# Ventilation requirements for refrigerating machinery rooms.

INTRODUCTION

Section 8.11.5 of ASHRAE Standard 15 (ASHRAE 2010) states:

"The mechanical ventilation required to exhaust an accumulation of refrigerant due to leaks of a rupture of the system will be capable or removing air from the machinery room in not less than the following quantity:

Q = 100[square root of G]

(Q = 70[square root of G])

where Q = the airflow in cubic feet per minute (liters per second) and G = the mass of refrigerant in pounds (kilograms) in the largest system, any part of which is located in the machinery room."

The Standard 15 User's Manual (Fenton and Richards 2002) states that the table of ventilation requirements, on which the formula is thought to be based, was likely developed before 1930. In fact, Brown (2005) shows that this table appeared first in the Explosive and Hazardous Trades of the New York Municipal Code, 1927 (Section 220 of Article 18) and also gives some other previous history in developing ventilation requirements. This table did not appear in the ASRE handbook until 1939 when it was adopted by the then B9 committee with modifications. The modifications consisted of higher ventilation rates for systems with greater than 1000 lbs of refrigerant (Brown 2005).

The Standard 15 equation has several fundamental flaws: 1) the equation does not account for variations in the maximum acceptable refrigerant concentration (or recommended concentration limit -- RCL); 2) the equation does not account for differences in refrigerant properties such as boiling point, vapor pressure, and molecular weight; and 3) it does not account for different room sizes.

In an effort to improve the Standard 15 equation, this research program was carried out. First, the history of accidents related to machinery room refrigerant leaks was reviewed to help determine conservative, but reasonable scenarios for future accidents. Next, a differential equation was developed to estimate the ventilation rate needed to limit the concentration of the refrigerant to the maximum limits for the proposed accident scenario. Finally, a simplified equation was derived for easy application and implementation into the standard.

The objective was to establish a technical rationale for specifying ventilation requirements in refrigeration machinery rooms to maintain safety and to develop equations or other methods to calculate required ventilation rates. To meet this objective, ventilation strategies and rates required to minimize health risks resulting from a leak were based on a realistic accident scenario, reflect the fundamentals of dilution, account for the recommended concentration limits and physical properties of each refrigerant, and are simple enough to use in standards and building codes for determining refrigerating machinery room ventilation requirements.

BACKGROUND

An overview of several databases reveals release scenarios, including quantities, duration, location in the system, and the cause of the release. Several have suggested their opinions on realistic release rates to be used to calculate the ventilation requirement.

Reported Accidents

Searching for historical refrigerant leaks yielded direction to define realistic and appropriate accident scenarios upon which ventilation requirements can be based. All accidents in North America and Europe on record were considered with accidents related to machinery room refrigerant leaks, including the magnitude of the leak, the cause, and any related damage to people or property included. Four main databases were searched:

* MARS (Major Accident Reporting System) (MAHB 2001): database on "major accidents" reported under Seveso, OECD and UN-ECE Managed by the Major Accident Hazards Bureau (MAHB). The current MARS database contains over 700 accidents and near misses collected since 1982 from the Member States of the European Union. Seventeen accidents were considered from examined between May 1991 and December 2001.

* NRC (The National Response Center) (NRC 2010): NRC has made available yearly data files containing data related to incidents (oil, chemical, radiological, bio-logical, and etiological releases anywhere in the United States and its territories). Table 1 shows the number of reports made to the NRC (2010) concerning the refrigerants suggested above as likely being in machinery rooms. The reportable quantity (RQ) (US EPA 2001) is the threshold quantity where the accumulated loss of a refrigerant within a 24 hour period must be reported to the NRC by law. Ammonia is the only refrigerant with a low enough reportable quantity to be meaningful. Therefore, almost all accidents are Ammonia. Some of the NRC reported accidental releases are also included in the ARIP database.
```Table 1. Reported Fixed Source Releases (NRC) from Jan 1, 2004
through Dec 31, 2009

Refrigerant Number Incidents Reported Reportable Quantity (lb)

R-11 1 5000

R-12 2 5000

R-22 13 *(1)

R-123 0 *(1)

R-134a 2 not listed

R-245f 0 not listed

R-407a 0 not listed

R-410a 0 not listed

R-417a 0 not listed

R-422a 0 not listed

R-290 (propane) 0 10,000

R-601 0 not listed

R-601a 0 not listed

R-717 (Ammonia) 6010 100

R-744 146 not listed

* (1) chemical subject to reporting under section 313 and section 6607
of the Pollution Prevention Act.
```

* ARIP (The Accidental Release Information Program) (US EPA 2010): ARIP is a database compiled by the EPA that focuses on accidental releases at fixed facilities that resulted in off-site consequences or environmental damage. All non-routine releases of oil and chemicals are required to be reported (to the NRC, Coast Guard or EPA regional offices). The US Environmental Protection Agency (EPA) compiles these reports into the Emergency Response Notification System (ERNS) database. Significant accidents are selected from the ERNS database under the Accidental Release Information Program (ARIP) and a questionnaire is sent out to the facility involved, consisting of 23 questions about the facility, the circumstances and causes of the incident, and the accidental release prevention practices and technologies in place prior to, and added or changed as a result of, the event. The ARIP database was searched for accidents involving indoor refrigerant releases for the timeframe of 1990 to 1999. Since the EPA draws information from the NRC database, most of the accidents found in the ARIP database also appear in the NRC database. However, the ARIP database provides more detailed information on the cause of the release as well as the release quantity and release duration. Therefore, the information extracted from the ARIP database was supplemented with the NRC data, if possible, to provide as much information as possible for each incident. In case the NRC database conflicts with the ARIP database, it is assumed that the ARIP database is more accurate. 223 accidents were recorded that met the criteria.

* OSHA (The Occupational Safety and Health Administration) (OSHA 2010): The OSHA database focuses on accidents affecting workers and included accident investigation summaries. The Accident Investigation Summary database details accidents affecting workers and has limited information on the cause of a chemical release, release duration or release quantity. A search was performed and accidents involving indoor refrigerant releases between January 1990 and February 2010 were extracted, yielding 94 applicable accidents. Eighty of the reported releases were for Ammonia. Limited information is provided on the quantities released.

* NIOSH (National Institute for Occupational Safety and Health) (NIOSH 2010): The center for disease control (CDC) and NIOSH report deaths from some accidents (mostly workplace) through the Fatality Assessment and Control Evaluation (FACE) program. Only one fatality could be found relating to a refrigeration leak. In 1992 in Alaska, an assistant ice rink manager died of asphyxiation while trying to stop a CFC-22 leak inside of a compressor room.

All of the databases listed were analyzed to determine accident scenarios that are conservative, yet reasonably likely to occur in standard practice. The ARIP/NRC databases are the most comprehensive. The release quantity, release duration, as well as the release parameters (gas, liquid, etc.) are listed for most accidents reported. Unfortunately, mostly Ammonia accidents are recorded (see Table 2.1 for explanation). The most frequent release quantifies vary between 100 and 10,000 lbs. The released quantities were binned and counted. The emission rate probability distribution is shown in Figure 1.

