Sam Burgess, CEO, SamsonVT
The deployment of condition monitoring is expected to grow strongly over the next few years. As a key enabler of predictive maintenance, it will be crucial to the reduction of maintenance costs.
Knowing when a piece machinery is going to fail can have huge financial implications. The U.S. Department of Energy estimates that a predictive approach can be up to 40% more cost effective than reactive maintenance.
But perhaps more significant is the reduction of downtime and the impact this has on manufacturing output. That ability to apply computer science so we know when to replace components – before they fail and potentially damage machinery more widely – is keeping production lines operational for longer.
A study by GE found that where operators have deployed predictive maintenance in the oil & gas industry (an early adopter of conditioning monitoring) unplanned downtime has been reduced by 36%.
Yet, despite these benefits, adoption in the manufacturing sector remains sluggish. This is due to the way technology providers have previously attempted to roll out this technology. Suppliers have too often demonstrated a lack of understanding when it comes to the operational realities of a production environment.
This has included computer scientists asking businesses to shut down their production lines so they can seed failure modes to train algorithms. Unsurprisingly, most manufacturers find this an unacceptable request. Early condition monitoring solutions also proved hard to scale across multiple sites due to a reliance on the experts needed to interpret the data.
Condition monitoring solutions have evolved, however, and it is now possible to gather information from operational machines so algorithms can learn what normal looks like. It is then possible for the equipment to detect anomalies – be that vibrations, sounds, temperature changes, etc.
The development of this type of scalable solution has been a crucial first step in enabling the broader adoption of predictive maintenance. The next big step, however, will be cultural acceptance within engineering.
It may take time for engineers to believe what a ‘black box’, and the computer science, is telling them. Especially when the technology is detecting problems that aren’t obvious to the naked eye. For example, a sensor might detect a vibration caused by a hairline crack that’s almost impossible to see.
We need to overcome this mistrust though, as the implications are significant. For instance, if an off specification bearing fails, the repair costs could be more than 100 times the cost of the part.
Given the benefits, there is now an inevitability that condition monitoring will become widely accepted in the future. Perhaps the biggest driver of this will be the growing number of equipment manufacturers offering an ‘as-a-service’ approach.
When equipment manufacturers are guaranteeing uptime under a SLA, in a ‘power by the hour’ style approach, we will see more businesses embedding condition monitoring technology during production.
In the short term, however, the deployment of black boxes is how we are likely to see condition monitoring grow. And with scalable solutions already available, there is no reason why we shouldn’t look to bring computer science and engineering maintenance together – and embrace this technology today.
