DOI: 10.18462/iir.gl2022.

0009

REFRIGERANT EGRESSION TO CONNECTED SPACES IN RELATION TO

HYDROCARBON CHARGE LIMITS
D. COLBOURNE*(a), A. L. VONSILD(b)

Re-phridge Ltd, c/o GIZ Proklima

(a)

Dag-Hammarskjöld-Weg 1-5, 65760 Eschborn, Germany

(b)
Vonsild Consulting
Kløvertoften 45, 7100 Vejle, Denmark
*Corresponding author: [email protected]

ABSTRACT
Safety standards specify allowable refrigerant charge limits for refrigeration systems based on the size of the
room that the system could leak into. However, in practice, rooms may be connected to other spaces by
means of doorways or other permanent openings. Such openings may potentially provide a larger effective
room area across which a leak could disperse, implying allowable charge limits could also be based on this
larger total area. There are formulae within current safety standards that relate “natural ventilation” opening
size to allowable refrigerant charge, but in many cases their lineage and underlying assumptions are unclear.
The present work addresses this topic of “refrigerant egression” to connected spaces primarily using
computational fluid dynamics, supported by measurements, to examine behaviour under a variety of
arrangements. Following analysis of the results, an alternative formula is proposed for egressed mass of
leaked refrigerant or conversely the minimum egression opening areas as required to permit an incremental
charge amount of hydrocarbon refrigerant.

Keywords: hydrocarbon refrigerant, R290, flammable, connected rooms, charge limit, refrigerant egression

1 INTRODUCTION
Allowable charge limits (ACLs), as prescribed by refrigeration, air conditioning and heat pump (RACHP) safety
standards, are often obstructive to wider use of hydrocarbon refrigerants (HCs). Usually, ACLs for occupied
spaces are based on the confined room area within which the refrigerant-containing parts are installed
(“Space1”) and the area of adjacent spaces (“Space2”) joined with permanent openings are usually neglected.
Conversely, such considerations are allowed for A2L refrigerants. Due to the hindrances imposed by ACLs for
HCs, any additional allowance can be beneficial to their wider application.

The general calculation for ACL (𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 ) [kg] takes the form (equation 1).
𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 = 𝐹𝐹 × 𝐿𝐿𝐿𝐿𝐿𝐿 × ℎ∗ 𝐴𝐴1 (1)
where 𝐹𝐹 is a non-dimensional “concentration factor” usually between 0.20 and 0.75 depending upon
flammability class and mitigation measures applied to the equipment, 𝐿𝐿𝐿𝐿𝐿𝐿 is lower flammability limit [kg m-
3
], ℎ∗ is some representative height [m] such as installation height of the refrigerant-containing parts, room
height, etc. and 𝐴𝐴1 is the floor area of Space1 [m2].

IEC 60335-2-40 (2018) contains formulas for so-called “natural ventilation” to other rooms and to the outside,
enabling larger ACLs – for A2Ls only – compared to a single (confined) space. Minimum opening area (𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 )
to Space2 is equation (2).
𝑚𝑚𝑡𝑡 −𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 𝐴𝐴 𝑀𝑀
𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 = 104×𝐿𝐿𝐿𝐿𝐿𝐿
�𝑔𝑔𝑚𝑚 𝑟𝑟 𝑀𝑀−29 (2)
𝐴𝐴𝐴𝐴𝐴𝐴

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
where; 𝑚𝑚𝑡𝑡 is the total (e.g., system) refrigerant charge [kg], 𝑀𝑀 is the molar mass of the refrigerant [kg kmol-
1
], 𝑔𝑔 is gravitational acceleration [9.81 m s-2], 29 is molar mass of air and 104 is a constant.

For “natural ventilation” to the outdoors, IEC 60335-2-40 (2018) sets 𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 according to equation (3), or
conversely it can be transposed for 𝑚𝑚𝑡𝑡 .
0.04
𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 = 0.14�𝑚𝑚𝑡𝑡 (3)
𝐿𝐿𝐿𝐿𝐿𝐿

where 0.14 is a constant derived from 𝑔𝑔 and the flow coefficient and 0.04 is a conversion factor from (generic)
hydrocarbon 𝐿𝐿𝐿𝐿𝐿𝐿 to 𝐿𝐿𝐿𝐿𝐿𝐿 of the applicable refrigerant.

EN 378 and ISO 5149 include another approach, also excluding A2 and A3 flammables (equation 4).
𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 = 0.0032 (4)
The general calculation for ACL takes the form of equation (5).
𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 = 𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄 × ℎ𝑅𝑅,1 𝐴𝐴1 (5)
where 𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄 is “Quantity Limit with Minimum Ventilation” and is effectively some 𝐹𝐹 × 𝐿𝐿𝐿𝐿𝐿𝐿, having been
assigned to various refrigerants and ℎ𝑅𝑅,1 is the room height [m] of Space1. A condition for using this concept
is that the lower sill of 𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 must be within 0.2 m of the floor. In the derivation of 𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄 there is also an
implicit assumption that an opening of at least 𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 is present at a high level to provide pressure
equalisation between Space1 and Space2.

The derivation of the equation (2) is detailed in IEC (2000) and is basically a transposition of the Bernoulli
equation, relying on the premise of a static pressure difference across the opening caused by density
(concentration) differences. The source of equation (3) is unknown and no source of the derivations of
equation (4) and (5) has been found (but can also be assumed to be a transposition of Bernoulli).

