Rolling bearing fault damage degree recognition method based on acoustic emission multi-parameter fusion

Abstract

The application provides a rolling bearing fault damage degree recognition method based on acoustic emission multi-parameter fusion, and relates to the field of rolling bearing fault diagnosis. Firstly, whether the bearing has a fault is preliminarily judged according to the acoustic emission characteristic parameter TAFI; then the acoustic emission multi-parameter count and the impact number are used to identify the fault type of the rolling bearing, and the different damage degrees of the rolling bearing are identified according to the characteristic parameter energy, the outer ring fault and the rolling body fault; finally, the introduced fault factor parameter is used to identify the different damage degrees of the typical fault rolling bearing, and the shortage of the characteristic parameter energy in identifying the damage degree of the inner ring fault is made up. The method involves acoustic emission parameter analysis, waveform flow envelope spectrum analysis and fault factor calculation analysis, and is a comprehensive bearing typical fault damage degree recognition analysis method, which can effectively identify the different damage degrees of the typical fault of the rolling bearing, and can provide certain method guidance for the typical fault damage degree recognition of the main shaft bearing and the intermediate bearing of the aero-engine.

Description

Technical Field

[0001] This invention relates to the field of bearing fault diagnosis technology, and in particular to a method for identifying the degree of damage in rolling bearing faults based on acoustic emission multi-parameter fusion. Background Technology

[0002] Bearing fault diagnosis and condition monitoring are crucial aspects of mechanical equipment fault diagnosis technology. Bearing fault diagnosis methods mainly include vibration testing, acoustic emission testing, temperature monitoring, lubricant testing, and clearance testing. Among these, acoustic emission testing technology, with its advantages of high signal frequency, clear characteristic signals, sensitivity to impact signals, and insensitivity to structural and rotational dynamic vibration noise, is widely used in online monitoring and fault diagnosis of rolling bearings. In actual field operations, simply knowing whether a bearing has failed and the type of failure is far from sufficient for preventative maintenance. Understanding the extent of damage from typical rolling bearing failures is of significant practical engineering importance for providing effective and appropriate maintenance.

[0003] Regarding methods for identifying the damage level of typical rolling bearing faults, domestic and international scholars mainly focus on analyzing vibration signals, with few applying acoustic emission signals. In methods using vibration signals to study the damage level of typical bearing faults, scholars emphasize different decomposition methods to effectively extract fault characteristic parameters, thereby distinguishing bearings with different damage levels and achieving identification of typical bearing fault damage. However, effective identification methods are limited, and the types of bearing faults that can be identified are also relatively narrow. Few studies discuss methods for identifying the damage level of rolling bearing faults based on acoustic emission parameter analysis. Acoustic emission parameters are abundant, including ring count, energy, amplitude, and impact count, each reflecting different bearing fault information and severity. While comparative analysis of a single acoustic emission parameter can effectively identify the bearing fault type, current comparative analysis of a single acoustic emission parameter is insufficient to effectively distinguish the magnitude of bearing fault damage for bearings with the same fault type but different damage levels. Therefore, there are still many shortcomings in identifying the damage level of rolling bearing faults based on acoustic emission parameters, and a more efficient and accurate multi-parameter fusion technology for identifying the damage level of bearing faults is urgently needed. Summary of the Invention

[0004] To address the aforementioned issues, this invention introduces a dimensionless fault factor that is fused with acoustic emission parameters, providing a method for identifying the damage degree of typical rolling bearing faults based on the fusion of multiple acoustic emission parameters. This method can effectively identify different damage degrees of typical rolling bearing faults, providing a theoretical basis for identifying the damage degree of typical rolling bearing faults.

[0005] This invention provides a method for identifying the degree of damage in rolling bearing faults based on acoustic emission multi-parameter fusion, comprising the following steps:

[0006] Step 1: Acquire the acoustic emission signal of the rolling bearing using an acoustic emission (AE) sensor;

[0007] Step 2: Use TAFI analysis to preliminarily determine whether the bearing has a fault;

[0008] TAFI is a feature parameter extracted based on the frequency, amplitude, and time characteristics of AE impacts. When TAFI detects the bearing condition, if the image is messy and disordered, no bearing fault is identified, and it can be preliminarily determined to be a healthy bearing. If the image is regular stripes, it can be preliminarily determined that the bearing is faulty. Then, step 3 is executed to further determine the specific fault type and damage degree based on some feature parameters in the acoustic emission signal (hereinafter referred to as acoustic emission parameters). The feature parameters include ring count (hereinafter referred to as count), impact count (hereinafter referred to as impact number), signal energy (hereinafter referred to as energy), etc.

[0009] Step 3: If the initial assessment indicates a bearing fault, diagnose the bearing fault by combining the acoustic emission parameter count versus time evolution graph, the acoustic emission parameter impact number versus amplitude distribution graph, and the acoustic emission parameter energy versus time evolution graph; the specific process is as follows:

[0010] The counting parameters in the acoustic emission signal are divided into intervals according to a preset time interval, and the count values ​​corresponding to each time interval are summed to obtain the count value Q corresponding to each interval. An experience graph f1 is plotted on the count value Q and time t. Based on the plotted experience graph f1, the range of the counting intervals in which different bearing faults are located is determined.

