Field of the Invention The invention relates to a bearing unit provided with sensors and processing means for detecting a shock load on the bearing. The bearing unit may be used for condition monitoring of wind turbines, underwater turbines and other rotating applications where the bearing is subject to shock loading. Background of the Invention

Condition monitoring of wind turbines is an important issue, in particular in the case of off-shore wind parks which are difficult to access by maintenance staff. Bearings for use in wind turbines, in particular main shaft bearings holding a rotor, have very large dimensions and have to support immense loads while reliably operating under a wide range of environmental conditions. Replacement is complicated and expensive and should be avoided as far as possible. It is therefore very important to provide reliable estimates of bearing lifetime, which is usually done using parameters measured by sensor systems on the bearing. It is known to provide temperature sensors, strain sensors, vibration- or acoustic emission sensors and acce I ero meters on bearing rings.

Various technologies using strain sensors targeting to measure load on the side face of a bearing have been proposed. The sensors correlate the strain/displacement on the bearing side face with individual roller loads by means of a mathematical model. The peak-to-peak values of the sensor signal correspond to the roller load.

However, it transpires that an important contribution to lifetime estimates of bearings and other components in wind turbines are shock/transient loads. These happen e.g. during emergency shutdown, swivelling of the wind turbine nacelle, manoeuvres, abrupt changes in wind conditions or turbulences. The frequency of the shock/transient loads is significantly higher than the roller pass frequency dominating the sensor signals corresponding to the roller load.

There is always a trade-off between signal processing performance, sensor quality and sensor quantity necessary to have the best possible performance/price ratio.

Reliable shock detection is further important to trigger emergency stops of a wind turbine or other machine in which the bearing is mounted. Accelerometers are used to detect shocks in many applications. However these do not allow quantification of the load on the bearing.

Another possibility would be to attach a strain gauge to a structure which is carrying the load, this could be a shaft or a surrounding component. However this is not possible with a bearing because the load is continuously changing in a macroscopic scale caused by the rollers passing by.

It is therefore an object of the invention to provide a bearing unit, which is suitable for reliably detecting shock loads or transient loads on the bearing.

Summary of the Invention

The invention relates to a bearing comprising an inner ring, an outer ring and at least one row of rolling elements arranged between the raceways in a cage, wherein a predetermined equidistant angular spacing of the rolling elements is maintained by the cage. The rolling elements may in particular be cylindrical, spherical or toroidal rollers. However, the invention is not limited to any particular type of rolling elements. The bearing further includes at least two sensors attached to one of the bearing rings at different predetermined angular locations. The sensors are configured to generate a signal having one signal component with a frequency of the rolling elements passing by the locations of the sensors. It is proposed that the predetermined angular locations of the at least two sensors are chosen such that phase angles of said one signal component are different. Due to the different phase angles, the components corresponding to the roller pass frequency in the two signals cancel each other out at least partially when the signals of the sensors are summed, such that the weight of the other components in the signal, in particular the contributions of transient or shock loads, is increased correspondingly.

The invention is suitable for use on main shaft bearings of wind turbines but may be applied to any kind of bearing. Preferably, the bearing has a small ratio of ring width to bearing diameter, the ratio being preferably below 0.1 .

Other applications include large-size bearing in general, slewing bearings for cranes, and slewing bearings for pods in ship thrusters.

In preferred embodiments including a number n of sensors, the phase angle of the i ^th sensor corresponds to i/n ^* 360°. Expressed otherwise, the locations of the sensors are equidistantly distributed over the period length of the rollers. In this context, "corresponds to" relates to an equivalence relation, wherein phase angles differing by integer multiples of 360° and angular differences between angular locations on the circumference of the rings differing by integer multiples of the pitch or spacing between adjacent rollers are considered equivalent. In cases where more than two sensors are used, it is possible to detect uneven deformations or higher multipole moments of the deformations of the ring.

In a preferred embodiment of the invention, only two sensors are used and the predetermined angular locations are chosen so as to correspond to an uneven integer multiple of one half of the angular spacing between the rolling elements such that the components with the frequency of the rolling elements have opposing phase angles in the signals of the at least two sensors. In embodiments with three sensors, these could be arranged with phase angles of 120° between adjacent sensors and in embodiments with four sensors, these could be arranged with phase angles of 90° between each two of adjacent sensors. The sensor locations need not be homogeneously distributed over the circumference of the bearing ring. What counts is the distribution over the period of the rollers within the equivalence relation defined above.