Lastly, IEC 60335-2-40 (2018) offers an option for situations using any A2L refrigerants, where a permanently
open doorway to another room enables the area of the entire connected space to be used for determining
𝑚𝑚𝑡𝑡 , i.e., in equation (1), 𝐴𝐴1 would be replaced with 𝐴𝐴1 +𝐴𝐴2 . The draft IEC 60335-2-40 (IEC, 2022) permits the
same for A2 and A3 refrigerants provided that the potential leak is above 1.6 m; otherwise only half of the
Space2 area may be used to calculate 𝑚𝑚𝑡𝑡 , i.e., 𝐴𝐴1 +0.5 × 𝐴𝐴2 . This rule was established through a sort of trial-
and-error approach, checking that resultant 𝐹𝐹 in Space1 did not exceed about 1.5 × 𝐹𝐹 as used for 𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 .

Despite the interest in this connected space concept, it is seldom addressed in the literature. One source
dedicated to examination of this topic (Hansen and Kujak, 2017), through CFD simulations showed that for
A2L refrigerants, a restriction the size of a doorway does not significantly impede distribution of refrigerant
across the total space. It was argued that for a “reasonably” sized opening, additional ACLs for HCs are
insignificant, or conversely, large opening areas are otherwise needed and thus impractical. In addition to
providing a thorough assessment of the problem, Laughman et al. (2016) applied CFD to examine the
dispersion behaviour of leaked (albeit non-flammable) refrigerant and its distribution to connected spaces.
In particular, the implications associated with over-simplified models and characterisation of leakage rates
were considered. Natural ventilation for the purposes of dispersing releases of flammable gases are
addressed in IEC 60079-10-1 (2021), Annex C.5. It addresses wind induced ventilation and buoyancy induced
ventilation (or “stack effect”) and a combination of both. However, the objective parameter is airflow rate,
rather than distribution of the released gas. Application of such methods were evaluated by Saunders and
Ivings (2005) primarily with CFD, yielding practical guidance for design purposes.

Whilst the term “natural ventilation” is often used, the mechanism addressed in the present work differs
from the conventional meaning. According to BS 5925 (1991) natural ventilation is the movement of air
through openings in the building fabric, due to wind or to static pressures created by differences in
temperature between the interior and exterior of the building, or to a combination of these acting together.
In the present context, the objective is not necessarily to quantify airflow, but rather the transfer of

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
refrigerant-air mixtures. Furthermore, the entire process could be taking place within two connected rooms
that are entirely sealed from the outside, thus being separate from any ventilation in the conventional sense.
Thus, the term adopted is “refrigerant egression”.

2 ASSESSMENT OF THE PROBLEM

From the available background information, a number of observations may be drawn, which this study seeks
to address:
 The origin, derivation (and arguably validity) of several formulae above is unclear.
 Consideration of equations (2), (3), (4) and (5) suggests some functional parameters may be neglected,
so further analysis of the mechanisms involved is needed.
 Since smaller openings and doorways may equally be used for refrigerant egression, involving the same
transfer phenomena, it seems appropriate that a single approach be used for both types of openings.
 Arguing that application to HCs may be impractical is a subjective judgement; any approach should be
applicable to all refrigerants.

First, the basic process is described. Refrigerant leaks within a porous RACHP unit housing in vapour phase
and begins descending to the floor of Space1. The plume entrains air from the surroundings and thus dilutes.
There is a permanent opening in a wall at floor level, linking to the connected Space2. If the opening area is
smaller than the cross-section of the room, there will be additional (to frictional losses) resistance to flow,
resulting in hold-up of some of the mixture in Space1.

ACL calculations employ a value of 𝐹𝐹 that ensures a significant flammable concentration does not persist
within Space1 (Colbourne et al, 2020). This premise must equally apply if ACLs are determined when
accounting for connected spaces. If the same 𝐹𝐹 is applied to the area of Space1 + Space2, then the hold-up
of mixture means that the resultant Space1 𝐹𝐹 will be larger than specified and correspondingly smaller in
Space2. 𝐹𝐹 is set broadly to allow for some deviation from simplified conditions; Hansen and Kujak (2018)
examined the influence of numerous “noise factors” on 𝐹𝐹. Practically, 𝐹𝐹 can be eroded or enhanced, such as
by congestion in the room, system overcharging, deviation from assumed (constant) leak rates, alternate
room configurations, thermal convention currents, movement of occupants, variations in temperature and
humidity, altitude, etc. – although the majority can be beneficial for improved mixing and dilution.

Assessing the circumstances for application of any generic approach, some common variables may be
considered:
 Type/size of opening: spaces connected with a permanent access route (e.g., doorway, hallway) or an
intentional opening (“hole”) smaller than a doorway.
 Size of spaces: a connected Space2 which is smaller or larger (including infinitely larger) than Space1.
 Lateral position of opening: centre of a wall or towards a side end of a wall.
 Vertical position of opening: at floor level or above floor level.
 Characteristics of release source: vertical and horizontal position, orientation (wall) of source, mass flow
rate, properties of the refrigerant being released, etc.

Situations where ACL is determined for connected spaced involving integral airflow (Colbourne and Suen,
2022a) are not presently addressed.

CFD was used to assess the problem in more detail, where this variety of circumstances can be conveniently
evaluated. A number of experiments 1 were carried out using connected space arrangements, against which
the CFD model was validated. The model was then used to simulate cases encompassing the variables listed
above. Outputs from CFD calculations were used to determine generalised rules for including connected
space area(s) in ACL calculations.

