The Fundamentals of

Fluid Flow Conditioning

Danny Sawchuk
Canada Pipeline Accessories
10653- 46th Streeet, Calgary, Alberta, T2C5C2
 2

Introduction

Flow conditioning is one of the most critical aspects dealing with any type of volumetric flow
metering. Flow conditioning is the final buffer between the flow meter and the upstream piping
layout. Flow conditioning is responsible for eliminating swirl, restoring profile symmetry, and
generating a repeatable, fully developed velocity flow profile. Modern advancements have
resulted in low uncertainty, high repeatability devices that are effective across a range of flow
rates. However, proper utilization of flow conditioners is still required to maximize the meters’
performance and diagnostics to ensure the most stable long-term flow measurement. All flow
conditioner technologies are not made equal, as commonly used designs such as AGA tube
bundles and straightening vanes can actually cause more measurement problems than they
resolve. This paper will focus on two main types of flow conditioners: perforated plate flow
conditioners, and straightening vanes such as tube bundles.

Over the past few decades, flow conditioner systems such as AGA-3 19 Tube Bundles and
straightening vanes have an extensive history of use in liquid and gas measurement systems.
They are even being used today in modern measurement scenarios that utilize ultrasonic flow
metering. Historically, the common belief was that due to their length, straightening vanes
must be very effective at swirl removal, and must result in an excellent measurement device
that offered a low pressure drop. Modern research has shown that this is quite the opposite. We
now know that their excessive length and low pressure loss makes them unsuitable for use in
precision measurement applications, both in liquid and gas phase scenarios.
 3

Installation Effects

An installation effect is any object, fitting, or obstruction in a pipe that leads to a fluid flow
disturbance. This can be any valve, tee, orifice plate, eccentric orifice plate, pipe wall rust,
misaligned gaskets, or intruding probe that causes a distortion in the fluid flow. This can also be
any series of piping direction changes (due to elbows, tees, or headers) that result in a bulk fluid
rotation (swirl).

Flow Profiles

The key to studying flow conditioner performance is to understand the behaviour of velocity
flow profiles. A velocity profile is a cutaway of the flow through the centre of the pipe. It shows
the distribution of fluid velocity in a gas or liquid pipeline as you cross from one pipe wall to
the other. The velocity profile is used to determine the state of the flow and to determine if
disturbances such as swirl or installation effects are present.

A fully developed flow profile is the perfect, symmetrical distribution of velocity from one pipe
wall to the other. It is the natural state that all fluid flows return to over long distances of pipe,
due to the influence of the pipe wall friction. A deviation from this fully developed profile means
that there are disturbances in the flow. Swirl will cause the velocity peaks to move closer to
the walls due to centripetal force, while disturbances such as tees, elbows, valves, gaskets, and
probes will cause a distinct and unpredictable lack of symmetry. Good flow conditioners return
the flow to the fully developed state in the shortest distance possible. Otherwise, at least 200
internal diameters of pipe may be required to fully ensure that no disturbances remain.
 4

Figure 1: Fully developed velocity flow profiles in a pipe cross section. Distribution is from one
pipe wall to the other through the center of the pipe

Figure 2: Example velocity flow profiles for non fully developed flow. Profiles shown are due to
two elbows out of plane.
 5

Tube Bundle Performance Deficits

Straightening vanes and tube bundles are flow-conditioning systems that are composed of vanes
or small tubes welded together. Their openings are usually the same size and they can be roughly
two pipe diameters long. It is commonly assumed that the length of a flow conditioning device
is directly linked to its effectiveness. Testing has actually proven this to be incorrect. In fact, the
length of the flow conditioner actually causes more harm than good. To better understand this,
the discrete behaviour of flow conditioners must be discussed.

There are three separate actions that a fluid flow undergoes when being flow conditioned:
(1) the initial pressure drop as the fluid is guided through the flow conditioner opening; (2)
the fluid travelling through the flow conditioner passages; and then (3) the fluid leaving the
flow conditioner and returning to pipe flow conditions. The initial pressure drop forces the
redistribution of fluid across the flow conditioner surface and will have the strongest impact on
the final velocity profile distribution of the fluid. If the flow conditioner has a low-pressure drop
with holes that are equally sized, as in the case of a tube bundle, it will be unable to eliminate
uneven velocity distribution in the pipe due to a distorted flow profile. The distorted distribution
that enters the flow conditioner would be exactly the same distribution once it re-enters the pipe.
A pressure drop is the first step to ensuring that the flow conditioner can effectively eliminate
disturbances and a proper velocity distribution in the pipe.
 6

