Dave Clark/Condition monitoring specialist

There are three methods for monitoring wind-turbine operations: wind-conditions monitoring, performance monitoring, and condition monitoring. The terms are thrown around so much that many in the industry are confused. Although the terms describe different tasks, they have the same goal: performance and reliability of the wind assets.

An automotive analogy may be useful. For instance, if your car is not getting optimal fuel mileage, you get a tune-up. If that doesn’t work, you take the car to a mechanic for a more detailed analysis. So diagnosing the mileage problem is analogous to monitoring wind conditions. A tune-up is analogous to examining SCADA data, or performance monitoring. And the mechanic is akin to the analyst who finds meaning in the vibration data. Here’s more detail on each.

Monitoring wind conditions

Suppose a turbine’s power production or output is less than its power curve says it should be. Does the problem concern wind conditions or the turbine? Wind conditions monitoring may provide a clue. A wind-resource assessment of the site is performed prior to construction and at least for a year. If the resource assessment is correct, the turbine should produce predictable power. If it doesn’t make rated power, (rated for the available wind) then you will need to benchmark the conditions versus the rated power, essentially plotting the wind versus the turbine output.

The Triton from Second Wind is one of several sodar-based (wind sensor using sound) remote wind sensors. The unit can measure wind speed and direction at hub heights to give operators data to gauge a wind plant’s efficiency.

This is done with wind-conditions monitoring this way: Examine the site’s meteorological data to gauge or plot against the turbine’s output. Look for data from a met tower or remote sensor, such as a sodar unit. “Our systems measure wind conditions that can be used for wind-resource assessments,” says Naomi Pierce of Second Wind, a company that makes portable wind-sensing systems. “The equipment can be used for wind-resource assessments and monitoring operating farms. In the latter case, the equipment monitors wind and wind conditions to gauge the performance of an operating wind turbine.”

Performance monitoring based on wind conditions measures the wind potential that can be plotted against the rated power of a wind asset or site, to determine the production efficiency. Wind performance should show how much power the site can produce. If the turbine does not produce expected power, there might be a problem with the turbine. This also involves power-purchase agreements and warranty implications.

Performance monitoring

Performance monitoring is a great tool for wringing maximum performance out of a farm or fleet. This monitoring involves taking the sensor data from the wind turbine and mining (examining) it for information as to why it’s underperforming. You look for reasons why the rated power was not produced.

A tremendous amount of data comes off of a single turbine. Its sensors monitor characteristics such as:

• Yaw and pitch position
• Temperature
• Wind speed and direction
• Generator speed

There are more aspects, but mining the SCADA data can determine where performance issues lie. Steve Brost, a CMS Engineer and turbine prognostics and health management analyst, says this about performance monitoring:

“We use a software tool called T2 for identifying deviations between two or more data populations.  We use it to identify deviations in trends by comparing the (multivariate) means of data populations and point out the one with the most consistent deviation.  We configure a threshold in the tool to identify the ‘highest hitters’, those that deviate most. Once the tool has identified a repeat offender, a value for that data population is collected into what we call a Cusum (cumulative sum), so we can focus on locations in the turbine with the highest probability of potential failure. For example, the hottest temperature might indicate damage or confirm damage through other measurements such as vibration. The Cusum characteristic is something of an alarm threshold or a filter we can set.

After taking all Cusum values across our fleet, we examine the highest values and cross reference the findings with vibration data to focus on locations that are alarming the most frequently in vibration and temperature or other SCADA trending data such as highest fault count.”

Suppose a turbine’s temperature points to a potential problem. This element is neither sufficiently specific nor nearly as predictive as vibration-condition monitoring. But a “hot” indicator is similar to a “check engine” light in a car. It signals a problem but not specifically what is causing it.

Statistically, every tenth turbine faced relevant damage each year. Costs for a planned repair are on average less than 30% compared to the replacement of a component. (Source: DEWI). Consequential damage can be prevented.

Condition monitoring

Nothing reduces turbine efficiency more than a failed component that halts its power production. Condition monitoring using vibration in particular, aims at identifying specific component degradations. Typically, this includes sensors placed on components that are costly to repair, fail with regularity, or both. Usually and regardless of model, such components include:

• Main bearings
• Low-speed shaft, gearbox
• Planetary section, gearbox
• High-speed shaft section, gearbox
• Generator, drive end
• Generator, non-drive end

Several operational parameters (which performance monitoring can detect) ultimately increase wear on drive-train components. Performance monitoring detects these abnormalities while condition monitoring detects component wear. Hence, it is crucial to perform both. The monitoring methods work together on performance-related failures.