[FIGURE 1 OMITTED]
```Table 2. Emission Rate Probability Distribution from ARIP/
NRC Database (Ammonia)

Probability of 1% 2% 5% 10%
Exceedance

Total Emission 2.99 kg/s 2.29 kg/s 1.13 kg/s 0.61 kg/s
Rate (metric)

Total Emission 395.9 302.4 149.7 80.4 lb/min

Rate (english) lb/min lb/min lb/min
```

From the databases, accidents seem more frequent before 1993, as shown in Figure 2. In 1992, OSHA issued the Process Safety Management (PSM) of Highly Hazardous Chemicals standard (OSHA 1992), which contains requirements for the management of hazards associated with processes using highly hazardous chemicals. For refrigeration systems the standard applies to systems that contain 10,000 pounds of ammonia. Any facility having the threshold quantity of ammonia must have a PSM program in place. All new facilities are required to develop a plan and have it in place prior to introducing ammonia above the threshold quantity.

[FIGURE 2 OMITTED]

Table 2 lists the total emission rate for specific probabilities of exceedance. These emission rates were obtained by linear interpolation. One percent of all accidents evaluated had a total emission rate greater than 2.99 kg/s (395.9 lb/min). Five percent of the accidents had a total emission rate greater than 1.13 kg/s (149.7 lb/min).

So far, no differentiation has been made depending on the release type (gas, liquid or both). Figure 3 shows the fraction of the different release types from the ARIP/NRC database. For 75% of the reported accidents, gaseous ammonia was released and only 13% of the accidents involved a pure liquid release. However, the boiling point of ammonia is 240K (-28[degrees]F). A liquid pool of ammonia would generate vapor due to heat transfer to the liquid from the floor in most machinery room settings. Therefore, even a purely liquid release would have a gaseous component. On the other hand, a liquid release could be mistaken for a gaseous release due to the fact that the sudden pressure drop at the release point causes a portion of the liquid jet to flash into vapor and the remaining liquid may evaporate almost immediately.
```Gas 75%

Liquid 13%

Gas & Liquid 12%

Figure 3 Release types from ARIP/NRC database.
Note: Table made from pie chart.
```

Figure 4 shows the percentages for different causes of the releases reported. Equipment failure is the major cause for accidents. Calculating the average emission rate for all release causes resulted in 0.27 kg/s (35.7 lb/min) for operator error, 0.27 kg/s (35.7 lb/min) for equipment failure and 0.16 kg/s (21.2 lb/min) for other. Therefore, no distinction can be made between the sizes of the release due to equipment failure or due to operator error.
```Operator Error 27%

Equipment Failure 68%

0ther 5%

Figure 4 Cause of release from ARIP/NRC database.
Note: Table made from pie chart.
```

The ARIP database also details the release location. The MARS and OSHA databases were also evaluated. Both the OSHA and the MARS databases have a large percentage of unknown locations. Even though no emission rate information is given, release locations are listed Figure 5 shows the fraction of occurrence of different release locations. The most common leak location is at a valve. The second most probable location is the piping.
```(a)
A 46%

B 23%

C 15%

D 4%

E 12%
(b)

A 23%

B 23%

C 18%

D 6%

E 6%

F 24%
(c)

A 27%

B 27%

C 6%

D 1%

E 7%

F 32%

Figure 5 Release location from (a) ARIP/NRC, (b) MARS, and (c) OSHA
databases. A: Valves; B: Piping, C: Compressor/ Process Vessel,
D: Pump, E: Other, and F: unknown.
Note: Table made from pie chart.
```

Although the total mass of the refrigerant is not as important as the emission rate, it plays a role in the analysis because it provides a bounding limit of the amount of refrigerant that can be released. This leads to an emission time, where increased ventilation may slow the increase in refrigerant concentration in the room, but until the leak stops or all the mass is released, the ventilation will not clear the room.

The upper limit is generally defined by the catastrophic failure of the largest vessel in the machinery room. The quantity of refrigerant stored in this vessel could all be released if a large rupture occurred at the bottom. The released refrigerant would consist of the initial flash gas followed by the remaining liquid. The liquid pool the machinery room floor would generate vapor due to heat transfer to the liquid from the floor. After a time, as the floor cools and approaches the equilibrium temperature at atmospheric pressure for the refrigerant, the vapor generation rate would decrease dependent on heat transfer from the air above the liquid surface. Another failure that could cause a large release would be a large rupture in a liquid supply line. If the flow is not stopped, a very large quantity of liquid could be expelled by the line. Several incidents have occurred in this manner resulting in the release of large quantities of ammonia.

In commercial buildings with refrigerating machinery rooms containing water chillers, the smaller release incidents are similar to those that occur with ammonia in industrial systems. These leaks are caused by valve packing failure, mechanical seal leakage, pipe and tube fitting failure, corrosion failure, or incorrect operation or maintenance procedures. Refrigerant storage vessels are not generally used in conjunction with water chillers and so the quantity of refrigerant in these systems (R-11, R-22, R-123, R-134a, etc.) is consider-ably less. Consequently, the largest release of refrigerant from water chillers is significantly smaller than the potential large release from an industrial facility.

Refrigerant Release Types

General. Brown (2005), Stoecker (1998), Richards (1986), and others suggest that a failure consisting of a rupture of a 1/2" high pressure, high temperature liquid line is the most probable design worst-case scenario. Depending on the refrigeration system's piping and valves, the leak rate can be computed by applying the fluid flow principles in conjunction with the thermodynamic properties of the refrigerant.

Seidl and Taylor (2005) examine a leak from a 0.25 in. (6.35 mm) hole. Using R-22 and temperatures and pressures of 40[degrees]F (4.4[degrees]C) and 83 psia (572 kPa) and 100[degrees]F (37.8[degrees]C) and 210 psia (1448 kPa), the leak rate was calculated to be 3.5 lb/min (0.026 kg/s) and 8.5 lb/min (0.064 kg/s), respectively. For a half in hole at the higher temperature and pressure, the leak rate was calculated to be 34 lb/min (0.257 kg/s). They recommend a leak rate of 15 lb/min (0.11 kg/s), arguing that similar non-toxic refrigerants will have similar leak rates and that a 0.25 in. (6.35 mm) hole is not an unreasonable estimate of a puncture or accidental drilling.

Refrigerant releases are generally jet flows that consist of only vapor, only liquid, or a mixture of vapor and liquid (i.e., two-phases). Ventilation systems are designed to dilute refrigerant vapor concentration to a selected value, so the anticipated refrigerant vapor leak rate selected by the designer determines the ventilation rate. The four variables that influence the vapor leak rate are: pressure, temperature, state of the refrigerant (vapor or liquid), and physical size of the hole or orifice. The variables that usually cause higher leak rates are high pressure, high temperature, large orifices, and/or a liquid release. Low pressure, low temperature, small orifices, and/or a vapor release usually are characteristic of lower leak rates. This is not completely the case in all situations, but it does correctly show general trends. For example, a leak of high pressure, high temperature liquid through a large orifice would be large. In contrast, a leak of low pressure, low temperature vapor through a small orifice would be relatively small. Liquids under high pressure provide a greater leak rate due to high density and the greater velocity of flow through the orifice. High temperature liquids also augment the leak rate due to the higher fraction of flash vapor generated. From another perspective, refrigerant liquid leaks are greater and thus more difficult to handle than vapor leaks through the same orifice, same pressure, and same temperature. Therefore, high pressure, high temperature, liquid leaks produce the largest vapor quantity (volume) and thus pose the greatest safety risk in machinery rooms.

