Begin your condition monitoring journey
Reliability engineering and in particular condition monitoring is a mature and complex engineering subject and for those not from a reliability background, establishing where to start can be challenging. Jim Fowlie, Director of Operations at Sensor-Works, explains how to approach the introduction of condition monitoring to a facility, the key steps that should be taken, and how to ensure the wider company benefits are appreciated.
A good place to start with condition monitoring is to understand how best to identify failure. There are many techniques available, and where possible, it is best to use a combination of approaches. Ultimately, the more information available in an investigation, the more accurate the outcome will be.
The widely used I-P-F Curve (Installation – Point of Detection – Point of Failure) highlights the instance when a specific method will start to identify an emerging issue. Ultrasonic techniques are the first to identify a failing condition. However, this is a relatively difficult and skilled technology to master, and it also does not lend itself to an online monitoring approach. The next and most widely used approach is vibration monitoring. This technology is mature and cost effective, and is used in both in person (usually done with ultrasonics) and with online (continuous) monitoring.
Figure 1. The I-P-F Curve shows the deterioration in a rotating element over the operating interval.
Generally, the earlier the detection of an issue and the appropriate intervention take place, the lower the cost to repair. For example, the catastrophic failure of an asset may well be uneconomical to repair and require replacement.
Vibration is considered the best operating parameter to judge dynamic conditions such as balance, alignment, bearing stability, and stresses applied to components. Vibration is a symptom, not the cause of the problem itself. It is inherent in all machinery, and an increase in vibration is generally a symptom that something is going wrong. Problems which manifest as a change in vibration include mechanical looseness, structural resonance, soft foundation, misalignment, rotor bow, or lost rotor vanes.
Vibration analysis as a method of diagnosing machinery problems can determine the severity of those problems and help with scheduling the most appropriate time to correct them.
Trend analysis
Extending beyond the principles of carrying out a single measurement, the condition-based maintenance concept is simple. Here, using technologies like those outlined above, physical parameters such as vibration, temperature, pressure, lubricant condition, etc., are measured and used to determine which combination provides the best indication of machine health.
From these baselines, alarm limits can be established that will trigger during ongoing routine monitoring.
Periodic or continuous monitoring readings are taken on the machinery. If a measurement exceeds its alarm limit, the system automatically detects the exception and produces plots and reports that help analyse the problem.
As the problem is likely detected early in its failure stage, the maintenance team has time to schedule the most efficient and effective repair prior to any component failure.
Ideally, the alarm limits are set high enough to minimise extraneous alarms, yet conservative enough not to miss a critical deviation in machine condition. And they can also be continuously adjusted as more is learned about the machinery from the predictive maintenance program.
Preparation
Before any condition monitoring program is implemented, preparatory work must be done. It is important to have an up-to-date asset register in place, including where assets are located, their serial numbers and model details. It is also strongly advised that an asset criticality review be carried out.
The higher the criticality of an asset, the more important it is to gain full visibility and predictability of the equipment’s health with continuous monitoring, ideally online.
For semi-critical assets or on assets deemed necessary, you can trend and apply exception reporting, only analysing and diagnosing machine problems with a more advanced portable setup for understanding the underlying sources of machine issues as and when appropriate.
Running a criticality review requires a systematic approach; Figure 2 is a sample of a flow chart that may be used in this process. Other process flows are used and can be similarly applied; the key is to have a structured, repeatable, consistent and purposeful outcome.
Figure 2. A criticality review should be structured, consistent and repeatable.
It is helpful if there is a documented history of machine failure, and if so, make good use of this information; if not, then the process of carrying out a machine assessment is a good approach. This gives a baseline of machine operating conditions. From this information, a decision on which machines to focus on in the first instance can be made.
Where there is no historical failure information, it may be worth undertaking machine assessments of assets throughout the plant as one-off vibration and temperature health checks. This will give a baseline and help to identify which machines to monitor in which phase of implementation.
It is also good practice to implement any condition monitoring system using a scalable approach. Start small and build the system and learn as you go, rather than having to rebuild each time an improvement is realised.
