Bearing fault measurement method combining bearing dynamics model and machine learning model

By combining the bearing dynamics model with the machine learning model, the vibration signals of various bearing components can be accurately distinguished, and real-time and high-precision fault diagnosis of the bearing can be achieved. This solves the problems of low fault diagnosis accuracy and poor adaptability in existing technologies and improves the operating reliability of rotating machinery.

CN120561711BActive Publication Date: 2025-10-17SOUTHEAST UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511052943.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-07-30
Publication Date
2025-10-17
Estimated Expiration
2045-07-30

AI Technical Summary

Technical Problem

Existing technologies have difficulty in accurately distinguishing the vibration signals of various bearing components, resulting in low fault diagnosis accuracy and poor adaptability under dynamic working conditions, making it impossible to achieve real-time monitoring and high-precision fault diagnosis.

Method used

Combining the bearing dynamics model with the machine learning model, signals are obtained through acceleration sensors and Hall sensors, filtered using fast Fourier transform and inverse fast Fourier transform, and combined with the bearing dynamics numerical solution model to calculate the motion frequency of each component. Finally, machine learning models such as CNN or SVM are used for fault identification.

Benefits of technology

It realizes real-time monitoring and high-precision fault diagnosis of various bearing components, improves the accuracy of fault location and type, saves manpower and financial resources, avoids unplanned downtime, and ensures the safe and stable operation of rotating machinery.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120561711B_ABST
    Figure CN120561711B_ABST
Patent Text Reader

Abstract

The application discloses a bearing fault measurement method combining a bearing dynamics model and a machine learning model. The specific steps are as follows: the vibration signal and the main shaft rotating speed signal of the bearing are monitored in real time through an acceleration sensor and a Hall sensor, the frequency components of the bearing vibration signal and the main shaft rotating speed signal are analyzed by using fast Fourier transform, and the bearing inner ring speed, the cage speed, the roller sliding speed, the roller rolling speed, the roller revolution and spin speed are calculated in combination with the bearing dynamics model. By comparing the speeds of the bearing components calculated with the frequencies analyzed by FFT, the actual vibration frequencies and the time domain vibration signals of the bearing components are distinguished, and finally the machine learning model is used for classifying and identifying the vibration signals. The vibration signals of the bearing sub-components are accurately distinguished, the real-time monitoring and high-precision fault measurement of the operation of each component of the bearing are realized, and the problems of low fault diagnosis precision and poor dynamic working condition adaptability of each component of the bearing system are solved.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rotating machinery condition monitoring, in particular to a rolling bearing fault diagnosis monitoring system and method. BACKGROUND

[0002] Bearing is a key component in rotating machinery equipment, and the safe and reliable operation of the bearing system is of great significance to the entire rotating equipment, and its operating state directly affects the performance and service life of the equipment.

[0003] The traditional bearing system operating state monitoring and fault diagnosis method mainly relies on vibration signal analysis and manual maintenance method, however, the traditional method is difficult to distinguish the vibration signals of each component of the bearing, and the fault diagnosis accuracy is low, and offline analysis is usually required, which cannot realize real-time monitoring; the manual experience maintenance method can solve the fault problem, but it needs to be stopped for maintenance, consumes a lot of manpower and financial resources, and cannot meet the real-time monitoring and intelligent demand.

[0004] In recent years, with the development of sensor technology and machine learning, the bearing fault diagnosis method has been significantly improved, however, the existing method is difficult to accurately separate the vibration signals of each component of the bearing, resulting in insufficient accuracy of fault diagnosis, and the adaptability in dynamic working condition is weak, and the diagnosis model cannot be adjusted in real time. SUMMARY

[0005] Technical problem: In order to overcome the shortcomings of the prior art, the present application provides a bearing fault measurement method combining bearing dynamics model and machine learning model, which solves the problems of low fault diagnosis precision of each component of the bearing system and poor adaptability in dynamic working condition, more accurately distinguishes the vibration signals of bearing sub-components, and realizes real-time monitoring and high-precision fault diagnosis of the bearing system.

