You are on page 1of 12

Heat Exchangers in Carbon Dioxide Cascade Systems

Part II. Practical Aspects of Carbon Dioxide Installations


% Initial evaporation, % Evaporation rate, % of R508A
1. A leakage above the liquid surface.
100
90
80
Initial liquid level
70
60
50
40
30
2a. If the CO2 tem- 2c. In a tank with a
20
10
perature drops to be- horizontal exit, both
0 low the triple point, vortex formation and the risk
dry ice can form and for dry ice entering the pipe are

R404A

R410A

R507A

R508A
CO2

NH3

R22

R23

R32

R41

R116

R125

R143a

R407C

R290

R1270

R218
sink to the bottom. decreased.

Figure 14. Leakage of refrigerants and 2b. Dry ice can then 3a. A leak below the liquid sur-
destroy pumps and face.
the formation of dry ice in CO2. block valves. 3b. If the leakage is here, the
The diagram is valid for a leakage above the liquid level. complete tank is drained, ex-
The absolute evaporation rate depends on insulation, re- cept for CO2, see text.
frigerant mass, shape of the vessel, etc. See also text.

3. Some design considerations for carbon dioxide installations.


3.1. Corrosion by carbon dioxide. somewhat devious though. Ice can get stuck, tighten
the leak and then release when the pressure increases.
Carbon dioxide itself is inert to practically all metals and
The positive aspect of it is that a large part of the filling
elastomers but some care should be taken:
might remain even for a leak at the bottom of a vessel.
♦ Metals. When CO2 is mixed with water, the carbonic acid, Carbamate formation. A special case is the formation
H2CO3, is formed. This is corrosive, especially if oxygen of ammonium carbamate when CO2 leaks into the am-
is present as well. Stainless steels are not affected but monia side in a cascade condenser/evaporator. As the
carbon steel, brass, copper and copper alloys are. The CO2 pressure is practically always higher than the am-
corrosive behaviour is impaired by the addition of corro- monia pressure, the leak is into the ammonia side. Am-
sive breakdown products of the oil. Stainless steel PHEs monium carbamate is corrosive and abrasive, i.e. it can
(PHEs) have not had any problem with corrosion due to destroy a compressor. It can be detected as white pow-
CO2, but there are cases of compressor breakdowns due der in the vapour after a direct expansion evaporator,
to too high water content. Thus, some precaution should e.g. by breaking a light circuit. The detection can be
be taken, §3.9. rapid and if action is taken quickly compressor break-
♦ Oil. The oil is not chemically affected by CO2 but CO2 down can be avoided. Carbamate dissolves readily in
dissolves in some oils and at a pressure decrease there water and it can be decomposed to ammonia and CO2
will be foaming. Oil can deteriorate by wear and tear by heating to above 60 °C. The temperature has to be
and high temperature and form corrosive products. Wa- kept until the gases have left the system otherwise they
ter and oxygen form corrosive organic acids with oil de- will recombine.
composition products.
Welded or semi welded PHEs are safer than HEs where
♦ Elastomers. CO2 will not corrode or affect these chemi- the CO2 channel is entirely surrounded by ammonia
cally but if high pressure CO2 diffuses into an elastomer channels. The reason is that leakage through a weld of
it can sometimes break this when the pressure is re- a PHE is to the exterior.
leased and the elastomer removed.
Flooded flow evaporator are probably safer as the car-
3.2. Leaks. bamate is dissolved in the liquid ammonia and rendered
relative harmless what regards the compressors. It will
A leak in a plant can be of either two types, figure 14:
accumulate in the separator-evaporator loop. The de-
♦ A leak at a vessel or its adjoining pipes above the tection will be more complicated though.
liquid level, figure 14.1. 3.3. Carbon dioxide close to the triple point.
There is an initial flashing down to the ambient pressure If a vessel with liquid CO2 is operating close to the triple
followed by an isobaric evaporation of the refrigerant. In point (-55.6 °C), a sudden pressure decrease could cause
a well insulated vessel the evaporation is slow. It is pro- the temperature to drop below this. Dry ice then forms and
portional to the temperature difference and inverse pro- as this it heavier than the liquid it sinks to the bottom of the
portional to latent heat. Ammonia has very little initial vessel. A lump of dry ice can then move along a pipe to
evaporation and the evaporation rate is slow, i.e. it e.g. a circulation pump or a valve and destroy or block this.
takes along time to empty a vessel. R508 leaves the
vessel quickly. This is due to large initial evaporation, If the exit from a vessel feeding a circulation pump is
large t to the outside and low latent heat. hydraulically correctly executed in order to prevent vor-
tices, this danger is substantially decreased, see figure
♦ A leak below the liquid level 14.2. A vortex free exit should be horizontal. In a vertical
In a leak (figure 14.3a) is between the lowest point and exit, at the bottom of a vessel, vortices can easily form es-
the liquid level, the vessel drains to the leak, followed pecially in CO2 with its low viscosity. There are vortex
by evaporation. If the leak is at the lowest point (figure breakers inserted at the exit but the author is not fond of
14.3b), the vessel drains completely. Carbon dioxide is inserts in vessels, which cannot be opened for inspection.
special. It solidifies, when it reaches the triple point and Note also, dry ice is heavier than liquid CO2 thus freeze
remains solid, i.e. it will not drain from a vessel. The ini- damages as in water circuits will not occur. See also figure
tial evaporation remains though. The behaviour is 11 for pressure increase in enclosed spaces.
1. The emergency valve opens, no
other actions. 5. At least two of critical com-
ponents.
2. Managed release of CO2.