1
These were not designed specifically for the current purpose, but were deemed sufficiently suitable.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
 3 EXPERIMENTAL WORK
3.1 General description
Initially, two series of experiments were carried out to help characterise concentration development
following a simulated leak from an indoor unit (IDU) where small openings and doorways connected the
adjacent spaces. Specifically:
 Series A: a small room connected to a large one by means of a doorway and short hallway.
 Series B: two rooms of equal area, connected with small openings.

Conditions for the tests are listed in Table 1, which comprised the majority of variables identified above.
Instrumentation included catalytic and infra-red gas sensors (all within ±3 accuracy), electronic mass flow
controller (±1%) and electronic balance (±1 g).

Table 1: Summary of leak simulation concentration test conditions

Release Mass flow
Series Test no. Configuration Equipment type
mass [g] [g min-1]
A 66 AHU in closet low mass flow 300 35
A 67 AHU in closet medium mass flow 50 50
Ducted split AHU
A 68 AHU in closet medium mass flow 400 50
A 69 AHU in closet with 1 m high plenum 410 50
B 22 Closed room 440 100
B 24 0.2 m × 0.8 m floor level opening Split AC wall unit 440 100
B 26 0.2 m × 0.8 m floor and high-level openings 440 100

Room layout plan views are shown in Figure 1 and Figure 2, indicating the position of units, gas sensors (red
dots) and pathways for refrigerant flow (blue arrows). Since these experiments were not conducted
specifically for the present study, instrumentation and positioning of sensors were not as comprehensive as
would be preferred, although as a means to validate the CFD model they were deemed adequate.

#2 Unit #2
Unit Louvre door
#3 #3
#5 #7
3.1 m

#4 #9
Openings
3.0 m

#1 #1 #4
Sensor tower, #5 – #12:
#6 0.00 m, 0.15 m, 0.30 m,
1.1 m

18.0 m

0.45 m, 0.60 m, 0.75 m,

4.5 m
#10 #11 #12
0.90 m, 1.05 m
1.3 m
4.5 m
0.9 m #8
4.0 m

5.8 m

Figure 1: Room arrangement for test series A Figure 2: Room arrangement for test series B

3.2 Series A
R290 was released from a floor-mounted ducted split air handling unit (AHU) installed within a closet with
double louvre doors. The closet is within a small room, with a permanent doorway to a short hallway, which
opens into a large room. Figure 3 shows concentrations from selected sampling points for test no. 66.
Typically, concentrations closer to the AHU are higher and values diminish substantially for more distant
sampling locations. For instance, concentrations within the hallway are approximately half of the values
directly in front on the AHU and those within the large room (away from the hallway entrance) tend to be
about half again. This is in contrast to the case of an empty room, where concentrations at a similar distance
(e.g., <5 m) from the AHU will be much closer especially as the release proceeds. Evidently, these openings
between rooms impose resistance to outflow.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
3.3 Series B
A single room was connected to a second room of the same size, where the doorway was sealed with plastic
sheeting except for certain sized openings, at the floor only (test 24) and at the top of doorway (test 26), plus
a baseline case with an enclosed room (test 22). Concentration data for selected sampling points are included
in Figure 4. Since incremental vertical concentrations were measured, mass of refrigerant within the space
could be approximated by integration. By the end of the release, both test no. 24 and 26 led to 10% and 15%
lower floor concentrations and 25% and 30% less refrigerant present in the room, respectively. Whilst having
the additional opening at ceiling level in test no. 26 did not provide a significant benefit, which was likely due
to other openings in Space2 providing exfiltration.

Figure 3: Concentrations at selected sampling Figure 4: Concentrations at selected sampling

points for test no. 66 (markers for measurements, points for test no. 26 (markers for measurements,
lines for CFD) lines for CFD)

4 CFD CALCULATIONS
4.1 Set-up
A series of CFD calculations were used primarily to examine the effect of the variables identified in section 3
on the egression of refrigerant from Space1 to Space2. Initially, the model was calibrated against the
experimental results; examples of corresponding concentration plots are superimposed in Figure 3 and Figure
4. Other than the physical geometry, one crucial parameter for achieving a better match 2 is good
approximation of unit exiting concentration (Colbourne and Suen, 2022b). The CFD model used the
OpenFOAM toolbox within Simflow package. Solver was rhoreactingbuoyant, turbulence model was LES
WALE, mesh dimensions were 0.001 – 0.1 m as appropriate, temporal and convective discretisation methods
were implicit Euler and limited linear, respectively. Other parameters were adjusted to better match the
measurements. For the analysis, the source conditions involved a wall-type IDU with a low-momentum dilute
uniform mixture exiting uniformly along the base of unit, to mimic the observations in Colbourne and Suen
(2022b).

Different cases were assessed, as listed in Table 2 and the variety of room arrangements are shown in Figure
5; again intended to reflect the variables identified in section 3. Areas for Space1 and Space2 ranged from 4
m2 to 66 m2. Openings were from 0.01 m2 to triple doorways (3 × 1 m wide × 2 m high). Unit (release) heights

2
Note that matching CFD outputs to measurements exactly does not necessarily represent “true accuracy”. For
instance, six different types of sensors have been used and whilst calibration at 50% LFL is within ±1%, comparison of
measurements during a test with all sensors in the same location leads to wide variations (up to ±30% of the mean).
This is due to differences in response time, sampling methods, porosity of sampling aperture, precision of outputs, etc.
Thus, some deviation has to be tolerated.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
(ℎ𝑢𝑢 ) were 0.5 m, 1.5 m and 2.0 m and room area ratios (𝐴𝐴1 :𝐴𝐴2 .) from 2:1 to 1:7. All cases used R290, except
a few with R32 and R1234yf to account for different refrigerant densities. Where openings were small, there
was one (of the prescribed area) at floor level and a second at ceiling level, i.e., total opening area was double
the stated value. With doorway openings mostly single (1 m × 2 m high), but also double and triple widths
were used. Constant leak mass flow rates were employed, although several were repeated with linearly
increasing and decreasing mass flow rates (on average, matching constant values).

