Method and computer programme for detecting the stationary state of a roller bearing and a roller bearing which may be analyzed thus

The present invention relates to a method and to a computer program product (also called computer program or software for short) as claimed in claim 1 and 19 , respectively, for detecting the stationary state of a roller bearing, and a roller bearing which may be analyzed with the aid of the aforementioned method, as claimed in claim 20 .

Roller bearings are used in every machine in the industrial field. Due to the continuously increasing demands on the service life and the operational reliability of such machines, there is an increase in demand for being able to determine whether the roller bearing is rotating or is actually stationary. This information can be obtained with difficulty, in particular, when the roller bearing is possibly rotating very slowly because it is then possible, in the case of signals which are recorded and evaluated for detecting a stationary state, to distinguish only with difficulty between stationary state and slow rotation due to the noise in the evaluation electronics or in the sensor arrangements and due to the fact that the signal gradients are often only small in this boundary area.

The present invention is based on the object of creating a method for detecting the stationary state of a roller bearing which detects with reliability but is as efficient as possible and which should also be implementable as a computer program. Furthermore, a roller bearing is to be created which can be or is connected to an evaluating device in which the detection of the stationary state is carried out reliably and as efficiently as possible.

This object is achieved by means of a method as claimed in claim 1 and by means of a computer program according to claim 19 , respectively, and a roller bearing according to claim 20 . Advantageous embodiments of the invention are the subject matter of the dependent claims.

The method according to the invention according to claim 1 provides for sampling the sinusoidal signal depending on the rotational position of the roller bearing during the rotation of the roller bearing in a number of intervals, calculating the average value of the samples in each interval and comparing the average value of the first interval with the average values of the subsequent intervals. The roller bearing is considered to be stationary for as long as the average values of the following intervals do not significantly differ from the average value of the first interval. As an alternative, the roller bearing can be considered to be stationary for as long as the gradient between the first average value and the respective average value of the further intervals does not significantly differ from 0. As a further alternative, a stationary state of the roller bearing can be assumed as long as either one of the two aforementioned conditions is met or cumulatively both of the aforementioned conditions are met.

According to the preferred method according to claims 2 to 8 , the respective average values are determined particularly efficiently and with very little storage space requirement, the comparison of the following average values with the first average value or the comparison of the gradients with the number 0 being carried out as comparison with a preferably adjustable threshold value.

The speed of the detection of a stationary state of the roller bearing can be increased when, according to claim 11 , the intervals adjoin one another, or according to claim 10 , are interlocked with one another, wherein the method can be made simple if, according to claim 9 , an equal number of sampling times are allocated to each interval.

The method according to the invention operates particularly effectively if each of the time intervals maximally covers a rotational angle of the roller bearing which corresponds to less than one quarter of the complete distance covered by a rolling body rolling over a sensor—that is to say an angle of rotation of 90°/z in the case of z rolling bodies.

If at least two sensor arrangements are provided in a roller bearing, the distance from one another of which is unequal to an integral multiple of the distance between two adjacent rolling bodies, the method according to the invention according to claim 18 can also be used for detecting a possibly occurring to and fro movement of the roller bearing—usually called “bucking”—by an angle of rotation of less than 360°/z when the signals of all sensor arrangements are evaluated. If more than z sensor arrangements are provided, this can be used for making the distance between the sensor arrangements from one another—in the circumferential direction of the roller bearing—smaller than the distance between two adjacent rolling bodies from one another—also in the circumferential direction of the roller bearing, as a result of which, when the signals of all sensor arrangements are evaluated in parallel, any bucking which may occur can be detected particularly reliably.

The roller bearing according to the invention according to claim 20 comprises an evaluating device by means of which the methods according to claims 1 to 18 or, as an alternative, the computer program according to claim 19 , which can be stored on a storage medium (e.g. RAM, ROM, CD, DVD, floppy disk, hard disk, flash memory etc.) and/or can be called up via a network, can be carried out. This takes into consideration that the evaluating device preferably constructed as ASIC in a chip only has limited computing capacities in its size due to the placement in the groove of the outer ring. The width of the chip is determined a priori by the width of the outer ring of the bearing. In addition, the chip is not too long in the circumferential direction of the outer ring and since otherwise the chip located in the groove would be bent disproportionately due to the curvature of the outer ring and thus a defect would threaten.

