Bearings are an integral device widely used in machinery that requires rotational, axial or linear movement to operate whilst restricting motion into a designed path, minimizing friction and stress. Many industries have machinery that requires some form of motion enabled by bearings. Such as:
Steel production facilities
Steel and iron use cold rolling mill machi-nery. The working rolls of the plant are the most extensively monitored. This is quite a challenging monitoring environment due to high temperatures, high and low-speed operation, as well as high contamination of water and debris.
Slow turning rolling element bearings require monitoring in this industry. Machinery such as crushers, stackers, conveyors, vibrating feeders, magnetic separators, slurry and vacuum pumps, classifiers, agitators and compressors.
Paper processing Industry
Papermaking machine bearings operate under very high temperatures and can be vulnerable to fractures of the inner ring, causing stoppages in production. Monitoring is done to determine the condition of the rolling-element bearings of the rolls, roll alignment, balance, and the condition of the electric motors and gearboxes.
Many machines in this industry require monitoring of bearing condition, inclu-ding crushers, mills, separators, roller presses, separators, conveyors, feeders, air compressors and fans. Most of which use rolling-element bearings powered by electric motors.
Thermal power industry
Gas and steam turbine generators and combined cycle plants require their most critical machines monitored. Dynamic rotating machines use high-speed rotating parts. The cost of failure in this industry is far-reaching.
Diesel engines, gas turbines and nuclear reactor powered ships utilize bearings in all areas, from crankshafts to pistons and pumps. Also, gearboxes fans and other motors onboard, such as alternators used to generate electricity.
Many aspects of ship engines utilize bearings, from crankshafts to pistons and pumps. Also, gearboxes fans and other motors. Almost all machinery that moves requires bearings. However, they are liable to degradation over time and many factors influence this. According to studies conducted by the Mobius Institute defects are encapsulated in the chart below.
We are not investigating these particular categories in detail but we can exclude the handling/Installation section for our purpose, as this is not applicable nor possible to prevent through direct monitoring. The remaining 84% accounts for issues occurring once installed. Monitoring is essential for the detection of faults caused by fatigue or stress due to excessive load or use. Incorrect lubrication, as well as insufficient levels, can cause friction which can lead to fractures among other defects. Also, contamination, both in the bearing housing or the lubrication can damage and reduce the life of bearings.
Monitoring can detect abnormalities caused by these and allow preparations for maintenance or replacement. Ultimately avoiding failure and saving money. The cost of bearings is highly variable and depends on position and if any associated damage has occurred as a result of the initial bearings then figures can rise quickly. In severe cases when bearing monitoring has been limited and total failure has occurred then the cost can be tremendous.
All of the above contain bearings and if not monitored, or repaired when needed. The results could be catastrophic failure. This is just the cost of the replacement machinery, of course, this varies, especially when only small replacements are needed. A failure of a bearing is not just the part itself, the cost to a- vessel owner is repair, fitting and downtime in dock and loss of earning whilst not operational. Needless to say, the process can become extremely expensive.
Why do bearings fail?
Bearings will fail for a number of reasons but the key take away is that ALL bearings will degrade at some point and if they are left unchecked, maintained or replaced WILL fail. Bearing failure can have overwhelming consequences for a business. Appropriate monitoring matters because false diagnosis can result in undue downtime, wasted time, money and resources. The correct diagnosis is also extremely important but can be challenging. Knowing the root cause of damage can help prevent future failures. According to ISO 15243, damages left undiagnosed can actually mask the underlying cause if left too long [ISO 15243]. Below is an example showing the progression of bearing damage that has become more severe, thus hiding the root cause:
Many cases of catastrophic failure in bearings have occurred throughout the industry that could have been avoided if effective condition monitoring and maintenance had been in place. The following case study of bearing failure has been taken from the Australian Transport Safety Bureau:
On 7 March 1997, the Polish flag general cargo vessel Lodz 2 was using one of its own cranes, discharging a general cargo of steel products, including bundles of steel pipes, from no. 2 hold and tween deck. At about 0740, the sixth load of steel pipes, for that morning, was being discharged onto the wharf by no.1crane, a 12.5-tonne capacity crane situated on the aft end of the forecastle on the ships centreline. The load, weighing approximately 8.6 tonnes, consisted of 18 lengths with diameters varying up to 273 mm. As the load reached the side of the ship, there was a violent jolt and a bang as the slew bearing failed, then the crane fell from its pedestal into the port tween deck of no. 2 hold.
The jib struck the port bulwark, setting it down and out from the ships side, while the body of the crane hit the inboard edge of the port hatch coaming, before rotating through 180∞ and finishing up, upside-down, in the tween deck. The driver had fallen with the cab of the crane, approximately 17 metres into the tween deck from the cranes position on its pedestal. The crane was severely damaged and the badly twisted jib had to be cut up to remove it from the ship.
