HO(CH 2O)m − 1H HooooI HO(CH 2O)m H
*
kMGm
Because of the higher boiling point and lower vapor pressure
+ H 2O; 2 ≤ m ≤ mmax (2) of higher polymers, it is assumed that in the vapor phase, only
monomers are present. The thermodynamic model includes the
mmax is the number of monomers of the longest polymer phase equilibria of formaldehyde, water, methanol, methylene
considered. The reaction rates for the production of methylene glycol, and hemiformal, described by eqs 9 and 10, and the
glycol and poly(oxymethylene) glycol are described by eqs 3 and chemical equilibria of the methylene glycol, hemiformal,
4, respectively. poly(oxymethylene) glycols, and poly(oxymethylene) hemi-
rMG = ctkMG(x Fx W − xMG/KMG) formals formation, described by eqs 11−16.
(3)
yP
i
= xiγi(T , x)PiSat(T ), i = F, W, MG1, M, HF1
rMGm = ctkMGm(xMGxMGm−1 − xMGmx W /KMGm) (4) (9)

5998 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

B the addition of small amounts of methanol to formaldehyde

ln PiSat(T ) = A +
T+C (10) solutions. The experimental results revealed that the effect of
chemical reactions in the vapor−liquid equilibrium becomes
yMG Pref
KMGgas = × more important at lower temperatures, and that, although in
yF yW P (11) most cases the addition of methanol results in lower form-
aldehyde volatility, for low temperatures, the addition of small
PFSatPW
Sat
quantities of methanol to a binary formaldehyde−water solution
KMG liq = KMGgas Sat may cause a considerable increase of the formaldehyde volatility.
PMG Pref (12)
xMGmx W γMG γW 5. MODEL DEVELOPMENT
m
KMGm = × The simulation and design of absorption and some other
xMGxMGm−1 γMGγMG (13)
m−1 multistage separation operations are often based on the
yHF Pref equilibrium stage model. The key assumption of this model is
KHFgas = × that gas and liquid streams leaving a packing stage or tray are in
yF yM P (14) equilibrium. An absorption column is modeled as a sequence of
such stages. In the case of reactive absorption, chemical
PFSatPM
Sat
reactions are accounted for via reaction equilibrium or rate
KHFliq = KHFgas Sat
PHF Pref (15) expressions implemented in the mass and energy balances. If the
reactions are very fast, the stage can be reasonably well
x HFmxM γHF γM described, assuming reaction equilibrium. In practice, thermo-
m
KHFm = ×
x HFx HFm−1 γHFγHF dynamic equilibrium is rarely achieved, and deviations are
m−1 (16)
accounted for by empirical parameters, for example, tray
In aqueous formaldehyde solutions, the concentration of efficiencies or HETP-values (height equivalent to one
unhydrated formaldehyde is very low because the equilibrium theoretical plate). However, this simplification often fails in
described by eq 1 is shifted far to the right. In more concentrated multicomponent systems, which is the case of reactive
solutions, more unhydrated formaldehyde is present, but then absorption.18,19
the polymerization reaction 2 intensifies, and the concentration The rate-based model is considered to be a more reliable
of methylene glycol becomes too difficult to quantify. The same approach to model separation processes when reaction kinetics
problem exists with the formation of hemiformal. Therefore, the have a considerable effect on the mass transfer.18−21 In this
oligomer distributions in the liquid and gas phases are approach, the actual rates of multicomponent mass and energy
determined from the mass balances related to the reactions in transfer are taken into account, coupled with the kinetics of the
the liquid and gas phases (eqs 17−22). occurring reactions. The stage mass and energy balances for each
phase are written separately. The equations are then numerically
x F̃ = solved for the total column height. Mass and energy transfer at
m m
x F + xMG + ∑mmax (m ·xMGm) + x HF + ∑mmax (m ·x HFm) the interface can be described by the two-film model, or by the
=2 =2
m m penetration/surface renewal model. The advantage of the two-
1 + xMG + ∑mmax
=2
(m ·xMGm) + x HF + ∑mmax
=2
(m ·x HFm) film model is that the fluid dynamics in the columns can be
(17) described by a broad spectrum of empirical correlations for the
holdup, pressure drop, interfacial area, and mass-transfer
xM
̃ = coefficients available in the literature.22
m
xM + x HF + ∑mmax x
= 2 HFm
5.1. Model Assumptions. The major model assumptions
m m are:
1 + xMG + ∑mmax
=2
(m ·xMGm) + x HF + ∑mmax
=2
(m ·x HFm)
• The absorption system operates at a steady-state;
(18)
• A plug-flow regime is considered for both the liquid and
x ̃W = 1 − x F̃ − xM
̃ (19) the gas phases, that is, no axial dispersion, as the calculated
yF + yMG + yHF Peclet numbers in the gas and liquid phase are above
yF̃ = 100,000;
1 + yMG + yHF (20) • The absorption columns operate adiabatically;
yM + yHF • The interfacial area is the same for mass and heat transfer;
yM̃ = • Except for monomeric formaldehyde, water, and meth-
1 + yMG + yHF (21) anol, only the reaction products of formaldehyde with
water and methanol in the liquid phase vaporize
yW̃ = 1 − yF̃ − yM̃ (22) (methylene glycol and hemiformal);
x̃ and ỹ are the overall mole fractions of formaldehyde, methanol, • The liquid in the distributor above the bottom packing of
and water in the liquid and gas phases, respectively. column T-1 and at the bottom of both columns is ideally
The gas phase was treated as a mixture of ideal gases, and the mixed, and the heat and mass transfer is negligible;
non-idealities in the liquid phase were taken into account by the • The effect of the caustic solution added in column T-2 is
UNIFAC group contribution method to calculate the liquid accounted for by fixing the pH in the liquid for both
component activity coefficients. This model was revised by columns;
Hasse and Maurer13,14 to account for the influence of slow • Because information on gas-phase reaction kinetics is not
reactions at temperatures below 320 K and later by Albert et available, it was assumed that reactions take place in the
al.,15−17 from the same research group, to verify the influence of liquid phase only. Some preliminary simulations were
5999 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Table 1. Chemical Equilibrium and Reaction Rate Constants

