Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
WEAR-OUT PATTERN RECOGNITION
Document Type and Number:
WIPO Patent Application WO/2010/069318
Kind Code:
A1
Abstract:
The present invention relates to a method of providing an estimate of remaining lifetime of a component in a wind turbine. The method may comprise providing a model for the component, obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time, and determining an estimate of remaining life-time of a component based on the model and the sensor signal. The present invention further relates to a system for providing an estimate of remaining life-time of a component in a wind turbine. The system may comprise a memory unit adapted for storing a model for the component, a data receiver adapted for obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time, and a processor adapted for providing an estimate of remaining life-time of a component based on the model and the sensor signals.

Inventors:
ZANG, Tie Ling (BkI 708 Clementi West Street 2 #07-303, 8 Singapore, 12070, CN)
SANDVAD, Ingemann Hvas (Musvitvej 14, Ringkøbing, DK-6950, SG)
Application Number:
DK2009/000257
Publication Date:
June 24, 2010
Filing Date:
December 16, 2009
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
VESTAS WIND SYSTEMS A/S (Alsvej 21, Randers SV, DK-8940, DK)
ZANG, Tie Ling (BkI 708 Clementi West Street 2 #07-303, 8 Singapore, 12070, CN)
SANDVAD, Ingemann Hvas (Musvitvej 14, Ringkøbing, DK-6950, SG)
International Classes:
G07C3/08
Foreign References:
DE19713583A11998-10-08
DE10353647A12004-10-21
DE10144076A12003-03-27
DE10315630A12004-10-28
Other References:
None
Attorney, Agent or Firm:
ZACCO DENMARK A/S (Hans Bekkevolds Allé 7, Hellerup, DK-2900, DK)
Download PDF:
Claims:
CLAIMS:

1. A method of providing an estimate of remaining life-time of a component in a wind turbine, the method comprising: providing a model for the component, obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time, and determining an estimate of remaining life-time of a component based on the model and the sensor signal.

2. The method according to claim 1 or 2, wherein the sensor registers temperature and/or vibration and/or a combination thereof.

3. The method according to any of the claims 1-4, wherein the method further comprises averaging and/or smoothing and/or filtering the registered sensor signal before fitting to the model.

4. The method according to claim 1 , wherein the model is based on empirical data obtained from identical or similar components.

5. The method according to any of the claims 1-4, wherein the component is a bearing or gear or any other moving component in the wind turbine.

6. The method according to any of the claims 1-5, wherein the method further comprises on the basis of the model and the sensor signal establishing if the component is in need or repair or needs to be replaced.

7. The method according to any of the claims 1-6, wherein the method further comprises transmitting a first type warning signal if the sensor signal fulfills a first criteria and/or a second type warning signal if the sensor signal fulfills a second criteria and/or a third type warning signal if the sensor signal fulfills a third criteria, the first, second and third type warning signal indicating different levels of severity.

8. The method according to claim 7, wherein the criteria includes the registered signal being more than 1 σ away from a predefined threshold.

9. The method according to any of the claims 1-8, wherein the method comprises obtaining a sensor signal in a plurality of time periods and on the basis of on the model and the plurality of sensor signals determining the estimate of remaining life-time of the component.

10. A system for providing an estimate of remaining life-time of a component in a wind turbine, the system comprising: a memory unit adapted for storing a model for the component, a data receiver adapted for obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time, and a processor adapted for providing an estimate of remaining life-time of a component based on the model and the sensor signals.

1 1. The system according to claim 10, further comprising a display unit for displaying the estimate of remaining life-time and/or a transmitter unit for transmitting an alarm to a technician or monitoring person.

12. The system according to claim 10 or 11 , wherein the processor is adapted for averaging and/or smoothing and/or filtering the registered sensor signal before fitting to the model.

13. The system according to any of the claims 10-12 wherein the system is configured to perform any of the steps defined in the claims 1-9.

Description:
Wear-out pattern recognition

The present invention relates generally to a method and a system for determine an estimate of component life-time. The method may involve wear-out pattern recognition.

A first aspect of the present invention relates to a method of providing an estimate of remaining life-time of a component in a wind turbine. The method may comprise providing a model for the component, obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time, and determining an estimate of remaining life-time of a component based on the model and the sensor signals.