[0011] The impact number parameter in the acoustic emission signal was statistically analyzed using the distribution map analysis method. A distribution map f2 of impact number and amplitude was plotted. Based on the plotted distribution map f2, the impact number range of different bearing faults was determined.

[0012] The energy parameters in the acoustic emission signal are divided into intervals according to a preset time interval, and the energy values ​​corresponding to each time interval are summed to obtain the energy value W corresponding to each interval. An experience graph f3 is plotted on the energy value W and time t. The degree of damage of the bearing outer ring and rolling element is determined based on the plotted experience graph f3.

[0013] For bearing inner ring failures, it is necessary to introduce failure factors to analyze the degree of bearing damage; the specific process is as follows:

[0014] Step 4: Select the acoustic emission waveform stream signal during a period of stable rotation speed in the test, apply EMD signal decomposition to perform envelope detection processing on the high-frequency acoustic emission waveform stream signal, and use Hilbert transform to extract fault information; the specific process is as follows:

[0015] Step 4.1: Determine all local extreme points of signal x(t), then connect all the maximum and minimum points with cubic spline functions to form upper and lower envelopes, calculate their average value function m1(t), and subtract m1(t) from x(t) to obtain h1(t) as the signal to be processed;

[0016] h1(t)=x(t)-m1(t) (1)

[0017] Step 4.2: Use formula (2) to perform the first screening of the signal h1(t) to be processed, and obtain the first iteration signal h. 1,1 (t):

[0018] h 1,1 (t)=h1(t)-m 1,1 (t) (2)

[0019] In the formula, m 1,1 (t) represents the average curve of the upper and lower envelopes formed by connecting all the maximum and minimum points in h1(t);

[0020] After k rounds of filtering, the first fundamental modal component h is obtained. 1,k Let c1(t) = h 1,k (t);

[0021] h 1,k (t)=h 1,k-1 (t)-m 1,k (t) (3)

[0022] In the formula, h 1,k-1 (t) represents the signal of the (k-1)th iteration; m 1,k (t) indicates that h 1,k-1 Connect all the maximum and minimum points in (t) to form the average curve of the upper and lower envelopes;

[0023] Step 4.3: Use formula (4) to separate the basic modal components from the original signal;

[0024] r1(t)=x(t)-c1(t) (4)

[0025] Step 4.4: Using r1(t) as the new original signal, n signal components are obtained sequentially:

[0026]

[0027] In the formula, r n (t) represents the nth signal component, c n (t) represents the nth fundamental mode component;

[0028] Finally, x(t) is decomposed into:

[0029]

[0030] Step 4.5: Envelope detection is performed on the high signal-to-noise ratio high-frequency acoustic emission signal to obtain the envelope waveform. Then, Hilbert transform is used to de-envelope the signal and extract fault information. The Hilbert transform is defined as:

[0031]

[0032] Where x(t) is the original time-domain signal; Let x(t) be the Hilbert transform of the signal; Perform a convolution on x(t), and the impulse response of this convolution is:

[0033] Step 5: Introduce a dimensionless parameter fault factor to characterize the damage degree of rolling bearings with inner ring defects;

[0034] Step 5.1: When the outer ring is fixed, the formula for calculating the characteristic frequency of the bearing inner ring failure is:

[0035]

[0036] In the formula, D b Let d be the diameter of the rolling element. m D is the inner diameter. m D is the outer diameter. c Let be the pitch circle diameter, and 2D c =d m +D m θ is the angular contact angle, and z is the number of rolling elements. i The characteristic frequency of bearing inner ring failure, in Hz, f r Reference axis rotational frequency;

[0037] Based on the operating conditions, the corresponding speed and bearing geometric parameters are substituted into formula (8) to obtain the theoretical value of the bearing inner ring fault characteristic frequency;

[0038] Step 5.2: Compare the fault information extracted in Step 4 with the theoretical fault characteristic frequency obtained in Step 5.1 to determine the power peak D corresponding to the i-th (i = 1, 2, 3...) times the characteristic frequency. i And calculate the average power within the frequency band between the (i-1)th and (i+1)th harmonics of the fault characteristic frequency. in The calculation does not include the peak power D corresponding to the (i-1), i, and i+1 times the fault characteristic frequency. i-1 D i and Di+1 ;

[0039] Step 5.3: Calculate the fault factor γ for each harmonic, and characterize the damage state of different defects in the bearing based on the value of the fault factor;

[0040]

[0041] The beneficial effects of this invention are:

[0042] 1. The main idea of ​​the method provided by this invention lies in the identification process of typical fault damage degree of rolling bearings, which involves parameter analysis, waveform flow envelope spectrum analysis and fault factor calculation analysis. It is a comprehensive method for identifying and analyzing the damage degree of typical bearing faults.

[0043] 2. The acoustic emission parameter TAFI analysis used in the method provided by this invention can preliminarily determine whether a bearing is faulty, and analyze the data of identified bearing faults, thus reducing the amount of data. By analyzing the differences in the numerical range of acoustic emission characteristic parameter counts and impact numbers between different faulty bearings and healthy bearings, different types of rolling bearing faults can be effectively diagnosed.