Cd1_ 1m×4m Cd1_z 1 m × 10 m

16 m2 C6_ 16 m2 16 m2 Cd8_ 16 m2
8 m2 8 m2
16 m2

Cd7_ Cd4_
Cd2_
4 m2 Ch1_-4,
Ch9_-4 16 m2
16 m2 16 m2 28 m2 4 m2 8 m2
9 m2

Cd3_
Ch1_, Ch4_, Ch5_, Ch1_-25, Cd3_x
8 m2 Ch6_, Ch9_ Ch9_-25
2 25 m2 8 m2 24 m2
24 m 2 2 2
9m 9m 9m

Cd3_y
8 m2 Cd5_
24 m2 Ch7_
Ch8_ 32 m2
18 m2 9 m2

Figure 5: Selection of different connected space geometries trialled. Grey box is source of release.
Underscore in case reference (“_”) if for unit height, H or L.

4.2 Output
For the current work the main parameters of interest were:
 Evolution of total refrigerant mass in Space1 and Space2;
 Transient mass flow rate of refrigerant through the opening;
 Average concentration development at incremental heights in Space1 and Space2;
 Average velocity (any direction) at incremental heights in Space1.

Figure 6: Mean floor concentration in Space1 and Figure 7: Mean floor concentration in Space1 and
Space2 with connecting holes Space2 with connecting doorways

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
 Table 2: Summary of cases for CFD simulation conditions and key outputs
Case 𝑚𝑚̇𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 [g 𝑚𝑚̇𝑜𝑜 [g ̅
𝐶𝐶𝑓𝑓,1,𝑚𝑚𝑚𝑚𝑚𝑚 ̅
𝐶𝐶𝑓𝑓2,𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
𝐴𝐴1 [m2] 𝐴𝐴2 [m2] ℎ𝑢𝑢 [m] ℎ𝑜𝑜 [m] 𝑤𝑤𝑜𝑜 [m] ∆𝑡𝑡𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 [s] 𝑚𝑚𝑡𝑡 [g]
reference min-1] min-1] (g m-3] (g m-3]
Ch6L 9 9 0.5 0.2 0.2 23 240 94 3.0 23.2 5.7
Ch6H 9 9 2.0 0.2 0.2 23 240 94 1.8 11.3 2.8
Ch1L-4 9 4 0.5 0.4 0.4 23 240 94 4.9 21.9 14.6
Ch1L 9 9 0.5 0.4 0.4 23 240 94 6.2 21.8 10.1
Ch1L-25 9 25 0.5 0.4 0.4 23 240 94 6.5 22.2 5.9
Ch1H-4 9 4 2.0 0.4 0.4 23 240 94 4.3 11.8 7.2
Ch1H 9 9 2.0 0.4 0.4 23 240 94 4.7 11.3 5.3
Ch1H-25 9 25 2.0 0.4 0.4 23 240 94 4.9 11.4 3.3
Ch4H 9 9 2.0 0.6 0.6 23 240 94 7.1 11.2 6.4
Ch4L 9 9 0.5 0.6 0.6 23 240 94 7.6 20.8 12.1
Ch5L 9 9 0.5 0.8 0.8 23 240 94 7.9 20.1 12.8
Ch5H 9 9 2.0 0.8 0.8 23 240 94 8.0 10.7 6.7
Ch7H 9 - 2.0 - - 23 240 94 - 11.4 -
Ch7L 9 - 0.5 - - 23 240 94 - 23.9 -
Ch8H 18 - 2.0 - - 23 240 94 - 7.8 -
Ch8L 18 - 0.5 - - 23 240 94 - 15.8 -
Cd9H-4 9 4 2.0 2.0 1.0 23 240 94 5.2 10.8 8.6
Cd9H 9 9 2.0 2.0 1.0 23 240 94 7.9 10.5 6.9
Cd9H-25 9 25 2.0 2.0 1.0 23 240 94 11 10.2 5.2
Cd9L-4 9 4 0.5 2.0 1.0 23 240 94 5.0 20.7 16.6
Cd9L 9 9 0.5 2.0 1.0 23 240 94 8.2 19.8 13.3
Cd9L-25 9 25 0.5 2.0 1.0 23 240 94 10 19.4 8.8
Cd1H 16 16 1.5 2.0 1.0 53 720 622 28 23.8 19.3
Cd1Hz 16 16 1.5 2.0 1.0 53 840 725 30 25.3 18.3
Cd2H 16 16 1.5 2.0 1.0 53 640 565 24 23.0 19.3
Cd2H dbl 16 16 1.5 2.0 2.0 53 640 565 25 22.5 19.8
Cd2H trp 16 16 1.5 2.0 3.0 53 640 565 26 22.1 19.6
Cd3H 8 24 1.5 2.0 1.0 53 720 635 38 24.4 19.7
Cd3Hx 8 24 1.5 2.0 1.0 53 640 565 37 25.8 20.7
Cd3Hy 8 24 1.5 2.0 1.0 53 640 565 36 25.7 19.3
Cd3H lhr 8 24 1.5 2.0 1.0 57 640 565 40 29.7 23.3
Cd3H lhr hec 8 24 1.5 2.0 1.0 54 640 565 38 29.8 23.2
Cd3H R32 4 28 1.5 2.0 1.0 149 3000 7450 112 143 134
Cd3H R12yf 8 28 1.5 2.0 1.0 305 3100 15760 228 276 264
Cd4H 16 - 1.5 2.0 1.0 53 320 285 - 21.6 -
Cd5H 32 - 1.5 2.0 1.0 53 640 565 - 21.8 -
Cd6H 8 - 1.5 2.0 1.0 53 160 145 - 22.1 -
Cd7H 16 16 1.5 2.0 1.0 53 640 565 43 25.8 19.5
Cd8H 16 16 1.5 2.0 1.0 53 640 565 36 24.8 19.6
Cd1HQT 8 24 1.5 2.0 1.0 85 240 340 35 24.8 12.4
Cd2HQT 16 16 1.5 2.0 1.0 85 240 340 29 23.4 14.4
Cd3HQT 16 16 1.5 2.0 1.0 85 240 340 47 27.0 16.9
Cd3HQT R32 16 16 1.5 2.0 1.0 1214 240 4860 872 198 162
Cd6HQT 8 - 1.5 2.0 1.0 38 240 150 - 20.1 -
Cd1L 16 16 0.5 2.0 1.0 53 240 210 17 30.2 11.2
Cd2L 16 16 0.5 2.0 1.0 53 220 195 16 28.2 13.1
Cd3L 8 24 0.5 2.0 1.0 53 220 195 29 35.9 16.4
Cd4L 16 - 0.5 2.0 1.0 53 90 80 - 20.3 -
Cd5L 32 - 0.5 2.0 1.0 53 180 160 - 18.7 -
Cd6L 8 - 0.5 2.0 1.0 53 45 40 - 23.2 -
Cd7L 4 28 0.5 2.0 1.0 53 180 160 38 36.8 18.4
Cd11L 8 64 0.5 2.0 1.0 53 400 355 33 37.4 11.7
Notes: “Cd”: case with doorway, “Ch”: case with ‘holes’ (one at floor level, one at ceiling), “H”: high level unit, “L”: low
level unit, “lhr”: release from left-third of unit length, “hec”: unit very high exit concentration, “dbl”: double-width door,
̅
“trp”: triple-width door, 𝐶𝐶𝑓𝑓2,𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 is the mean floor concentration in Space2 at cessation of the leak.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
Mean floor concentrations (𝐶𝐶𝑓𝑓̅ ) in spaces with smaller openings and doorways are shown in Figure 6 and
Figure 7, respectively. Graphs include data for single rooms, to be used as a comparison. Figure 8 and Figure
9 include average mass flow of refrigerant through the opening over the leak duration.

