Maintenance Sets the Pace in Food Manufacturing’s Embrace of IIoT

Automated analytics and condition monitoring tools enable cloud-based software to make predictive maintenance affordable and practical.

By Kevin T. Higgins, Managing Editor

In ways big and small, food manufacturing is tapping into the promise and potential of internet-based tools and applications that address some of the most nettlesome problems in plant operations.

At least two makers of AC induction motors are introducing moderately priced sensors for monitoring temperature, current and vibration, a capability reserved in years past for large machinery that would have catastrophic consequences if they went down. One of these condition-monitoring tools, the ABB Smart Sensor, soon will be able to continuously relay data to the cloud for analysis and reporting of maintenance alerts.

Those devices support maintenance’s embrace of the industrial internet of things (IIoT). Cloud-based computerized maintenance management software (CMMS) is supplanting asset management systems hosted on a local server. Cost is the initial attraction, but cloud-based CMMS opens up new possibilities for improving  production throughput, machine uptime and other tangible benefits. As a consequence, maintenance is leading food manufacturing’s charge onto the IIoT.

Maintenance techs are mobile, and a CMMS that enables mobility via personal devices for work order distribution and commentary input has obvious appeal. But the potential for automated analysis of the raw data from field devices, particularly vibration sensors, may finally move food and beverage plants from scheduled and reactive maintenance to predictive maintenance.

The migration of consumer electronics tools into industrial operations is allowing plant managers make an end run on IT to inch toward predictive maintenance. Cell phone technology, data encryption and open source programs are helping maintenance capture data on machine health, patch it up to a remote server and receive real-time feedback while making an end run on IT restrictions.

Cloud-based CMMS liberates maintenance from the tyranny of a local server, suggests Greg Perry, senior maintenance & reliability consultant with Fluke Digital Systems, the umbrella for Fluke Corp.’s cloud-based CMMS, digital sensors for condition monitoring and SCADA system for maintenance. Besides short-circuiting IT concerns about data security, a standalone reliability platform avoids the “siloing of yourself from the rest of the world” that occurs with a locally hosted system, says Perry.

Fluke’s bundled service is called Accelix (www.accelix.com), a centralized depository of information regarding machine health. .Work orders, parts histories, uploaded photos: The platform hosts an expanding record of every touch point with a plant asset, with the CMMS feeding back information as needed. “Everyone has access and can enter information,” he says. “Everyone is speaking the same language and using it as their CMMS historian.”

Los Angeles-based Herbalife International (www.herbalife.com) deployed Fluke’s CMMS in 2014, when the nutritional foods company relocated East Coast operations into a 792,600-sq.-ft. plant in Winston-Salem, N.C. Two additional U.S. manufacturing facilities now are on the platform. In all, 165 employees have access to the database, including quality assurance and other non-maintenance staffers with read-only privileges.

One of the system’s distinctions is the ability to query technician commentary with a text analysis tool. Thousands of comments are inputted by technicians every day to describe the cause and solution to a work order. “We have more issues than we have technicians,” laments Laura Philips, maintenance planner/scheduler at Herbalife’s Winston-Salem facility.

Ordinarily, those comments would lie dormant, but the system allows managers to download them to a spreadsheet and upload the file to voyant-tools.org, an open source reading and analysis tool for digital text. Voyant counts the frequency that words are used. That helps maintenance zero in on problem areas of specific types of machines.

At December’s International Maintenance Conference, Perry and Phillips provided an example of how the system works. Querying the CMMS, maintenance was able to determine the five assets that accounted for the most labor hours and the five that generated the most work orders. A bottle labeler ranked second and first, respectively, on those lists, but there was no indication of what the problems were. Using the key word function, it was determined that a label magazine was the major culprit. A redesigned magazine resolved the issue.

Other elements of Accelex are under consideration, in particular vibration sensors for large motors and the SCADA system for real-time communication of machine health.

Data-cruncher extraordinaire

A shortage of data has never been an issue in manufacturing automation. PLCs and field devices generate enough data in a day to keep a human being busy for a lifetime, combing through it for patterns and anomalies. Reducing human involvement in analysis or eliminating it completely is the goal of many IIoT enabling efforts.

Model predictive controls are designed to do just that, but they tend to be machine-centric and fail to account for all the factors that impact a process, suggests Humera Malik, CEO of Canvass Analytics (www.canvass.io), a Toronto predictive analytics firm. Her firm’s software is able to consider many more input variables, processing perhaps 20 million data points every 20 seconds in a digitized environment to provide “holistic information about a process,” Malik says.