[0044] 3. The acoustic emission characteristic parameter energy selected by the method provided by the present invention is highly sensitive to different degrees of damage to rolling bearings. The difference in the energy range of wire cutting and pitting defects of the outer ring and rolling element faulty bearings is very obvious. The energy can effectively identify the degree of damage of the outer ring and rolling element faulty rolling bearings.

[0045] 4. The method provided by this invention introduces a dimensionless parameter fault factor to characterize the damage degree of rolling bearings with different defects. By using the difference in the numerical values ​​of the fault factors of wire cutting and pitting defects at 1-5 times the frequency, it effectively identifies the different damage degrees of rolling bearings with inner ring faults, makes up for the inadequacy of characteristic parameter energy in identifying the damage degree of inner ring faults, and realizes the identification of the damage degree of typical rolling bearing faults. Attached Figure Description

[0046] Figure 1 This is a flowchart of the rolling bearing fault damage degree identification method based on acoustic emission multi-parameter fusion in an embodiment of the present invention;

[0047] Figure 2 These are the TAFI time-lapse graphs of healthy bearings and typical faulty bearings in embodiments of the present invention.

[0048] Figure 3 The following are time-lapse graphs showing the counts of different bearing failure types in embodiments of the present invention; wherein, (a) is a time-lapse graph showing the counts of wire-cut defect bearings, and (b) is a time-lapse graph showing the counts of pitting defect bearings.

[0049] Figure 4 The following are amplitude distribution diagrams of impact number of bearings with different faults in the embodiments of the present invention; wherein, (a) is an empirical diagram of impact number versus amplitude of bearing with wire cutting defect, and (b) is an empirical diagram of impact number versus amplitude of bearing with pitting defect.

[0050] Figure 5 The following are energy versus time evolution diagrams for different bearing failure types in embodiments of the present invention; wherein, (a) is the energy versus time evolution diagram of a wire-cut defect bearing, and (b) is the energy versus time evolution diagram of a pitting defect bearing;

[0051] Figure 6 The following is a three-dimensional intuitive diagram of the acoustic emission parameters in the embodiment of the present invention; wherein, (a) is the average count of typical faulty bearings under different damage states, (b) is the average energy of typical faulty bearings under different damage states, and (c) is the total number of impacts of typical faulty bearings under different damage states.

[0052] Figure 7 The following are envelope diagrams for different fault types under wire cutting and pitting defects in the embodiments of the present invention: (a) is the outer ring fault envelope spectrum of wire cutting defect, (b) is the outer ring fault envelope spectrum of pitting defect, (c) is the rolling element fault envelope spectrum of wire cutting defect, (d) is the rolling element fault envelope spectrum of pitting defect, (e) is the inner ring fault envelope spectrum of wire cutting defect, and (f) is the inner ring fault envelope spectrum of pitting defect.

[0053] Figure 8 This is a definition diagram of the dimensionless parameter fault factor of the acoustic emission waveform stream signal in an embodiment of the present invention;

[0054] Figure 9 These are the fault factors for various damage states of the bearing in the embodiments of the present invention; (a) is the fault factor under the damage state of the outer ring of the bearing, (b) is the fault factor under the damage state of the rolling elements of the bearing, and (c) is the fault factor under the damage state of the inner ring of the bearing.

[0055] Figure 10 This is a diagram showing the definition of acoustic emission signal characteristic parameters in an embodiment of the present invention. Detailed Implementation

[0056] This invention combines acoustic emission parameter analysis and waveform flow analysis with the introduction of a dimensionless fault factor to analyze acoustic emission signals of different damage levels in typical rolling bearing faults, achieving a multi-parameter acoustic emission fusion-based method for identifying the damage level of typical rolling bearing faults. First, typical rolling bearing fault types are identified through acoustic emission parameter counting and impact number analysis. Then, the damage level of typical rolling bearing faults is identified through single acoustic emission parameter energy analysis combined with the introduced dimensionless fault factor. Ultimately, this achieves the desired effect of identifying the damage level of typical rolling bearing faults based on multi-parameter acoustic emission fusion.

[0057] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0058] This embodiment uses real experimental data for analysis. The typical rolling bearing fault simulation test device consists of a variable speed motor, coupling, bearing housing, shaft, mounting bearing, test bearing, and rotor. The acoustic emission data acquisition system is AEwin for PC12, and the AE sensor is installed vertically on the bearing housing of the test bearing. The faulty bearing selected for analysis is the NJ204EM cylindrical roller bearing, and the bearing parameters are shown in Table 1. Bearing faults are simulated by artificially machining wire cutting and pitting defects on the outer ring, inner ring, and rolling elements of the rolling bearing. The wire cutting defects are rectangular grooves with a width and depth of 1 mm, and the pitting defects are cylindrical grooves with a diameter and depth of 2 mm. The operating condition is 840 r / min. During the test, the acoustic emission signal sampling frequency f of the rolling bearing is... s =1MHz.

