Page 39 - Gi March 2019 - Flipbook

igem news – yppc

TABLE 1 CRITICALITY FACTORS

generate projected failure curves from of failure, I looked at our pipeline as a Applications of models
eight years of data. whole to predict total leakages on our Current
To build my model, I used the network each year, as per Figure 3. By So far, the implementation of this
Kaplan-Meier estimator of likelihood combining this model with criticality, risk model has helped Powerco to
of survival. This model is traditionally we can see the likely yearly impact of prioritise renewal projects, addressing
used for integer populations, so leakages on our business. certain asset classes as being faulty, for
fits our individual assets well. example, polyethylene pipe installed
Additionally, I stretched the model Future development prior to 1985 in Figure 4. These projects
to fit our linear pipe assets by So far the risk model is only designed tend to have several groups of assets
assuming each 10m length of pipe for distribution pipelines. The next requiring renewal. The risk model
could be considered as one individual step is to apply this model to all other allows prioritisation within these
asset. This is a very conservative Powerco gas distribution assets. high-risk asset types. We now prioritise
length to consider as leaks tend to The current models will be validated based on which areas are likely to
be pinhole or small cracks. Even so, yearly, comparing the actual number cause most harm, instead of simply
with this conservative model our of failures with that predicted by the prioritising streets that have had high
oldest polyethylene pipe only has a models, then updating the models with historical leakage rates. This moves
probability of failure of 3.5 per cent more recent data and/or changes in our planned renewal projects from a
per year. Powerco considers this very reasoning. I also plan to collaborate reactive “chasing-the-leaks” approach
low relative to the regulatory expected with other distribution businesses to a more preventative method of asset
lifetimes and thus indicates our to increase the data at my disposal, renewal by targeting areas that appear
pipelines are in good condition. particularly where the Powerco data most likely to leak due to similarities
Next, with the finding that individual appears incomplete, to make the to assets that have leaked in the past.
pipeline assets have minimal likelihood likelihood models more robust. We expect that this new method will

39

IGEMNews_YPPC.indd 2 14/02/2019 14:27