Printer Friendly

Investigation of Low GWP Flammable Refrigerant Leak from Rooftop Units.


New alternative refrigerants have gained popularity as a replacement to their conventional CFC, HCFC and HFC refrigerants. Many of these alternative refrigerants are flammable and listed under ISO 817 [1] and ASHRAE 34 [2] as A2L, A2 or A3. To determine the boundaries of flammability, two quantities are defined in terms of volume fraction of refrigerant in air: Lower Flammability Limit (LFL) and Upper Flammability Limit (UFL). If a mixture of air and refrigerant has a concentration between these limits, it can ignite under certain pressure and temperature conditions. Several studies have been conducted numerically and experimentally to investigate flammable refrigerant leaks from different types of refrigeration and air conditioning equipment. Kataoka et al. [3] studied catastrophic leak scenarios for a leak of 0.33 lb (150 g) of C[O.sub.2] (where C[O.sub.2] was used as a surrogate for flammable refrigerants). They also studied other flammable refrigerant leaks (R-290, R-32 and R-152a) for the sake of developing a new method for calculating the maximum allowable charge. A leak from a wall mounted, indoor room air conditioner (RAC) unit was studied numerically and experimentally for different refrigerants (R1234ze, R1234yf, R32 and R290) by Okamoto et al. [4]. They found that for indoor units, combustion does not occur if the unit is wall mounted in the case where there is no ignition source inside the unit. They also found that for a leak from a floor mounted indoor RAC unit, the risk of combustion is higher in R1234yf than R32. Li [5] experimentally investigated leaks from a split RAC unit mounted at 1.8 m high using R290. He found that the leak rate has a predominant effect on the refrigeration concentration compared to the leak height. The unit fan has a major effect on reducing the concentration inside the room which helps in mitigating the risk of ignition resulting from the leak. Due to their higher density than air, when a leak occurs, refrigerants descend toward the floor. Colbourne and Suen [6] investigated, experimentally, leaks from RAC equipment. They found that the factors that improve mixing and lower floor refrigerant concentration are: (1) smaller refrigerant charge, (2) high room air velocity, (3) directing the discharged air downwards, (4) higher release source location, (5) using a longer post-ventilation period. The authors extended the work into developing empirical correlations to determine the maximum allowable charge, release height, air discharge height and direction, mixing effectiveness, air discharge area and flow rate and desired post-ventilation time when a leak incident occurs [7]. Liu et al. [8] modeled an R290 leak from point source in the ceiling to find the dependency of the concentration distribution on leakage rate, ventilation and room size. They did not find an effect of the room size on the refrigerant concentration. Moreover, they found that ventilation can intensify mixing which in turn reduces the concentration distribution significantly. Jia et al. [9&10] experimentally studied R32 leaks with a total charge of 1.54 lb (0.7 kg) from a wall mounted AC inside a 9.5 x 12.8 x 9.0 ft (2.9 x 3.9 x 2.75 m) room. It was found that the R32 concentration increased to reach a peak of 16% close to the leak source then decreased with time. The leaked R32 descended near the ground, and then diffused horizontally. The dispersion behavior with the variation of leakage rate, ventilation rate and the height of the wall undercut was closely investigated by Laughman et al. [12]. They stated that the refrigerant dispersion inside an occupied space is complex due to the presence of several temporal and spatial scales. They found that the well-mixed assumption is not an accurate model for low-rate refrigerant leaks. Nagaosa [12] concluded in his numerical work that the gas flow velocity depends strongly on the gas density. Zhang et al. [13] experimentally studied R290 leaks with 0.84 lb (382 g) charge from a wall mounted unit at 7.2 ft (2.2 m) inside a 15.7 x 11.8 x 8.5 ft (4.8 x 3.6 x 2.6 m) room. Similar to other studies, they found that the flammable range of release of R290 is located close to the leak source. They also found that the most dangerous situation is when R290 is ignited and the refrigerant continues burning. This will result in the possibility of igniting the unit plastic case which will generate a lot of smoke which adds to the risk on room occupants' safety.

The main difference in leaks from an RTU is the presence of ducting systems which add to the complexity of the problem. In this paper, we are going to study the catastrophic leak scenario of R32 refrigerant from a 5 ton (17.6 kW) RTU which serves a 30 x 40 ft (9.1 x 12.2 m) conference room with high heat load, due to the potential for large internal heat gain, right after the unit stops operating.