An early project for the two-year-old firm involved the cogeneration power system at Ingredion Inc.’s London, Ontario, manufacturing facility. Gas turbines generate electricity and steam, and suboptimal performance worked against corporate goals to cut greenhouse gas emissions 10 percent by 2020. Energy inputs account for approximately 10 percent of finished-goods cost, and rising energy prices have been identified as a business risk.

The plant’s chief engineer set the parameters for ideal cogen operation, and Canvass’s AI platform was integrated with a digital control system from Honeywell, as well as the controllers of the Caterpillar turbine and other equipment. Because the platform runs on the IIoT, data standards that make the platform “agnostic to the environment” are imposed, Malik explains. More importantly, the predictive model automatically adapts as more data is inputted.

Instead of being tied to a dashboard of performance data and adjusting controls accordingly, the chief engineer now is free to focus on more value-added projects, Malik adds.

Outsourcing analytics is another solution. Machine OEMs like Buhler Aeroglide (www.buhlergroup.com/drying) in Cary, N.C., are moving in that direction, building on predictive controls that already use elements of IIoT.

Buhler already had introduced a moisture control system that relies on microwave sensors tied to controls algorithms to monitor moisture content of product exiting dryers and roasters. Deviation from setpoint moisture levels results in inconsistent product, excess energy costs and, in the case of roasters that provide a kill step, a food safety risk. Buhler’s system automatically adjusts dryer temperatures when moisture levels drift and provides real-time remote access to what is occurring via web-connected devices.

The system stops short of identifying what changed to cause the drift in the first place, allows Paul McKeatham, head of digital services. To provide that insight, the firm recently launched its Digital Services Group to provide that insight and to validate the accuracy of sensor measurements.

“You need sensors to collect data, but the industry already has the ability to remotely monitor operations,” notes McKeatham. Monitoring is passive, however. “We’re trying to develop a complete customer IIoT experience.

“People lose confidence in sensors if they’re not calibrated,” he continues. “We become the champion” by assuming responsibility for confirming calibration. Human oversight extends to “crunching the numbers and spotting trends,” McKeatham adds. “Most of the value depends on how fast you analyze data.”

Smarter analytics

Smart phone technology is the model for a hybrid analytics system recently introduced by Rockwell Automation (www.rockwellautomation.com), Milwaukee. The goal of Project Scio, states Khris Kammer, information partner & competency manager, is “taking manual analytics, automating it, and adding machine learning algorithms and other apps as needed.”

Depending on project scope, analysis is done locally or in the cloud. Either way, a lack of data management expertise in house isn’t a barrier to analysis.

Network switches, advanced controllers and other plant floor devices already are capable of polling data on machine performance over time. Machine-specific problem-solving is best done locally, Kammer maintains, and Scio enables that. For more in-depth, cross-plant analysis, Scio edits out extraneous data to get at meaningful information, such as machine state before a crash. That information then is transported to the cloud for root cause analysis.

“When it comes to manufacturing automation, the amount of data generated is truly staggering,” he observes. Sending all machine data to the cloud would be overkill and counterproductive, which is why “orchestration”—data filtration down to meaningful information—is necessary. Scio does the editing that otherwise might involve a data scientist.

Less is never more when it comes to data for precision maintenance. Working in conjunction with Microsoft’s Azure cloud-computing platform, Tetra Pak Inc. (www.tetrapak.com/us) is collecting condition-monitoring data from 5,000 of its filling machines at food plants around the world to move processors from scheduled to predictive maintenance.

Thermal and vibration sensors are being added to servo motors on fillers already in the field, complementing the sensors already on board, according to Paul Grainger, technical key account manager in Tetra Pak’s Denton, Texas, office. Data are pooled in a London service center, where machine performance is compared to similar fillers worldwide.

Algorithms compare performance and “raise the flag in Europe if a machine is deviating from the performance of the others,” he says. “The value comes from the algorithms and the ability to benchmark against similar equipment,” regardless of ownership.

If corrective action requires troubleshooting beyond the scope of plant personnel’s competence, the OEM can dispatch a service technician with a HoloLens to the site. Microsoft’s Skype service links the technician to engineers in London who can see what the technician sees and provide the field worker with audio and visual assistance to correct the problem.

It may be the most powerful example of leveraging the IIoT to improve machine performance in food manufacturing to date, though similar big-data approaches are sure to follow. Comparing machines’ performance with their digital twins globally, regardless of ownership, looked like a pipe dream a few years ago. Thanks to internet connectivity, it’s happening now.

Show Comments
Hide Comments

Join the discussion

We welcome your thoughtful comments.
All comments will display your user name.

Want to participate in the discussion?

Register for free

Log in for complete access.

Comments

No one has commented on this page yet.

RSS feed for comments on this page | RSS feed for all comments