3. Emergency
cooler unit.

4. A liquid receiver design for the


vapour pressure at maximum am-
bient temperature.

Figure 15. Plant shut down.

3.4. Plant shut-down. 3.5. Condensate subcooler/vapour superheater.


In case of a compressor shut down, the pressure in the Sometimes subcooling of the condensate is done by su-
CO2 circuit starts to increase. There are some different perheating the suction vapour; mainly for three reasons:
responses to this, see figure 15 as well:
♦ Capacity increase. When the condensate is subcooled
1. No response, the emergency valves release CO2 when less refrigerant evaporates after the expansion valve,
the set pressure is reached. The CO2 is then replaced. i.e. a higher liquid fraction remains. As it is the liquid,
2. A managed release of CO2, similar to 1 but all pres- which gives the capacity, the capacity increases.
sures and temperatures are carefully monitored. This On the other hand, the vapour density at the suction
can mean a saving of the lost CO2. inlet decreases. As the volume flow is constant, the
3. A special emergency cooling unit starts and condenses mass flow decreases, which decreases the capacity.
the vaporized CO2 Less refrigerant circulates but with a higher liquid frac-
4. Pump the liquid CO2 to a vessel, which can stand the tion. The capacity thus depends on the balance be-
highest possible pressure.. tween increase of the liquid content and decrease of
5. The HP system is built with a redundancy e.g. at least the total flow.
two each of the critical components. At the AC temperature program, 2/40 °C, some refrig-
4 & 5 can obviously only be used in case of planned shut erants – e.g. R404A and R507A – gain, others – espe-
down but not in case of a power outage. cially NH3 and R22, loose and the capacity decreases.
In the temperature program studied here, -40/-10 °C, all
except the unimportant R116 and R508A, loose, but the
2. The condensate and vapour pipes decrease is for most refrigerants unimportant.
are running together inside the insula- ♦ Drying of the oil. Some compressor manufacturers re-
tion. quire that the suction inlet superheat is in the order of
18 to 25 K in order to evaporate as much refrigerant as
1. Dedicated condensate sub- possible from the oil droplets. Especially R1270 and
cooler/vapour superheater R1290 but also carbon dioxide are very soluble in the
oil, which otherwise might enter the compressor diluted.
♦ The superheat is controlled after the condensate sub-
cooler/vapour superheater, i.e. the evaporator can run
with little superheat or even wet. This increases the
evaporator performance.
Two methods may be used, see figure 16:
1. A dedicated condensate subcooler/vapour superheater
is used for larger superheats. PHEs are suitable but in
case of some low pressure refrigerants, the low vapour
density causes to many channels as the number of
Condenser
channels are controlled by the pressure drop. Carbon
dioxide with its large vapour density does not have this
problem.
2. If the site permits, the vapour and condensate lines
could run together inside the insulation. This is a cheap
way of getting a sufficient superheat for oil drying but it
Figure 16. Condensate subcooler/ should not be used to control the superheat at the suc-
Vapour superheater. tion inlet as the response time will be far too long.
Cascade condenser with two separate condenser circuits.
12 13

4 1a 1b

11 16 PCond

17 8
5 .

6 PLR 7 2b 18 2a 15
3

“Through” liquid receiver (TLR). “Surge” liquid receiver (SLR).

Place all inlets at one end of the


LR and all exits at the other.
14a. To DX evaporators.
To DX evaporators.
9. Drain. 10. Drain. 14b. To flooded evaporators.

Figure 17. Vents, drains, compressor connections and equalization lines.