Liquid leaks are possible with the following combination of temperatures and pressures: high pressure and high temperature, high pressure and low temperature, and low pressure and low temperature. With high pressure and high temperature, the flash vapor comprises approximately 20% - 40% of the total leaked mass flow rate depending on the actual pressure and temperature. Flashing of the vapor cools the remaining liquid to its thermodynamic equilibrium temperature at atmospheric pressure. The refrigerant supplying the leak also drops in temperature according to its pressure. When the upstream pressure drops to atmospheric pressure, the leak rate through the orifice is nearly zero. However, the refrigerant may still exist in the source vessel or pipe depending on the location of the leak. The liquid that has leaked out will absorb heat from surfaces that it comes in contact with and consequently vaporize while cooling that surface. If sufficient liquid is leaked, it may pool on the floor of the machinery room and after producing large vapor quantities cooling the floor, produce only a small quantity of vapor. At that point, the heat vaporizing the refrigerant liquid only comes from small amount of thermal energy still left in the floor and the surrounding air.

If the liquid is at a low temperature, then the flash vapor quantity is considerably less (on the order of 10% by mass) and so the vapor produced is less. But if the pressure is high, the leaked liquid is rapidly replenished and the leaving liquid produces vapor cooling the surfaces it comes in contact with in the machinery room. Low pressure liquid will also produce flash vapor at a similarly low fraction but will not be vigorously replenished as with a high pressure liquid.

The quantity of vapor that will leak from an orifice depends only on the orifice size and the upstream pressure assuming the flow is not choked. The upstream temperature only influences the leak rate by how it affects the density of the upstream vapor. However, if the upstream pressure is greater than the critical pressure (and depending on the hole size), then the flow is choked and is not dependent on the upstream pressure. Under chocked flow conditions, upstream temperature and pressure changes will influence the flow rate in the same manner that it influences the upstream vapor density.

Liquid Jet Releases. When the hole causing the release is at or below the liquid vapor interface, then liquid will exit. The pressure inside the vessel or pipe resulting from the thermodynamic equilibrium condition pushes the liquid through the hole. With refrigerants, the sudden pressure drop causes a portion of the liquid jet to flash into a vapor. Consequently, this is a two-phase flow. Because flashing of liquid to vapor will generally occur to some extent with refrigerants, purely liquid releases are not considered except to predict the liquid flow approaching the hole or rupture.

The flow of liquid from a vessel, such as a reservoir tank, through a hole depends on the pressure difference between the inside and outside of the vessel and on the gravity head devel-oped by the liquid above the hole (AIChE 1996):

[E.sub.l] = [c.sub.o] [A.sub.h] [[rho].sub.l] [[2((p-[p.sub.o])/[[rho].sub.l]) + 2g[H.sub.l]].sup.1/2] (1)

The shape of the vessel also influences the flow rate of a liquid through a hole. For a vertically oriented vessel, the liquid height may be computed using (AIChE 1996).

[H.sub.l] = 4[V.sub.l] / [pi][d.sup.2] (2)

When the vessel is horizontal, then (AIChE 1996),

[H.sub.l] = d/2 (1 - cos[[theta].sub.l]) (3)

[V.sub.l] = L[d.sup.2]/4 ([theta] - sin(2[[theta].sub.l]) / 2) (4)

Reservoir tanks are commonly located outside refrigerant machinery rooms. Therefore, the liquid jet release out of a pipe is a more appropriate scenario. The liquid flow out of a pipe depends on the difference between the liquid pressure and the ambient pressure, as well as the head loss due to fittings (such as elbows, tees and valves) and wall friction (ASHRAE 2009). Thus,

[E.sub.l] = [A.sub.h][[rho].sub.l][[2((p-[p.sub.o])/([[rho].sub.l]([f.sub.f]L/D + [[SIGMA].sup.K])))].sup.1/2] (5)

The friction loss depends on the Reynolds number Re of the flow and on the roughness height, H, of the pipe wall surface (ASHRAE 2009):

[f.sub.f] = 8[[[[(8/Re)].sup.12] + 1/[(A+B).sup.1.5]].sup.1/12] (6)

[A.sub.h] = [[2.457.ln(1/([(7/Re).sup.0.9] + (0.27[epsilon]/D)))].sup.16] (7)

B = [(37530/Re).sup.16] (8)

Re = 4[E.sub.l] / [[rho].sub.l][pi]Dv (9)

Because the friction loss f depends on the mass emission rate, [Q.sub.l], the mass emission rate has to be calculated iteratively. For high Reynolds numbers, i.e. fully turbulent flow, the friction factor becomes independent of the Reynolds number.

Two-Phase Jet Releases. The release of liquid refrigerant involving a significant pressure drop will result in a portion of the liquid flashing to vapor. If it is assumed that the expansion of the liquid is adiabatic, that air does not mix with the expanding fluid, and that the expansion process is reversible, then the amount of flashing liquid can be calculated by an isentropic energy balance (API 1996):

[E.sub.f] / [E.sub.l] = [S.sub.l1] -[S.sub.l3] / [DELTA][S.sub.vap] (10)

If the kinetic energy of the fluid expansion is neglected, the isenthalpic balance can be used to calculate the amount of flashing liquid. The heat of vaporization is obtained entirely from the enthalpy of the liquid being released. It can be shown that the fraction of liquid flashed is (API 1996):

[E.sub.f] / [E.sub.l] = [c.sub.p](T - [T.sub.b]) / [H.sub.vap] (11)

The release of liquid refrigerant from high pressures involves a substantial change in kinetic energy, therefore the isenthalpic balance approach might be inappropriate for the calculation of the amount of flashing liquid. Neglecting the kinetic energy of the fluid expansion causes an over prediction of flashing liquid. According to the Manual for Modeling Hypothetical Accidental Releases to the Atmosphere (API, 1996) Equation 11 should be used in cases where a saturated liquid is released. For refrigerant blends, the flashing process is much more complex and calculations must include information concerning the vapor-liquid equilibrium conditions of the blend.

Releases of two-phase flow from vessels and pipes involving choked flow remain an area of research. While empirical relationships have been proposed, they only apply to simple situations: refrigerants composed of single components. One approach that has been used for single component two-phase flows is the homogeneous equilibrium flow model (HEM) which is based on the assumptions:

* Homogeneous liquid-vapor mixture

* Thermal equilibrium between liquid and vapor phases

* No slip between liquid and vapor phases

* Isentropic expansion process

The HEM has been applied in determining the choked flow rate of flashing liquid releases, but refrigerant releases have not been found in the literature. Several researchers have used this approach to predict laboratory results, but only (Sallet 1990) has predicted flow rates of several refrigerants including: R-290 (propane), ammonia, carbon dioxide, and R-22.

The monograph (AIChE 1996) reports another approach for predicting two-phase choked flows where analytical expressions are employed. The thermodynamic conditions must be appropriate allowing for an all liquid flow equation similar to the Bernoulli equation.

Evaporating Pool. For a two-phase release, liquid can form a pool on the floor while at the same time a super cooled vapor is released. The emission rate for the two mechanisms will be additive. One equation that is considered for the liquid spill is provided below (US EPA 1992):

Q = 6.94 x [10.sup.-7](1 + 0.0043[[[T.sub.2]-273.15].sup.2]) x [U.sup.0.75 x [A.sub.s] x [M.sub.W] x [V.sub.p] / [V.sub.Ph] (12)

[V.sub.ph] = exp(76.858 - 7245.2/[T.sub.2] - 8.22ln([T.sub.2]) + 0.0061557 [T.sub.2] (13)

The vapor pressure is calculated using the Clausius Clapeyron equation:

[V.sub.p] = 10.1325 * exp ([H.sub.vap][M.sub.w]/R (1/[T.sub.b] - 1/[T.sub.2])) (14)

The surface area and surface velocity of the liquid pool can all be adjusted to obtain specific evaporation rates for alternative emission scenarios and room sizes.