Further advantages, features and peculiarities of the invention are obtained from the subsequent description of preferred but not restricting embodiments of the invention by means of the schematic drawings, which are not true to scale, in which:

FIG. 1 shows a schematic representation of a roller bearing according to the invention,

FIG. 2 schematically shows various possibilities of the arrangement of strain gauges in a Wheatstone bridge circuit in the outer groove of a roller bearing according to the invention,

FIG. 3 a to FIG. 3 c in each case show an interference-free section of a signal to be evaluated, with various possibilities of the arrangement of the sampling intervals, FIG. 3 a showing adjoining sampling intervals, FIG. 3 b showing spaced-apart sampling intervals and FIG. 3 c showing interlocked sampling intervals,

FIG. 4 shows a flow chart of a first embodiment of the method according to the invention,

FIG. 5 shows a flow chart of a second embodiment of the method according to the invention,

FIG. 6 shows a flow chart of a third embodiment of the method according to the invention,

FIG. 7 shows a flow chart of a fourth embodiment of the method according to the invention, and

FIG. 8 shows a flow chart of a fifth embodiment of the method according to the invention.

Initially, a roller bearing according to the invention will be described by means of a preferred embodiment and then advantageous embodiments of the method according to the invention for detecting the stationary state will be described in detail by using flow charts.

FIGS. 1 and 2 represent the main components of the roller bearing 20 according to the invention, a so-called intelligent bearing. The representation is symbolic and is used for illustration but is not to be considered as restrictive.

The intelligent bearing is intended to provide the user with information whether the roller bearing (hereinafter also abbreviated as “bearing”) 20 is rotating or not. Such a bearing 20 comprises an inner ring 21 and an outer ring 22 , at the outside of which a circumferential longitudinal groove 23 is provided. Between the inner ring 21 and in the outer ring 22 , rolling bodies 24 are arranged so that the inner ring 21 is rotatable with respect to the outer ring 22 . To pick up the data, sensor arrangements 26 —also called “sensors” 26 in brief—are used which, in the preferred exemplary embodiment, are in each case four strain gauges 31 to 34 , combined to form a Wheatstone bridge circuit, which are accommodated in an external longitudinal groove 23 in the outer ring 22 and the resistance of which changes by the rolling bodies 24 rolling over. Furthermore, a circuit board 28 arranged in the groove 23 is shown schematically which establishes the conductor connections between the individual strain gauges 31 to 34 of each sensor 26 and the conductor connections between the sensors 26 and the evaluating devices 50 described later. The direction of movement of the rolling bodies 24 is designated by an arrow A in FIG. 2 .

The groove 23 , and thus the circuit board 28 , is located in the entire circumference of the outer ring 22 with sensors 26 arranged at equidistant intervals (strain gauges 31 to 34 ) and a corresponding evaluating device 50 , comprising, for example, two evaluating units 50 and 51 , for each sensor as will be explained in greater detail later. The resultant sensor signal 40 is intended to be evaluated in a suitable manner by the evaluating device 50 , preferably electrical circuits in the form of application specific integrated circuits, so-called ASICs. In this arrangement, the method according to the invention running in the ASICs for detecting a stationary state must be dimensioned in such a manner that a consistent on-line evaluation of the data is possible in spite of the restricted construction space and thus the restricted chip size. The complete unit of sensor and intelligent evaluating hardware thus represents a so-called “smart sensor” which makes it possible to supply a possible stationary state of the bearing 20 to the potential customer in real time.

For the embodiment of the method according to the invention, the geometric arrangement of the sensors 26 in the outer ring 22 must be taken into consideration since this determines the variation of the signal 40 delivered by the sensor 26 and thus the procedure for dealing with the sensor signal 40 using signal theory is also predetermined. The four strain gauges (hereinafter also abbreviated by SG) 31 to 34 of a sensor 26 , interconnected to form a Wheatstone bridge, are arranged in the longitudinal direction of the groove 23 in such a manner that their distance from one another corresponds to half the rolling body distance (compare FIG. 2 ). This ensures that rolling bodies 24 always roll over two strain gauges 31 and 33 or 32 and 34 , respectively, of a bridge at the same time. The first strain gauge 31 of the following Wheatstone bridge in the groove 23 is again located at half the rolling body distance from the last strain gauge 34 of the preceding bridge. This arrangement results in z/2 sensors 26 for z rolling bodies in the bearing. As an alternative, fewer or even more than z sensors can be present, for example z/2−1, z2+1, z, z+1 or 2 z sensors.