Some general observations are made:

 For units at 2.0 m, the maximum 𝐶𝐶𝑓𝑓̅ in Space1 (𝐶𝐶𝑓𝑓,1,𝑚𝑚𝑚𝑚𝑚𝑚
̅ ) reduce approximately linearly by about 10% per
m of opening area.
2

̅
 For units at 0.5 m, 𝐶𝐶𝑓𝑓,1,𝑚𝑚𝑚𝑚𝑚𝑚 reduces approximately linearly by about 30% per m2 of opening area.
 Once the opening reaches a certain size, any subsequent increase has diminishing benefit
 The size of Space2 seems to have minimal effect (also reported by Laughman et al., 2016), where halving
̅
or doubling the size of Space2 (relative to Space1) affects 𝐶𝐶𝑓𝑓,1,𝑚𝑚𝑚𝑚𝑚𝑚 only within ±2% (in favour of the larger
Space2).
 Once the release ceases, subsequent equalisation occurs, except with large openings (doorways) where
the mass already equalises almost by cessation of the release.

Figure 8: Mass flow rate through opening for cases Figure 9: Mass flow rate through opening for cases
with holes with doorways

One important observation from Figure 8 and Figure 9 is the effect of the wave development. As the
spreading plume hits a room wall, the pressure impact forms a return wave that rolls back and thereafter
cycles between reducing and increasing velocity pressure across the opening and eventually transfers into
the connected space too. It seems to happen regardless of the opening size and leak mass flow; rising, falling
or constant. This wave behaviour is also responsible for the negative mass flows, where the momentum of
the wave causes transfer of some mixture back from Space2 into Space1.

