Living With And Learning From Your Data

Big Data can be too big for some. Getting a grip on it—and its value—means separating wheat from chaff, say experts, and acting on revealed trends.

A January advertisement for SAP claimed that “complexity” costs the world’s top 200 companies $1.2 billion annually. It goes on to say that “Simple saves.” And while some maintenance professionals might find SAP’s plug for simplicity amusing, the business-management software giant is on the mark—not just for the eye-catching dollar amount, but for the cause of those wasted dollars.

Definitions of “complexity” vary, of course, but in the current manufacturing environment, one factor maintenance pros increasingly view as a complexity contributor is data. Suddenly, there seems to be a data surplus. You can’t live without it—the challenge, in fact, has always been to obtain more data—but technology has now met the challenge, and then some.

“With the continued adoption of industrial automation systems, device and equipment data is originating from a variety of technology platforms,” says Juan Collados, Principal Applications Consultant for Schneider Electric. “This includes SCADA and distributed control systems, safety management systems, manufacturing execution systems and mobility applications, to name just a few. Add to that the Internet of Things, where billions of devices and machines are becoming interconnected on a global basis, and we can see why we now have an overabundance of data.”

ABB’s Kevin Starr, Director of Product Management for Process Automation Service, likens manufacturers’ exposure to data as taking a drink from a fire hydrant. “You get really wet,” he says, “but you don’t know what hit you.”

Context drives value

For Starr, Collados and others tasked with making sense of data for clients or crew, determining exactly what does hit you is the real issue, and cannot be a random action. The same technology that provides the quantity can manage and guide it, they say, but it’s first necessary to know exactly what is valuable for your operation. Data is too often “served up without specific context,” says Collados. “Acquiring quality data and transforming it to actionable information, therefore, becomes a focal point in enabling an effective asset-management strategy.”

Here, “quality” means data that has value in its ability to provide useful interpretation of equipment status and trends. But with so much data pouring in, value takes on new meaning, too, says Gil Acosta, Director of Engineering Services at eMaint, a New Jersey-based provider of CMMS software-as-a-service solutions. Asked what he tells clients who are looking for guidance on how to handle the abundance of incoming data, Acosta says he guides them “into finding the metrics of importance. I’m careful to use the word ‘importance’ because it’s easy to get caught up in the standard measurements out there. And if you get too involved in them, they may not be that meaningful at your place of business.”

Metrics that should be reviewed, suggests Acosta, include Mean Time to Repair (MTTR) and Mean Time Between Failure (MTBF). “These are old, reliable measurements that are still useful, but don’t always tell the whole story any more.” They carried more weight, he says, when reliable data was in short supply. “They were a way of getting a small sample size and turning it into information. Now, with the massive amount of data that we have, we have way more calculations available to us than just MTTR or MTBF. We have trend analysis.”

Trend analysis can be as simple as a “change in slope,” says Acosta. It can occur over any period of steady data input of important measurements, as opposed to the former need to catch changes during the brief windows of observation that were typical in traditional data-capture techniques. Important measurements are anything condition-related: temperature, pressure, hours of operation, amperages and others. “And with many condition data points, you can start to monitor trends,” says Acosta, “and see changes in the condition of that asset, making MTTR and MTBF less useful.”

Most of the maintenance pros Acosta meets quickly agree on where such data detail can lead. “It’s the kind of information they have been missing for years and would love to suddenly have,” he says. “Things like cost of ownership, energy consumption, labor consumption, parts consumption. So when they learn how their work-order system collects that data, and how easy it is to get those reports out of the system, the process itself almost becomes an afterthought because it’s so simple. It’s the concept of starting with the end in mind,” he says, “and I can’t tell you how many times I’ve used that phrase. I tell folks to concentrate on the data you want to extract from the system to help you make better decisions. Invariably, I’ll get three or four things right away. Then I’ll say, let’s work backwards. Let’s make sure the data source is there to answer that.”

Let your system do the work

Starr uses a simple analogy to describe ABB’s current approach to providing data context from automated systems. “We’ve put in devices that assimilate the information and sort it,” he says. “And if you can imagine the old game Stadium Checkers with the marbles that would fall into holes at different levels, that vision is really accurate because you have all these levels now, and so much information, but nobody can see anything, so you have to filter it so it catches.”

Some call this process “data reshaping,” says Starr, which “basically means that you’re looking at statistical computations and probability of wave patterns in the data that correlate with known problems. This allows you to then sort it into smaller bins a human can look at and interpret. That’s the new skill I see evolving and is very much needed.”

When it comes to automated systems, this skill is sometimes more easily handled at the vendor rather than plant level. Starr therefore recommends that every time you add a layer of automation to make your life easy, you should add a service component that makes sure that level of automation gives you the right answer. ABB offerings in this regard include its ServicePro Service Management System, a global, real-time database of proven maintenance best practices and schedules.

“ServicePro,” Starr explains, “allows us to look at failure rates from the factory, replacement times, how long should it take, how often should it fail, and we make comparisons with global figures to this particular site. If a part fails on average once a year, but at your plant it fails once a month, we know something’s wrong. One of the issues with data,” he adds, “is that you can make it read whatever you want. But when you have a global average and you have thousands of items—we’re now managing 600,000 parts—you get updates every day” and, ideally, a relatively clear path to problem-solving.

