Yet another O2 heater coil had failed. It had gone from the original precise helical shape to looking like a flat, wound garden hose. Our operators called the coil “pan-caked.” Pipe supports failed under creep. A number of engineering reports were written, including one by me, proposing solutions based on endless files of bone-yard measurements. In the end, we hired a reliability specialist from a refinery. Stan Stacy worked closely with me, eventually allowing us to develop a simple model that predicted coil failure. We couldn’t, however, achieve our ultimate goal of extending service life because we couldn’t prevent operations from overheating the coils.
Such situations confound predictive maintenance (PdM). The best that can be done is to predict failure and establish a reasonable mean time between failure (MTBF). So, where can a reliability specialist have a bigger impact?
A specialist lacks detailed knowledge of day-to-day operations. How equipment is used often is more important than the idealized concept considered by process engineers and the vendor.
At another company I worked at, we used magnetic-drive pumps for the first time with pickle liquor; we never expected the iron in the solution to demagnetize our cores. Early pump failures dismayed our vendor, which said it was impossible. The operators caught the problem first.
One approach may be to put simple tools in the hands of maintenance staff or even operators. The reliability expert then serves as mentor and guru. All of this takes a serious investment of resources, both material and personnel. The last time I checked, a good thermograph gun cost almost $15,000. Given the maintenance and calibration costs of testing instruments, and training, it may be better to rely on contractors unless overhead costs such as per diems are too expensive. A contractor may be especially useful for elaborate testing such as vibration analysis, which has proven highly useful for rotating equipment; however, expertise comes with a high cost. For more about such tests, see June's “Bolster your condition monitoring toolbox.
The unspoken question is: Is a PdM program necessary for all situations?
If the technology is new or unfamiliar, the simple answer is: Absolutely. For older, established processes, the decision is harder. Let’s consider the role of your reliability specialist. His first chore will be establishing failure curves for major equipment; the “bathtub” curve and increasing wear curve probably are the most common. After identifying the mode of failure, the specialist will work with a process group to create several lists: 1) a maintenance schedule; 2) ideas for extending equipment life; 3) equipment that can benefit from continuous monitoring; 4) assets that may best be treated as throw-aways; and 5) unreliable equipment that could be replaced by substituting a unit operation. Lists are prioritized by criteria: most likely, most expensive and most hazardous. Of course, safety trumps cost.
The maintenance schedule assigns work into categories of daily, weekly, monthly, quarterly, annual, etc. Simple ideas such as verifying bolt tightness — 60% of rotating equipment downtime is caused by loose fasteners — and checking lubrication can be immediately implemented. Other ideas require capital. Failures may seem random but usually aren’t; constant monitoring is required to determine failure mode. Emerging developments such as passive wireless sensors may be useful. Throwaway equipment can reduce maintenance. For instance, sometimes it’s cheaper to discard a pump than to repair it, especially if you have high labor costs.
Finally, it’s important to recognize when a process won’t work. If a unit operation depends on unreliable equipment, it’s time to consider an alternative. Engineering decisions are often based less on what has the most promise than on whether a technology has been employed elsewhere. Who invented the tray tower and why hasn’t it been replaced with something more efficient?
A preventive maintenance (PM) program may be unnecessary or outgrow its usefulness in some circumstances. Here’re a few examples: 1) new equipment rental; 2) newly commissioned plants; 3) where equipment truly follows a random failure mode; 4) when a plant will be abandoned; or (5) if a company can’t afford PM. Be careful! Several modes are mistaken for random failure ― get complete data during the entire life of equipment. PM is wasted for random failure.
Where a company can’t afford the expense of a proper PM program, consider a planned component replacement (PCR) schedule. A consultant can help you establish such a program. Periodic review is crucial if this schedule is to be successfully implemented. So, too, is careful recordkeeping. For example, our roaster ran with the same thermocouples for seven years before they began to fail. I doubt if anyone but me thought to write an amendment to our PCR schedule. Barebones maintenance depends on good documentation. Without it, PCR maintenance can devolve into run-to-failure maintenance — which is to say, no maintenance at all.