3.6. Vents, drains, compressor connections and Inerts cannot pass this lock and are vented at the con-
equalization lines at condensers. densate exit (13). Vent (7) has a double error; inerts can-
not be vented from here and it is too close to the shell.
Figure 17 shows some dos and don’ts when arranging the
condenser piping. I. The refrigerant leaves through the pump (14b) to the
flooded flow evaporators and returns partly vaporized at
A. Connection of the compressor discharge to the con-
(15). The two-phase mixture separates and the liquid re-
denser inlet (1a) versus to the liquid receiver (2a, b).
turns to the pump (14b)
If the hot gas from the compressor passes the liquid re-
ceiver (3) it heats up the liquid, but its temperature is The discharge vapour from the compressor (2a or 2b)
lowered, which reduces the stress on the condenser. of the DX system enters the liquid receiver as well.
The drawback is that a refrigerant close to the bubble The vapour flows, via the equalization line (16), back to
point can cause cavitation in the pumps and in general the condenser inlet (1b) for recondensation.
a loss of capacity. If the vapour connection is at (2b) However, the pressure in the SLR is lower than at the
there is no larger heating of the condensate but a cer- condenser inlet, there is the condenser pressure drop,
tain dampening of pressure variations occurs Pcond, and possibly others.
B. Another method to dampen excessive pressure and/or
At start up, the pressure is equal in all points 1b, 3 &16
temperature variations is to connect a muffler, a vessel
– 18. The liquid levels in the pipe (18) and the vessel
or the like (4), which can impart inertia to the flow.
(3) are equal. Once the cold water enters at the lower
C. A drain from this vessel (4) to the liquid receiver should port, the vapour starts to condense and the pressure
be closed (5) during normal operation. See §3.6 K. decreases to reflect the liquid temperature in 17. The
D. Vents (or drains) should never be placed directly on a lower pressure sucks vapour from the inlet port and the
pipe (6-10) or a vessel, particularly not at low tempera- pressure drop but it also cause the liquid to mount in
ture operation. Moisture can enter from the outside, the pipe (18) until there is a balance between the result-
freeze and block the valve. ing liquid column and the pressure drop.
E. The same is valid for safety valves. The process is the same as drinking through a straw
F. In general, place safety valves, drains, vents, equalization from a glass of water. The pressure is lower at the con-
lines valves, etc. well away from vessels and main pipes. denser exit (the mouth) than in the liquid receiver (the
water surface in the glass) and refrigerant (water)
G. The discharge vapour in (11) proceeds in a straight flow mounts in the condensate pipe (the straw).
from the discharge exit, possibly via a muffler (4), into con-
denser and finally into the “through” liquid receiver (TLR). J. Theoretically, the two-phase flow from the flooded
There might be pressure drops along the flow path but evaporators could enter directly to the condenser but it
these will not disturb the flow. is very difficult to design a condenser, where all the liq-
uid and vapour a distributed equally from channel to
As the inlet to the TLR is flush with the shell, thus free channel. More important, the additional liquid gives an
access to the vapour space, the condensate drips into extra resistance to the heat transfer
the TLR together with possible inerts. The inerts can be
vented from the TLR (12) but also at the condenser exit, K. Note! A faulty placed or open equalization line is a
as shown for the SLR (13). frequent cause of underperforming condensers.
H. The other liquid receiver is of the “surge” type (though it L. The error can be insidious. The equalization line was
is not exactly a true “surge” SLR). The condensate pipe maybe not installed deliberately as such, it is simple a
ends at the very bottom of the SLR, well below the liquid connection, through various pipes and vessels from the
surface. condenser inlet to the liquid receiver.
4. Detection of inert gases, a 6. Venting of CO2 The inlet vapour temperature is substantially de-
vibrating needle. 50 °C creased, from 50 °C to 0 °C in a desuperheater.
Note, the superheat is decreased as well.

2. t to inlet
5. Venting and re-
covery of HFCs. The pinch point is
approaching.
0 °C

5K -10 °C
3. t to cold
side. -14 °C

1. Detection 7. Can CO2


and venting of be vented into
ammonia ammonia wa- -10 °C
ter? 5K
-14 °C
To waste water treatment

Figure 18. Detection of inerts and Fig. 19. The temperature program in a
venting. cascade unit.

3.7. Detection of inerts and venting, figure 18. 3.8. Temperature difference in a cascade unit.
1. Ammonia is the easiest refrigerant what regards detec- Figure 19 shows the temperatures in a cascade conden-
tion and venting. Connect a hose to the vent, dip the ser evaporator. The vapour enters superheated at 50 °C,
other end as deep as possible into a bucket of cold wa- condenses at -10 °C and leaves slightly subcooled at may-
ter. The result is clear be -11 °C. The cooling refrigerant evaporates at -14 °C
Bubbles emerge => Inerts are present and superheats with 5 K to -9 °C.
No bubbles emerge => Inerts are not present If the vapour temperature is decreased, e.g. by a desuper-
Bubbles or no bubbles, practically no ammonia smell is heater or mixing with vapour from a flooded evaporator,
noticeable. Other refrigerant are harder to detect. There the temperature difference to the evaporating refrigerant
are mainly three methods to indicate inerts: decreases and it might even be impossible to keep the
condensing or evaporating temperatures. The “pinch
2. A temperature difference, between condensate out and point” is then approached and the evaporator size ap-
cooling medium in, of less than a couple of degrees. proaches infinity.
3. There is a large temperature drop from the saturation Compare also with an evaporator, § 2.8, figure 13.
temperature in to the condensate temperature out.
Note! It is practically impossible to distinguish between a 3.9. Carbon dioxide quality.
condensate flooding and inerts by using methods in 2 or There are various carbon dioxide qualities, differing mainly
3. See also 6 below. in the water content. Check with the compressor maker of
the CO2 quality and oil type to be used.
4. A vibrating needle in an undampened pressure gauge,
type Bourdon, indicates the presence of inert gases. ♦ A compressor in the CO2 cycle. R744, Refrigerant
quality 4.0 (Ref. 3) with < 10 ppm O2 and < 10 ppm H2O
If inert gases are suspected, venting has to be done. should be used. This is expensive, though.
5. In case of H(C)FCs, venting should only be done into a ♦ Note that some oils, e.g. ester oils are hygroscopic and
recovery unit. their use is thus somewhat questionable.
6. Carbon dioxide can be vented directly to the atmos- ♦ Pump circulation (without compressor). Practically
phere but to an outside location. Note, valve well away any CO2 can be used. PHEs has been used for many
from the vessel and no pipes after the valve. decades in treating all type of CO2 qualities, including
Carbon dioxide does not actually need any detection, it with a high water content without any problem. Other
can be vented and the result checked. However, if there components, e.g. valve & pumps, could be more sensi-
is a lot of inerts, the venting can take quite some time. tive. As for compressors: check with makers for a suit-
The author once vented an ammonia system for four able CO2 quality.
hours. Thus some detection is useful. 3.10. Fouling in carbon dioxide circuits.
A vent placed on the upper side of the condensate exit
Fouling usually ends up in the evaporators, especially
can be used to detect if too low capacity is due to flood- flooded evaporators, which then should be inspected regu-
ing or inert gas presence. When the valve is opened: larly and cleaned if necessary.
♦ If liquid droplets leave, flooding is likely. A source of fouling in the refrigerant circuit is oil and its
♦ If no liquid droplets leave, inerts are likely. decomposition products. As no oil is completely insoluble
It can be difficult to detect liquid droplets, though in carbon dioxide, it is flushed away be the liquid CO2 .
Fouling is usually a sign of excessive wear and tear or cor-
7. Question? Can CO2 be absorbed if vented into
rosion somewhere else and excessive fouling should thus
a bucket with ammonia water and thus detected?
entail an investigation to the causes.
Single section cascade condenser/evaporator Double section cascade condenser/evaporator