The bearing unit of the invention further comprises signal processing means configured to sum up signals of the at least two sensors and to detect a shock load acting on the bearing, based on the sum signal. The summing may be done by software using digitalized signals or using an electronic circuit.

Suitably, the sensors used in the inventive bearing unit are sensors which can detect elastic deformation of the bearing ring due to the passage of the rolling elements. In one embodiment, the sensors are formed as fibre-Bragg gratings in an optical sensing fibre, which is attached to an outer circumferential surface of the bearing outer ring. The optical sensing fibre may also be attached to an inner circumferential surface of the bearing inner ring. Preferably, the optical sensing fibre is provided in a circumferential groove in the inner/outer bearing ring.

In a further embodiment, strain gauges are used, whereby the sensors may be provided in a sensor package having a housing and optionally further sensors such as a vibration sensor, a temperature sensor, an acoustic emission sensor, an accelerometer or the like. The at least two strain gauges may be mounted to a circumferential surface of one of the bearing rings, or to a side face of one of the bearing rings. Preferably, the at least two strain gauges are provided in a recess of the bearing ring.

When the at least two sensors are attached to an axially oriented side face of a bearing ring, the bearing preferably has angled raceways, tapered raceways or spherical raceways. In other words, the rolling element bearing is a type of bearing that is adapted to withstand not only radial loads, but also axial loads. In such bearings, deformation of the bearing ring due to the passage of rolling elements is more readily detectible at the axial side face of a bearing ring. The sensorized bearing unit as descried above lends itself to the implementation of a method for detecting a shock load on a rolling element bearing. The invention proposes that the method comprises the steps of:

· providing a ring of the bearing with at least two sensors, located at different predetermined angular locations, wherein the sensors are configured to generate a signal having one signal component with a frequency of the rolling elements passing by the locations of the sensors, wherein the predetermined angular locations of the at least two sensors are chosen such that phase angles of said one signal component are different;

• monitoring the signals of the at least two sensors;

• summing the signals in order to obtain a signal sum;

• evaluating the signal sum so as to detect a shock load if the signal sum exceeds a predefined threshold value.

Suitably, the method may further comprise storing or forwarding a signal indicating the detected shock load.

The above embodiments of the invention as well as the appended claims and figures show multiple characterizing features of the invention in specific combinations. The skilled person will easily be able to consider further combinations or sub-combinations of these features in order to adapt the invention as defined in the claims to his or her specific needs. Brief Description of the Figures

Fig. 1 is a schematic representation of a bearing according to the invention; Fig. 2 is a series of graphs showing different types of load signals obtained by sensors of a main shaft bearing of a wind turbine;

Fig. 3 is a series of graphs showing the signals obtained by sensors of the bearings according to Fig. 1 and of a sum thereof; and Fig. 4 illustrates a network of wind turbines equipped with bearings according to Fig. 1 and a control- and monitoring server.

Detailed Description of the Embodiments

Fig. 1 is a schematic representation of a bearing unit 10 according to the invention. The bearing 1 0 is a double row taper roller bearing with an outer ring 10a and an inner ring 10b which have tapered raceways. The bearing 10 is designed for use as a main shaft bearing for a wind turbine and has a very large diameter of up to 2 - 4m. At least one row of rollers 12 is arranged between the raceways of the rings 10a, 10b in a cage 14, wherein a predetermined equidistant angular spacing δ ₂ of the rolling elements 1 2 is maintained and defined by the cage 14. The invention is not limited to any specific kind of rolling element bearing and the bearing 1 0 can have one or two rows of rolling elements and the rolling elements can be balls or spherical, cylindrical or toroidal rollers.

Two strain sensors 1 6, 1 6' are attached to an axially oriented side face of the bearing outer ring 10a at different predetermined angular locations, wherein the sensors 1 6 ,1 6' are configured to generate a signal having one signal component with a frequency of the rollers 12 passing by the locations of the sensors 16 ,1 6'. The latter frequency is sometimes referred to as the roller pass frequency.

The predetermined angular locations of the at least two strain sensors 16, 1 6' are chosen such that phase angles of said one signal component are different. In other words, the angular spacing δι of the sensors 16, 16' differs from an integer multiple of the angular spacing δ ₂ between the angular locations of the rollers 1 2, which is equal to the 360° divided by the number of rollers.

