This presentation explores the ways in which modern ultrasonic technology can square the circle of maintaining simplicity for those who need it while at the same time providing sophistication for those that want it.

Introduction

There should be little doubt left in the world of Predictive Maintenance that ultrasound is an important element of a complete PdM programme. Unfortunately, there is. Why? It is clear that there is a growing recognition of the power of ultrasound, but the ability to find problems quickly, easily and without recourse to a computer seems to be a little hard for some to accept.

The problem tends to be more associated with contact or structure borne ultrasound than with airborne. Ultrasound is the undisputed champion for finding compressed air leaks, steam leaks, vacuum leaks and electrical defects such as corona, tracking and arcing.

It seems that when the world of PdM talks about structure or contact ultrasound it is almost always in the limited context of bearings. There is more to life than bearings. There is a whole lot more to PdM than bearings too. How often do you read that?

The typical maintenance engineer is interested in far more than just bearings. Valves, steam traps, hydraulic systems, cylinders, sealing, tightness testing, conveyors, actuators and many more such items are just as important as motors, gearboxes, pumps and bearings. I suppose this might be summarised as "not all problems rotate".

In a time of tough economic times, the return on investment is a very important metric. It stands to reason then that the more problems you can fix with the one tool, the greater will be your ROI. The challenge then is broadening the application range within your facility.

In this presentation we will compare the methods and relative merits of using ultrasound to find problems in a simple way, in a loosely structured way and in the highly structured regimen of routine data collection.

A review of the technology

Ultrasound has been with us a long time. It was discovered that bats used ultrasound for navigation in the late 18th Century. Instruments using ultrasound for airborne leak detection have been with us since the early ‘60s. The general acceptance of this technology compared with vibration and infrared is still relatively poor - which is surprising since there are problems in the world of maintenance which cannot be found with vibration or infrared and can be found with ultrasound. There are also problems which can be found more easily with ultrasound than either vibration or infrared.

What does ultrasound do? Effectively, ultrasound can "hear" friction, turbulence and impacts. If you have performed an FMEA (and that's an unfortunately big "if" in most predictive departments) and that exercise shows up potential defects whose characteristics are friction, turbulence or impacts, then ultrasound should be the most important tool in your toolbox. Ultrasound is high frequency sound - technically any frequency above 20kHz, though a typical system of the kind used here is working around the 40kHz region. Sound will travel through gases, liquids and solids. Ultrasound is no different. Therefore we have applications using ultrasound in all three media.

Since ultrasound is by definition inaudible to the human ear, signal processing is required to make the inaudible, audible. This is achieved using a technique known as heterodyning - the vibration world likes to use the alternative word "demodulation" but the principle is the same. Heterodyning works because of the trigonometric fact that

What this means is that if I multiply (called mixing) two sinusoidal signals together I create two cosine signals whose frequencies are the sum and the difference of the two original signals.

A typical circuit for this type of process looks like this:

In this example, the mixer frequency is set to 38.4kHz and there is a 2kHz filter on the output of the filter. In this manner, not only can I capture the difference frequency but also the sum frequency.

A variety of sensors are available for most good quality ultrasound systems which allow the user to work in airborne, submerged or contact mode.

Working with ultrasound

Perhaps the greatest single function which ultrasound provides is the ability to listen to what you are measuring. This great function has been frowned upon by some of the vibration fraternity because diagnosis based upon listening to something is apparently not diagnosis - unless you are an "old timer" that is.

Another aspect to consider in parallel at this point is what I like to call the "data reaction time": how long does it take to get from collecting data to circulating information. If you are using an outside service Company, you will understand this concept. If data is collected today, when will you know about the problems you have? Will it be today or tomorrow or even next week?

Tools such as ultrasound and infrared thermography offer an almost instantaneous data reaction time where vibration in the hands of a "trained monkey", as the vibration industry marketeers like to suggest is possible, is most definitely not instantaneous. Of these two technologies, ultrasound is going to give you a much higher level of detection rate, and data reaction time when it comes to bearing condition than vibration.

It is interesting how many vibration Companies have recognized this problem over recent years and now provide an audio output from the vibration data collector - though once again of little interest to the monkey!

Listening is the way we have found problems for centuries - why should electronics suddenly make that redundant? With ultrasound, electronics enhances the ability by giving us a broader frequency range to listen to.

In many plants that I have visited around the world, time signals and spectra might as well be hieroglyphics for all the information they convey. Playing a sound to someone however, like looking at thermal image, tends to be immediately stimulating - and when presented with an associated "good" condition, almost always conveys the message that a problem has been identified.

Setting up an Ultrasound regime

Ultrasound can be deployed in three ways:

1. survey an area looking for non-trendable problems
2. organised routine data collection of dB values
3. organised routine data collection of dB values and time signals

With so many applications to cover, it is important that some thought is devoted during the startup phase to the best way to cover each application on site.

Non-trendable problems:

There are many areas where failures are binary rather than trendable and so collecting Gbs of data has little or no benefit. Here are a few examples:

• Compressed air leaks
• Steam leaks
• Vacuum leaks
• Heat exchanger inspections
• Tightness testing
• Corona, arcing and tracking detection (also in category 3)
• Steam trap testing (also in category 3)
• Basic identification of some rotating defects

Trending dB values:

This is the classical area of application for bearings - particularly important when operating an on-condition lubrication strategy using ultrasound. The sad thing is that too many people still do not understand dBs - even at the time of writing this paper there was a learning presentation on the internet suggesting that the 12dB increase from ~60 to ~72dB was an increase of 20% instead of the x4 it really is.