5 CALCULATION METHODOLOGY
Based on these observations, a methodology was developed in order to estimate the mass flow of refrigerant
through an opening. First, the mass permitted in a single enclosed room is 𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 and the mass of refrigerant
that flows into the Space2 is 𝑚𝑚2 . Therefore, the total permitted charge for the system is (equation 6).
𝑚𝑚𝑡𝑡 = 𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 + 𝑚𝑚2 (6)
𝑚𝑚2 must flow through the opening into the connected space and is considered over the leak duration (∆𝑡𝑡𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 )
only (equation 7).
𝑚𝑚2 = 𝑚𝑚̇𝑜𝑜 ∆𝑡𝑡𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 (7)
where 𝑚𝑚̇𝑜𝑜 is the average mass flow through the opening. Further to equation (1), the ACL is equation (8).
̅
𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 = 𝐹𝐹𝐶𝐶𝑓𝑓,𝑚𝑚𝑚𝑚𝑚𝑚 ℎ𝑢𝑢 𝐴𝐴1 (8)

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
where the concentration factor, 𝐹𝐹 [-], is specified within the applicable standard. For example, for A2 or A3
̅
refrigerants under quiescent conditions 𝐹𝐹 = 0.35 and 𝐹𝐹 = 0.50 for A2Ls and usually 𝐶𝐶𝑓𝑓,𝑚𝑚𝑚𝑚𝑚𝑚 = 𝐿𝐿𝐿𝐿𝐿𝐿. Observation
of the CFD outputs shows that under most circumstances, flow through the opening is solely a function of
the spreading plume velocity. Thus, volume flow of mixture through the opening is simply (equation 9).
𝑉𝑉𝑜𝑜̇ = 𝑐𝑐𝑑𝑑 𝑢𝑢�𝑠𝑠𝑠𝑠 (ℎ𝑜𝑜 𝑤𝑤𝑜𝑜 ) (9)
where ℎ𝑜𝑜 and 𝑤𝑤𝑜𝑜 are the height with width of the opening [m]. Mass flow of the refrigerant within the
mixture passing through the opening is therefore equation (10).
𝑚𝑚̇𝑜𝑜 = 𝐶𝐶(ℎ) × 𝑉𝑉𝑜𝑜̇ (10)
where 𝐶𝐶(ℎ) is the vertical concentration distribution of the refrigerant relative to the hight of the opening.

Influence of Space2 conditions on the outflow may be considered. If outflow was predominantly due to the
mixture static pressure difference, presence of mixture in Space2 may be deemed to inhibit the flow.
However, since velocity pressure of the plume dominates, the mixture in Space2 is unlikely to have much
influence. To confirm, further CFD cases were ran with a mechanically imposed plume of pure air (to mimic
leaking refrigerant) and nearly identical velocity profiles were observed flowing through the openings. Thus,
relative densities either side of an opening have negligible effect on egress mass flow rate. This is further
supported by Figure 8 (Ch1_ and Ch9_), showing for most of the leak period 𝑚𝑚̇𝑜𝑜 with different connected
spaces are almost identical. After leak cessation, differences arise primarily due to the wave effects,
evidenced by the departure occurring first with the smaller Space2 case, where the return wave approaches
the opening sooner than the cases with larger Space2.

From equation (9), the key parameter is velocity of the spreading plume. To do this a correlation for a
descending negatively buoyant plume was employed (Etheridge and Sandberg, 1996; Colbourne and Suen,
2004) and adjusted for the representative volume element of the Space1 floor (ℎ′ 𝐴𝐴1 ), where spreading
initially takes place (equation 11).
1/6 1 1 1/3
𝑢𝑢�𝑠𝑠𝑠𝑠 = 𝑎𝑎ℎ𝑢𝑢 �𝑚𝑚̇𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 � − � ℎ′ 𝐴𝐴1 � (11)
𝜌𝜌𝑎𝑎 𝜌𝜌𝑟𝑟
where 𝑎𝑎 is a constant, 𝑚𝑚̇𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 is assumed mass flow rate of the leak [kg s-1], 𝜌𝜌𝑎𝑎 and 𝜌𝜌𝑟𝑟 are the density [kg m-
3
] of the air and refrigerant vapour, respectively at ambient conditions and ℎ′ is some representative height
of the flow layer. The trend expressed by equation (11) is similarly reflected by the concentration front flow
measurements reported in Colbourne and Suen (2022a), where higher release position and mass flow lead
to greater front velocity.

Figure 10: Average velocity during the release Figure 11: Correlation between equation (11) and
period over the horizontal plane at various vertical average velocity from Figure 10 over different
increments representative heights

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
From selected CFD cases, Figure 10 shows spreading velocity averaged across the entire horizontal plane, at
0.03 m vertical increments, further averaged over the release period. Figure 11 shows the correlation
between CFD outputs and equation (11), based on four different representative heights. Using ℎ′ = 0.20 m
provides the best-fit (and also corresponds to minimum velocities in Figure 10), meaning 𝑎𝑎 = 1.42.
Subsequently, opening area in equation (9) must also encompass this representative flow height, thus 𝐴𝐴′𝑜𝑜 =
ℎ′ × 𝑤𝑤𝑜𝑜 .

Considering equation (10), 𝐶𝐶(ℎ), needs be characterised. For all cases, vertical concentration profiles below
0.3 m were almost linear, even with low ℎ𝑢𝑢 = 0.5 m. Whereas with R32 and “R12yf” 3 these vertical profiles
were almost uniform since release duration is considerably longer than that for R290, allowing longer
diffusive mixing to better homogenise. In general, 𝐶𝐶(ℎ) can be represented (equation 12) by an adjustment
̅ ).
to the average floor concentration in Space1 (𝐶𝐶𝑓𝑓,1
̅
𝐶𝐶(ℎ) = 𝜑𝜑 × 𝐶𝐶𝑓𝑓,1 (12)
where the best match between CFD results was with 𝜑𝜑 = 0.91.

Quantifying equation (10) with (9), (11) and (12), predicted mass flow through the opening can be compared
with CFD outputs (Figure 12), which encompasses all variables identified in section 3. In most cases, predicted
𝑚𝑚̇𝑜𝑜 matches the CFD results within a factor of two and by reducing the constant by 25% errs under-prediction
of calculated 𝑚𝑚̇𝑜𝑜 to below the CFD results.

