Diagnostic method of a rotational speed and load desensitization filter for rolling bearing diagnosis

Abstract

The present invention relates to the technical field of bearing fault diagnosis, and specifically relates to a diagnostic method for a rotational speed and load desensitization filter for rolling bearing diagnosis, including obtaining the rated power of the motor, bearing structure parameters, bearing rotational speed frequency, time-domain signal, and sampling frequency of a bearing public dataset; determining the characteristic frequency of bearing faults according to the bearing structure parameters and the bearing rotational speed frequency; making an amplitude spectrum based on the time-domain signal and the sampling frequency; designing a filter based on the characteristic frequency to obtain a desensitization filter; multiplying the amplitude spectrum by the desensitization filter to obtain a filtered amplitude spectrum; using the filtered amplitude spectrum to obtain training samples for modeling and then diagnosing measured samples to obtain a diagnostic result, solving the problem that the training samples and test samples have large differences in feature distribution, reducing the diagnostic effect of the model obtained by training.

Description

Technical Field

[0001] The present invention relates to the technical field of bearing fault diagnosis, and particularly to a diagnosis method for a speed and load desensitization filter for rolling bearing diagnosis. Background Art

[0002] In current rolling fault diagnosis research, public rolling bearing data sets are often used in modeling.

[0003] However, in actual diagnosis, the samples are affected by speed and load, resulting in a large difference in feature distribution between the training samples and the test samples. The diagnostic ability of the model obtained from the training samples is greatly affected, and the diagnostic effect is poor when using the above model for diagnosis. Summary of the Invention

[0004] The purpose of the present invention is to provide a diagnosis method for a speed and load desensitization filter for rolling bearing diagnosis, aiming to solve the problem that the difference in feature distribution between the training samples and the test samples is large, reducing the diagnostic effect of the trained model.

[0005] To achieve the above purpose, the present invention provides a diagnosis method for a speed and load desensitization filter for rolling bearing diagnosis, including the following steps:

[0006] Obtain the rated power of the motor, bearing structure parameters, bearing speed frequency, time-domain signal, and sampling frequency of the bearing public data set;

[0007] Determine the characteristic frequencies of bearing faults according to the bearing structure parameters and bearing speed frequency;

[0008] Based on the time-domain signal and sampling frequency, make an amplitude spectrum;

[0009] Design a filter based on the characteristic frequencies to obtain a desensitization filter;

[0010] Multiply the amplitude spectrum by the desensitization filter to obtain a filtered amplitude spectrum;

[0011] Use the filtered amplitude spectrum to obtain training samples for modeling and then diagnose the measured samples to obtain a diagnostic result.

[0012] Among them, the characteristic frequencies include inner race fault frequency, outer race fault frequency, rolling element fault frequency, and cage fault frequency.

[0013] Among them, the specific method of making an amplitude spectrum based on the time-domain signal and sampling frequency is:

[0014] Sample the time-domain signal based on the sampling frequency to obtain the number of sampling points;

[0015] After removing the direct current from the number of sampling points and adding a Hanning window, an amplitude spectrum is obtained.

[0016] Among them, the specific method of designing a filter based on the characteristic frequency to obtain a desensitized filter is as follows:

[0017] Determine the center frequency and characteristic frequency band of the filter based on the characteristic frequency;

[0018] Determine the amplitude of the center frequency to obtain a desensitized filter.

[0019] Among them, the specific method of determining the amplitude of the center frequency to obtain a desensitized filter is as follows:

[0020] After removing the direct current from the time-domain signal and adding a Hanning window, a power spectrum is obtained;

[0021] Select the amplitude of the first harmonic of the power spectrum;

[0022] Calculate the amplitude of the characteristic frequency based on the amplitude of the first harmonic to obtain a characteristic amplitude;

[0023] Calculate the amplitude of the first harmonic and the characteristic amplitude according to the time-domain signals at the same point under the powers of 1HP, 2HP, and 3HP in the rated power of the motor to obtain a desensitized filter.