For example, a rated power output may be affected by a misalignment between the gearbox and generator, or a failing high-speed shaft. Yaw deviation, the difference between wind direction and nacelle direction, can also affect power output by causing undue drive train loading. Condition monitoring detects the wear.

As you would expect, not all failures or shortcomings are performance-monitoring detectable or correctable. For instance, looseness, misalignment, imbalance, and most all early detection on gear and bearing failures are impossible with factory sensors. So check the fuel efficiency on your turbine, give it a tune-up, and avoid a trip to the mechanic. Although methods for monitoring wind turbine performance have different names, all work toward the same thing: peak wind-turbine performance.

WPE

Nick Harper
Applications Manager
Blade Sensing Systems
Moog Inc.
www.Moog.com

Cold weather presents special problems for wind turbines. Inside the nacelle, low-viscosity lubricants keep the gearbox turning and enclosure seals to keep moisture and ice off electronic components. But outside the nacelle, things are different. Ice easily forms on turbine blades possibly adding hundreds of kilograms, which degrades performance and shortens working life. About 65% of wind turbines in North America are in areas where icing is possible and likely. Also because wind farms are often in remote locations, shutdowns are occasionally necessary when icing conditions are present. Then, turbines require a visual inspection before a restart.

The problem with ice on a working turbine is that it can be thrown, and when not, it causes additional drive-train loads often in excess of design loads. For this reason, many turbines are shut down when ice buildup threatens and restarts come only after an inspection confirms the ice is gone. This is a difficult practice in remote locations or at night.

Ice on turbine blades is fairly obvious when the unit is clearly visible. But at night or in remote locations, the question of ice on blades becomes a matter of conjecture. Weather conditions right for ice is one indicator, but there are better ways to detect it and protect turbines.

Traditional ice detection uses meteorological equipment but this does not detect ice on blades. It simply measures conditions for icing, so it does not give operators enough warning or time to take action – such as shutting down the turbine to prevent damage.

Detecting ice
A recent solution to the problem of blade icing includes ice-detection sensors and controls. The unit is part of a rotor monitoring concept, called RMS, for wind turbines. It lets operators monitor wind-turbine performance to detect ice on blades and avoid damage. RMS uses optical strain sensors mounted inside at the root of each blade. They work optically so the sensors are immune to lightning and electromagnetic induction effects.

The ice detection system provides information for the operator to shut down the turbine when ice loads exceed the specs or present a danger of throwing ice. The system also detects when ice has been shed.

The ice-detection system works two ways. The left of Signals from ice shows the system working under normal operation conditions. The monitoring system measures the bending moment of the blade as it rotates, generating a sinusoidal wave pattern. As ice builds on the blade, the amplitude changes. When the trace goes above a predetermined threshold, the monitoring system shuts the turbine down.

The right side of Signals from ice shows a static signal trace. At this time, the turbine is closed down but not stopped. The rotor is still turning slowly while the system is working in the frequency domain. It is relying on the low wind speed, about 3 m/s, to excite the natural frequency of the blade. This natural frequency changes as ice builds on the blade and changes its mass.

The photos and traces show how ice looks to a technician and the detection system.

The buildup of ice correlates with stored data. The right of Reading the graph tracks a period from October to December. At the test site, there was no ice buildup for the first 70 days. At about day 72, moderate ice was verified. On day 78, a significant amount of ice was noted. After that, there is a reduction in mass to the point where it was safe to restart the turbine.

Now consider two turbines, A and B, on the same wind farm. The red and blue traces show strong correlation. Notice the apparent noise signal level indicating about a 100 to 150 kg addition to one blade. This is significantly reduced by providing blade pitch information.

The ellipses identify two severe ice events detected by the system and confirmed by inspection. Without ice, mass values should be zero. In this case, both turbines have inherent noise of 100 to 150 kg because pitch angle feeds were not provided. Overall system accuracy is maximized by supplying a pitch and angle feed from the turbine controls, along with blade mass and center of gravity.

Experience shows that to increase the overall accuracy of the ice detection system, blade pitch position from the turbine’s control system is combined with the optical data.

