Rolling bearing service life prediction method based on hybrid sensor monitoring characteristics and physical damage model
By combining monitoring data with physical damage models and utilizing Hertz contact theory and wavelet envelope spectrum eigenvalue monitoring, the inaccuracy of rolling bearing life assessment in existing technologies has been solved, enabling accurate prediction of rolling bearing operating status and remaining life, and providing an online monitoring technology approach for rolling bearing failures.
Patent Information
- Application Number
- CN202510615606.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-14
- Publication Date
- 2025-11-07
AI Technical Summary
In the current technology for assessing the service life of rolling bearings, data-driven methods require a large amount of experimental data and are detached from mechanical mechanisms, while methods based on physical models are difficult to simulate crack damage evolution, resulting in inaccurate assessments and poor generalizability.
By combining monitoring data with physical damage models, contact stress is calculated using Hertz contact theory to establish a damage mechanics model. Early damage is monitored by combining wavelet envelope spectrum eigenvalues of vibration data, and the bearing operating status and remaining life are comprehensively evaluated.
It enables accurate assessment of the operating status of rolling bearings and prediction of their remaining life, provides an online monitoring technology for rolling bearing failures, detects early failures in advance, and ensures safe operation of equipment.
Smart Images
Figure CN120907832A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of rotating machinery fault diagnosis, and particularly relates to a bearing operation life evaluation method combining monitoring data and a physical damage model. BACKGROUND
[0002] Rolling bearings are one of the most common components in rotating machinery equipment. In some cases, they are the core and key components of the machinery equipment, for example, once the main bearing in an aero-engine fails, it will cause great harm to the safety of the airplane operation. Therefore, evaluating the operation life of the rolling bearing is a prerequisite for ensuring the safe and good operation of the equipment.
[0003] At present, various methods are applied to the evaluation of the operation life of the bearing, mainly focusing on data-driven methods based on observation data and physical model-based methods. However, the data-driven method needs a large amount of test data and training model, and it is divorced from the mechanical mechanism of the actual failure of the rolling bearing, and the simulation of the failure process does not have good generalizability. The physical model-based method is very difficult to observe the evolution and expansion of the crack damage of the rolling bearing in the actual process, and it is also very difficult to link the crack expansion condition with the vibration signal. Therefore, the application provides a bearing operation life evaluation method combining monitoring data and a physical damage model, which is based on the observation data obtained by the vibration sensor to evaluate the operation state of the bearing in real time, and is used for the correction of the physical model. The damage mechanics physical model is used to simulate the damage evolution of the bearing, and finally the residual life of the rolling bearing is predicted by combining the respective advantages of the two. An important technical approach is provided for realizing the operation state and life evaluation of the rolling bearing. SUMMARY
[0004] The purpose of the application is to provide a bearing operation life evaluation method combining monitoring data and a physical damage model. The method can accurately and effectively evaluate the operation state and residual life of the bearing.
[0005] The technical scheme of the application is:
[0006] characterized in that it comprises:
[0007] Step 1: Based on the rolling bearing operation condition, the contact stress suffered by the rolling bearing during operation is calculated based on the Hertz contact theory, and the damage evolution of the bearing is simulated based on the damage mechanics model to estimate the operation life of the bearing when the initial damage occurs, which is denoted as T0, T0 is the number of cycles.