An example with z sensors, of which only two sensors 29 and 30 are shown, is also shown in FIG. 1 . In this arrangement, four SG (designated by different symbols) in each case form one sensor, that is to say 36 , 37 , 38 and 39 form the sensor 29 and 46 , 47 , 48 and 49 form the sensor 30 . The distances of the individual SGs within a sensor are considerably reduced and, therefore, the distance from one sensor to the next can also be greatly reduced in appropriate manner. This makes it possible to detect any “bucking” of the bearing 20 , that is to say a to and fro movement by relatively small angles of rotation which otherwise would possibly not be achievable with greater sensor distances. The sensors 29 and 30 shown are then followed by other sensors, not shown.

A further possible arrangement of the sensors is obtained by the fact that the eight SGs of two sensors which are located behind one another (considering sensors 29 and 30 here for the illustration), in deviation from the representation in FIG. 1 , are interconnected in such a form, i.e. the sensors are interlocked with one another in the manner of a comb, that the SGs are allocated to the sensors in the longitudinal direction of the groove 23 in the following manner: a first sensor comprises SGs 39 , 37 , 49 and 47 and a second sensor comprises SGs 38 , 36 , 48 and 46 . Other such sensors can be provided but are not shown. As a result, two phase-shifted edges (signals 40 ) are generated in a very small space by rolling over which are especially well suited for the evaluation of to and fro movements.

It is also possible to interconnect only two SGs (for example the first and the second SG), instead of with a third and fourth SG, respectively, with in each case one resistor which is not strain-sensitive to form in each case one Wheatstone half bridge. Instead of the SGs mentioned, other sensors, such as piezoelectric or magnetic sensors, can also be used. This may result in a different ratio of sensors to rolling bodies.

FIG. 2 schematically and by way of example shows for a sensor the geometric distribution of strain gauges and rolling bodies 24 in the groove 23 of the outer ring 22 . When rolling over SG 1 31 and SG 3 33 , the bridge output voltage U A (t) rises with increasing deformation of the strain gauges caused by the roller bearing pressure, the supply voltage U B remaining constant. The cross-connected strain gauges in the bridge circuit are synchronously roiled over or deformed as a result of which the rise in output voltage is further increased. As the process continues, the strain gauges SG 1 31 and SG 3 33 are increasingly relieved until they have reached again their original resistance value and the bridge is almost balanced. When subsequently strain gauges SG 2 32 and SG 4 34 are rolled over, the same variation in the output voltage occurs, only with reversed sign due to the polarity of the bridge circuit so that finally a wave-shaped approximately sinusoidal variation is produced.

In FIGS. 3 a to 3 c , a small section of a typical variation of the signal 40 after the zero transition is shown for representing the different possibilities which is why the signal 40 appears to be linear even though it is actually sinusoidal. For the representation, it is assumed that the first sampling interval J 1 in each case begins exactly at zero transition even though this is not the case, as a rule, since the position of this first sampling interval J 1 depends on the starting time of the method.

According to the first possibility, shown in FIG. 3 a , the intervals J k adjoin one another. That is to say, interval J 1 with N sampling times t 1 to t N adjoins the interval J 2 with N sampling times t N+1 to t 2N , and this, in turn, adjoins the interval J 3 with N sampling times t 2N+1 to t 3N etc. Thus, all sampling times are used for detecting the stationary state.

According to the second possibility, shown in FIG. 3 b , in contrast, the sampling intervals are at a distance from one another. In the first time interval comprising L sampling times, sampling occurs at the N sampling times of the first sampling interval J 1 from to t N , and after that the process waits until sampling time t L until the remaining L-N sampling times have elapsed before the second interval J 2 is reached. It is only then that sampling occurs in this second interval J 2 from sampling time t L+1 again at N successive sampling times up to t L+N and then the process waits until sampling time t 2L . This procedure continues in the following intervals. FIG. 3 b shows that L is greater than N in this example. Such an arrangement of the intervals can be advantageous, for example, with an extremely slow rotation of the roller bearing 20 .