equilibrium and rate constants references
2579
ln KMG = − 2.325 + Siling and Akselrod23
T
kMG = (1 + 870 × 10−pH + 6.3 × 10−8 × 10 pH) × e8.962 − 1913/ T (K) Schecker and Schulz24
560.9
ln KMG2 = 0.01449 + Hahnenstein et al.25
T
kMG2−5 = (1 + 247.3 × 10−pH + e−0.838 − 3102/ T × 10+pH)e23.12 − 8551/ T Hahnenstein et al.26
460.4
ln KMG3−5 = − 0.1084 + Hahnenstein et al.25
T
3512
ln KHF = − 1.902 + Hahnenstein et al.26
T
* = 10.9 − 4939
ln kHF Hahnenstein et al.26
T
503.2
ln KHF2−5 = − 0.3476 − Hahnenstein et al.25
T
kHF3−5 = (1 + 1413 × 10−pH + e−5.574 − 2228/ T × 10+pH)e21.79 − 10190/ T Hahnenstein et al.26

2mmax
performed, considering the gas-phase reactions using the
R iLb = ∑ υi , jr jLb
kinetic expressions obtained for the reactions in the liquid
phase. As the difference between both cases was marginal, j=1 (26)
it was concluded that the reactions in the gas phase could Additionally, the condition for the summation of the
be neglected; component molar fractions in the liquid phase has to be fulfilled.
• It is assumed that the longest polymers have five repeating n
units (mmax = 5), meaning that there are a total of 10 ∑ xiLb = 1
(2mmax) reactions in the liquid phase. The chemical i=1 (27)
equilibrium and the reaction rate constants for these For the bulk-gas phase, similar equations are used, but they do
reactions are presented in Table 1. not include chemical reactions
5.2. Bulk-Phase Equations. The mass and enthalpy ∂ Gb
balances to calculate the axial compositions and temperature 0=− (y G) − NiGba intAc ; i = 1, ..., n
profiles in the liquid phase are given by eqs 23 and 24, ∂z i (28)
respectively. Because of the presence of chemical reactions in the ∂
liquid phase, the equations include the reaction source term for 0=− (GH Gb) − qGba intAc
∂z (29)
mass in eq 23 and energy in eq 24.
n
∂ Lb ∑ yiGb = 1
0= (xi L) + [NiLba int + ϕLR iLb]Ac ; i = 1, ..., n
∂z i=1 (30)

ÄÅ ÉÑ
(23)

ÅÅ ÑÑ
Additionally, pressure drop in the gas bulk phase is taken into

ÅÅ Lb int Ñ
(LH ) + ÅÅÅq a + ϕL ∑ r j ( −ΔHR, j)ÑÑÑÑAc
account via the relevant correlations, depending on the type of

