Rolling mill gearbox fault feature extraction method based on adaptive order analysis

Abstract

The application discloses a rolling mill gearbox fault feature extraction method based on adaptive order analysis, wherein according to a gear order bandwidth, a gear vibration component low-pass filter cutoff frequency is calculated, a bearing vibration component band-pass filter center frequency is searched in a high-frequency spectrum peak, and a band-pass filter bandwidth is calculated; a gear vibration component is obtained through low-pass filtering, and a bearing vibration component is obtained through band-pass filtering and envelope demodulation; angular domain resampled signals are averaged, gear sideband amplitude and modulation ratio are taken as features to identify gear faults, and the ratio of a bearing fault feature amplitude to a rotation frequency amplitude is taken as a feature to identify bearing faults; according to the vibration performance of gear and bearing faults, the method designs feature indexes, has clear physical meaning, can eliminate the influence of working condition changes and noise components, and realizes accurate identification of rolling mill gearbox faults.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of equipment fault diagnosis, and particularly relates to a rolling mill gear box fault feature extraction method based on adaptive order analysis. BACKGROUND

[0002] The rolling mill gear box usually works in a variable speed condition due to the rolling process, and the vibration signal thereof is a typical non-stationary signal. The existing vibration signal analysis of the gear box mainly relies on spectrum analysis. Since the speed variation will affect the frequency values of the vibration components in the vibration signal, and the amplitude values of the vibration components will also change accordingly, considering the limitation of the field conditions, the order analysis is the most feasible. Although the order analysis has been theoretically studied, for the actual diagnosis demand of the rolling mill gear box, the precision and the calculation efficiency need to be considered. The complexity of the structure of the rolling mill gear box and the harshness of the working environment make the vibration signal contain not only the fault features of the parts, but also a large amount of interference noise, so that the signal-to-noise ratio in the angular domain vibration signal is low, which brings difficulties to the signal analysis and fault identification.

[0003] Since the gear is located inside the gear box, it is impossible to directly install a vibration measuring point, and the vibration monitoring point of the rolling mill gear box is usually arranged on the bearing seat or the box outside the gear box. The signal of one measuring point contains not only the vibration information of the gear, but also the vibration information of the bearing. To realize the fault diagnosis of the gear and the bearing, the key is to extract the respective fault features from the vibration signal of the measuring point.

[0004] The fault feature information of the bearing is reflected in the high-frequency resonance band region. If the order resampling method is used, a very large angular domain resampling rate needs to be set to ensure subsequent analysis, which will bring huge calculation complexity and is not suitable for online monitoring and diagnosis. The speed of the rolling mill gear box is low, and the gear box is a one-stage transmission, so the rotation frequency of the gear and the meshing frequency are often in the low frequency region, while the resonance frequency band of the bearing fault appears in the medium and high frequency region. Therefore, the vibration components representing the fault information of the two types of faults can be separated by filtering, that is, the gear fault feature information is extracted by low-pass filtering the vibration signal, and the rolling bearing fault feature information is extracted by band-pass filtering and envelope demodulation of the vibration signal. In addition, considering the vibration noise interference in the field, the frequency order related to the reference shaft rotation frequency in the order spectrum is fixed, while the noise is random. Therefore, by means of the time domain averaging processing idea, the angular domain averaging processing is performed on the angular domain vibration signal of the rolling mill gear box, which can reduce the influence of the noise component, extract the periodic features related to the fault, and further identify the possible fault damage of the rotating parts.

[0005] When the order analysis is applied to the rolling mill gear box, the method needs to be designed according to the characteristics of the rolling mill vibration signal, and the feasibility, precision and calculation efficiency need to be considered. The related technology and application have not been reported. SUMMARY

[0006] The technical problem solved by the present application is to provide a rolling mill gearbox fault feature extraction method based on adaptive order analysis, which utilizes rolling mill gearbox vibration signals, and through adaptive filtering of gear and bearing vibration features, identifies gear faults based on gear sideband amplitude and modulation ratio, and identifies bearing faults based on the ratio of bearing fault feature amplitude to rotational frequency amplitude, eliminating the influence of working condition changes and noise components, and achieving accurate fault identification.