The friction, due to the wall roughness, will act on the fluid and eventually restore the fully
developed velocity flow profile if the pipe is long enough. This takes place in any pipe and
does occur within the flow conditioners as well; the wall friction in each fluid passage acts on
the fluid and attempts to form discrete flow profiles within the flow conditioner itself. As a
result, the length of the flow conditioner will actually lock in the profile distortions due to the
formation of discrete, fully developed flow patterns. The pressure drop due to the significant
fluid/wall contact will also become a concern at lower viscosities. Pressure loss due to wall
friction is typically not a concern in high-pressure natural gas applications (Reynolds numbers of
approximately 5,000,000 to 30,000,000) but quickly becomes a problem when the fluid viscosity
is increased.

Once the fluid leaves the flow conditioner, it then recombines into a total, bulk velocity profile.
This is where the problems of the tube bundles become immediately evident, as all the separate
miniature flow paths recombine into one. Because of the low initial pressure drop, this means
that the volume, velocity, and energy of the flow through each passage is not equal. When
these flows recombine, they result in a single, highly distorted profile, as each passage is not
contributing equally to the flow. This means that the tube bundle itself becomes an installation
effect, just like a valve, tee, or any other disturbance that results in a flow measurement error.

Figure 3: Tube bundle exposed to minor swirl. A 30D meter run is required to ensure fully
developed flow.
 7

Figure 4: Tube bundle exposed to major swirl. A 40D meter run is required to ensure fully
developed flow. A 20D meter run has a heavily inverted flow profile due to higher flow energy
being stuck in outer tubes in bundle.

Figure 5: Tube bundle exposed to severe swirl (theoretical flow behaviour due to swirl generator).
Using a 20D meter run, the flow profile is extremely inverted.
 8

Pipe Wall Friction

A fundamental of fluid dynamics and flow measurement is that the pipe itself is a flow
conditioner. Pipe wall friction and fluid viscosity are the most critical variables determining the
shape of a fully developed flow profile in any gas or liquid pipeline. Flow profiles are generated
by the pipeline wall drag. The fluid pushes down the pipeline as quickly as possible while the
wall friction resists this motion, creating the arc of a flow profile as the velocity is reduced
near the pipe wall. The smoother the pipe wall, the flatter the profile, as the drag is reduced.
Rougher pipe walls have pointier profiles due to the higher drag. Viscosity has the same effect
on the flow profile shape. Higher viscosity fluids have a greater resistance to flow and result
in more wall drag and a peaky flow profile, while lower viscosity fluids result in less drag and
a flatter profile. This means that higher viscosity flows or rougher pipes restore their fully
developed flow profile much more quickly than lower viscosity flows or smoother pipes. The
increased pipe wall drag, due to the viscosity and wall roughness, is dragging the flow back into
a fully developed state while also eliminating swirl.

Figure 6: Formation of a fully developed flow profile due to the influence of wall friction.
 9

Figure 7: Fully developed flow profiles in a laminar flow, Re ~ 800.

Figure 8: Fully developed flow profiles in a turbulent flow, Re ~ 30,000,000.

 10

Perforated Plate Flow Conditioning

A perforated plate flow conditioner is a solid metal disk with a carefully designed hole pattern.
It is typically only a few inches thick. It is effective for three reasons: (1) a tested hole layout,
(2) an abrupt aggressive pressure drop, and (3) the short length.

First, the hole layout is designed to result in an effective redistribution of the fluid flow. This
helps balance the fluid so that it can reach the fully developed flow profile distribution more
quickly.

Second, the pressure drop helps properly balance the fluid flow across the flow conditioner and
ensure that each hole moves the amount of fluid that it was designed for. This step removes any
flow distortions or asymmetry. The fluid is also accelerated through the holes in the plate; this
acts as a filter and removes the rotational vector from the swirl.

Third, the short length ensures that the fluid doesn’t have time to redevelop flow profiles within
the flow conditioner itself. The symmetrical, swirl free flow leaves the flow conditioner and
it recombines into a bulk, swirl free, fully developed flow profile after 5 – 8 internal pipe
diameters.