The system consists of a 5 ton (17.6 kW) RTU and is connected to a ducting system which distributes the cooled air inside the room as seen in figure (1). The ducting system has four outlets connected to 2 x 2 ft (0.61 x 0.61 m) air diffusers in addition to two return grills mounted on the ceiling. The duct system dimensions are shown in figure 1. A numerical model was built to run the computational fluid dynamics (CFD) simulations using CONVERGE 2.3. The simulation was divided into two stages. First, the ducting system was simulated to predict the mixing behavior between air and refrigerant. The CFD simulation predicts the air/refrigerant flow rate entering the room at each of the four air diffusers. In the second stage, the room was simulated and the air/refrigerant mixture diffused into the room was adopted from the first stage of the simulation. The 5 ton (17.6 kW) RTU system has a total R32 charge of 15 lb (6.8 kg). Typical 5 ton RTU have lower charges, XbXut we wanted to simulate an extreme case. The leak scenario investigated involved the full rupture of a 1/4 inch (0.635 cm) pipe located after the evaporator coil with a refrigerant pressure of 150 psi (1.03 MPa) at the leak point. The refrigerant temperature was assumed to be 45[degrees]F (280 K) since the unit just stopped running. Orifice equation was used to calculate the refrigerant leak velocity assuming a discharge coefficient of 0.8. The leak velocity was calculated as 689 ft/s (210 m/s) and the mass flow rate of the leak was calculated as 0.47 lb/s (0.213 kg/s). The total charge was leaked in 31.5 seconds. The simulation was transient and ran for a total of 600 seconds. Room temperature was 80[degrees]F (300 K) and the initial temperature of the air inside the duct was 58[degrees]F (288 K). The refrigerant was assumed to be an ideal gas. Navier-Stokes and energy equations were solved and standard k-[epsilon] turbulence model was used to simulate the flow turbulence and gravity force was included in the momentum equation. Pressure Implicit with Splitting of Operators (PISO) scheme was used for pressure-velocity coupling. The time step was variable and ranged between [10.sup.-5] to 3x[10.sup.-3] seconds and the maximum iterations per time step was 98 with a residual of 1 x [10.sup.-4]. The simulation was run on a cluster using 88 cores.

Numerical Validation

We were not able to find in the literature a similar experimental work that we can use to validate the numerical work. Therefore, we were able only to validate the numerical approach with an experimental leak scenario from a mini split unit by Okamoto et al. [4]. Figure 2 reveals the comparisons of volume fraction of R32 between our modeling results and experimental data from Okamoto [4]. The details about the case can be found in the paper. The base grid size is 1 inch (2.54 cm) and the total number of cells of the model is about 1 Million cells. The time step for the simulation is 0.02 second. The inlet velocity is about 0.18 ft/s (0.055 m/s). The transient pressure-based solver with gravity (g = 32.2 ft/[s.sup.2] (9.81 m/[s.sup.2])) was employed. Energy equation was turned on. A standard k-[epsilon] model with enhanced wall treatment and buoyancy effects was utilized. The species model was used to simulate the species transport. In the mixture, the employed models included ideal gas for density, mixing law for specific heat, ideal gas mixing law for thermal conductivity and viscosity, and a constant mass diffusivity value of 3.1e-4 [ft.sup.2]/s (2.88e-5 [m.sup.2]/s). The inlet and outlet were set as mass flow inlet with a mass flow rate of 0.0092 lb/s (0.00417 kg/s) and 100% R32 leakage while the pressure outlet was 100% air with zero gauge pressure, respectively. PISO scheme was used as pressure-velocity coupling. Figure 2 shows that the simulations in the present work successfully capture the concentration as a function of time. Six monitor points were stacked vertically starting with point 1 at the floor level. Although, the comparisons show some over prediction of R32 concentration at lower sampling points (i.e. points 1 to 4) from present modeling, it does capture the concentration at points 5 and 6 well. All numerically predicted data are within the error bar of the experimental data. This validation shows that our model was able to simulate and capture the refrigerant leak dispersion inside the room with acceptable accuracy.

A grid dependency test was conducted as seen in Figure 3 which shows the % volume fraction of R32 at two different monitor points (Point 3: (X, Y, Z) = (10, 7, 5) ft [(3.05, 2.13, 1.52) m] and 9: (X, Y, Z) = (10, 23, 5) ft [(3.05, 7.01, 1.52) m]) (see Figure 5) for three grid base sizes with total number of cells of 2, 3 and 4 Million. The figure shows that the convergence is achieved at 3M size grid for both monitored points which has a 1.25 inch (3.175 cm) base grid. This grid was adopted to conduct the numerical simulations.


