Device and method for processing a digital signal

Abstract

A device is proposed for processing a continuous signal sampled at a fixed sampling rate to obtain a digital signal, wherein the continuous signal comprises state monitoring data of a rotating element, the rotating element being subject to changes in rotational speed. The device includes: - Segmentation means (15), - Spectral analysis agents (16), - Filter media (17), - first processing equipment (18), - a storage (19), - second processing equipment (20) and - Scanning device (21).

Description

[0001] The present invention relates to devices for processing a digital signal and methods for processing a digital signal.

[0002] Condition monitoring algorithms, for example for detecting a bearing fault, such as a fault in the inner raceway of the bearing, require a constant rotational speed of the bearing to function properly.

[0003] In general, condition monitoring algorithms perform a spectral analysis of the rotating bearing to detect faults based on specific tones and harmonics.

[0004] Rotational speed changes of the bearing during signal measurement blur the spectral tones, reducing the ability to identify the frequencies and harmonics that may be associated with faults, such as faults in an inner raceway of the bearing or faults in the outer raceway of the bearing.

[0005] Without access to a speedometer or other direct speed measuring devices, the difficulties in correcting speed fluctuations can be further exacerbated by complex speed fluctuation profiles, such as profiles with multiple acceleration and deceleration phases and time-varying changes in tone amplitude.

[0006] Consequently, the present invention is intended to correct harmful effects of changes in rotational speed.

[0007] According to one aspect, a method for processing a continuous signal sampled at a fixed sampling rate to obtain a digital signal, wherein the continuous signal comprises state monitoring data of a rotating element, the rotating element being subject to rotational speed changes.

[0008] The procedure includes: - (a) Segmenting the time-stamped samples of the digital signal into multiple frames, - (b) Transforming each of the multiple frames into the frequency domain and providing arrays containing absolute values ​​of the order of magnitude and frequency of each transformed frame, - (c) for each transformed frame, determine significant spectral peaks from background noise and their peak mid-frequencies, - for each pair of frames of a set of frames of the multiple frames: (d) Determining all permissible frequency ratios of each significant spectral peak of a first frame of the pair with respect to each significant spectral peak of a second frame of the pair, (e) Clustering the frequency ratios into clusters with similar frequency ratios, (f) Determining the cluster that has the largest number of frequency ratios and a rate-change coefficient from the frequency ratios of the cluster that has the largest number of frequency ratios, (g) Assigning a timestamp value to the velocity change coefficient determined in step (f), which is determined according to at least one timestamp value of the first or second frame of the pair, - (h) Storing the rate-of-change coefficients and their timestamp values ​​in an array, - (i) Generating a velocity profile from the velocity change coefficients and their timestamp values ​​stored in the array, and - (j) re-sampling the digital signal in the radian range according to the velocity profile.

[0009] The method allows for the condition assessment of systems and / or sensors that do not have the possibility of direct speed measurement.

[0010] The method allows the use of any existing algorithm designed for constant speed conditions that is industry-accepted, reliable, and well-understood.

[0011] The process further avoids repeated data collection attempts that would have to be discarded simply due to changes in speed.

[0012] Instead, such data can now be used effectively. Suppressing repeated attempts to collect data allows for the conservation of power consumption in a system implementing the method, for example, a wireless system that includes a power source such as a battery.

[0013] Advantageously, step (i) includes: - for each timestamp value, select a representative rate of change coefficient from the rate of change coefficients stored in the array, - Multiplying each representative velocity change coefficient by the average velocity of the rotating element to obtain velocity values, - Obtaining arrays of representative rate-of-change coefficients as a function of their timestamp values, and - Applying interpolation to the selected velocity change coefficients to obtain the velocity profile.

[0014] Preferably, selecting the representative rate of change coefficient for each timestamp value includes: - Clustering the rate-of-change coefficients associated with the timestamp value, and - Applying a statistical method to clusters to determine the representative rate of change coefficient.

[0015] Advantageously, the method involves applying a Hanning window to each of the multiple frames before transforming each of the multiple frames into the frequency domain.

[0016] Preferably, the method includes padding each frame of the multiple frames with zeros before transforming each frame of the multiple frames into the frequency domain.