In the cost-saving embodiment of Fig. 1 , only two sensors 1 6, 1 6' are used. In this case, the predetermined angular locations are chosen so as to correspond to an uneven integer multiple of one half of the angular spacing δ ₂ between the rolling elements 1 2 such that the components with the frequency of the rolling elements 12 have opposing phase angles in the signals of the at least two sensors 1 6, 16'.

The sensors 16, 16' are formed as strain gauges and are embedded in pertinent recesses in the outer ring 1 0a of the bearing 10 and are integrated in a sensor housing containing basic signal processing and energy harvesting means as well as a wireless or wired communication interface for communication the direct results or the pre-processed results to an outside monitoring unit 18. The monitoring unit 18 may be arranged in a nacelle of the wind turbine or in a remote server 20 (Fig. 4) for maintenance data. The monitoring unit 18 is a signal processing means configured to sum up signals of the at least two sensors 16, 16' and to detect a shock load acting on the bearing 10, based on the sum signal as shown in further detail in Fig. 3. The monitoring unit 1 8 implements a method for monitoring a wind turbine having a bearing 10 as shown in Fig. 1 . The method comprises the steps of summing the signals in order to obtain a signal sum; evaluating the signal sum so as to detect shock load; storing information on the shock load in a storage unit of the monitoring unit 18 and forwarding a signal relating to the shock load to a remote maintenance data server 20 (Fig. 4) storing data used for calculating the expected remaining bearing lifetime. A shock load is detected once the sum signal exceeds a predetermined threshold value Θ, which may depend on an average value of the signal in the last few minutes or second. In possible embodiments of the invention, the detection of a shock load may lead to an emergency stop of the windmill and to a swivelling of the windmill head. The threshold for the emergency stop can be set to a higher value than the threshold Θ used for identifying the shock events which are accounted for in the context of the lifetime calculation.

Fig. 2 is a series of graphs showing different types of load signals obtained by sensors 16, 16' of a main shaft bearing of a wind turbine. Periodic oscillating signal components as illustrated in the uppermost graph exist mainly due to the passing rollers at the roller pass frequency and at lower frequencies due to a mass imbalance of the rotor and due to the wings passing the tower. The load signal of the sensors further comprises stochastic components due to regular hydrodynamic and aerodynamic forces as illustrated in the graph in the middle and transient events caused a stop of the windmill or a swivelling of the windmill head as illustrated in the graph on the bottom of Fig. 2. The transient events result in a peak with a width of the order of several seconds. Shock loads can be caused e.g. by abrupt blasts of wind, forces from plunging breaking waves or seismic events and have a shape of a very sharp peak and happen on even smaller timescales. Fig. 3 is a series of graphs showing the signals obtained by sensors of the bearings according to Fig. 1 and of a sum thereof. The uppermost graph is a signal of the first sensor 16 and the second graph from the top is a signal of the second sensor 16'. Both sensor signals contain a basic sinusoidal component of the signal caused by the passing rollers, wherein these components have opposing phases for the two components.

The lowermost graph is a sum of the two signals from the sensors 16, 16'. Due to the opposite phases, the basic roller-pass components of the signals cancel each other out such that the weight of the remaining components, in particular of a peak in the signal stemming from a shock event as illustrated on the right-hand side of the graphs in Fig. 3, increases such that the evaluation of the remaining components and the detection of transient or shock loads is facilitated. The threshold Θ is illustrated as well. The visualization of sensor signals in Fig. 3 shows that the sensor signal of the sensor 16 and the sensor signal of sensor 16' can be summed up and will then partially cancel each other out. A shock load will lead to a peak in both sensor signals and in the sum signal, where it can then be reliably detected and quantified.

Fig. 4 illustrates a network of wind turbines 22 equipped with bearings 10 and monitoring units 18 according to Fig. 1 and a control- and monitoring server 20. Once the sum signal obtained by the monitoring unit in the nacelle of the wind turbine 22 based on the signals of the two sensors 16, 16' exceeds the threshold value Θ, the monitoring unit 18 produces a signal indicating a shock event, which is then sent with accompanying information (wind turbine identifier, time, and strength of the shock) to the remote monitoring data server 20. For very strong shocks, the corresponding wind turbine 22 is stopped and swivelled out of the wind.