ÅÅ ÑÑ
2mmax
∂ column internals.
ÅÅÇ ÑÑÖ
Lb Lb
0=
∂z j=1 ∂
P = f (L , G , ρL , ρG , ...)
∂z (31)
− Q removed (24)
5.3. Film Equations. In the film region, it is considered that
The term Qremoved accounts for the heat removed by the mass transfer occurs by diffusion along the normal coordinate to
internal coils in column T-2. As there is no cooling in column the interface, η. Mass transfer is coupled with chemical reactions,
T‑1 and the columns are considered to operate adiabatically, and the differential component balance is given by eq 32 with the
Qremoved is equal to 0 in T-1. The enthalpies of reaction are boundary conditions (33).
calculated from the chemical equilibrium constants according to Lf
eq 25 1 ∂Ni
0= − R iLf ; i = 1, ..., n
δ L ∂η Lf (32)
ΔHR, j = −RBj (25)
xiLf |ηLf = 0 = xiint , xiLf |ηLf = δ L = xiLb; i = 1, ..., n
where R is the universal gas constant, and Bj is taken from
equation ln Kj = Aj + Bj/T, for the calculation of the chemical (33)
equilibrium constants. In diluted systems, the diffusional interactions can be
The production rate of each component, RLb i , is calculated neglected, and the multicomponent diffusion is described
27
using eq 26. using the effective diffusivity, Deff
i (eq 34).

6000 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

ctLf Dieff ∂xiLf However, the liquid phase in a stage is not at chemical
NiLf = − + xiLf NtLf ; i = 1, ..., n − 1 equilibrium; thus, in this work, the gas-phase compositions at
δ L ∂η Lf
the interface are determined similarly as by Winkelman.3 The
(34) overall compositions of formaldehyde, methanol, and water at
The differential energy balance in the liquid film takes into the interface are calculated using eqs 17−19; these values are
account the heat of reaction released (eq 35). then introduced into a submodel with the Maurer method,10 in
which the equilibrium molar fractions, xi and yi, the chemical
∂q Lf equilibrium constants, Kigas and Kiliq, and the partial pressures,
0= + ΔHrLf
∂η Lf (35) PSat
i , are calculated using eqs 9−22, with the revised parameters
obtained by Albert et al.17 The activity coefficients, γj, are
Equation 36 for the heat flux in the liquid film includes a calculated with the UNIFAC method.29 Assuming that activity
conductive and a convective term. coefficients do not vary significantly, partial pressures are then
λ Lf ∂T Lf
n corrected for the real liquid compositions at the interface.
q Lf = − + ∑ (NiLf HiLf ); i = 1, ..., n At the interface, the following continuity equations are valid,
δ L ∂η Lf i=1 stating that the mass and energy fluxes in the gas and liquid
(36) phases are equal
The gas film equations for mass and heat transport are similar NiGf |ηG,f = δ Gf = Niint = NiLf |ηLf = 0 ; i = 1, ..., n (45)
to the equations for the liquid film but without the reaction
terms.
qGf |ηGf = δ Gf = q int = q Lf |ηLf = 0 (46)
1 ∂NiGf
0= Gf
; i = 1, ..., n To solve the set of equations presented above, the axial and
δG ∂η (37) normal coordinates are discretized, and the balances are
transformed into algebraic equations. With 17 components
yiGf |ηGf = 0 = yiGb , yiGf |ηGf = δG = yiint ; i = 1, ..., n considered, the model has 86,064 algebraic equations and is
(38) implemented into the commercial simulator, gPROMS
ModelBuilder 5.1.1.
Gf
ctGf Dieff ∂yi 5.5. Other Model Parameters and Physical Properties.
NiG,f = − + yiGf NtGf ; i = 1, ..., n − 1 The mass-transfer coefficients, specific interfacial area, volu-
δG ∂ηGf
metric hold-up, and pressure drop are included in the model
(39)
using the appropriate correlations according to the type of
1 ∂q
Gf section inside the column. The references for these correlations
0= are shown in Table 2.
δG ∂ηGf (40)
n Table 2. Correlations Used to Calculate the Necessary Model
λG,f ∂T Gf
qGf = − + ∑ (NiGf HiGf ); i = 1, ..., n Parameters and the Pressure Drop
δG ∂ηGf i=1 bubble-cap
(41) β-rings valve trays trays
The liquid and gas film thicknesses, δL and δG, respectively, are mass-transfer Onda et al.30 Scheffe and Andrew32
calculated via eqs 42 and 43 using empirical correlations for the coefficients Weiland31
mass-transfer coefficients, kL and kG. interfacial area Onda et al.30 Scheffe and Andrew32
Weiland31
DL volumetric hold-up Billet and Bennett et al.34 Bennett et
δL = Schultes33 al.34
kL (42) pressure drop Strigle35 Perry’s36 Perry’s36

