Posted: 25 October 2017
No Comments

A new era of AI

National Grid’s gas business is applying the latest ideas in artificial intelligence (AI) to improve how it carries out inspections on its pipework. This exciting technology will make the classification of corrosion more consistent and help the business make smarter investment decisions, as Richard Waine, Senior Engineer Asset Strategy, explains.

Share Share on Facebook7Share on LinkedIn224Tweet about this on TwitterShare on Google+0Email this to someone
Article:

A new era of AI

A new era of AI

National Grid is using artificial intelligence to support its vital programme of pipe coating inspections.

"Through a Network Innovation Allowance (NIA) funded project, we’ve been exploring whether AI can help our technicians standardise our inspection process and make our results more consistent"

Richard Waine, Senior Engineer Asset Strategy.

Insight:

The first phase of development is complete, with the AI system achieving more than 80% accuracy in recognising the types of asset regularly inspected by our technicians.

Source: National Grid.

At National Grid, our pipelines must be in good condition for us to flow gas securely across our transmission network. As with all large infrastructure owners, corrosion inspection and monitoring forms a key part of effective asset management.

Richard Waine, Senior Engineer Asset Strategy.

Richard Waine, Senior Engineer Asset Strategy.

Pipe-coating systems help protect our pipework from corrosion at around 450 above-ground installations (AGIs). It’s important we carry out regular inspections to assess its condition and make sure our network continues to work safely and efficiently. These inspections are carried out by our team of trained technicians. Human nature, however, can lead to some inconsistencies. When multiplied over a large number of inspections, these can affect how we view the overall health of our assets and have an impact on the decisions we make on where best to invest.

Through a Network Innovation Allowance (NIA) funded project, we’ve been exploring whether AI can help our technicians standardise our inspection process and make our results more consistent.

Training and educating a machine

We’re using the latest machine-learning and vision technology to train an algorithm – a set of mathematical instructions that allow a machine to learn for itself – to recognise the different equipment types and categories of corrosion found on our network. This is done using a library of tens of thousands of photographs, along with knowledge gathered from technicians and engineers, which are then processed by our algorithm.

The first phase of development is complete, with the AI system achieving more than 80% accuracy in recognising the types of asset regularly inspected by our technicians.

The next step, which we’ll be carrying out in the coming months, is to train the algorithm to classify different types of corrosion. To do this, we’ll need to take over 15,000 new photographs of different corrosion categories across all types of above-ground pipework. This is far more complex than just snapping away with your smartphone.

Our algorithm can only learn if the photos are taken clearly and carefully from the same perspective that a human would inspect the corrosion. It needs to know the precise intent of the human to capture the visual information if it’s to make a correct decision on its condition.

As work on the algorithm continues, we’ll be developing a computer-based app alongside it. The app will allow our technicians to easily take and organise photos, quickly conduct assessments and generate automated above-ground asset condition reports.

Building a reliable, cost-effective network

We believe that this new technology is an opportunity to improve our data collection and standardise how corrosion condition is categorised. With this improved information, we’ll be able to better focus our maintenance and investment decisions.

Another advantage is the continued learning of the algorithm. So even after the project finishes, the system will go on improving its level of accuracy. Ultimately, it will help us build a safer, more reliable and cost-effective network for our customers and the end consumer.

Read more about the innovations that are changing the way Gas Transmission does business in the National Grid Gas Transmission Innovation Annual Summary. 

A prototype of the AI app will be showcased on NGGT’s stand at the Low Carbon Networks and Innovation (LCNI) Conference on 6 and 7 December at The International Centre, Telford. So if you’re going along, be sure to drop by.

LCNI 2017: your ticket to the future