Ultrasound inspection offers a unique dual-purpose proposition for condition monitoring, contends to Alan Bandes, UE Systems, VP marketing: it offers both a stand-alone inspection technology and an effective screening tool that can speed up the inspection process and help inspectors determine effective follow-up actions for mechanical, electrical and leak applications. He adds: “Whether you refer to proactive inspections as predictive maintenance or condition monitoring, the goal is the same; to note a deviation from a normal or baseline condition in order to determine whether or not to take corrective action in a planned orderly manner and to prevent an unplanned incident.”
Ultrasound simply means frequencies higher than the audible limit for human hearing, which is approximately 20 kHz. Conventional vibration analysis typically operates at significantly lower frequencies. For example, common AC motors running at 1,800 rpm and 3,600 rpm will produce characteristic unbalance and angular misalignment frequencies of 30 Hz and 60 Hz respectively.
When many people think of ultrasound, they think of the non-destructive testing (NDT) used to identify cracks, porosity or delamination. The NDT application of ultrasound uses an ultrasonic transducer which sends bursts of highly-directional ultrasound into the part being tested. An ultrasound detector can then measure the distance to a free edge by recording the time taken for the signal to be returned. When used for condition monitoring, only a detector is required. The ultrasound is emitted by the components of the machinery being tested.
Low-frequency vibration typically produces whole-body motion of mechanical components. As frequency increases, deformation of components starts to occur and resonance may be encountered if the driving frequency coincides with a resonant frequency of the structure. At the even higher frequencies of ultrasound, which characteristically feature shorter wavelengths, sound waves propagate through solids as deformation at the scale of the material’s crystal lattice.
The high frequency of ultrasound does not mean that only very high frequency disturbances are detected. Often, the actual fault condition may occur much less frequently, but the carrier frequency is very high. A simple way to understand this is to consider different drums. A bass drum has a low carrier frequency and a snare drum has a high carrier frequency, but they can both be used to keep the same rhythm. Considering how this applies to condition monitoring, a bearing defect might produce a burst of noise at a regular interval of 112 Hz. This could drive a 112 Hz vibration. But it is also possible to listen to the actual sound being produced, which could be in the ultrasonic spectrum, over 20 kHz. Often these bursts of ultrasonic noise are detectable before vibration is noticeable.
TIME SERIES AND SPECTRAL PLOTS
There are two main ways that vibration and ultrasound data can be viewed. The first is a time series plot, in which the raw data is presented as a waveform, with time along the X-axis and amplitude along the Y-axis. This can be useful to spot any patterns and observe the randomness of the signal. Spectral plots are obtained by performing a Fast Fourier Transform (FFT) to identify different underlying frequencies (see also www.is.gd/uwosem). The spectral plot shows frequency on the X-axis and amplitude on the Y-axis. Spikes on the spectral plot indicate the specific frequencies driving vibration, while a relatively flat spectral plot indicates white noise.
In ultrasonic condition monitoring, a consistent level of white noise generally indicates correctly operating machinery. This can be seen in a time waveform plot but is probably most clear in the FFT results displayed on a spectral plot.
Because ultrasonic condition monitoring is generally listening for an deviation from white noise, rather than for specific characteristic frequencies, simply listening to it can often be effective. Rather than plotting the signal, software can demodulate it into the audible spectrum so that a user can listen to it, a process also known as heterodyning.
DETECTING METAL TO METAL CONTACT
Ultrasound is particularly good at detecting metal-to-metal contact, or friction, which typically results in frequencies around 30 kHz, within a 4 kHz bandwidth. These sounds may be relatively low amplitude, not producing any noticeable vibration in the structure, yet still be clearly detectable in the ultrasound spectrum. “This makes ultrasound ideal for detecting the early stages of lubricant breakdown, before significant damage is done, allowing relubrication,” says Andy Mellor, managing director of Pragmatic Maintenance & Reliability.
Metal-to-metal contact may occur – and be detected by ultrasound – when levels of lubricant drop too low. One tool that takes advantage of this ability of the technique to rapidly detect lubricant condition is the ultrasonic grease gun. These lubricating devices have an ultrasonic listening device attached. When the grease gun is inserted into the grease fitting on a machine, any friction-related vibrations can then be detected. Grease is then inserted, one shot at a time. After each shot, the operator waits for the grease to work into the machine and then check the ultrasound reading again. When optimal lubrication is achieved, no further grease should be inserted. Using an ultrasonic grease gun in this way helps to prevent under- or over-lubrication. This reduces energy consumption and increases bearing life. It also reduces the risk of grease leaking and causing damage in other components.
