Things To Keep In Mind
Ultrasound listens to frequencies above the range of human hearing – most frequently in the region between 30-40kHz. Ultrasound is great for listening to friction and to impacting. If I carry out an FMEA exercise and determine that friction or impacting are likely indicators of failure, then ultrasound is the ideal detection tool.
Clearly, the detection of these types of behavior is not limited to rotating equipment, and so, neither is ultrasound. It should be equally clear that some defects will be trendable, in other words there will be a progressive deterioration, and some will be binary, the defect is either there or not.
The dBµV scale continues to be misunderstood. The depth of this misunderstanding is best indicated when people talk about trending. Only a few weeks ago, I heard about a presentation where, to paraphrase, it was stated that 72dB was 20% higher than 60dB. This is probably incorrect. In the case of an instrument using a dBµV scale and a decibel calculation of 20log10 (V1/V0), this 12dB difference would correspond to a factor of 4, which is 300% not 20%.
When we look at a trend we automatically perform these ratio calculations between the new reading and the previous reading. This ratio calculation is invalid when using a decibel scale. It is the absolute dB change which is important. When using a 20log scale, the difference between 20 and 26 is the same as the difference between 60 and 66, namely 6dB which is a doubling of amplitude.
To make matters worse, only a few suppliers in the world of ultrasound actually quote the reference value upon which their dB scale is based. This allows for a significant amount of confusion – a confusion which has, sadly, been exploited by some salesmen. It goes like this: take a reading with two instruments and the one with the higher reading must be more sensitive. Not necessarily so. If I were to take a measurement with an instrument measuring in dBµV and then re-scaled the instrument to read in dBnV, my voltage reading would increase by a factor of 1,000 which is 60dB. Have I changed the sensitivity? No, I haven’t. All I have done is re-scale my reading. So if one system is measuring in dBµV and the other is measuring in dBbananas, how can you possibly conclude that one is more sensitive than the other, just because the reading is higher? Neat trick, but it really is just that – a trick.
Another thing to keep in mind is variable frequency ranges. Lots of manufacturers allow the user to change the mixer frequency setting in the heterodyning circuitry (read Part 1 if that meant nothing to you). This was, and still is, most useful for trying to tune out the narrow frequency range of a parasitic airborne noise that is drowning out the broader frequency range generated by an air leak. The air leak will have energy of this much broader region, so my moving away from the parasitic noise, you should still be able to pick up the air leak.
To apply the same approach to a contact measurement is much more dangerous. First, the frequency response of the typical contact sensor is not as wide as the typical airborne sensor – in fact some designs these days are resonant giving a stable, high sensitivity of a fixed and narrow bandwidth. Secondly, the transmission of the high frequency ultrasound can vary with frequency and, lastly, the frequency range of the ultrasound source is unknown. This all means that comparison of measurements taken at different frequencies is not a good idea. Some ultrasound systems allow you to change the frequency range in a discovery mode only while others allow you to store data captured at these uncontrolled frequencies. Combine this lack of control with multiple users and you have recipe for confusion.
Ultrasound is evolving. There is considerable interest these days in the diagnostic capabilities of ultrasound. Apply signal processing principles to ultrasound and you have a very powerful analysis tool which will complement vibration analysis.
Valves (Blockage, Leakage and Cavitation) — Valves play an important part in plants and valves come in all sorts of shapes and sizes. There are some common failure modes in valves which are detectable using contact ultrasound - namely, blockage, leakage and cavitation.
Ultrasound is great for listening to friction, turbulence and impacting. When fluid flows in a pipe or through a valve, the fluid molecules rub against the side wall of the pipe or valve. This rubbing, synonymous with flow, generates ultrasound.
By applying a contact ultrasound sensor to a pipe, and depending upon the material that pipe is made of, it is possible to hear fluid flow in that pipe. By implication therefore, the absence of flow – a blockage – is also audible (or not, if you see what I mean).
The ability to detect very low levels of flow, leaks, depends upon many factors. The material of the pipe or valve, the true sensitivity of the ultrasound system, the pressure in the fluid and the type of valve will all have an impact upon the audibility of an internal valve leak.
Cavitation in valves, just as in pumps, can be devastatingly destructive. Cavitation is associated with the formation and then collapse of vapor bubbles in a flowing fluid. It is the collapse of the bubbles which is destructive. Physicists tell us that when a bubble collapse it creates a jet. These jets can have velocities of 100m/s (over 200mph) which will generate pressures at the impact surface in excess of 1,000N/mm². These events take place in fractions of a second. The cumulative effect is damaging to the impact surface, the valve or the impellor of a pump. These events however do generate quite large ultrasonic signatures, probably making valve cavitation the easiest of the three defects to detect.
