The concept of intelligent maintenance prognostics is more a trending of information, as opposed to predictive maintenance, which is more trending of data, explains Jay Lee, director of the National Science Foundation’s (NSF’s) multi-university Center for Intelligence Maintenance Systems (CIMS, www.imscenter.net), at the University of Cincinnati.
Moin Shaikh says prognostics represents an engineering discipline focused on predicting the future condition of a component and/or a system of components. “Prognostics is process- and device-centric, while predictive maintenance is only device-centric,” adds Shaikh, marketing manager with the Process Automation Systems Group, for vendor Siemens Energy & Automation Inc. (www.siemens.com), in Spring House, Pa.
But you need actionable information, emphasizes Lee, the Ohio Eminent Scholar in Advanced Manufacturing at the University of Cincinnati. “We [at CIMS] use smart software or algorithms to convert the data.” A common analogous real-world example is weather forecasting, he says. Predictive data might be ambient temperature and wind speed, while prognostics information would be the likelihood of rain or snow, for instance.
It’s through similar industrial forecasting that prognostics add value to end-users. “It reduces the cost of unscheduled maintenance, and, for that matter, also reduces scheduled maintenance due to having better knowledge about how long a mechanism will last,” says Ralph Maguire, principal engineer for the automation supplier Bosch Rexroth Corp. (www.boschrexroth-us.com), Hoffman Estates, Ill. It optimizes productivity, and ultimately, the financial bottom line, by moving from a reaction-based maintenance mode to a proactive mode, Sinha notes. “It helps avoid unplanned downtime,” Shaikh says. “Prognostics shows how much savings occurred,” adds Lee.
To fulfill all those potentials, prognostics produces four types of actionable information, Lee explains. One is degradation. “That’s about performance change. It’s from 0 to 1, where O is not acceptable and 1 is the best.” The second type is minicomponent, each with its own degradation that’s shown in what is called a “radar chart.” These three-dimensional views allow end-users to see when components degrade, he explains.
The third type is pattern. It states, according to Lee, “multiple degradation into one combined symptom.” For example, a person goes to a medical doctor and complains of a parched throat. “The doctor will judge you, based on what you have,” Lee remarks. However, “if you have a fever, it might be flu. But if your nose is stuffy and sinuses stopped, you might have an allergy or cold.”
Risk is the fourth type of actionable information prognostics offers, Lee states. Using a risk chart, “this means that before degradation occurs, before you take action, you must know the impact.” Cost and criticality comprise impact, he explains, “and you take action based on the highest impact that could occur.” For example, if given a flickering light bulb vs. a malfunctioning air-conditioning (A/C) unit, you’ll likely check the A/C unit first, “because the downtime of a light bulb failure is less than the downtime of the A/C unit.” Lee notes that the cost component is the total downtime cost.
These risk charts help people plan, Lee emphasizes. But to provide optimal information, he advises, charts need to be done on every component in a system. To do that, the NSF’s CIMS has developed 20 algorithms, notes Lee. Using them, “once you have a transparent view of the health vs. cost, then you can have a much better way to optimize, strategize and prioritize maintenance.” And therein lies the payoff. “If we can have information-centric maintenance related to product, quality, productivity, safety, delivery and also supplier quality,” says Lee, “then the intelligence of maintenance can impact much higher value.”
C. Kenna Amos, firstname.lastname@example.org, is an Automation World Contributing Editor.
Center for Intelligence Maintenance Systems
Siemens Energy & Automation Inc.
Bosch Rexroth Corp.