CO2 -10 °C
NH3 -15 °C
CO2
-40 °C
CO2 -40 °C

DX evaporators CO2 -10 °C

Liquid receiver Liquid receiver & LR & flooded


Flooded evaporators DX evaporators evaporators.
with pump circulation -10 °C

Fig. 20. Single or double carbon dioxide circuits.


A. Single circuit. B. Double circuit.
Advantages: Less costly, only one liquid receiver. Advantages: CO2 of a lesser quality but cheap can be
The same CO2 quality – but expensive - is used in the flooded circuit.
used in both the DX and the flooded, The flooded section is oil free, important for
pumped circuit. the performance of the evaporators.
Disadvan- Better quality – more expensive – CO2 is Disadvan- It is easy to mix up the CO2 qualities for the
tages: necessary. tages: DX and the flooded circuits.
Oil in the flooded circuit will foul the evapo- Expensive, two liquid receivers and two con-
rators and decrease the performance. denser sections are necessary.

3.11. Carbon dioxide filling. ♦ Be aware of dry ice formation if evaporation just above
Follow the gas manufacturer’s instruction. In general, the the triple point (-56.6 °C).
filling should start from the gas phase until a pressure well ♦ There is no problem with distribution of liquid carbon
over the triple point (5.2 bar) is reached otherwise dry ice dioxide to parallel connected evaporators as is the case
can form and block valves and pipes. When the pressure with pump circulation, see § 3.15. The high pressure
is reached continue the filling from the liquid phase. drop in the parallel legs compared to the header pipes,
ensures an almost perfect distribution.
3.12. One or two liquid receivers.
Figure 20 gives an overview of one or two liquid receivers. 3.14. The flooded evaporator.
The advantages and disadvantages can be summed up as: The condenser-liquid receiver has been treated in chapter
♦ The CO2 in the isothermal system is not very sensitive 3.6 and 3.7. Here we will give some points on the design
to the water content and more important, it is oil free, an and installation of flooded evaporators, either the cascade
important point for the operation of heat exchangers evaporator – in flooded flow mainly ammonia – or a proc-
ess cooler with CO2 as refrigerant, see figure 21.
♦ The compression cycle needs CO2 of a higher quality,
The main aspects are on PHEs, but most information are
which is more expensive. valid for other evaporator types as well.
♦ As a compression cycle usually contains oil, an oil
management system is necessary. Insoluble oil is also ♦ In flooded flow the refrigerant leaves the evaporator wet,
detrimental for the operation of heat exchangers as it sometimes as little as ten weight percent is vapour.
can cover the nucleation sites, which are responsible ♦ The driving force is a liquid column L, which has to
for a major part of the boiling heat transfer coefficients. overcome the pressure drops in the drop leg Pd,
♦ It is questionable to use two qualities of a refrigerant in evaporator Pe and return leg Pr. It can do this as the
the same plant. Sooner or later, the qualities will be
mixed up, with compressor break down as a result.
♦ The benefit of two separate circuits is thus nullified, e.g.
two emergency cooling circuits are necessary,
Separator
♦ As the circuits are separate, they can operate at differ-
ent temperature levels, here one is condensing at
-10 °C and the other at -15 °C, albeit the temperatures
should not be too different. Pr(eturn) leg
♦ Another – better - option is two separate cascade units.

