Industry maintenance requirements

Wind turbines are typically subject to high and varying loads, as well as extreme weather conditions. This can accelerate machine component degradation or at least make it unpredictable at best. In such a situation, adequate levels of machine uptime can only be achieved by rapid response times and fast service to correct the problems. Such an approach is difficult to implement, given the remote locations and the need for cranes and other external factors. This is even more critical for offshore wind parks where turbines cannot be reached during adverse weather conditions.

For this reason condition monitoring plays a vital role in a successful operation and maintenance strategy for wind turbines. A condition monitoring solution dedicated to the special requirements of monitoring a wind turbine is needed to take into account the special construction of these machines and their operation:

  • Low rotational speeds
  • Complex gearbox structure
  • Non-rigid foundations
  • Compact construction
  • Continuously changing wind conditions

Early fault detection with proper alarm management by Brüel & Kjær Vibro

Such a system not only has to accurately detect a number of developing faults that are unique to wind turbines, but it also has to avoid the flood of alarms that can often occur while monitoring the wind turbines. Some of these alarms can be due to changing operating conditions, and therefore the condition monitoring system requires an effective power class monitoring strategy for reducing these types of false alarms. Often there are many alarms generated for the same, single fault due to the compact construction of the wind turbine drive train, so an intelligent alarm management system is also needed to reduce these types of redundant alarms.

Early fault detection with proper alarm management significantly improves the accuracy and reliability of the condition monitoring strategy, but it is still not enough for many applications. Fault diagnosis expertise is necessary for not only for identifying the type and location of the fault, but also for determining its severity and establishing a lead-time to failure. Such expertise is hard to come by, and many operators do not have it.

These are some of the primary underlying concepts from which the Brüel & Kjær Vibro wind turbine condition monitoring system solution was built upon.

Application solutions

With more than 60 years in the monitoring business, we have gained solid expertise in detecting and diagnosing potential failure modes at an early stage of development for all types of machines and applications, including wind turbines.

As our monitoring strategy is primarily focused on the drive train portion of the wind turbine, most of the potential failure modes are associated with the rolling element bearings, gears and the generator.

Although the fault detection techniques for these components are not new, the means to detect these faults present challenges due to the wind turbine design and operation concept. The widely variable operating conditions coupled with non-rigid foundations, compact construction, complex gearbox and low rotating speed require an entirely different monitoring approach for wind turbines.

Important monitoring functionality that enables a wind turbine monitoring system to detect and isolate developing machine faults includes:

  • Process bins (see below)
  • Alarm management (see below)

Vibration measurements, process parameters and speed/phase information on the machine components.

  • Gearbox 1st, 2nd & 3rd Stage Failure Modes:
    Gear Defects: Looseness, Gear Wear, Gear Tooth Faults
    Bearing Defects: Lubrication, Misalignment, Looseness, Race and Cage Defects
    Shaft Problems: Unbalance, Misalignment
  • Low Speed Shaft & Main Bearing Failure Modes:
    Rotor, Unbalance, Bent Shaft, Lubrication defects, Bearing Looseness, Bearing Defects
  • Generator DE & NDE Failure Modes:
    Lubrication Defects, Bearing Misalignment, Bearing Looseness, Bearing Defects, Shaft Unbalance and Misalignment, Bent Shaft, Looseness, Coupling Faults, Support Structure Changes, Electrical Faults
  • Structural Failure Modes:
    Support Structure Changes, Gearbox Suspension Defects, Resonances

System for distinguishing a change in vibration due to process changes, such as active power, from that due to developing faults.

The effectiveness of a monitoring system lies in the ability to detect developing faults at an early stage, without generating false alarms, despite speed and load changes. Our wind turbine monitoring system enables the user to identify relevant process classes where the vibration signature is distinct, in order to set up individual alarm limits that correspond to these process classes, as shown in the figure below. This prevents incorrect alarming.

  • False alarms – Caused by vibration changes due to operation changes, not machine faults
  • Absence of alarms – No alarm is generated despite the presence of a developing fault, because the alarm limits were too high

Intelligently reduces the flood of alarms that often occur when monitoring wind turbines, due to variations in operating conditions. 

As there can be 1000+ alarm limits per turbine, one of the greatest challenges of monitoring a wind turbine is dealing with a potentially large number of alarms and determining which alarms are relevant. There are also many transient events during the operation of a wind turbine, such as gusting and yawing. These transient events can actually cause the exceedance of an alarm limit for a shorter period of time even though there is no component fault. Another area of concern is localizing a developing fault. A fault that has been detected by a sensor on a particular bearing, for example, can also be picked up by other nearby sensors. In this case there is only one machine fault but many alarms.

Brüel & Kjær Vibro introduced an intelligent alarm management system that scans this flood of alarms and filters out those that are not relevant. It also filters out alarms due to transient effects. It filters the alarms such that one physical fault has only one alarm.

The alarm information is graduated into five severity levels in order to provide the customer’s service department with a lead-time estimate on a developing fault, as shown below. New alarms are generated only when a new severity level is reached. There are therefore only a maximum of 4 alarms for the “lifetime” of the fault. Severity is first estimated automatically by the Alarm Manager, and then a final assessment is made by a diagnostic expert.