The evaporation of a liquid pool is a time varying process. Initially, the pool is very small; therefore the evaporation rate is insignificant. However, as time progresses, the liquid pool becomes larger, increasing the evaporation rate. Eventually, the pool is large enough that equilibrium is reached and the evaporation rate approaches the liquid release rate (total mass emission rate minus emission rate of flashing liquid). The limiting size of the pool is the room floor area or area of any dike that is used.

Vapor Jet Releases. The release of vapor jets can be approximated well assuming an isentropic expansion (reversible and adiabatic). The estimates provided yields reasonable temperatures of the initial vapor cloud and velocity of the vapor entering the cloud. For vapor releases that result in a large change in velocity, the isentropic approximation has been shown to be preferred over the isenthalpic approximation (Moran and Shapiro, 2000).

A vapor jet forms upon the development of a small hole generating a path for the refrigerant to flow through and enter the surrounding air in the machinery room. When the pressure in the vessel or pipe containing the refrigerant vapor exceeds the critical pressure, the flow velocity is sonic (equals the speed of sound) through the hole and the flow is defined as "choked." Regardless of how much greater the refrigerant pressure is above the critical pressure the flow rate is constant. Only the density inside the vessel or pipe (through temperature and pressure) influences the flow rate. Assuming the refrigerant vapor behaves as an ideal gas, the critical pressure ratio is

p / [p.sub.o] [greater than or equal to] [((k+1)/2).sup.k/(k-1)] (15)

Values for k generally range from 1.2 to about 1.5 and the pressure ratio defined by Equation 16 is approximately 2 for most gases and vapors.

The choked flow condition for an ideal gas or vapor based on isentropic flow is given by

E = [c.sub.o][A.sub.h][[[[rho]k(2/(k+1)).sup.(k+1)/(k-1)]].sup.1/2] (16)

Values for the discharge coefficient vary from 0.6 to almost one and are due to non-ideal flow affects. Observe that if the pressure is above the critical pressure in a vessel, the temperature will decrease somewhat causing the density to decrease in turn decreasing the flow rate.

When pressures are less than the critical pressure ratio, the condition of choked flow no longer holds and the mass flow rate is dependent on the pressure inside the vessel of pipe. The flow rate is given by

E = [c.sub.o][A.sub.h][[2p[rho](k/(k-1))([([p.sub.o]/p).sup.2/k]-[([p.sub.o]/p).sup.(k+1)/k])].sup.1/2] (17)

Note that the mass flow rate is less than that for choked flow.

For a given release, the rate of change of pressure inside the vessel or pipe depends on the specific situation governing the release. Either an adiabatic or an isothermal flow process may be appropriate. An example of an isothermal flow would result when a small leak occurs in a pipeline because the cooling that occurs by the expansion is counteracted by the friction heating and heat transfer through the pipe from the outside. Thus, the temperature in the pipe would remain constant; however, with relatively large vapor release flow rates, the expansion process cools the internal volume of the pipe or vessel and the adiabatic process is the appropriate choice. Moreover, if the pipe or vessel is insulated, then the adiabatic process is again the appropriate choice. In those situations when the isothermal process holds, the inside temperature is constant. With the adiabatic flow process, the pressure decreases as the refrigerant is released. The pressure may be estimated by using conveniently small time increments and calculating the mass released over the each time step. For each subsequent time step, subtract the mass that escaped from the mass inside the vessel or pipe recalculating the pressure. In this manner, even the case of isothermal flow can be accommodated as well as heat transfer to or from the vessel or pipe applying reasonable values for the convection heat transfer coefficient.

Further elaboration on calculation methods for vapor releases are presented in the monograph (AIChE 1996) where procedures first developed by Wilson (1981) for releases from pipe ruptures are discussed. The concept developed involves defining the initial vapor mass in the pipe and treating the release as an isothermal vapor release characterized by an expression consisting of the sum of two exponential terms. The exponential terms are actually time constants that depend on the physical size of the pipe and the size of the rupture.

RELEASE SCENARIOS

The purpose of this section is to present the method used to develop a realistic vapor release rate was for a machinery room based on the type of refrigerant and upon historical information on accidental releases.

There are five factors that affect the vapor release rates as previously discussed: size of the release orifice, state of refrigerant (gas or liquid); properties of the refrigerant; pressure and temperature. Different combinations of these factors will produce different vapor release rates. Below are the combinations ordered from high to low vapor release potential as indicated by Brown (2005):

* Release Scenario 1: Liquid, high pressure, high temperature

* Release Scenario 2: Gas, high pressure, high temperature

* Release Scenario 3: Liquid, high pressure, low temperature

* Release Scenario 4: Liquid, low pressure, low temperature

* Release Scenario 5: Gas, low pressure, low temperature

Of most interest is the worst-case realistic scenario and provide background information for use in selecting a realistic release rate to be used for ventilation room design. Figure 6 shows a schematic of a typical refrigeration cycle identifying where these release scenarios could occur in the process.

[FIGURE 6 OMITTED]

Release Scenario 1: Liquid, High Pressure, High Temperature

A liquid, high pressure, high temperature release is a two-phase flow. The pressure inside the vessel or pipe resulting from the thermodynamic equilibrium condition pushes the liquid through the hole. The sudden pressure drop at the release point causes a portion of the liquid jet to flash into a vapor. The remainder of the liquid pools on the machinery room floor at the boiling point temperature (Brown, 2005) after which it slowly warms to ambient temperature. The vapor emission rate is therefore calculated in 3 steps: 1) the total liquid mass emission rate out of the hole is determined; 2) the fraction of flashing liquid is computed; and 3) the evaporation rate from the liquid pool surface is calculated.

Several assumptions have to be made concerning the release size, release area, release temperature and air velocity over refrigerant pool. The liquid refrigerant pool size was taken to be the equilibrium pool size (evaporation rate was limited to be not more than liquid release rate minus emission rate of flashing liquid), with a liquid depth of 0.394 in. (10 mm) (US EPA 1992). During the release out of the pipe, the liquid will expand and cool to the boiling point. For refrigerants with boiling points below ambient conditions, the liquid pool will gradually become warmer but will most likely never reach ambient temperature before it is entirely evaporated. The liquid pool will also cool the surrounding air. Therefore, the pool temperature, T2, in Equation 13 was taken to be the greater of 32[degrees]F (0[degrees]C) or the boiling point temperature of the specific refrigerant. This is a conservative assumption for those chemicals that have boiling points lower than 32[degrees]F (0[degrees]C) in that the liquid will remain cooler than 32[degrees]F (0[degrees]C) for some period of time. For example calculation purposes, the air velocity over the spill is taken to be 50 ft/min (0.254 m/s). This would correspond to a ventilation rate of 20,000 cfm, which is recommended by Brown (2005) for all machinery rooms, and a room cross-sectional area of 400 [ft.sup.2] (37.2 [m.sup.2]). A lower ventilation rate and a smaller room size would give a similar air velocity which is the only important input for the evaporation calculation.

To determine the total vapor emission rate, the mass flow of the flashing liquid (using the isentropic balance) is added to the evaporation rate.

A potential failure location of a valve is directly behind a storage vessel of liquid high pressure and high temperature refrigerant, for example a defective liquid line valve behind the receiver tank.

A simplified short pipe release without wall friction, valve loss coefficients and gravity head minimizes required assumptions about the refrigeration system. This release scenario is solely dependent on the diameter of the hole and the difference in pressure inside and outside of the pipe. Therefore, this scenario is simple but still realistic.