Internal Leaks in Cylinders — Following on from valve leakage, comes internal leakage in hydraulic and pneumatic cylinders. Operational failure of cylinders can range from being a nuisance to being the cause of catastrophic failure and perhaps even the loss of life.
Cylinder inspections in situ, in operation, is therefore highly desirable, but in many cases, access is limited, or the plant is noisy. So there is no easy way of knowing if a cylinder is passing.
An ultrasonic detector with a contact sensor does not pick up the ambient audible noise. It if focused purely on a narrow range of inaudible ultrasound. The hissing sound of a passing cylinder is quite distinct against this background noise.
Some companies have become so successful using contact ultrasound in this way that they have prepared detailed written procedures for their maintenance people to follow. Their test procedures not only cover troubleshooting methods, but also recommendations for routine data collection on cylinders which allows the maintenance team to build up that all-important sensitivity to change. In this way, the onset of a leaking cylinder is handled within their predictive maintenance regime in much the same way as they would look after their bearings.
Steam Traps — Steam traps are a commonly used automatic valve. Steam systems have a very important role to play in the operation of many plants as well. Despite this important role, so many organizations have little in-house maintenance in place to maintain these assets.
Frequently, the only care the steam system gets is an annual inspection by an outside contractor or supplier. Of course, this is better than nothing, but routine inspection of steam traps by your own maintenance people can save a fortune in terms of steam loss, water consumption, chemical consumption and energy.
It is not uncommon to find production processes faltering because the steam heating or cooking system cannot reach or maintain a required temperature. Once again, the same ultrasonic inspection tool with the same contact sensor comes to the rescue.
Contact ultrasound, in conjunction with temperature measurement, is the industry standard method of inspecting steam traps. Listening to the ultrasound signal generated by a healthy steam trap, you will hear the trap collect some mixture of condensate and air and then you will hear the steam trap discharge. The amount of time it takes for an individual trap to collect and to purge varies widely. Some traps will cycle more than once per minute while other traps may take over 10 minutes to collect. A detailed knowledge of the steam system being tested is, therefore, vital to the success of this inspection procedure.
Traps tend to fail in fairly simple modes – stuck open or stuck shut. There are some subtleties of course, but we will save that discussion for another time. The common aspect of both of these failure modes is the lack of change when you are listening. A passing trap will generate a constant loud noise caused by the turbulence and noisy steam. A blocked trap will make almost no noise – just the mechanical noises carried in the pipework itself.
As with all predictive processes, success lies in sensitivity to change. The more data there is defining the normal operating condition, the more likely you are to spot a small change indicating the onset of failure.
One common criticism of this inspection procedure is that it is subjective – that it requires an “experienced” ear to discern good and bad operation. There is a small amount of truth to this. The process has been subjective thus far, primarily because of the absence of any methods for objectively recording the dynamic signal associated with the collect and purge cycle. The newest generation of ultrasound technology has evolved to overcome this problem. Using advanced signal processing allows this type of instrument to record events of virtually any desired length. This scalable, digital data acquisition means that an entire cycle, or even multiple cycles, can be recorded and then objectively compared. For example, in Figure 1, you can see that the peak amplitude is virtually the same, but the nature of the collect and purge cycles is totally different. With these measurements, we now have baselines and objective peer comparison to add to just listening.
Bearing Lubrication — In some respects, it seems that bearing lubrication has not altered much in the last two millennia–Roman historians document using animal fat to lubricate cart wheels for instance. Sadly, while lubricant technology has improved quite significantly from animal fats to organic oils to synthetics, via additives on the way, the way we use these highly sophisticated compounds has not really changed at all. Excessive lubrication probably kills more bearings than under lubrication. It certainly increases rather than decreases the friction in the bearing.
For many years now, Ultrasound has been providing a way forward – a means of greasing rolling element bearings on demand and to ensure that the bearing being greased receives only the amount of grease required and no more.
This is an ideal application for trending ultrasound. As friction increases in the bearing, the ultrasound produced will also increase. It is not entirely so simple, because an increase in friction can be caused by too much, just as easily as by too little, lubrication.
Paul Klimuc, a well-known personality in the world of ultrasound, has a wonderful analogy which compares lubrication with walking in a swimming pool. To paraphrase, there are three states:
1. The pool is empty, you can walk up and down but you cannot slide.
2. There is a thin film of water on the bottom, now you slip when you
walk – you are aquaplaning, and you might even be able to slide
from the shallow end to the deep end.