[0059] Table 1 Geometric parameters of the test bearings

[0060]

[0061] like Figure 1 As shown, a method for identifying the degree of damage in rolling bearing faults based on acoustic emission multi-parameter fusion includes the following steps:

[0062] Step 1: Acoustic emission signals from the rolling bearing are acquired using an AE sensor, and relevant bearing parameters are measured simultaneously. These parameters include: contact angle α, number of balls Z, and rolling element diameter D. b , pitch diameter D c For detailed parameters, please refer to Table 1.

[0063] Step 2: Use TAFI analysis to preliminarily determine whether the bearing has a fault;

[0064] Generate a TAFI time-lapse graph. The TAFI time-lapse graph generated in this embodiment is shown in the attached figure. Figure 2 As shown in the attached image. If the image appears cluttered and disorganized, as shown in the attached image. Figure 2 (a) indicates that no bearing fault was detected, and the bearing can be preliminarily determined to be healthy; if the image shows regular stripes, as shown in the attached image... Figure 2 (b) If the bearing is found to be faulty, it can be preliminarily determined that the bearing is faulty. TAFI analysis can be used for online bearing inspection to determine whether the bearing is faulty, and further processing can be performed on the test data of the bearing fault identified by TAFI analysis.

[0065] Acoustic emission parameter analysis is a method that analyzes the relationship between acoustic emission parameters and bearing fault type and fault damage by statistically analyzing the changes of parameters such as ring count, energy, amplitude, and impact count with factors such as rolling bearing fault type and fault damage degree.

[0066] like Figure 10 As shown, the specific acoustic emission signal characteristic parameters are described below:

[0067] (1) Ring count: The number of oscillations of the signal exceeding the threshold, used for evaluating acoustic emission activity. The ring count of an acoustic emission signal reflects the amplitude of the signal to some extent and is often used for assessing acoustic emission activity.

[0068] (2) Impact Count: Any signal that exceeds the threshold and causes a channel to acquire data is called an impact. It reflects the total amount and frequency of acoustic emission activity and is often used for acoustic emission activity evaluation.

[0069] (3) Amplitude: refers to the amplitude of the highest amplitude in the signal. Its value is not affected by the threshold size and is often used to evaluate the intensity of acoustic emission or as a measure of the rate of signal decay.

[0070] (4) Signal energy: refers to the area under the signal detection envelope, reflecting the signal strength, and its value is not affected by the threshold size.

[0071] Step 3: Analyze the bearing failure type based on the count and impact number in the acoustic emission signal;

[0072] Experience graph analysis is a method that analyzes the changes of acoustic emission parameters over time to obtain the bearing condition and development trend, and can be used to evaluate the activity of bearing faults.

[0073] This embodiment combines the acoustic emission parameter count versus time evolution graph, the acoustic emission parameter impact number versus amplitude distribution graph, and the acoustic emission parameter energy versus time evolution graph to diagnose bearing faults. The specific analysis process is as follows:

[0074] The acoustic emission raw data test time was divided into intervals of 0.5 seconds, and the count values ​​corresponding to each time interval were summed and plotted to determine the counting interval range of different bearing faults. The time evolution graph of the counts of different faulty bearings in this embodiment is shown in the attached figure. Figure 3 As shown. For the bearing under test, the fault type of the bearing is determined by comparing the data distribution of the measured acoustic emission parameter counts in the time-lapse graph with the distribution results of the parameters for different typical faults of the bearing.

[0075] Counting is a relatively sensitive parameter among acoustic emission parameters, exhibiting high sensitivity to rotational speed, load, fault size, and fault type. Due to the high density of acoustic emission signal acquisition, the fault characteristics are not clearly distinguishable in the time-lapse analysis of the raw data. Therefore, the test time of the raw data is divided into intervals of 0.5 seconds, and the count values ​​corresponding to each time interval are summed to obtain the processed data for plot analysis. The time-lapse graphs of bearing counts for different faults are shown below. Figure 3 As shown.

[0076] The time evolution graph of the wire EDM defect count is shown below. Figure 3 As shown in (a), the count ranges for bearing outer ring, rolling element, and inner ring failures are 200-500, 100-300, and 20-15, respectively, while the count for healthy bearings is below 20; the count of pitting defects versus time evolution is shown in the graph. Figure 3 As shown in (b), the count ranges for bearing outer ring, rolling element, and inner ring faults are 100-300, 30-80, and 10-30, respectively, while the count for healthy bearings is below 20. When a bearing has a fault, the count for outer ring faults increases significantly, while the counts for rolling elements and inner rings also increase slightly. Therefore, the outer ring fault count > rolling element fault count > inner ring fault count > healthy bearing count. By analyzing the count against the time history graph, the bearing fault type can be effectively determined.

[0077] pass Figure 3 A comparison of the values ​​in (a) and 3(b) shows that the number of failures in the outer ring, rolling elements and inner ring of the bearing with wire cutting defects is very small compared with that of the bearing with pitting defects, making it impossible to analyze the degree of damage to the bearing failure.