[0006] Technical scheme: In order to achieve the above purpose, the technical scheme adopted by the bearing fault measurement method combining bearing dynamics model and machine learning model of the present application is:

[0007] Step 1: obtaining bearing vibration signals through an acceleration sensor and obtaining bearing spindle speed signals through a Hall sensor, to determine the basic parameters required for bearing dynamics model calculation;

[0008] Step 2: using fast Fourier transform to analyze the bearing vibration signal frequency and bearing spindle speed signal of step 1, to determine the real-time speed of the bearing spindle and the main frequency component in the vibration signal;

[0009] Step 3: Establish a bearing dynamics numerical solution model, and input the real-time rotating speed of the bearing spindle in step 2 into the bearing dynamics numerical solution model to calculate the rotating angular velocity of the bearing inner ring, the orbital angular velocity of the cage, the relative sliding velocity between the roller and the raceway, the linear velocity of the roller in the rolling direction, the orbital angular velocity of the roller, and the self-rotation angular velocity of the roller around its own axis;

[0010] Step 4: According to the comparison of the motion frequencies of the various motion sub-components of the bearing and the frequency components of the bearing vibration signal in step 2, the real-time rotating frequency of the bearing spindle, the real-time rotating frequency of the bearing inner ring, the real-time orbital frequency of the cage, the orbital frequency of the roller, the self-rotation frequency of the roller, and the sliding and rolling frequencies of the roller are determined.

[0011] Step 5: According to the motion frequencies of the bearing sub-components determined in step 4, a frequency error interval of f±n*f is designed with the motion frequency as the midpoint, the bearing vibration signal spectrum in step 2 is filtered, and the time-domain vibration signals of the bearing inner ring, the cage, the roller orbit, and the roller spin are accurately distinguished using inverse fast Fourier transform; 0<n<1.

[0012] Step 6: Using a machine learning model to classify and identify the time-domain vibration signals of the bearing inner ring, the cage, the roller orbit, and the roller spin obtained in step 5, the real-time monitoring and high-precision fault measurement of the operation of each component of the bearing are realized.

[0013] wherein,

[0014] In step 1, the determination of the basic parameters required for the bearing dynamics model calculation includes: bearing positioning and pressure preloading force, bearing spindle speed, bearing and bearing seat size, bearing and bearing seat material, lubricant performance parameters.

[0015] In step 2, the determination of the real-time rotating speed of the bearing spindle and the main frequency components in the vibration signal includes: real-time frequency of the bearing spindle, real-time frequency of the inner ring, real-time frequency of the cage, real-time frequency of the roller orbit and self-rotation.

[0016] In step 3, the establishment of the bearing dynamics numerical solution model includes: five-degree-of-freedom model of the bearing inner ring, force and torque balance equation set of the roller, differential equation of the cage dynamics, geometric deformation relationship between the roller and the raceway, bearing trackless control model, bearing grease lubrication model.

[0017] In step 4, the comparison of the frequency components of the bearing vibration signal is performed by inputting the real-time rotating speed of the spindle into the bearing dynamics model to calculate the frequencies of the bearing inner ring, the cage, and the roller orbit and self-rotation, sorting all frequency theoretical values and main frequencies in the vibration signal in ascending or descending order, and minimizing the sum of squares of the frequency difference between all frequency theoretical values and corresponding vibration signal frequencies.

[0018] In step 5, the bearing vibration signal spectrum filtering is to keep the frequency points unchanged within the frequency error interval, and set 0 outside the interval.

[0019] In step 6, the machine learning model is a convolutional neural network (CNN) or a support vector machine (SVM).

[0020] In the bearing dynamics numerical solution model, the bearing inner ring five-degree-of-freedom model is:

[0021]

[0022] In the formula, F x is the axial load, F y is the radial load along the y-axis direction, F z is the radial load along the z-axis direction, M y is the torque around the y-axis, M z is the torque around the z-axis, Z is the number of rollers, j is the jth roller, i is in contact with the inner raceway, Q ij is the inner contact load of the jth roller, a ij is the inner contact angle of the jth roller, F ij is the inner drag force of the jth roller, y j is the roller position angle, is the inner raceway curvature center trajectory radius, r i is the curvature radius of the inner raceway generatrix.

[0023] In the bearing dynamics numerical solution model, the roller force and torque balance equation set is:

[0024]

[0025] In the formula, e is in contact with the outer raceway, Q ej is the outer contact load of the jth roller, F ej is the outer drag force of the jth roller, a oj is the outer contact angle of the jth roller, F cj is the centrifugal force of the jth roller, f ej is the outer raceway friction force of the jth roller, f ij is the inner raceway friction force of the jth roller, M gj is the gyroscopic moment of the jth roller, D is the roller diameter.