Many devices make use of vibrational analysis for monitoring, for instance, accelerometers. Traditional accelerometers are seismic transducers and rely on piezoceramics, either lead zirconate titanate or single crystals (e.g. quartz, tourmaline). Measuring acceleration via its piezoelectric crystals which convert vibration into an electrical signal is known as the piezoelectric effect.
These devices function based on the crystals’ natural frequency and can sense a wide range of defect frequencies [typically Hz-20 kHz].
Accelerometers are designed to pick up vibrations as an indicator of defects. Excessive vibrations can indicate a serious problem that requires maintenance before a catastrophic failure occurs. Complex software is then used in conjunction to diagnose machinery faults. This is the basis of VA.
AE is used for effectively monitoring different machine conditions; balanced, unbalanced, misaligned and defecting bearings. Monitoring equipment using AE works on the principle of using transient elastic waves to detect the rapid release of strain energy caused by deformations or damage within the surface of materials. Strains, stresses and impacts produce transient stress (elastic) waves which can be effectively detected and occur in the very early stages of bearing degradation.
Although traditional acceleration and AE techniques are both used for condition monitoring they apply different methods, cover different aspects of the same problem. They are both reliable within a specific operational range and have significant overlap in capability. In some cases, it is less clear which is more capable. However, extensive studies in both areas have been conducted and each has unique benefits but also various limitations. Both of these aspects will be considered in this paper. The table below summarises some common defects that occur in bearings:
Operationally Effective Ranges: Accelerometers have an operational effectiveness within a low to high-frequency range ~ 10 Hz - 30 kHz.
This is suitable for displacement, velocity and acceleration measurements for general vibrational analysis. AE is based on frequencies much higher and with a larger range than what’s capable with VA, AE will operate within ~20 kHz-1 mHz , but more usefully around ~20 kHz-1 mHz.
The immediate advantage of this is that mechanical noise is no longer present in this range. Although frequencies of this level mean that AE is slightly less effective for general monitoring because defects in the advanced stages produce lower frequencies and can sometimes saturate the AE signal, it does mean that the ability to detect defects at a very early stage is possible.
Defect frequencies are highly dependent on machinery designs and component geometry and are also a function of bearing geometry, pitch, roller diameter and the relative speed between raceways. Fortunately AE is less geometrically sensitive. It is important to be able to distinguish between machine noise and defect noise. In order to establish defects using VA, a “base-line” needs to be established from the initial conditions of the machinery and used to indicate an irregular condition by monitoring an overall trend that should highlight anomalies.
This is not an easy task and requires highly technical knowledge as well as being very time-consuming. One drawback when monitoring the general condition of the entire machine is that the overall trend level may not show any significant change if the machine fault is not severe, or the signal is insensitive to the fault. With AE the detection of acoustic pulses are usually independent of any machine noise interference, especially at the high operating frequencies which is ideal when baseline frequencies are not known. This gives confidence that defects will not be missed.
Moreover, experiments conducted to test bearing operation at low speeds s [10-100 rev 〖min〗^(-1)] found that AE is sensitive to loading conditions [McFadden, 1984]. A change in load at constant speed causes an increased output of AE signals. The experiment concluded that typical AE transducers are less suitable to general purpose monitoring, whereas, VA is very suitable in this area.
However, it has been shown that AE is still effective at low speeds [Alshimmeri, 2017] and can effectively detect bearing defects when previously it was assumed that AE was mainly effective at high speeds. It seems that excessive loading will cause disruptions in the obtained signal as it can affect signal attenuation. It has been found that the attenuation of signal remained constant at higher loads and can now be distinguished fairly easily. The effects of attenuation can vary, as with 3d structures, in general, the amplitude of a propagating acoustic wave will decrease by about 50% every time the distance is doubled from the source.
Signal output will depend on a few design factors. Distance between bearings and sensor placement, including an air gap can also affect output signal. Acoustic waves propagate significantly slower compared to metallic materials within the air gap. The result is a damped signal which may fall below the sensors detection threshold. This will depend on the strength of the source, distance, material and even sensor placement.
Bearing defects are essentially random but occur in natural stages and will usually worsen due to dynamical strains on the bearing components. Generally speaking, advanced stage defects will produce high intensity but low-frequency signals. The very early stages of wear occur as a result of minuscule abnormalities, such as very small contaminants in lubrication, or small stresses and deformations in the material. Defects at this stage can only be detected in the very high-frequency range which is covered by AE monitoring.
Contaminants in lubrication will cause mechanical impacts that would traditionally be checked separately using other methods such as listening to frequency changes in conjunction with structural monitoring or with post-processing, such as the Shock Pulse Method (SPM). This will pick up the mechanical shock wave produced when mechanical contact has occurred. This method uses slightly more complicated processes and again needs specialist knowledge.
