“Young Charley” (photo above) in Part 4 (April/May 2014) is gone and now his job is being done with vibration sensors (accelerometers), computers, etc. We call this procedure condition monitoring (CM), condition-based maintenance (CBM), or machinery health monitoring (MHM). The general idea behind CM/CBM/MHM is:
Costs are high when no maintenance is done, as shown in the left peak in Figure 1. We run machines until they fail and replace machines as needed.
Costs are high when we maintain on schedule (preventive maintenance), whether anything is wrong or not. Too often, disassembly and reassembly create damage. That’s the right side peak in Figure 1.
Costs are lower when we maintain only when a known need exists, as the dip in Figure 1 reflects.
Figure 1: Concept of CBM/MHM
How do we know when need exists? Lacking “Young Charley,” we must use more modern (often computer-based) warnings and diagnoses. Figure 2 mentions some of the techniques. Vibration warning and diagnosis is popular and is our focus here.
In MHM, we basically compare today’s vibration spectrum (or “signature”) with other spectra obtained using the same instruments at the same locations last week, last month, or last year, or when the machine was first installed or immediately after an overhaul. We try to deduce internal changes (deterioration) from changes in the spectrum.
The goal of MHM is predictive maintenance. We want to be warned when overhaul or replacement of worn or damaged parts is becoming necessary.
As tooth condition deteriorated in the gear shown in Figure 3, the machine’s vibration signature changed, alerting maintenance that overhaul was necessary on the next weekend shutdown. Replacement gears and other parts were procured by Friday afternoon. That plant was lucky. Had the gear failed, the plant would have had an unscheduled shutdown at a cost of millions of dollars per day.
Figure 3: Gear failure
Digital Recording Useful
Could that change if vibration had been detected by the oversimple readout instrument suggested by Figure 4? No. Nor by the combined sensor and readout of Figure 5.
Figure 4: Ordinary voltmeter is of little value for MHM
Figure 5: Indicator containing accelerometer (Photo courtesy of IDCON)
Overall vibration, as displayed on any meter, would probably not change very much. The signature, the vibration spectrum, is needed to reveal subtle, but significant changes. To whom? To trained personnel, to a highly specialized analyzer, or to a specially programmed computer.
We have replaced “Young Charley” and his stethoscope with more modern data collectors, which usually contain digital recorders cable-connected to an accelerometer. These are carried from machine to machine, perhaps monthly, so each machine’s signature is recorded on schedule. Next, all machines’ signatures are downloaded into the maintenance department computer’s larger memory for storage and later comparison with earlier signatures. Spectral changes alert maintenance personnel to possible problems.
Figure 6 shows Equipment Reliability Institute (ERI) student Kevin White measuring vibration on an idler gear bearing of a Minton dryer at Fletcher Challenge Canada’s Crofton Pulp Company. Crofton’s predictive maintenance program includes analyzing dryer felt and dryer bearings for faults.
Figure 6: Checking paper production machinery (Photo courtesy of Crofton Pulp & Paper, Crofton, BC, Canada and Wilcoxon Research, Gaithersburg, MD. Photo by Douglas Schwartz)
Shafts turn slowly here: 24 RPM to 45 RPM. High temperature (+300° F on dryer bearings) vibration sensors are permanently installed and hardwired to monitor 100+ inaccessible bearings. (Drive gears prevent easy access to bearing locations.) Data is gathered at five junction box locations.
Instead of climbing over greasy surfaces 30+ feet above the ground, analysts (see Figure 7) use nine, permanently mounted accelerometers to gather data from important, but inaccessible points. Accelerometers are wired to a junction box. Analysts simply turn a switch at the box to cycle through the sensors. They monitor additional accessible points using magnet-mounted sensors and a handheld data collector. In total, they monitor up to 50 data points on a medium-sized press and up to 80 points on a large press. Stud-mounted sensors are installed on seven of the 14 forge hammers at the Tonawanda Forge, Tonawanda, New York, plant of the American Axle and Manufacturing Company. In Figure 7, a near molten billet is about to be stamped into a netshaped gear assembly for a major automaker.
Figure 7: Forging hammer vibrations (Photo courtesy of IMI Sensors, division of PCB Piezotronics, Depew, NY)
Let’s eliminate the walk-around job. Rather than carry the data collector and accelerometer from machine to machine throughout a large factory, let’s permanently attach accelerometers, and possibly other sensors, to vital machines. And let’s permanently install cables between all accelerometers and the central maintenance computer.
Is this costly? Yes. Permanent wiring also makes it difficult to move machines to other locations. Or a vehicle may need monitoring. In these situations, go wireless. Consider Figures 8 and 9 using wireless radio links between the machines and the central maintenance computer.
Figure 8: Wireless sensor
At the left in Figure 9 is a 2.4 GHz transceiver connected to two sensors. Batteries are drained only when the transceiver is interrogated. At right is a base station, which receives signals from up to 32 transceivers. The base station connects into the RS-232 port on the maintenance department’s computer.
Figure 9: Concept of a wireless sensor system (Image courtesy SKF)
The frame of an active machine is where sensors are attached to detect motion. At certain locations on that machine, there must be dynamic strain. Development work, mostly piezoelectric, is underway to harvest energy from that strain to keep batteries charged.
One expert is able to examine signatures and trends from a remote location, possibly from another continent. This expert can serve several plants, saving much travel time. The machines can alert this person when a problem arises. He or she can interrogate any machine of concern and, if necessary, dispatch parts and skilled labor to arrive within 24 hours.
MHM provides us with a good reason to learn about spectrum analyzers and spectrum analysis beyond the scope of this series. Over the past 50 years, as machinery speeds and vibration frequencies have increased, early analyzers have become technically obsolete, although many are still in use.
Think for a moment about the two spectrum analyzers each of us carries in his/ her head – the cochlea, as shown in Figure 10. Quoting Richard D. Lee’s letter to Electronic Design magazine, “The inner ear has resonant fibers in the cochlea that stimulate attached nerve endings at a level corresponding to the fiber response amplitude. An array of nerves scans the outputs from these fibers and develops what amounts to a Fourier transform of the received sound. If a single frequency is received, only one fiber has a max response. The array locates it and reports that to the brain.”
Figure 10: Spectrum analyzer in cochlea
Consider the similarity of the mechanical spectrum analyzer in Figure 11. Here, the slight vibration of a slightly unbalanced motor shaft running at 1,750 RPM is sensed. One particular reed, having an fn of 1,750 cpm, responds.
Figure 11: Mechanical spectrum analyzer
An array of reeds is seen in Figure 12. Individual reeds are hand-tuned by filing. The multi-reed mechanical spectrum analyzer might be considered a parallel analyzer. Each reed continually monitors a given slice of the mechanical spectrum.
Figure 12: Array of reeds
Wayne Tustin is founder and president of the Equipment Reliability Institute (ERI) in Santa Barbara, California. ERI operates worldwide, though Wayne, now 91, has cut back on overseas travel. Wayne has authored a 33 iBook series, available at Apple Bookstore’s http://goo.gl/xNWC49. Several expand upon the five parts in this Uptime series. Book #1 is free. www.equipment-reliability.com