Heat Recovery in the Cross-Heat Exchanger

Before entering the reboiler, some heat can be recovered from the hot lean amine stream coming from the desorber in a cross-heat exchanger. Therefore, the temperature of the rich amine stream that enters the reboiler is not that of the amine leaving the absorber (T0 = 60 °C has been fixed so as to reach a compromise between the different cases presented in Section 3) but a higher value TXTX is defined by the temperature of the hot lean amine stream TH and by the area in the cross-heat exchanger. In this work, an ideal TX is assumed to be TX = 105 °C following Xu et al. (54)

In a previous study, Lin and Rochelle (55) had performed a parametric optimization of the cross-heat exchanger in the context of CO2 capture from which this analysis will draw heavily. As those authors pointed out, optimum operation of this unit is dependent on solvent properties such as heat capacity and viscosity. However, their findings were applied to the context of reducing capital and operational costs. Conversely, our analysis will be divided into two different scenarios, case C and case D.

  • Case C: This scenario assumes that the CO2 capture plant already exists, meaning that its capital costs are fixed. The existing capture plant has been designed for traditional solvents, so that the area available for heat transfer will be evaluated for TX = 105 °C when using aqueous absorbents. Once this area is calculated, it will be kept constant while studying different organic diluents. As will be seen in this approach, TX is typically lower than 105 °C, which translates into more sensible heat expenses in the reboiler.

  • Case D: In this scenario, we consider that a new heat exchanger of a different area could be installed to deal with the heating of the rich amine up to TX = 105 °C. This unit will be bigger than that used for warming up the aqueous solvent although calculation of how this impacts the capital costs of the CO2 capture plant is beyond the scope of this study. Overall, the results of this analysis will be more optimistic than those of case C.

The amount of heat that can be transferred in the cross-heat exchanger is given by eq 31a−b.


In eq 31aUHX is the overall heat transfer coefficient, AHX is the area available for heat exchange, and ΔTlm is the log mean temperature difference between rich and lean streams. Equation 31b is also true, since the heat transferred in the unit is responsible for raising the temperature of the rich stream from T0 up to TX. This stream has a mass flow rate of q and a heat capacity of CP. Equality between eqs 31a and 31b results in eq 32 for the exit temperature of the rich stream.


Following the approach of Lin and Rochelle, (55) one can consider this equipment to be a plate-and-frame-type exchanger (PHE), and the following derivations will be based both on that work and on Dhar’s. (56) Assuming a PHE with plate spacing DHX, plate width WHX, plate height HHX, and number of plates NHX, the mass flow rate of the rich solvent q can be calculated by eq 33. Meanwhile, eq 34 expresses the area available for heat transfer. With eqs 3234, one can obtain eq 35.


Lin and Rochelle (55) have assumed that the resistance to heat transfer in the PHE plates is negligible. Furthermore, they propose that the convective heat transfer coefficients h of both rich and lean streams can be averaged. Following this approach, eq 36 suggests that the global heat transfer coefficient of the PHE is half the convective heat transfer coefficient evaluated at the average temperature between rich and lean amine streams.


Substitution of eq 36 into eq 35 yields eq 37.


The convective heat transfer coefficient h can be obtained using the three nondimensional numbers—Nusselt (Nu), Reynolds (Re), and Prandtl (Pr) (eqs 3840)—and a suitable correlation, expressed generically in eq 41, where CNum, and n are empirical parameters. Notice that, in these equations, the characteristic length of the PHE is 2·DHX or 2 times the spacing between plates.


With eqs 3841, the convective heat transfer coefficient h can be approximated by eq 42. Finally, eq 43 shows how the exit rich solvent temperature TX can be calculated. Notice in eq 43 that the first few terms of that expression are related either to the dimensions of the equipment (DHXLHX) or to how one chooses to operate the process (u, ΔTlm), whereas the solvent properties ρ, CP, λ, and η appear in the end. As for the empirical parameters CNum, and n, a good overview of their values has been cataloged by Lin and Rochelle. (55) A common thread among the alternatives is that m and n are always numbers between 0 and 1 and m > n, meaning that TX increases with λ and decreases with ρ, CP, and η for a fixed process configuration and operational conditions.

The choice between the models presented by Lin and Rochelle (55) should not be a big concern in this analysis. It has been shown by those authors that, although the estimated Nu can vary greatly depending on the chosen correlation, its dependency on Re and Pr does not change too much among different models. Therefore, for comparisons between aqueous and water-lean solvents, the Nusselt correlation one selects is not that relevant. In the remainder of the calculations in this section, it is assumed that the model of Okada et al. (57) for PHE with a corrugation angle of 30° can be employed, meaning CNu = 0.157, m = 0.66, and n = 0.4. Also following Lin and Rochelle, (55) one can consider DHX = 0.002 m and ΔTlm = 5 °C.
According to Lin and Rochelle, (55) the fluid velocity inside the cross-heat exchanger is typically between u = 0.32 and 0.42 m·s–1. For a fixed amount of energy transferred, higher values of u will demand a larger AHX, whereas lower velocities require smaller equipment. For the sake of comparison, it is assumed that u = 0.4 m·s–1.
Table 5 shows the results of case C and case D with different water-lean solvents based on 30 wt % MEA when compared with the aqueous amine solvent. The parameters ρ, CP, λ, and η were obtained as functions of temperature for all organic diluents along with MEA, as shown in Section S1, and then treated with the mixing rules presented in Section S2. The effects of CO2 loading in these properties have been neglected for the present analysis.
Table 5. Comparison between the Performance of the PHE Using Different Hypothetical Solvents Based on 30 wt % MEA
name case C: TX calculated (°C) case D: AHX increment (%) for TX = 105 °C
acetone 89 +56
benzaldehyde 81 +112
butanol 80 +118
2-butanol 80 +116
tert-butyl alcohol 78 +134
cyclohexanol 73 +209
cyclohexanone 79 +124
cyclopentanone 91 +43
dimethyl sulfoxide 79 +126
dimethylformamide 85 +77
ethanol 83 +89
ethylene chloride 87 +71
ethylene glycol 78 +127
glycerol 71 +256
heptanol 76 +173
hexanol 77 +156
isoamyl alcohol 78 +136
isobutyl alcohol 77 +153
isopropyl alcohol 79 +132
methanol 90 +49
methyl ethyl ketone 85 +75
nitrobenzene 79 +128
N-methyl-2-pyrrolidone 77 +152
pentanol 78 +144
propanol 80 +113
propionitrile 88 +58
propylene carbonate 79 +124
pyridine 85 +78
sulfolane 76 +171
water 105 +0
As evidenced in Table 5 and already predicted by Yuan and Rochelle, (15) shifting from aqueous to water-lean solvents has an enormous impact on the performance of the PHE. This has three main reasons. The first one is the high viscosity of a good number of organic diluents. The second is that all of these diluents have heat capacities lower than that of water when taken on a mass basis. The third, vastly overlooked, is the thermal conductivity of the diluent. These three factors strongly affect the performance of the PHE, and we have carried out an analysis on the impact of each different parameter in Section S5. The takeaway is that, for water-lean solvents based on N-methyl-2-pyrrolidone or sulfolane for example, this implies either accepting that the rich amine can only be heated up to about 77 °C instead of 105 °C (case C) or that the cross-heat exchanger must be more than twice its original size (case D).

Leave a Comment