Cadent have set up a Gas Governor Improvement Plan to review the condition of their <7bar Governor asset group. In the first instance there is a need to confirm the location and condition of the buildings (conditions covering multiple attributes). Typically, data on the condition of these assets would be captured by a field operative who would need to travel to each asset and manually capture key attributes. This would need to be completed regularly to ensure asset data and condition monitoring is up to date.
A holistic view of Governor asset data is required, including location, asset health and asset risk data to make better asset management decisions. Some key problem areas that need to be identified across the governor asset group within the West Midlands are:
· Issues with the location data of Governor assets and proactive identification of assets compared to historical records
Notes on Completion: Please refer to the appropriate NIA Governance Document to assist in the
completion of this form. The full completed submission should not exceed 6 pages in total.
· Physical location information related to the assets (e.g. Compound, footway, verge etc.)
· Kiosk details and general state of repair
· Vandalism of assets e.g. Graffiti
· Correct vents on kiosks and state of repair of the vents
· Labelling present and correct on compounds or kiosks
· Impact protection of assets where required in safety critical locations
· Identification of changes to the environment to enable proactive site husbandry e.g. vegetation encroachment on assets
· Encroachment of buildings on to our assets
· Identification of standing water and possible water ingress points
Objectives
The key objectives of this project are:
· To survey and provide 360 HD Imagery of 175 Governors in the Stafford area
· To produce an AI model based of the HD imagery captured and asset attribute analysis
· To test the AI model’s ability to provide a holistic view of Governor asset data, combining location, asset health and asset risk data
· To provide a simple system and dashboard to access data and imagery
· Data curation and enrichment; providing data in a format that is useful to Cadent and can be integrated with current asset data to improve completeness and add value
· Ensure data can be easily extracted from systems and autonomously identify and detect changes to various asset data attributes
· Provide an understanding if an AI model can be used as an innovative way of working, lower costs improve efficiency and provide valuable services
Learnings
Outcomes
As a result of the project, the following outcome were delivered or noted:
- 360 degree HD imagery was captured for 175 governors in the Stafford and Newcastle-Under-Lyme areas. This consisted of 718 images of Gas Governor assets and a total of 3,477 individual labelled data points
- The imagery demonstrated that for a large proportion of Cadent’s governor assets, a survey on general condition would be possible from roadside imagery however there were also a significant were not accessible from the roadside or the view was obstructed by enclosures or fencing. For these assets an in-person survey would still be required
- This project has provided a good understanding into the possible uses and limitation for an AI model when used to detect certain attributes from imagery. Key examples were the limitation to the precision of the AI at lower sample sizes. The sample size would need to be significantly larger for most of the attributes listed in the scope of this project to determine if an AI model could be reliable at detection of issues
- The AI model testing also demonstrated the challenges associated with our asset base when using imagery from the roadside. In particular the model was not able to identify attributes on governor houses when the housing was inside a complex or behind a fence as demonstrated in the Gaist – Gas Governor AI Report (pg. 6).
Lessons Learnt
The after-action review highlighted the following lessons learnt:
· The legal delays with NIA agreement highlighted the challenges around project where there is some crossover with current supply contract in place. It may have been useful for the supplier to be provided with a summary of what the NIA agreement does and does not cover in relation to these existing contract/agreement
· When reviewing the results of the project, the outputs highlighted the importance of an in-person scoping session when deciding the functional spec for a project like this. Unfortunately this could not take place due to the pandemic and as a results there was minor details missed or slight confusion in some of the attribute classes captured.
