Learnings

Outcomes

Extensive data collation and augmentation activities were completed to demonstrate the process by  which corrosion modelling of both buried and non-buried pipework components could be completed to  support tactical and strategic planning activities at an above ground installation (AGI). 
 
A minimum data specification for corrosion modelling was developed and validated through the use of  statistical exploratory analysis and machine learning techniques, to determine which information was of  greatest use to understand corrosion defect initiation and growth. 
 
Different parameters have been shown to be of some importance for defect initiation and growth but  these include, as might be expected, a mix of intrinsic factors (e.g. material grade, diameter, age),  environmental factors (including pit location) and operational factors (e.g. pressure class, plant  runtime/cycling). Some site-specific factors were also found to be important, in particular geospatial  location. This was included to recognise the importance of capturing historical defect clusters that were  not fully explained by other operational or environmental factors included in the model. 
 
The limitations of the current site model and asset register were also explored and partly overcome by  geospatial analysis and manual augmentation through inspection of site drawings. Model datasets were  augmented with publicly available borehole data to estimate soil corrosivity, and data-driven modelling  was supplemented with the use of selected industry norms and empirical data derived from the literature. 

The completion of the current development of a more comprehensive site model and the mastering of asset identification at component level will improve the ability to model corrosion in future and reduce the need for significant data manipulation. 
 
A machine learning model has been developed to predict the number of new defects arising at CM/4 Grade 4 to and indicate the year at which subsequent interventions may be required. This was a deliberate choice at as the majority of CM/4 visual defect data is held for components which are defective at Grade 4 or worse, and, as such, a model training set was available for these defect severities. 
 
On non-buried pipework, a method for determining the times at which existing defects reported by CM/4 are likely to move from one category to the next category has been described in detail. 
 
On buried pipework, a probabilistic approach to calculate the likelihood of failure was adopted. As part of this process, two different approaches have been used to predict the number of corrosion defects on the below ground pipework. The first exclusively uses In-Line Inspection (ILI) data from neighboring feeders plus information on soil corrosivity derived from on-site borehole data; the second also uses indicator data from analysis of the onsite Close Interval Potential Survey (CIPS) dataset. 
 
If new cathodic protection systems are installed in the near future  and this, along with any 
future coating surveys of the buried pipework will give a better indication of the actual level of corrosion defects on the buried pipework. If Project GRAID is also used to complement these surveys then further validation and a likely gain in accuracy will be obtained. 
 
 
Due to the significant levels of commercially sensitive information within the final technical reports, we are unable to publish these on the Smarter Networks Portal. If you wish to find out more detail on this project, please get in touch with the team at: box.GT.innovation@nationalgrid.com  

Lessons Learnt

Effective asset management relies on striking the best balance of cost, risk and performance, 
allowing maximum value to be derived from the assets we control. Each of these factors 
requires a detailed understanding of what assets we have, their condition and their likely 
performance over time under different stresses. 
 
This project has been useful in driving a deep test of the alignment, accessibility and granularity 
of asset data held by NGGT. The need to understand the condition and performance at a 
component level has revealed some gaps that must be addressed if the outputs of this project 
are to be of value beyond the immediate timeframe. Lessons Learnt can be summarised as: 
 

• Creating a comprehensive asset register suitable for modelling 
• Enhancing the use of visual inspection data
• Enhancing the use of survey data on buried pipework
• Recording inspection and repair activities to update risk projections