DG
δG =
kG (43) Physical properties, such as viscosity, surface tension, and
thermal conductivity, are obtained using Aspen Plus V8.8
DL , DG are the liquid and gas average diffusivity, respectively, through the Thermo CAPE-OPEN Standard properties with the
defined by eq 44.28 Di,j is the binary diffusion coefficient. formaldehyde−methanol−water Data Package provided by
n−1 n AspenTech. In this package, the gas phase is modeled using
∑i = 1 ∑ j = 1 (xi + 1 × 10−4)(xj + 1 × 10−4)Di , j the Hayden-O’Connell model, whereas the liquid phase is
D̅ = n−1 n modeled using the UNIFAC model. This property package is
∑i = 1 ∑ j = 1 (xi + 1 × 10−4)(xj + 1 × 10−4)
valid for a temperature range of 0−100 °C, molar fractions of
(44) formaldehyde of 0−0.6, and pressure of 0−3 bar.28
1 × 10−4 is a weighting parameter to provide stability when
compositions change, especially in reactive systems when some 6. RESULTS
compositions may go to zero at the boundary.28 6.1. Model Validation. To validate the absorption system
5.4. Interfacial Equations. The thermodynamic equili- model, two simulation studies were performed. In simulation
brium at the gas−liquid interface is used to connect the gas and study 1, design data provided by the industrial unit licensor were
liquid phase compositions. The equilibrium relationships are used as input values for the system shown in Figure 2. The
based on the method presented by the group of Maurer10 (eqs simulation results were then compared with the expected values
9−22). from the licensor. However, during the normal operation of the
6001 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

absorption unit, the temperatures of the liquid leaving the heat CH3OH F50, −3.2 °C in Tgas at the top of T-2 and −1.6 °C in
exchangers, E-1 and E-2, as well as the temperature of the Tliq at the bottom of T-2).
process water entering column T-2, are much higher than the Figure 4 presents the calculated profiles of temperatures and
design values to prevent the formation of paraformaldehyde at formaldehyde concentrations in the gas and liquid phase in
lower temperatures. Hence, in simulation study 2, the average
temperatures of these streams measured in the industrial unit
were used instead.
In Table 3, the results of simulation study 1 are compared with
the design data provided by the licensor (εrel 1); whereas the

Table 3. Simulation Results Compared with the Licensor

Data
simulation εrel simulation εrel
study 1 1 (%) study 2 2 (%)
WCH2O in F50 (%) 54.5 0.89 55.8 5.34
WCH3OH in F50 (%) 0.53 15.2 0.38 58.9
% recovered CH2O 99.8 0.02 99.4
Tgas top of T-2 (°C) 25.2 0.87 29.2 10.2
Tliq bottom of T-2 37.6 7.48 59.6 2.61
(°C)
Fgas top of T-2 89.1 0.16 89.5
(ton/h)
FF50 (ton/h) 15.6 0.92 15.2

results of simulation 2 are compared with the average values

measured in the industrial unit when they are available (εrel 2).
The relative deviations between the simulated (vSim) and
experimental (vexp) values, as well as the formaldehyde and
methanol overall mass fractions presented in Table 3, are
calculated using eqs 47−49.
|vexp − vSim|
εrel = × 100
vexp (47)
Figure 4. Temperature and formaldehyde concentration profiles along
WCH2O (%) = the height of the columns for simulation study 2. The error bars
represent the range of operational values observed in the industrial unit.
x F̃ MWF
× 100
x F̃ MWF + xM
̃ MWM + (1 − x F̃ − xM
̃ )MWW simulation study 2, where the error bars represent the range of
(48) values measured in the industrial unit. The different sections of
the absorption system are clearly visible in the observed profiles,
WCH3OH (%) especially in the liquid-phase concentration profile of form-
xM
̃ MWM aldehyde.
= × 100 The discontinuities observed at the dimensionless heights z =
x F̃ MWF + xM
̃ MWM + (1 − x F̃ − xM
̃ )MWW 0.28 and 0.61 follow from the configuration of column T-1 with
(49) multiple feeds and outlets.
At the industrial absorption unit, the flow and composition of The concentration of formaldehyde in the liquid phase
the gas leaving column T-2 were not measured; hence, in Table follows the typical operational profiles of the industrial
3, the relative deviations of these values were not calculated for absorption unit, whereas the gas and liquid temperatures fall
simulation study 2. Also, in the available operation history, there within the range of the typical values measured in the industrial
is no registration of an operational day on which the flow of the unit, except for the outlet temperature of the liquid from column
inlet gas to column T-1 is similar to the design value. Therefore, T-1. This difference could be related to the values of reaction
it is impossible to compare the value of the flow of F50 from the enthalpies used in this work, particularly, the enthalpy of the
industry with the one from simulation study 2, so that relative reaction of formaldehyde to methylene glycol in the liquid
deviation was not determined as well. phase, which has a significant influence on the liquid
In the case of simulation study 1, the expected and computed temperature. The values for the enthalpy of reaction of the
results are in accordance. Only the concentration of methanol in hydration of formaldehyde available in the literature vary from
the product has a higher relative deviation, mainly because this is −21.4 to −43.51 kJ mol−1.37,38 The best simulation results were
a very small number. In the case of simulation study 2, although obtained with the lowest value.
the relative deviations are in general higher than those for According to the results of simulation study 1, about 93.6% of
simulation study 1, the absolute deviationswith the exception the formaldehyde in the gas phase is removed in column T-1,
of the weight percentage of methanol in the stream F50 that is whereas in column T-2, only 6.1% of the initial formaldehyde is
more than half of the average operating valueare within an removed but that corresponds to about 600 kg/h of extra
acceptable range (2.3 wt % of CH2O F50, −0.005 wt % of recovered formaldehyde. On the other hand, in simulation study
6002 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Figure 5. Effect of the outlet temperature of formaldehyde solutions from heat exchangers E-1 and E-2 on Fout/Fin, wt % F50, and wt % F30.