Figure 12: Average mass flow over the release period, comparing CFD results and calculation method for
both holes and doorways

Combining equations (7), (9), (10) and (11) and reducing 𝑐𝑐𝑑𝑑 × 𝑎𝑎 × 𝜑𝜑 to a single constant, 𝑏𝑏 = 0.79 (or 𝑏𝑏 = 0.6
to underpredict), the mass of refrigerant flowing into Space2 by cessation of the release is (equation 13).
1/6 1 1 1/3
̅ 𝐴𝐴′𝑜𝑜 ℎ𝑢𝑢 �𝑚𝑚̇𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 �
𝑚𝑚2 = 𝑏𝑏𝐶𝐶𝑓𝑓,1 − � ℎ′ 𝐴𝐴1 � ∆𝑡𝑡𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 (13)
𝜌𝜌𝑎𝑎 𝜌𝜌𝑟𝑟
Since the conventional charge limit formula assumes the maximum floor concentration to be the 𝐿𝐿𝐿𝐿𝐿𝐿, 𝐶𝐶𝑓𝑓,1 ̅ is
set at 𝐿𝐿𝐿𝐿𝐿𝐿. Whilst measurements with HCs found experimental 𝐹𝐹 to be higher, usually 0.5 to 0.8 (Colbourne
and Suen, 2021), use of the IEC (2022) value 0.35 results in a higher 𝑚𝑚̇𝑜𝑜 and thus a greater 𝑚𝑚2 . In practice,
̅ , then 𝑚𝑚̇𝑜𝑜 will be lower so less refrigerant will travel to Space2. However,
when a leak results in a lower 𝐶𝐶𝑓𝑓,1
̅
using a resultant greater mass within Space1 for 𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 and transposing equation (8) for 𝐶𝐶𝑓𝑓,𝑚𝑚𝑚𝑚𝑚𝑚 shows that
̅
with the better mixing of the release, 𝐶𝐶𝑓𝑓,1,𝑚𝑚𝑚𝑚𝑚𝑚 will always be below 𝐿𝐿𝐿𝐿𝐿𝐿. In other words, better-than-
anticipated mixing in Space1 cannot inadvertently result in the floor exceeding 𝐿𝐿𝐿𝐿𝐿𝐿.

3
R1234yf (molar mass = 114 kg kmol-1) is absent from the fluid library and therefore R12 (121 kg kmol-1) was selected
but assigned the LFL of R1234yf and hereafter designated “R12yf”.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
Whilst the CFD results indicate the size of Space2 (and its concentration build-up) have little influence on 𝑚𝑚̇𝑜𝑜 ,
as a constraint, the mass permitted to travel to the connected space may be limited to equation (14).
𝑚𝑚2 ≤ 𝐹𝐹 × 𝐿𝐿𝐿𝐿𝐿𝐿 × ℎ𝑢𝑢 𝐴𝐴2 (14)
where 𝐴𝐴2 is the area of Space2. With ℎ𝑢𝑢 referring to the unit in Space1, it is evident that equation (14) will
result in a smaller 𝑚𝑚2 than could be tolerated since the effective 𝐹𝐹 will always be larger than in Space1.
Nevertheless, this current form is proposed for convenience.

One further aspect may be discussed, which is for when Space2 is the outside (considering outdoor
temperature and wind). Lower outdoor air temperature and corresponding higher air density will potentially
produce inflow through an opening and thus inhibit any egressing refrigerant-air mixture. Similarly, wind
blowing at the opening will create an opposing dynamic pressure against egressing mixture. Sample cases
were evaluated to examine these circumstances.

Taking case ‘Ch1H’, with a 60 g min-1 R290 release from a unit at 2.0 m, behaviour was observed with an
outdoor temperature 30 K lower (-3°C) than the indoor (27°C). The process was surprising: no refrigerant-air
mixture was able to flow out through the lower opening and in fact the cold air forms a layer on the floor of
Space1, resulting in an “inversion” where the mixture accumulates at an elevated region of the Space1 and
eventually egresses from the upper opening. For this particular case, less than 10% of the mass egressed
from Space1 by cessation of the release. The same arrangement (case ‘Ch1H’) was evaluated for the
(isothermal) effect of external wind blowing uniformly and perpendicularly at the outside wall. Ordinarily,
30% of the refrigerant mass would egress from Space1 by cessation of the release. With a 0.25 m s-1 wind,
only 19% of the refrigerant egresses and with a 1.0 m s-1 wind, only 4% of the mass. Whilst these results
represent a somewhat unrealistic scenario (i.e., a step change in external/Space2 conditions), it does raise
questions over the suitability of employing such an approach with outdoor spaces.

6 CONCLUDING REMARKS
The intention of this study was to develop a robust approach to estimate the additional refrigerant charge
that may be permitted in a system, when permanent openings to connected spaces allow egress of leaked
refrigerant and thus lowering floor concentrations in Space1. Whilst a variety of methods currently exist in
RACHP safety standards – albeit currently limited to A2L refrigerants – their origins and rationale are
relatively unclear. However, since the methods employed within IEC 60335-2-40, EN 378 and ISO 5149 are
likely based on static pressure difference, rather than spreading plume velocity, they are not
phenomenologically correct. On one hand, in the cases examined above, plume spreading velocity appears
to dominate over the effects of mixture static pressure, so the methods in the standards may lead to larger
holes than necessary and may be regarded as “safe”. On the other hand, it is shown that only the lowest 0.2
m of an opening is “useful”, so those that rely on a greater height for their opening area may be inadequate.
Commensurately, accompanying requirements within the standards stipulate that “the lower edge of the
lower opening shall be a height of 0.2 m or less from the floor”. Yet, if openings are not below 0.2 m, then it
is unlikely that sufficient refrigerant will flow into Space2. Additionally, when openings are to the outdoors,
the variety of external conditions can lead to vastly differing effects and can potentially negate the intended
benefits of such an opening.

