Building Management / Plant Maintenance / Technology / Smart Industry

Predictive Maintenance Is Coming to Food Manufacturing

Data collection and analysis from a broader base of installed equipment should help both processors and their OEMs.

By Kevin T. Higgins, Managing Editor

Machine downtime, by and large, is a random event, which explains why failure prediction remains an elusive goal in food manufacturing.

Predictive maintenance certainly is possible, as practices in other industries demonstrate. The big obstacle is cost: Establishing an in-house predictive maintenance program “can easily exceed $100,000 for start-up alone,” according to authorities at SKF Inc., Lansdale, Pa. It also requires another precious commodity: time. Time-based maintenance routines are not a panacea for random breakdowns, but they are evidence of a plan to reduce their frequency.

The cost of infrared thermometers, ultrasonic probes, vibration monitors and other tools of condition monitoring are declining, although price points still may exceed the budgets of smaller food and beverage manufacturers. Smart sensors that provide similar feedback may make affordability a non-issue, but if they simply generate more data, they won’t move the industry any closer to predictive maintenance.

Predicting failure for a specific machine may be a challenge beyond the scope of vibration analyzers and temperature probes. A broad sample is needed for reliable prediction. Equipment manufacturers theoretically could determine the early warning signs of a breakdown, but OEMs seldom are able to benchmark performance once machines leave their shops. That’s beginning to change.

After two years of data collection with a computerized maintenance management system (CMMS), deli-meat processor West Liberty Foods had amassed an impressive database on the performance of 50 high-speed slicing machines. Analyzing that data was another matter, so fault signals and repair data were exported to an Excel spreadsheet and turned over to the slicing machines’ OEM. The vendor was able to identify three maintenance procedures that addressed many downtime events.

Tetra Pak Inc. ( is taking a more expansive approach with a remote monitoring service involving its global installed base of more than 5,000 packaging machines. It begins with beefed up sensing capabilities on machines, augmenting the sensors already on board. Two or three additional sensors are retrofitted onto each servo motor, according to Paul Grainger, technical key account manager at Tetra Pak’s U.S. headquarters in Denton, Texas. With anywhere from two to 20-plus servos on each machine, it isn’t a stretch to call those packaging systems smart machines.

Sensor data are uploaded to a cloud server and analyzed by service experts in London. When heat or vibration from a component begins to move into a certain range, the manufacturer is alerted to the predicted time to breakdown, based on machine failures across the network. If the manufacturer subscribes to Tetra Pak’s plant care solutions service, a technician with a Microsoft HoloLens is dispatched, linking the technician to the London experts and guiding the pre-emptive maintenance.

HoloLens adds a Buck Rogers flair to the service. Field technicians communicate in real time with the London experts through Skype, which also is owned by Microsoft. Grainger downplays the optics, however. “It’s not the hardware investment,” he says. “The value comes from the algorithms we’ve developed, the ability to benchmark against similar equipment and the dedicated personnel’s competence in interpreting the data.”

In a pilot test, a U.S. processor of fluid milk quantified a 48-hour time savings in recovering from a downtime event, thanks to a HoloLens-equipped field technician. With the headgear giving experts a real-time view of the machinery and the tech viewing diagrams and other support materials, root-cause analysis was simplified and accelerated. Otherwise, an “escalation process” of identifying the problem area, matching it to the appropriate expert, and flying that individual to the site would have delayed getting the machine back on line by two days.

The service is limited to Tetra Pak equipment, although it is evolving to “a plant-wide approach” to include robotic machinery and other sophisticated equipment, notes Grainger. The firm is “exploring possibilities” of extending the service to other OEMs’ equipment and technical-service centers.

OEM feedback loops

Schneider Electric USA ( is developing a similar “augmented reality” system using Google glasses, tablets or similar devices, according to Eric Lemaire, food & beverage segment marketing manager. As with the Tetra Pak system, it aims to curtail fishing expeditions and instead target maintenance activities at likely problem areas.

“We’re working with OEMs to help improve basic maintenance of their machines,” says Lemaire. He describes the effort as being in the proof-of-concept stage as Schneider collects data on machine performance and breakdown precursors such as electric draw.

Milwaukee-based Rockwell Automation ( has similar ambitions. By early 2018, it expects to launch a subscription service aimed at OEMs to monitor the health of their machines in the field and improve overall reliability and productivity.

“The cloud capability is a prerequisite for allowing OEMs to participate in the machine-health monitoring process,” says Todd Smith, manager of a suite of analytics products the automation firm is introducing. How a machine is operated can have a big impact on performance, he points out, and part of Rockwell’s focus is “to compare machines to each other to understand what makes those machines act so differently.”