3.13. Dry expansion evaporators. L


The installation and selection of expansion valves and Liquid head
other components do no differ from other high pressure
Pe(vap) Evaporator
refrigerants, there are sufficient components available. Pd(rop
The main points to consider for carbon dioxide are: Leg) H
♦ Flash gas before the expansion valve. Liquid carbon Pump head
dioxide is sensitive to overheating of the condensate,
less so for a moderate pressure decrease in a sub-
cooled condensate. Fig. 21. The separator-evaporator.
A B
D

E
F

Side placed (A) Top placed (B) or horizontal E. The pump is placed in a pit F. A downwards loop makes
with three bends. exit (C), both with two bends. to reduce cavitation danger. back flow more difficult.
Inclined (D) is questionable.
G
L
K
H

Avoid inlet from the top (G)- P with out An ejector inlet (J) is questionable as L. Short but wide separator. The design is
benefits - as well as letting the flash there will be maldistribution in the expensive and the separation efficiency is
vapour enter the liquid (H). evaporator. A simple tube (K) is useless. questionable, see also (O).

M N

M. Two symmetrical return legs. At N. Two symmetrical exits, joined before O. Avoid asymmetric return legs. A long and
least the return leg at the movable entrance to the separator. It is probably slender separator is cheaper than (L) and the
frame plate should have a flange. more expensive than (M). flow is more stream lined with better efficiency.

Figure 22. Separator placement.


density in the drop leg is much higher than the two- quid enters the evaporator again, too much vapour is
phase density in the evaporator and the return leg. produced and the process repeats.
♦ It is also possible to add a pump to give the natural cir- A good design rule is to maximize the return leg pres-
culation a boost. sure drop to 25 % of the total.
♦ Natural circulation is normally used if the evaporator ♦ Pumped flow systems are less critical. The pump can
and separator are in the immediate vicinity. Pumped be rated for a fairly large circulation. If the pressure
flow is used for far away evaporators. drops in the system turn out to be larger than expected,
♦ The optimal circulation rate - inverse of the exit vapour the circulation decreases but this is already taken into
fraction – ranges from less than 1.1 to 10 for CO2 and consideration and most evaporators are fairly flexible.
1.15 to 2 for ammonia. It varies with the thermal duty, Figures 22 and 23 shows some do’s and don’ts for sepa-
evaporator type and pipe length. As the evaporator is rators-evaporator systems.
usually the most critical component the manufacturer ♦ If design conditions permit, a horizontal exit as in (22C)
should be consulted to get a proper circulation rate. allows very large load variations. An inclined return leg
♦ The design of the return leg is especially critical in case as in (22D) should be avoided, especially for low pres-
of natural flow, also called thermosiphon. A too small sure drop or expected very low load as instable flow
pipe diameter, too many bends or too long pipes could could result. Top inlet (22G) means an extra lift and
lead to a far too small circulation with an impaired heat should be avoided. If the liquid level is used for control,
transfer. A too large pipe diameter could mean that the injection of the flash gas in the liquid body (22H) is un-
vapour cannot lift the liquid in the vertical sections. suitable as the level will be unstable.
♦ If the pressure drop in the return leg is too large there is ♦ Ejector designs (22J) have in general proved to be un-
a danger of oscillation; the vapour cannot leave the re- satisfactorily as the two-phase mixture tends to distrib-
turn leg as fast as it is produced and the liquid is ute unevenly over the plate pack. Especially unsuitable
pushed out backwards into the drop leg. is a simple pipe (22K) in the inlet; back flow is assured.
When no more liquid in the evaporator, the vapour flow ♦ A long and slender separator is cheaper and performs
decreases, the vapour leaves the return leg and the li- better than a short and wide (22 L & O).
♦ Figure 23 shows some design elements for a thermosi-
phon separator-evaporator loop.
1b
♦ Place all inlets at one end and all exits at the other. 7
♦ The separator can be divided into a separator (7) and a 1a
receiver part (1, 1c). The liquid filling is minimized and if
the liquid level is used to control an expansion valve, 1 1c
the operation will be more stable.
♦ Use a horizontal exit (1a, 1b & 1c) to the drop leg if
possible. Vortex formation is suppressed and in case of 4
CO2, dry ice is less likely to enter the pipes.
♦ A valve (2) in the drop leg can be used to suppress un- 6
stable evaporation.
♦ Never put a control valve in the return leg. If a stop valve
is necessary use a ball or gate valve, globe valves are
questionable as the pressure drop is too high. 6 5
♦ In case of insoluble oil, heavier than the refrigerant,
drain the oil at the lowest point (3). Oil separation is im-
proved if the pipe inclines slightly upwards. 2
♦ Especially for a low pressure vapour, the return leg (4)
should be as smooth as possible. The pipe exit should 3
be slightly pointing downwards, here by a 45 ° cut of a
pipe (5). Elbow bends (6) are very suitable. Fig. 23. Separator elements.
♦ Figure 23 shows a two pass design on the liquid side. It
is used for cooling of water/brine to a temperature close the UCs and the attached pipe work are equal, then the
to the evaporation temperature. The liquid inlet and ex- distribution could be as required. The danger is that if
its are at the top. In this way the evaporating refrigerant the capacity of lowest placed UC decreases suddenly,
meets roughly the same liquid temperature when enter- the pressure drop decreases as well. It could the steal
ing the channels at the bottom CO2 from especially the highest placed UC.
A plate heat exchanger is one of the few evaporator ♦ Better is to join the exits above the UC as in figure 24B.
types, which can cool water close to the freezing point If the pressure drops in the UC is lower than the corre-
without evaporator damage should freezing occur. sponding head H, the flow is better distributed over the
♦ Design of a two-phase circuit is difficult but an ex- UC and capacity changes are easier accommodated.
perienced evaporator manufacturer should be able If the capacity in one of the UC - # 3 - decreases to zero,
to assist in this. i.e. no evaporation at all, the pressure drop changes.
It increases if the friction pressure drop is small and the
3.15. Pumped flow evaporators. static two-phase pressure drop is large as this part is
The previous chapter dealt mainly with a separator in the replaced by liquid with higher density.
immediate vicinity of flooded flow evaporators – pumped It decreases if the friction part is large and the static
or natural circulation - typically plate or tube evaporators two-phase part is small. The friction part goes to zero
for cooling of brine. It is also usually a compressor system. and replacement of the two-phase mixture with liquid
The refrigerant, ammonia, carbon dioxide or other is has no importance if this pressure drop is small.
evaporated, compressed and condensed in a normal
compressor cycle. There will be no flow in the non evaporating UC and in
the pipe a static liquid column – LC - forms, which cor-
This chapter deals with pumped flow evaporators where responds to the pressure drops in the other UCs.
evaporating carbon dioxide is used instead of brine. The
general properties of carbon dioxide in this application was The other UCs get a little more refrigerant, which usu-
described in § 2.4. The evaporators – unit coolers, plate ally has no adverse effect.
freezers, tubes in an ice rink, freeze driers etc. are further If very large height difference between the UC, there
away and the carbon dioxide has to be pumped apprecia- might be a temperature penalty in shape of a boiling
ble distances. point increase. In such a case, the UC should be di-
It is usually – but must not be - an isothermal circuit, see vided into groups, each group fed by its own pump.
figure 20. The carbon dioxide evaporates and condenses ♦ One method to even out the refrigerant distribution is to
at basically the same pressure except for the small pres- arrange the UC with symmetrical exits as in figure 24C.
sure differences necessary for the circulation. ♦ Figure 24D shows circuit, where the UC are arranged
Figure 24 shows some basic layouts of the separator- with asymmetric exits, i.e. the refrigerant in the further-
pump-evaporator circuit. Distribution of a fluid between most UC has a longer way to travel and has to pass
parallel connected vessels of different types is a difficult various bends & connections. This creates pressure
problem; for a single phase fluid and still more so for a drops in both the feed and return leg. The driving head
two-phase fluid. Only some points can be treated here. in the furthermost UC is then less than in the first.
♦ As figure 24 implies, each group of heat exchangers There is neither any appreciable vertical distance as in
should be fed by its own pump, at least if the groups figure 24B to even out pressure drop variations.
are far apart, at different heights or of different types. A high pressure drop at each UC inlet is a common
method of distributing a fluid between parallel channels.
♦ In figure 24A a number of unit coolers are placed at dif-
This can be done by disks with tailored holes at the inlet
ferent heights. The exits from the UC join a common
or better, as the figure shows, with valves.
header below the UC. If the pressure drop for all
To cascade condenser. LC