[0007] To solve the above technical problems, the rolling mill gearbox fault feature extraction method based on adaptive order analysis comprises the following steps:

[0008] Step one, obtain the vibration signals of the gearbox measuring point, the real-time speed curve, and the analysis parameters of the gear and bearing, wherein the analysis parameters include the gear tooth number z, the bearing inner ring fault feature coefficient c i , the bearing outer ring fault feature coefficient c o , the bearing rolling element fault feature coefficient c b , and the bearing cage fault feature coefficient c c ;

[0009] Step two, set the gear highest meshing frequency multiple k g , calculate the gear order bandwidth o gm according to the gear tooth number z,

[0010] o gm =k g ×z

[0011] wherein the gear highest meshing frequency multiple k g should be no less than 5;

[0012] Step three, calculate the highest speed s max from the obtained real-time speed curve, and calculate the gear vibration component low-pass filter cutoff frequency f g according to the highest speed s max and the gear order bandwidth o gm ,

[0013] f g =o gm ×s max

[0014] Step four, low-pass filter the vibration signals according to the gear vibration component low-pass filter cutoff frequency f g to obtain the vibration components required for gear analysis;

[0015] Step five, in the amplitude spectrum of the vibration signals, take the gear vibration component low-pass filter cutoff frequency fg For the lower limit, search the maximum amplitude in the amplitude spectrum, and the corresponding frequency is the center frequency f of the bearing vibration component band-pass filtered b ;

[0016] Step six, set the bearing maximum fault characteristic frequency multiple k b , according to the bearing maximum fault characteristic coefficient max(c i ,c o ,c b ,c c ) and the highest speed s max Calculate the band-pass filter bandwidth f bm ,

[0017] f bm =k b ×max(c i ,c o ,c b ,c c )×s max

[0018] Wherein, the bandwidth refers to a single width on one side, and the total bandwidth of the band-pass filter is twice the bandwidth, which should at least ensure that the 3rd harmonic of the bearing fault characteristic frequency is analyzed, then the bearing maximum fault characteristic frequency multiple k b =3;

[0019] Step seven, according to the bandwidth f bm and the filter center frequency f b , the vibration signal is band-pass filtered, and the Hilbert envelope demodulation is carried out, and the bearing analysis required vibration component is obtained;

[0020] Step eight, respectively, the gear vibration component and the bearing vibration component are angular domain resampling processed, according to the gear low-pass filter cutoff frequency f g and the bearing band-pass filter bandwidth f bm , the gear and bearing angular domain resampling rate is calculated respectively,

[0021]

[0022] In the formula: o gs is the gear vibration component angular domain resampling rate, o bs is the bearing vibration component angular domain resampling rate, s min is the lowest speed;

[0023] Step nine, the angular domain resampling signal of the gear and bearing vibration component is angular domain averaged, and the signal sequence after equal angle sampling is recorded as y(i), i=1,2,3…,N, and the sampling order is o s , then the total number of revolutions R is:

[0024]

[0025] If the reference axis rotation frequency period is r, that is, r is taken to average y(i), the angular domain average algorithm is:

[0026]

[0027] In the formula: is the signal sequence after angular domain averaging, m = 1, 2, … L, is the number of segments into which the signal after equal-angle sampling is evenly divided, Downward nearest integer value, L = r·o s , the number of sampling points in each segment is the same, n is the serial number of the signal segment;

[0028] Step ten, Fourier transform is performed on the angular domain average resampled signals of the gear vibration component and the bearing vibration component respectively to obtain the gear order spectrum and the bearing order spectrum;

[0029] Step eleven, extract the rolling mill gearbox fault feature, calculate the feature for identifying gear fault,

[0030]

[0031] In the formula, ER is the feature for identifying gear fault, is the amplitude at the modulation rotation frequency order interval k times, k takes the value of -2, -1, 1, 2, is the amplitude at the 1, 2, 3 times meshing frequency order, is the amplitude at the maximum meshing frequency order;

[0032] Calculate the feature for identifying bearing fault,

[0033]

[0034] BR o , BR i , BR b , BR c are the fault features of the bearing outer ring, inner ring, rolling body and cage respectively, are the amplitudes at the 1 times fault feature frequency order of the bearing outer ring, inner ring, rolling body and cage respectively, is the amplitude at the rotation frequency order.

