Pipeline commissioning: purging air with gas

Since nitrogen is used in process industries to displace air from process vessels and complex piping prior to introducing a flammable gas, people in the pipeline industry thought more about the risk and it became more common to separate the air from the gas using a large slug of an inert gas such as nitrogen.

This introduced significant complexity into the operation, requiring large tankers of liquid nitrogen, vaporising equipment and fuel to be transported to very remote locations.

Mechanical separation (a pig), is an alternative, but rarely used because the velocity during a purge is usually too high.

A few pipelines such as the Moomba to Wilton pipeline were left filled with water following construction.

The water was displaced using gas to propel a relatively complex pig train to separate the water and gas.

The effort to dry the pipeline was considerable and this approach has rarely been used since.

When it became necessary to introduce gas into the 800 km long Eastern Gas Pipeline (EGP) a range of options were considered, including; nitrogen slugs, a separating pig train and a return to the early practice of simply displacing the air with gas.

The review concluded that the quickest, simplest and safest approach was to displace the air with gas.

The operation was successful, and subsequently a similar approach was used on the Tasmanian Gas Pipeline and the SEA Gas Pipeline.

More recently, pipeline commissioning has reverted to using an inert gas slug to separate the air and gas.

While this approach may have been appropriate for those pipelines, there are concerns that the simple gas purge approach is not considered because of a basic lack of understanding of the process, and indeed the potential ignition risk.

This article uses the EGP purge process as an example that purging air without the complexity of an inert gas slug, or mechanical separation of air and gas with a pig, is a safe and predictable process.

AGA Purging Principles and Practice

Research undertaken by the Gas Research Institute (GRI) was published in 1997 (GRI-97/0104 Pipeline Purging Principles and Practices Research) together with a software application, PURGE.EXE, designed to model the purge process.

The EGP commissioning team used both this report and PURGE.EXE in evaluating options for purging air from the pipeline.

Much of the GRI research was incorporated into the 2001 revision of the AGA Purging Principles and Practice.

The research was undertaken by South West Research Institute (SWRI) (this report is still available from the Gas Technology Institute, although unfortunately the PURGE.EXE application, developed for Windows XP, appears to no longer be available).

Ignition risk

The risk that the gas-air mixture at the purge interface ignites (explosively) within the pipe is often quoted as the reason for separating the gas and air with an inert gas slug.

In new, steel pipelines this risk is vanishingly small because there is no ignition source in the pipeline.

There may be a small risk in an existing pipeline containing pyrophoric dust.

The SWRI research concluded that in the unlikely event of interface ignition, the maximum overpressure in the pipe is approximately 10 times the pressure in the pipe at the time of ignition – for this reason the AGA Purging Principles document recommends limiting the maximum pressure in the pipe during a gas purge to 100 psig (thus limiting the maximum pressure should combustion occur to ~1,000 psig, the typical maximum pipeline pressure in North America at the date of the research).

This can be readily confirmed by calculating the volume change when methane is burnt, using the equation
CH4 + 2O2 = CO2 + 2H2O.

The plume ignition risk is also low, provided simple safety procedures are followed.

Moreover, the high plume velocity at the vent provides momentum to propel the plume above most credible ignition sources (like the vent from a pipeline blowdown).

Hydraulics – the length of the mixed air-gas interface

There are two distinct flow regimes within a pipeline – laminar flow, which is characterised by a parabolic velocity profile across the pipeline cross section, and turbulent flow, which is characterised by a constant velocity profile across the pipe cross section, except for a thin boundary layer at the pipe wall where the velocity transitions to zero at the wall.

In laminar flow, the fluid in the central region moves at a higher velocity than that in the outer region, causing continuous longitudinal mixing of the pipe contents.

In turbulent flow the essentially constant velocity profile prevents longitudinal mixing.

There is a minor mixing at the batch interface resulting from a combination of dispersion across the interface, and from the velocity change in the boundary layer.

The interface volume can be calculated from the pipeline flow and product properties, while the constant velocity profile enables the interface position to be accurately tracked from the flow rate.

This phenomena is used in many multi-product pipelines to separate batches of different products, eliminating the need to provide positive separation using, for example, batching pigs.