H
From cascade
condenser.
#3

A. Unit coolers with exits at different heights. B. Unit coolers with exits at the same height.

C. Symmetrical exits. D. Unsymmetrical exits.


Figure 24. Arrangement of pumped flow systems.
Both methods are questionable, especially the disks, ♦ The return leg should be downwards inclining – 0.5 % -
which cannot easily changed as fine tuning of the pres- but this might be difficult to keep in practice. Note how-
sure drops might be necessary. ever, that an ice rink has more than 100 m of two-phase
The major drawback is that for the disks/valves to be ef- flow in perfectly horizontal pipes and with any problems.
fective, the pressure drops have to be fairly large in or- ♦ The optimal circulation rate can vary considerably, to
der to be effective. the point that the term “optimal” has no meaning.
A high pressure drop means a danger of flashing after The special design of PHEs – plates are removed or
the disk/valve, especially if the liquid carbon dioxide has added in parallel to increase the size – means that
been heated somewhat during the transport. when the circulation rate increases, the K-value in-
If the UC, or the freeze dryer, the plate freezer, the PHE crease and the number of plates can be reduced but
is not designed for a vapour fraction at the inlet a se- the pressure drop increases.
vere maldistribution could occur in the unit. Thus, natural circulation evaporators usually operate
Liquid systems use three way valves, which by-pass with a very low circulation rate, in the order of 1.1 to 2.0.
part of the liquid but that is questionable in two-phase Forced flow evaporators can operate for a little higher
system as a by-pass simply means extra circulation. circulation rate, 1.1 to around 3.
Note also that the pressure resistance of the valve has However, if the circulation is too high, the pressure drop
to be comparable to that of the evaporator, otherwise in all parts of the system becomes too large and the
there will be no or only an erratic control function. evaporation temperature increases, see § 2.8. This is
♦ A train of evaporators as in figure 24, could be com- especially serious for lower evaporation temperatures.
posed of various types of evaporators and with different This implies that the circulation rate should be lower,
capacities. The manufacturers are consulted and then the lower the evaporation temperature is.
submit the specifications. Most likely the pressure drops Unfortunately, there is a conflicting requirement, the liq-
will differ from the specifications. uid volume fraction should not be too low. For a given
Then, these are assembled to a circuit and the pump is circulation rate, it decreases with decreasing temperature.
started. However, the pressure drops between parallel This is an area, which has not been investigated very
legs in a well design circuit, e.g. figure 24C have to be well but as a reference we can take an ammonia ther-
equal but most likely they are not. The system solves mosiphon evaporator operating around 0 °C, which is
this by redistributing the liquid carbon dioxide until the usually laid out for an exit vapour fraction around 0.8,
pressure drops become equal. Thus one item can be circulation 1.25, but the actual circulation is probably
starved while others are overfed. This can be solved by: larger as the driving head usually is larger than the cal-
o Instead of the exact, optimal circulation of liquid car- culated pressure drop. Thus use a circulation 2.5.
bon dioxide, the circulation rate is increased. This is an The liquid volume fraction out is then 0.75 % and this
easy and fairly safe method of assuring that each item value is used to calculate the circulation rates for CO2.
should obtain at least its proper amount of refrigerant.
At -10 °C: Circulation: 1.12 (Ref. 1) 1.6
o Adjustment of the inlet valves as in figure 24D. It At -40 °C: “ 1.36 (Ref. 1) 2.4
probably needs a lot of time consuming tinkering with
the valves. (Ref. 1) recommends about the double circulation but
an overfeed is advantageous, see the previous section.
o Request from the manufacturer of each item, how
much refrigerant has to be fed to the item in order to ♦ In the end, the evaporator manufacturer should be con-
give the requested nominal pressure drop. sulted. Only he knows under which conditions an evapo-
rator gives its best performance.
1100 Density, kg/m3