Equation 2 simplifies to:

[E.sub.l] = [c.sub.o][A.sub.h][[rho]sub.l][[2((p-[p.sub.o])/[p.sub.l])].sup.1/2] (18)

The EPA recommends a discharge coefficient [c.sub.o] = 0.6, however, this coefficient depends on the flow conditions out of the orifice, with [c.sub.o] = 1 representing flow that is not hindered by the shape of the orifice (such as a severed pipe).

During the course of the release the pressure and temperature inside the storage vessel are kept constant (US EPA 1992). In reality, the loss of liquid refrigerant out of the valve causes evaporation inside the vessel and therefore a reduction of the temperature and pressure of the system. The resulting total liquid emission rate would therefore decrease slightly faster over time, than the one calculated using the assumption of constant pressure and temperature, which is therefore the simpler and more conservative approach.

As described above, the flashing emission rate of the R-134a release is calculated using the isentropic method. The remaining liquid R-134a forms a pool. The pool size is time dependent and calculated using an air velocity of 50 ft/min (0.254 m/s).

The rate of flashing liquid is proportional to the total liquid release rate and decreases slowly over time. Shortly after the release, the pool size is small and therefore the pool evaporation rate is insignificant. However, over time the pool grows until equilibrium is reached, where the amount of liquid refrigerant supplying the pool is equal to the evaporation rate, resulting in a total vapor emission rate that is equal to the total liquid emission rate.

The above described calculations were repeated for other hole diameters. Figures 7 and 8 show the dependence of the various calculated emission rates on the diameter of the release orifice for R-134a and Ammonia, respectively. The total liquid mass emission rate, the vapor emission rate due to flashing, the equilibrium liquid pool evaporation rate and the total vapor emission rate are plotted with the hole diameter on a log-log scale for two examples: Ammonia (R-717) and 1,1,1,2-Tetrafluoroethane (R-134a). The vapor released from flashing and evaporation of a 0.25 in. (6.35 mm) hole represents approximately 95% (5% would exceed this) of the accidents found in the databases. A 0.5 in. (12.7 mm) hole exceeds 99% of the accidents.

Release Scenario 2: Gas, High Pressure, High Temperature

This was initially rated as the second worst vapor release potential. A high pressure, high temperature gas release is possible in the piping between the compressor and the condenser (see Figure 6). If the pressure in the pipe exceeds the critical pressure, the flow velocity equals the speed of sound through the hole and the flow is "choked." If that is the case, the vapor emission rate is constant regardless of how much greater the pressure inside the pipe is above the critical pressure. The choked condition is therefore the limiting case for this scenario. This scenario has significantly lower emission rates than observed (see Table 3.2) and will not be considered as a reasonable case from which to base the room ventilation rate.

Release Scenario 3: Liquid, High Pressure, Low Temperature

This scenario is very similar to Scenario 1 described. A high pressure, low temperature liquid refrigerant release is possible between the heat exchanger and the expansion valve. All equations used in Scenario 1 can be applied to this scenario as well. Insignificant differences in the resulting emission rates are expected for this scenario compared to Scenario 1, because the total liquid emission rate is solely driven by the pressure difference between the inside and outside of the vessel. The lower temperature of the refrigerant causes a slightly lower flashing rate and a lower evaporation rate. However, once the system reaches equilibrium (i.e. the evaporation rate equals the total liquid emission rate minus the flashing rate) the temperature difference between the two scenarios will have no effect and this scenario will provide the same emission rates as Scenario 1.

[FIGURE 7 OMITTED]

[FIGURE 8 OMITTED]

Release Scenario 4: Liquid, Low Pressure, Low Temperature

A low pressure, low temperature liquid release of refrigerant is possible between the expansion valve and the evaporator. If the pressure inside the pipe or vessel drops below the ambient pressure, the release is governed by the weight of liquid above the orifice and no release may occur until the pressure is equalized. Due to the lower temperature compared to Scenario 1, the rate of flashing liquid is significantly lower. The equilibrium pool size is smaller than that for Scenario 1 because the overall liquid emission rate is lower. Since the calculated emission rates are significantly lower than scenario 1, this scenario has also be ruled out for use in determining the ventilation rate.

Release Scenario 5: Gas, Low Pressure, Low Temperature

This scenario is very similar to Scenario 2. A low pressure, low temperature gaseous refrigerant release is possible between the evaporator and the compressor. All equations used in Scenario 2 can be applied to this scenario as well. The resulting vapor emission rate is expected to be lower than that calculated in Scenario 2, because the total vapor emission rate is solely driven by the pressure difference between the inside and outside of the vessel. A reduction in that pressure difference causes a reduction in the emission rate. If the pressure inside the pipe or vessel falls below the critical pressure, the condition of choked flow no longer holds, so the vapor flow rate is dependent on the pressure inside the vessel or pipe. The total mass flow rate for un-choked flows is less than that for choked flows. Since the calculated emission rates are significantly less than Scenario 1, this scenario will also be ruled out for use in determining the ventilation rate.

Recommended Release Scenario for Ventilation Room Design

To account for a likely worst-case scenario, it is recommended that the high temperature, high pressure release be used. Although the two-phase release is the most realistic, accounting for all the variables for the evaporating pool, including surface velocity, pool depth, pool size (floor area), and heat transfer considerations becomes very difficult and full of ambiguous assumptions. In addition, the emission rate for the evaporating pool can become very large. Hence, it is assumed that a drain is provided to capture any liquid that would accumulate. All vapor emissions will be created by the flashed vapor from a 0.5 in. (12.7 mm) hole. Only taking into account the flashing vapor (no evaporation from potential liquid), for Ammonia, this represents an accident scenario with only a 5% probability of exceedance which seem like a reasonable probability for design purposes (including evaporation, this scenario would yield a 1% probability of exceedance). It is further assumed that similar probabilities would occur for other refrigerants.

VENTILATION RATE

The purpose of ventilating machinery rooms is to dilute the air in the room such that the recommended concentration limit (RCL) in Tables 1 and 2 of ANSI/ASHRAE Standard 34 (ASHRAE 2010) is not exceeded. However, depending on the actual leak rate, ventilation may, or may not, dilute the refrigerant concentration to the RCL or lower using the current Standard 15 method. This section will presents methods whereby a designer can specify a ventilation rate to ensure the concentration will be below the RCL value. In subsequent paragraphs, a derivation of the equations needed to estimate the concentration in a machinery room due to an accidental release of any refrigerant is presented.