[0017] Advantageously, step (d) further includes the deletion of any frequency ratio that is not contained within a predetermined interval containing permissible frequency ratios.

[0018] According to another aspect, a device for processing a continuous signal is proposed, which is sampled at a fixed sampling rate to obtain a digital signal, wherein the continuous signal comprises state monitoring data of a rotating element, the rotating element being subject to changes in rotational speed.

[0019] The device includes: - Segmentation means designed to segment the time-stamped samples of the digital signal into multiple frames, - Spectral analysis instruments trained to transform each of the multiple frames into the frequency domain and to provide arrays that include absolute values ​​of the order of magnitude and frequency of each transformed frame, - Filtering means designed to identify significant spectral peaks from background noise and their peak mid-frequencies for each transformed frame, - first processing means that are trained to perform the following for each pair of frames of a set of frames of several frames: • to determine all permissible frequency ratios of each significant spectral peak of a first frame of the pair in relation to each significant spectral peak of a second frame of the pair, • to cluster the frequency ratios into clusters with similar frequency ratios, • to determine the cluster with the largest number of frequency ratios and a rate-change coefficient from the frequency ratios of the cluster with the largest number of frequency ratios, • to assign a timestamp value to the velocity change coefficient determined in step (f), wherein the timestamp value to the velocity change coefficient is determined according to at least one timestamp value of the first or second frame of the pair, - a memory designed to store the rate-of-change coefficients and their timestamp values ​​in an array, - second processing means trained to generate a velocity profile from the velocity change coefficients stored in the array and their timestamp values, and - Sampling means designed to resample the digital signal in the radian range according to the velocity profile.

[0020] Preferably, the first processing means are further developed to apply a Hanning window to each frame of the multiple frames before transforming each frame of the multiple frames into the frequency domain.

[0021] Advantageously, the first processing means are further developed to pad each of the multiple frames with zeros before transforming each of the multiple frames into the frequency domain.

[0022] A storage device is proposed based on one aspect.

[0023] The storage device includes: - a bearing which is provided with an inner ring and an outer ring which are suitable for concentric rotation relative to each other, - a sensor designed to measure the vibrations of the inner ring or the outer ring, and designed to provide a continuous signal, - a sampler configured to sample the continuous signal at a fixed sampling rate, and configured to deliver the digital signal comprising the successive samples, and - a device, as defined above, designed to process the digital signal.

[0024] Other advantages and features of the invention become clear upon examination of the detailed description of the embodiments, which are in no way limiting, and the accompanying drawings, in which: Fig. 1 schematically represents a machine according to the invention, Fig. 2 schematically represents an example of a device for processing a digital signal according to the invention, Fig. Figure 3 schematically illustrates a method for processing a digital signal according to the invention, Fig. Figure 4 schematically represents an example of clusters of rate-change coefficients according to the invention. Fig. Figure 5 schematically illustrates an example of the speed profile according to the invention, and Fig. Six schematic examples of a spectrum supplied by spectral analysis agents for a bearing that is subject to velocity changes during measurement, according to the prior art and according to the invention.

[0025] It will be on Fig. 1 is referenced, which schematically represents a partial longitudinal section of machine 1.

[0026] The machine 1 comprises a housing 2 and a shaft 3, which is supported in the housing 2 by a rolling bearing 4 (e.g. roller bearing or ball bearing).

[0027] The rolling bearing 4 is provided with an inner ring 5 fixed to the shaft 3 and with an outer ring 6 fixed in the bore of the housing 2. The outer ring 6 surrounds the inner ring 5 radially. The inner and outer rings 5, 6 rotate concentrically relative to each other.

[0028] The rolling bearing 4 is further provided with a series of rolling elements 7, which are inserted radially between the inner and outer raceways of the inner and outer rings 5, 6. In the example shown, the rolling elements 7 are balls. Alternatively, the rolling bearing can include other types of rolling elements 7, for example, rollers. In the example shown, the rolling bearing includes a series of rolling elements 7.

[0029] Alternatively, the rolling bearing can comprise several rows of rolling elements.

[0030] A sensor 8 is mounted in the housing 2 to measure vibrations of the bearing 4, which is subject to changes in rotational speed.

