Tough economic times increase the pressure on maintenance engineers to maintain uptime and avoid outages. But a predictive approach, based on analysing data from condition monitoring, means you can carry out maintenance when it’s really needed.

Who says that machine needs lubrication? The manual, the way we’ve always done it, or the machine itself? In an era when we all need to do more with less, it pays to listen to your assets when it comes to maintenance.

In the past, maintenance was sometimes just about responding to machine failures. Today, few maintenance plans rely on simply reacting to breakdowns, but this still has its place in some circumstances, according to maintenance engineering academic Dr Moray Kidd.

“Often people think that reactive maintenance is a practice that is no longer appropriate. But for items that are less critical, run to failure may still well be an option,” he says.

“Nowadays, preventative maintenance is the more common practice, where assets are maintained based on typical operating hours or another more appropriate metric of time, depending on how it’s operated,” Dr Kidd adds.

Even when a schedule is being followed, maintenance engineers have always formed their own judgments about when to intervene based on physical monitoring of machines in their care.

But Dr Kidd says that for critical assets, some form of condition monitoring is essential. By using digital devices to monitor the condition of key components inside an asset, you get to see how it performs in real life.

Measuring levels of vibration, for example, can give early warning of component failure, and condition-based maintenance is widely used in many UK factories, according to the Institution of Mechanical Engineers.

Taking maintenance to the next level
The next step up is to move to predictive maintenance based on analysis of condition monitoring data using sophisticated artificial intelligence (AI) algorithms. These build up a complete picture of the whole factory and all the machines within it in real time.

While condition monitoring can alert you to an impending component fault, predictive maintenance allows you to learn when components need replacing, Dr Kidd points out.

On a practical level, AI analysis might show you that some assets run quite happily well beyond the recommended service interval. And because you have a real-time view of their performance, you will still be alerted to imminent failure.

Over time, the picture that the digital system builds up will allow you to plan maintenance to fit in with the actual performance of your assets. This will help avoid breakdowns and give you the opportunity to schedule downtime when it’s least disruptive to production.

The technology may be new, but the idea of maintaining assets when they need it, based on operational observations, is not, says Richard Jeffers, Managing Director of RS Industria at RS Group.

“The concept of looking at the outputs of your actions and using that to drive future actions is not new. We’re not suddenly much cleverer in the 21st century than they were in the previous one,” he says.

“What has fundamentally changed in maintenance is the recognition that data can drive your maintenance strategy.”Richard Jeffers, Managing Director, RS Industria, RS Group

“What has fundamentally changed in maintenance is the recognition that data can drive your maintenance strategy. If you understand failure modes, failure theory and the leading indicators of failure, then you can start to use that understanding to drive your maintenance strategy in a structured fashion.”

A key factor has been a reduction in the cost of the technology used to extract and analyse the data, he adds. “You no longer need human beings to collect and interpret that data – we can allow the technology to do it.

“But what hasn’t changed is the need to create a culture where you derive real value from the data,” he says. “You’ve still got to get people to respond to that data and to value the engineer that lives in a calm environment because he’s in control, more than the hero that runs in to fix things.”

Training the algorithm
Professor Olga Fink of the Swiss National Science Foundation, an expert in “smart maintenance”, says that AI algorithms can now be trained to recognise a healthy production line so they know when something is going wrong.

Dr Kidd agrees. “Industry 4.0 and the Industrial Internet of Things have revolutionised what we do. Access to data through the Cloud opens up a whole host of opportunities and is a real game changer,” he says.

“If we have lots of good, healthy data, we can train machine-learning algorithms to understand what good looks like so that when things become abnormal, we can say – today is not a good day; we don’t know why, but intervention is required.”

Such sophisticated tools highlight how far we have come in maintenance engineering. Where once an engineer had to stand beside the machine to check its performance, today they don’t need to be in the same location, or even the same country.

However, Naim Kapadia of the UK Manufacturing Technology Centre cautions that although you don’t need to be in the factory to monitor operations, you do still need to be able to respond promptly to alerts.

“You can have a machine that was fine in the morning, but in the afternoon it starts to behave a bit differently. Then you can start looking at the data and see what’s changed. This early detection allows you to react before it fails,” he says.

Clearly, data analysis is key to effective preventative maintenance. But according to André Kreul, a Risk Engineer at global insurer Swiss Re, only 1% of the data produced by condition monitoring is ever analysed by AI-driven predictive tools.

“There are still companies that are gathering large volumes of data, without doing the interventions based on what they are seeing.”Dr Moray Kidd, Maintenance Engineering Academic

It’s a phenomenon that Dr Kidd has seen too. “Sadly, there are still companies that are sat back gathering large volumes of data, even with some quite complex machine learning, but without doing the interventions based on what they are seeing,” he says.

“They don’t see a difference with the scores on the doors at the end of the year and so they become increasingly frustrated that it’s not impacted the performance. But that’s because they’re not acting on that data and those insights,” he adds.

Nevertheless, those companies that have introduced predictive maintenance have experienced significant reductions in downtime – up to 50% in one case quoted by Kreul at a recent seminar – and savings running into millions.

Power from predictive maintenance
Predictive maintenance will transform maintenance engineering. It puts power in the hands of maintenance engineers to ensure that plants and machinery function efficiently at all times. What’s more, it underscores the strategic importance of maintenance engineers to the success of the whole organisation.

For information on support RS can provide, please visit RS Maintenance Solutions here.