Detector Activation of Ventilation System

Standard 15 (ASHRAE 2010) outlines the requirement that every refrigerating machinery room contains a detector that activates an alarm and mechanical ventilation in case of a refrigerant leak. The alarm is to be activated at a refrigerant concentration level no higher than the threshold limit value (TLV) time weighted average (TWA). The threshold limit value-time-weighted-average (TLV-TWA) is the average concentration for a normal workday exposure (40 hours a week, 8 hours a day) without any adverse effect. Table 3 lists the TLV-TWA for different refrigerants. In the following analyses and examples the TLV-TWA shall be used as the thresh-old for the activation of the ventilation. For those chemicals where a TLV-TWA is not available, 10% of the RCL is used as the detector threshold value.
```Table 3. Refrigerant Quantities to Calculate Ventilation Rate

Refrigerant RCL (a) Concentration
Setpoint (b)
ppm g/[m.sup.3] lb/Mcf ppm mg/[m.sup.3]
(v/v) (v/v)

R-12 18,000 90 5.6 1,000 4,945

R-22 59,000 210 13 1,000 3,537

R-23 41,000 120 7.3 4,100 12,000

R-32 36,000 77 4.8 3,600 7,700

R-123 9,100 57 3.5 910 5,700

R-124 10,000 56 3.5 1,000 5,600

R-125 75,000 370 23 1,000 4,909

R-134a 50,000 210 13 1,000 4,173

R-143a 21,000 70 4.5 1,000 3,437

R-152a 12,000 32 2.0 1,200 3,200

R-170 7,000 9 0.54 1,000 1,230

R-245fa 34,000 190 12 3,400 19,000

R-290 5,300 10 0.56 1,000 1,804

R-404A 130,000 500 31 1,000 3,992

R-407C 76,000 270 18 7,600 27,000

R-410A 130,000 390 25 13,000 39,000

R-507A 130,000 520 32 1,000 4,043

R-600a 4,000 10 0.6 400 960

R-717 (c) 320 0 0.014 25 17

R-1270 1,000 2 0.1 500 861

Refrigerant m [PHI] [Q.sub.max]
lb/Mcf sec cfm L/s

R-12 0.31 39 0.33 44,400 21,000

R-22 0.23 32 0.34 23,400 11,100

R-23 0.75 24 0.61 95,700 45,200

R-32 0.49 28 0.34 73,600 34,800

R-123 0.36 161 0.049 2,550 1,210

R-124 0.35 51 0.3 48,600 23,000

R-125 0.31 28 0.54 24,700 11,700

R-134a 0.27 40 0.35 19,800 9,310

R-143a 0.22 33 0.46 94,400 44,600

R-152a 0.2 50 0.29 85,200 40,200

R-170 0.077 39 0.64 873,000 412,000

R-245fa 1.2 98 0.14 3,580 1,690

R-290 0.12 53 0.4 372,000 176,000

R-404A 0.25 32 0.5 14,600 6,860

R-407C 1.7 31 0.4 22,200 10,500

R-410A 2.5 27 0.42 18,300 8,640

R-507A 0.26 31 0.5 14,500 6,820

R-600a 0.06 87 0.29 164,000 77,200

R-717 (c) 0.0011 45 0.21 10,100,000 4,730,000

R-1270 0.054 48 0.4 2,310,000 1,090,000

(a.) From ANSI/ASHRAE Standard 34-2010 (ASHRAE 2010). If
disputes arise, Standard 34 takes precedent.

(b.) TLV/TWA value or 1/10 RCL if TLV/TWA not available

(c.) R-717 see IIAR Standard 2 (IIAR 2008) for required
ventilation rates
```

Refrigerant Concentration in a Room

A mass balance equation is used to solve for the concentration in a room as a function of the sources of the contaminant and the losses, mainly ventilation. Several assumptions are made while deriving and solving this mass balance:

1. The room is well-mixed. When the contaminant refrigerant is released into the enclosed room and is gas, the emitted mass is instantaneously diluted in the volume of the room. Without much more sophisticated tools to account for densities, mixing rates, and other parameters to deter-mine stratification and the concentration gradient, this assumption will underestimate the concentration in part of the room and overestimate it in others. For example, if the refrigerant is heavier than air and the ventilation rate is initially small (low mixing), the concentration would be higher at the floor level and lower near the ceiling. For this reason, many emergency exhaust intakes are located nearer to the floor. Once the ventilation rate is increased, better mixing will most likely occur.

2. The contaminant is inert. There are no reactions, either homogenous or heterogeneous, in the volume of the room.

3. There is no nucleation or condensation. Once the contaminant refrigerant is gas, no nucleation or condensation occurs.

For a volume of space, V, the change in concentration, C, is equal to the emission rate, E, minus the ventilation, Q, of the contaminant out of the room:

V dC(t)/dt = E(t) - QC(t) (19)

Dividing all terms by the volume yields a first order differential equation.

dC(t)/dt + [lambda]C(t) = E(t)/V (20)

By multiplying all terms by [e.sup.[lambda]t] and using the reverse chain rule, an intermediate solution for C(t) is found:

[e.sup.[lambda]t] dC(t)/dt + [e.sup.[lambda]t]C(t) = [e.sup.[lambda]t]E(t)/V (21)

d/dt[C(t) [e.sup.[lambda]t]] = [e.sup.[lambda]t]E(t)/V (22)

C(t) = [e.sup.-[lambde]t][integral[e.sup.[lambde]r]E(t)/V dt] (23)

Assuming a constant emission rate, E(t) = [E.sub.0], and an initial condition of C(0) = [C.sub.0], the analytical solution for C(t) is:

C(t) = [E.sub.0]/Q(1-[e.sup.-[lambde]t]) + [C.sub.0] [e.sup.-[lambde]t] (24)

For most other emission rates, a numerical approach is needed to find a solution.

The analytical solutions consist of two fundamental mechanisms. The decay of the concentration due to ventilation (the only loss considered in this approach) is given by the [C.sub.0] [e.sup.-[lambde]t] (24) term. The second term(s), with the (1-[e.sup.-[lambde]t]) term or variants of it, represent the approach of the concentration towards an asymptote that represents a steady-state concentration. This term is multiplied by the source emission to deter-mine the magnitude.

For each unique set of source and loss terms (emission rate and ventilation rate), the equation should be reconsidered with a new C0 and E0 and any other parameter that may have changed. Each time the emission or ventilation rates change, the system attempts to reach a new equilibrium. For example, for a short time at the beginning of the leak, the emergency exhaust ventilation has not yet started, so the concentration rises in the room. When the ventilation rate increases, the steady-state concentration value decreases.

Assuming a constant emission rate, at some point a steady-state concentration is reached. This also constitutes the worst case scenario. For a constant emission rate, the concentration of the refrigerant in the room is independent of the room volume and is only a function the emission rate and the RCL as shown below:

C(t) = [E.sub.0]/[lambde] = [E.sub.0] /Q (25)

Therefore, in order to calculate the ventilation rate for the worst case scenario (steady-state for a given leak), the formula will be:

Q = [E.sub.0] /RCL (26)

where RCL is the recommended concentration limit given by Standard 34 and [E.sub.0] is determined by the accident scenario.

The steady state solution is conservative because it only examines the maximum emission rate while not being able to consider such factors as the room volume, time varying emission rate and variable ventilation (i.e., low ventilation when concentration is below the TLV and high thereafter). By calculateing the concentration as a function of time, these factors can be considered.

Although a numerical integration of the Equation 25 through various implicit or explicit methods would yield a pure numerical solution, it can be used for each time step, resetting the value for the initial concentration, [C.sub.0], and the emission rate, [E.sub.0], correlated to each time step, n. The super-script n denotes the time step.

[C.sup.n] = [E.sup.n]/[Q.sup.n](1-[e.sup.[[lambde].sup.n]([t.sup.n] - [t.sup.n-1]])]) + [C.sup.n-1] [e.sup.[[lambde].sup.n]([t.sup.n] - [t.sup.n-1]])]) (27)

The release duration is defined by the total mass of the refrigerant. The emitted liquid, which either flashed into vapor or spilled as liquid and then evaporated, is limited by the reservoir of available refrigerant. Therefore, the flashed vapor ends when the liquid emission is done and the liquid pool stops growing and diminishes in size while the refrigerant evaporates. Figure 9 shows a sample plot of the emission rate, showing the total liquid emission rate, the flashing emission rate, and the evaporation emission rate. The figure demonstrates the relationship between the various mechanisms contributing to refrigerant vapor in the room.