[0026] In the bearing dynamics numerical solution model, the cage dynamics differential equation is:

[0027]

[0028] In the formula, F yjThe inertia force of the jth roller and the cage pocket along the y direction, F zj The inertia force of the jth roller and the cage pocket along the z direction, M xj The inertia moment of the jth roller and the cage pocket around the x direction, G cage The gravity of the cage, F cy And F cz The force of the cage in the y axis and the z axis generated by the oil film pressure, M cx The friction moment of the cage by the fluid dynamic oil film, Z is the number of rollers;

[0029] The geometric deformation relationship between the roller and the raceway in the bearing dynamics model is:

[0030]

[0031] In the formula, A 1j And A 2j The axial and radial distance of the curvature center of the inner and outer raceways of the jth roller after being loaded, BD is the initial distance of the curvature center of the inner and outer raceways, alpha 0 The initial contact angle, delta a The axial displacement of the curvature center of the inner raceway after being loaded, delta r The radial displacement of the curvature center of the inner raceway after being loaded, theta is the angular displacement of the inner ring after being loaded;

[0032] The control equation of the posture angle of the roller is introduced in the bearing trackless control model:

[0033]

[0034] In the formula, beta is the posture angle of the roller, M bi The spin friction moment of the roller in the inner raceway, M bo The spin friction moment of the roller in the outer raceway, gamma o The dimensionless geometric parameter of the roller and the outer raceway, gamma i The dimensionless geometric parameter of the roller and the inner raceway, gamma' is a dimensionless parameter,

[0035] Alpha i The inner contact angle, alpha o The outer contact angle;

[0036] The bearing grease lubrication model in the bearing dynamics model is:

[0037]

[0038] In the formula, tau is the shear stress of the lubricating grease in the contact ellipse, tau y The yield shear stress of the lubricating grease, The plastic viscosity of the lubricating grease, n is the rheological index, eta bη is the dynamic viscosity of the lubricating grease at the current temperature, η is the shear strain rate of the lubricating grease between the roller and the raceway, ref η is the dynamic viscosity of the lubricating grease at the reference temperature, T ref T is the reference temperature, f ref η is the viscosity-temperature parameter of the lubricating grease, T b T is the current temperature.

[0039] Beneficial effects: the application provides a bearing fault measurement method combining bearing dynamics model and machine learning model, compared with the prior art, the advantages are that: on the basis of the existing machine learning model, the vibration frequency of each sub-component of the bearing system is calculated in real time by combining the bearing dynamics analysis, the time domain vibration signals of the bearing main shaft, inner ring, rolling body and outer ring are accurately distinguished, the dynamic working condition of the bearing is tracked in real time based on multi-sensor analysis, the manpower and financial cost of manual maintenance is saved, the unplanned shutdown of rotating mechanical equipment is avoided, and the real-time state monitoring and fault diagnosis technology of rotating machinery has important significance. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The bearing fault measurement method combining the bearing dynamics model and the machine learning model provided by the application is a processing flow chart;

[0041] Figure 2 The bearing dynamics model calculation flow chart;

[0042] Figure 3 The machine learning model fault diagnosis flow chart. DETAILED DESCRIPTION

[0043] The following will be described in detail with reference to the accompanying Figures 1-3 The application will be further described in detail. It can be understood that the specific embodiments described herein are only used to explain the application, and not to limit the application.

[0044] Reference Figure 1 The application provides a bearing fault measurement method combining a bearing dynamics model and a machine learning model, which comprises the following steps:

[0045] Step 1: obtaining the bearing vibration signal through the acceleration sensor and the bearing main shaft rotating speed signal through the Hall sensor, and determining the basic parameters required for bearing dynamics model calculation;

[0046] Step 2: analyzing the bearing vibration signal spectrum and the bearing main shaft rotating speed signal of step 1 by using fast Fourier transform, and determining the main frequency component in the bearing main shaft real-time rotating speed and vibration signal;

[0047] Step 3: Establish a bearing dynamics numerical solution model, and substitute the real-time rotating speed of the main shaft in step 2 into the bearing dynamics numerical solution model to calculate the rotating angular velocity of the bearing inner ring, the orbital angular velocity of the cage, the relative sliding velocity between the roller and the raceway, the linear velocity of the roller in the rolling direction, the orbital angular velocity of the roller, and the self-rotation angular velocity of the roller around its own axis;

[0048] Step 4: According to the comparison of the motion frequencies of the various motion sub-components of the bearing in step 3 and the frequency components of the bearing vibration signal in step 2, the real-time rotating frequency of the bearing main shaft, the real-time rotating frequency of the bearing inner ring, the real-time orbital frequency of the cage, the orbital frequency of the roller, the self-rotation frequency of the roller, and the sliding and rolling frequencies of the roller are determined.