It has been found that lubrication containing contaminants as small as 500 μm, which would ordinarily be missed using traditional methods produces an AE pulse, detectible by AE monitoring. This is a clear indication of its usefulness of AE in this area compared to traditional methods
Although sources of defects are random in nature, the most important AE source is in the form of cracks because they produce some of the highest frequencies. Cracks cause a displacement of the surrounding material which increases in magnitude and will usually evolve into more serious problems. It has been shown that vibrational analysis will not pick up defects at this stage, as it is beyond the frequency range. Other aspects of monitoring are arguably just as important as crack formation for the prevention of total failure but usually, occur in the later stages when the damage has already occurred.
Once raw data has been obtained using data-acquisition systems, post-processing takes place in order to clarify findings. A significant difference between the two techniques in question happens at this stage. VA requires the signals to be processed downstream, manually or semi-automatically where complex post-measurement data needs to be interpreted by a specialist. Raw data is commonly analysed using Fast Fourier Transforms (FFT) to determine the sine wave frequencies and respective amplitudes. The peaks in frequencies define the type of fault whilst the amplitudes indicate the severity. This seems unnecessary complex in comparison to the automatic processing at sensor level provided by AE devices.
AE devices commonly use algorithms to derive acoustic emission parameters of Distress® and db levels which give an indication of bearing health and overall noise respectively. This provides immediate clarification of defects and is easier to interpret. AE requires approximately 10 seconds of continuous monitoring at a consistent running speed, whereas FFT based vibrational analysis typically requires 60 – 120 seconds of measurement to establish the same level of knowledge about the defect.
Although analysis of the frequency spectrum using vibrational data appears to be more in-depth and possibly even the best indicator when looking at overall machine conditions, the defect frequency may be close to the frequency excited by the components of the machine, which can lead to misinterpretations.
The totality of information provided by vibrational analysis exceeds that of AE and can be very useful when trying to determine the precise location and nature of advanced defects, however, data acquired using AE can also be analysed in a similar way in order to gain more information, but is usually unnecessary considering AE devices automatically derive vital information automatically at sensor level.
When using AE and more advanced post-processing a detection threshold needs to be established which is based on the number of times the amplitude (determined from the wavelet) exceeds a pre-set voltage within a given time. This threshold (voltage limit) can be adjusted, or floating, depending on the circumstances and machine specifications. Signals below the threshold are discarded allowing faster processing with less data storage. In reality, there are a variety of rotating machines with an even larger variety of operating conditions making the decision to set a lower limit a challenge for AE monitoring.
However, studies have shown that varying values of this threshold that lie within the range of 20%, 30% and 40% of the signal amplitude, are insensitive and doesn’t affect results significantly. Tests have been conducted and show that the probability of misdiagnosis is proved to be very remote, indeed. This indicates that the threshold is mainly effective at removing excess data, thus reducing processing time.
Operationally speaking, both of these monitoring systems can be complementary. However, vibrational analysis can be influenced significantly by the natural frequency of the machinery and VA is only useful when combined with data acquisition systems and complicated post-processing. Both VA and AE is useful when investigating low rotational speeds but AE is very effective at detecting deformation due to the formation of cracks or friction which occur in the very early stages that are independent of the machines dynamics. It has been shown that defects give off an AE pulse well before they can be detected by vibrational analysis.
Due to the general evolution of defects in machinery, damage must have already occurred for it to produce a detectible vibrational anomaly which sits in the operational frequency of suitable for VA. VA is very useful for detecting more advanced stages of defects and for general monitoring of rotational and non-rotational machinery but, is clearly limited when damage occurs beyond the detectable limit and for defects in the very infant stages. Ultimately AE provides the longest period of monitoring.
The importance of being able to operate at very high frequencies is paramount for early detection of defects and for detecting defects when noise levels are high [as is always the case]. Using AE operating at very-high-frequency levels ensures very early detection and enables detection even with high operational noise from the machinery. This allows maintenance teams to schedule appropriate action that minimises operational disruption whilst avoiding catastrophic failure.
Extensive research has been conducted over the past few decades into vibrational analysis and more recently into AE. Overall, it is apparent that VA coupled with complex post-measurement data analysis is useful for general degradation monitoring but can only detect defects that are in the advanced stages, effectively missing arguably the most important stages of defects. VA also requires experienced operators to use and interpret raw data, an unnecessary expense to a company.
AE technology and post-processing has effectively been de-skilled due to its simplicity, meaning it can be used by most operators with relative ease, again reducing the cost for training/hiring specialists, but also, empowering current monitoring staff. I think the research is compelling and clear, early detection and simple diagnosis is key to avoiding catastrophic failure, which is what AE monitoring provides.
Looking forward, it’s clear that AE is far superior for monitoring, it is now a well-established, tried and tested method of defect detection, proven time and time again. Preventative maintenance and early warning systems work and have shown to be more effective. AE is a technology that is making headway in the industry and is a smart choice when it comes to time-saving cost-effective monitoring. The earlier we know that a problem exists the better equipped we are at solving it. AE has a proven track record and can give practitioners and companies confidence that they have the greatest protection in place.