[FIGURE 9 OMITTED]

The evaporation from the pool creates difficult nonlinearities in the solution because it introduces additional variables that must be resolved, such as the depth of the pool, air velocity at the surface of the pool, and the size of the pool, which necessitates a floor area. Comparing a leak from a 0.25 in. (6.35 mm) hole with pool evaporation and from a 0.5 in. (12.7 mm) hole without evaporation yielded similar gas emission rates for most refrigerants. Therefore, the simplification to increase the hole size but ignore the evaporation from the liquid pool was made for the selected accident scenario. Further, it is recommended that a drain be present in the room to remove liquid from the floor.

For each time step, the liquid emission rate is broken up into the portion that flashes and the liquid pool, which evaporates, but is ignored. This total vapor emission rate is used to calculate the concentration in the room. The maximum concentration is found and a routine solves for the ventilation rate such that the maximum concentration does not exceed the RCL. Several of the variables that can be changed are: the total refrigerant mass, the hole diameter for the accident scenario, the temperature and pressure of the refrigerant in the vessel, the volume of the room, and the RCL and TLV-TWA values (which are dependent on the refrigerant).

The steady-state solution yields the easiest calculation for the required ventilation rate, but also is the most conservative. It may be a starting point for the calculation, but a time varying analysis that accounts for the emission rate not reaching equilibrium gives a much more accurate solution. The transient analysis ends up calculating the required ventilation rate by using the concentration in the room at the maximum emission rate. Depending on a number of variables, including the mass of the refrigerant, the room size, and how the liquid pool is contained, the ventilation requirement may be higher than previously presented in Standard 15 (ASHRAE 2010) by using the [cube root of (100)]G equation (for lbs of refrigerant). For lower masses and larger rooms, the ventilation rate may be lower and the RCL may not even be reached.

Safe Room Volume

The safe room volume, as defined here, is the volume of room that is needed in order for the concentration of the leaked refrigerant to remain below the RCL if all the mass was leaked and mixed instantaneously. It also takes into account the flashed portion of the liquid release from the accident scenario. The safe room volume may be defined as:

[V.sub.s] = G[PHI]/RCL (28)

The values for [PHI] are found in Table 3.

The safe room volume fraction, f, is defined as the ratio of the room volume to the safe room volume (V/[V.sub.s]). This non-dimensional variable incorporates the refrigerant mass, room volume, RCL, and accident scenario. If the safe room volume fraction, f, is greater than unity, no additional ventilation will be needed if there is an accidental release.

Detector Delay Time

The time between the start of the refrigerant release and initiation of the ventilation fan may have a big impact on the ventilation rate needed. Increased ventilation is needed to slow down the increase in concentration in the room. When the reservoir of available refrigerant is exhausted, the ventilation will continue to flush the room. An example of this effect on concentration is shown in Figure 10, which uses the same parameters for R-134a as the accident scenario. The concentration rises until all the available refrigerant has leaked out, whereupon the ventilation purges the room of the refrigerant. The goal is keep the maximum concentration below the RCL at all times. If left too long, the concentration will exceed the RCL. Ventilation will slow the increase in concentration and eventually purge the room, but the RCL will have already been exceeded.

The maximum delay time was found for which some amount of ventilation could prevent the concentration from exceeding the RCL. This ventilation rate is excessively high. Figure 11 shows that delaying ventilation increases the ventilation requirement dramatically as it gets to the maximum delay time. To limit the amount of ventilation required, a detector delay time has been provided such that it is 50% of the maximum detector delay time. All maximum ventilation rates ([Q.sub.max]) are calculated at this point.

[FIGURE 10 OMITTED]

[FIGURE 11 OMITTED]

For each refrigerant and machinery room, the allowable detector delay time, in seconds, is calculated using m (Table 3) and f:

[t.sub.d] [less than or equal to] mf for f < 1 (29)

Recommended Ventilation Rate

By calculating the required ventilation rate as a function of the safe room volume fraction, f, the required ventilation rate fitted to the data (see Figure 12) becomes:

Q = 0 for f > 1(30)

Q = [Q.sub.max](1 + 0.3f - 1.3[f.sup.2]) for f < 1(31)

The values for [Q.sub.max] are provided in Table 3.

DISCUSSION

[FIGURE 12 OMITTED]

It is useful to compare this method of calculating the ventilation requirement with that of the Standard 15 (ASHRAE 2010). Since the approach in this work contains many more variables than just the refrigerant mass, the trend of the ventilation requirement is not quite as predictable as the square root of the mass. In general, the ventilation requirement goes down when the mass of the refrigerant is lower, when the volume of the room is larger, when the refrigerant is not as volatile, and when the RCL is higher. To compare the two methods, a comparison has been calculated using chiller rooms that have been designed, as listed in Table 4. In some cases, the ventilation requirement goes up. In others, it decreases. For a few rooms, the safe volume fraction is above unity, so no additional ventilation is needed to keep a leak of the entire mass from exceeding the RCL in the room.
```Table 4. Comparison of Proposed Ventilation Requirement with
Current Standard 15

Chiller Room Refrigerant Mass [PHI] RCL f [Q.sub.max]
Volume

[ft.sup.3] lb lb/Mcf cfm

3300 134a 124 0.35 13 0.99 19,800

7956 134a 400 0.35 13 0.74 19,800

4730 123 750 0.049 3.5 0.45 2550

33,895 123 1050 0.049 3.5 2.31 2550

5299 134a 355 0.35 13 0.55 19,800

36,229 134a 2300 0.35 13 0.59 19,800

8788 134a 760 0.35 13 0.43 19,800

20,925 134a 760 0.35 13 1.02 19,800

29,700 134a 625 0.35 13 1.77 19,800

Chiller Room Q [this Q [Standard
Volume work] 15]

[ft.sup.3] cfm cfm

3300 521 1114

7956 10,140 2000

4730 2222 2739

33,895 0 3240

5299 15,181 1884

36,229 14,464 4796

8788 17,603 2757

20,925 0 2757

29,700 0 2500
```

CONCLUSIONS

The current recommendation for ventilation in refrigeration machinery rooms found in Standard 15 only accounts for the mass of the refrigerant in the room and does not have a strong scientific rationale for its existence. This work presents a ventilation requirement for refrigerant machinery rooms in the case of a release. The approach is simple enough to be codified or put in a standard and includes the major variables that would affect the ventilation requirement: mass of the refrigerant in the room, volume of the room, refrigerant properties, a concentration limit, and an appropriately conservative and realistic accident scenario. Further work is needed to expand the refrigerants with RCL values as well as better defining some of the refrigerant properties in Standard 34 to complete this work for all refrigerants.

ACKNOWLEDGMENTS

This work was funded by ASHRAE research project 1448-RP through TC 4.3 - Ventilation Requirements & Infiltration.

