Cutting factory downtime in half with machine-learning powered prognostics

As manufacturers continue to automate their factories and connect them with intelligent sensors, the data collected is arming them with critical information on the health and remaining life of their machinery, enabling scalable predictive maintenance.

The idea of understanding the health of machinery, otherwise known as condition monitoring, is nothing new. As a principle, it’s been around for generations but online technology saw widespread adoption in the 1990s in the aerospace industry. Unfortunately performing it at industrial scale has remained a pipedream for everyone other than the privileged few who have been able to afford the sky-high costs of scaling it up and employing an army of expensive monitoring and data experts to extract the critical insights.

Prognostics goes beyond condition monitoring by showing the remaining useful life of machinery. As manufacturers delve further into Industry 4.0, they are increasingly desiring these techniques to reduce costly downtime. Fortunately, the democratisation of enabling technologies, such as cloud computing and the internet of things, is now set to accelerate these from their limited-scalability phases.

“Prognostics software takes in data from Industry 4.0 machinery and automatically builds a picture of machine health as well as the remaining useful life,” says Alex Hill, chief commercial officer at Senseye, a young UK company whose machine learning-based software for predictive maintenance already helps several Fortune 100 companies prevent downtime from machine failure on their production lines.TImes Advertorial v1 RGB smaller

“By automating condition monitoring and prognostics analysis, you can do predictive maintenance at scale, so you know which machines are currently healthy and which aren't, and you know which ones will be healthy or not in a given timeframe.”

Senseye helps manufacturers avoid downtime and save money by automatically forecasting machine failure. Its unique machine learning algorithms allow it to be used on any machine from any manufacturer, taking information from existing industrial IoT sensors and platforms to automatically diagnose failures and provide the remaining useful life of machinery.

If unplanned downtime happens, for example, in the automotive industry, every minute costs £50,000, Mr Hill says. “It's an incredible amount of money that our customers want to avoid losing, if they have a downtime event they can never make that back. By using our technology, industrial companies have reduced downtime by 50% in three months. It's incredibly quick to see results.”

Central to Senseye’s approach is cloud-based technology that doesn’t require experts in condition monitoring or data science to operate. It’s designed to be used by the maintenance team, not IT staff.

Previously, predictive maintenance was so difficult and expensive to do at scale because of the need for these skills. By not relying on human expertise to analyse the data, manufacturers can now enjoy predictive maintenance at scale, going from a few machines to a few thousand without requiring more resources.

“We do all of the difficult analysis work for them and they just get information about what machines are causing problems and how long they're likely to live,” says Mr Hill. “We connect with whatever data source is already being used so installation can take anything from half an hour to two weeks, depending on what's in place, making the whole setup process as painless as possible.”

Driving servitization

In the coming years, predictive maintenance is not only set to drastically reduce factory downtime but it will also enable servitization in industrial sectors. The concept has already transformed the aerospace industry by changing the business model of many aircraft manufacturers, for example, from physically offloading assets to selling and delivering flights hours.

The evolution of servitization will mean that industrial manufacturers will begin to sell machine capability rather than machines themselves. For example, instead of the manufacturer selling a welding robot, it will sell a number of welds per hour, moving from a model of selling hardware to offering products as a service.

“Delivering a product as a service requires a high degree of automated data analysis when you're looking at the industrial scale of hundreds or tens of thousands of machines,” says Mr Hill. “Predictive maintenance, without large deployment or ongoing costs and with an ROI of months, now really is achievable.”