An in-field factor that lends support for a service like this is equipment age. Certain older control systems can make data-capture difficult. “For example,” says Starr, “if the system is not OPC-compliant—and there are many out there like this—you can really be flying blind. A lot of [manufacturers] are trying to compete in this century with old technology. And if you don’t maintain these assets, at some point you’re going to be out of business because you just can’t see the information.” While this stance usually requires him to explain why new systems are so much better, Starr says this is an easy task. “The new systems have fail safes, redundant servers and decoupled processes for collecting data so you can’t harm them by trying to extract data [as in some older systems]. And when customers see what can be done, they’ll often say they want this every day, which means they’re talking about an integrated solution.”

Collados concurs that the newer your equipment, the better your data analysis can be. But he also recommends creating system and data solutions that don’t depend on a single vendor’s approach. “Maintenance professionals should ensure their data collection and analysis solutions are vendor-agnostic to both the sources of data, control and safety platforms, and the business systems (CMMS\EAM) being utilized to manage industrial assets,” he says. “Ease of use leading to end-user adoption is critical, where applications should provide an end-user experience that offers a clearly perceived benefit relative to the implementation and operability investment. Applications should be uniquely intuitive and offer standard configuration ease-of-use concepts such as wizards, drag & drop and templates. Equally important, is support from all levels of management in implementing a clearly understood asset-management strategy.”

When data speaks

Getting management support is clearly an ongoing challenge for some maintenance operations. With modern data elements, the process can be much easier, provided data meanings are properly collected and conveyed. “Too many maintenance professionals treat every asset in the plant the same way,” says Acosta. “This is often because it’s so obvious to them when an asset needs replacing that they don’t bother to calculate the return on investment. But not everything is a bald tire. They need to be able to say it’s costing X per month to maintain it, the number of PMs it needs is up, the warranty has expired and it’s near failure. They then have to be able to say, if you give me X to replace this, I’ll give you a return of some sort.”

The response to this approach is typically an approval for funding, says Acosta. But too often the story is not presented that way and the request will be “added to the list.” Noting that the data to support this type of story is probably already in the CMMS, Acosta adds that “you must do the homework to understand what the investment part of it is, and then interpret what you’ll get back. For example, how will current costs change once I make the investment? Answering that means tracking the labor, the parts, the oil consumption and everything else that goes into maintaining an asset.”

And that’s where Big Data can simplify the entire process. Not only can it rapidly take maintenance and reliability teams many steps forward on their continuous-improvement path, it can help bridge the skills-shortage gap most operations now face. “It used to take a person five years to get to the point where they could really maintain a site,” notes ABB’s Starr. “Now, with the tools we have, we can take somebody who is relatively new to the industry and in six months to a year, they’ll be doing work that took me 10 years to figure out. So we’re moving in the right direction.”

The 3 Main Types of Data

Data-collection-and-analysis training requirements can be described from the following three distinct, but interrelated perspectives, says Juan Collados, Principal Applications Consultant for Schneider Electric.

Disconnected and Stranded Assets
Equipment and devices outside of the automated control and safety network, where a mobility solution can be effectively implemented. This often comes in the form of operator rounds or planned inspection activities supported by mobile data-capturing capability. Data is either manually collected or automatically entered through handheld device such as an infrared camera. Condition-based data for this asset base also often originates from third-party services, such as oil spectroscopy and analysis contractors. Regardless of its origin, the data is still relatively raw. However, it can be monitored with rules- and template-based condition management applications to make it truly actionable.

Instrumented Equipment and Components
This asset base is typically within the control and safety network and can provide valuable maintenance-relevant data that can be transformed into actionable information through condition-management solutions. This data is typically stored in process historian platforms and can be collected through a variety of “data source” communication protocols such as ODBC (Open Database Connectivity) or OPC. Condition-based maintenance rules utilizing analysis tools such as thresholds, statistical process control or even simple expressions can then be associated with the collected data. The analysis should yield the desired results, typically in the form of notifications to maintenance and operations, automatic generation of contextual event-driven maintenance work orders or requests, and an ongoing optimization of the overall maintenance plan.

Intelligent Devices and System
This asset base generally resides within the instrumentation layer, but can include larger equipment and systems. Both can provide critical process and maintenance-relevant data, including, but not limited to, information relating to its current state and health through its self-diagnostics capabilities. It includes devices, such as smart-valve positioners and transmitters (level, pressure, etc.), as well as major equipment, such as HVAC systems in a data center or cleanroom. This type of asset provides advanced data broadcast capability directly through the base-level automation network. It utilizes a variety of real-time digital communication fieldbuses, such as HART, Foundation Fieldbus and Profibus, to name a few. Despite its complex nature, device manufacturers now provide rich human-machine interface applications that empower end-users to easily interface with this type of device or equipment. Vendor-neutral HMIs based on open standards are also available and increasingly relevant since one HMI can be used with any device regardless of the original manufacturer. In many cases, smart-device manufacturers provide not only the quantity of data required to manage the asset, but also offer richness to information quality and context.

Don’t miss Juan Collados’ free Webinar “Lower Your Maintenance Costs Through a Condition-Based Management Approach,” Thursday, Feb. 19. For information or to register, visit efficientplantmag.com/SchneiderWebinar.