[0008] Hertz contact theory is well known in the art, which is the basic theory to study the contact mechanics behavior of elastic solids; from Hertz contact theory, according to the load of the bearing, the stress distribution P of the two-dimensional contact area of the bearing rolling body and raceway, the contact half-width b can be calculated, and a two-dimensional finite element model with a length of 10b and a depth of 6b is established to simulate the bearing damage evolution process. The damage accumulation equation of the bearing is:
[0009]
[0010] where Δτ represents the change range of the shear stress suffered by the material; D is the damage degree of the material unit, N is the cycle number; the material parameters m and σ r are related to the bearing steel material; according to the test (reference: Shigeo, Shimizu, Kazuo, et al. Probabilistic Stress-Life (P-S-N) Study on Bearing Steel Using Alternating Torsion Life Test [J]. Tribology Transactions, 2009, 52(6): 807-816), m = 11.1 and σ r = 5979 MPa can be obtained. d is the differential symbol; assuming that the damage increment ΔD of the bearing is constant, then its cycle life ΔN k under each damage increment can be represented as:
[0011]
[0012] where subscript i, crit represents the maximum value of dD / dN of the bearing material in the i-th cycle block;
[0013] The damage degree of the bearing in the new cycle is updated as:
[0014]
[0015] where j represents the number of each unit in the model; is the damage degree increment of the j-th unit; at the beginning of the next cycle, then:
[0016]
[0017] represents the damage degree of the j-th unit in the i-th cycle, represents the damage degree of the j-th unit at the beginning of the i+1-th cycle, is the elastic modulus of the j-th unit in the i-th cycle, represents the elastic modulus of the j-th unit at the beginning of the i+1-th cycle.
[0018] When the damage degree D of a material unit in the bearing model accumulates to 1, the initial damage of the bearing occurs, and the cycle number at this time is recorded as N0, which is substituted into the running speed to obtain T0, and the specific calculation formula is as follows:
[0019] T0=N0 / (V×Z)
[0020] Wherein, V is the bearing speed, and Z is the number of rolling elements.
[0021] Step 2: Collect vibration data during the operation of the bearing, obtain the effective value trend from the vibration data, and the calculation method of the effective value is well known in the art, that is, the root mean square value of the data; and construct a wavelet envelope spectrum characteristic value to monitor the running state. Extract the first significant change greater than the abnormal threshold point, recorded as the early damage occurrence time T1. The effective value of the vibration data is the root mean square value of the data, which is a well-known concept to those skilled in the art, and will not be described here.
[0022] The vibration observation signal is decomposed by 5 layers using db8 wavelet basis, and 5 detail signals and 1 reconstruction signal are obtained after decomposition. The envelope spectrum analysis is performed on the above 6 signals respectively, and 6 envelope spectra are obtained. Assuming that the fault characteristic frequency of the bearing is f faul , there are characteristic spectrum peaks near each order harmonic frequency, the envelope spectrum analysis bandwidth is f e , the envelope spectrum is W(f), and the number of W(f) spectrum lines is N e , then the average value S ea of the envelope spectrum is:
[0023]
[0024] Let S ed be the average value of the spectrum lines at each order harmonic frequency of the fault characteristic frequency in the envelope spectrum, and the number of spectrum lines of the fault frequency in the envelope spectrum is n e , then
[0025]
[0026] A dimensionless time-frequency feature can be constructed:
[0027]
[0028] The obtained ΔS e is the constructed wavelet envelope spectrum characteristic value, and its size reflects whether there is a significant bearing fault characteristic in each decomposition signal. In order to avoid the error between the theoretical fault characteristic frequency and the actual fault characteristic frequency, the search range of the fault characteristic frequency is ±5Hz in the calculation process.
[0029] The early failure threshold S of the observation data is determined by using 3σ criterion, wherein σ is the variance of the effective value of the vibration data. When the effective value reaches S, it indicates that the bearing has early abnormality, and the specific formula is:
[0030] S = μ + 3σ (8)
[0031] Wherein, μ is the average effective value of the bearing vibration observation signal in the normal running state, and σ is the variance of the effective value of the vibration data.
[0032] Step 3: The starting damage time is determined by comprehensively comparing the observation data and the bearing damage physical model. If T0≤T1, it indicates that the early abnormal state of the bearing cannot be found from the vibration observation signal, and the initial damage of the bearing has been indicated based on the damage mechanics physical model at the T0 moment. In order to ensure the safe operation of the equipment, the residual life of the bearing is predicted from the T0 moment, the damage evolution process of the bearing is simulated, and the surface spalling is formed. The spalling size width 2b is set as the failure threshold, that is, when the damage size width is greater than 2b, the failure state is reached. The time from the T0 moment to the failure prediction is the residual running life of the bearing, which is recorded as T d . The complete running life of the bearing is T0+T d . If T0>T1, it indicates that the early failure of the bearing is found in the vibration observation signal. At this time, although the damage has not occurred in the physical damage model, in order to ensure the safe operation of the equipment, the initial damage of the bearing is adjusted to the physical model, and the residual life is predicted. The time from the T1 moment to the failure prediction is the residual running life of the bearing, which is recorded as T v . The complete running life of the bearing is T1+T v .
