Brogan Mortan / Chief Technology Officer / NRG Systems / www.nrgsystems.com
There are many different condition-monitoring systems available that provide information on the current health of your wind turbine. Unfortunately there is no yardstick to evaluate and compare the performance of these different systems. Germanischer Lloyd maintains a certification process for vibration-based and oil debris condition-monitoring systems (CMS) based on a specification, but the process does not directly address performance.
While there are several facets to CMS performance, such as timeliness of detection and what action to take based on the information provided, system accuracy is the least understood. Several specific aspects to accuracy must be considered: component coverage, fault-mode coverage, fault isolation, and detection accuracy.
The goal of a condition monitoring system is to provide insight into the current health of components in your turbine. However, each CMS covers different components. The gears and bearings in the planetary section of the gearbox provide an excellent example. Many vibration-based systems cannot detect faults in this section of the gearbox, while an oil-debris sensor generally does detect them. When determining what CMS to deploy, take an inventory of the components in your drivetrain that have been driving the largest maintenance costs and ensure that the CMS you are considering can detect faults in them.
Fault mode coverage
When determining which components are most critical to monitor, it is important to first consider their most common fault modes. While an oil-debris monitor can see wear debris in the oil from bearing and gear pitting, it cannot detect cracks in gears and shafts. An oil monitor will miss a gear tooth just days away from shearing off and putting the turbine out of commission. Conversely, many vibration-based systems are poor at detecting early stage pitting on gear faces, while oil debris systems can detect these faults. Understanding the fault modes you see in your fleet and matching those to the appropriate condition monitoring system will serve you well.
Fault isolation refers to a condition-monitoring system’s ability to determine which component has faulted. This is important because a CMS is designed to help make operations and maintenance activity more efficient, so it is critical to narrow down what needs replacing. When an operator is determining whether to replace a component they will typically verify the fault with a borescope inspection. It is difficult and time consuming to visually inspect every component in the gearbox. Even under ideal conditions some faults can be overlooked. If the condition monitoring system can isolate the fault, the borescope technician can verify the damage more quickly and reduce the likelihood of overlooking a fault. Be sure to ask condition-monitoring vendors how their systems isolate failed components.
Condition monitoring systems typically classify components as either ‘faulted’ or ‘unfaulted’. The methods used for this classification are inferred by processing various sensor outputs (e.g. vibration data, oil particle counts, SCADA anomalies) and often involve some level of human interpretation and therefore, potential human error. Given the complex environments and varied conditions in which these systems work, it is critical to reduce errors that come from the inevitable uncertainly of the processed sensor outputs.
Two different kinds of classification errors can occur. The first is when the CMS classifies a component as ‘unfaulted’ when that component actually is faulted. This is commonly referred to as a ‘missed detection’. The second type of error occurs when a component is classified as ‘faulted’ but no fault exists. This is the more familiar ‘false alarm’. The indications of component condition are inherently noisy due to the varying operating conditions. Therefore it can be difficult distinguishing between an unfaulted and a faulted component. For this reason the rates at which these two errors occur are typically inversely correlated.
The figure Probability versus component condition indicator, shows an example of the trade-off between missed detections and false alarms. In practice, setting thresholds is even more difficult because you don’t typically have measurements of what a faulted component looks like (red distribution in the graph). In the best case, CMS thresholds are set based on knowing what an unfaulted components looks like (green distribution in the graph) and a predefined probability of false alarms. It is important to understand how a condition monitoring vendor will set alarm thresholds to minimize false alarms while ensuring the system is still sensitive enough catch all faults.
Selecting an appropriate condition monitoring system for your wind farm is a complicated task. It’s often like comparing apples and oranges. By understanding system accuracy issues with potential CMS vendors you can start to evaluate which systems will suit your needs best. WPE