[0024] A diagnostic method for a speed and load desensitized filter for rolling bearing diagnosis according to the present invention includes obtaining the rated power of the motor, bearing structure parameters, bearing speed frequency, time-domain signal, and sampling frequency of the bearing public dataset; determining the characteristic frequency of the bearing fault according to the bearing structure parameters and bearing speed frequency; making an amplitude spectrum based on the time-domain signal and sampling frequency; designing a filter based on the characteristic frequency to obtain a desensitized filter; multiplying the amplitude spectrum by the desensitized filter to obtain a filtered amplitude spectrum; using the filtered amplitude spectrum to obtain training samples for modeling and then diagnosing the measured samples to obtain a diagnostic result. The present invention can effectively remove the sensitivity of the background signal and noise in the public dataset to the samples, reduce the correlation between the training samples and the measured samples, so as to achieve the ability to construct an effective model using the public dataset for actual accurate diagnosis, solve the problem that the training samples and the test samples have large differences in feature distribution, and reduce the diagnostic effect of the model obtained by training. Description of the Drawings

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1 It is a flowchart of a diagnostic method for a speed and load desensitization filter used in rolling bearing diagnosis provided by the present invention.

[0027] Figure 2 It is a flowchart for making an amplitude spectrum based on the time-domain signal and the sampling frequency.

[0028] Figure 3 It is a flowchart for designing a filter based on the characteristic frequencies to obtain a desensitization filter.

[0029] Figure 4 It is a schematic diagram of the filter. Detailed implementation manners

[0030] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0031] Please refer to Figures 1 to 4 , the present invention provides a diagnostic method for a speed and load desensitization filter used in rolling bearing diagnosis, including the following steps:

[0032] S1 Obtain the rated power of the motor, bearing structure parameters, bearing rotational speed frequency n (also known as the fundamental frequency 1X), time-domain signal, and sampling frequency f of the bearing public dataset s ;

[0033] S2 Determine the characteristic frequencies of bearing faults according to the bearing structure parameters and the bearing rotational speed frequency;

[0034] Specifically, the characteristic frequencies include the inner race fault frequency f I , the outer race fault frequency f O , the rolling element fault frequency f B , and the cage fault frequency f T .

[0035] Inner race fault frequency:

[0036] f I = n × f BPI (1)

[0037] Outer race fault frequency:

[0038] f O = n × f BPO (2)

[0039] Rolling element fault frequency:

[0040] fB = n × f BS (3)

[0041] Cage fault frequency:

[0042] f T = n × f FT (4)

[0043] where f BPI , f BPO , f BS , f FT are the inner race, outer race, ball spin, and cage frequencies of the bearing respectively, and can be calculated according to the bearing model using relevant formulas.

[0044] S3 generates an amplitude spectrum based on the time-domain signal and the sampling frequency;

[0045] The specific method is as follows:

[0046] S31 samples the time-domain signal based on the sampling frequency to obtain the number of sampling points;

[0047] Specifically, the number of sampling points is 2048 points.

[0048] S32 removes the DC component from the number of sampling points and applies a Hanning window to obtain the amplitude spectrum.

[0049] Specifically, a windowing judgment is performed on the sampled data after removing the DC component. If the sampled data after removing the DC component does not require windowing, the amplitude spectrum is obtained through the sampled data after removing the DC component. If the sampled data after removing the DC component requires windowing, a Hanning window is applied to the sampled data after removing the DC component to obtain the amplitude spectrum.

[0050] S4 designs a filter based on the characteristic frequency to obtain a desensitized filter;

[0051] The specific method is as follows:

[0052] S41 determines the center frequency and characteristic frequency band of the filter based on the characteristic frequency;

[0053] Specifically, four center frequencies are calculated according to the bearing model and the bearing rotational speed frequency n in the bearing structure parameters using formulas (1)-(4) respectively.