[0033] Step 4: When the effective value reaches 5 times of the normal running state, the bearing stops running and is judged as failure. The prediction result is compared with the actual running life, wherein the effective value is the root mean square value of the vibration data.
[0034] In at least one embodiment of the present application, the present application is applied to the full life fatigue test of the rolling bearing and the effectiveness and accuracy are verified. The present application has at least the following beneficial technical effects:
[0035] The bearing running life evaluation method provided by the present application can effectively evaluate the running state of the rolling bearing and predict the residual life thereof, and provides an important technical approach for realizing the online monitoring of the rolling bearing failure. BRIEF DESCRIPTION OF DRAWINGS
[0036] The above and / or other aspects of the present application will become apparent from the following detailed description of the application taken in conjunction with the accompanying drawings.
[0037] Figure 1 is a specific scheme flow chart of the present application.
[0038] Figure 2a is a physical diagram of ABLT-1A type fatigue testing machine.
[0039] Figure 2b is a schematic diagram of ABLT-1A type loading system.
[0040] Figure 3 is a bearing after test with flaking of inner ring Figure 4 is the change of effective value during the whole life test of 6206 bearing Figure 5a is the change trend of wavelet envelope characteristic value of inner ring Figure 5b is the change trend of wavelet envelope characteristic value of outer ring Figure 6 is the prediction result of initial damage occurrence of bearing based on physical model Figure 7 is the prediction result of residual life of bearing based on physical model DETAILED DESCRIPTION
[0041] As shown in Figure 1 , the present application provides a bearing operation life evaluation method combining monitoring data and physical damage model, comprising the following steps:
[0042] Step 1: based on Hertz contact theory, the contact stress suffered by the rolling bearing during operation is calculated from the rolling bearing operation condition, and the damage evolution simulation of the bearing is carried out based on damage mechanics model, so as to estimate the operation life when the initial damage of the bearing occurs, which is recorded as T0.
[0043] Step 2: the vibration data during the operation of the bearing are collected, the effective value trend thereof is obtained from the vibration data, and the wavelet envelope spectrum characteristic value is constructed to monitor the operation state. The first significant change of the characteristic value greater than the abnormal threshold point is extracted, which is recorded as the early damage occurrence time T1.
[0044] Step 3: the initial damage time is determined by comprehensively comparing the observation data and the physical model of bearing damage. If T0≤T1, it indicates that the early abnormal state of the bearing cannot be found from the vibration observation signal, while the physical model based on damage mechanics at T0 time has shown that the initial damage of the bearing has occurred. In order to ensure the safe operation of the equipment, the residual life prediction of the bearing is started from T0 time, the damage evolution process of the bearing is simulated until the surface flaking is formed, and the flaking size width 2b is set as the prediction failure threshold. The time from T0 time to failure prediction is the residual operation life of the bearing, which is recorded as T d . The complete operation life of the bearing is: T0+Td If T0>T1, it means that the early failure of the bearing is found in the vibration observation signal, at this time, although the damage has not occurred in the physical damage model, in order to ensure the safety of the equipment operation, the physical model is adjusted to the initial damage of the bearing, and the remaining life prediction is carried out. The time from T1 to the failure prediction is the remaining running life of the bearing, which is recorded as T v The complete running life of the bearing is T1+T v .
[0045] Step 4: When the vibration effective value reaches 5 times of the normal running state, the bearing stops running, and is judged as failure. Compare the prediction result with the actual running life.