3. The pool is full, you cannot slide on the bottom and the drag
caused by the water makes walking very difficult.
This simple analogy beautifully illustrates the difference between a dry bearing, an optimally lubricated bearing and an over lubricated bearing.
There is a need for trending in this application. Where does the myth that all bearings consume grease at the same rate originate? Nobody really knows and yet this myth is continually propagated in industry. If it is a myth, how does time-based lubrication work? How much grease does that specific bearing need? It’s an impossible question to answer without measurement. If you have not calibrated your grease guns, how much grease are you putting in? It’s starting to look a little bit too random, isn’t it?
Consider the ultrasound alternative. Measurement on each bearing, trending the results and only applying grease to those bearings which measurably need it. To go one step further, to answer the big “how much?” question, you can use the same ultrasound system.
By listening to a bearing with an ultrasound system, you can hear and measure the benefit of the grease going into the bearing. With a simple procedure you can follow the improvement down to that optimum point, that sweet spot, where you have an optimally lubricated bearing.
Many might say, “It takes too long to do that!” Oh, Really? Consider the time and the cost spent in incorrectly, infrequently and/or sporadically over greasing bearings with the consequent consumption of grease and time spent replacing bearings which have prematurely failed. Compare this with the time spent gathering routine ultrasound data, trending that data and applying the optimum amount of grease only to those bearings that need it.
Does it really take too long? No, not really. You will save time and money with the consequent reduction in grease consumption and the extended bearing life.
Using the signal analysis approach to bearing lubrication allows us to produce time signals like the one shown in Figure 3 that I recorded during one ultrasound lubrication procedure implementation training exercise. How’s that for a before and after?
Slow Speed Bearings — This topic was covered in some depth in the Aug/Sept issue of this magazine. To summarize, since ultrasound is listening to friction and to impacting, the rotational speed of the shaft is not really significant. In my own experience, bearings with rotational speeds below 1rpm can be inspected and defects identified with ultrasound in a matter of minutes, compared with the hours required to do the same with specialized vibration analysis equipment.
Here once again, the use of scalable time signals gives us the opportunity to perform a peer comparison to identify the potential defects.
Motors, pumps, gearboxes, soft foot (before and after) — Apart from the (hopefully now) obvious applications for contact ultrasound, there are many other potential defects which can be identified or investigated.
There is a lot of work underway using ultrasound, particularly using dynamic ultrasound, as an inspection tool for gearboxes. Once again, we are listening for two possible problems – friction (tooth rubbing) and impacting (chipped tooth). The ability of ultrasound to listen only for those high frequency phenomena, while ignoring the rest of the background noise generated in the gearbox, gives this approach a clear advantage over vibration.
I have already discussed cavitation in this article – a common destroyer of pump impellors. In Part 1 (June/July Issue), I also discussed using airborne ultrasound to listen to the rubbing in a misaligned coupling and the chatter of a loose coupling.
The two signals in Figure 4 highlight another advantage to scalable dynamic data – the before and after. This is actually a soft foot problem on a fan bearing. As usual, this boils down to a bad design. The bearing was on a plinth, there was no hole cut in the long side of the plinth, only in the short side underneath the pulley. This was the non-driven bearing, roughly 1 meter away from the front face where the opening was. One man with normal length arms could therefore not reach underneath with a wrench to hold the retaining nut of the bearing housing and at the same time tighten the bolt from above – the result: soft foot.
While performing a contact ultrasound inspection of the fan bearings (again during an implementation training exercise), an abnormal noise could be heard on this bearing – a noise not present on the drive end bearing. It was not the usual hissing or crackling noise associated with poor lubrication. It was a clatter, and obviously a periodic noise. Capturing the time signal and analyzing it showed the repetition frequency related to the shaft rotational speed. Closer inspection with feeler gauges showed that the clatter was caused by impacts of the loose bearing housing either against the bolt head or against the plinth. The repair was a two-man job. But it was only tightening a bolt.
A further scalable time signal and we have an example of one of the nicest comparisons there is in the predictive world – a before and after comparison of a successful repair.
Electrical Inspections — Even within the electrical distribution world, contact ultrasound has been recognized as a useful tool.
Some years ago, EPRI produced a document which indicated applications for ultrasound in high voltage electrical applications and particularly associated with transformers. One of these applications was incredibly simple – use an ultrasound system with a contact probe to listen for the impacts generated by loose parts in the tank.
Ultrasound has an important role to play in supporting infrared inspections in electrical panels too. Airborne ultrasound is frequently used to listen for arcing or tracking inside a panel or cubicle. If that panel is watertight however, it is unlikely that there will be an air gap through which such an ultrasound signal could escape. In which case, it is recommended to use contact ultrasound either on the door or on the side of the panel to hear what is happening inside.
