Condition monitoring, whilst now highly technical and adapted for specific industries (with Aerospace & Defence arguably leading the way) had very humble beginnings, essentially listening to a machine and trying to understand if it ‘felt right’ or not. Humans seem to have an inherent talent for this (I’m sure many of us have experience with our cars, feeling that something isn’t quite right or doesn’t sound like it should!) but pure intuition isn’t a good way to scientifically implement predictive maintenance and avoid unplanned downtime.
Blood, sweat and tears
Condition monitoring today is largely focused on rotating equipment, driven significantly by the advancements made in the Aerospace & Defence industries to improve safety and reliability.
Chinook helicopter similar to that of the 1986 crash
These advancements came with a heavy price; in 1986 a Boeing Chinook helicopter used for ferrying workers to offshore oil platforms crashed with the loss of all 45 people on board. The subsequent investigation revealed a bearing failure in the main gearbox which allowed the two rotors to collide; something which could have been identified early had condition monitoring been present.
Subsequent activities and recommendations from the British Civil Aviation Authority were to mandate that any craft with more than 9 people on board must have online condition monitoring, with that, a whole Health and Usage Monitoring (HUMS) sector expanded rapidly.
With an obvious safety benefit, condition monitoring became ubiquitous in the Aerospace and Defence industries. This new data source enabled the transformation of operations and the emergence of new business models, to the point where entire products such as engines have become servitized (Rolls Royce, power by the hour). The much-maligned F-35 Joint Strike Fighter is also notable for its use of integrated health monitoring, with sensors woven into the composite materials from which it is constructed.
Types and effectiveness
It’s important to note that many of these techniques can be used offline (e.g. captured using handheld meters and then recorded offline) we are mainly discussing ‘Online’ monitoring to describe the use of real-time automated capture by dedicated equipment (e.g. without requiring a maintenance engineer to walk around and place a sensor in various positions).
Time-horizon: High – Weeks to months (depending on usage)
Vibration condition monitoring has come a long from ‘does it feel more shaky than usual’. Modern implementations don’t just give a raw energy content (Root Mean Squared) value but will derive various other condition indicators such as peak to peak (distance between the maximum and minimum amplitudes), kurtosis (shape of the amplitude distribution) and even ball bearing energy. More information on various condition indicators is available in this excellent paper by Junda Zhu et al.
Vibration is most often used for rotational components (bearings, shafts, gearboxes) and well suited to complex assets, unfortunately it requires significant post-processing to extract value and requires a high sample rate to be most effective. These factors make it one of the most expensive methods of determining machine condition.
Time-horizon: Low – Hours to Days
This can be as simple as bolting a thermocouple onto something or monitoring using thermal imaging. It’s well suited for monitoring solid-state electrical components but poor for mechanical as by the time a temperature delta shows, the damage is likely too profound for any notifications to be helpful.
Time-horizon: Medium – Weeks to Months
Current monitoring is effective not only when monitoring electrical motors but for understanding the systems to which they are attached. Behaviour of the motors as well as any mechanical issues (gearbox, bearing) can be identified through current monitoring if taken at a high enough sample rate. The cost is attractively low (orders of magnitude lower than vibration monitoring) but accuracy and specificity lags behind, although this might not be important in most industrial applications.
Time-horizon: Medium-high – Weeks to Months
Whilst at its most basic it is inexpensive (some microphones and a system to capture data). Truly accurate acoustic emission detection requires similar post processing to vibration monitoring. The time horizon is very good and it performs best in high frequency applications.
Oil debris monitoring
Time-horizon: Medium-low – Weeks
This involves taking a sample of gearbox / motor oil and inspecting it for debris. This can be with something as simple as a magnetic sump/inspection plug which will attract metallic particles. Oil debris monitoring can be done online using special sensors but is most commonly a manual (and labour intensive process). The cost drivers are time (manual labour and potentially downtime) as well as any lab analysis. Unfortunately, once debris is visible in the oil, significant damage has already occurred.
Time-horizon: High – Months – Years
Ultrasound is particularly well suited to inspection of pumps / sealed systems for leak detection but can also be used to monitor things like bearings and valves. It’s versatile but requires significant processing and analytics to produce information useful for end-user maintainers and operators. The speed of detection is extremely quick, with NASA stating that ultrasonic bearing monitoring can actually provide the earliest warning of failure.
Tying it all together
As sensors and processing costs reduce, hardware becomes more and more of a commodity, spelling the end of days for manual condition monitoring. Technology that used to cost millions for Aerospace and Defence companies can now be used in the Smart Factory and is by far the best use of the industrial internet of things / Industry 4.0 as it helps you to reduce unplanned downtime and operational costs.
Automatic condition monitoring – Senseye
Senseye is a ground-breaking automated condition monitoring and prognostics product that predicts the unpredictable, using any condition monitoring source. It automatically forecasts machine failure and helps you to avoid unplanned downtime and save money.