[0031] The sensor 8 can be attached to a hole in the housing 2.

[0032] In one variant, the sensor 8 can be attached to a different location on the machine, for example near the outer ring 6 or in the vicinity of the housing 2.

[0033] Sensor 8 provides a continuous signal S8, which represents the operation of a rotating element.

[0034] The rotating element can be the bearing 4 and the sensor 8 provides the continuous signal S8, which represents the vibration of the bearing 4, to an input of a probe 9.

[0035] The probe 9 delivers a digital signal S9, which contains time-stamped sample values ​​x p of the continuous signal S8, which is sampled at a fixed sampling rate, to an input 101 of a device 10 for processing the digital signal S9, where p is an integer.

[0036] The bearing 4, the sensor 8, the probe 9 and the device 10 form a bearing device.

[0037] A memory (not shown) can store the output signal S9 and delivers the output signal S9 to the device 10.

[0038] An output 102 of the device 10 can be connected to implementation means 11 which implement at least one constant-speed time domain algorithm from a first output signal S102 supplied by the device 10 to the first output 102, for example to implement an envelope fault detection algorithm.

[0039] The output 102 of the device 10 can be further connected to second implementation means 12 to perform a spectral analysis of the output of the device 10.

[0040] The second implementation means 12 implement, for example, a fast Fourier transform.

[0041] The first and second implementation means 11, 12, for example, are each made from a processing unit that implements the algorithm.

[0042] A processing unit 13 implements the sensor 8, the probe 9 and the device 10.

[0043] Fig. Figure 2 shows a schematic example of device 10.

[0044] The device 10 comprises a first storage device 14, segmentation means 15, spectral analysis means 16, filter means 17, first processing means 18, a second storage device 19, second processing means 20 and scanning means 21.

[0045] The first memory 14 is designed to store the digital signal S9, which contains the time-stamped sample values ​​x p includes those received at input 101 of device 10.

[0046] The first memory 14 is connected to an input of the segmentation means 15.

[0047] The segmentation means 15 further include a first output which is connected to an input of the spectral analysis means 16.

[0048] The segmentation means 15 are designed to measure the successive samples x p of the digital signal into multiple frames F1, F2... F k-1 to segment, where k is an integer. The frames have identical sizes.

[0049] The segmentation means 15 can further be used to define frames F1, F2... F k-1 , F k to pad with zeros and a Hanning window on each frame F1, F2... F k-1 , F k to apply to multiple frames.

[0050] The successive samples x p are grouped in the p frames.

[0051] The integer p is determined according to the resolution expected to capture changes in the machine's speed in each frame.

[0052] The segmentation tools 15 are intended to divide frames F1, F2... F k-1 , F k to supply the input of the spectral analysis medium 16.

[0053] An output of the spectral analysis agent 16 is connected to an input of the filter agent 17.

[0054] The spectral analysis agents 16 are intended to perform a spectral analysis of frames F1, F2... F k-1 , F k to perform, in order to provide arrays that contain the absolute values ​​of the frequencies and orders of magnitude of each frame F1, F2... F k-1 , F k include.

[0055] Padding the frames F1, F2... F with zeros k-1 , F k and / or the application of the Hanning window improve the spectral resolution.

[0056] The spectral analysis instruments 16, for example, implement a fast Fourier transformation algorithm.

[0057] An output of the filter medium 17 is connected to inputs of the first processing medium 18.

[0058] The filter elements 17 are designed to remove significant spectral peaks from the background noise and their peak mid-frequencies of the transformation frames F1, F2... F k-1 , F k to determine from the arrays the absolute values ​​of frequencies and orders of magnitude of frames F1, F2... F k-1 , F k include.

[0059] The filter means 17 implement, for example, a noise carpet filter to suppress noise from the frequency spectrum of frames F1, F2... F k-1 , F k The filter is used to retain only those components that can be, for example, +10 dB above the local spectral carpet level. The significant spectral peaks are components that can be +10 dB above the local spectral carpet level.

[0060] The first processing means 18 include a memory 22 designed to store arrays containing the peak mean frequencies of each significant spectral peak of frames F1, F2... F k-1 , F k along the number of frames.