A second aspect of the present invention relates to a system for providing an estimate of remaining life-time of a component in a wind turbine. The system may comprise a memory unit adapted for storing a model for the component, a data receiver adapted for obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time, and a processor adapted for providing an estimate of remaining life-time of a component based on the model and the sensor signals.

The present invention will be discussed in more detail with reference to the appended figures illustrating exemplary embodiments of the present invention wherein:

Fig. 1 is a schematic illustration of a chart, Fig. 2 is a schematic illustration of a first graph illustrating temperature variations, Fig. 3 is a schematic illustration of a second graph illustrating data and a model, Fig. 4 is a schematic illustration of a third graph illustrating data and a model, Fig. 5 is a schematic illustration of a fourth graph illustrating alarm levels, Fig. 6 is a schematic illustration of a method, Fig. 7 is a schematic illustration of first system, and Fig. 8 is a schematic illustration of a second system.

When a wind turbine is operating it is desirable to avoid stops due to broken or worn- out components. The present invention is contemplated to allow an optimized planning of repair and replacements of components. It is an object of the present invention to provide a method and a system for determining an estimate of the remaining life-time of a component in a wind turbine. Further, it is an object of the present invention to provide a system for determining an estimate of the remaining life-time of a component in a wind turbine. It is still a further object to provide a method and system for determining an estimate of the remaining life-time of several components in a wind turbine.

Using recorded data from wind turbines and recorded failures and faults, a model based on these statistics may be defined or established. The model is contemplated to allow a system to compare the current state of a wind turbine component to the model and based on this to determine or establish an estimate of time to failure or remaining life-time for that wind turbine component.

Fig. 1(a) depicts a schematic chart 10. The chart 10 schematically illustrates a model which is established to represent a specific performance or evaluation of health status of a component by selected sensor signals from sensors registering information regarding that component. The model may be related to a regression equation of the selected parameters and may be built through the recorded data of those parameters recorded from operational wind turbines. The histogram may represent the distribution of deviation between the model response, for example, generator bearing temperature, and real recordings in a specific period of time. Chart 10 shows three histograms, each depicts the model by using the data of a group of the same wind turbines, the data of wind turbines operating in the same park and the data of a specific wind turbine which is running in that park.

The model may be used in a system configured for monitoring sensor data from a sensor positioned at or on a component in a wind turbine. For example, a model of generator bearing temperature may be provided for a specific type of wind turbine, shown in Fig. 1 as the solid line 16. This model is based on a large number of measurement and observations yielding the establishment of the solid line 16. Also, this model for a wind park with a certain composition of wind turbines, e.g. a number of wind turbines may be provided, and illustrated as the histogram 18. Further this model for a single turbine, e.g. a wind turbine in the above-mentioned wind park, may be provided and is illustrated as the histogram 20 in Fig. 1 (a) and Fig.1 (b).

The models established also provide the possibility to establish alarm thresholds. One alarm threshold could e.g. define that an alarm should be raised if measurements from one wind turbine are 1 , 2 or 3 standard deviations from the model distribution originating from the park as a whole where the wind turbine is located. An alarm threshold is illustrated in Fig. 1 as the line 22.

In Fig. 1 , the difference between Fig. 1 (a) and Fig. 1 (b) is that each chart 10 is generated from data set in different time frame. Chart 10 in Fig. 1 (a), for example, is plotted based on empirical data that were recorded two months before the chart 10' in Fig. 1 (b). In another meaning, the component remaining life-time may be predicted by identifying how far a specific turbine's histogram deviates from the alarm 22 with comparison to the recorded patterns that reflect different remaining life-time.

In some surveillance systems it is only possible to detect the occurrence of a fault. Advantageously, the method and system according to the present invention is contemplated to provide the possibility of predicting when a fault will occur and thereby allow a technician to take action before the fault actually occurs. The model may provide a prediction of single component remaining life-time, which is also contemplated to be advantageous when planning repairs and component purchase, i.e. planning of when to purchase and stock replacement components, etc. Further, the method and system may provide estimations for a multitude of components in one or more wind turbines.

In one embodiment, a sensor registers temperature of a generator bearing. The temperature curve 24 is illustrated in Fig.2. The X-axis represents time and the Y-axis represents temperature.

