Zach Wright
Engineering Analyst
Dr. Ashley Crowther
VP Engineering for Wind
Romax Technology Inc.,
U.S. Wind Technology Center
Boulder, Colo.
www.romaxtech.com
Gearbox reliability has been an issue for the wind industry since its inception. Premature failures increase operational costs for owners and warranty costs for OEMs. This seems a bit puzzling because gearboxes in other industries are reliable workhorses. Better design, manufacturing, and development will let the wind industry produce reliable gearboxes as well.
To aid such efforts, the National Wind Technology Center (NWTC) has been running the Gearbox Reliability Collaborative project (GRC) for the Department of Energy. The project brings together gearbox engineers, manufactures, owners, expert consultants, academics, and other partners from the wind industry in an effort to collectively improve gearbox reliability. The project aims to understand the causes of poor reliability, and decrease the costs associated with operations and maintenance of these major drivetrain components.
Project goals
Although the project uses a generic gearbox, significant steps have been made toward project goals. This article focuses on one aspect of the project: an engineering assessment of the function, strength, and durability of the gearbox.
One GRC goal is to instrument two gearboxes to measure tooth-root strains, bearing temperatures, bearing-raceway strains, gearbox motion, carrier motion, torque, speed, and other parameters. These allow a more detailed understanding of the internal behavior of the machinery under operation. This has been a successful phase of the project with tests performed at the NWTC 2.5-MW dynamometer facility and in the field at the Ponnequin Wind Farm in Colorado.
As part of the GRC analysis team, we validated CAE gearbox models against measured planet-to-ring gear-load distributions using RomaxWind software. The measurements are taken from strain gauges at three positions around the circumference of the ring gear. The three graphs in Normalized micro strain compare a model versus test results, and show good correlation at the three positions. The system’s physics and many parameters within gearbox software must be accurately modeled for such good correlation. A few parameters include clearances in the planet-carrier bearings and planet bearings, gear geometry and microgeometry, whole system structural deflections leading to certain gear and bearing misalignment. Additional parameters include weight loading of components, external loading (rotor-bending moments and torque), bearing and gear Hertzian contact stiffness, and deflections and flexibility of bearing raceways and gear blanks.
CAE models are used for gearbox design and assessment. From results in normalized graphs (only one shown) it is clear the planetary-stage ring gear has a poor load distribution with peak loading moving from the upwind to downwind side of the gear face width as the carrier rotates. This causes higher stresses in the gears, leads to premature failure, and also results in time-varying loaded contact distributions and stresses in the rolling-element bearings that support the planet gears. The software showed (reported in previous literature) that rotor bending moments, rotor weight, main-bearing mounting arrangement, carrier-bearing clearances, and the lead slope corrections on planet gears are key contributors to poor loading. Taking these factors into account when designing gearboxes ensures good load distribution and load sharing in the planetary sets. Thanks to the GRC program, these issues have been well disseminated to the industry.
Validating model results to test data leads to confidence in gearbox engineering assessments. Such assessments are made using industry standards and recommendations from certification bodies, as well as engineering know-how and good practice. Today’s analytical tools allow simulating the full spectrum of loading that a gearbox experiences in operation, and determines any design short-comings. Then it’s possible to calculate gear and bearing misalignments, strength, fatigue life, and vibration responses. Still, turbine loads are a worry for gearbox designers. Parallel efforts to improve turbine-load simulations have helped gearbox reliability. Institutions such as NREL and OEMs are performing more rugged and expensive testing by installing hardware on test rigs capable of dynamic torque. Driving such tests from measured field data helps close the loop between real turbine loads and gearbox design.
The GRC gearbox
A full design assessment has been performed on the GRC gearbox. This article provides a snapshot of key results. Design for good alignment is critical. System deflections of the support structure and shafts must be captured by the analysis software. This is important in wind-turbine gearboxes because the machinery must be lightweight to maintain a relatively low nacelle mass.
Graphs for bearing and gear misalignments show values calculated in RomaxWind under rated torque conditions. Bending moments include rotor and blade weight, and support of the main bearing. Allowable misalignments for functioning cylindrical and taper roller bearings are typically 1-μ radian. Values fall within this limit. However, misalignments predicted for the planet bearings (0.7-μ rad) are close to allowable limits so an additional load in the system, such as overload torque or large rotor-bending moments, may cause edge loading, or pinching, or both of the roller leading to premature bearing failure.
Gear misalignments are excessive, particularly in the sun-to-planet gear mesh. The planetary stage in this design is oversensitive to rotor loads. When the gearbox is in the turbine, the rotor and blade mass introduce a bending moment to the planet stage. In the dynamometer test (such as results in Normalized microstrains) the bending moment differs (it is now attributable to shaft and coupling masses) and may also be applied through a hydraulic system capable of applying nontorque loads to the test article.
Misalignments differ in the two cases: gearbox installed in a turbine versus the gearbox installed in a dynamometer. Consequently, engineers must design and test with the differing misalignments in mind.
For this particular gearbox, a torque-only-load spectrum of 31 load cases is available including positive and negative values with durations from a 750-kW turbine. This was typical for older machines. Modern methods use loads in several degrees of freedom, including rotor bending moments and forces as well as torque. Depending on the machine elements or structures under strength-and-fatigue analysis, the loading is processed several ways into extreme loads, or fatigue spectra in multiple dimensions. Many load cases are then run to assess the components versus the strength and fatigue limit criteria.
