Predicting machine maintenance needs in food plants is a commendable goal, but accomplishing it is no small task.

When consultants and software sellers start talking about maximizing asset utilization and reducing downtime with Web-enabled tools, maintenance engineers can be excused if their eyes start to glaze.

If condition-based monitoring of equipment in pursuit of predictive maintenance was a slam dunk, a lot more companies would be doing it. Power plants do it by necessity, and airlines have moved from a preventive to a predictive maintenance mode simply to avoid having to ground half their fleets at any given time.

But successful applications in process industries are few and far between. Firms that write the software that is supposed to capture vital data from controllers and specialized monitoring equipment and turn it into alarms that trigger staff action estimate one in 20 clients are practicing predictive maintenance. That figure probably overstates adoption. It counts firms that monitor some critical equipment while continuing preventive schedules and reactive maintenance on less important assets.

Material purchases alone for maintenance, repair and operations (MRO) in U.S. manufacturing are pegged at $300 billion, and predictive maintenance has a role in carving out some of the cost there and in staff time spent on unnecessary work. Computerized maintenance management systems (CMMS) serve as the platforms to reduce costs by extracting predictive data from intelligent devices on the plant floor, but examples of successful food applications are hard to come by.

Even more rare are instances where condition-based monitoring is seamlessly integrated with CMMS and the enterprise asset management systems that upper management focuses on. By making the case for how integration can produce significant yields, suppliers of process control systems and instrumentation for process industries hope to create a lot more.

Incremental improvements

A survey by condition-based monitoring specialists Entek IRD suggested only 5 percent of plant maintenance is predictive in nature. Another 25 percent is preventive, and at least half of that work is unnecessary. Most of the rest is corrective, despite the fact that it costs more than three times as much as predictive steps would have cost.

Vibration analysis is a fundamental tool in condition-based monitoring. It can serve as a predictor of failure in any rotating machinery, and Milford, Ohio-based Entek IRD specialized in monitoring equipment for high-speed compressors and other critical equipment. Last year the firm was acquired by Rockwell Automation and renamed Rockwell Automation Entek. Now it's working with its Milwaukee parent to better integrate monitoring tools with CMMS and control devices in food plants.

The trend among OEMs is to include monitors and other smart devices in the equipment they offer. Sophisticated control panels on compressors were considered pricey add-ons in years past, recalls William "Joe" Buck, corporate vice president of engineering at Engineered Refrigeration Systems, Mobile, Ala. "Now everybody agrees they can't operate without them." The latest innovation is on-board vibration monitors. Those detectors cost about $9,000 per compressor. "That's not a trivial investment," he allows. Nonetheless, "in today's economy, that's not going to slow down adoption."

"A lot of companies try to buy the technology and implement it on their own, but it's important to have a partner to get the most out of condition monitoring," according to Ken Bever, Rockwell Automation Entek's product portfolio manager. "If a company puts out $150,000 to move toward a predictive maintenance model, they have to be confident that they're going to get their money back in 12 months or two years."

Bever and others advocate a phased approach to predictive maintenance. An extruder or other critical piece of equipment is a typical starting point to begin continuous or periodic monitoring to predict failure. If vibration data indicate deviation from a normal profile, that information still must interface with CMMS to trigger a warning message and, preferably, recommend a specific course of action.

"In the food industry, you don't necessarily have the number of maintenance people and staff to execute even a preventive maintenance program," he acknowledges. The key to a workable maintenance plan is to prioritize the assets: at one extreme, repairing on an as-needed basis makes perfect sense, while preventive maintenance is a reasonable strategy for much of the remaining asset base. Predictive maintenance then is applied to the most critical units.

Studies by ARC Advisory Group suggest 35 percent of indirect costs in food production are attributable to maintenance. That ratio can be reduced to 17 percent, Bever says, but only incrementally.

"We have worked with customers who tried to jump to predictive without have a basic maintenance management system in place," he adds. "It's important to start off by establishing your asset registry, identifying your critical equipment and then implementing your maintenance management system. Many times a company will have purchased CMMS software but not implemented it."

At a major tobacco processing plant, predictive maintenance is performed by a team of technicians armed with vibration monitors, thermographic imaging devices and other specialized equipment, but those efforts are limited to critical equipment. Configuring and calibrating the field instruments in the plant's distributed control system is MRO's focus, and an asset management system is in place to accomplish that.

"We have some folks who do true predictive maintenance," explains Bruce Cameron, senior electrical engineer at the plant, "but in the case of less critical equipment, I'm just happy to get what is broken fixed."

Cameron's plant underwent a $500 million expansion three years ago. The project represented an opportunity to install a software system that could pull data from controls throughout the plant to help maximize plant assets. "The audit trail of as found/as left data, pass/fail status, and other data are picked up and tell me what was done and how it affected performance," Cameron says. Three shifts of maintenance engineers keep the plant running, and the skill levels of those workers vary. The ability to "manage all the numbers in the transmitter and use it as a diagnostic tool" is critical, he adds.