So if we can take blade-pitch information on the turbine, we can reduce the noise level. Even without it, we can clearly see that both turbines correlate well with each other – they are both icing at about the same rate. At 40 days, turbine A had ice for about half the time of turbine B. This tells that the operator was able to start Turbine A before B by two days. This, of course, improves the revenue for that particular machine.

When the RMS is functioning on a working turbine, it can generate the blue traces. When signal top a predefined threshold, the controls halt the turbine, but the RMS continues to track ice build up through the frequency domain, the red plots.

Data in Time domain and frequency analysis presents a comparison between a turbine rotating (blue trace) and one stopped (red data). Notice that the turbine on Day 2 is working without ice, but slowing over a period of 1 to 1.5 days. The blue curve is going up indicating ice build up on the blade.

At some point, the selected threshold is exceeded and the turbines will be stopped. The red information indicates that Turbine A is halted, but the system still provides useful data. The mass still increases on the blade, up for two more days, but then the data becomes flat again. After about day six, it reverts to zero. So the system works when the turbine is operating, but it is just as useful when it is not.

A closer look
The rotor monitor includes several components. First, there are four independent sensors at the root of each blade. These are installed at four positions corresponding to the leading edge, trailing edge, pressure surface, and suction surface. The sensors that connect to the interrogator unit, are rugged, well protected, and will last the life of the blade. The interrogator unit (OEM1030), mounts in the hub of the turbine. It can also mount in the root of the blade, but anywhere in the rotating part of the turbine would do.

There are several ways to transmit data out of the hub. One is a GRPS modem, which lets operators send data to the U.K. for analysis. Another way is to send the data by slip ring into the nacelle where it could be connected to the SCADA system for direct control of the turbine. So as the ice builds to a certain level, the SCADA can call for a shutdown and restart when the ice clears.

The RMS from Moog includes four optical strain gages in each blade, an interrogator unit in the hub, and at least one of two ways for transmitting signals across the turning hub.

The other main feature of the system is part of the RMS. Using the same system and additional software supplied by Moog, the system can interrogate the data and deduce information associated with blade damage and rotor imbalance. This can reduce the unit’s overall payback period.

Ice detection can be implemented on any wind turbine. It has been globally installed in many and we have not yet received word of any sensor failure. From experience, the system annually provides about $15,000 in additional revenue. In September 2009, GL certified the system.

WPE

Alex Nino, LUDECAwind, www.ludecawind.com

The green arrows indicate sensor (accelerometer) locations for standard wind turbines.

The wind energy industry is growing significantly in generating capacity and volume. As demand increases, so does output power of wind turbines. OEM’s across the globe are having extreme difficulty testing newer and larger wind turbines because conditions vary from country to country, thus producing a negative impact on system availability and reliability. In 2003, as OEMs were inroducing relatively large 1+MW turbines, their gearboxes began failing at a staggering rate. In Europe, system insurers introduced a revision in their policies and cancelled all old contracts. A revised clause stipulated that “all roller bearings in the drive train must be replaced after 5 years or 40,000 operating hours, unless a suitable Condition Monitoring Systems (CMS) is installed.” Insurers found value and protection in the function and fault diagnostics of condition monitoring systems.

Function diagnosis refers to the measurement and data collection of functional and operating parameters, and overall vibration values. This information is required for the proper analysis and long-term operation of rotating machinery. Fault diagnosis is the determination (cause) of damage conditions of machinery and its components.

According to DIN ISO, condition-based maintenance for wind-power plants aims to maintain, visually inspect, measure, and analyze the condition of the turbines and perform required repairs. However, how can we measure and evaluate vibration components of wind turbines when they have been excluded from many of these international standards? For example, ISO 10816-3 explicitly excludes wind-power plants.

Condition-monitoring systems have made it clear that wind turbines are complex machines in which overall vibration values must be systematically determined and evaluation references made available. Points to consider when writing evaluation guidelines for wind-power plants include:

• Function and structural design of wind turbines and their components
• Interaction between the individual drivetrain components (modules) being tested
• Information and experience regarding the possible faults and damages occurring in the individual testing modules during operation and their economic impact
• Knowledge of operation-related and machine-related vibration influences, the diagnosis procedures that must be adhered to and their respective limits

The recent VDI 3834, established and released in 2009, takes into consideration the special requirements for evaluating wind turbine components. The guideline is set for turbines ranging from 100 kW to 3 MW.