2, only 78.8% of formaldehyde is removed in column T-1 and

20.6% in column T-2. This is explained by the fact that the
higher temperatures promote the vaporization of formaldehyde.
Still, column T-2 is able to nearly absorb the rest of the
formaldehyde from the gas phase.
6.2. Sensitivity Analysis. The validated model was used to
carry out a comprehensive analysis of the influence of some
operating parameters. The studied parameters include the
temperature of the liquid streams leaving heat exchangers E-1
and E-2 (Figure 5), the cooling duty in column T-2 (Figure 6),

Figure 7. Effect of the process water flowrate on Fout/Fin, wt % F50, and

wt % F30.

Figure 6. Effect of the cooling duty in column T-2 on Fout/Fin, wt % F50,

and wt % F30.

the flowrate of process water entering column T-2 (Figure 7), Figure 8. Effect of F50 recycling fraction on Fout/Fin, wt % F50, and wt
and the recycling factor of product stream F50 (Figure 8). For % F30.
the first simulation, the design data for these operating
parameters were used. Then, with the exception of the stream performance indicators; therefore, the results are not shown
F50 recycling factor, changes of ±10, 20, and 30% were here.
implemented, and their influence on the performance indicators The results presented in Figure 5 were obtained from two
was studied. These performance indicators include the relative separate studies in which one of the heat exchanger outlet
amount of formaldehyde in the tail gas from column T-2 (Fout/ temperatures was varied, whereas the other remained constant.
Fin) and the formaldehyde concentration in the high and low Although the decrease of the temperature leaving E-1 enhances
concentrate product (wt % F50 and wt % F30). Additionally, the the absorption unit’s efficiency, it also reduces the content of
influence of the temperature of the gas fed to column T-1 was formaldehyde in stream F50, so that attention to the product
studied, but no significant changes were observed in the specifications is needed. The temperature of the outlet stream
6003 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