According to the third possibility shown in FIG. 3 c , the sampling intervals are interlocked with one another. In the example shown in FIG. 3 c , this is shown for the case of Lot. This means nothing other than that the second interval 32 already begins with its first sampling time t L+1 at the second sampling time t 2 of the first interval J 1 , that is to say, this sampling time t L+1 is identical with the sampling time t 2 . The second interval J 2 ends with the sampling time t L+N which corresponds to the first sampling time t N+1 after the first interval J 1 . Due to this extreme overlap of the intervals, it may be possible to achieve a very rapid detection of a stationary state.

The said values of N and L can be selected arbitrarily, i.e. adapted to the respective peculiarities of the bearing 20 to be detected and, as a rule, are determined empirically, wherein the interval width can vary from system to system.

In the following, advantageous embodiments of the method according to the invention are described by means of flow charts.

In FIG. 4 , a simple method is shown as first embodiment, in which the sampling intervals adjoin one another. In step S 101 , the parameters N are input as a number of the sampling times per interval and a threshold value ε. After that, the starting values for the run variables n and k are in each case set to 1 in step S 106 . After that, the method is capable of processing the samples x n , x n+1 etc. from step S 108 onward. In step S 110 , the starting value S 1 for the first sum, from which the first average value MW 1 of the samples in the first sampling interval J 1 is later calculated, is set to 0. In step S 112 , the current sample x n is added to this starting value S 1 . Following this, the run variable n is incremented by 1 in step S 114 and in step S 116 a check is made whether the end of the first sampling interval has not yet been reached, in other words, whether n≦N. If the end of the first sampling interval has already been reached or exceeded (‘NO’ branch), the first average value MW 1 is calculated in step S 120 by dividing the sum S 1 of the samples by the number N of samples. Otherwise (“YES” branch), the method goes back to step S 112 and is continued until the end of the first sampling interval J 1 is reached and the first average value MW 1 can be calculated. After that, the run variable k is incremented by 1 in step S 122 and the kth sum S k is set to 0 in step S 124 .

After that, the current sample is added to the current sum S k in step S 126 —similar to step S 112 . After that, the run variable n is incremented by 1 in step S 128 and in step S 130 —analogously to step S 116 —a check is made whether the end of the kth sampling interval has not yet been reached, in other words whether n≦kN. If the end of the kth sampling interval has already been reached (“NO” branch), the kth average value MW k is calculated in step S 132 by dividing the sum S k of the samples of the kth interval J k by the number N of samples of this interval J k . Otherwise (“YES” branch), the method goes back to step S 126 and is continued until the end of the kth sampling interval is reached and the kth average value MW k can be calculated.

Finally, the first average value MW 1 is subtracted from the kth average value MW k in step S 134 and a check is made whether the amount of this difference is greater than the threshold value c initially input. If this is so, which corresponds to the “YES” branch, an output flag, that the roller bearing 20 is rotating, is set in step S 150 . If necessary, this information can then be output and/or displayed on a display 60 . If the threshold vale ε has not been exceeded, the method goes back to step S 122 and is continued at least until the end of the next sampling interval J k+1 is reached and the (k+1)th average value MW k+1 can be calculated. To ensure continuous monitoring of the roller bearing 20 for a possible stationary state, it can be provided that, after step S 150 , the method goes back to step S 122 and is continued.

A further advantageous, second embodiment of the method according to the invention is shown in FIG. 5 . In this case, in step S 102 , it is required, in addition to inputting the step S 101 , to input the parameter L which fixes the position of the intervals relative to one another. The further method initially proceeds according to FIG. 4 as up to and including step S 124 . In step S 225 , n is then set to the value of (k−1)L+1 in order to jump over the L-N samples since passing through step S 116 . After that, the known summation is again performed in step S 126 and in step S 128 , the run variable n is incremented by 1 after which a check is made in step S 230 whether the end of the kth sampling interval has not yet been reached, in other words whether n≦(k−1)L+N here. If the end of the kth sampling interval has already been reached, the kth average value MW k is again calculated in step S 132 by dividing the sum S k of the samples of the kth interval J k by the number N of the samples of this interval J k . Otherwise, the method goes back to step S 126 and is continued at least until the end of the kth sampling interval is reached and the kth average value MW k can be calculated. The further course of the method from step S 134 onward takes place as in the first embodiment.