[0049] Step 5: According to the motion frequencies f of the bearing sub-components determined in step 4, a frequency error interval of f±n*f is designed with the motion frequency as the midpoint to filter the bearing vibration signal spectrum in step 2, and the inverse fast Fourier transform is used to accurately distinguish the time-domain vibration signals of the bearing inner ring, the cage, the roller orbit, and the roller self-rotation; 0<n<1.

[0050] Step 6: Using a machine learning model to classify and identify the time-domain vibration signals of the bearing inner ring, the cage, the roller orbit, and the roller self-rotation obtained in step 5, to realize real-time monitoring and high-precision fault diagnosis measurement of the operation of each component of the bearing.

[0051] The present application constructs a bearing fault diagnosis measurement method combining a bearing dynamics model and a machine learning model to solve the problem that bearing fault diagnosis technology cannot accurately distinguish the vibration signals of each component of the bearing, resulting in inaccurate fault positioning. The frequency analysis of the bearing vibration signal is compared with the motion frequencies of each component calculated by the bearing dynamics model, and then the vibration signals of each component of the bearing are accurately distinguished, and finally the time-domain and frequency-domain data of each component of the bearing are input into the machine learning model for fault diagnosis. Experimental verification shows that this method can realize real-time monitoring and high-precision fault diagnosis of each moving component of the bearing, solving the problems of low accuracy of fault location and fault type diagnosis and insufficient real-time monitoring of internal motion data in the existing technology in the field of bearing fault diagnosis. Compared with the prior art, the present application improves the accuracy of bearing fault location and fault type, thereby ensuring the safe and stable operation of the rotating equipment and improving production efficiency and enterprise competitiveness.

[0052] The following will be described in detail with examples of the present embodiment:

[0053] The present application proposes a bearing fault diagnosis measurement method combining a bearing dynamics model and a machine learning model, which can be implemented according to the following steps:

[0054] Step 1: Data acquisition stage;

[0055] Step 2: Data preprocessing stage;

[0056] Step 3: Solve the bearing dynamics numerical solution model to determine the theoretical motion frequency of each component of the bearing;

[0057] Step 4: Determine the real-time motion frequency of each component of the bearing;

[0058] Step 5: Determine the time domain vibration data of each component of the bearing;

[0059] Step 6: Use machine learning model to diagnose faults in time domain and frequency domain data of each component of the bearing;

[0060] The specific process is as follows:

[0061] (1) Data acquisition

[0062] A Hall sensor of model Honeywell SS49E is installed on the non-driving end of the 7005C angular contact ball bearing spindle, and an acceleration sensor of model PCB 352C3 is installed on the bearing seat. Real-time acquisition of bearing vibration signals and spindle speed signals is performed, and data is uploaded to the cloud server through a wireless transmission module of model XBee S3B.

[0063] (2) Data preprocessing

[0064] The vibration signal and spindle speed signal spectrum of step (1) are analyzed using fast Fourier transform to determine the real-time speed of the bearing spindle and the main frequency components in the vibration signal.

[0065] (3) Solve the bearing dynamics numerical solution model to determine the theoretical motion frequency of each component of the bearing

[0066] According to the basic parameters of the 7005C angular contact ball bearing (as shown in Table 1) and the bearing preload (F x ,F y ,F z ,M y ,M z ), a bearing dynamics numerical solution model is established, which can be simplified as formulas (1), (2), and (3). The real-time frequency of the spindle, the bearing basic parameters, and the bearing preload in step 2 are brought into Figure 2 The bearing dynamics model calculation process is iteratively solved to calculate the bearing inner ring speed, cage speed, roller sliding speed, roller rolling speed, roller revolution speed, and roller spin speed;

[0067] Table 1 Bearing parameters

[0068]

[0069] The formula of the established five-degree-of-freedom model of the 7005C angular contact ball bearing inner ring can be expressed as:

[0070]

[0071] The formula of the established 7005C angular contact ball bearing roller force and torque balance equation set can be expressed as:

[0072]

[0073] The formula of the established 7005C angular contact ball bearing cage dynamics differential equation can be expressed as:

[0074]

[0075] In the formula, F x is the axial load, F y is the radial load along the y-axis direction, F z is the radial load along the z-axis direction, M y is the torque around the y-axis, M z is the torque around the z-axis, Z is the number of rollers, j is the jth roller, i is in contact with the inner raceway, e is in contact with the outer raceway, Q ij is the inner contact load of the jth roller, α ij is the inner contact angle of the jth roller, F ij is the inner drag force suffered by the jth roller, ψ j is the roller position angle, is the inner raceway curvature center trajectory radius, r i is the curvature radius of the inner raceway generatrix, Q ej is the outer contact load of the jth roller, F ej is the outer drag force suffered by the jth roller, α oj is the outer contact angle of the jth roller, F cj is the centrifugal force suffered by the jth roller, f ej is the outer raceway friction force suffered by the jth roller, f ij is the inner raceway friction force suffered by the jth roller, M gj is the gyroscopic torque suffered by the jth roller, D is the roller diameter, F yj is the inertia force of the jth roller and the cage pocket along the y-axis direction, F zj is the inertia force of the jth roller and the cage pocket along the z-axis direction, M xj is the inertia torque of the jth roller and the cage pocket around the x-axis direction, G cage is the gravity of the cage, F cy and F cz are the components of the resultant force acting on the cage in the y-axis and z-axis directions generated by the oil film pressure, M cx is the friction torque of the cage by the hydrodynamic oil film.

[0076] (4) Determine the real-time motion frequency of each component of the bearing

[0077] According to the comparison of the motion frequency of the bearing inner ring, cage and roller in step (3) with the main frequency component of the bearing vibration signal in step (2), sort the corresponding descending order of all frequency theoretical values and main frequencies in the vibration signal, and determine the real-time frequency of the bearing spindle, the real-time frequency of the bearing inner ring, the real-time frequency of the cage, the revolution frequency of the roller, the spin frequency of the roller, the sliding and rolling frequency of the roller according to the minimum variance of all frequency theoretical values and corresponding vibration signal frequencies in numerical value;

[0078] (5) Determine the time domain vibration data of each component of the bearing

[0079] According to the motion frequency f of the bearing sub-components determined in step (4), design a frequency error interval of f±n×f with the motion frequency as the midpoint, filter the bearing vibration signal spectrum in step 2, and accurately distinguish the time domain vibration signals of the bearing inner ring, the cage, the roller revolution and the spin (0<n<1) using inverse fast Fourier transform;

[0080] (6) Use machine learning model to diagnose the time domain and frequency domain data of each component of the bearing

[0081] According to the time domain signals and frequency domain data of the bearing inner ring, the cage, the roller revolution and the spin obtained in step (5), use the support vector machine model for fault classification and identification, and realize real-time monitoring and high-precision fault diagnosis of each component of the bearing;

[0082] Wherein the fault feature input of the support vector machine model includes time domain features and frequency domain features,

[0083] Time domain features: mean, variance, peak value, kurtosis, impact pulse;

[0084] Frequency domain features: frequency band energy, main frequency, spectral entropy;

[0085] Preprocess the known fault data set, convert the fault category to a numerical label, use Gaussian radial basis function as the kernel function, divide the data set into training set and test set according to 7:3, input the feature data and corresponding label of the training set into the model for training, use the test result to evaluate the trained model, calculate the prediction accuracy of the model at different fault positions, and evaluate the performance of the model. Use the trained model to diagnose the fault of new data.

[0086] The application discloses a bearing fault diagnosis measurement method combining a bearing dynamics model and a machine learning model, compares and analyzes the frequency analysis of bearing vibration signals and the movement frequencies of each component calculated by the bearing dynamics model, and then accurately distinguishes the vibration signals of each component of the bearing, finally inputs the time domain and frequency domain data of each component of the bearing into the machine learning model for fault diagnosis. The application provides a new idea and method for early bearing fault diagnosis. The application solves the problems of low fault position and fault type diagnosis accuracy and insufficient real-time monitoring of internal movement data of the bearing in the prior art in the field of bearing fault diagnosis. Compared with the prior art, the application improves the accuracy of the bearing fault position and fault type, thereby guaranteeing the safe and stable operation of the rotating equipment, improving the production efficiency and enterprise competitiveness.