NOMENCLATURE

A = area

[c.sub.0] = discharge coefficient

[c.sub.p] = specific heat at constant pressure

[c.sub.v] = specific heat at constant volume

d = diameter

E = mass flow rate

[f.sub.f] = friction factor

f = safe volume fraction, V/[V.sub.S]

G = mass

g = acceleration due to gravity

H = height

[H..sub.vap] = enthalpy of vaporization (the difference between [H.sub.v] and [H.sub.t] at the boiling point and ambient pressure)

p = pressure of the liquid inside the vessel,

K = fitting loss coefficient

k = ratio of specific heats [c.cub.p]/[c.sub.v]

L = length

M = mass

[M.sub.w] = molecular weight

m = detector delay time influence factor (Table 3)

Re = Reynolds number

Q = volume flow rate

R = universal gas constant

S = entropy

[DELTA][S.sub.vap] = entropy of vaporization (difference between [S.sub.v] and [S.sub.t] at the boiling point and ambient pressure)

T = temperature

t = time

U = air velocity

V = volume

[V.sub.p] = vapor pressure (at [T.sub.2])

[V.sub.ph] = vapor pressure of hydrazine (at [T.sub.2])

[V.sub.s] = safe room volume, M[PHI]/RCL

Greek

[epsilon] = roughness height

[lambde] = air exchange rate, Q/V

v = kinematic viscosity

[THETA] = angle relative to the upward vertical direction formed by the liquid surface remaining around the circumference of the tank

[rho] = density

[PHI] = percentage of the liquid that flashes, found in Table 3

Subscripts

b = boiling

d = detector delay

f = flashing

h = hole

l = liquid

v = vapor

s = surface

0 = ambient or initial

1,2,3 = states

max = maximum

DISCUSSION

Joy Kohler, Engineering Manager, Johnson Controls, York, PA: The report cited only one example of fluorocarbon leaks in machinery rooms: a fatality accident in Alaska. Most of the data is from ammonia systems, which are significantly different from fluorocarbon systems due to the self-alarming nature of ammonia and the typical design of ammonia systems. What effort was made to locate such data? Certainly there must be accident data for such systems.

Scot K. Waye: The major databases that were searched for refrigerant leaks in machinery rooms are the following: MARS, NRC, ARIP, OSHA, and NIOSH. Table 1 of the paper lists the number of accidents reported to the NRC for different refrigerants. The number of fluorocarbon refrigerant releases reported for the indicated time period are very few, especially as compared to ammonia. The same is true for the OSHA data-base. No accidents involving fluorocarbon refrigerants were found in the MARS database, and only one incident was found in the NIOSH database (the Alaska incident). Summaries of the incidents reported in the NRC database are in the appendix of the full report for ASHRAE RP-1448. [Editor's note: final reports of ASHRAE Research Projects are available for free download to members at www.ashrae.org.] Efforts were made to find additional databases and sources for more reported leaks involving non-ammonia refrigerants, but these failed to provide additional incidents.

For many refrigerants, apart from Ammonia, the quantity released that triggers a mandatory report is high, which we believe causes an underreporting of incidents for fluorocarbon leaks. For some refrigerants, no reportable quantity is listed at all. The report for the Alaska incident was due to a fatality, causing an investigation and a report. For some cases, it is presumed that without injury or fatality a report of the leak was not mandatory under the current requirements; this also causes underreporting. We acknowledge that better reporting and more data representing fluorocarbon refrigerant leaks would enhance understanding of the issue, but this data was not avail-able to us through public and open sources. Although much of the data from actual releases was for ammonia, trends of potential leak locations and causes may be examined. Additionally, a threshold probability for emission rate exceedance can be estimated for worst-case emission scenarios.

REFERENCES

AIChE (American Institute of Chemical Engineers). 1996. Guidelines of Vapor Cloud Dispersion Models, 2nd Edition. New York: Center for Chemical Process Safety.

API (American Petroleum Institute). 1996. Manual for Modeling Hypothetical Accidental Releases to the Atmosphere. API PUBL 4628.

ASHRAE. 2009. 2009 ASHRAE Handbook--Fundamental. Atlanta: American Society of Heating, Air-Conditioning and Refrigeration Engineers, Inc.

ASHRAE. 2010. ANSI/ASHRAE Standard 15-2010, Safety Standard for Refrigeration Systems. Atlanta: American Society of Heating, Air-Conditioning and Refrigeration Engineers, Inc.

ASHRAE. 2010. ANSI/ASHRAE Standard 34-2010, Designation and Safety Classification of Refrigerants. Atlanta: American Society of Heating, Air-Conditioning and Refrigeration Engineers, Inc.

Brown, R. 2005. Machinery Room Ventilation for Industrial Refrigeration Systems: A Rational Engineering Approach. Technical Paper No. 5. IIAR Ammonia Refrigeration Conference and Exhibition. Acapulco, Mexico.

Fenton, D. L. and W.V. Richards. 2003. User's Manual for ANSI/ASHRAE Standard 15-2001 Safety Standard for Refrigeration Systems. ASHRAE Special Project SP-93. Atlanta: American Society of Heating, Air-Conditioning and Refrigeration Engineers, Inc.

IIAR. 2008. ANSI/IIAR 2-2008 (Addendum A), Americation National Standard for Equipment, Design and Installation of Closed-Circuit Ammonia Mechanical Refrigerating Systems. Alexandria: International Institute of Ammonia Refrigeration.

MAHB (Major Accident Hazard Bureau). 2001. eMARS -- Major Accident Reporting System. www.emars.jrc.ec.europa.eu/?id=4.

Moran, M.J. and H.N. Shapiro. 2000. Fundamentals of Engineering Thermodynamics, 4th Ed. New York: John Wiley and Sons.

NIOSH (National Institute for Occupational Safety and Health). 2010. Fatality Assessment and Control Evaluation (FACE) Program. www.cdc.gov/niosh/face/.

OSHA (Occupational Safety & Health Administration). 1992. Process Safety Management (PSM) of Highly Hazardous Chemicals Standard. 29 CFR 1910.119.

OSHA (Occupational Safety & Health Administration). 2010. Fatality and Catastrophe Investigation Summaries. www.osha.gov/pls/imis/accidentsearch.html

Richards, W.V. 1986. How Codes and Regulations Affect the Design of Your Plant. IIAR Ammonia Refrigeration Conference and Exhibition. Innisbrook, FL.

Sallet, D.W. 1990. Critical two-phase mass flow rates of liquefied gases. Journal of Loss Prevention in the Process Industries 3:38-42.

Seidl, R. and S.T. Taylor. 2005. Sizing Exhaust System for Refrigerating Machinery Rooms. Taylor Engineering, Alameda, CA.

Stoecker, W. F. 1998. Industrial Refrigeration Handbook. New York: McGraw-Hill Companies, Inc.

US EPA. 1992. Workbook of Screening Techniques for Assessing Impacts of Toxic Air Pollutants (Revised). Office of Air Quality Planning and Standards.

US EPA. 2001. Consolidated List of Chemicals Subject to the Emergency Planning and Community Right-To-Know Act (EPCRA) and Section 112[R] of the Clean Air Act. EPA 500-B-01-003. Office of Solid Waste and Emergency Response (5104).

US EPA, USA. US EPA. 2010. www.epa.gov/emergencies/tools.htm#arip

Wilson, D.M. 1981. Along-wind diffusion and source transients. Atmospheric Environment, 15:489-495.

This paper is based on findings resulting from ASHRAE Research Project RP-1448.

Scot K. Waye, PhD, PE Member ASHRAE

Ronald L. Petersen, PhD Member ASHRAE

Anke Beyer-Lout

Scot K. Waye is a senior engineer, Ronald L. Petersen is a principal and vice president, and Anke Beyer-Lout is a project scientist at CPP, Inc. in Fort Collins, CO.
COPYRIGHT 2012 American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.

Author: Printer friendly Cite/link Email Feedback Waye, Scot K.; Petersen, Ronald L.; Beyer-Lout, Anke ASHRAE Transactions Jul 1, 2012 10902 Laboratory testing of a fabric air dispersion system. Secondary control strategies for cycling of air conditioning and refrigeration systems. Machinery Magneto-electric machines Refrigerants Safety regulations