Intelligent

Get to know your provider options before putting machine data to work

The lure for manufacturers of looking into smarter manufacturing by way of an automated, intelligent reading of machine data has never been stronger—and there has never been more help available from so-called data experts, who seem to be thick in the ground in recent years. Some are better than others, of course—but, given the number of them, who should a manufacturer trust? How does one begin?

The quality of machine data help one may get from an integrator, consultant, or OEM can vary based not only on what a customer needs or asks for, but also on factors such as education and domain experience that influence the provider. Before hiring one of these data gurus for help implementing Industry 4.0 intelligent software, it pays to check around, ask questions, learn more, and really get to know the provider to ensure it's a match.

Smart Manufacturing talked to three providers about a specific smart-manufacturing task: implementing a predictive maintenance function using data collection and analysis. Although the basics were the same, the three illustrated different attitudes and thoughts about how they help their customers.

Preston Johnson, senior solutions manager, asset integrity and reliability at CB Technologies, relies heavily on science.

"I tend to take a physics-based approach to applications supporting predictive maintenance and related business processes in industrial environments," he said. "For example, I leverage my experience in industrial instrumentation and certifications in vibration analysis and reliability engineering. These experiences and education provide a set-up story for predictive maintenance, and tracking the ROI metrics we improve through predictive maintenance activities."

Studies have shown that predictive maintenance is cheaper than maintenance based on time and usage, Johnson pointed out. The real utility is being able to predict far enough out to allow for the work to be done within normal schedules and expected supply chains.

IoTco CEO Mo Abuali's approach is: Think big, start small, scale fast.

Abuali explained that IoTco, short for IoT Company, provides a turnkey solution from the sensor and data acquisition to the analytics to help the customer assess and train their team so they can scale the software. "Our guarantee around predictive maintenance and also around predictive quality is to bring domain knowledge, implement in less than 90 days, scale fast using the client's internal team and bring ROI from A.I. and predictive analytics in three to six months."

At CNC/motion control OEM Heidenhain, customers said they needed an easy, plug-in solution to monitor our machines to see if the spindle is running, what programs are running, and how efficient the machine is running, explained Gisbert Ledvon, director of business development of machine tools. "Customers tell us, ‘we’re not IT experts, we just want to take advantage of the data the machine has without hiring an IT guy.’ And that's what Heidenhain developed the StateMonitor for."

StateMonitor is a software platform enabling remote monitoring and data evaluation during real-time manufacturing.

Heidenhain recently released a new version of StateMonitor and also introduced PlantMonitor, which aggregates monitoring and analysis production data from multiple sites. To use PlantMonitor, at least one machine has to be equipped with a Heidenhain TNC control/StateMonitor software combination at each site, but machine tools using non-Heidenhain controls can be included and monitored.

"We looked at the most common protocol in use in the United States, and it's MTConnect," Ledvon said. "Internationally it's OPC-UA. We’re now offering these interfaces so customers can connect other brands of machines very easily and connect data from them using StateMonitor to bring that data into a nice, clean analytics. And that can be put on your cell phone, it can be on a tablet, so you can collect the data wherever you are."

StateMonitor provides the plug-and-play ability Heidenhain's customers asked for.

"You can plug it right into the control and it will collect all the data the Heidenhain CNC already collects," said Ledvon. "We take that out of the control and put it into nice analytics, you get graphics, and you can analyze your shop and see where your bottlenecks are."

Designed to highlight data usage and encourage time-saving workflows using charts and graphs, the new StateMonitor V1.4.0 offers several benefits. This includes the ability to collect current tool usage data, which helps avoid premature tool replacement; customize the machine status by adding additional conditions and changing to a preferred one when the machine is considered productive; and show monitored signals on its machine status page.

Ledvon explained that the monitored signals capability means StateMonitor can keep track of as many as five sensors per machine, with the resulting data used for predictive maintenance. In addition, operators can input recommended preventive machine maintenance (based on time and/or usage) into the program.

"Then, when you do your process planning, it will indicate you planned this machine to run for the next 50 hours, but remember there's a maintenance to do in the meantime," he said. "You can either override or do the maintenance first."

An operator can also add supplemental information to support an output level.

"The other thing different in State Monitor compared with other monitoring software is we’re allowing the operator to input into the system," Ledvon said. "Our system allows the operator, right on the machine with the ‘FN38’ function, to put information in the analytics. He can report, ‘I couldn't run the machine because I didn't have a program,’ or ‘I couldn't run the machine because I didn't have parts, or the tool broke.’ So the operator doesn't feel like he is monitored and judged without his voice being heard as to why his productivity isn't good."

Heidenhain not only accommodates shops with machines from other vendors, it also has a feature for more tech-savvy operators.

"You can export all the data StateMonitor collects in an Excel format or an Oracle environment, or SQL, or Microsoft, and then you can do whatever you want in the big scheme of Industry 4.0," he said.

Abuali said manufacturers are looking for four things: technical value, domain knowledge, return on investment, and velocity (or speed) to implement and scale.