[0061] The first processing agents 18 further include first determination agents 23, cluster agents 24 and allocation agents 25.

[0062] An output from the first processing unit 18 is connected to a second storage unit 19.

[0063] An input of the second processing means 20 is connected to the second memory 19 and an output of the second processing means 20 is connected to an input of the sampling means 21.

[0064] An output of the scanning means 21 is connected to the output 102 of the device 10.

[0065] Fig. Figure 3 represents an example of a method for processing the digital signal S9, which is implemented by the device 10.

[0066] In step 30, the sampler 9 delivers the digital signal S9, which contains the time-stamped sample values ​​x p includes a continuous signal S8 supplied by sensor 8.

[0067] The time-stamped sample values ​​x p They are stored in memory 14.

[0068] In step 31, the segmentation means 15 determine the frames F1, F2, ..., F k-1 , F k .

[0069] In step 32, the spectral analysis tools 16 transform each frame of the multiple frames F1, F2, ..., F k-1 , F k into the frequency domain and provide the arrays that include absolute values ​​of frequencies and orders of magnitude of each transformed frame.

[0070] In step 33, the filter means 17 determine the significant spectral peaks and their peak mid-frequencies of the transformed frames F1, F2... F k-1 , F k from the arrays, the absolute values ​​of frequencies and orders of magnitude of frames F1, F2... F k-1 , F k include, and provide arrays that contain the peak mid-frequencies of each significant spectral peak of frames F1, F2... F k-1 , F k along the number of frames.

[0071] The arrays supplied by the filter means 17 are stored in the memory 22 of the first processing means 18.

[0072] In step 34, a set of frames consisting of the multiple frames F1, F2...F k-1 , F k defined. The sentence can be part of several frames F1, F2... F k-1 , F k or all of the frames of the multiple frames F1, F2... F k-1 , F k include.

[0073] In step 35, pairs of frames from the set of frames are determined.

[0074] Each pair of frames includes a first frame F i and a second frame F j , where i and j are distinct integers.

[0075] For each pair of frames F k , F j The first determination methods 23 determine in one step 36 all permissible frequency ratios of each significant spectral peak of the first frame F i of the pair with reference to each significant spectral peak of the second frame F j of the couple.

[0076] The frequency ratios represent speed fluctuations (accelerations and decelerations) of the rotating element between the first and second frame F i , F j .

[0077] Some frequency ratios that are representative of speed fluctuations may be unrealistic.

[0078] The first determining means 23 eliminate any frequency ratio that is not contained within a predetermined interval containing permissible frequency ratios.

[0079] The cluster means 24 cluster the frequency ratios into clusters with similar frequency ratios.

[0080] Each cluster has the same width, centered on a different frequency value.

[0081] The cluster means 24 determine the cluster with the largest number of frequency ratios.

[0082] The cluster means 24 further determine a rate of change coefficient from the frequency ratios of the cluster that has the largest number of frequency ratios.

[0083] The coefficient of velocity change is determined using various static methods, such as the mean frequency of the cluster that has the largest number of frequency ratios, the physical center, the center of mass, or the weighted mean frequency ratio.

[0084] The allocation means 25 assign a timestamp value to the coefficient of rate of change.

[0085] The assigned timestamp value is determined according to at least one timestamp value from the first or second frame F. i , F j determined by the couple.

[0086] The assigned timestamp value is, for example, equal to the midpoint timestamp of the second frame F. j .

[0087] The rate-of-change coefficients and their timestamp values ​​of the pairs of frames of the set of frames are stored in an array that is stored in the second memory 19.

[0088] In step 37, the second processing means generate a speed profile from the velocity change coefficients stored in the array and their timestamp values.

[0089] For each timestamp value, a representative rate of change coefficient is selected from the rate of change coefficients stored in the array.

[0090] The second processing means 20 can select a first velocity change coefficient β1 stored in the array, representing velocity changes between the first frame F1 and the second frame F2; a second velocity change coefficient β2 stored in the array, representing velocity changes between the second frame F2 and the third frame F3; ...; a k-1 velocity change coefficient βk-1 stored in the array, representing velocity changes between the k-1 frame Fk-1 and the k-Frame F k represented. Two consecutive frames comprise consecutive samples x p .

