In many factories with discrete PLC controls, 4-to-20-mA loop signals are part of the system. Now, there are new vibration monitoring systems that tap into this network, and report on a variety of vibrational situations, without requiring resident technical staff to assist in data interpretation. Too, these systems allow plant engineers to use a hybrid of online and offline approaches. In short, online vibration monitoring is installed on critical plant machines, while non-critical machines are left to occasional checks with handheld data collectors.
Let's back up a minute and review the two most common vibration monitoring systems. The first is portable analyzers. Their advantage is that they can connect to machines when and wherever a problem is suspected. So, regardless of their initial cost, many machines can benefit from one unit. These analyzers can also be used to diagnose any vibration issue related to balance, alignment, bearing conditions, or gear mesh, for example. Many models even include a tachometer input that facilitates field balancing software.
During the fifty-plus years that vibration analysis has been conducted on rotating machinery, portable analyzer technology has become increasingly smaller, lighter, and more powerful. In fact, most units are now handheld devices. However, the entry cost to an in-house vibration program has been increasing. Twenty years ago, data collectors were simpler devices, and full software/hardware suites cost less than $10,000. Computers were more expensive, but this was offset by relatively inexpensive maintenance contracts. Now, annual software maintenance costs have escalated, and initial equipment and training costs can exceed $40,000.
Portable analyzers require at least one skilled analyst to perform the critical diagnosis of machinery. Additionally, the data collection is a labor-intensive operation. But with their portability, it is possible to tailor the frequency of machine visits to the threat of machinery and severity of mechanical faults.
Online dynamic monitoring
During the 1950s and 1960s, Bently Nevada Corp. (now of GE Energy, Minden, Nev.) pioneered turbomachinery vibration monitoring systems, based on eddy-current proximity probes. Today, no self-respecting company would dare operate these critical systems without having such probes installed — and no business insurance carriers would insure an operation that didn't have these safeguards against catastrophic damage in place.
Online monitoring systems have made great strides elsewhere too. Today, most online systems consist of piezoelectric accelerometers for vibration probes, and specialized computer-based data acquisition systems for gathering, storing, and archiving FFT data. These FFT-based data systems exist apart from plant computer controls, because generated data files are so massive and specialized. Only trained vibration analysts can review, interpret, and analyze the data collected. Too, system installation costs are high because of the capital cost of the hardware and installation labor — about $1,000 per vibration point.
But the real advantage of online dynamic vibration monitoring is that it acquires data almost continuously. Its hardware system architecture usually includes a single acquisition channel that is multiplexed across many vibration sensors, for a scan rate varied by a system scheduler. Usually, a full set of time and spectrum data is required for each monitored point about every three to six hours. The result is four to eight data sets each day, for a mind-boggling 1,000 to 2,000 data sets acquired for each point every year.
Through the 1990s this was one of the great limitations of online systems: When managing this data, older computer systems had to sort, discard, and move portions offline to cap data-set file sizes. But with modern computing memory, this is no longer a problem. The cost reduction in memory has also enabled online systems that continually change memory to only store information if some sort of catastrophic machine fault occurs. This pre-event memory of machine conditions just prior to catastrophic failure is powerful information that can be instrumental in assisting analysts determining the root cause of a failure.
Resulting data is passed through sorting-based algorithms established by a skilled vibration analyst. Problems are automatically flagged 24 hours a day by the system and identified on the screen of a control computer without human intervention. So, online dynamic vibration monitoring provides mechanical fault detection at the earliest detectable time.
Even so, because of their disadvantages, online systems are usually installed on only a fraction of machine points that could benefit from continuous monitoring.
Hybrid: 4-to-20 mA sensors
Now, let's discuss hybrid vibration monitoring. An alternative to FFT-based online monitoring systems is vibration transducers that incorporate signal processing to produce a conditioned vibration signal. These devices can be directly powered by PLC systems and produce an overall signal proportional to machine vibration levels. Another variation is DIN-rail-mounted modules that use the traditional dynamic accelerometers of portable data collectors and online FFT-based vibration systems. Again, both of these generate 4-to-20-mA process loop signals that serve as inputs to PLC systems.
This is useful, because DCS/PLC systems are already widely used to control plant operations and provide control loop functions, data trend displays, and long-term data archiving. The technical expertise to maintain these control systems exists in many plants already; in fact, most plants depend on such systems for their operation. Adding overall vibration data to these systems is a logical, low cost extension of the existing system: It only requires a vibration sensor/transmitter and an available 4-to-20 mA analog input channel on the PLC. Too, the cost to add a channel of vibration data is low — less than half the cost of an on-line monitoring point.
Interpreting their signals
Because in most applications, 4-to-20-mA vibration sensors report overall machine vibration energy, ISO 10816 can serve as a guideline for machine conditions based on published levels. Or, if users know manufacturer-recommended maximum vibration levels, those can be used to establish allowable vibration limits.
4-to-20 mA-based vibration data requires no complex analysis, because current level from the sensors is directly proportional to vibration. Now, these overall levels do not change significantly in the early stages of mechanical fault development. That said, they usually change significantly prior to actual failure.