[0035] Because the adaptive order analysis-based rolling mill gearbox fault feature extraction method of this invention adopts the above-mentioned technical solution, namely, this method calculates the low-pass filter cutoff frequency of the gear vibration component based on the gear order bandwidth, searches for the center frequency of the band-pass filter of the bearing vibration component in the high-frequency peak, and calculates the band-pass filter bandwidth; obtains the gear vibration component through low-pass filtering, and obtains the bearing vibration component through band-pass filtering and envelope demodulation; averages the angular domain resampled signal, identifies gear faults using the gear sideband amplitude and modulation ratio as features, and identifies bearing faults using the ratio of the bearing fault characteristic amplitude to the rotational frequency amplitude as a feature. This method designs characteristic indicators based on the vibration performance of gear and bearing faults, which have clear physical meaning, can eliminate the influence of operating condition changes and noise components, and achieve accurate identification of rolling mill gearbox faults. Attached Figure Description

[0036] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments:

[0037] Figure 1 The flowchart is shown below for a method for extracting fault features of rolling mill gearboxes based on adaptive order analysis.

[0038] Figure 2 This is a schematic diagram of the rolling mill gearbox structure;

[0039] Figures 3a-3f The diagram shows the rolling mill vibration acceleration signal, speed curve, gear vibration component angular domain average resampled signal waveform and order spectrum, and bearing vibration component angular domain average resampled signal waveform and order spectrum as specific examples of this method. Detailed Implementation

[0040] Implementation, for example Figure 1 As shown, the method for extracting fault features of rolling mill gearboxes based on adaptive order analysis of the present invention includes the following steps:

[0041] Step 1: Acquire vibration signals, real-time speed curves, and analysis parameters of the gears and bearings at the gearbox measuring points. These parameters include the number of gear teeth (z) and the bearing inner ring fault characteristic coefficient (c). i Bearing outer ring fault characteristic coefficient c o Fault characteristic coefficient c of bearing rolling elements b and bearing cage failure characteristic coefficient c c ;

[0042] Step 2: Set the maximum gear meshing frequency multiple k g Calculate the gear order bandwidth o based on the number of teeth z of the gear. gm ,

[0043] o gm =k g ×z

[0044] wherein the highest gear meshing frequency multiple k g should be no less than 5;

[0045] Step three, calculating the highest speed s from the obtained real-time speed curve max , according to the highest speed s max and the gear order bandwidth o gm , calculating the gear vibration component low-pass filter cutoff frequency f g ,

[0046] f g = o gm × s max

[0047] Step four, according to the gear vibration component low-pass filter cutoff frequency f g , low-pass filtering the vibration signal to obtain the vibration component required for gear analysis;

[0048] Step five, in the amplitude spectrum of the vibration signal, taking the gear vibration component low-pass filter cutoff frequency f g as the lower limit, searching for the frequency corresponding to the maximum amplitude in the amplitude spectrum, and taking the frequency as the center frequency f b of the bearing vibration component band-pass filtering;

[0049] Step six, setting the bearing highest fault characteristic frequency multiple k b , according to the bearing maximum fault characteristic coefficient max(c i , c o , c b , c c ) and the highest speed s max calculating the band-pass filter bandwidth f bm ,

[0050] f bm = k b × max(c i , c o , c b , c c )× s max

[0051] wherein the bandwidth refers to a single width on one side, and the total bandwidth of the band-pass filter is twice the bandwidth, which should at least be able to ensure that the 3rd harmonic of the bearing fault characteristic frequency is analyzed, then the bearing highest fault characteristic frequency multiple k b = 3;

[0052] Step seven, according to the bandwidth f bm and the filter center frequency f b of the band-pass filtering, band-pass filtering the vibration signal and performing Hilbert envelope demodulation to obtain the vibration component required for bearing analysis;

[0053] Step eight, angular domain resampling processing is carried out on the gear vibration component and the bearing vibration component respectively, and the gear low-pass filter cutoff frequency f g and the bearing band-pass filter bandwidth f bm The gear and bearing angular domain resampling rates are calculated respectively,

[0054]

[0055] In the formula, o gs is the gear vibration component angular domain resampling rate, o bs is the bearing vibration component angular domain resampling rate, s min is the lowest rotating speed;

[0056] Step nine, angular domain average processing is carried out on the angular domain resampling signals of the gear and bearing vibration components respectively, and the signal sequence after equal-angle sampling is denoted as y(i), i = 1, 2, 3…, N, and the sampling order is denoted as o s The total rotating number R is:

[0057]

[0058] If the reference shaft rotating frequency period is r, that is, y(i) is averaged every r revolutions, the angular domain average algorithm is:

[0059]

[0060] In the formula, is the signal sequence after angular domain average, m = 1, 2, …L, is the number of segments into which the signal after equal-angle sampling is evenly divided, L = r·o s is the same number of sampling points in each segment, and n is the serial number of the signal segment;