[0091] The representative velocity change coefficients are the velocity change coefficients β1... βk-1.

[0092] In another embodiment, the second processing means 20 can determine and select the rate change coefficients, which can represent rate fluctuations between two non-consecutive frames, such that, for example, a rate change coefficient β1,3 stored in the array, representing rate fluctuations between the first frame F1 and the third frame F3, ... a rate change coefficient β2,k stored in the array, representing rate fluctuations between frame F2 and frame F k represented.

[0093] In another embodiment, the second processing means 20 determine a matrix of rate-change coefficients by successively comparing all frames with all other frames stored in the second memory 19.

[0094] Several rate-of-change coefficients are associated with each timestamp value.

[0095] The second processing means 20 cluster the rate change coefficients associated with each timestamp value.

[0096] Fig. Figure 4 represents an example of clusters of rate-of-change coefficients.

[0097] Each point represents a coefficient of change of velocity of the matrix.

[0098] The rate-of-change coefficients associated with time t1 are clustered in a cluster C1; analogously, the rate-of-change coefficients associated with time tn are clustered in a cluster Cn, where n is, for example, between 2 and 15.

[0099] Of course, the rate-of-change coefficients can be clustered into more than fifteen clusters or fewer than fifteen clusters.

[0100] For each cluster C1 to C15 (timestamp value), the representative rate of change coefficient is determined by various statistical procedures, such as the mean of the rate of change coefficient values ​​of each cluster, curve fitting of the rate of change coefficient values ​​of each cluster, or other statistical procedures.

[0101] Once the representative velocity change coefficients have been determined, the second processing means 20 multiply each representative velocity change coefficient by the average velocity of the rotating element to obtain velocity values.

[0102] The second processing means 20 determine arrays of representative rate-of-change coefficients based on their timestamp values ​​and apply interpolation to the selected rate-of-change coefficients to obtain the rate profile. The interpolation can be linear, polynomial, or curvilinear, for example, depending on the density of the representative rate-of-change coefficients.

[0103] Fig. Figure 5 provides an example of the speed profile.

[0104] The points represent the representative coefficients of rate of change associated with the timestamp values ​​t1 to t15.

[0105] In step 38, the scanning means 21 sample the digital signal S9 in the radian range again according to the velocity profile determined in step 37.

[0106] Fig. Figure 6 shows an example of the spectrum of oscillations supplied by sensor 8 without velocity compensation (dashed line) and with velocity compensation (solid line), where the digital signal S9 was sampled again by the scanning means 21.

[0107] The tones are easily identifiable in the vibration spectrum with velocity compensation, while the tones are smeared in the vibration spectrum without velocity compensation.

[0108] Device 10 allows for machine condition assessment for systems and / or sensors that do not have the possibility of direct speed measurement.

[0109] The device 10 allows the use of any existing algorithm designed for constant speed conditions that is accepted, reliable and well understood in the industry.

[0110] Device 10 further avoids repeated data acquisition attempts that would have to be discarded solely due to changes in speed.

[0111] Instead, such data can now be used effectively. Suppressing repeated attempts to acquire data allows the power supply of a system comprising the sensor 8, the probe 9, and the device 10, for example, a wireless system with a power source such as a battery, to be optimized.