[0078] The impact count signal of the bearing acoustic emission was analyzed using the distribution map analysis method, and an impact number versus amplitude distribution map was plotted. The impact number versus amplitude distribution maps for different faulty bearings in this embodiment are attached. Figure 4 As shown, faulty bearings and healthy bearings are distinguished based on the maximum number of impacts. For the faulty bearing under test, the fault type of the bearing is determined by comparing the data distribution of the measured acoustic emission parameter impact number amplitude distribution diagram with the distribution results of this parameter for different typical faults of the bearing.

[0079] Depend on Figure 4It can be seen that the maximum impact counts for wire-cut defects in the outer ring, rolling elements, and inner ring of bearings, and for healthy bearings, are 134, 97, 52, and 22, respectively. For pitting defects, the maximum impact counts for the outer ring, rolling elements, and inner ring of bearings, and for healthy bearings, are 125, 80, 24, and 5, respectively. The maximum impact count of faulty bearings is significantly higher than that of healthy bearings. The maximum impact count is highest for bearings with faulty rolling elements. This is because rolling elements are rotating objects, resulting in a greater number of impacts than the stationary outer ring. Therefore, the maximum impact count shows the order: rolling element faults > outer ring faults > inner ring faults > healthy bearings. This can be used to distinguish between different faulty and healthy bearings using the impact count amplitude distribution diagram. Figure 4 As can be seen from (a) and 4(b), the graphic distributions are extremely similar, and the number of impacts is very close in value. The rolling bearings with different degrees of damage have very little distinguishability on the impact number amplitude distribution map.

[0080] The acoustic emission raw data test time was divided into intervals of 0.5 seconds, and the energy values ​​corresponding to each time interval were summed to plot the acoustic emission parameter energy versus time evolution diagram. The acoustic emission parameter energy versus time evolution diagrams for different faulty bearings in this embodiment are attached. Figure 5 As shown, the distribution of acoustic emission parameter energy values ​​over time in the time-series graph is used to determine the bearing outer ring and rolling element faults and the degree of damage. For the bearing under test, the distribution of the measured acoustic emission parameter energy values ​​over time in the time-series graph is compared with the distribution of the parameters for different typical faults of the bearing to determine the bearing fault type and the degree of damage.

[0081] The acoustic emission parameters of the wire-cut defect, energy versus time, are shown in the attached graph. Figure 5 As shown in (a), the energy ranges for bearing outer ring and rolling element failures are 400-1200 mV·μs and 200-800 mV·μs, respectively; the acoustic emission parameters of pitting defects versus time evolution diagrams are attached. Figure 5 As shown in (b), the energy ranges of bearing outer ring and rolling element faults are 30-90 mV·μs and 10-50 mV·μs, respectively; the energy of bearing inner ring faults with wire cutting and pitting defects, as well as healthy bearings, are all below 10 mV·μs; when a healthy bearing is running, the acoustic emission energy is relatively low, while the acoustic emission energy of bearings with outer ring and rolling element faults is significantly higher than that of healthy bearings at the same rotational speed. Because the inner ring moves with the bearing rotation, its acoustic emission fault signal must be refracted and reflected by mechanical components such as the rolling elements, cage, outer ring, and bearing housing before it can be transmitted to the sensor, resulting in significant energy attenuation. The energy of inner ring faults and healthy bearings is very close, therefore, the degree of movement of different faulty bearings varies, with the outer ring exhibiting the strongest movement and the inner ring the weakest. Figure 5A comparison of the values ​​in (a) and 5(b) shows that the energy of the bearing outer ring and rolling element failures caused by wire cutting defects is much higher than that of the bearing with pitting defects. The energy versus time evolution diagram can effectively identify the different degrees of damage to the outer ring and rolling elements of the rolling bearing.

[0082] By comprehensively analyzing the acoustic emission parameter count versus time evolution graph, the acoustic emission parameter impact number versus amplitude distribution graph, and the acoustic emission parameter energy versus time evolution graph, bearing faults can be determined.

[0083] Based on the data from the specific analysis of the acoustic emission characteristic parameters mentioned above, the count, average energy, and total number of impacts of typical faulty bearings under different damage states were statistically analyzed, as shown in Table 2. A three-dimensional visual graph of the acoustic emission parameters was also plotted, as shown in the attached figure. Figure 6 As shown in (a), (b), and (c).

[0084] Table 2. Statistical analysis of acoustic emission parameters of typical faulty bearings under different damage conditions.

[0085]

[0086] From the appendix Figure 6 As shown in (a), (b), and (c), taking wire EDM defects as an example, the average count, energy, and total number of impacts of bearing outer ring, rolling element, and inner ring faults are all higher than those of healthy bearings. Furthermore, there are differences in the average count, energy, and total number of impacts among typical faults. Numerical comparison can effectively distinguish typical bearing faults; therefore, the average count, energy, and total number of impacts can serve as effective means of diagnosing typical bearing faults. The differences in average count for outer ring faults with different damage states, average energy for inner ring faults with different damage states, and total number of impacts for rolling element faults with different damage states are not significant. Based on the above, it can be concluded that acoustic emission characteristic parameter analysis can serve as an effective means of bearing fault diagnosis, but it still has limitations in identifying bearing damage states.