A further third embodiment of the method according to the invention will now be described with reference to FIG. 6 . Instead of the aforementioned threshold value ε, a threshold value δ is now input in addition to the parameters L and N in step S 103 . After that, the method proceeds as in FIG. 5 up to the calculation of the kth average value MW k in step S 132 . After step S 132 , the gradient grad k between the average value MW k and MW 1 is calculated in step S 333 by subtracting the first average values MW 1 from the kth average value MW k and dividing the difference by the distance (k−1)L of the two average values Subsequently, a check is made in step S 335 whether the value of the gradient grad k is greater than the threshold value δ initially input. If this is so, it is assumed that the roller bearing 20 is rotating and in step S 150 a corresponding output flag is set. Otherwise, the method goes back to step S 126 in order to determine the gradient grad k+1 in the (k+1)th interval, and to compare its value with the threshold value δ in step S 335 . To ensure here, too, that the roller bearing 20 is continuously monitored for a possible stationary state, it can also be provided that, after step S 150 , the method goes back to step S 122 and is continued.

FIG. 7 shows a fourth embodiment of the method according to the invention. In step S 104 , a threshold value E and a threshold value δ are input apart from the parameters L and N. After that, the method proceeds as in FIGS. 5 and 6 up to the calculation of the kth average value MW k in step S 132 . After step S 132 , two branches are carried out in parallel. On the one hand, the value of the difference of the average values MW 1 and MW k is compared with the threshold value ε in step S 134 as in FIGS. 4 and 5 and from the result, a rotation of the roller bearing 20 is assumed, if applicable and, on the other hand, the value of the gradient grad k determined in step S 333 is compared with the threshold value δ in step S 335 and from the result a rotation of the roller bearing 20 is assumed, if applicable. In other words, this means that in step S 150 , an output flag for the rotation of the roller bearing 20 is set when one of the two conditions of steps S 134 and S 335 is met which corresponds to the OR operation on these. The method can thus take into consideration two different criteria for detecting a rotation of the roller bearing 20 . If none of the two conditions of steps S 134 and S 135 is met, the method is continued in every case from step S 122 onward. Here, too, the method can also be continued, if desired, when a rotation of the roller bearing 20 is found in step S 150 .

In FIG. 8 , which shows a fifth embodiment of the method according to the invention, a cumulative consideration of the two conditions of steps S 134 and S 335 —corresponding to an AND operation—as prerequisite for the roller bearing 20 to be considered as rotating, is shown. From step S 104 to step S 132 , the method proceeds as in the fourth embodiment.

After step S 132 , the gradient grad k is calculated in step S 333 as in the fourth embodiment after which the amount of the difference of the average values is compared with the threshold value ε in step S 134 , whereupon the value of the gradient grad k is compared with the threshold value δ in step S 335 . It is only when both comparisons of steps S 134 and S 335 have led to the “YES” result that the output flag for the rotation of the roller bearing 20 can be set in step S 150 . Naturally, as an alternative, step S 333 can also be carried out only after step S 134 or the order of the steps S 335 and S 134 can be mutually exchanged.

The aforementioned threshold values E and 6 are usually force-dependent, that is to say, they depend on the force with which the rolling bodies 24 act on the sensor arrangements 26 and 29 , respectively. These threshold values ε and δ can be selected either according to theoretical information or can be empirically determined.

In an exemplary application, a signal having a typical frequency of 1 Hz is sampled with a frequency of 80 kHz. Accordingly, the times t m of the sampling are apart by 12.5 μs. A preferred width of the time intervals J k with the samples x m (x 1 , . . . , x 2000 ) is obtained for the case of precisely adjoining intervals as N=2000 values.

Although this has not been explicitly mentioned in the aforementioned five embodiments of the method according to the invention, sums or intermediate sums S k which are not (no longer) needed after the calculation of the associated average value MW k , and average values MW k (except the first average value MW 1 ) and/or gradients grad k which are not (no longer) needed after comparison with the corresponding threshold values ε and δ, respectively, may be deleted without storage, if necessary.

In the preceding text, it was assumed that an identical number of samples N is allocated to each sampling interval. In individual cases, however, it may be appropriate to vary the number of sampling points per sampling interval in the course of the process by intervention from outside or to allow it to be changed or adapted by the method itself. Furthermore, the calculation of the respective average values and gradients is not restricted to the advantageous method described but can also take place in another suitable manner.

It must be noted that the features of the invention described with reference to individual embodiments such as, for example, individual steps from the flow charts or constructional peculiarities can also be present in other embodiments unless otherwise specified or inherently barred for technical reasons.