from E-2 has a greater influence on the concentration of reactor for the production of formaldehyde39,40 to simulate an
formaldehyde in the stream F30 outlet. The decrease of the existing industrial formaldehyde production unit in order to
outlet temperature in E-2 reduces the flowrate of stream F30 study the complete process and to enhance its competitiveness
leaving column T-2 and being recycled to the top of column T-1. in the market through the increase of its production capacity and
This effect increases the concentration of formaldehyde in this cost reduction.
stream, but, at the same time, it also declines the absorption
capacity of both columns. In practice, this effect can be
circumvented via increasing the recycling fraction of the outlet
■ AUTHOR INFORMATION
Corresponding Author
stream F50, which will reduce the concentration of form- Henrique A. Matos − Centro Recursos Naturais e Ambiente,
aldehyde in stream F30, as can be seen in Figure 8. Instituto Superior Técnico, Universidade de Lisboa, 1049-001
As can be seen in Figure 6, increasing the heat removal from Lisboa, Portugal; Phone: +351 21 841 1639;
column T-2 has a positive influence on the absorption system’s Email: [email protected]
efficiency. This is due to a reduction of the liquid temperature in
column T-2 and the consequent decrease in evaporation of Authors
formaldehyde, water, methanol, methylene glycol, and hemi- Catarina G. Braz − Centro Recursos Naturais e Ambiente,
formal to the gas phase. Instituto Superior Técnico, Universidade de Lisboa, 1049-001
The increase of the process water flowrate in column T-2 Lisboa, Portugal; orcid.org/0000-0001-7859-2294
improves the performance of the absorption system (see Figure Nicole Lutters − Chair of Fluid Process Engineering, Paderborn
7), but it decreases the concentration of formaldehyde in the University, 33098 Paderborn, Germany
liquid. ́
Jorge Rocha − EuroResinas, Indústrias Quimicas S.A.−Uma
Finally, Figure 8 shows that the increase in the recycling Empresa do Grupo Sonae Arauco, 7520-064 Sines, Portugal
fraction of stream F50 also increases the efficiency of the ́
Ricardo Alvim − EuroResinas, Indústrias Quimicas S.A.−Uma
absorption unit, but it decreases the formaldehyde content in Empresa do Grupo Sonae Arauco, 7520-064 Sines, Portugal
stream F30. Eugeny Y. Kenig − Chair of Fluid Process Engineering, Paderborn
It is worth noting that generally the increase of the absorption University, 33098 Paderborn, Germany; Gubkin Russian State
capacity of the column system is accompanied by a reduction of University of Oil and Gas, 119991 Moscow, Russian Federation
formaldehyde concentration in the product, which is undesir- Complete contact information is available at:
able because of additional storage requirements and transport https://pubs.acs.org/10.1021/acs.iecr.9b07013
costs.
Notes
7. CONCLUSIONS AND FUTURE WORK The authors declare no competing financial interest.

■
The absorption of formaldehyde is a critical stage in its
production. The reactive nature of formaldehyde results in a ACKNOWLEDGMENTS
high level of process complexity, making detailed and accurate The authors greatly acknowledge EuroResinas and FCT support
modeling essential for the correct description of absorption through the project UID/ECI/04028/2013 and the Ph.D. grant
systems. In this work, a predictive rigorous rate-based model number PD/BDE/113539/2015.

■
capable of describing the effect of chemical reactions on the mass
and heat transport rates by accounting the relevant reaction and NOMENCLATURE
diffusion kinetics for all components in both the bulk and film
liquid phase was developed. The model validation was Symbols
performed by comparing the simulation results with the data Ac column cross-section, m2
provided by the process licensor and with the typical operational aint specific gas−liquid interfacial area, m2 m−3
values observed in the industrial unit. Except for the temperature ct total concentration, kmol m−3
and the methanol concentration in the product stream F50, Deff effective diffusion coefficient, m2 s−1
good agreement between the simulated and experimental data G gas flowrate, kmol s−1
could be obtained. In further applications of the model, in H molar enthalpy, kJ mol−1
particular, during studies of the dynamic column behavior, it is ΔHR enthalpy of reaction, kJ mol−1
very important to assure that the profiles correctly represent the K chemical equilibrium constant
reality. k rate constant of the formation, s−1
A comprehensive study of the influence of operating k* rate constant of the degradation, s−1
parameters on the absorption system behavior was carried out. L liquid flowrate, kmol s−1
The results suggested that the absorber performance is N molar flux, mol m−2 s−1
improved with lower temperatures of the liquid stream entering P pressure, kPa
the top section of column T-1, lower liquid temperatures in PSat
i saturation pressure, kPa
column T-2, and higher process water rates in column T-2 and q heat flux, kJ m−2 s−1
recycling rates in the bottom section of column T-1. It was Qloss heat loss, kJ s−1
observed that the increase in the efficiency of the columns r reaction rate, kmol m−3 s−1
absorption is accompanied by a decrease in the content of RLbi production rate of component i, kmol m−3 s−1
formaldehyde in the liquid. Hence, the recommendations T temperature, K
presented above should be limited by the required product vexp experimental value
specifications and the storage space available. vsim simulation value
In future work, the final absorption system model will be used W overall mass fraction
together with a previously developed model of an oxidation x mole fraction in the liquid phase
6004 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