Storing and comparing time signals:

This would be the approach most similar to vibration data collection. The difference being the range of application:

• Corona, arcing and tracking detection
• Steam trap testing
• Bearing condition
• Gearbox analysis
• Advanced analysis of some rotating defects

Since this is a paper and not a book, it is impractical to cover all of the potential applications in any detail. So instead I intend to look at just two from each of the three deployment methods and how it does not take a great deal of technology to find a problem, but having more technology provides a diagnosis.

Identification of rotating defects

We concentrate so much on bearings, don't we? When vibration is used to inspect a coupling however, we must take vibration readings on both sides of the coupling - both radially and axially - in order to go back to our computer to look for problems. If you have had any IR training, you will know that a misaligned coupling generates heat. This heat is caused by friction. To do this, I need to be able to have line of sight of the coupling - not always easy to do.

With an ultrasound system I can listen to friction. So, with an ultrasound system, I can hear a misaligned coupling. All that I need to be able to do is to position my ultrasound sensor in a way which gives me a good air path to the coupling - a flexible sensor is extremely useful for this. No routes, no trends, no analysis really needed, I can just walk up to a coupling and hear the friction - the compression of the spider - caused periodically by the misalignment. How quick is that? How efficient?

If I do have a more powerful analytical set-up, I can capture the ultrasound signal and store it. When I analyse that signal I can see the repetition of the squeezing event:

This sound, this signal, is totally from that of a loose coupling. In this case I have fretting and rattling which is not periodic but random in nature:

Another very simple example is soft foot. I can of course find this with vibration. But once again, no routes, no trends, no analysis is really needed, I can just walk up to a motor or bearing and hear the clatter - the impacting at the top and bottom of the movement - caused periodically by the looseness. How quick is that? How efficient? If I do have a more powerful analytical set-up, I can capture the ultrasound signal and store it. When I analyse that signal I can see the repetition of the event:

One thing which has not been easy to do with ultrasound until quite recently, is to record a dynamic signal in a calibrated manner which will allow historic comparison with a previous reading. Using this newer technology it becomes easier to prove the efficacy of a repair by recording a before (above) and after (below) condition on the same scale.

On condition HV electrical cleaning

In the application of predictive maintenance to more and more areas of our maintenance lives, there is a simple cycle rule which works very well.

To fully implement and do more than pay lip service to your predictive maintenance program, you will realise that there will inevitably be a conflict between your planned maintenance system and your predictive system. How many times do we see perfectly healthy machinery or equipment being pulled apart because a clock says so?

The measurement cycle approach suggests that we should not be fixing something when we do not have measurement data that indicates a problem. Furthermore, we should be using measurement to confirm that any repair we think we have performed has actually been successfully completed.

One occasion where this planned maintenance problem costs a lot of money is when it comes time to clean high voltage switchgear. The traditional approach is to isolate and power down an area and send in a cleaning crew to clean, check and inspect connections and components. This work may have been carried out on an annual basis in the past but with cost pressures, companies are trying to extend this costly intervention beyond 12 months.

The savings realised by the utility would make cleaning a live unit compared with isolating very significant, depending on the location of the switchgear. In some cases it can take a full day to complete an order to operate just to isolate the gear. Coupled with the potential loss of revenue to the utility, loss of man power to perform the switching and the customer complaints generated by
the loss of power makes cleaning electrical apparatus, when live, very attractive to a utility.

Our measurement cycle approach above is now being combined with the use of portable ultrasound systems and infrared cameras to deliver innovative, targeted cleaning using dry ice on live equipment - only on the live equipment that actually needs to be cleaned. This approach means that cleaning can be removed from the planned maintenance approach and dealt with predictively and incisively - a targeted approach to HV electrical maintenance. If this wasn't good enough, there is a secondary benefit in that dry ice cleaning does not create chemical waste.

Once again, the measurement approach can be either simple or detailed. Using a parabolic dish, a cone or even just an airborne sensor, corona and tracking - both invisible to IR - can easily be identified. Using the same calibrated measurement approach and comparing historical data does however allow us to prove whether a cleaning operation has been fully successful or whether there is some residual problem which will require more detailed repair efforts.

Steam trap testing

Around the world, the combination of ultrasound and temperature measurement is recognised as the way to test the condition of steam traps.

Despite this global use, there are no standards as such for the acceptable behaviour of a steam trap and the analysis of condition is generally subjective. There are several historical reasons for this, but one of the key reasons has been the lack of a system of data collection - similar to a vibration data collector which could capture long duration time events with scalable, repeatable, comparable data. This type of measurement is not normally possible with a vibration data collector because a data collector can only capture data within the constraints of its signal processing capabilities. Feeding the 2kHz bandwidth of the headphones output of the typical ultrasound system into the 12,800-line equivalent vibration data collector will give a typical time signal recording length of 6.4 seconds. Nowhere near long enough.

Now that such technology does exist, it is now possible to capture the entire charge and discharge cycle digitally, review it, store it, compare it with historical recordings and with known "good" state traps in order to develop objective analysis. Trying to develop adjectives and adverbs to describe the sound heard can be quite difficult - recording a signal for comparison, trending and training is going to provide a much easier approach to identifying defects. The two recordings here are both over 20 seconds in length and show the ease with which this comparison can be made:

With the right hardware and software it is not only possible to record these signals, but to also play back the recorded signals on your PC.

Summary

In this presentation we compared the methods and relative merits of using ultrasound to find problems in a simple way, in a loosely structured way and in the highly structured regimen of routine data collection.

Learning point takeaways

1. You don't need routes to find problems

2. Implementing ultrasound has virtually immediate returns on investment

3. Routine data collection, trending and analysis of static and dynamic ultrasound is now possible.

Article submitted by: By Thomas J Murphy C.Eng. Managing Director, Reliability Team, Manchester, UK