Debunking four condition-based maintenance misconceptions
Condition-based maintenance, applied to mission critical and non-critical assets, can be an effective strategy for spotting decreasing performance and for triggering maintenance so that work can be completed before the equipment fails. Even with CBM in place, it’s advisable that maintenance practitioners brush up on the skills and instrumentation that will help turn performance information into proactive maintenance tasks. Among the many misinterpretations of CBM, here are four that we’d like to lay bare, along with suggestions on what to do instead.
MYTH: “It’s more effective to perform maintenance on time rather than condition.”
Many people are inclined to believe that performing Time Based Maintenance (TBM) on a piece of equipment will effectively reset its life to 100 per cent, improve reliability and mitigate or prevent equipment failures. Multiple-failure studies dating back to the 1940s, during World War II, have shown this theory to be false and ineffective. Based on these studies it was found that failures can be categorized into different failure patterns that can either be random in nature, or degrade over time and eventually fail. The truth is that about 80 per cent of total equipment failures will be non-age related (random) compared to 20 per cent found to be age related or having a wear-out failure pattern. These results raise a question: “If 80 per cent of failures are random and could occur at any time, how can we effectively schedule value-added maintenance activities?”
In a perfect world all failures would be age related, as this would provide the ability to plan and schedule repair right before the failure occurred and makes time-based maintenance the preferred maintenance strategy. However, in reality, with only 20 per cent of our failures being age related, time-based maintenance should only add value to a small portion of the total failures we’ll experience. Saying that TBM will protect us from 20 per cent of failures is also being generous, as most PMs are general in nature and are based on the original equipment manufacturer’s instructions. Depending on the owner/operator company, an operational readiness program may be developed and RCM studies may have been completed on critical assets to ensure PMs are designed relative to the types of failures that the equipment will experience. Alternatively, if a plant is already operating, it may have optimized its PMs (PMO) to ensure each task adds value by combating a particular failure mode that has been experienced by the plant, over its operating life, and removing tasks that waste time and lack value. The reality is that not a lot of companies go through the rigorous process of fine-tuning their PMs prior to or while in operation. So, if we can expect about 80 per cent of failures to be random, then TBM will not be effective at addressing these failures.
In these cases, adopting a condition-based maintenance (CBM) strategy using appropriate predictive maintenance technologies will assist in extending equipment life and increase equipment uptime. This is done by using predictive technologies that will identify particular failures in the early stages prior to loss of function and well before catastrophic failure. In scenarios where predictive technologies aren’t being used, failures are generally noted by an operator during daily workarounds when he or she hears the ominous sound of a bearing rumble or smoke coming from a bearing. By this time there is insufficient time to properly address the failure without causing panic, unplanned downtime, or potential production loss.
Early fault detection is key to allow the CBM team to increase the frequency of data collection to closely monitor faults as they develop, while at the same time providing direction to the maintenance team so the repair can be planned and scheduled. Doing this can extend the useful life of the equipment and the repair work will be executed in a controlled environment, which not only saves on maintenance costs but also increases equipment uptime and reduces the probability of an HSE incident as well.
Intrusive maintenance is known to cause breakdowns (maintenance-induced failures) rather than prevent them, so wouldn’t it be nice to have confidence that when you do have to perform intrusive maintenance, it’s based on the operating condition and health of your asset rather than the clock.
MYTH: “Adopting a condition-based maintenance strategy will stop failures from occurring.”
A common misconception is that having a condition-based maintenance program will prevent failures from occurring. This is not the case as CBM is employed to detect failures – it cannot stop them. Integrating multiple predictive maintenance technologies, such as lubricant and vibration analysis or thermography into an organized and structured program can paint a clearer picture about the health of an asset.