Duct Simulation

The CFD simulation prediction of the refrigerant flow through the duct outlets is shown in Figure 4. Outlet 1 which is the closest to the leak location has negligible R32 flow rate for the first 60 seconds of simulation. For the other three outlets, the refrigerant flow rate rapidly increases approximately after 5 seconds from the start of the leak and reaches a peak value of approximately 0.14 lb/s (0.064 kg/s) for outlet 2 and 0.11 lb/s (0.05 kg/s) for outlets 3 and 4. The flow rate at these three outlets slowly decreases afterwards until time = 31 seconds which is the total refrigerant leak time. A sharp decrease in flow rate in outlets 2, 3 and 4 then follows. After 35 seconds from the start of the leak, the four outlets have low refrigerant flow rate (approximately 0.005 lb/s (2.3 g/s)). The reason for this behavior is the high momentum of R-32 flow inside the duct which builds up rapidly in the first few seconds of the leak. This high momentum drives the flow to the end of the duct (outlets 2, 3 and 4) while gravity has relatively negligible effect to drive the flow towards outlet1. The rapid decrease in the refrigerant flow rate at outlets 2, 3 and 4 is expected around this time (i.e. t = 32 seconds) since there is no charge left inside the unit after 31.5 seconds. The relatively constant and low refrigerant flow rate from outlets 2, 3 and 4 after 32 seconds is due to remaining R32 travelling inside the duct.

Room Simulation

The room simulation was run for ten minutes. We assumed a symmetry plan along the Y-Z plane at the middle of the X-axis. Hence we only had to simulate 1/2 of the room as shown in Figure 5. Outlets 2 and 3 have the highest R32 mass flow rate compared to outlet 1 and 4, as such we modeled half the room where outlets 2 and 3 are located to present the worst-case scenario. Different monitor points in the domain were used to monitor the % R32 volume fraction to characterize the diffusion of the refrigerant during ten minutes of physical time. The coordinates of the nine monitor points are shown in Figure 5.

Figure 6 shows the % volume fraction of R32 at all monitoring points. The figure shows that the volume fraction is almost zero, at all points, for about the first 10 seconds then followed by a rapid increase in the volume fraction for points 1-3 and 7-9 which are located under outlet 2 and 3. The volume fraction for these points ranges approximately from 18% to 23% which is within the flammable limit of R32 which is 14.3%. R32 volume fraction slowly decreases with time untill 32 seconds which is approximately the time for the RTU charge to be completely leaked. R32 volume fraction rapidly decreases then to less than 5% and stays in this range for the rest of the simulation. For points 4, 5 and 6 which are located under the return grill (there is no return air during the leak incident), the % R32 volume fraction stays below 5% for the whole ten minutes of simulation. It can be seen from this figure that R32 volume fraction profile for the highest six monitor points is similar to that for R32 mass flow rate as seen in Figure 4.

The total space volume where R32 exceeds LFL is calculated and presented as a function of time in Figure 7. The volume profile follows the % volume fraction as seen in Figure 6 where it is almost zero for the first 10 seconds then followed by rapid increase where the total volume reaches a peak of 8.6 [ft.sup.3] (0.24 [m.sup.3]) at 17 seconds then the volume decreases slowly till time reaches 32 seconds. After that, the volume rapidly decreases and becomes zero around 35 seconds from the leak start. The figure shows that for 26 seconds, a definite volume inside the room contains R32 volume fraction that is higher than LFL.


A catastrophic leak incident from a 5 ton (17.6 kW) RTU with a 15 lb of R32 refrigerant charge was numerically simulated. The ducting system was simulated separately to predict the air/refrigerant flow rates from the four outlets. The conclusions can be summarized as follows:

- The ducting system plays a major role in the distribution of the leaked refrigerant from RTUs and it was noticed that there is about 5 seconds delay between the start of the catastrophic release and the actual release of the refrigerant into the space. Furthermore, it was noticed that the refrigerant almost bypassed the air register closer to the refrigerant source due to the high momentum.

- R32 volume concentration exceeded the LFL at six monitoring points located below outlets 2 and 3. R32 volume fraction ranges from 18% to 23% at these points.

- The monitoring points located under the return grill away from the outlets showed R32 volume fraction less than 5%.

- The total volume where R32 volume fraction is higher than LFL was calculated. The maximum volume where R32 concentration exceeded LFL was 0.18% of the entire space.