[0054] Considering factors such as frequency ambiguity, the fault characteristic frequency harmonic is designed to be 0.8 times the cage characteristic frequency, i.e., 0.8f T :

[0055] f 11 = f T - 0.8f T , f 12 = f T+0.8f T ; (5)

[0056] f 21 = f O -0.8f T , f 22 = f O +0.8f T ; (6)

[0057] f 31 = f B -0.8f T , f 32 = f B +0.8f T ; (7)

[0058] f 41 = f I -0.8f T , f 42 = f I +0.8f T ; (8)

[0059] If the above frequency bands cross, take the upper and lower limits of each frequency band at the crossing point.

[0060] S42 determines the amplitude of the center frequency to obtain a desensitization filter.

[0061] The specific method is as follows:

[0062] S421 performs DC removal on the time-domain signal and then adds a Hanning window to obtain a power spectrum;

[0063] Specifically, for the time-domain signals of 4 single faults respectively, after DC removal, a Hanning window is selected to calculate the power spectrum of the time-domain signal;

[0064] S422 selects the amplitude of the fundamental frequency of the power spectrum;

[0065] Specifically, the amplitude of its fundamental frequency of 1X is H 1X .

[0066] S423 calculates the amplitude of the characteristic frequency based on the amplitude of the fundamental frequency to obtain a characteristic amplitude;

[0067] Specifically, calculate f according to formulas (1)-(4) and 1X T , f O , f B , f I of the amplitude H T , H O , H B , H I .

[0068] S424 calculates the fundamental frequency amplitude and characteristic amplitude according to the time-domain signals at the same point under the rated powers of 1HP, 2HP, and 3HP of the motor, and obtains a desensitization filter.

[0069] Specifically, based on the time-domain signals at the same point under the powers of 1HP, 2HP, and 3HP provided by the data set, the above five amplitudes are calculated respectively, which are:

[0070] H 1X 1 ,H T 1 ,H O 1 ,H B 1 ,H I 1 ;H 1X 2 ,H T 2 ,H O 2 ,H B 2 ,H I 2 ;H 1X 3 ,H T 3 ,H O 3 ,H B 3 ,H I 3 ;

[0071] For each power, calculate the ratio of H T ,H O ,H B ,H I to H 1X , and then average it three times to obtain the average value of the ratio, which is the coefficient of the four characteristic frequencies of the filter:

[0072]

[0073] S5 multiplies the amplitude spectrum by the desensitization filter to obtain a filtered amplitude spectrum;

[0074] The amplitude parameter of the final desensitization filter is:

[0075] h T =c t h 1x (9)

[0076] h O =c oh 1x (10)

[0077] h B = c B h 1x (11)

[0078] h I = c I h 1x (12)

[0079] From formulas (9)-(12), the dynamic desensitization filter for each sample can be determined. The amplitude-frequency parameters and frequency parameters of this filter are both functions of the rotational speed of the bearing. After filtering the samples, modeling can be carried out and then diagnostic work can be performed on the measured samples.

[0080] S6 Use the filtered amplitude spectrum to obtain training samples for modeling and then diagnose the measured samples to obtain a diagnostic result.

[0081] Specifically, the filtered amplitude spectrum can be used to obtain frequency-domain characteristic parameters, or perform an inverse Fourier transform to obtain time-domain characteristic parameters, or use time-frequency signals to obtain characteristic parameters for modeling and then diagnose the measured samples.

[0082] The specific method is as follows:

[0083] S61 Construct a training model;

[0084] S62 Use the training samples to train and test the training model to obtain a diagnostic model;

[0085] Specifically, the training samples are the desensitized training set and test set.

[0086] S62 Use the diagnostic model to diagnose the measured samples to obtain a diagnostic result.

[0087] Specifically, the rotational speed and load of the samples are effectively desensitized, so that when modeling and testing, the frequencies that are sensitive to non-rolling bearing faults are attenuated.