As a plant operates its equipment tells the story about how it’s feeling, much like a doctor listening to patient’s symptoms. The doctor hears the symptoms, then runs diagnostic tests, such as blood work and X-rays, to assist in diagnosing the illness or problem. The same principle applies to CBM. As the equipment operates, it produces different forces or vibrations, it’s oil condition changes, heat is generated – all of which are symptoms of the machinery’s “illness” or problem.
Failure detection will only be effective if the program is structured the correct way, meaning: (1) the sampling frequency or time between samples will allow the failure to be detected, (2) the right technologies are being used by trained employees who know how to configure the data acquisition equipment to ensure value-added data is collected, (3) the data collected is consistently measured under the same or close to the same operating conditions from the same measurement points and (4) the analysts know how to identify failures based on the data captured. A correctly functioning CBM program should provide early failure detection and tracking, which allows the maintenance team advanced notice to ensure the repair is completed in a planned and scheduled manner.
If the goal is to stop failures from occurring rather than just detecting them, we need to move past using CBM and integrate reliability-based methodologies into our program to provide a holistic approach. Using different forms of reliability analysis tools will provide an understanding of the causes and effects of the failures being experienced. Understanding why a failure is occurring is the key to eliminating it by addressing it at the root cause level. Having a mature CBM program in place is a great achievement and provides major cost savings. However, wider benefit will come from failure prevention rather than failure detection, that is, higher inherent reliability and availability, lower overall operating costs, and reduced spares consumption.
MYTH: “Not all vibration data collected should be considered effective for CBM.”
Are all vibration data collected analyzed or particularly useful? Vibration measurements can be collected in many forms and the effectiveness of the data depending on how the signal is processed. Some data acquisitions systems on the market today used in conjunction with online condition monitoring software lack the data-sampling rate to allow any form of accurate high-frequency fault detection. Not being able to sample fast enough will not give the high-frequency resolution required to detect problems, such as typical gearbox faults. It’s possible that having an online continuous condition monitoring system in a plant can give a false sense of security to those who do not fully understand the fundamentals of vibration. It is believed that any and all failure characteristics can be detected with this “state of the art” system, so this data collected is the basis for effective CBM.
Much the same can be applied to measuring an overall vibration measurement. Many equipment packages used in industry today are available with vibration monitoring capabilities provided from factory installed transducers and control systems. These systems will normally provide a simple overall measurement used for equipment protection capabilities rather than condition assessment. To a maintenance group that is unfamiliar with proper vibration analysis techniques and fundamentals this may provide a false sense of security in that the equipment package is monitoring vibration and is therefore perceived as an effective form of CBM.
Having an overall vibration magnitude set point used for equipment protection only protects the machine from catastrophically failing. It can neither provide early fault detection weeks, or even months ahead of time, nor accurately tell you what part of a bearing has the fault or which gear tooth is broken. Effective CBM is about collecting detailed, specific useful data that can tell you a story about the machine’s condition; overall measurements are very generic in nature and do not reveal useful information. An overall reading can tell us that there is an increase in vibration at a particular point on a machine. Let’s say a “warning” alarm has been activated and our piece of equipment is now operating in alarm. When we sit down and look at the data we know that there are numerous things that can occur that will cause a change in vibration. It could be a faulty transducer, the transducer may need to be calibrated, it could just be a bad reading, or maybe the change in vibration could actually be real. There may have been a process change resulting in increased load or maybe there truly is a problem with the machine. This data leaves many questions unanswered and makes it ineffective at accurately diagnosing the fault without further troubleshooting, site visits or intrusive investigation.
Comparing the same scenario, but instead of using the overall data, we do a little more signal processing and we derive some common plots used, such as a spectrum and time waveform to assist us. The advantages of a spectrum are being able to see how the vibration force is dispersed over a defined frequency range and a time waveform provides a time-based sample of the vibration and shows how the machine responds from one point in time to the next. Using the spectrum we clearly see what forcing frequency or frequency range the vibration is coming from, which can tell us a lot about what fault condition exists on the machine. When we look at our spectrum we see non-synchronous harmonics (impacting) with what appears to be operating speed side bands. Immediately we think, “That looks like the characteristics of an inner-race bearing fault but, to confirm, I’ll take a look at my time waveform where I can visually see the impacts and can calculate the same non-synchronous frequency between the impacts.” Also, we can confirm our operating speed side banding, which is caused by amplitude modulation from the defect coming into and out of the load zone, which can be visually seen in the waveform. As time passes, the peak-to-peak amplitude of the time waveform will coincide with the defect to load zone position.