Measuring methods

Varying characteristics of wind turbine operation and wind conditions require collecting vibration data using piezoelectric accelerometers which can measure frequencies from 0.1 Hz to 6 kHz, as defined by VDI 3834. Other important criteria for proper data collection is a minimum load of 20%. Because of natural fluctuations in wind loads, the VDI 3834 specifies measurement periods from 1 to 10 minutes. Such long evaluation periods provide a stable and meaningful root mean square (rms) vibration data for slow rotating compenents.

Characteristic valuesThe VDI guideline separates drivetrain components into their main groups and assigns overall vibration values to the most important ones. These component-specific vibrations can be classified and identified. The VDI 3834 is based on statistical analysis of vibration measurements from more than 450 wind turbines, and defines threshold values in terms of vibration velocity in mm/s and vibration acceleration in m/s2 for the generator, gear, main bearing, nacelle, and tower. VDI 3834 also recommends warning and alarm thresholds. Threshold values are defined as component-specific frequency bands.

Detecting generator faults early calls for monitoring vibration magnitudes. The graphical trend displays the increase in vibration amplitude over time and response. As can be seen, the turbine was shutdown for repair. Turbine start up began after more than month out of production.

Assessing and reducing vibrations

Level 1 monitoring differentiates between remote monitoring of these overall vibration values and the remote monitoring of characteristic diagnosis values. The practice is not new to general industry. Vibration values from ISO 10816-3 are used to monitor the general vibration conditions while other guidelines use frequency-based or order-based characteristic trending values. Based on overall vibration values, it is now possible to assess and compare the vibration levels of wind turbines. The early detection and reduction of elevated vibration levels will extend wind turbine operational life.

Corrective measures

The required measures can be identified by means of condition diagnosis. Diagnosis specialists use amplitude spectra, envelope spectra, time signals, or spectra, or both to detect unusual vibration signals, identify dominant excitations, and evaluate frequency-specific trends using the water fall display function.

The polar plot shows the balancing result (amplitude and phase) of the wind turbine (initial unbalance, trim corrections, and final result). The chart graph displays the DIN ISO 1940-1 standard for rotor unbalance. In the wind power industry, the selected tolerance grade is G16 (as pointed about by the red line). For example, a wind turbine running 12 rpm the permissible residual unbalance to abide by is 16,000 g-mm/kg

Here are a few examples on how to increase the uptime and availability of wind turbines using results from vibration analysis:

Detecting vibration results from a generator faults: The accompanying illustration Detecting generator faults shows the trend of a wind-turbine generator in which an increase in vibration amplitude indicates an ongoing machine fault several weeks in advance. After replacing the generator, overall vibration values returned to normal. It should be noted that such vibration changes only arise if the affected drive train component is dominant in the frequency band. Overall vibration values do not rise when the damage is not dominant in the amplitude spectra.

Identifying deviations in the drivetrain alignment: Telemonitoring of a wind turbine in How alignment reduces vibration shows an increase in overall vibration values. Frequency analysis showed additional vibrations from poor misalignment between the generator and high-speed side of the gearbox. The machine was then aligned using accurate laser-alignment equipment with defined alignment targets. Overall vibration values dropped significantly.

Reducing blade imbalance: Unbalanced rotor blades can lead to rotational excitations and a load increase on bearings and other drivetrain components. Although frequencies are relatively low in wind turbines, resultant amplitudes can range up to 100 mm/s. Measurements must be taken with linear vibration sensors and longer measurement periods, as prescribed by VDI 3834. Applying the recommended G16 balancing grade for rotor blades, Permissible residual unbalance shows the resultant dynamic field balancing. In this example, other vibrations caused by imbalance, were reduced to the point where the difference was noticeable across the entire nacelle.

he graphical trend plot displays the vibration velocity before and after an alignment of the high speed side of the wind turbine.

These examples illustrate the targeted use of measuring and testing techniques that make it possible to reduce vibrations in working wind turbines. VDI 3834 lets manufacturers and operators assess the vibration condition of wind turbines and reduce them by implementing specific corrective measures in order to reach new threshold values.

These conditions, when left uncorrected, manifest themselves in damaged components along with associated collateral damage. This is the typical pay-a-little-now-or-a-lot-later scenario. The task is certainly one for condition monitoring, but this topic could be called “correctible detection”. Here’s an overview of three common correctible conditions along with how they can be detected and corrected.

