A new generation of smart, value for money solutions looks poised to open up the market for predictive and preventive maintenance, says Michelle Knott.
The trouble with classic preventive maintenance is that it’s impossible to plan the perfect interval between scheduled maintenance activities.
“It isn’t the perfect answer because you’ll never know if the frequency that you’ve chosen is correct,” explains Geoff Walker, director at Faraday Predictive.
“If it’s too far apart the equipment will break down, but if it’s too close you’re doing more work and spending money when you don’t need to. Also, every time you touch equipment you can harm your reliability.”
Predictive maintenance instead relies on spotting when something is about to go awry and fixing it shortly before it impacts on the process.
This should unlock enormous savings by minimising the work and cost of maintenance while slashing downtime and improving the reliability of critical equipment.
However, two drawbacks have hindered the widespread uptake of predictive maintenance.
First there’s the cost of deploying the condition monitoring (CM) equipment needed to collect enough data to make accurate predictions.
If it’s too far apart the equipment will break down, but if it’s too close you’re doing more work and spending money when you don’t need to
Geoff Walker, director, Faraday Predictive.
Second, there simply aren’t enough CM specialists to trawl through the mountains of data and carry out the necessary analysis.
“It would not surprise me if only 20-30% of all the manufacturers out there deploy CM to any extent. And they’ll only be doing it on 20% of their most critical assets, because if they tried to widen it they’d need far too many specialist engineers,” says John Chappell, director of international business development for AVT Reliability.
Now a new generation of smart, cost-effective CM solutions looks poised to bust the market wide open.
For instance, at the time of writing, Faraday Predictive was planning to launch an initial, portable version of its new CM technology.
A version using fixed sensors, complete with Ethernet-based communications, is on the cards for 2018, while the eventual goal is to deploy a third option using fixed low-cost sensors that link back to Faraday’s server via the mobile phone network. Users will be able to access the results via a web portal.
Faraday Predictive is the recently rebranded reincarnation of Artesis that has been offering CM technology for the past decade. Walker says that the company has been working with mathematicians at the University of Cambridge to completely overhaul the algorithms and code underpinning its systems.
Bearing up
In some ways the Faraday approach resembles the most common CM technique – vibration monitoring.
However, rather than monitoring vibrations of motors and other machinery, the system’s Model-Based Voltage and Current (MBVI) algorithms model systems by monitoring the relationship between the voltage and the current being drawn.
Sensors can be fitted to existing wiring, rather than on the actual machinery, making installation much simpler.
MBVI can reveal an astonishing amount, according to Walker: “If you put more load on a motor it will draw more current. With a varying load the current will vary in step with the load [while the pattern of the alternating voltage remains constant].”
The simplest example is probably a bearing fault, which will result in a slight variation in the torque required to turn the shaft. If it’s one bearing, you might expect to see a corresponding current fluctuation once per rotation.
“Each problem has a characteristic frequency and the magnitude of any distortion says something about how bad it is,” he explains.
Close eye on the situation
Data analystics specialist Senseye is also taking a novel approach to CM. This time, the company’s focus is using data analytics to perform ‘prognostics’, according to co-founder Alex Hill: “In an ideal world, we’ll be taking data that the users already collect. The customer may already have invested in a middleware system to pull a lot of data from their machinery. We’re compatible with all the ‘big boys’ and many smaller systems.
“Ideally, we come along and use that existing data to build models of that machine’s behaviour. Of course, we’re not quite in an ideal world, so the customer may have to look for an Internet of Things middleware solution to suck all that data out of the machines and put it somewhere, so we may end up being part of a wider solution.”
Senseye aims to automate the data analysis that would otherwise demand the scrutiny of CM specialists.
“You’d previously have had someone looking through reams of very boring data looking essentially for the same patterns. You can only really manage tens of assets like that, possibly getting into the hundreds if you’re very skilled,” says Hill.
“For one customer we’re monitoring well over 1,000 machines now, which wouldn’t have been doable before without a vast amount of money. That’s why CM technology has been very much restricted to aerospace and defence until now.”
Leap forward
As a start-up, Senseye focused initially on discrete manufacturing such as automotive, but discussions are currently underway about using the technology in continuous processing, such as oil and gas and water, confirms Hill.
The system can use a variety of data, such as vibration, current or torque. Hill says it takes Senseye around two weeks of automated learning to build its internal model of a new site.
“It learns by looking at the data coming in from machinery and what the maintainers are doing and correlating the two. “So if it spots something strange and that item needed attention soon after, the next time it happens the equipment probably needs attention again. We can use that to give an indicator of likely useful life.”
The system is cloud-based and the Senseye app should alert on-site personnel when something needs doing.
AVT Reliability is part of AESSEAL and a more-established name in CM for the process industries. It too is on the cusp of offering what it hopes will be a game-changing leap forward system.
The company’s web-enabled, cloud-based Machine Sentry system enables general maintenance technicians to carry out the data gathering, while AVT’s software and high-level CM specialists (whether from AVT or the end user) carry out the analysis remotely.
Chappell says that Machine Sentry is initially being offered with a portable data collection system: “At the moment the hardware system physically collects temperature and vibration data and the software schedules and manages data from other sources, such as thermal imaging and oil analysis. It correlates all the results to identify the failure mode.
Our online system will still give an overall alarm but it also allows us to see the data and we can tell the customer what the actual problem is. That’s where the added value comes in
John Chappell, director of international business development, AVT Reliability
“Our software has an app capability so it connects to any Android phone or tablet. If the user already owns one, they don’t need another device. Just download the app, purchase a sensor that Bluetooths to that device, and they’re up and running.”
The next step is a fixed solution, with permanently installed sensors sending a stream of data to Machine Sentry’s wireless gateway via Bluetooth. The gateway then connects to AVT, or the customer may choose to access the data themselves. This version should be available by the end of the year or Q1 2018, says Chappell.
The final iteration will be an online version in which Machine Sentry’s sensors talk directly to the end user’s SCADA system.
“Lots of fixed systems will just say ‘something’s vibrating in a certain part of the machine’ and give you an overall alarm. But in the case of a motor attached to a coupling attached to a pump, the machine knows there’s something vibrating but it doesn’t know what the problem is.
“Our online system will still give an overall alarm but it also allows us to see the data and we can tell the customer what the actual problem is. That’s where the added value comes in,” says Chappell.
