ºÃÉ«ÏÈÉúTV

By 2025, the amount of data generated globally by IoT devices will reach 79.4 zettabytes (ZB) — that’s 79 trillion gigabytes. And nowhere is this data explosion creating more of a challenge than in the infrastructure sector.

Almost every asset we’re managing, seeing acquired or building now comes with connected sensors built in — or the ability to add them at a trivial cost. They’re combining this data intelligence from maintenance reports and asset inspections. As a result, many infrastructure operators find themselves inundated with terabytes of data daily, covering almost every aspect of their assets’ status and operations.

It is true that aging assets with no sensors do face a barrier to entry because of the logistics of retrofitting an entire stock. This is a problem in itself because people are being left behind in an intelligence-led world.

But more and more asset owners are now adding sensors to their buildings and are unlocking access to reams of data as a result.

Without the right data schema, strategy and operating model, organisations are unable to use this data productively and much of it gets lost. Asset owners are, frankly, struggling to keep pace, often employing whole control rooms of analysts who spend their days watching real-time, or near real-time, data – often error alerts – to try to find priority maintenance tasks. And they still often miss the important things. In short, they become data blind.

It’s almost impossible to efficiently pre-empt unplanned failure in this way. And in the worst-case scenario, the cost of increasing the number of data workers will have been offset by cutting the number of engineers and technicians working on site.

This means that even when the data team correctly identifies a problem in a timely manner, there may not be enough skilled workers on hand to address the issue within the right timeframe.

How machine learning can help

The goal of intelligent operations and maintenance (O&M) is to put the right information, into the right hands, at the right time, so that asset owners can make the right decisions. Machine learning capabilities can now enable this by both homing in on the most valuable data to collect from each asset and processing this information more effectively. By concentrating only on those variables that have a real impact on key performance metrics — usually those relating to the avoidance of unplanned downtime — infrastructure operators can unlock new value from existing assets and make better strategic and tactical decisions.

Understanding the most valuable data points you wish to track will also inform where you place the sensors in the first place. Most organisations will find it isn’t financially viable to stick sensors everywhere, especially if they are retrofitting them on older assets. A strategic approach to data collection will always win over a scattergun method of placing sensors either in random places or everywhere.

The next step is to ensure that data is getting pushed to the right applications. Typically, you need a dashboard and alerting system that allow analysts to see important variables at a glance and which alert operators when key variables pass pre-set threshold values. This allows the company to react quickly to any asset that might be at risk of trending off peak performance.

It’s also important to ensure that the right data is collected at the right time — and that it triggers alerts specifically when it is contextually relevant to the company’s O&M and wider business objectives. This is where it helps to have experts, not just in data-driven maintenance but also in the specific field in which the client operates — whether that’s the energy sector, transportation, water infrastructure, or something else.

While traditionally, we tend to think of efficiency in terms of money, time and savings – with digital the enabler to doing things at better cost and pace – enhanced data and machine learning capabilities are also now allowing asset owners to track things like climate benefits, net zero goals and sustainability – all of which are becoming key performance metrics being measured against.

If we get intelligent O&M right, we will be able to predict the performance of assets to a much greater degree and really start to rely on them, instead of being out of control and constantly firefighting. Within a relatively short time, it can have a measurable impact on performance indicators, such as total return on investment, meantime between failure and the sustainability of company operations. By providing into asset performance, this kind of intelligent, data-driven approach to O&M can also help to improve future investment decisions and to increase returns on capital investment.

Gareth Vest is asset and information management director at Atkins and Neil Walker is director for asset management systems at Atkins