Figure 9: Flow profiles downstream of a perforated plate flow conditioner, with straight pipe
upstream.
 11

Figure 10: Flow profiles downstream of a perforated plate flow conditioner, with elbows upstream.

Figure 11: Flow profiles downstream of a perforated plate flow conditioner, with tees upstream.
 12

Orifice Meters & Differential Pressure Measurement

An orifice meter is a very simple type of meter. The flow is forced through an obstruction
known as the primary element, and the fluid pressure is measured before and after the element.
The square root of the difference in this pressure is approximately proportional to the mass flow
rate and velocity of the fluid. This behaviour is applicable to all types of differential pressure
measurement. For example, venturi meters, cone meters, wedge meters, and multi-hole orifice
plates. The fundamentals for each type are exactly the same. However, the specifics of the tap
location and primary element shape are what determine the rangeability of the flow meter, the
permanent pressure recovery, and the resistance to errors caused by flow distortions.

Due to the limited source of input information, orifice meters can be very sensitive to velocity
profile distortions and swirl. Lower beta ratio orifice installations are less sensitive to velocity
profile distortions as the increased pressure drop of the orifice plate helps act as a flow
conditioner, and results in a better-averaged state of the flow profile. The fundamental problem
with differential pressure (dP) measurement is that all of the data is typically being retrieved
from a single pair of pressure taps. This reduces the amount of information available, and makes
the meter very sensitive to local pressure effects based on the physical location of the taps.
 13

Swirl is one of the primary concerns when dealing with orifice meters and dP measurement for
a variety of reasons. First, the rotational velocity component of the swirl causes a measurement
error. The increased velocity directly over the pressure taps results in a local reduction in
measured pressure that does not correspond to the overall bulk pressure behaviour. Second,
in severe swirl instances, the centripetal force of the fluid pushes the fluid energy to the pipe
wall, resulting in a deviation from fully developed energy distribution. Third, the interaction of
the swirl with the measurement primary element itself can cause unpredictable fluid dynamic
behaviour. Severe pressure drops force the swirl to behave in unexpected ways, further
complicating the flow measurement.

Figure 12: Visualization of velocity contours as flow passes through an orifice meter.

Figure 13: Visualization of velocity trajectories as flow passes through an orifice meter.

Figure 14: Visualization of velocity trajectories as flow with heavy swirl passes
through an orifice meter.
 14

Velocity Velocity Swirl P1 P2 Density K deltaP

(fps) (m/s) (Degrees) (kPa) (kPa) (ρ) (kPa)

50 15.24 0 5598181 5458170 36.84 32.73 140011

50 15.24 20 5591067 5471339 36.74 28.06 119728

50 15.24 45 5560318 5497955 36.68 14.64 62363

Figure 15: Showing how the addition of swirl can change the output of an orifice meter due to the
local effects of the velocity at the pressure taps.

Ultrasonic Meters

Ultrasonic Meters (USM) function by bouncing an ultrasonic sound pulse from one side of the
pipe to the other. This is done along a predetermined path, and then repeated on the reverse
side of the path. The path is at an angle so the pulse will either be travelling with or against the
axial bulk velocity flow. The time for the sound pulse to travel from one side of the pipe to the
other is then measured. As one path is being carried with the bulk fluid flow, it will have a lower
transit time across the pipe, where the path that is opposing the fluid flow will have a higher
transit time. The difference in these times is used to calculate the average velocity measured
across that transit path.

Problems exist with this design, however, as the only variable that an ultrasonic meter is able to
consider is this transit time. There are no other inputs that can be analyzed. This means that the
velocity flow profile that the path is measuring has to be assumed to be symmetrical and evenly
distributed across the pipe. The USM path has no way to determine what the distribution of
velocity is in the pipe. This also causes issues when dealing with swirl. A single USM path has
no way to distinguish if the velocity influencing its travel time is due to the axial bulk velocity,
or if it is a transverse component due to swirl.
 15

To combat this, modern Ultrasonic Meters are built with multiple signal paths. These are
positioned at different distances across the pipe to help generate an overall picture of the pipe
flow by combining the results from each path. These positions are chosen based on selecting
areas of the pipe where a majority of the flow volume and swirl may be found, to make sure the
meter is accessing as much critical flow information as possible. They may also be in locations
that are isolated from shifts in Reynolds numbers, to help reduce meter error due to changes in
pressure or viscosity. The paths may also be oriented in different planes to help combat swirl or
any other issues that would generate transverse velocities across the pipe.