[0061] Step ten, Fourier transform is carried out on the angular domain average resampling signals of the gear vibration component and the bearing vibration component respectively, to obtain the gear order spectrum and the bearing order spectrum;

[0062] Step eleven, the rolling mill gearbox fault feature is extracted, and the feature for identifying the gear fault is calculated,

[0063]

[0064] In the formula, ER is the feature for identifying the gear fault, is the amplitude at the modulation rotating frequency order interval k times, the value of k is -2, -1, 1, 2, is the amplitude at the meshing frequency order interval 1, 2, 3 times, is the amplitude at the maximum meshing frequency order;

[0065] characteristics of identifying bearing faults are calculated,

[0066]

[0067] BR o , BR i , BR b , BR c are fault characteristics of bearing outer ring, inner ring, rolling element and cage respectively, are amplitude values at fault characteristic frequency order of bearing outer ring, inner ring, rolling element and cage respectively, is amplitude value at rotation frequency order.

[0068] The method will be described in detail below with specific examples.

[0069] Figure 2 Fig. 1 is a schematic diagram of a structure of a gear box of a finishing mill F2, four measuring points are arranged on bearing seats and two measuring points are arranged on the gear box body, and the gear teeth number of the input shaft gear of the gear box is 22.

[0070] Fig. 3(a) is a vibration acceleration signal of bearing monitoring alarm of the measuring point on the transmission side of the input shaft of the gear box of the finishing mill F2 in a certain rolling process, the sampling frequency is 20 kHz, and the time interval of the discrete rotation speed signal is 0.01 s. It is found that the lower shaft transmission side bearing outer ring is damaged. Fig. 3(b) is a rotation speed curve obtained by segment fitting of the discrete rotation speed signal.

[0071] The bearing inner ring fault characteristic coefficient is 13.218, the highest rotation frequency of the input shaft is 1.1397 Hz and the lowest rotation frequency is 0.8512 Hz obtained from the rotation speed curve. According to these parameters, the parameters of the feature extraction algorithm can be determined, wherein: the low-pass filter cutoff frequency of the gear is set to 126 Hz, the angular domain sampling rate of the gear vibration component is set to 380; the center frequency of the bearing band-pass filter cutoff frequency is set to 510.8 Hz, the bandwidth is 46 Hz, and the angular domain sampling rate of the bearing vibration component is set to 140; the length of 8 power frequency periods (i.e. r = 8 rotations) is taken as the period for angular domain averaging.

[0072] Fig. 3(c) and (d) are the waveform and order spectrum of the angular domain average resampling signal of the gear vibration component, and from Fig. 3(d), the 1 times, 2 times, 3 times, 4 times and 5 times meshing frequency orders of the gear can be clearly seen. The amplitudes of the 1 times, 2 times and 3 times meshing frequency orders in the gear order spectrum are 0.0040, 0.01796 and 0.0020 respectively, and the amplitude at the 2 times meshing frequency order is the largest, so the 2 times meshing frequency order is selected as the analysis object. The rotational frequency order interval around the 2 times meshing frequency order and the amplitude at the 2 times rotational frequency order interval are 0.0017, 0.0014, 0.0020 and 0.0034 respectively, and the sum of the sideband amplitudes is 0.0085. The calculated sideband amplitude sum modulation ratio is 0.4733, the diagnostic threshold is set to 1, and it is judged that the running state of the gear is normal.

[0073] Fig. 3(e) and Fig. 3(f) show the waveform and order spectrum of the angular domain average resampling signal of the bearing vibration component, and in Fig. 3(f), the amplitudes at the 10.75 order and the 21.75 order are prominent, the 10.75 is the 1 times fault characteristic order of the transmission side bearing outer ring, and the 21.75 is its 2 times order. The amplitude of the rotational frequency order in the bearing order spectrum is 0.0035, and the amplitude at the 1 times fault characteristic order of the bearing outer ring is 0.0109, and the calculated outer ring fault characteristic amplitude ratio is 3.1143, and the diagnostic threshold is set to 2, and it is judged that the bearing has outer ring fault. The diagnostic result is consistent with the actual situation, which shows that the method effectively extracts the gear and bearing fault characteristics in the rolling mill gearbox vibration signal and accurately identifies the fault.