The temperature curve 24 will vary depending on the load, wind speed, lubrication system, temperature, cooling system etc. The temperature is contemplated to be related to the state of the component. A more worn or deteriorated component is contemplated to have a higher temperature, as indicated in the left-hand part of the curve 24 in Fig. 2.

The model includes recorded data from a number of measurements and other data obtained empirically. The data recorded by one or more sensors, e.g. such as the data forming the basis of the curve in Fig. 2, is fitted to the model. Before the data is fitted, the data may undergo a filtering or smoothing process, allowing a better fit to the model. The process is contemplated to filter or smooth out the relative large variations in the curve of Fig. 2, and that these large variations do not effect the establishment of the current state of the component. Also it is contemplated to allow a better and faster fitting of the data by smoothing or filtering the data in advance.

Fig. 3 illustrates the data 26 and the model 28.

Fig. 4 illustrates the model 28 and the fitted data set 30. The punctured line 32 illustrates the time of observation. From the model 28, it is possible to determine the estimated time of failure 34 by the fitted data 30 and the observation time. The difference between the time of observation 32 and the estimated time of failure 34 is the estimated remaining life-time of the component.

In one embodiment, the model 28 may be a generic reference for life-time curves for all generator bearings. The model 28 may be a measure or prediction for the same type of failure/wear-out for similar components. The model may be expanded or improved by adding empirical data sets and observations registered in relation to components in other wind turbines. The more events of worn-out components that are included in the model, the more mature the model is. Each monitoring system according to the present invention may also register data and transmit or store this data for allowing a central data processing station to include the data in a new or updated model. Through this way, the model may be improved or strengthened continuously.

In an advantageous embodiment, a monitoring system will record the generator bearing temperature periodically, in an embodiment each 10 minutes, perform a computation of the current state and plot the generator bearing temperature vs. time curve, e.g. as information to an operator or maintenance technician or the like. In an embodiment the information may be used in turbine generator control for extending the component life-time by reducing acting load on the component.

The monitoring system may compare the observed or recorded data, shown as the curve 30 in Fig. 4, with the model 28. If the recorded data 30 differs from the approximately constant value, i.e. the more flat part of the curve 28, the system compares the data to the model to determine if the data fits a wear out pattern. If it is established that the data fits a wear-out pattern, the system calculates the estimated remaining life-time. Different components may have different types of wear-out patterns so that the system may be adapted or configured to compare the data to a number of patterns or models to identify the most likely model.

As mentioned above, one or more alarm thresholds may be defined. For example, a level 1 alarm could be raised if the overall temperature starts to increase and/or if a turbine temperature distribution is 1σ away from the overall or park distribution.

Another alarm, e.g. a level 2 alarm, could be raised if the overall temperature or component life-time has reached a certain limit and/or when the single turbine distribution is 2σ away from the overall or park distribution.

A further alarm, e.g. a level 3 alarm, could be raised if the overall temperature has reached a certain limit and/or is going to reach its maximum component lifetime and/or when the single turbine distribution is 3σ away from the overall or park distribution.

Fig. 5 schematically illustrates different levels of alarm settings in different ways. In Fig. 5, three thresholds 36, 38 and 40 are illustrated as punctured lines, where 36 is a level 1 threshold, 38 is a level 2 threshold and 40 is a level three threshold.

A monitoring system as described above may be added or connected to an existing wind turbine or wind turbine park. This could be advantageous as the system may detect faults that would not otherwise be detected before a component or the entire wind turbine fails.

Fig. 6 schematically illustrates steps of one embodiment of a method according to the present invention.

The method of providing an estimate of remaining life-time of a component in a wind turbine may comprise the steps illustrated by the boxes in Fig. 6. In a presently preferred embodiment, the method comprises providing a model for the component 42, obtaining a sensor signal from a sensor registering information regarding the component and registering the sensor signal for a first period of time 44. Further, it is to determine an estimate of remaining life-time of a component based on the model and the sensor signals 46. In a further embodiment, the method may include filtering or smoothing module 48 that preprocesses the recorded data, as illustrated by the punctured line in Fig. 6.

Still, in further embodiments, the method may include registering data in a plurality of time periods. The method may in such embodiments include comparing or using the plurality of measurements for determining the estimated remaining lifetime of the component.