[0046] Step 1 includes: according to the Hertz contact theory, the stress distribution P of the two-dimensional contact area of the bearing rolling body and the raceway, the contact half-width b can be calculated according to the load borne by the bearing, and a two-dimensional finite element model with a length of 10b and a depth of 6b is established to simulate the damage evolution process of the bearing. The damage accumulation equation of the bearing is:
[0047]
[0048] Where Δτ represents the change range of the shear stress borne by the material; D is the damage degree of the material unit, N is the cycle number; Material parameters m and σ r are related to the bearing steel material; According to the test (reference: Shigeo, Shimizu, Kazuo, et al. Probabilistic Stress-Life (P-S-N) Study on Bearing Steel Using Alternating Torsion Life Test [J]. Tribology Transactions, 2009, 52(6): 807-816), m=11.1, σ r =5979MPa. d is the differential symbol; Assuming that the damage increment ΔD of the bearing is constant, the cycle life ΔN k of each damage increment can be represented as:
[0049]
[0050] Where subscript i, crit represents the maximum value of dD / dN of the bearing material in the ith cycle block;
[0051] The damage degree of the bearing in the new cycle is updated as:
[0052]
[0053] Where j represents the number of each unit in the model; is the damage increment of the jth element, and at the beginning of the next cycle, we have:
[0054]
[0055] is the damage of the jth element in the ith cycle, is the damage of the jth element at the beginning of the (i+1)th cycle, is the elastic modulus of the jth element in the ith cycle, is the elastic modulus of the jth element at the beginning of the (i+1)th cycle.
[0056] When the damage D of a material element in the bearing model accumulates to 1, the initial damage of the bearing occurs, and the cycle number at this time is recorded as N0, which can be converted to T0by substituting the running speed.
[0057] In step 2: use db8 wavelet basis to decompose the vibration observation signal for 5 layers, and get 5 detail signals and 1 reconstruction signal. The envelope spectrum analysis is performed on the above 6 signals respectively, and 6 envelope spectra are obtained. Assuming that the fault characteristic frequency of the bearing is f faul , and there are characteristic spectrum peaks near each order frequency, assuming that the envelope spectrum analysis bandwidth is f e , the envelope spectrum is W(f), and the number of W(f) spectrum lines is N e , then the average value S ea of the envelope spectrum is:
[0058]
[0059] Let S ed be the average value of the spectrum lines at each order frequency of the fault characteristic frequency in the envelope spectrum, and let n e be the number of spectrum lines of the fault frequency in the envelope spectrum, then
[0060]
[0061] A dimensionless time-frequency feature can be constructed:
[0062]
[0063] The obtained ΔS e is the constructed wavelet envelope spectrum feature value, and its size reflects whether there is a significant bearing fault feature in each decomposition signal. To avoid errors between the theoretical fault characteristic frequency and the actual fault characteristic frequency, the search range near the fault characteristic frequency is ±5 Hz during calculation.
[0064] The specific formula of the early anomaly threshold 3σ criterion is:
[0065] S = μ + 3σ (8)
[0066] Wherein, μ is the average effective value of bearing vibration observation signal under normal operation state, and σ is the variance of vibration data effective value.
[0067] Embodiment:
[0068] The application adopts 6206 type single-row deep groove ball rolling bearing to carry out accelerated fatigue test verification. The test relies on ABLT-1A type fatigue testing machine, can carry out rolling bearing full life fatigue test and rolling bearing fault simulation test, and is suitable for high speed condition, and the speed can reach 12000r / min. Fig. 2 is a schematic diagram of ABLT-1A type fatigue testing machine and the loading mode of the tested bearing. Its loading mode is hydraulic loading, which converts the weight of the weight disc into hydraulic pressure to load the bearing in the radial and axial directions. The maximum radial load capacity is 30kN, and the axial load capacity is 10kN. The specific parameters of 6206 rolling bearing are shown in table 1:
[0069] Table 1
[0070]
[0071] In order to accelerate the bearing failure, and make the rolling bearing speed close to the real aero-engine speed, the test speed is 11500r / min, 13kN radial load acts on the weight, and the hydraulic pressure is transmitted to the test head. As shown in Fig. 2, each bearing is subjected to a radial load of 6.25kN. The sampling frequency in the test is 50kHz, and a sample is stored every 0.1 hour. The length of each sample is 32000. After 36 hours of operation, the bearing test is stopped because the effective value is greater than 5 times the effective value during normal operation, and serious spalling is found in the inner ring. At this time, the whole cycle life of the bearing is 2.23x10 8 . Figure 3 The inner ring spalling bearing after the test.