The monitoring solutions employed can be a simple route-based (as part of an in-person walk around a facility) or an online (and continuous) schedule-based implementation, or, as is often the case, a combination of both based on the criticality of the machinery.
At the very minimum level, a facility manager might want to know when something is about to fail. This can be done with periodic vibration and temperature checks, usually monthly but generally no longer than quarterly, carried out by operators or maintenance personnel. These tend to be cost-effective and easy to understand when used with common traffic light concepts.
Detailed analysis
Relating back to the reliability process flow, it is possible to understand what complexity of surveillance is needed. Basic vibration measurements as outlined above will indicate when a failure condition has begun to occur. For more detailed insights, a spectral analysis can show what the failure mode is and possibly what component in the asset is causing the issue.
A trend of the vibration level indicates the severity stage of a problem. The frequency (Spectrum) indicates the source of the problem.
Usually, a combination of both is done: trending with alerts and alarm levels set to trigger more in-depth spectral analysis. The setting of the levels for alerts and alarms can and will vary. The use of standards such as ISO 20816 is a reference to start; however, if the history of a machine’s operating condition is known and the point of failure is understood, this can be a better reference. A common practice is to trigger an alert when a 10% increase from the baseline occurs, with an alarm signal when a 30% increase is measured.
Sensor technology
Most sensors used in today’s vibration monitoring systems are accelerometers, which give a live ‘acceleration over time’ signal. The sensors will generally use either MEMs or piezo technology: MEMs has a different sensing frequency range – usually DC to 6kHz (for the better devices) – and a piezo sensor usually has a 0.2Hz to 10KHz range.
To understand the accelerometer output, each rotating element produces a sine wave at a defined frequency. However, all of these rotating elements combine to produce a signal that is very difficult to understand as it is outputted from the vibration sensor.
By employing a mathematical technique known as the Fourier Transform, the raw signal can be converted from the time (or spatial) domain into the frequency domain. This method takes a complex waveform and breaks it down into a sum of sine and cosine functions, each with its own amplitude and phase.
Using this approach, the live signal can be displayed as a number of peaks at the rotational (or running) speed of each element of the machine, along with the interaction, to display conditions such as imbalance, misalignment, gear mesh, baring issues, and even lubrication problems as demonstrated in Figure 3.
(or spatial) domain into the frequency domain. This breaks it down into a sum of sine and cosine functions,
each with its own amplitude and phase relative to a specific element of the machine.
With this much more detailed information, an emerging issue can be more readily identified, and the appropriate remedial action can be triggered.
Mounting vibration sensors
One of the challenging areas of condition monitoring implementation is how and where to mount the sensors.
Ideally, a vibration sensor should be mounted inline and closest to the load zone of all rotating supports, i.e. the bearings. Where analytics are required (spectrum analysis), the sensors are typically mounted in the horizontal, vertical and axial positions, or alternatively, triaxial sensors are available for semi-critical and small to medium-sized assets.
For a basic monitoring setup, a single sensor on the drive end in vertical orientation is ideal. If misalignment is a concern, then an axial sensor gives useful information.
It is helpful to have both the drive end and the non-drive end monitored. However, if this adds too much from a cost perspective, then the drive end is the best to monitor as this is likely to be the part of a motor that can experience the highest amount of loading, especially on a belt or chain coupling arrangement.
It is best to mechanically fix the sensor using a stud mount, which gives the greatest transfer of energy (a good time to do this is during a refurbishment or repair). Where this is not so practical, a suitable adhesive will work. A magnetic mount can also be used for temporary or workaround operations. Split face magnets with suitably strong pull force are preferred. I would recommend that you avoid the use of a stinger, however, if this is needed due to the material of an asset being non-magnetic, then use as short a stinger as possible which reduces the bending effect on longer probe devices.
Once the condition monitoring journey has begun, it is critical to record progress, document what worked, calculate the savings and present this to management. I would also recommend that users record what did not work so well and learn from it.
It is important that within a company, this approach is seen as a value to the business and not a cost. Done correctly it will bring savings; it is estimated the for every 1% of improved operational equipment effectiveness 2% profit is generated on the company’s bottom line and this should be widely understood.