[0087] The degree of failure of the bearing outer ring and rolling elements is determined by the data distribution of the acoustic emission parameter energy values ​​in the time history graph. The energy range of the bearing outer ring and rolling element failures with wire cutting defects is 400-1200mV·μs and 200-800mV·μs, respectively. The energy range of the bearing outer ring and rolling element failures with pitting defects is 30-90mV·μs and 10-50mV·μs, respectively. The energy of the bearing inner ring failures with wire cutting and pitting defects and healthy bearings is below 10mV·μs.

[0088] pass Figure 5The numerical comparison between (a) and (b) shows that the energy of the bearing outer ring and rolling element failures due to wire cutting defects is much higher than that of the bearing with pitting defects. The energy versus time history diagram can be used to effectively identify the different damage levels of the rolling bearing outer ring and rolling element failures. When the bearing inner ring fails, it is necessary to introduce a failure factor to analyze the bearing damage level.

[0089] Step 4: Select the acoustic emission waveform stream signal during a 1-second stabilization process of the rotational speed in the experiment. Apply EMD to decompose the signal and perform envelope detection processing on the high-frequency acoustic emission waveform stream signal. Use Hilbert transform to extract fault information. The inner ring fault envelope diagram of wire cutting and pitting defects obtained in this embodiment is shown in the attached figure. Figure 7 As shown in (a) and (b).

[0090] Empirical Mode Decomposition (EMD) is a time-frequency analysis method that requires no prior knowledge and is suitable for processing nonlinear and non-stationary acoustic emission signals. This method can decompose any signal into several fundamental mode components (IMFs) and a remainder sum.

[0091] The specific process of step 4 is as follows:

[0092] Step 4.1: Determine all local extrema of the signal x(t), then connect all the maxima and minima using a cubic spline function to form upper and lower envelopes, calculate their average curve m1(t), and subtract m1(t) from x(t) to obtain h1(t).

[0093] h1(t)=x(t)-m1(t) (1)

[0094] Step 4.2: Treat h1(t) as the signal to be processed, and repeat the above operation.

[0095] h 1,1 (t)=h1-m 1,1 (t) (2)

[0096] After k rounds of filtering, h 1,k (t) becomes the fundamental modal component;

[0097] h 1,k (t)=h 1,k-1 (t)-m 1,k (t) (3)

[0098] In the formula, h 1,k-1 (t) represents the signal of the (k-1)th iteration; m 1,k (t) indicates that h 1,k-1 Connect all the maximum and minimum points in (t) to form the average curve of the upper and lower envelopes;

[0099] Step 4.3: Decompose the first fundamental modal component c1(t), and denote c1(t) = h 1,k (t);

[0100] Separate the fundamental modal components from the original signal.

[0101] r1(t)=x(t)-c1(t) (4)

[0102] Step 4.4: Using r1(t) as the new original signal, repeat the above steps to obtain the following results.

[0103]

[0104] When r n When (t) is essentially monotonic or sufficiently small, the decomposition can be stopped. Finally, we obtain:

[0105]

[0106] Step 4.5: Envelope detection processing is performed on the high signal-to-noise ratio high-frequency acoustic emission waveform stream signal to obtain the envelope waveform. Then, Hilbert transform is used to de-envelope the signal and extract fault information. The Hilbert transform is defined as:

[0107]

[0108] Where x(t) is the original time-domain signal; Let x(t) be the Hilbert transform of the signal; Perform a convolution on x(t), and the impulse response of this convolution is:

[0109] (1) Find the Hilbert transform pair of the signal, that is, make the signal produce a 90° phase shift;

[0110] (2) An analytic signal is constructed by taking the original signal as the real part and the Hilbert transform pair as the imaginary part;

[0111] (3) Obtain the envelope of the signal by finding the modulus;

[0112] The envelope signal is low-pass filtered and the envelope spectrum is obtained by performing a fast Fourier transform.

[0113] like Figure 7 As shown in (a), peaks appear at 61Hz, 122Hz, 184Hz, 245Hz, and 306Hz in the envelope spectrum of the outer ring fault of the wire cutting defect. These peaks are compared with the theoretical fault characteristic frequency f. o The comparison with 60.91Hz confirms that the peaks in the graph are all main peaks within a local area near the theoretical fault characteristic frequency and its harmonics. Figure 7As shown in (b), peaks appear at 60Hz, 121Hz, 181Hz, 241Hz, and 301Hz in the envelope spectrum of the pitting defect's outer ring. These peaks are compared with the theoretical fault characteristic frequency f. o The comparison with 60.91Hz confirms that the peaks in the graph are the main peaks in the local area near the theoretical fault characteristic frequency and its harmonics. A slight error in the speed adjustment during the experiment caused a small difference in the fault frequency, but they remained very close. Therefore, it can be determined that the outer ring of the rolling bearing failed under both strong and weak damage conditions.