x̃ overall liquid phase molar fraction (13) Hasse, H.; Hahnenstein, I.; Maurer, G. Revised Vapor-liquid
y mole fraction in the gas phase Equilibrium Model for Multicomponent Formaldehyde Mixtures.
z axial coordinate, m AIChE J. 1990, 36, 1807−1814.
(14) Hasse, H.; Maurer, G. Vapor-Liquid Equilibrium of Form-
Greek Letters aldehyde-Containing Mixtures at Temperatures below 320 K. Fluid
γ activity coefficient Phase Equilib. 1991, 64, 185−199.
δ film thickness, m (15) Albert, M.; Hahnenstein, I.; Hasse, H.; Maurer, G. Vapor−Liquid
εrel relative deviation Equilibrium of Formaldehyde Mixtures: New Data and Model
η film coordinate, m Revision. AIChE J. 1996, 42, 1741−1752.
λ thermal conductivity, W m−1 K−1 (16) Albert, M.; García, B. C.; Kreiter, C.; Maurer, G. Vapor-Liquid
ϕ volumetric hold-up, m3 m−3 and Chemical Equilibria of Formaldehyde-Water Mixtures. AIChE J.
1999, 45, 2024−2033.
Subscripts, Superscripts, and Abbreviations (17) Albert, M.; Coto García, B.; Kuhnert, C.; Peschla, R.; Maurer, G.
F formaldehyde Vapor-Liquid Equilibrium of Aqueous Solutions of Formaldehyde and
Gb gas bulk Methanol. AIChE J. 2000, 46, 1676−1687.
Gf gas film (18) Taylor, R.; Krishna, R. Multicomponent Mass Transfer; Wiley:
HF hemiformal New York, 1993.
HFm poly(oxymethylene) hemiformal with m CH2O segments (19) Kenig, E. Y.; Górak, A. Reactive Absorption. Integrated Chemical
i index of components Processes; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, FRG,
int interface 2005; pp 265−311.
j index of reactions (20) Seader, J. D. The Rate-Based Approach for Modeling Staged
Separations. Chem. Eng. Prog. 1989, 85, 41−49.
Lb liquid bulk
(21) Errico, M.; Madeddu, C.; Pinna, D.; Baratti, R. Model Calibration
Lf liquid film for the Carbon Dioxide-Amine Absorption System. Appl. Energy 2016,
m number of repeating units in the polymer 183, 958−968.
M methanol (22) Kenig, E. Y.; Kucka, L.; Górak, A. Rigorous Modeling of Reactive
MG methylene glycol Absorption Processes. Chem. Eng. Technol. 2003, 26, 631−646.
MGm poly(oxymethylene) glycol with m CH2O segments (23) Siling, M. I.; Akselrod, B. Y. Spectrophotometric Determination
n number of components in the fluid of Equilibrium Constants of the Hydration and Protonation of
W water Formaldehyde. Russ. J. Phys. Chem. 1968, 42, 1479−1482.