Misalignment
It’s common between the gearbox output (or high-speed shaft) and the generator input shaft. Causes include a flexible bedplate, large temperature variations, cantilevered mounting, thermal growth, and others. It is prevalent enough that some manufacturers recommend shaft alignment at prescribed intervals while others specify laser-based tools for aligning this portion of the turbine. “Because of the dynamic movement and flexibility of the turbine, alignment tolerances are much broader than what we see in standard industrial applications,” says alignment expert Paul Berberian of Alignment Supplies Inc. “Sometimes acceptable wind-turbine-alignment specifications are 2 to 4 times higher than those found elsewhere.” When left uncorrected, high-speed-shaft bearings and generator input-shaft bearings suffer and fail at a faster rate that when corrected. In some cases the coupling also bears the brunt of neglect and fails.

Detection
Vibration analysis detects misalignment remotely and over a progression indicated by a specific vibration signature detailing the fault. Misalignment saps production and performance. The graphs in Before and after balancing shows a vibration reading or FFT spectrum between a misaligned gearbox and generator coupling. The vibration signals come from remote monitoring. The lower signal is partly due to aligned shafts.

Correction
An alignment using lasers has largely replaced older, more time consuming methods. The alignment tool mounts across the coupling with two lasers pointing at laser detectors. The shaft is rotated 180° or less and three points are taken to define a circle. The lasers represent the shaft centerlines so the difference in the two lasers provide a measure of misalignment which is corrected at least two ways:
• The generator’s mount-ing pads can be shimmed up or down to correct mis-alignment in those directions.
• The generator’s “jack bolts” can be moved side to side to correct misalignment in those directions.

Looseness
If you have ever been in a wind turbine, you have noticed things move around up there. Because of the nacelle’s yawing and pitching, things may come loose. When parts break, it is usually due to being under repetitive or fatigue loads they were never intended to handle. Such failures can be real photo opportunities.

Detection
Vibration analysis detects looseness as multiple frequency peaks. One can remotely isolate the source of vibration without an up-tower climb. This is a decent alternative to climbing and checking everything that could be loose.
Correction
This one is easy – tighten what’s loose to the specified torque.

When loose things rotate, they generate many noticeable vibration peaks.

Imbalance
The image Before and after balancing shows vibration signals from those periods. Several components are susceptible to imbalance in a typical wind turbine. These include generator fans, generator rotors, and couplings to name a few. The causes of an increase in imbalance can range from simple things such as debris build-up on blades to material degradation, and possibly damage in the field.
Blade imbalance was once a major issue on smaller and much older wind turbines, mostly kilowatt-class machines. A common solution was to add balance weights of up to a few pounds. As the industry has grown, so have blade manufacturing methods. Much improved quality control allows creating matched blade sets weighing within close tolerances.

Misalignment comes in several variations, each of which is detectable and correctable. Parallel misalignment, for example, can produce vibration signals of this sort.

Detection
Using vibration analysis, you can see imbalance clearly and its degree. There are a few different types of imbalance, each having a unique vibration signature. There are also a few different types of balancing dependent upon the machine and type of imbalance detected.
Correction:There is a three step procedure for balancing in a single plane. Getting started requires a vibration analyzer, tachometer (or strobe light), vibration accelerometer, and a constant running speed. It also requires eliminating any looseness and misalignment prior to attempting to balance a component. The three-step method first calls for a:

• Baseline run. This is where vibration from the imbalance is initially measured, creating a baseline.
• A trial run then attaches a weight as a trial to induce a 30-30 rule. This means you are looking for a 30% increase or reduction in vibration, or a 30% shift in phase. This also means the trial weight has had some measurable affect on the imbalance. A calculated correction weight is placed on the machine, completing the initial balance job.

Angular misalignment can produce signals of this sort.

• Lastly, trim runs are performed to finalize balancing when an acceptable imbalance is not yet within tolerance. Several trim runs may be needed before the balance signal is within specs. ISO standards provide acceptable balancing specs.

Remember two simple things: Detect and correct. Tools and technology can detect these correctible conditions, then choose the appropriate corrective action for each of the three conditions.

The previously mentioned faults are avoidable and detectible using basic technologies readily available to wind farm owners, O&M providers, and manufacturers. Of course this discussion is meant to be a cursory overview of the process, not a detailed description. The details should be part of critical asset management tasks. Correcting the problems described will improve a fleet’s overall reliability and wring maximum performance from it. WPE

David Clark/Director/Turningpoint Inc., El Dorado Hills, Calif./turningpointwind.com