1000

900 CO2
Flooded system
800 PAO
compressor
Temperature, °C
700
-30 -20 -10 0 10 20 E. Oil return
F. CO2 vs. oil densities.
DX compressor Expansion valve

F
B
Liquid receiver Separator C. Oil eva-
porator

D
A

Figure 25. Oil return in a flooded flow system.


A. An insoluble oil heavier than the refrigerant should be Heating medium for this usually the high pressure con-
drained at the lowest point of the loop. densate but any suitable heat source can be used.
B. An insoluble oil lighter than the refrigerant should be D. Another possibility could be to pump a small stream
drained from the surface. This implies a constant liquid back to the liquid receiver. If 1 % of the feed stream is
level. The separator is thus just a separator and not also pumped back, the oil concentration in the flooded loop will
a liquid receiver. This function is taken by the receiver to be 100 times the concentration in the liquid receiver.
the left. The liquid level is kept low to minimize the filling.
A hanging liquid receiver can also be used, figure 22 E. E. If the UC are used for the oil return, the oil collects in
the oil vessel at the DX evaporators. Oil then has to be
C. A soluble oil has to be separated by evaporating the pumped back to the flooded flow compressor or this will
liquid refrigerant. Here it is done by an oil evaporator. be starved on oil.

3.16. Oil return. Carbon dioxide has very little density difference to oil.
At low temperatures it is heavier than the oil but lighter
♦ Dry expansion. There is a straight flow from the com-
at higher temperatures, see figure 25 F. The oil density
pressor to the condenser, expansion valve, evaporator
varies with viscosity range and between manufacturers,
and back to the compressor and the oil follows this.
thus a check should be made for a specific oil.
The minimum vapour velocity for oil transportation in a
vertical pipe depends mainly on the vapour & oil density, The little density difference means that it might be diffi-
viscosities, oil surface tension and pipe diameter. A cult to separate it by gravity. More effective separators
consistent design method is difficult to find. A pressure can be used, e.g. a lamellae separator.
drop of 5 kPa/meter vertical pipe is a reasonable simple
formula. The pressure drop per length is proportional to Another possibility is to use an oil evaporator for insolu-
the shear forces at the wall, an important parameter. ble oils as well.
♦ Flooded flow in cascade system. If it is connected to Note that the oil draw-off point, whether soluble or in-
a cascade condenser/evaporator as in figure 20B, thus soluble oil, has to be in the refrigerant-oil loop. If it is
no compressor in the system, there should be no oil in drawn off from a stagnant point, e.g. close to D, there is
the system at all. a danger that there is little or no oil in the draw-off
♦ Flooded flow in parallel with a DX system, see figure stream and oil continues to build up in the circuit.
20A. The common liquid receiver feeds both pumped Figure 25 D shows how the DX system could be used
flow evaporators and expands to feed DX evaporators. for the oil return. There are some considerations:
The subsequent compressor will release oil into the
system, which will enter both the flooded and the DX o If the flooded system is large and the DX small, the
evaporators. The DX evaporators serve as oil evapora- strain on the DX evaporators could be too large.
tors, i.e. oil will never accumulate in the liquid receiver.
o The flooded flow compressor could be starved on oil
♦ Flooded flow in a compressor system, see figure 25. if no oil return to the compressor is provided from the
The liquid CO2 from the common liquid receivers ex- DX system.
pands in normal expansion valve to feed a number of
DX unit cooler and in a level controlled expansion valve o There is another possibility for oil return not shown in
to feed the flooded circuit. The oil return from the DX the figure. Liquid is drawn off from the separator at a
system is straightforward as described above. suitable point and fed directly to a DX evaporator. If
the evaporation pressure is lower than in the flooded
♦ In the flooded circuit, the oil remains, as only vapour loop, no pump is necessary.
leaves the circuit. It thus has to be removed as shown
in figure 25. The strain on this DX evaporator will be large though.
Tsat = -10 °C/Tsup = 49 °C

3b 4d 4c 4b 5a 78 °C

70 °C 5b

54 °C 5c
Ts = -10 °C/3 K subcooling

54 °C 78 °C 5d. Data for the defrosting circuit.