[0072] Figure 4 The effective value change in the process of this bearing full life test, Figure 5a The change of the inner ring wavelet envelope characteristic value. Figure 5b The change of the outer ring wavelet envelope characteristic value. As can be seen from the above images, the vibration monitoring method based on the effective value has a significant change at about 30 hours, while the inner ring wavelet characteristic value has a significant rise at 13 hours (T1 = 8.1x10 7 ), and the outer ring wavelet characteristic value has no obvious change trend in the whole life period, which shows that the observation data monitoring method based on wavelet characteristic value proposed in the application can find the early failure of the bearing more in advance compared with the traditional effective value.
[0073] Figure 6For the prediction result of initial damage occurrence of bearing based on the physical model of damage mechanics, it can be seen that material units under the contact subsurface of bearing have occurred damage (red represents damage degree D=1), and the prediction cycle number of initial damage occurrence of bearing is T0=9.4×10 7 Compared with the wavelet characteristic value, it is slightly lagged, so the remaining life prediction is selected from T1. The physical model of damage mechanics is adjusted, T1 is taken as the initial damage occurrence time of bearing, and the damage degree D of the maximum damage degree unit in the model at this time is assigned to 1, and the final spalling formation result is shown in Figure 7 At this time, the subsurface and the obvious propagation crack are formed, and extend to the surface, and the model is distorted. The crack width is 2.1b, and the predicted remaining cycle life is T v =7.3×10 7 Therefore, the whole running life of bearing is 1.54×10 8 Although it is less than the test result 2.23×10 8 However, in fact, the spalling width obtained by the test is much larger than 2.1b, and the bearing damage state determined by the effective value threshold has a lag, and after the initial surface spalling formation, it also experiences some running time until the test is stopped, so the error of the bearing life evaluation result provided by the present application and the test result also includes the cycle number of wear expansion on the raceway after the surface spalling formation. Overall, the results are in the same order of magnitude range, so the method of the present application can effectively master the fault information of bearing in advance, and provides a safe and reliable technical approach for online monitoring of rolling bearing failure.
[0074] The present application provides a bearing running life evaluation method combining monitoring data and physical damage model, and there are many methods and ways to realize the technical scheme, and the above description is only the preferred embodiment of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the principle of the present application, some improvements and refinements can be made, for example, different state index methods are used to evaluate early fault information, and these improvements and refinements should be regarded as the protection scope of the present application. The components not explicitly described in the embodiment can be realized by the existing technology.
Claims
1. A method for assessing the operational lifetime of a bearing in combination with a data and physical damage model, characterized in that Comprise: Step 1: from the rolling bearing operating condition, based on Hertz contact theory, the contact stress of the rolling bearing is calculated, and the damage evolution simulation of the bearing is carried out based on the damage accumulation equation of the bearing, the running life of the bearing when the initial damage occurs is calculated, recorded as T0; Step 2: collect the vibration data of the bearing during operation, obtain the effective value trend from the vibration data, and construct the wavelet envelope spectrum characteristic value to monitor the running state, extract the characteristic value of the first significant change greater than the abnormal threshold point, recorded as the early damage occurrence time T1; Step 3: compare T0 and T1, if T0≤T1, then from T0, the residual life of the bearing is predicted, otherwise if T0>T1, then from T1, the residual life of the bearing is predicted, and the life prediction method adopts the damage accumulation equation based on the bearing; Step 4: when the vibration effective value reaches 5 times of the normal running state, the bearing stops running, and the prediction result is compared with the actual running life result.