In an advantageous embodiment of the present invention, the sensor registers temperature and/or vibration and/or a combination thereof. The estimate may be based on one or a combination of the signals from the sensors. In some embodiments, the estimate may be based on several sensors registering the same type of signals on the same component. This is contemplated to be advantageous when determining remaining life-time of large components, or components having several sub- components, or component having different failure modes.

In a still more advantageous embodiment, the method may further comprise averaging and/or smoothing and/or filtering the registered sensor signal before fitting to the model. As described above, this may be used as a preprocessing of the data. It is contemplated to be advantageous to preprocess the data. The preprocessing may provide an easy fit to the model and/or a faster calculation of the estimate of remaining life-time.

In an even more advantageous embodiment, the model is based on empirical data obtained from identical or similar components. The model may be built on a number of observations of other similar components. The model may be updated or improved by including new data. The new data may originate from the component being monitored. The model may include different information relating to the component and may include different types of information relating to the present state of the component, e.g. current number of revolutions or speed. The model may also include information regarding surroundings, e.g. ambient temperature and/or wind speed around the wind turbine.

In specific embodiments, the component may be a bearing or gear or any other moving component in the wind turbine. It may be further advantageous that the method may further comprise establishing on the basis of the model and the sensor signal if the component is in need or repair or needs to be replaced. This information is contemplated to be of use e.g. in planning future repairs, or even immediate need of repair or replacement. The method may result in the establishment of a plan for repair of a number of components and/or a component or components in a number of wind turbines.

Also the method may even further comprise transmitting a first type warning signal if the sensor signal fulfills a first criteria and/or a second type warning signal if the sensor signal fulfills a second criteria and/or a third type warning signal if the sensor signal fulfills a third criteria, the first, second and third type warning signal indicating different levels of severity. These criteria may include temperature or vibrations or the like. In a specific advantageous embodiment, the criteria may include the registered signal being more than 1 σ away from a predefined threshold, e.g. a temperature distribution. Other thresholds are mentioned above.

Fig. 7 schematically illustrates a system 50 for providing an estimate of remaining lifetime of a component in a wind turbine. The system 50 comprises a memory unit 52 adapted for storing a model for the component. Also the system 50 comprises a data receiver 54 adapted for obtaining a sensor signal from a sensor 56 registering information regarding the component and registering the sensor signal for a first period of time. The system 50 further comprises a processor 58 adapted for determining an estimate of remaining life-time of a component based on the model and the sensor signals. The processor may be configured or adapted to perform steps of the method mentioned above.

The elements in the system 50 are electrically connected. The system 50 may also comprise other elements not illustrated for the system 50 to work, e.g. power supply etc.

Fig. 8 schematically illustrates a system 50' similar to that of Fig. 7 identical numerals are used for similar components.

The system 50' further comprises a transmitter 60. The transmitter 60 may transmit data from the system 50' to an external unit, e.g. a central computer system monitoring a wind park. The transmitter 60 may also send data to a computer unit planning repairs. The transmitter 60 may also be configured to transmit data to the turbine controller of a wind turbine to adjust one or more operation parameters of the turbine in order to extend the component's lifetime and have the replacement or maintenance done in a scheduled periodical maintenance cycle of the wind farm or during a low wind season.

Advantageously, the system may further comprise a display unit for displaying the estimate of remaining life-time and/or a transmitter unit for transmitting an alarm to a technician or monitoring person. This is contemplated to allow a service technician or person monitoring or inspecting the status of the wind turbines to determine if one or more components need to be repaired. If the system is connected to a transmitter, the information may be transmitted to a remote location. The information may be transmitted to a central computer unit or the like. The transmitter may also be adapted or configured to transmit data relating to the sensor signal.

In one embodiment, the processor is adapted for averaging and/or smoothing and/or filtering the registered sensor signal before fitting to the model.

The system according to the present invention may further be configured or adapted to perform any of the steps mentioned in relation to the method above.

The above method and system are contemplated to provide several advantages including an overall improved quality and/or life time of the wind turbine.

Further the method and the system are able to provide precise predictions for when the component will go into wear-out process and when it will be worn out so as to avoid immediate preventive exchanges and it can also help predict if a component may operate until next service visit or may last for several years.

Still further service may be provided to customers with turbines that are running out of warranty, thereby allowing better planning of preventive repairs and updates.