This work was sponsored by the U. S. Department of Energy's Building Technologies Office under Contract No. DE-AC05-00OR22725 with UT-Battelle, LLC. The authors would also like to acknowledge Mr. Antonio Bouza, Technology Manager--HVAC&R, Water Heating, and Appliance, U.S. Department of Energy Building Technologies Office.


[1] ISO, 2014, ISO 817:2014 Refrigerants - Designation and Safety Classification

[2] ASHRAE, 2013, ASHRAE 34-2013 - Designation and Safety Classification of Refrigerants

[3] Kataoka, O., Yoshizawa, M., and Hirakawa, T., 2000, "Allowable charge calculation method for flammable refrigerants," Proc. International Refrigeration and Air Conditioning Conference, Purdue University, West Lafayette, IN.

[4] Okamoto, H., Hattori, T., Dang, C., and Hihara, E., 2014, "Leakage of Mildly Flammable Refrigerants into a Room," Proc. International Refrigeration and Air Conditioning Conference, Purdue University, West Lafayette, IN.

[5] Li, T., 2014, "Indoor leakage test for safety of R-290 split type room air conditioner," International Journal of Refrigeration, 40, pp. 380-389.

[6] Colbourne, D., and Suen, K. O., 2003, "Equipment design and installation features to disperse refrigerant releases in rooms--part I: experiments and analysis," International Journal of Refrigeration, 26(6), pp. 667-673.

[7] Colbourne, D., and Suen, K. O., 2003, "Equipment design and installation features to disperse refrigerant releases in rooms--part II: determination of procedures," International Journal of Refrigeration, 26(6), pp. 674-680.

[8] Liu, Q., Zhang, H., Liu, Y., Huang, H., Zhang, X., Li, Z., and Yao, W., 2013, "Influencing Factors of Flammable Refrigerants Leaking in Building Air- conditioning System," Procedia Engineering, 62, pp. 648-654.

[9] Jia, L., Jin, W., and Zhang, Y., 2017, "Experimental study on R32 leakage and diffusion characteristic of wall-mounted air conditioners under different operating conditions," Applied Energy, 185, Part 2, pp. 2127-2133.

[10] Jia, L., Jin, W., and Zhang, Y., 2015, "Analysis of Indoor Environment Safety with R32 Leaking From a Running Air Conditioner," Procedia Engineering, 121, pp. 1605-1612.

[11] Laughman, C., Grover, P., and Nabi, S., 2015, "A study of refrigerant dispersion in occupied spaces under parametric variation," Proc. ASHRAE Annual Conference, Atlanta, GA.

[12] Nagaosa, R. S., 2014, "A new numerical formulation of gas leakage and spread into a residential space in terms of hazard analysis," Journal of Hazardous Materials, 271, pp. 266-274.

[13] Zhang, W., Yang, Z., Li, J., Ren, C.-x., Lv, D., Wang, J., Zhang, X., and Wu, W., 2013, "Research on the flammability hazards of an air conditioner using refrigerant R-290," International Journal of Refrigeration, 36(5), pp. 1483-1494.

Ahmed Elatar, PhD

Viral Patel, PhD


Omar Abdelaziz, PhD


K Dean Edwards, PhD

Van Baxter, ASHRAE Director-at-Large

Ahmad Abu-Heiba, MSc


Mingkan Zhang, PhD


Ahmed Elatar and Mingkan Zhang are postdoctoral research associates at Oak Ridge National Laboratory, Oak Ridge, TN. Ahmad Abu-Heiba and Viral Patel are research associates at Oak Ridge National Laboratory, Oak Ridge, TN. K Dean Edwards and Van Baxter are senior researchers at Oak Ridge National Laboratory, Oak Ridge, TN. Omar Abdelaziz is the Building Equipment Research Group Leader, at Oak Ridge National Laboratory, Oak Ridge, TN.

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (
COPYRIGHT 2018 American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc. (ASHRAE)
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:global warming potential
Author:Elatar, Ahmed; Patel, Viral; Abdelaziz, Omar; Edwards, K. Dean; Baxter, Van; Abu-Heiba, Ahmad; Zhang
Publication:ASHRAE Conference Papers
Article Type:Report
Date:Jan 1, 2018
Previous Article:Comparing Atmospheric Stability Versus HVACR Equipment Chemical Stability of New Low GWP Olefin Based Refrigerants.
Next Article:Performance of Low GWP Refrigerant R-516A in an Air-Cooled Chiller.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters