Companies in the energy and utilities industry are asset intensive by nature. Producing, extracting, storing, and eventually delivering to the consumer requires an immense network — production facilities, transmission, distribution, and more — which needs to be carefully monitored and maintained.

In the face of this complex and often expensive challenge, it makes sense that all energy companies, regardless of size, invest in some form of asset management and predictive maintenance to optimize assets. Some companies rely on internal experience and historical benchmarks to predict when an asset will need to be repaired, maintained, or replaced. But forward-thinking companies make use of data and statistical analysis to make these projections more accurate and spend no more than absolutely required to balance availability, reliability, and efficiency.

Here’s why asset management and predictive maintenance based on good data will improve this balance, and how to implement it.

Simply put, predictive maintenance is adding operating and performance intelligence to the process of anticipating when assets are going to degrade or break.

Why invest in a rigorous process for predictive maintenance? To start with the obvious, choosing to rely on reactive maintenance only can lead to disastrous outcomes — workplace injuries, financial losses, and insurance claims resulting from assets that fail due to improper or poorly timed maintenance.

Consider the example of a pipeline rupture, or a fire in a facility or adjacent lands due to poor utility asset management. Beyond the safety, environmental and financial implications, immense reputational damage can occur to the company.

Because of the ever-present risks associated with faulty equipment or assets, companies in these asset-intensive industries are known to take an overly conservative approach to predictive maintenance. While this abundance of caution is a good thing — far better than the alternative — it can be unnecessarily expensive and wasteful if you don’t use the proper tools and process.

When you take a deliberate approach and build predictive models for asset maintenance, you gradually remove the guesswork from the process. Today’s predictive models can be built on accessible and user-friendly cloud platforms; they leverage your existing and historical data, integrate assumptions, and run it all through an algorithm that gives you a clearer picture of the status of your assets.

Using these tools to monitor your assets ultimately saves you time and money while maintaining or improving availability, reliability, and efficiency.

Consider the example of a power company serving a municipal area, with countless transformers on the ground and lines connecting people to electrical power. Every so often, it sends a crew to perform a visual inspection that all equipment is running properly. And they almost always are when this inspection is undertaken.

When deciding how often to send the field staff to perform regular checks, this company relies on a somewhat arbitrary schedule, or possibly a schedule enhanced by historical data and incidents. Perhaps a veteran employee, who has observed several maintenance issues over a long career, is able to make more educated assumptions about when to conduct repairs. The company risks running into problems if these veteran employees leave before passing on key knowledge, or if it switches to a new type of equipment the veteran employees are unfamiliar with.

If this sounds similar to how your company approaches predictive maintenance, it’s likely time to re-evaluate your process. But where to start?

The first step is to conduct a discovery session, so you know how much operating and performance data you already have. Many energy companies are collecting, or have the ability to access, massive amounts of data they may not even know about. Any device equipped with sensors should be constantly providing data and input; take for example a fracking diesel engine that contains sensors that measure temperature and vibration.

And you’re not limited to the data you already have — there are innovative methods that allow you to gather new data on a one-time or recurring basis. These methods include digital twinning (the process of creating a digital replica of a physical object), and intelligent pipeline pigging (using a smart pipeline integrity gauge (PIG) with probes and sensors that can tell you about the physical state of a pipeline).

All these inputs and data serve as the starting point for building a more informed and better predictive model. The statistical model will have assumptions incorporated in it, especially when it’s first built. It will not be perfect. But over time, as that model gets re-trained, the advanced analytics become more accurate and useful.

One of the main barriers preventing small and mid-sized energy companies from adopting modern predictive maintenance is cost. Large energy companies investing in proprietary software for predictive modelling can and do spend millions in licensing fees.

Unable to afford those costs, many smaller companies instead rely on the manual land labour-intensive processes described above — decisions based on historical knowledge and guesswork rather than good data. But advances in technology have created a middle ground that’s accessible to all.

Using cloud technology such as Microsoft Azure, you can set up sophisticated predictive models in a matter of weeks. There’s no need for expensive software or massive on-site technical investment. You can begin to build a modern system using the data you already have and pay only a reasonable monthly cost for your cloud consumption bill.

Forward-thinking executives in energy and utilities are already making these investments and improving asset availability, reliability, and efficiency as well as safety and environmental performance.