DETECTING BEARING DEFECTS
In vibration analysis, bearing faults are diagnosed using knowledge of rotational speed and bearing precession rates. Using this information it is possible to predict the frequency of vibration that a bearing fault should cause. However, at the earliest stage of fault development, the defect may not impact consistently with every rolling element. This means that a consistent frequency is not produced, and the fault cannot be easily seen in a spectral plot. Simply listening to the demodulated ultrasonic spectrum, however, the fault will be clearly audible as crackling and popping. This can also be seen in a time series waveform plot.
Explains Andy Page, principal, Allied Reliability Group: “Some maintenance personnel follow a pre-set schedule under which bearings are replaced every one or two years. There are much better ways to determine when a bearing needs to be replaced, though, and one of the most reliable methods lies in identifying the presence of subsurface fatigue – small cracks, generally around 5-6 microns deep, which can be seen with the naked eye once the bearing is removed, disassembled, and cleaned. Subsurface fatigue is the first sign that a bearing is on its way to fatigue failure. While this wear and tear still cannot be seen or felt, the energy given off by the fatigue is ultra-high frequency, so ultrasound instrumentation picks it up very well.”
LEAK DETECTION
Leaks in pipework carrying gases and liquids can also be easily detected using ultrasound. Preventing leaks can produce significant savings. Even for compressed air, the cost of generation is often high, meaning that removing leaks is cost-effective. For other substances, there may also be significant environmental impacts caused by leakage; for example, natural gas is an extremely powerful greenhouse gas (see also article, pptk).
External leaks are best detected using airborne sensors, to literally listen to the sound of the leak. Because very high-frequency sound is highly directional, it is possible to separate these signals from background noise and to accurately locate the leak.
Leak detection in valves uses contact sensors. Typically readings are taken at five points: two upstream of the valve, two downstream of the valve and one on the valve body. If the downstream measurements show more noise than the valve body, this generally indicates a leak. The upstream measurements are made as a check, as sometimes noise from another source may be detected.
Although vibration monitoring is a well-established means of detecting and diagnosing problems with rotating equipment, ultrasound can often provide earlier detection. The level of training required to successfully detect faults is typically also lower. Ultrasound has additional applications such as checking lubricant level and detecting leaks.
“Ultrasonic acoustic emissions provide high signal to noise ratio, even from slow moving equipment, and require a lower level of skill to effectively utilize, but this method is not diagnostic of, and doesn’t enable proactive corrections for, conditions such as unbalance and misalignment,” concludes Mellor. “Vibration analysis has a lower signal-to-noise ratio for slow moving equipment, and requires a higher skill level to utilize effectively. However, it provides spectrum analysis, diagnostics and proactive corrections for many conditions. These are complementary methods that each have their place.”
BOX: OVEN BEARING FAILURE DETECTION
Slow speed bearings produce low levels of ultrasonic noise – but an ultrasound instrument can still reveal faults by allowing the bearing’s signature in the audible frequency range to be analysed, according to Christopher Hallum, UE Systems regional manager, UK & Ireland.
An inspection with an ultrasound instrument was carried out on a site with a newly-installed oven dryer. This was a large drum oven, about 20m long by 5m wide. It was rotated by four large motors, each of them having two large sets of bearings. The motors were rotating at a speed of around 7-10 rpm, so that this was a case of extremely slow-speed bearings.
An ultrasonic instrument was used to inspect all the bearings – almost all of them presented a smooth sound and a 0dB reading, except for one. On one of the bearings, the ultrasonic instrument was displaying 2dB instead of 0db. Also, the audible sound heard through the headphones was different: it was not smooth as in the other bearings and it presented a repetitive knocking sound.
Analysis of a grease sample from the bearing in question found particles of metal, confirming that the bearing had been damaged. On replacement, it was found to be in a poor condition.
Hallum points out that in applications where bearings run at normal speeds (above 25 rpm), ultrasonic recordings of noise levels can be compared against a baseline noise level, measured in decibels. But in slow-speed applications, defects are harder to detect by ear, so their identification requires recording of measurements and offline spectrum analysis of those recordings by computer.