[0074] Through the above application description, the method can not only effectively detect the fault characteristic frequency under variable speed, but also separate the vibration components of gear fault and bearing fault, establish characteristic indexes respectively, and realize automatic diagnosis of rolling mill gearbox fault.

Claims

1. A method for extracting fault features of rolling mill gearboxes based on adaptive order analysis, characterized in that... This method includes the following steps: Step 1: Acquire vibration signals, real-time speed curves, and analysis parameters of the gears and bearings at the gearbox measuring points. These analysis parameters include the number of gear teeth (z) and the bearing inner ring fault characteristic coefficient (c). i Bearing outer ring fault characteristic coefficient c o Fault characteristic coefficient c of bearing rolling elements b and bearing cage failure characteristic coefficient c c ; Step 2: Set the maximum gear meshing frequency multiple k g Calculate the gear order bandwidth o based on the number of teeth z of the gear. gm , oh gm =k g ×z Among them, the highest gear meshing frequency multiple k g It should be no less than 5; Step 3: Calculate the maximum speed s from the obtained real-time speed curve. max According to the maximum speed s max and gear order bandwidth o gm Calculate the low-pass filter cutoff frequency f of the gear vibration component. g , f g =o gm ×s max Step 4: Based on the low-pass filter cutoff frequency f of the gear vibration component g The vibration signal is low-pass filtered to obtain the vibration components required for gear analysis. Step 5: In the amplitude spectrum of the vibration signal, use the low-pass filter cutoff frequency f of the gear vibration component. g As the lower limit, search for the frequency corresponding to the maximum amplitude in the amplitude spectrum, and use this frequency as the center frequency f of the bandpass filter for the bearing vibration component. b ; Step 6: Set the bearing's highest fault characteristic frequency multiple k b According to the bearing's maximum failure characteristic coefficient max(c i ,c o ,c b ,c c ) and maximum speed s max Calculate the bandpass filter bandwidth f bm , f bm =k b ×max(c i ,c o ,c b ,c c )×s max Here, bandwidth refers to the single-sided width, and the total bandwidth of the bandpass filter is twice that. This bandwidth should be able to guarantee that the third harmonic of the bearing fault characteristic frequency can be analyzed. Therefore, the multiple of the highest bearing fault characteristic frequency k is... b =3; Step 7: Based on the bandwidth f of the bandpass filter bm and the filter center frequency f b The vibration signal is bandpass filtered and demodulated using Hilbert envelope to obtain the vibration components required for bearing analysis. Step 8: Perform angular domain resampling processing on the gear vibration components and bearing vibration components respectively, based on the low-pass filter cutoff frequency f of the gear. g and bearing bandpass filter bandwidth f bm Calculate the resampling rate for the gear and bearing angular domains respectively. In the formula: o gs For the gear vibration component angular domain resampling rate, o bs The bearing vibration component angular domain resampling rate, s min This is the lowest speed. Step 9: Perform angular domain averaging on the angular domain resampled signals of the gear and bearing vibration components respectively, and denote the signal sequence after equal-angle sampling as y(i), i = 1, 2, 3, ..., N, and assume the sampling order is 0. s Then the total number of revolutions R during sampling is: If the reference axis rotation period is r, that is, r rotations are used to extract y(i) for averaging, then the angular domain averaging algorithm is as follows: In the formula: It is the signal sequence after angular domain averaging, m=1,2,…L, It is the number of segments into which the signal, after being sampled at equal angles, is divided. This means rounding down to the nearest integer value, L = r·o s , where n is the number of identical sampling points in each segment, and n is the sequence number of the signal segment; Step 10: Perform Fourier transform on the angular domain averaged resampled signals of the gear vibration component and the bearing vibration component respectively to obtain the gear order spectrum and the bearing order spectrum. Step 11: Extract the fault characteristics of the rolling mill gearbox and calculate and identify the characteristics of gear faults. In the formula, ER represents the characteristics for identifying gear faults. It is the amplitude at intervals k times the modulation frequency shift order, where k takes values ​​of -2, -1, 1, and 2. The amplitudes at 1st, 2nd, and 3rd times the meshing frequency are... This represents the amplitude at the order of maximum meshing frequency. Calculate and identify the characteristics of bearing failures. BR o BR i BR b BR c These are the fault characteristics of the bearing's outer ring, inner ring, rolling elements, and cage, respectively. These represent the amplitudes at one fault characteristic frequency order for the bearing's outer ring, inner ring, rolling elements, and cage, respectively. This represents the amplitude at the frequency order.

Images