Procedurally, this is quite simple and akin to a visit with your general family practitioner where a stethoscope is used against your chest to listen to your heart and breathing. For electrical panels, a magnetic contact sensor works perfectly. Position the sensor in the centre of the panel and, with headphones in place, adjust the sensitivity (amplification) of your detector. Ultrasound produced by an electrical fault is typically an airborne phenomenon however we know from acoustics that sound, and ultrasound, can transfer from one medium to another. So tracking or arcing inside a panel starts out as airborne ultrasound, and the direct and reverberant components of this sound inside the panel make it possible for ultrasound sensors to capture part of the signal produced by the electrical fault as these tiny signals induce a corresponding vibration in the door or walls of the panel.
Normally the inside of panels should be quiet, or some behave in a rhythmic manner as contactors come in and out. Listen for tell-tale buzzing, crackling, and popping signals consistent with arcing, tracking, or even corona discharge if the voltage is high enough. Now measure the dBµV of the signal. The SDT170 detector I have used will read anywhere from -5dBµV to +5dBµV for a quiet panel and if there is something unusual going on inside expect readings as high as 15dBµV.
As a brief aside, many people ask me why their SDT devices read a negative value. There is a simple answer to that. The decibel reference for this manufacturer is published and well known as 1µV = 0dB. If the panel is quiet and the sensor were only producing ½µV for example, this would correspond to -6dBµV.
If a high reading is detected and anomalies heard in the headset you may want to capture a dynamic signal and analyze in the time and frequency domain to determine the type and severity of the problem. Since electrical panels are often named and categorized for the purpose of maintenance, why not set up a systematic survey of all panels and capture static (dBµV) data on an interval basis and dynamic (5-10 second signal recording) data for those displaying potential problems?
Contact ultrasound has a wide variety of applications in the world or maintenance. It can be applied at a very simple level - touch a probe to a surface and listen - or it can be applied at a much higher level as a diagnostic tool. Many times I encounter power users of other PdM technologies (Vibration/Infrared/ Oil Analysis), but find a general lack of understanding in the differences between contact ultrasound and contact vibration. My explanation starts by going back to understanding the benefit of FMEA in predictive maintenance programs. What is the point of unfocussed data collection? Why take readings to look for problems that may not be detectable using that particular method?
Vibration accelerometers are capable of measuring acceleration over a huge range of amplitudes and frequencies. In my early years in vibration measurement, I used to work with accelerometers with a mounted resonant frequency of 500kHz and a measurement range of over 100,000g. In the world of predictive maintenance, most people are using an accelerometer on a magnet with a mounted resonant frequency of perhaps 5kHz. How do you measure an event occurring in the 20-50kHz region with such a sensor?
The vibration world is stuck to the spectrum like glue. But Fourier’s mathematics were fundamentally based on periodic signals. It is quite common to see vibration measurements where the spectrum has clearly been wrongly applied. So, if your FMEA suggests that defects will be detectable as random, transient events or as high frequency noise, your trusty mag-mount sensor and FFT are not very likely to find them.
Ultrasound sensors measure sound pressure waves produced in air, liquids and solids. These sound pressure waves are the result of friction and impacting.
Understanding the difference between ultrasound and vibration measurements makes all the difference in the world when it comes to applying the right technology for the right task. Knowing that vibration is normally good for repetitive events, when and where should it be applied? Knowing that ultrasound is normally good for random or transient events, when and where should it be applied?
Are vibration and ultrasound competitors? No, not really. (Who uses a vibration data collector on valves, steam traps, hydraulic cylinders or loose part tests?)
Are there problems that I can only find with one or the other? Yes. Are there problems that I can find with either? Also yes. Are they complementary technologies? Absolutely yes.
As the technology continues to evolve, the ability to capture ultrasound signals and analyze them will open up more and more applications in rotating, linear and flow applications. I hope to write Part 3 on this subject in the near future.
Tom Murphy is an Acoustics graduate from Salford University and has 25 years experience in the world of industrial ultrasound and vibration measurement – 15 of those years have been involved with the use of Operating Deflection Shape techniques in the paper, printing, petrochemical, power generation, pharmaceutical and food industries. Tom is the Managing Director of Adash 3TP Limited, based in Manchester, England, a Company specializing in the application of vibration, infrared and ultrasonic technologies to improve maintenance. More info can be found at www.reliabilityteam.com and Tom can be contacted at +044 161 788 9927 or at email@example.com