This simple example shows how effective properly using the data can be and that one measurement of overall vibration is not too revealing. Not being able to break down the data into its different frequency and time-based properties provides little to no assistance to properly diagnose a machine fault. An overall measurement may be very effective when integrated as a machinery protection set point but it must not be confused with being an effective form of CBM.
MYTH: “Why you shouldn’t solely rely on vibration software to tell you when you have a bearing fault.”
Many if not all vibration analysis software on the market today come equipped with bearing databases where by you can select the type of bearings in your machine and your bearing fault frequencies are automatically calculated for you. The benefit of this is we don’t have to manually calculate our defect frequencies, Ball Pass Frequency Inner, Ball Pass Frequency Outer, Ball Spin Frequency and Fundamental Train Frequency, (BPFI, BPFO, BSF and FTF) potentially making a mistake and coming out with an incorrect forcing frequency and it also saves time. The software packages allow the forcing frequencies to be overlaid onto the spectrum or time waveform making it fairly easy to detect a bearing fault by quickly cycling through each frequency to determine if any of the higher frequency peaks are lining up with any of the forcing frequency markers generated in the software. This is a great feature and it certainly helps but it also can promote the wrong mindset when performing analysis if you don’t have a good understanding of your bearings; this mindset being: “If they line up then you have a fault and if they don’t then you’re all good.”
This mindset could be true if we were in a perfect world where no variables could change this forcing frequency. For example, if we knew that our machine speed never changed, required load was always consistent, every bearing has the same amount of wear, no slippage was occurring, the same model bearing and bearing manufacturer was used every single time you replaced the bearings, etc. all of which will affect the frequencies at which our faults will appear. Different bearing manufacturers can use a different number of rolling elements for the same model number bearing. This can be confusing as we know that bearing fault frequency equations are highly dependent on the number of rolling elements in the bearing, this changes the frequency at which your fault should appear. During any normal operating day there are many variables or conditions that change that will cause the fault frequency markers programmed in our software to not directly line up. readjust
The main thing to remember is to not solely rely on the software to tell you when you have a fault. There are a few criteria that can be used to provide a fairly accurate estimate as to whether a bearing defect exists or not. When you see consistent vibration at a non-synchronous (non-integer multiple of running speed) frequency in your spectrum you can be fairly confident that it’s a bearing fault. The type of bearing fault can also be determined as each fault generates certain characteristics and forcing frequencies, such as, bearing faults will most likely generate non-synchronous harmonics (impacts) in your spectrum with accompanying sidebands at running speed (1X) indicate a BPFI (inner race) fault. Non-synchronous harmonics with about half running speed (≈0.5X) sidebands normally indicate a BSF (rolling element) fault and just seeing non-synchronous harmonics with no side banding is most likely a BPFO (outer race) fault. (Note: this is just a general set of criteria used when the inner race is rotating. In cases where the outer race is rotating, simply switch the BPFO and BPFI.) By simply knowing the operating speed, this can be very helpful in self diagnosing a bearing fault when you have little to no details about the bearing geometry or manufacturer.
Ultimately, software capabilities are getting better and there will be times when the fault frequency markers line up and a bearing fault is detected automatically with very little effort. Variables and conditions are not always controlled, so some attention to detail and self-reliance must be employed as well.
Nick Fewer is a condition monitoring and reliability consultant at Aker Solutions, based in St. John’s, N.L. He can be reached at firstname.lastname@example.org.