The existing data has potential to provide various types of information. This paper reports on:

• Overall average incident rates - useful for benchmarking Australian industry performance against other countries
• Average incident rates for various location classes - a benchmark for quantitative risk assessment studies (QRA)

Scope & validity of data

The database currently contains about 120 Australian incidents which have been recorded since 1965. Incidents reported comprise coating damage, steel damage, leaks and ruptures. Causes of incidents include external interference, corrosion, ground movement, construction defects, material defects and lightning, although there are very few incidents in the latter categories. The data used for determining average incident rates has been restricted to events since 1985.

There is some doubt about the completeness of the database. Information is being sought from major pipeline operators to confirm and update records for about the last 20 years. So far confirmation has been received from operators representing about 14,000 km out of a total of 27,500 km of pipelines in Australia. Less than 10 incidents were new to the database. It is reasonable to expect that the remaining unreported incidents will be a small fraction of those already recorded, and that their addition will not change the overall conclusions that can be drawn from the existing data.

Total incidents, causes & consequences

Figure 1 shows the total incidents recorded in the database, grouped by five-year periods. It is likely that some as yet unreported incidents will be added to the database. However even if a couple of dozen new incidents are added they will not dramatically change the shape of the graph.

Causes of incidents are as shown in Figure 2. As is generally well known, external interference is by far the most common cause of pipeline damage and accounts for 76 per cent of all incidents. Corrosion is the next most common cause, but the large number of corrosion incidents in the years up to 1970 were all on a single pipeline and hence are not representative. Remaining causes (and corrosion since 1970) have resulted in very few incidents.

The most common consequence of a pipeline incident is deformation of the pipe steel (which includes scratches, gouges and dents), accounting for 64 per cent of incidents. There have been been six ruptures (6 per cent) and 27 leaks (25 per cent), although 1 rupture and 10 leaks were from the pre-1970 period on the single unrepresentative pipeline referred to earlier.

The consequences do not include deaths or injuries – there have been none recorded in connection with pipeline operations.

Average incident rates & overseas comparison

The graphs of total incidents took no account of the substantial growth in the Australian pipeline industry over the last 40 years. Figure 4 repeats the data for total incidents and also shows the growth in cumulative length of pipeline in Australia, which is now about 27,500 km compared to only about 2,600 km in 1970.

Combining this data provides a more meaningful representation of pipeline incident data, expressed as incidents per 1,000 km per year (see figure 5).

The apparent improvement in the safety performance of Australian pipelines may not be quite as clear-cut as the graph indicates, since the high rate pre-1970 is anomalous as noted previously, and it is likely that the incident rate in more recent years will increase somewhat when unreported events are added. Nevertheless, it is most unlikely that unreported events would be sufficient to change the overall downward trend.

In order to compare these incident rates with overseas data it is convenient to reduce them to a single number. Because of the variability of the data it was judged reasonable to take the average over the last four half-decades (ie. 1986 onwards). The graphs compare Australian safety performance with overseas data from Ref 1. Note that all overseas data is for loss of containment (LOC) incidents only (see figure 6).

The Australian incident rate is very much lower than overseas, by about an order of magnitude. The difference is probably due to a combination of remoteness (few threats), relatively young pipelines (modern design and coatings) and good management.

This comparison also provides a clear indication that it is not valid to apply overseas data to Australian risk and safety studies.

Location class

It is obvious that higher rate of incidents can be expected in more developed areas, and therefore that the overall average incident rates discussed so far are not the whole story.

To break down the incident rates by location type requires not only data on the location class for each incident but also the total length of each location class in Australia. However this information is not readily available so the analysis and conclusions here are based largely on estimates. Only two location classes have been considered: R1 representing undeveloped areas and combined R2/T1 representing semi-rural and suburban areas (because the borderline between R2 and T1 areas can be hard to estimate without detailed maps).

R2/T1 areas contain only about 10 per cent of total Australian pipeline length. However they account for 60 per cent of pipeline incidents so clearly have a much higher incident rate. For external interference events only the average incident rates broken down by location class are:

• R1 0.05 per 1000 km-yr
• R2/T1 0.48 per 1000 km-yr

The difference in incident rate between rural and more developed areas is a factor of about 10. It is quite clear that any use of the incident data for predictive risk studies must take account of location class.

Comparison with QRA studies

Quantitative risk assessment studies have been required for some pipelines in Australia. Two such studies for recent pipelines have predicted failure rates as follows (per 1000 km-yr):

• QRA 1, for rupture: 0.2
• QRA 2, for penetration: 0.2

Both were for pipelines in R1 areas, and were based on UK and European data. For comparison, the following incident rates can be extracted from the Australian database for R1 areas over the period 1986-2004 (per 1000 km-yr):

• All Incidents 0.05
• Loss of Containment 0.017
  • Leak 0.014
  • Rupture 0.003

The comparison shows that QRA studies based on overseas data estimate failure rates for Australian R1 conditions that are far too high, by one to two orders of magnitude. QRA studies for R2/T1 conditions may overestimate failure rates by a factor of up to 10.

Conclusions

Australian pipelines have a much better safety record than Europe and the US. There have been no recorded injuries or fatalities and the incident rates are an order of magnitude lower than overseas.

External interference is the dominant cause of pipeline incidents in Australia, and therefore provides the greatest scope for further improvement in safety performance. This reinforces the importance of conscientious risk assessment and implementation of protective measures. The fact that very few incidents are due to corrosion or construction and material defects shows that these aspects of the Australian pipeline industry are well managed.

Incident rates in developed areas (R2 and T1) are an order of magnitude higher than in remote rural areas (R1).

Quantitative risk assessments based on overseas data produce failure rates that are one to two orders of magnitude higher than can be justified by the Australian experience. Design based on such overly conservative results may incur additional costs without any significant reduction in risk, and is therefore not in the interests of the industry or community.

The numerical results presented here may change as further information is received. However these conclusions are based on such large differences in numerical values that they are most unlikely to be affected by refinement of the data.

References

1. W. Guijt, Analyses of incident data show US, European pipelines becoming safer, Oil & Gas Journal, Jan 26 2004

Acknowledgements

Craig Bonar (Pipelines Manager, GasNet) was co-author of the original paper presented at the APIA Convention, and as chairman of POG has been largely responsible for facilitating access to the existing database and for upgrading the database to improve future incident recording.