2. A method of assessing the operational lifetime of a bearing in combination with a data and physical damage model according to claim 1, characterized in that, In step 1, according to the load borne by the bearing, the stress distribution P of the two-dimensional contact area of the bearing rolling body and raceway, the contact half-width b are calculated based on Hertz contact theory, a two-dimensional finite element model with a length of 10b and a depth of 6b is established to simulate the damage evolution process of the bearing. The damage accumulation equation of the bearing is: where Δτ represents the range of shear stress variation experienced by the material; D is the damage level of the material element, N is the number of cycles; material parameters m and σ r are related to the bearing steel material, m = 11.1, σ r = 5979 MPa, d is the differential operator; assuming that the damage increment ΔD of the bearing is constant, its cycle life ΔN k under each damage increment can be represented as: Wherein, subscript i, crit represents the maximum value of dD / dN of the bearing material in the i th cycle; The damage degree of the bearing in the new cycle is updated as: where j denotes the number of each cell in the model; is the damage increment for the jth cell in the ith cycle; at the beginning of the next cycle, then, we have: denotes the damage degree of the jth cell in the ith cycle, denotes the damage degree of the jth cell at the beginning of the i+1 cycle, is the elastic modulus of the jth cell in the ith cycle, denotes the elastic modulus of the jth cell at the beginning of the i+1 cycle, When the damage degree D of a material unit in the bearing model accumulates to 1, the initial damage of the bearing occurs, and the cycle number at this time is recorded as N0, which is substituted into the running speed to convert the running life T of the bearing at the initial damage occurrence 0, , and the specific calculation formula is as follows: T0=N0 / (V×Z) Wherein, V is the speed of the bearing, Z is the number of rolling bodies.
3. The method of claim 2, wherein, In step 2, the vibration observation signal is decomposed by 5 layers using db8 wavelet base, 5 detail signals and 1 reconstruction signal are obtained, envelope spectrum analysis is performed on the above 6 signals respectively, 6 envelope spectrums are obtained, assuming that the fault characteristic frequency of the bearing is f faul , there are characteristic spectrum peaks near each order frequency, assuming that the envelope spectrum analysis bandwidth is f e , the envelope spectrum is W(f), assuming that the number of W(f) spectrum lines is N e , then the average value S ea of the envelope spectrum is: Let S ed Let the average value of the spectrum line at each order of the fault characteristic frequency in the envelope spectrum be S e Then A dimensionless time-frequency feature is constructed: The obtained ΔS e That is, the constructed wavelet envelope spectrum characteristic value, the size reflects whether there is a bearing significant fault feature in each decomposition signal. In the calculation process, the search range of the fault feature frequency is ± 5 Hz. The early failure threshold of the observation data is S, when the effective value reaches S, it indicates that the bearing has early abnormality, the specific formula is: S=μ+3σ (8) Wherein, μ is the average effective value of the bearing vibration observation signal under normal running state, and σ is the variance of the effective value of the vibration data.
4. The method of claim 3, wherein, In step 3, the starting damage time is determined by comprehensively comparing the observation data and the physical model of the bearing damage. If T0≤T1, it means that the early abnormal state of the bearing cannot be found from the vibration observation signal, but the physical model based on damage mechanics at T0 time has shown that the bearing has occurred initial damage, in order to ensure the safe operation of the equipment, the residual life of the bearing is predicted from T0 time, the damage evolution process of the bearing is simulated until the surface spalling is formed, and the spalling size width 2b is set as the failure threshold for prediction; the time from T0 time to failure prediction is the residual operating life of the bearing, recorded as T d . The complete operating life of the bearing is: T0+T d . If T0>T1, it means that the early failure of the bearing is found in the vibration observation signal, at this time, although the damage has not occurred in the physical damage model, in order to ensure the safety of the equipment operation, the physical model is adjusted to the initial damage of the bearing, and the remaining life prediction is carried out, and the time from T1 to the failure prediction is the remaining operation life of the bearing, which is recorded as T v , the complete operation life of the bearing is T1+T v .