As the wind industry evolves into larger turbines installed in more remote locations, the operations and maintenance (O&M) costs, as well as the cost of downtime, have increased. This presents a unique set of challenges in the effort to continue reducing the levelized cost of energy. The effective use of condition monitoring systems (CMS) can help address this through early and reliable fault detection.[adright zone=’190′]
When deciding which CMS technology to purchase, it is essential to know your drivetrain’s failure mode and understand the strengths and weaknesses of each CMS technology. This also applies to existing hardware and maintenance methods, such as borescope inspections, lubrication sample analysis and SCADA analysis. The table shown in Figure 1 is a good starting point for generating a CMS capability matrix for a specific gearbox model and CMS hardware brand. A good understanding of the service history of the specific gearbox model can also be helpful in deciding the best choice in CMS monitoring method. It is also suggested that one create a summary chart of the borescope inspection access for each gear and bearing in a turbine’s drivetrain.
It is also important to determine how quickly a specific component can progress to failure. Each type of failure mode has a different rate of progression. Main bearings rotate slowly and can operate for months or even years after a severe failure is detected (e.g., macropitting). Gear tooth failures can be very sudden, as a crack can grow large enough to liberate a tooth in a matter of seconds.
Wear debris sensor
There are multiple sensor technologies within the category of wear debris sensors. To be clear, these are not for monitoring the health of the gear oil, but rather, they are for detecting specific gearbox failure modes.
One of the most common sensors in the wind industry is termed a particle counter, which typically uses an inductive coil sensing method. There are also optical style sensors that can be referred to as particle counters or oil cleanliness sensors, depending on their particle size resolution. The simplest type is the chip detector, consisting of a magnetic head that switches state when ferrous particles have accumulated. The sensor mounting method within the gearbox lubrication system can vary across wear debris sensors. The full-flow, or inline, mounting is in the primary lubrication return line prior to the filter.
Other sensors can detect smaller particles more consistently when mounted in a partial-flow configuration, also referred to as offline or kidney loop. In addition to cost, there are many factors to be considered when outfitting a gearbox with a wear debris sensor: particle size detection, turbine interfacing, analysis requirements, false-positive track record, and the ability to distinguish between ferrous and non-ferrous metallic particles.
The CMS capability matrix in Figure 1 summarizes how wear debris sensors are effective in detecting specific gearbox failure modes, such as bearing spalling, in which debris is produced over a series of months. They are less effective for early-stage crack detection when fine cracks do not generate significant debris.
Here, vibration will provide the earliest warning if the analysis work is done well. Wear debris detectors are not capable of monitoring grease-lubricated components or detecting failures outside of the gearbox, such as broken blade bolts, main bearing damage or generator issues (e.g., bearing failures or loose stator wedges).
When setting alarm limits for wear debris sensors, it is important to understand not only your failure mode, but also the detection behavior of your sensor. Typical alarm criteria are the following: i) total cumulative count; ii) daily maximum; and iii) increase in slope (rate). Alarms can be tailored to specific gearboxes based on familiarity with failure modes (e.g., gear tooth inclusions, planet bearing macropitting and spline wear).
In the back-to-back sensor testing on Romax Insight’s roller bearing test rig, each sensor responded differently to the introduction of debris-generating deflects. Artificial damage was introduced first as a 4 mm isolated macropit feature, followed later by a larger damage extending the entire bearing raceway.
Figure 2 summarizes the results of the test. For example, Sensor C (Green) detected only 200 um ferrous particles, while Sensor B (Blue) detected as small as 25 um but only in a portion of the oil flow (kidney loop). Some sensors report the ISO 4406 cleanliness code or may not report cumulative particle count, such as Sensor A (Black), which must be interpreted differently, especially when setting alarm criteria.
By using a wear sensor or low-cost chip detector in combination with vibration-based CMS, a very thorough coverage will be obtained.
Considering again the CMS capability matrix (Figure 1), there is the obvious strategy of combining multiple CMS methods. A look at the planet bearing column shows how vibration monitoring is effective across all drivetrain components but is graded moderate for planet bearings. These failures require a lot of expertise to be effective (e.g., rotating frame, sampling methods, accelerometer placement and advanced analysis methods). The spalling planet bearing often passes under the radar of sophisticated vibration monitoring systems unless they are well configured and the monitoring engineers know their failure’s vibration behavior. Field service companies are advancing the capabilities of up-tower gearbox repairs to include planetary stage rebuilds, and insurers in North America are beginning to offer different terms for CMS-equipped turbines. Both of these industry trends can change the significance of early detection of planet bearing damage. The most comprehensive monitoring of planet bearings would utilize three CMS methods: a wear debris sensor, or low-cost chip detector; a vibration system; and targeted borescope inspections.
Let’s consider the two example cases of comprehensive CMS in Figure 3, Case 1.
An effective method is using vibration analysis to perform a targeted borescope inspection. A comprehensive inspection of a couple of gears or bearings has a higher likelihood of discovery of early-stage damage than a blanket inspection due to access and time constraints. When vibration analysis points to a problem at a certain location, a focused inspection can then take place. An example of this involved vibration analysis detecting an inner race defect on a bearing previously considered inaccessible with a borescope. Armed with this fleet-wide vibration warning, the inspection team developed a method of removing gearbox covers to gain borescope access and successfully discovered multiple bearings with early-stage axial cracks that were able to be repaired up-tower.
Case 2: Many wind farm owners are only aware of main bearing failures after SCADA temperature alarms alert them of the issue, which usually corresponds to the final stages of bearing deterioration. At this stage, the gearbox may be receiving damage due to the main shaft shifting backward as a result of roller/raceway wear, subjecting the gearbox to damaging thrust loads. Combining SCADA temperature data with vibration data and grease analysis gives owners a more comprehensive toolset to detect main bearing damage and degrading lubrication conditions early on. With this information, repair costs can be better forecasted, prioritized and ultimately reduced through minimizing downtime and sharing the cost of crane mobilization with other planned repairs. Figure 3 provides a case study in which advanced vibration fault detection algorithms provided more than one year’s warning on a main bearing failure when the first debris dents appeared on the inner race.
The effective use of CMS can significantly reduce wind plant O&M costs and increase availability. However, for a fleet containing varying gearbox models, a one-size-fits-all approach may not be the most cost-effective. A good understanding of the failure mode and service history of the specific gearbox model can be helpful in deciding the best choice in CMS hardware and health monitoring methods.
The cost of CMS methods varies greatly, which makes return on investment calculations all the more challenging. The end users have been demanding lower-cost CMS, and the industry is beginning to respond. Vibration equipment is starting to employ technology similar to that of a smartphone rather than a desktop computer. When considering wear debris sensors, an affordable alternative to the particle counter is a chip detector that simply indicates when debris has accumulated on a magnet within the gearbox lubrication system. A chip detector in combination with an affordable vibration monitoring system provides excellent detectability of the critical failure modes across the major components in a wind turbine.
Jesse Graeter is lead technical engineer and Becki Meadows is consulting engineer at Romax Technology. They can be reached at firstname.lastname@example.org and email@example.com.