Compressors: 3 working, 84.4 kW each
Unit coolers: 3 * 4, 21.1 kW/each
UC. Type: AlfaCubic BL403C
3a
Electric defrost: 12.5 kW (If required)
Compressor when defrosting.
Power: 3.5 kW
Suction: -12.5 °C/24.7 bar
1 Discharge: -10 °C/26.5 bar
Defrost cap.: 12.5 kW
Superheater: 9.0 kW

2
Ts = -40 °C/0 K subcooling

Brazed PHEs are suitable for


all the positions 4b, 4d & 5c

4a

Figure 26. Defrosting of unit coolers.

3.15. Defrosting. The data are shown in 5d. With a power input of 3.5 kW,
The condenser in a LT typically operates well below 0 °C. the defrosting capacity is 12.5 kW, which is delivered to
Defrosting the LT unit cooler by condensing the hot gas the unit cooler 78 °C in and 54 °C out. With the excel-
from the compressor is thus not possible. For carbon diox- lent heat transfer properties of CO2, this could very well
ide particularly there are some possibilities, see figure 26: be sufficient to defrost a unit cooler in sufficient short
time.
1. Electric defrosting, suitable for lower capacity systems.
o The advantages of this system are:
2. Glycol from the HT system. A special defrosting circuit is
o The maximum design pressure is equal to the con-
necessary in the unit coolers
denser design pressure.
3. A special high pressure compressor, which increases
o No particular extra components, except the super-
the pressure to condensation pressure of 10 °C. (Ref. 2)
has reported good experience with this arrangement. heater are necessary (plus the standby compressor).
The disadvantage is that all components in the circuit o A computer simulation where condensing R507A at
must be designed for an operating pressure of 45 bar. 10 °C was compared with 26 bar CO2 cooled 78 to
4. Instead of increasing the pressure by compressing a 54 °C gave a slightly lower heat transfer coefficient for
gas, it can be done by increasing the pressure of the CO2 than for R507A, but the temperature difference
liquid refrigerant (4a), followed by evaporation (4b), for CO2 was almost three times as large, giving a very
separation of vapour and liquid (4c) and superheating large advantage for CO2.
(4d). (Ref. 1) has reported good experience with this ar- There are some restrictions:
rangement. As before, all components have to be de- • The heat source temperature has to be sufficient high
signed for 45 bar. in order to heat the refrigerant vapour, to 70 °C in the
In theory it might be possible defrost a unit cooler by using case studied. An ammonia compressor in the HT cir-
the hot gas only, i.e. with no condensation. Unfortunately, cuit can easily supply this temperature level, either
the defrosting time will be too long. The arrangement the oil or a special glycol circuit.
shown here could overcome this problem. • The discharge temperature can be lifted higher by in-
5. In the figure are shown three operating and one stand- creasing the compression ratio but this increases the
by compressor, all equal. The standby compressor is compressor power consumption as well.
used for defrosting. The unit cooler to be defrosted (5a) • The compressing ratio is outside the range specified
is connected to this compressor (5b) and both are dis- by the manufacturers but this probably is due to that
connected from the system. In the loop, just before the such low ratios are usually not required rather than
compressor, is a superheater (5c) installed as well. any technical difficulties, but this has to be checked.
The compressor has two functions: • The higher the compressor efficiency is, the lower the
♦ Act like a pump to circulate the vapour. discharge temperature will be and the more heat has
to be supplied by the superheater. This is an advan-
♦ Lift the temperature after the vapour superheater.
tage as this heat probably is cheaper than the elec-
tricity for the compressor motor.
Acknowledgements.
For the content of this paper, I have had the invaluable help
of colleagues within Alfa Laval, in both Italy and abroad, but
especially Göran Hammarson, to whom all I express my
thanks.
We also thank Friosol AG in Switzerland for the use of the
cover photography .

References.
1. Design Consideration when Using Carbon Dioxide
in Industrial Refrigeration Systems.
Angus Gillies, B. Eng., C. Eng. &
David Blackhurst BSc(Hons), C. Eng.
Star Refrigeration Ltd.
Glasgow UK.

2. Introducing a New Ammonia /CO2 Cascade Concept


for Large Fishing Vessels.
Per Skærbæk Nielsen and Thomas Lund.
York Refrigeration, Marine and Controls
Viby J, denmark
2003 Ammonia Refrigeration Conference & Exhibition.
IIAR, Albuquerque, New Mexico, March 16 – 19, 2003

3. Safety leaflet Carbon Dioxide, Linde/AGA.

You might also like