[0088] A diagnostic method for a speed and load desensitization filter for rolling bearing diagnosis according to the present invention includes obtaining the rated power of the motor, bearing structure parameters, bearing speed frequency, time-domain signal, and sampling frequency of a bearing public dataset; determining the characteristic frequency of bearing faults according to the bearing structure parameters and bearing speed frequency; making an amplitude spectrum based on the time-domain signal and sampling frequency; designing a filter based on the characteristic frequency to obtain a desensitization filter; multiplying the amplitude spectrum by the desensitization filter to obtain a filtered amplitude spectrum; using the filtered amplitude spectrum to obtain training samples for modeling and then diagnosing measured samples to obtain a diagnostic result. The present invention can effectively remove the sensitivity of background signals and noise in the public dataset to samples, reduce the correlation between training samples and measured samples, so as to achieve the ability to construct an effective model using the public dataset for actual and accurate diagnosis, solve the problem that there is a large difference in feature distribution between training samples and test samples, and reduce the diagnostic effect of the model obtained by training.

[0089] The above-disclosed is only a preferred embodiment of a diagnostic method for a speed and load desensitization filter for rolling bearing diagnosis according to the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the invention.

Claims

1. A diagnostic method for a speed and load desensitization filter used in rolling bearing diagnosis, characterized in that, Including the following steps: Obtain the motor rated power, bearing structure parameters, bearing rotational speed frequency, time-domain signal, and sampling frequency of the bearing public dataset; Determine the characteristic frequency of the bearing fault according to the bearing structure parameters and the bearing rotational speed frequency; Generate an amplitude spectrum based on the time-domain signal and the sampling frequency; Design a filter based on the characteristic frequency to obtain a desensitized filter; Multiply the amplitude spectrum by the desensitized filter to obtain a filtered amplitude spectrum; Use the filtered amplitude spectrum to obtain training samples for modeling and then diagnose the measured samples to obtain a diagnosis result; The specific method for designing a filter based on the characteristic frequency to obtain a desensitized filter is as follows: Determine the center frequency and characteristic frequency band of the filter based on the characteristic frequency; Determine the amplitude of the center frequency to obtain a desensitized filter; The specific method for determining the amplitude of the center frequency to obtain a desensitized filter is as follows: Remove the DC component from the time-domain signal and then apply a Hanning window to obtain a power spectrum; Select the amplitude of the fundamental frequency of the power spectrum; Calculate the amplitude of the characteristic frequency based on the amplitude of the fundamental frequency to obtain a characteristic amplitude; Calculate the amplitude of the fundamental frequency and the characteristic amplitude according to the time-domain signals at the same point under the powers of 1HP, 2HP, and 3HP in the motor rated power to obtain a desensitized filter, specifically: According to the time-domain signals at the same point under the powers of 1HP, 2HP, and 3HP provided by the dataset, calculate the amplitude of the fundamental frequency and the amplitudes of the four characteristic frequencies respectively, which are: H 1X 1 ,H T 1 ,H O 1 ,H B 1 ,H I 1 ;H 1X 2 ,H T 2 ,H O 2 ,H B 2 ,H I 2 ;H 1X 3 ,H T 3 ,H O 3 ,H B 3 ,H I 3 ; Among them, the four characteristic frequencies include the inner race fault frequency, the outer race fault frequency, the rolling element fault frequency, and the cage fault frequency; Calculate H for each power T ,H O ,H B ,H I The ratio with H 1X is then averaged three times to obtain the average value of the ratio, which is the coefficient of the four characteristic frequencies of the filter:

2. The diagnostic method of the rotational speed and load desensitized filter for rolling bearing diagnosis according to claim 1, wherein The specific method for generating an amplitude spectrum based on the time-domain signal and the sampling frequency is as follows: Sample the time-domain signal based on the sampling frequency to obtain the number of sampling points; Remove the DC component from the sampled data and then apply a Hanning window to obtain an amplitude spectrum.

Images