A comparison of the approach developed in the current study with the existing (standard) methods is
provided in Table 3. This shows that equation (2) offers fairly consistent opening sizes, but overestimate when
units are positioned at higher levels. Whilst equation (3) yields a better match to the new method, it should
be noted that such openings to the outside should not be used, based on the observations above. Equation
(4) leads to excessively small holes.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway
 Table 3: Comparison of new proposed and existing methods applied to R290
𝐴𝐴𝑜𝑜,𝑚𝑚𝑚𝑚𝑚𝑚 [m2]
ℎ𝑢𝑢 [m] 𝐴𝐴1 [m2] 𝑚𝑚𝐴𝐴𝐴𝐴𝐴𝐴 [kg] 𝑚𝑚2 [kg] 𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡 [kg] 𝑤𝑤𝑜𝑜 [m]
Eq. (2) Eq. (3) Eq. (4) Eq. (13)
0.5 15 0.100 0.050 0.150 0.08 0.06 0.0032 0.13 0.65
1.0 15 0.200 0.100 0.299 0.12 0.08 0.0032 0.12 0.58
1.5 15 0.299 0.150 0.449 0.15 0.10 0.0032 0.11 0.54
2.0 15 0.399 0.399 0.798 0.34 0.13 0.0032 0.15 0.77
0.5 25 0.166 0.083 0.249 0.14 0.07 0.0032 0.11 0.55
1.0 25 0.333 0.166 0.499 0.20 0.10 0.0032 0.10 0.49
1.5 25 0.499 0.249 0.748 0.24 0.12 0.0032 0.09 0.46
2.0 25 0.665 0.665 1.330 0.56 0.17 0.0032 0.13 0.65

ACKNOWLEDGEMENTS
The authors would like to acknowledge project Proklima of the Deutsche Gesellschaft für Internationale
Zusammenarbeit (GIZ) GmbH for supporting this work.

REFERENCES
BS 5925: 1991 Code of practice for Ventilation principles and designing for natural ventilation.
Colbourne, D., Suen, K. O., 2004. Appraising the flammability hazards of hydrocarbon refrigerants using
quantitative risk assessment model. Part I: modelling approach. Int. J. Refrig., Vol 27, pp.774–783.
Colbourne, D., Suen, K. O., 2021. Hydrocarbon refrigerant charge limits for quiescent rooms. Int. J. Refrig.,
vol. 125, pp. 75–83.
Colbourne, D., Suen, K. O., 2022a. Airflow to disperse refrigerant leaks from hydrocarbon refrigeration
systems. Int. J. Refrig., https://doi.org/10.1016/j.ijrefrig.2022.02.008.
Colbourne, D., Suen, K. O., 2022b. Concentrations of hydrocarbon refrigerant exiting from RACHP
equipment enclosures. Proc. 15th Gustav Lorentzen Conf., Trondheim, Norway.
Colbourne, D., Suen, K.O., Li, T.-X., Vince, I., Vonsild, A., 2020. General framework for revising class A3
refrigerant charge limits –a discussion. Int. J. Refrig.
EN 378: 2016. Refrigerating systems and heat pumps — Safety and environmental requirements.
Hansen, W. B., Kujak, S., 2018. Review of methodology to set safety factors for A2L flammable refrigerant
being adopted by product safety standards. Proc. 1st IIR International Conference on the
Application of HFC Refrigerants, Birmingham, UK.
Hansen, W. G., Kujak, S., 2017. Defining Room Area for Connected Spaces for Flammable Refrigerants.
Paper LB-17-C023. Proc. ASHRAE, USA.
IEC 60079-10-1 2021. Explosive atmospheres. Classification of areas – Explosive gas atmospheres.
IEC 60335-2-40: 2018. Specification for safety of household and similar electrical appliances. Safety.
Particular requirements for electrical heat pumps air-conditioners, and dehumidifiers.
IEC, 2000. IEC 60335-2-40, Ed. 3, Am.3: Particular Requirements for electrical heat pumps, air conditioners
and dehumidifiers. Flammable refrigerants. Background information for CD. Formulas for
flammable refrigerants. Document 61D/87/CD.
IEC, 2022. Final Draft International Standard (FDIS) of IEC 60335-2-40, edition 7. Document 61D/491/FDIS.
ISO 5149: 2014. Refrigerating systems and heat pumps — Safety and environmental requirements.
Laughman, C. R., Nabi, S., Grover, P., 2016. A Numerical Study of Refrigerant Dispersion in Single and
Multiple Connected Spaces. TR2016-018. Proc. ASHRAE Winter Conference.
Saunders, C. J., Ivings, M. J., 2005. Natural ventilation of offshore modules. Health and Safety Laboratory,
Health and Safety Executive, Report 402, UK.

15th IIR-Gustav Lorentzen conference on Natural Refrigerants | June 13-15 | Trondheim, Norway