■
(24) Schecker, H.-G.; Schulz, G. Untersuchungen zur Hydratation-
REFERENCES skinetik von Formaldehyd in wäßriger Lösung. Z. Phys. Chem. 1969, 65,
221−224.
(1) Gerberich, H. R.; Seaman, G. C. Formaldehyde. Kirk-Othmer (25) Hahnenstein, I.; Hasse, H.; Kreiter, C. G.; Maurer, G. 1H- and
Encyclopedia of Chemical Technology; Wiley-Interscience, 2005; Vol. 12, 13C-NMR Spectroscopic Study of Chemical Equilibria in Solutions of
pp 929−951. Formaldehyde in Water, Deuterium Oxide, and Methanol. Ind. Eng.
(2) Formaldehyde. IARC Monographs on the Evaluation of Carcinogenic Chem. Res. 1994, 33, 1022−1029.
Risks to Humans; International Agency for Research on Cancer, 2006; (26) Hahnenstein, I.; Albert, M.; Hasse, H.; Kreiter, C. G.; Maurer, G.
Vol. 88, pp 39−325. NMR Spectroscopic and Densimetric Study of Reaction Kinetics of
(3) Winkelman, J. G. M. Absorption of Formaldehyde in Water; Formaldehyde Polymer Formation in Water, Deuterium Oxide, and
University of Groningen, 2003. Methanol. Ind. Eng. Chem. Res. 1995, 34, 440−450.
(4) Brásio, A. S. R.; Aseeri, A. S.; Romanenko, A. Nonlinear Model (27) Hüpen, B.; Kenig, E. Y. Rigorous Modelling of NOx Absorption
Predictive Control of the Formaldehyde Production via Catalytic
in Tray and Packed Columns. Chem. Eng. Sci. 2005, 60, 6462−6471.
Oxidation of Methanol. In ISA Conference 2013Saudi Arabia, 2013.
(28) Aspen Plus User Guide, version 10, 2000.
(5) Braz, C. G.; Matos, H. A.; Mendes, A.; Rocha, J.; Alvim, R. Model
(29) Gmehling, J.; Rasmussen, P.; Fredenslund, A. Vapor-Liquid
of a Formaldehyde Absorption System Based on Industrial Data.
Equilibriums by UNIFAC Group Contribution. Revision and
Computer Aided Chemical Engineering; Elsevier, 2018; Vol. 43, pp 25−
Extension. 2. Ind. Eng. Chem. Process Des. Dev. 1982, 21, 118−127.
30.
(30) Onda, K.; Takeuchi, H.; Okumoto, Y. Mass Transfer Coefficients
(6) Hasse, H.; Maurer, G. Kinetics of the Poly(Oxymethylene) Glycol
Formation in Aqueous Formaldehyde Solutions. Ind. Eng. Chem. Res. between Gas and Liquid Phases in Packed Columns. J. Chem. Eng. Jpn.
1991, 30, 2195−2200. 1968, 1, 56−62.
(7) Brandani, V.; Di Giacomo, G. Effect of Small Amounts of (31) Scheffe, R. D.; Weiland, R. H. Mass-Transfer Characteristics of
Methanol on the Vapour-Liquid Equilibrium for the Water-Form- Valve Trays. Ind. Eng. Chem. Res. 1987, 26, 228−236.
aldehyde System. Fluid Phase Equilib. 1985, 24, 307−333. (32) Andrew, S. P. S. Aspects of Gas-Liquid Mass Transfer. Alta
(8) Maurer, G. Phase Equilibria in Chemical Reactive Systems. Fluid Tecnol. Chim. 1961, 153.
Phase Equilib. 1986, 30, 337−352. (33) Billet, R.; Schultes, M. Predicting Mass Transfer in Packed
(9) Brandani, S.; Brandani, V.; Di Giacomo, G. A Physical Theory Columns. Chem. Eng. Technol. 1993, 16, 1−9.
Superimposed onto the Chemical Theory for Describing Vapor-Liquid (34) Bennett, D. L.; Agrawal, R.; Cook, P. J. New Pressure Drop
Equilibria of Binary Systems of Formaldehyde in Active Solvents. Ind. Correlation for Sieve Tray Distillation Columns. AIChE J. 1983, 29,
Eng. Chem. Res. 1991, 30, 414−420. 434−442.
(10) Maurer, G. Vapor-Liquid Equilibrium of Formaldehyde and (35) Strigle, R. F.; Strigle, R. F. Packed Tower Design and Applications:
Water-Containing Multicomponent Mixtures. AIChE J. 1986, 32, 932− Random and Structured Packings; Gulf Publishing Company, 1994.
948. (36) Perry, R. H.; Green, D. W. Perry’s Chemical Engineers’ Handbook,
(11) Brandani, S.; Brandani, V.; Di Giacomo, G. Vapor-Liquid 8th ed.; McGraw-Hill: New York, 2008.
Equilibrium of Water-Trioxane-Formaldehyde Mixture. Fluid Phase (37) Winkelman, J. G. M.; Voorwinde, O. K.; Ottens, M.; Beenackers,
Equilibria 1991, 63, 27−41. A. A. C. M.; Janssen, L. P. B. M. Kinetics and Chemical Equilibrium of
(12) Brandani, S.; Brandani, V.; Tarquini, I. Vapor-Liquid Equilibrium the Hydration of Formaldehyde. Chem. Eng. Sci. 2002, 57, 4067−4076.
of Formaldehyde Mixtures Containing Methanol. Ind. Eng. Chem. Res. (38) Detcheberry, M.; Destrac, P.; Meyer, X.-M.; Condoret, J.-S.
1998, 37, 3485−3489. Phase Equilibria of Aqueous Solutions of Formaldehyde and Methanol:

6005 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006
Industrial & Engineering Chemistry Research pubs.acs.org/IECR Article

Improved Approach Using UNIQUAC Coupled to Chemical

Equilibria. Fluid Phase Equilib. 2015, 392, 84−94.
(39) Braz, C. G.; Matos, H. A.; Mendes, A.; Rocha, J. F.; Alvim, R. P.
Model of an Industrial Reactor for Formaldehyde Production with
Catalyst Deactivation. Comput.-Aided Chem. Eng. 2017, 40, 121−126.
(40) Braz, C. G.; Mendes, A.; Rocha, J.; Alvim, R.; Matos, H. A. Model
of an Industrial Multitubular Reactor for Methanol to Formaldehyde
Oxidation in the Presence of Catalyst Deactivation. Chem. Eng. Sci.
2019, 195, 347−355.

6006 https://dx.doi.org/10.1021/acs.iecr.9b07013
Ind. Eng. Chem. Res. 2020, 59, 5996−6006