Claims

[1] Method for processing a continuous signal (S8) sampled at a fixed sampling rate to obtain a digital signal, wherein the continuous signal comprises state monitoring data of a rotating element (4) wherein the rotating element is subject to changes in rotational speed, the method comprising: - (a) Segmenting the time-stamped samples of the digital signal (S9) into multiple frames (F1, F2) k ), - (b) Transforming each frame (F1, F k ) of the multiple frames into the frequency domain and providing arrays that include absolute values ​​of the order of magnitude and frequency of each transformed frame, - (c) for each transformed frame, determine significant spectral peaks from background noise and their peak mid-frequencies, - for each pair of frames of a set of frames of the multiple frames (F1, F k ): (d) Determining all permissible frequency ratios of each significant spectral peak of a first frame of the pair with respect to each significant spectral peak of a second frame of the pair, (e) Clustering the frequency ratios into clusters with similar frequency ratios, (f) Determining the cluster that has the largest number of frequency ratios and a rate-change coefficient from the frequency ratios of the cluster that has the largest number of frequency ratios, (g) Assigning a timestamp value to the velocity change coefficient determined in step (f), wherein the timestamp value to the velocity change coefficient is determined according to at least one timestamp value of the first or second frame of the pair, - (h) Storing the rate-of-change coefficients and their timestamp values ​​in an array, - (i) Generating a velocity profile from the velocity change coefficients and their timestamp values ​​stored in the array, and - (j) re-sampling the digital signal in the radian range according to the velocity profile. [2] Method according to claim 1, wherein step (i) comprises: - for each timestamp value, select a representative rate of change coefficient from the rate of change coefficients stored in the array, - Multiplying each representative velocity change coefficient by the average velocity of the rotating element to obtain velocity values, - Obtaining arrays of representative rate-of-change coefficients as a function of their timestamp values, and - Applying interpolation to the selected velocity change coefficients to obtain the velocity profile. [3] Method according to claim 2, wherein selecting the representative rate of change coefficient for each timestamp value comprises: - Clustering the rate-of-change coefficients associated with the timestamp value, and - Applying a statistical procedure to clusters to determine the representative coefficient of rate of change. [4] Method according to any one of claims 1 to 3, comprising applying a Hanning window to each frame of the multiple frames prior to transforming each frame of the multiple frames into the frequency domain. [5] Method according to any one of claims 1 to 4, comprising padding each of the multiple frames with zeros before transforming each of the multiple frames into the frequency domain. [6] Method according to any one of claims 1 to 5, wherein step (d) further comprises deleting each frequency ratio that is not contained in a predetermined interval that has permissible frequency ratios. [7] Device for processing a continuous signal (S8) sampled at a fixed sampling rate to obtain a digital signal (S9), wherein the continuous signal comprises state monitoring data of a rotating element (4) wherein the rotating element is subject to changes in rotational speed, the device comprising: - Segmentation means (15) configured to divide the time-stamped samples of the digital signal (S9) into multiple frames (F1, F2) k ) to segment, - Spectral analysis instruments (16) trained to analyze each frame (F1, F2) k) to transform the multiple frames into the frequency domain and provide arrays that include absolute values ​​of the order of magnitude and frequency of each transformed frame, - Filter means (17) designed to determine significant spectral peaks from background noise and their peak mid-frequencies for each transformed frame, - first processing means (18) which are trained to perform for each pair of frames of a set of frames of the multiple frames (F1, F2) k ): • to determine all permissible frequency ratios of each significant spectral peak of a first frame of the pair in relation to each significant spectral peak of a second frame of the pair, • to cluster the frequency ratios into clusters with similar frequency ratios, • to determine the cluster with the largest number of frequency ratios and a rate-change coefficient from the frequency ratios of the cluster with the largest number of frequency ratios, • to assign a timestamp value to the velocity change coefficient determined in step (f), wherein the timestamp value to the velocity change coefficient is determined according to at least one timestamp value of the first or second frame of the pair, - a memory (19) configured to store the rate-change coefficients and their timestamp values ​​in an array, - second processing means (20) which are configured to generate a velocity profile from velocity change coefficients and their timestamp values ​​stored in the array, and - Sampling means (21) configured to resample the digital signal in the radian range according to the velocity profile. [8] Device according to claim 7, wherein the first processing means (18) are further developed to apply a Hanning window to each frame of the multiple frames before transforming each frame of the multiple frames into the frequency domain. [9] Device according to claim 7 or 8, wherein the first processing means (18) are further configured to pad each frame of the multiple frames with zeros before transforming each frame of the multiple frames into the frequency domain. [10] Storage device comprising: - a bearing (4) which is provided with an inner ring (5) and an outer ring (6) which are suitable for concentric rotation relative to each other, - a sensor (8) configured to measure the vibrations of the inner ring (5) or the outer ring (6), and configured to provide a continuous signal (S8), - a sampler (9) configured to sample the continuous signal (S8) at a fixed sampling rate, and configured to provide the digital signal (S9) comprising the successive samples, and - a device (10) according to one of claims 7 to 9, which is configured to process the digital signal (S9).

Images