In a multi-product pipeline, the batches are usually sequenced so that mixed interface fluid can be blended into the following uncontaminated batch without degrading it.

Various publications discuss prediction of the interface length in liquids pipelines – these provide useful background to understanding the phenomena.

Similar concepts apply to the interface length calculation for gases, although the velocity (Reynolds number) variation during the purge increase the calculation complexity.

These methods incorporate an estimate of the interphase dispersion along the travel length.

Eastern Gas Pipeline Purge example

Gas was used to purge the Eastern Gas Pipeline (DN 450), a continuous operation between the inlet compressor station at Longford and an intermediate station at Kembla Grange, a distance of 713 km.

Once gas was received, the Kembla Grange vent isolation valve was closed and the pipeline packing operation followed.

The purge operation was designed using the principles and recommendations from GRI-97/0104 Pipeline Purging Principles and Practices Research.

Additional modelling was undertaken using WINFLOW/WINTRAN hydraulic modelling software, licensed to Duke Energy.

The initial study concluded that the additional complexity introduced using a nitrogen slug to separate the air and gas was not justified by the ignition or operational risk – and neither was the use of a batching pig train.

Consequently, design effort concentrated on developing a gas purge procedure.

For purging the pipeline using gas, the key principles are:

The purge gas flow must be continuous:

1. Continuous purge flow is essential when purging with gas – if the gas flow is lost, the mixed gas-air interface volume increases rapidly, primarily through buoyancy driven mixing
(gas density is approximately half that of air).

For this purge operation, gas supply was assessed as being secure since it was drawn from the Longford Gas Plant delivery pipeline used to supply the Victorian market. Had there been a concern about the supply reliability, the purge design may have been changed to use either positive separation with a pig, or a slug of inert gas.

2. The maximum pressure should not exceed 10 per cent of the pipeline maximum allowable operation pressure (MAOP).

Simulation studies were undertaken to assess the purge performance using controls recommended from the research, namely:

  • Purge at constant inlet pressure; and,
  • Purge at constant inlet flow.

The constant pressure approach was discarded because the gas supplier advised that the variable flow required through the purge operation was not desirable, and it was considered more difficult to control than simply maintaining a constant flow.

It was decided to vent air at the Mila and Oallen scraper stations (kP291 and kP564 respectively).

This limited the volume of gas mixture discharged at Kembla Grange, while limiting the pipeline pressure and allowing physical tracking of the interface.

The purge process commenced with the pipeline isolation valves at Kembla Grange, Oallen and Mila stations closed, and the pipeline vent valve at Mila open.

Gas was introduced at Longford, and the flow adjusted to 15,000 scm/h (and adjusted as required).

Commissioning staff were located at the Mila vent, equipped with gas detectors.

Once detected, the gas concentration in the vent was monitored until the mixed interface was discharged.

The vent valve was closed and pipeline isolation valve opened, together with the vent valve at Oallen.

The process was repeated until gas was received at Kembla Grange.

The pipeline purge was completed in 32.5 hours.

Once gas was received, the vent valve was closed and gas flow increased to pack the pipeline with the linepack needed to sustain the intended flow rate.

The application PURGE.EXE developed by SWRI is limited to a pipeline length of 999,999 feet, requiring it to be used iteratively.

Figures 2 to 6 show the output from the purge simulation at an inlet flow of 15,000 scm/h for the 291 km pipeline section between Longford and Mila.

The calculation predicts the mixed gas volume is approximately 5000 ft3 and approximately 5100 ft long, while the prediction shows that the pipeline velocity at the time that the mixed volume reaches the vent is approximately 40 ft/s.

From this, the time for the mixed gas-air volume (from 100 per cent air to 100 per cent gas) to be discharged through the pipeline vent is approximately 127 seconds (2.1 minutes).

Figure 7 presents pressure data taken from supervisory control and data acquisition (SCADA) during the purge, together with the prediction from the SWRI application PURGE.EXE.