[0114] Envelope spectrum analysis was performed on the acoustic emission signals of bearings with wire cutting and pitting defects in the rolling elements to obtain the envelope spectrum of the rolling element failure due to wire cutting and pitting defects, such as... Figure 7 As shown in (c) and (d), peaks appear at 30Hz, 60Hz, 90Hz, 120Hz, and 150Hz in the envelope spectrum of the rolling element failure due to wire cutting defects. These peaks are compared with the theoretical fault characteristic frequency f. b Comparing the frequency to 30.25Hz, it can be determined that the peaks in the figure are all main peaks in the local range near the theoretical fault characteristic frequency and its harmonics. Peaks appear at 30Hz, 60Hz, 90Hz, and 120Hz in the envelope spectrum of pitting defects in rolling elements. These peaks are consistent with the theoretical fault characteristic frequency f. b The comparison with 30.25Hz confirms that the peaks in the graph are all main peaks in the local area near the theoretical fault characteristic frequency and its harmonics. Therefore, it can be determined that the rolling elements of the bearing have failed under both strong and weak damage conditions.

[0115] like Figure 7 As shown in (e) and (f), peaks appear at 99Hz, 198Hz, 297Hz, 396Hz, and 495Hz in the envelope spectrum of the bearing signal with inner ring failure due to wire cutting. These peaks are compared with the theoretical fault characteristic frequency f. i The comparison with 99.04Hz confirms that the peaks in the graph are all main peaks within a local area near the theoretical fault characteristic frequency and its harmonics. Figure 7 As shown, peaks appeared at 99Hz, 197Hz, 296Hz, 396Hz, and 494Hz in the envelope spectrum of the inner ring failure of pitting defects. These peaks were compared with the theoretical fault characteristic frequency f. i The comparison of 99.04Hz indicates that the peaks in the figure are the main peaks in the local area near the theoretical fault characteristic frequency and harmonics. Therefore, it can be determined that the inner ring of the rolling bearing has failed under both strong and weak damage conditions.

[0116] Step 5: Introduce a dimensionless parameter fault factor to characterize the damage degree of the rolling bearing with inner ring defects. The definition diagram of the dimensionless parameter fault factor of the acoustic emission waveform signal in this embodiment is attached. Figure 8 As shown;

[0117] The specific process of step 5 is as follows:

[0118] Step 5.1: Based on the rotational speed conditions in this embodiment, substitute the corresponding rotational speed and bearing geometric parameters into formula (8) to obtain the theoretical values ​​of typical fault characteristics of rolling bearings. The theoretical values ​​of typical fault characteristic frequencies of rolling bearings in this embodiment are shown in Table 3.

[0119] When the outer ring is fixed, the formula for calculating the characteristic frequency of bearing inner ring failure is:

[0120]

[0121] In the formula, D b Let d be the diameter of the rolling element. m D is the inner diameter. m D is the outer diameter. c Let be the pitch circle diameter, and 2D c =d m +D m θ is the angular contact angle, and z is the number of rolling elements. i The characteristic frequency of bearing inner ring failure, in Hz, f r Reference axis rotational frequency;

[0122] Table 3 Theoretical values ​​of bearing fault characteristic frequencies at the same rotational speed.

[0123]

[0124] Step 5.2: Compare the fault information extracted in Step 4 with the theoretical fault characteristic frequency obtained in Step 5.1 to determine the power peak D corresponding to the 1st to 5th harmonics. i And calculate the average power within the frequency band between the 1st and 5th harmonic fault characteristic frequencies. And calculate the fault factor γ for each harmonic according to equation (9);

[0125]

[0126] Step 5.3: Determine the damage status of the bearing based on the fault factor values;

[0127] In this embodiment, the fault characteristic frequency, harmonic peak value, and nearby average peak value are shown in Table 4, and the fault factors of typical faulty bearings under different damage states are shown in Table 5.

[0128] Table 4. Fault characteristic frequencies and corresponding peak values ​​and nearby average peak values ​​for their harmonics

[0129]

[0130]

[0131] Table 5 Failure factors of typical bearings under different damage states

[0132]

[0133] The failure factors of typical bearings with different damage states were calculated using the data in Table 4 and formula (9), as shown in Table 5. The values ​​of the failure factors are used to characterize the damage state of different defects in the bearing. Since the damage state of wire cutting defects is much greater than that of pitting defects, wire cutting defects belong to the severe damage state, while pitting defects belong to the weak damage state.

[0134] Since no obvious fifth harmonic of the fault characteristic frequency appears in the acoustic emission signal envelope spectrum of the rolling element bearing with pitting defects, the average power value within the frequency band corresponding to the fifth harmonic of the fault characteristic frequency is equated to the average power value between the frequencies corresponding to its two adjacent peaks, i.e., the fault factor value is 1. Fault factor curves are plotted based on the fault factor data for different fault types and damage states in Table 5, as shown below. Figure 9 As shown.

[0135] After envelope spectrum analysis of acoustic emission signals from typical faulty bearings with wire cutting and pitting defects, the fault characteristic frequencies and the corresponding powers of their harmonics were extracted. The 1st to 5th harmonic fault factors of typical faulty bearings under different damage states were then calculated and analyzed. For example... Figure 9 It is evident that the failure factors for wire EDM defects on the outer ring, rolling elements, and inner ring are all greater than those for pitting defects, with the first-harmonic failure factor showing a particularly significant difference, approximately twice the value. Since wire EDM defects represent a severe damage state, while pitting defects represent a minor damage state, the failure factor, as an indicator of bearing damage status, can effectively distinguish different damage states of typical faulty bearings through its numerical differences, compensating for the inadequacy of energy-time evolution diagrams in identifying the degree of damage to the inner ring.