The data from these paths is combined to generate a complete picture of what is occurring in the
flow. Based on an understanding of the shape of the velocity flow profile at the path location,
each path can be properly weighted for its contribution to the final average calculation of flow
velocity. Each signal path is essentially one slice of the final velocity profile vehicle. The more
slices available, the higher the resolution of the final picture.

The figures on the next page help to illustrate this relationship. Figure 16 shows the average
velocity profile that is present at the center of a sample flowing pipe, while Figure 17 shows
the profiles that a two path Ultrasonic Meter would see at this same location. Once we do a
weighting correction to adjust the outputted values due to their location, we find that USM Path
1 varies from the bulk velocity flow by -5.99%, while USM Path 2 varies by 7.62%. Taking the
path at the center of the pipe only results in a 1.2% deviation.

The problem is that if we cannot observe the state of the flow profiles, and we do not know what
the actual bulk mass flow rate and velocity is for this corresponding pipe flow, then how do we
know which of these profiles is the correct representation of what is actually occurring in the
pipe?
 16

Figure 16: Average velocity profile at USM location, taken from the center of pipe.

Figure 17: Velocity flow profiles for a theoretical two path Ultrasonic Meter.
 17

Liquids and Pressure Drop

Tube bundles are still used in modern applications due to the assumption that perforated plate
flow conditioners should not be used in liquids. This incorrect assumption mainly came about
due to the fact that a majority of flow conditioning testing and published data is from high
pressure, natural gas applications. Testing has been done with natural gas for convenience and
financial feasibility. In the case of the tube bundles, the pressure drop data being used was
extrapolated from the gas testing data, thus the significant pressure drop of the tube bundles in
liquid applications was not being properly represented in published results. Tube bundles were
already being used in liquid applications for decades before modern plate flow conditioners had
been invented, thus the suitability of tube bundles for liquid applications was never questioned.
In the case of plate style flow conditioners, a large amount of the testing was also done in gas
applications, therefore this led to a belief that they were not developed for liquids or were
suitable for those circumstances.

Flow conditioner behaviour in gas is transferable to liquids, however pressure drop performance
must be acknowledged. This is because the only functional difference from a measurement
perspective between a gas and a liquid is in the viscosity, molecular composition, and density of
the fluid. Fluid phase is only a concern when dealing with thermodynamics, heat capacity, and
enthalpy of a fluid. Fluid dynamics and flow measurement are only concerned with Reynolds
number and viscosity. As long as the fluid is single phase (multiphase measurement in general
is a very complex and problematic concept) it will always behave the same, independent of
whether it is a gas or liquid, oil, water, air, or natural gas. Pressure loss due to piping and flow
conditioners will indeed change, but this is simply due to the different densities and viscosities
being used. As a result, properly designed flow conditioners can be used in all fluid types, not
just natural gas.
 18

Figure 18: Fully developed flow profiles downstream of a perforated plate flow conditioner in
laminar flow, Re ~ 800.

Figure 19: Fully developed flow profiles downstream of a perforated plate flow conditioner in
turbulent flow, Re ~ 30,000,000.
 19

Figure 20: Showing the effectiveness of a plate based flow conditioner at eliminating swirl
across a range of Reynolds numbers.

Tube bundle use in liquids has also been more predominant due to misunderstandings of the
pressure drop effects across a range of Reynolds numbers. It is typically accepted that the
pressure loss coefficient of straightening vanes and tube bundles is roughly 0.75 – 1.25. This
is roughly half of the commonly accepted pressure loss coefficients that are used for perforated
plate flow conditioners, which is commonly found to be around 2.0.

The problem with these values is that the true pressure loss coefficients have yet to be
determined through experimental testing. Most of the lab work that was done to study
these values was done in high-pressure natural gas; a high Reynolds number, low viscosity
application. The resultant data was then extrapolated to not only cover the entire turbulent range
of fluids, but to cover the laminar region as well. This is incorrect as changing the Reynolds
number, and specifically the viscosity, can have extremely significant effects on the pressure loss
coefficient and the pressure drop.