Here, analysis results are provided for the torque-only spectra and extreme torque loads, with inclusion of the bending moment from rotor and blade mass.
Bearing and gear fatigue-damage charts tell more about life expectancies. For the bearings, the calculation relies on standard, DIN ISO 281 Supplement 4, and for gears, ISO 6336. Bearings are typically rated to L10 life, that is 100% damage is equivalent of 10% of the bearings failing from rolling contact fatigue over the design life, 20 years for turbines. For gears, the rating considers bending and contact fatigue for 1% failure over the design life. Additional failure modes such as scuffing, micropitting, and skidding are assessed with other methods. Less well developed methods assess micropitting and skidding. The industry is doing more research and developing better standards in these areas.
Results also show the planet bearings are undersized for the application, and the high-speed bearings are marginal from a fatigue perspective. The contact and bending fatigue for the gearing is satisfactory. The illustration Contact stresses – high speed shaft shows the loaded contract distribution. The cylindrical-roller bearing has only a few rollers loaded due to its clearance. The taper-roller bearings are preloaded with a spring so all its rollers are in contact. However, axial thrust from the gear mesh unloads one bearing (HSS-B) while it increases the other bearing (HSS-C), evident from the loaded contract distributions. The taper angles on the bearings, the helix angle on the gears, and the amount of spring preload controls the load sharing between these bearings. The bearings and gears all have sufficient static-capacity margin (results not shown).
The engineering assessment of the GRC gearbox includes many design studies. Two studies deserve comment with regard to their influence on turbine reliability. Study One looks at the planet carrier bearings and the influence of rotor-bending moments about the horizontal and vertical plane. This drivetrain uses a three-point mount. That is, a spherical main bearing (1 point) supports thrust and radial loads while the gearbox trunnion mounts (2 points) handle moment loads. The figure shows how stress at the carrier bearings can rise considerably with significant moment loads. Extreme loads may damage bearings.
Study Two looks at an issue that plagues many currently installed gearboxes: Bearing-race slip, and wear between a bearing outer race and bearing bore that leads to a significant clearance between these elements. Wear leads to increased misalignment on the shaft, resulting in poor tooth-load distribution across the face width of the gears. This in turn, leads to gear-tooth wear and eventual failure.
A measure of the quality of the tooth-load distribution is the KHß factor (as defined in the ISO 6336 standard). It’s essentially a tooth peak load divided by mean load. The factor is used in gear ratings. The larger the factor, the more concentrated the tooth load. RomaxWind software allows considering bearing clearance so it calculates a gear-load contact distribution. For this gearbox, the influence of clearance in the intermediate shaft bearing has been charted against the KHß factor for the high speed and intermediate gears. The result helps explain how the wear issue leads to poor tooth contact and eventually breaking teeth in many gearboxes. The development of this type of CAE model (in original design, from prints or from measurements on tear down) in combination with engineering root cause analysis and gearbox know-how, often leads to design solutions. Also important to the refurbishment market, it allows reengineering and rebuilding gearboxes without plaguing them with the same issues.
The GRC continues with efforts in several areas. For example, industrial and university participants are validating software tools against a wealth of measured data that has been recorded for gearboxes. Companies that provide condition-monitoring equipment are applying their techniques to measurements to detect faults inside gearboxes. In addition, turbine owners are contributing gearbox teardown results to an NREL reliability database that is bringing together current statistics on the U.S. fleet. These efforts by NREL and the collaborative continue to help the industry improve gearbox reliability.
WPE
Filed Under: Uncategorized
George Fleming says
This gearbox research is remarkable, but it assumes that there must be at least one planetary stage in a wind turbine gearbox, starting on the low speed end. Three serious disadvantages of the planetary design are:
1. Load imbalance between the planets
AGMA recommends de-rating of the planet gears by 30% to account for this imbalance. The article refers to this load-sharing problem, but does not indicate whether it has a solution.
2. Size restriction because of the requirement for a ring gear
As the planetary stage rating increases, so must the diameter of the ring gear. The ring gear must be heat treated, and it should have helical teeth to reduce noise and improve load distribution. It becomes very expensive to produce such ring gears, and the number of manufacturers capable of it becomes smaller with each increase in diameter. At some point it won’t be worth the expense, even if it is possible. We have probably reached that point.
3. Design problems of the ring gear
The ring gear experiences fatigue from the passing planets. It is difficult to make the hardened ring gears in large wind turbine gearboxes resistant to this fatigue.
The article above states that the ring gear experiences poor fore-and-aft load distribution in wind turbines. I presume that the cause is mainly the periodic change in rotor shaft bending moment as the rotor blades pass the tower.
The periodic change in shaft bending also indicates a periodic change in the various shaft alignments. The article indicates that correct shaft alignment is critical to achieving long gearbox life.
It is true that the change in rotor bending moment is specific to wind turbines, not the planetary gearbox design. The question is whether the planetary design, regardless of its other faults, is best suited for accommodating this periodic change in bending moment, which is inherent in horizontal axis wind turbines.
Regarding the GRC gearbox, which we know can be improved, but we don’t know by how much, the article states:
“Gear misalignments are excessive, particularly in the sun-to-planet gear mesh. The planetary stage in this design is oversensitive to rotor loads.”
Final comments
Research indicates that the multi-branch parallel shaft gearbox design is equal to the planetary type in power density, weight and volume. It is probably more efficient and reliable, and less expensive. It has a much greater rating potential because it does not use ring gears of any kind, only external gears. Development of larger wind turbines requires this technology.