"This means they are wanting to implement complex A.I. solutions in a complex setting but they want to do it in under three months," he said. "They’ve already been through long and expensive MES [manufacturing execution system] and ERP [enterprise resource planning] implementations."

IoTco looks to interoperate with, or to complement, a customer's existing maintenance environment, Abuali said. A customer might have a computerized maintenance management system (CMMS) for predictive maintenance activities. In that case, he or she wants a machine and its A.I. to alert the CMMS or even auto-generate a work order before the machine fails. In that way, the shop has time to optimize spare parts ordering and work within its maintenance crew's regular schedule.

In other words, Abuali explained, companies want it all: improving labor efficiency, improving spare parts, and improving scheduling. IoTco's belief is not to just implement predictive maintenance software, but to really integrate it within a customer's maintenance team, schedules and work orders.

"That's why it's very important to create meaningful alerts, meaningful notifications—which we call predictive alerts—and also to generate those predictive work orders [such as] maintenance work orders and spare parts requests directly in their maintenance software," he said. "I think that's a big challenge today: There's no interoperability. We don't want to implement siloed solutions."

To overcome the lack of interoperability, IoTco looks to use APIs (application programming interfaces), software intermediaries that allow two applications to communicate.

"In today's world, if I’m predicting that a machine or a component like a spindle or a bearing is going to fail within the next two weeks, for example, you can create rules and you can use that API and auto-generate work orders and requests in pretty much any maintenance system," Abuali said.

At the start of a predictive analytics project for a customer, IoTco goes in and inspects all machines, including legacy assets. It selects the critical machines and critical components on the machine, such as spindles, ball-screws, pumps, and motors. All sensors and hardware are off-the-shelf; IoTco doesn't make anything proprietary.

"And then we help the customer select something affordable and scalable, and we send a person to install the sensors, providing a turnkey, white-glove service," Abuali said. "And then we train the trainer so their internal team can install additional sensors and scale on their own."

To keep costs down and accelerate A.I. implementations, the company has a library of 30+ different applications, or solution templates.

"The fact that IoTco had solutions that were already prepackaged for the equipment we have in our facility, that was a big part of our selection process," said Tyrone Ellis, aftermarket engineering leader at Trane Technologies in Columbia, S.C., in a case-study video.

The real benefit of being able to predict when machinery maintenance needs to be done is it turns a potential problem into one that's easy to solve.

"That allows me to schedule my crews without overtime and order my parts [amid] supply chain constraints," Johnson said. "That problem is easier if I have a longer time horizon."

To start, he likes to use machine condition monitoring thresholds that are based on industry best practices and his own experience with those technologies to give him some idea if there might be something wrong with a machine. He’ll look at any available data and ask how the machine is performing.

"Now I can point out where we can detect machine health degradation; we’re always going to have some degradation," he said. "And you can track and trend in case something changes, but the idea is, you want to capture and have a good solid view of your equipment health that allows you enough of a time horizon to plan before your next outage."

To help assess equipment health, Johnson uses five types of sensors: motor current signature analysis, vibration, ultrasound, oil analysis, and temperature.

"If you look at the current that's actually going into the motor and how that's being used, you can see an electrical motor problem starting to develop," he said. "Motor current signature analysis, I think that it's starting to grow in popularity especially with variable speed drives and smart motor controls. A lot of those have this kind of technology built into them; and then for a lot of the older ones there's some very cost-effective technology you can add.

"Vibration usually comes out of my pocket next, after I exhaust all of my available ‘free’ data,’" Johnson added. "Vibration can tell you more about the failure modes than other technologies. I know if my vibration goes above two inches per second my machine's really dancing and I’m probably going to see a functional problem soon."

Johnson differentiates two types of ultrasound—mechanical and airborne. He uses mechanical ultrasound to detect the "snap, crackle and pop" sounds metal makes as it starts to move due to wear and fatigue.

He offered an analogy: "Sometimes in the winter or the summer your house will creak a little bit, and the same thing happens with metal," he said. "In a lot of machines, an ultrasound in the right place can detect those crackings. That's something vibration [monitoring] won't see until a little bit later in time." However, mechanical ultrasound is not as widely used as vibration monitoring.

Where ultrasound really saves manufacturers money is with airborne ultrasound, he said. In industry, airborne ultrasound instruments are used to inspect pressurized air or steam delivery lines to check for costly leaks.

For oil analysis, historically a shop had to take a sample of oil and send it to a lab, kind of like doing blood work. Now sensors are available that will make the same measurements on a daily, hourly, or minute-by-minute basis and feed the results into a data lake.

Johnson calls temperature his confirmation sensor. "I think I see something in my vibration that tells me my bearings are acting up, how's my temperature?" he said. "Temperature's also going up, so now I’ve got two sensors that are kind of telling me the same thing and that helps when I’m trying to make a case."

Then it's a matter of subject matter expertise, or domain knowledge, to determine how fast failure will happen.

Just as machine informatics aid in predicting needed maintenance, data about prospective providers can help forecast which provider is best for an individual customer.

Connect With Us