An astute mechanical maintenance department can generally pinpoint problems once alerted to a developing fault. Otherwise, some designers install multiple sensors on machinery sets, with those additional sensors tuned to different frequencies. With enough band-limited sensors or band-limited signal processing, specific problems can be diagnosed automatically.
Sensors for 4-to-20 mA loops
Transducers for vibration sensing offer several monitoring options. They can forward-loop output as a processed acceleration signal, or (more commonly) as a velocity vibration signal. In either case, at the heart of these transducers are piezoelectric acceleration-sensing elements and electronic circuitry, powered by 24-Vdc plant power. The internal electronic circuitry allows the sensor to be configured in several ways.
Output can be calibrated in true root-mean-square (RMS) amplitude or in (RMS-derived) peak output. The latter is accurate for sinusoidal signals. Other units offer true peak output, suitable for detecting early stages of bearing wear, for example. Sensors can also be customized with built-in high and low-pass filters to allow engineers to focus on specific ranges of frequencies. With these filtered versions, users can more accurately identify specific vibration problems.
Some sensors offer both 4-to-20 mA processed output and a buffered dynamic signal output from the built-in piezoelectric accelerometer. These enable traditional vibration analyses with sensors already used to trend overall levels — for two monitoring functions. Some models also have a third option for temperature measurement; this is a powerful combination for confirming developing mechanical faults.
4-to-20 mA loop-powered sensors are simple to install. A single threaded stud is inserted into a hole drilled and tapped in the machine case. Shielded, twisted-pair wires feed 24-V power to the sensor, and then carry 4-to-20-mA signals to the DCS/PLC analog input.
Making use of what's there
In some plants, accelerometers are already permanently installed on machines to aid the vibration data collection process. These are often located on equipment in remote areas, or tucked away for safety reasons. Unless the data collector is actively collecting data, these sensors sit idle. Well, these preexisting sensors can be used for gathering 4-to-20-mA vibration signals, too.
Vibration transmitter modules can accept regular accelerometer signals as input and provide the raw data as output on a BNC connector. The same transmitter provides the necessary electronics to convert the signal into a 4-to-20-mA signal, making it available to the plant PLC. In this way, existing sensors are used for collecting routine route data on a periodic basis while providing continuous 4-to-20 mA data to a PLC. DIN-rail mountable, units can be installed close to the machinery in NEMA 4 boxes.
Transmitter modules offer more power for signal processing than the standalone sensor versions. They implement stronger filtering with more filter options, so they are capable of trending entire vibration spectra or a small range within a spectrum. Multiple modules can be daisy chained to one accelerometer to process different frequency bands, to allow monitoring of both acceleration and velocity.
Five technologies are used in applying PdM. So if you add these systems to your factory floor, who makes sure they get used? All require skilled practitioners.
u Vibration analysis requires technicians with years of experience, and full-time analysts with the knowledge to correctly diagnose problems. Both positions are increasingly difficult to fill.
v Technicians can be trained to operate thermal imaging (infrared) cameras in a relatively short time. The analysis of images is less complicated and more intuitive than analyzing vibration spectra. The personnel that operate imaging cameras can do so on a part-time basis, often as a complementary task to acting as factory electrician or mechanical technician.
w A similar case can be made for motor circuit evaluation. Here too, operators can collect data on a part-time basis.
x Ultrasonic equipment is even easier to learn to use. The interpretation of ultrasound data is intuitive and can easily be learned. The training period is short and the techniques are not easily forgotten.
y Oil analysis is a little different than other PdM technologies. Properly collecting oil samples is not complicated; however, its analysis is a different matter. Complete oil analysis requires rather exotic and expensive equipment, such as gas and infrared spectrometers. Particle analysis also requires proper training and the use of microscopes.
Taking it one step further
Control systems in plants have evolved into modern computer control systems: Many older local relay-ladder controls have given way to miniature PLC units, and integrated, computerized distributed control systems (DCSs) maintain equipment operation and production flow.
As these real-time plant controls continue to evolve, so do maintenance operations. It used to be that maintenance was performed only when equipment actually failed — an approach called breakdown maintenance. But now, regular maintenance and refurbishment is more common. Called periodic maintenance or preventive maintenance, the idea is to have most plant equipment operating most of the time. This carries the expectation of some equipment failures, but that is part of the tradeoff to minimize maintenance expenses.
With the increased use of the Fast Fourier Transform (FFT) mathematical method in the 1960s, portable spectrum analysis equipment spread. For the first time, the mechanical condition of an operating machine could be deduced. Early users often hauled these systems around in vans equipped with large dc-to-ac converters, and then ran long cables to equipment to make measurements. Though clunky, these systems laid the groundwork for today's predictive maintenance (PdM) of machines.
When truly portable FFT-based data collectors became available in the 1980s, vibration tools for diagnosing machinery faults proliferated, especially in petrochemical, paper, and power industries. Instead of waiting for a machine to fail before working on it (as in breakdown maintenance) or performing maintenance on a machine regardless of its condition (as in preventative maintenance) the idea of performing maintenance on equipment only when it exhibits faults — predictive maintenance — has taken hold. Now, using PdM to perform maintenance on machines when they exhibit signs of mechanical failure is called condition-based maintenance.