Although his plant handles tobacco, much of the drying and blending equipment is identical to that found in a food processing facility. The process controls are similar to what can be found in most other process facilities. Allen Bradley controls coexist with Rosemount units, Rockwell software presides over certain discreet equipment, and Cameron needed open-architecture software that could interface with all of them. He opted for AMS from Fisher-Rosemount.

"We're kind of a batch-continuous process, doing changeovers on the fly, and management trusted my judgment in selecting a smart maintenance system," Cameron says. "Management understands the buzz words like 'smart transmitters,' but it's up to the people in the plant to make these systems work."

Management support essential

Several factors conspire to make that task a challenge. Open architecture will simplify matters in the future, but it does nothing to resolve incompatibility issues with previously purchased controls. "A vendor can give you a real good price on a single-project basis, but if it comes with a closed system you're then locked into that and are pretty much at the mercy of that supplier," argues Jim Masterson, marketing and sales director at Fisher-Rosemount in Burnsville, Minn.

"Plants are running devices from Honeywell, Fisher-Rosemount, Allen-Bradley and legacy systems. There is a mishmash of control devices," agrees Courtney Millwood, a vice president with DataStream Systems Inc. "Somebody has to integrate them all if predictive maintenance is going to be done. Unfortunately, there aren't many integrators out there, and they charge an arm and a leg. The solution to the integration nightmare is probably five years off.

"It's a monumental task for anyone," Millwood concludes. "Until the boardroom takes notice of the advantages, you're not going to see change. It's not just about predictive maintenance, it's a holistic approach to managing your assets."

Another barrier is the lack of integration between operations and maintenance. Maintenance and operations remain separate departments at 43 percent of food companies, Bever points out. "A lot of this is basic engineering principles applied to maintenance, but adopting them requires changing the mindset at a company," he says. "It's important that the people side of the equation be managed, as well as the technology side."

A delicate aspect of the human element was touched on at Food Engineering's Food Automation 2000 conference. Workers are suspicious of automation initiatives like predictive maintenance, pointed out Ron Hatfield, maintenance manager at Pepsi Bottling Group's Riverside, Calif., plant. If the purpose of automation systems is to reduce maintenance, workers understandably will view the technology as a threat to future employment. That can undermine support for implementing the systems.

Hatfield also touched on another impediment to maintenance automation. Despite significant strides, maintenance technology still lags the technology for automating production, he noted. His efforts to automate some tasks were seriously hindered by the absence of a HMI component for maintenance in the asset management system.

His plant uses MP2, an asset management program from Greenville, S.C.-based DataStream. The firm boasts an installed base of more than 60,000 plants, making it the largest asset management software vendor in the world. H.J. Heinz and other major food firms use MP2. But DataStream may take a back seat in food and beverage to Maximo.

A distant second to DataStream with an installed base of 8,000 customers, Maximo is marketed by MRO Inc. (formerly PSDI), a firm that has forged partnerships with many of the food industry's software and controls suppliers. "Connecting these systems isn't free," acknowledges Milton Bevington, MRO's director of asset management, and a certain amount of customization is a given. "Software doesn't know anything; maintenance engineers know a lot," he points out. "The purpose of the software is to incorporate the knowledge of the engineer into the automated maintenance system."

"Predictive maintenance isn't a thing, it's a process," Bevington adds. "It can be as simple as counting the number of times a fuel filter is replaced and getting an alert when the fifth change occurs, or it might rely on a technique like spectrographic oil analysis. It is the job of the controls to feed data on temperature, pressure, on/off status and so on to a PC. It's the job of a system like Maximo to tell you maintenance must be done. If it's a good system, it tells you before there is a failure."

The final frontier

Until food manufacturers get a better handle on CMMS, efforts to adopt preventive maintenance strategies will face an uphill struggle. Without CMMS to move plant-level information to interested parties throughout the organization, the potential of process-related data will remain tantalizingly out of reach.

Bridging the gap requires a commitment from top management, and that can be difficult to attain. "Capital spending on maintenance is something that's hard to sell," reflects Buck. "If somebody wants to look good, he'll come into a plant and cut maintenance and other costs to become the lowest-cost provider. But he better be able to move on to another job in a few years when things start to break down."

"Maintenance is almost a philosophy, and it has to be bought into at the highest level if it is to become a tactical advantage," adds Rockwell Automation's Don Lovell, a onetime maintenance manager at Kellogg Co. "If it's not part of the corporate vision, it may not get the attention it needs."

"In a lot of ways, maintenance is the last frontier in capital investment, but companies are waking up to the fact that it can be a gold mine in cost savings and improved output," agrees Bever. But with corporations cutting back on greenfield projects, boosting capacity at existing facilities is becoming more important. If more capacity is needed and a new plant isn't an option, better asset utilization is the answer, and predictive condition-based monitoring is the first step along that path.