IIoT / Vibration Analysis / Smart Manufacturing / Industrial Safety

Checking your work: Maintenance validation via smart sensors

In this edition of What Works, connected vibration sensors make an immediate impact on building management.

By Christine LaFave Grace, managing editor

It’s the kind of moment that geeks out a reliability engineer: Sean O’Connor, a reliability engineer with Jones Lang LaSalle, was on his desktop computer at his home base in New Jersey, examining recent vibration data from a sensor recently installed on a fan at a JLL-managed life sciences facility in Redwood City, CA. On the phone with his counterpart in California, O’Connor witnessed a sudden drop in vibration on the fan.

“I said, ‘Why did you guys turn (the fan) off?’ ” O’Connor recalls. “He said, ‘We didn’t; we just changed the running speed.’ That’s how much the vibration instantly dropped. It was actually a pretty cool moment, from a technology perspective.”

It’s that real-time functionality and insight – the ability to collaborate remotely because everyone’s viewing the same continually updating data at the same time, and the ability to identify immediate consequences of any action taken – that has made O’Connor something of an evangelist for the use of smart technologies to enable a more-effective, more-efficient approach to maintenance.

As a real-estate services and investment management firm, JLL handles building management for companies whose expertise is not in that area. When JLL took over management of the Redwood City site, it undertook an evaluation of maintenance practices – what maintenance was being performed at what intervals, and did those strategies align with how the space would be used by the life-sciences company, a new tenant? In addition, O’Connor notes, “One of the first things we did was walk through an asset criticality study to determine which pieces of equipment really represent the largest risk in terms of disrupting the client’s business.” The finding of that study: “Almost the entire top 20 list was HVAC-related equipment, so that gave us a targeted focus for what we wanted to devote resources toward.”

JLL also sought to expand the use of predictive maintenance technologies for the space. There was no formal vibration monitoring program in place when JLL took over the site, and “we knew with the amount of rotating equipment we had that vibration analysis was going to be a cornerstone of our program,” O’Connor says. JLL evaluated options ranging from a basic walk-around monitoring approach to online continuous monitoring systems, and it was during that process that JLL connected with Silicon Valley IIoT startup Petasense. As O’Connor describes it, what set Petasense apart from other offerings that JLL was considering was the company’s expertise in both software and vibration analysis. Petasense’s wireless, triaxial vibration sensor, the Mote, relies on IoT platform provider Electric Imp for secure device-to-cloud connectivity, scalability, and life-cycle management. Petasense and Electric Imp clouds collect sensor data at user-defined intervals for trending and analysis with the aid of machine-learning algorithms.

Petasense seeks to help companies move from a time-based maintenance approach to “a completely predictive maintenance approach, where you go out to the machine only if the software tells you, ‘Hey, there’s a problem’; otherwise you don’t waste your time and resources,” says Petasense co-founder and CEO Abhinav Khushraj.

Find it faster, fix it faster

For O’Connor, the tech-centric approach to vibration monitoring has more than proved its worth. The aforementioned vibration issue with the fan – a critical fan for the facility – was spotted within a few days of the sensors going live in 2017, he says. The fan had been running close to or at a critical speed, but the problem previously was undetected.

“When you run a fan through its full range of speeds, there are going to be some rough spots that you’re going to have to get through,” O’Connor notes. “The problem is, if you don’t lock those rough patches, those critical speeds out of programming, sometimes (the fans) settle on it. And you wouldn’t even realize it, because this is a big, enclosed air handler. If you were to walk past it, you might not know up until it gets really bad that there’s anything wrong with it.” But a sensor logging the vibration data picked up a vibration spike, allowing the JLL reliability team to take action.

Ultimately, thanks to the machine-learning algorithms’ data trending, “We were able to redefine procedures and change running speed, and we have not had these problems since,” O’Connor says.

Further, from a predictive (PdM) and even prescriptive (RxM) maintenance perspective, “We are beyond excited about what the possibilities hold,” he says. “Taking that conditional data (provided by the sensors) and then also drawing in operational data – running speed, flow rate, anything else that’s being tracked and correlating the two ... To have a human do that is (like a) needle in a haystack,” he comments. Now, “Algorithms are going to be alerting us to this stuff.”

Using sensors to check technicians' work

JLL sees benefits of the smart sensors, too, beyond helping avoid equipment breakdowns, customer service issues, and costly downtime – namely, in validation of maintenance work performed. “We picked up misalignment of a belt-driven piece of equipment,” O’Connor explains. “We put in a work order, had someone go out there and correct it, and when we checked the data after the piece of equipment was put back into service, we realized the vibration was actually higher. So they did something to it; we’re not exactly sure what, but the vibration was higher.”

The incident helped spur a push to close training gaps and better communicate specific expectations for maintenance work, O’Connor says. The site has since started to incorporate use of the Petasense mobile app, allowing a technician using a mobile device to conduct post-repair testing at the point of maintenance. “They don’t turn it back over to operations – they’re using it as a cause-and-effect of their work,” he says. “Any time a PM or a repair is made, we’re revalidating.”

Use of vibration monitoring sensors and, down the road, additional IoT tools also supports PM optimization and more effective use of labor, O’Connor adds. Assigning sensors tasks currently done by personnel can free up technicians to do higher-value work, he says.

Reliability, in a universal language

One of the challenges in getting support for maintenance investments is that business managers sometimes struggle to understand the criticality of identified maintenance issues, especially relative to other priorities within and outside the maintenance department. The Petasense system seeks to eliminate some of that uncertainty by providing a health score for each machine it is monitoring based on the machine-learning algorithms. A score of 10 indicates the machine is in excellent condition; a score of 4 or below indicates that there’s a problem with the machine that requires urgent action.

With that data, discussions with those responsible for allocating resources to address problems are that much easier, O’Connor says. “Not everyone is expected to be a vibration analyst,” he says. “So when we bring issues to the director of operations or the director or manufacturing and say, ‘Hey, we have this really low-priority (to them) issue that you need to be worried about, they don’t really care; they don’t understand; they’re not vibration experts.” But, he says, “Boiling it down to a health score is a common means of communicating.”

Looking ahead in 2018, JLL plans to expand deployment of the Petasense sensors with Electric Imp’s IoT platform to other locations for the same life-sciences customer and to additional JLL clients. For O’Connor, connecting asset performance data to those most directly responsible for maintaining asset performance –  technicians – is crucial to be able to truly optimize a maintenance program.

“Everyone talks about closing gaps and improving your predictive maintenance program,” he says. “What it boils down to is you have one group of people that use calibrated, highly sensitive instrumentation processed through software to say what’s good or bad. And then we write a two-sentence work order and give it to someone who has none of that. (Often) they’re going to the piece of equipment when it’s off and all they have are their five senses. That is a significant gap, and that’s where a lot of programs get hung up.” But, he continues: “If we can easily put that data into the hands of the person that has the means of fixing it and give them a little more insight, the efficiency gain...it’s going to be through the roof.”