Claims

1. A method for identifying the degree of damage in rolling bearing faults based on acoustic emission multi-parameter fusion, characterized in that, include: Step 1: Collect the acoustic emission signal of the rolling bearing; Step 2: Use TAFI analysis to preliminarily determine whether the bearing has a fault; Step 3: If the bearing is initially determined to be faulty, the bearing fault is diagnosed by combining the time-to-sound emission parameter count, the amplitude distribution of the sound emission parameter impact number, and the time-to-sound emission parameter energy. If the bearing is determined to be faulty in the outer ring or the rolling element, the degree of damage is further determined based on the time-to-sound emission parameter energy. If the bearing is determined to be faulty in the inner ring, a fault factor needs to be introduced to characterize the degree of damage in the inner ring. If step 3 determines that the bearing inner ring is faulty, a fault factor needs to be introduced to characterize the degree of damage to the inner ring; specifically, it is described as follows: Step 4: Select the acoustic emission waveform stream signal during a period of stable rotation speed in the test, apply EMD decomposition signal to perform envelope detection processing on the high-frequency acoustic emission waveform stream signal, and use Hilbert transform to extract fault information; Step 5: Introduce a dimensionless parameter fault factor to characterize the damage degree of rolling bearings with inner ring defects; specifically, it is described as follows: Step 5.1: When the outer ring is fixed, the characteristic frequency of the bearing inner ring failure. The calculation formula is: (8) In the formula, The diameter of the rolling element, The inner diameter is... The outer diameter is... Let be the pitch circle diameter, and , Angular contact angle, The number of rolling elements. Reference axis rotational frequency; Based on the operating conditions, the corresponding speed and bearing geometric parameters are substituted into formula (8) to obtain the theoretical value of the bearing inner ring fault characteristic frequency; Step 5.2: Compare the fault information extracted in Step 4 with the theoretical fault characteristic frequency obtained in Step 5.1 to determine the power peak corresponding to the i-th multiple of the characteristic frequency. And calculate the average power within the frequency band between the (i-1)th and (i+1)th harmonics of the fault characteristic frequency. ,in The calculation does not include the power peak corresponding to the (i-1), i, and i+1 times the fault characteristic frequency. , and ; Step 5.3: Calculate the fault factor for each harmonic. The damage state of different defects in the bearing is characterized by the value of the failure factor. （9）。 2. The method for identifying the degree of damage in rolling bearing faults based on acoustic emission multi-parameter fusion according to claim 1, characterized in that, Step 3, combining the time-lapse graphs of acoustic emission parameter counts, the amplitude distribution graphs of acoustic emission parameter impacts, and the time-lapse graphs of acoustic emission parameter energy, is used to diagnose bearing faults. Specifically, it is described as follows: The counting parameters in the acoustic emission signal are divided into intervals according to a preset time interval, and the count values ​​corresponding to each time interval are summed to obtain the count value Q corresponding to each interval. An experience graph f1 is plotted on the count value Q and time t. Based on the plotted experience graph f1, the range of the counting intervals in which different bearing faults are located is determined. The impact number parameter in the acoustic emission signal was statistically analyzed using the distribution map analysis method. A distribution map f2 of impact number and amplitude was plotted. Based on the plotted distribution map f2, the impact number range of different bearing faults was determined. The energy parameters in the acoustic emission signal are divided into intervals according to a preset time interval, and the energy values ​​corresponding to each time interval are summed to obtain the energy value W for each interval. An experience graph f3 is plotted on the energy value W and time t. The degree of damage to the bearing outer ring and rolling elements is determined based on the plotted experience graph f3.

3. The method for identifying the degree of damage in rolling bearing faults based on acoustic emission multi-parameter fusion according to claim 1, characterized in that, Step 4 includes: Step 4.1: Determine the signal Identify all local extrema, then connect all the maxima and minima using cubic spline functions to form upper and lower envelopes, and calculate their average values. ,use minus get As a signal to be processed; (1) Step 4.2: Use formula (2) to process the signal. The first screening is performed to obtain an iterative signal. : (2) In the formula, Indicates will Connect all the maximum and minimum points in the curve to form the average curve of the upper and lower envelopes; After k rounds of filtering, the first fundamental modal component is obtained. ,remember ; (3) In the formula, This represents the signal from the (k-1)th iteration. Indicates will Connect all the maximum and minimum points in the curve to form the average curve of the upper and lower envelopes; Step 4.3: Use formula (4) to separate the basic modal components from the original signal; (4) Step 4.4: Put As the new original signal, n signal components are obtained sequentially: (5) In the formula, This represents the nth signal component. This represents the nth fundamental mode component; Finally Decomposed into: (6) Step 4.5: Perform envelope detection processing to obtain the envelope waveform, and use Hilbert transform to deenvelop the signal and extract fault information.

Images