The other problem is the assumption that a tube bundle must have a lower pressure drop,
because it has a higher open area than a perforated plate flow conditioner. This is only partially
correct.
 20

19-Tube Bundle Perforated Plate

Wall Thickness (Inches) 0.125 0.250 0.298
Pipe Diameter (Inches) 11.938 11.938 11.938 11.938
Pipe Area (Inches^2) 111.932 111.932 111.932 111.932
Flow Area (Inches^2) 94.957 79.848 74.493 53.450
Porosity (%) 84.84% 71.34% 66.55% 47.75%

Figure 21: Showing the difference in open flow area with various AGA-3 19 tube bundles and a
perforated plate flow conditioner.

Figure 22: A CAD cutaway of an AGA3 19 tube bundle, illustrating the open flow area.

Figure 23: A CAD cutaway of a perforated plate flow conditioner, illustrating the
open flow area.
 21

19-Tube Bundle Perforated Plate

Wall Thickness (Inches) 0.250
Pipe Diameter (Inches) 11.938 11.938
Pipe Area (Inches^2) 111.932 111.932
Flow Area (Inches^2) 79.848 53.450
Porosity (%) 71.34% 47.75%
Published K Factor Per ISO 5167 0.75 2
CFD K Factor, Re 30,000,000 1.39 2.020

Figure 24: Illustrating the different pressure drop characteristics of a 19 tube bundle vs a
perforated plate flow conditioner.

The lower flow area accounts for a lower pressure drop in fluids with extremely low viscosity.
This is because the wall friction becomes a negligible portion of the overall pressure drop. In
fluids with higher viscosities, it cannot be ignored.

hfriction = f L V 2

d 2g
Figure 25: The relationship between piping friction losses, duct diameter, duct length and fluid/
pipe friction factor.
 22

Figure 26: The Moody Diagram, used for calculating friction factor using Reynolds Number and
the pipe wall roughness.

The pressure losses due to wall friction can be calculated using the hfriction equation. This
equates the pressure head loss as a function of the pipe friction factor. Friction factor is
calculated using the moody diagram and the Reynolds number of the flow. The Reynolds
number takes into account the density, velocity, and viscosity of the flow. Changing the
viscosity will have the largest effect on the Reynolds number, as the difference in viscosity
values between different fluids can change by orders of magnitude.

The moody diagram above shows that as the Reynolds number is decreased, due to increasing
the viscosity, the friction factor also increases. Once the fluid reaches the laminar regime, the
friction factor suddenly increases even more quickly.
 23

The calculation of head loss due to friction primarily deals with closed ducts of a fixed diameter
and fixed length. This becomes an issue when dealing with different types of flow conditioners.
Specifically, due to the way a tube bundle is constructed, tube bundles have a significantly
higher surface area than perforated plates (because the flow is passing over both the inside and
the outside surfaces of the tubes). If we consider per unit length (1”), a perforated plate flow
conditioner has a surface area of roughly 128 square inches, while an AGA-3 19 tube bundle has
a surface area of roughly 287 inches.

This surface area problem is compounded by the fact that tube bundles are significantly longer
that plate style flow conditioners. This was purposely designed so that the tube bundle device
could compensate for the reduced flow conditioning effectiveness from the lower pressure drop.
Typical plate style flow conditioners have a thickness that is only 10 – 20% of the pipe inside
diameter. Tube bundles are 2 to 3 times the length of the pipe inside diameter. As a result, this
gives tube bundles an additional 20 – 30 times increase in the wetted flow area. In scenarios
where the wall roughness becomes critical, this results in a significant increase in the pressure
drop due to the large increase in surface friction.

The data shown in Figure 27 was determined using the best-case scenario when it comes to
a tube bundle, as the AGA-3 has large allowances on the final dimensions and quality of the
finished product. It was assumed that the pipe’s exterior was as smooth as the interior, and
that no welds were present. It was also assumed that there were no blockages in the exterior
passages, that the holes were fully chamfered on the inlet and outlet, and that there was no
upstream flange plate. The following changes could significantly affect the resulting pressure
drop:

1. AGA-3 has allowances on the range of the wall thickness of the tubes. This study used a
thickness of 2% of pipe ID. AGA-3 allows for thicknesses of up to 2.5% to be used.
Thicker tube walls would significantly reduce the flowing area, thus increasing the initial
pressure drop in high Reynolds number fluids. The physical welding of the tubes together
can increase the friction factor and resultant pressure losses on the exterior of the tubes.
 24

2. Blocking the inlets of the exterior passages will reduce the overall exterior surface area,
but increase the initial entrance losses. AGA-3 states that in pipe sizes of 4” or less, the
exterior passages can be sealed. Doing so would reduce the surface area by about 65%, but
also reduce the inlet open area to 43.2% of the total pipe area (compared to the open area of
47.7% for the perforated plate). So, while the friction losses due to the surface area would
be reduced, the initial pressure drop will increase across the entire Reynolds number range.

