Final (?) post in the series on ALARP …

Estimating failure frequency is part of AS 2885 risk evaluation anyway, and finding some sort of quantitative estimate of frequency is also a necessary part of the maximum justified spend concept.

The AS 2885 risk matrix is entirely qualitative – it just has text descriptions for frequencies ranging from Frequent (expected to occur several times) down to Hypothetical (theoretically possible but has never occurred on a similar pipeline). Sometimes the qualitative frequency descriptions work OK, but for a long time I’ve been supplementing them with some order-of-magnitude numerical frequency guidelines. My version of the frequency table looks like this:

I didn’t invent these out of thin air. The 1997 revision of AS 2885 had a companion guideline document known as SAA HB 105 (Guide to pipeline risk assessment in accordance with AS 2885.1). It was superseded by AS 2885.1-2007 and the numerical frequency guidelines it contained didn’t survive the transition. But I’ve kept using this adaptation of them anyway (adapted because the 1997 risk matrix had six frequency steps, not five as we now have).

A couple of clarifications:

- The frequencies apply over what I call the “exposure period” – either the life of the pipeline, or the duration of the project if the safety management study is for a one-off activity such as hot tapping or building a road over an existing pipeline. They are not annual frequencies
- To be pedantically correct all of the numerical values should be considered as frequencies, even though the terminology in the table above refers to both number of events and probability.

But are these numerical guidelines actually helpful? Personally I find it quite useful, in a subjective way, to put approximate numbers on the Frequent and Occasional ratings, and to some extent the Unlikely rating. Below that it becomes hard to develop an intuitive appreciation of what the numbers really mean. In this range historical incident statistics start to become useful though.

There haven’t been enough pipeline incidents in Australia to provide a basis for conventional quantitative risk assessment. But there is enough data to provide approximate overall incident rates that can form a basis for estimates that feed into the risk matrix or the maximum justified spend calculation.

Since 2001, the average rate of pipeline damage due to external interference has been:

- R1 (remote rural) areas 0.1 incidents per 1000 km.yr
- T1 (suburban) areas 0.4 incidents per 1000 km.yr

(Not enough data to obtain meaningful rates for R2 and T2 location classes.)

Only about one quarter of damage incidents result in a loss of containment, and only a quarter of those are ruptures rather than leaks. But these last figures need to be used with caution because the sample size is getting *really* small – three leaks and just one rupture.

It’s possible to get an order-of-magnitude estimate of failure frequency by judiciously adjusting these rates depending on the nature of the threat under evaluation and the protective measures in place. Large element of judgement here! But precision doesn’t matter – order-of-magnitude estimates are all that’s necessary for both the risk matrix and the maximum justified spend.

It’s interesting to bring up the HB105 document. Even this year I have had people asking me about it.

At the time it was released, it was a useful companion to the AS2885.1 risk assessment process. I think even today it still serves as a useful guide – provided the Risk Assessment meets current code requirements.

Most of the bits of HB105 that were still relevant and useful got transferred into various appendices in AS 2885.1-2007, with the notable exception of the frequency guidelines.

Pingback: “Hypothetical” frequency | Pipelines OZ

Pingback: Risk matrix selection | Pipelines OZ