The following notes:

  • The SCADA pressure measurement at Mila and Oallen did not report pressures higher than 50 kPa (a commissioning error).
  • The Mila pressure shows a relatively large pressure drop after gas arrival, once the vent valve is closed and the mainline valve at that site is opened. The effect is reflected at each other measurement point. The effect was not seen at Oallen, probably because of this instrument error.
  • The interface arrival time at Mila predicted by the PURGE application (Figure 5) is 700 minutes (11.67 hours). Gas arrived at Mila 11.62 hours after the purge commencing.
  • The PURGE application requires a reasonable estimate of the pipeline roughness to reasonably reflect the actual flow. The EGP was a new pipeline, and the construction team applied considerable effort in the drying and cleaning phase to burnish the inner pipe surface. The predicted arrival time assumed a pipe roughness of 20 microns.
  • Gas concentration was assessed using portable detectors set to monitor nitrogen (not methane). Gas arrival was reported at the time that the nitrogen concentration started to drop, and 100 per cent gas was reported at the time that the nitrogen concentration was essentially zero. The instruments did not provide data logging capability, so the recorded data reflects human and sampling errors.
  • At Mila, 100 per cent gas was reported eight minutes after it was first detected (longer than the 2.1 minutes predicted from the PURGE application). Records at Oallen indicate that the time between first and 100 per cent gas was 15 minutes (unreasonably high, and probably reflects the time taken to complete the valve changeover).

Conclusion – EGP Purge

The EGP purge showed conclusively that, provided a constant gas flow can be guaranteed for the duration of the purge, purging air from a pipeline using gas is simple, predictable and, above all, safe.

It showed that the model PURGE.EXE provided an accurate prediction of the hydraulics during the purge process, and a good estimate of the mixed air-gas volume to be discharged.

This methodology was subsequently used successfully to purge and pack the onshore Tasmanian Gas Pipeline sections and the SEA Gas pipeline.

Purge Planning – Future Pipelines

As mentioned earlier, the application PURGE.EXE is no longer available with the SWRI report, and GRI has advised that there is no intention to revise it for current desktop operating systems.

(Note: The writer is attempting to determine whether SWRI is prepared to update the application).

Transient gas hydraulics applications can be used to predict the interface arrival time provided the model used is carefully developed with an understanding of the computational methodology of the application.

Table 1 illustrates predictions from the application used by the writer (FLOWTRAN) for the Longford-Mila pipe section.

Two purge options are considered to assess the application’s capacity to predict the arrival of the first gas, and purge of the last of the nitrogen:

  • The pipe was filled with a simulated air mixture at atmospheric pressure and purged with a typical Longford gas.
  • The pipe was filled with nitrogen, and purged with methane.

The calculated first gas arrival shown in the table is the time that the nitrogen concentration falls by 2 per cent from the pre-purge concentration, and the “˜all gas’ arrival is the time that the calculated nitrogen concentration falls to 2 per cent.

The purge was also tracked using the “˜pig tracking’ option in the application, assuming zero slip.

The table shows:

  • The pig tracking option provides a good prediction of the arrival of the mid-point in the air-gas mixture when a simulated air mixture is purged with a typical gas mixture. Purging nitrogen with methane under-predicts the mixture arrival time.
  • The model significantly over-predicts the mixture volume (approximately 100 minutes, compared with 8-10 minutes). Other transient hydraulic models may calculate the mixture differently and give different results.
  • More calculation nodes did not significantly change the predicted arrival time, except that the model with 308 nodes did not calculate the “˜pig’ arrival time.


Purging a new pipeline (or refilling a section of an existing pipeline) using gas without an inert slug or mechanical separation, is a safe and predictable operation which reduces the cost and complexity of the activity.

Transient hydraulic models can be used with care to reasonably predict the arrival time of the gas-air mixture, although they may not provide a reliable estimate of the mixture volume that must be purged.

Designers using these tools should carefully evaluate the model predictions.

Those who have access to an application that can execute Windows XP executable files can also use PURGE.EXE to support the prediction.


1. The fully turbulent (plug) flow for this pipe requires a velocity of approximately 1 m/s.
2. The predicted pipeline velocity at the end of the DN 450 pipe section (40 ft/s) suggests that the plume velocity from the DN150 vent will be in the order of 100 m/s – providing significant momentum to propel the mixed air-gas interface well beyond ignition sources.

Send this to a friend