This study was performed using a tube bundle that has a length twice that of the pipe’s nominal
pipe size (NPS). AGA-3 states that this applies to tube bundles in pipes larger than NPS 4”.
AGA-3 states that tube bundles in NPS 2” pipe will have a length of 3 x NPS, which results in a
surface area increase of 50%. Tube bundles in pipes greater than NPS 2” and less than/equal to
NPS 4” will have a length of 2.5 x NPS, which gives a surface area increase of only 25%.

In a 2” tube bundle, with the exterior outlet passages closed, the open area is 43.2%, with a
flowing surface area that is still 19 times greater than that of a perforated plate. In a 3” tube
bundle, the open area is the same but the flowing surface area is 16 times greater than that of the
perforated plate. In these circumstances, the tube bundles will still have a significant amount of
pipe wall friction, but now they must also deal with much higher entrance losses as well.

Figure 27: Comparing the pressure drop of an AGA-3 19 Tube Bundle to a perforated plate
flow conditioner across a wide range of Reynolds numbers.
 25

Conclusion

Proper flow conditioning is highly recommended for all measurement systems, especially those
with differential pressure and ultrasonic flow meters. To ensure that measurement accuracy
and repeatability is maximized, a flow conditioner should always be used. A properly designed
perforated plate flow conditioner was not developed to benefit a particular meter type. Instead,
it is designed to improve the fundamental fluid dynamics of the pipe flow itself. The plate style
flow conditioner is designed to create a fully developed flow that is free of swirl, distortions, or
any other flow asymmetry. The flow conditioner is installed upstream from a meter, and the flow
conditioner’s performance is completely independent from the meter.

It is also important to note that a proper flow conditioner is effective in a wide range of fluids
and Reynolds numbers. Flow conditioning is not typically fluid specific, and perforated plate
flow conditioners can be used in many fluids other than natural gas. It was previously thought
that tube bundles should be used in liquid applications due to the lower pressure drop when
compared to perforated plate flow conditioners. This belief came about because a majority of the
current pressure drop data used to compare the performance of the devices was collected from
testing in high-pressure natural gas and air. The behaviour of these devices changes significantly
as the wall roughness and flow viscosity of the application changes. Recent studies have shown
that in viscous fluids, tube bundles can have significantly higher pressure drops than perforated
plate flow conditioners.
 26

While orifice meters or other differential devices and ultrasonic flow meters use different
techniques for measuring fluid flow, they have a significant concept in common: they are both
volumetric devices. Both meters are simply measuring the volumetric state of the pipe flow.
They also consider two significant variables: flow profile quality and swirl. For both metering
types, accuracy is directly linked to the quality of the upstream flow. A fully developed flow
profile will always maximize meter accuracy, as does flow with little to no swirl.

Combining a fully developed flow profile with minimal swirl helps guarantee that the meter is
in the best possible circumstances for repeatable, accurate measurement. A perforated plate
flow conditioner ensures that these are the conditions that the flow meter will always see,
effectively cutting it off from all upstream installation effects that would cause unpredictable
errors to an unprotected meter. The plate style flow conditioner will eliminate the swirl in the
flow and restore the fully developed flow profile, helping the meter to ignore the severity of the
disturbances upstream.
 27
References

Y. Cengel. J. Cimbala. “Flow In Pipes” in Fluid Mechanics: Fundamentals and Applications, 1st
Edition. New York, NY: McGraw-Hill, 2006, pp. 321-386

F. White. “Viscous Flow In Ducts” in Fluid Mechanics, 5th Edition. New York, NY: McGraw-
Hill, 2003, pp. 343-426

“Part 2: Specification and Installation Requirements” in AGA Report No. 3, 4th Edition, 2nd
Printing. Washington, DC: American Gas Association, 2003, pp 16-17.

AGA Report No 3-2:2003, “Specification and Installation Requirements”

ISO 5167-1:2002, “General Principles and Requirements”

10653 - 46 Street S.E.

Calgary, Alberta
Canada T2C 5C2
Phone: 403-236-4480
Fax: 403-236-0019
Email: [email protected]