A selective approach will be taken to data collection for both analytic and economic reasons. Over-archiving obscures the information value. Data filtering is essential, and the service will focus on a limited number of variables. Additionally, the cost of cloud storage depends on how much is stored. The monitoring service will limit machine data to three months, subject to change depending on OEM willingness to pay.

Regardless of whether data is stored in a local historian or the cloud, analysis tends to be machine-centric and geared more to preventive maintenance than process improvement, argues Bert Baeck, CEO of Trendminer (, a Houston-based software firm with roots in chemical processing. Contrary to conventional wisdom, process issues are more likely to cause plant shutdowns or slowdowns than equipment failures, insists Baeck. Instead of building a data model for every machine in the plant, process engineers should adapt a process analytics model to identify patterns that foretell disruption.

Trendminer has developed a data scientist in a box that integrates data from control systems, lab systems and annotations by operators and engineers to deliver time series data over a specific timeline and visualize related plant events. It is a data modeling approach similar to business analytics that moves from descriptive to diagnostic to predictive and finally prescriptive remedies. “It’s all fingerprints” of future failure, he says.

“Every food company will need data science at a corporate level, but that’s not practical on a plant level,” Baeck observes. Trendminer wants to fill the expertise gap with a software solution based on process data.

Sunshine in the Cloud

Fluke Corp. and SKF are two of the largest vendors of condition-monitoring devices, and both are rolling out new tools that are more economical and practical for use in food and beverage manufacturing.

Both firms have developed mobile, sensor-based kits that can profile performance on multiple pieces of equipment on a temporary basis. Both are leveraging advanced technologies like ultrasonics to detect high-frequency sounds and electric discharges. Fluke’s point of distinction was last year’s acquisition of eMaint Enterprises (, a CMMS provider based in Bonita Springs, Fla.

As with many other CMMS software packages, eMaint’s platform is cloud-based, which helps when integrating Fluke’s smart machine technology. On-premise CMMS packages create unnecessary information bottlenecks, according to Greg Perry, a senior consultant with eMaint, “and it’s too political in nature. IT wants control of systems on a local server, so it’s going to be shackled by the best interests of IT.”

CMMS provides “a snapshot in time” and an important record of events, including those essential to demonstrate compliance with food safety regulations, Perry points out. Condition monitoring, on the other hand, “can catch the bad actor when it’s misbehaving.” Used in tandem, they help boost overall equipment effectiveness.

Cloud-based CMMS is emerging as must-have technology with younger maintenance workers and engineers, he continues. The first thing those professionals do when exposed to cloud-based CMMS is to go to settings to learn how they can filter and view data in a format most comfortable to them, says Perry. “The talent that’s replacing the stubby pencil and pad mentality demands data that is instant, accessibility that is instant.”

Eric Whitley endorses that view. “A cloud-based CMMS provides the ability to manage a CBM (condition-based monitoring) process from a mobile platform,” writes the business development manager of Leading2Lean (, the Wellington, Nev., technology firm that supplied West Liberty’s CMMS. “Using mobile technology to input condition data in a real-time manner allows for instant updates and instant launch of preventative measures when conditions are met.”

Data access and retrieval is particularly important in food production, adds Whitley. Spoilage and food safety risks grow when machine downtime disrupts production. “It’s important that food producers have the highest confidence that the data used for a CBM process is solid in order to avoid the negative ramifications of machine failure.”
For all the improvements in machine health it offers, technology alone won’t improve uptime and product outcomes. Proper installation and use of components is fundamental, and some company cultures are simply more scrupulous about the basics of machine maintenance.

“The first line of defense is diligence, and some plants, unfortunately, are less diligent than others,” reflects Steve Katz, owner of Emerson Bearing ( in Boston. “Successful people are diligent and have a plan.”

Bearing failure is the most common cause of machine breakdowns. Telltale signs like metal discoloration are tips that a bearing won’t reach its predicted B10 life — the point at which 10 percent of a particular type of bearing will fail — but inappropriate maintenance or inappropriate bearing selection predetermine failure. Over-lubrication, shaft tolerances that are too tight and failure to upgrade bearing selection after repeated breakdowns are the root cause of many downtime events, Katz says.

“The vast majority of bearings are over-engineered,” says Katz, whose firm is not affiliated with Emerson Electric. “If the right bearing isn’t matched to the application, that’s when problems occur.”

Exposure to chemicals, temperature extremes and washdown environments contribute to premature failures. All are characteristic of many food plants. One Katz client affixes simple heat-sensitive paper labels to some bearings to monitor when temperatures spike to critical levels. It’s a simple way to monitor machine health, one more arrow in the uptime quiver.