[0087] It should be noted that the above merely describes the preferred embodiments of the present application. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and various obvious changes and substitutions can be made by those skilled in the art without departing from the scope of the present application. Therefore, although the implementation of the present application is described in detail from the above technical solutions, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the claims attached hereto.

Claims

1. A bearing fault measurement method combining a bearing dynamics model with a machine learning model, characterized in that: The following steps are involved: Step 1: Obtain the bearing vibration signal through the acceleration sensor and the bearing spindle speed signal through the Hall sensor to determine the basic parameters required for the bearing dynamics model calculation; Step 2: Use fast Fourier transform to analyze the bearing vibration signal frequency and the bearing spindle speed signal in step 1 to determine the real-time speed of the bearing spindle and the main frequency components in the vibration signal; Including: bearing spindle real-time frequency, inner ring real-time frequency, cage real-time frequency, roller revolution and spin real-time frequency; Step 3: Establish a numerical solution model for bearing dynamics. Substitute the real-time speed of the bearing spindle in step 2 into the numerical solution model for bearing dynamics to calculate the angular velocity of the bearing inner ring, the orbital angular velocity of the cage, the relative sliding velocity between the roller and the raceway, the linear velocity of the roller along the rolling direction, the orbital angular velocity of the roller, and the angular velocity of the roller spinning around its own axis. Step 4: Based on the comparison of the motion frequencies of the various moving sub-components of the bearing and the frequency components of the bearing vibration signal in step 2, determine the real-time rotation frequency of the bearing main shaft, the real-time rotation frequency of the bearing inner ring, the real-time revolution frequency of the cage, the revolution frequency of the roller, the roller spin frequency, and the roller sliding and rolling frequency; Step 5: Based on the motion frequency f of the bearing sub-component determined in step 4, a frequency error range of f±n*f is designed with the motion frequency as the midpoint to filter the bearing vibration signal spectrum in step 2, and the inverse fast Fourier transform is used to accurately distinguish the time domain vibration signals of the bearing inner ring, cage, and roller revolution and spin; <n<1; Step 6: Use the machine learning model to classify and identify faults in the time domain vibration signals of the bearing inner ring, cage, and roller revolution and spin obtained in Step 5, thereby achieving real-time monitoring of the operation of each bearing component and high-precision fault measurement.

2. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 1 is characterized in that: In step 1, the basic parameters required for the calculation of the bearing dynamics model are determined, including: bearing positioning constant pressure preload, bearing spindle speed, bearing and bearing seat dimensions, bearing and bearing seat materials, and lubricant performance parameters.

3. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 1 is characterized in that: In step 3, the numerical solution model of bearing dynamics is established, including: a five-degree-of-freedom model of the bearing inner ring, a group of roller force and torque balance equations, a cage dynamic differential equation, a roller and raceway geometric deformation relationship, a bearing trackless control model, and a bearing grease lubrication model.

4. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 1, characterized in that: The frequency components of the bearing vibration signal in step 4 are compared by inputting the real-time spindle speed into the bearing dynamics model to calculate the bearing inner ring frequency, cage frequency, roller revolution and spin frequency, and sorting all frequency theoretical values ​​and the main frequencies in the vibration signal in ascending or descending order, and the sum of the squares of the differences between all frequency theoretical values ​​and the corresponding vibration signal frequencies is minimized in terms of numerical value.

5. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 1 is characterized in that: In step 5, the spectrum filtering of the bearing vibration signal is to keep the frequency points unchanged within the frequency error interval and to set the frequency points outside the interval to 0.

6. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 1, characterized in that: In step 6, the machine learning model is a convolutional neural network (CNN) or a support vector machine (SVM).

7. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 3 is characterized in that: In the bearing dynamics numerical solution model, the five-degree-of-freedom model of the bearing inner ring is: Where F x is the axial load, F y is the radial load along the y-axis, F z is the radial load along the z-axis, M y is the moment about the y-axis, M z is the moment around the z axis, Z is the number of rollers, j is the jth roller, i is in contact with the inner raceway, Q ij is the internal contact load of the jth roller, α ij is the internal contact angle of the jth roller, F ij is the internal drag force on the jth roller, ψ j is the roller position angle, is the radius of the inner raceway curvature center trajectory, r i is the curvature radius of the inner raceway generatrix.

8. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 3 or 7, characterized in that: In the bearing dynamics numerical solution model, the roller force and torque balance equations are: Where e is the contact with the outer raceway, Q ej is the external contact load of the jth roller, F ej is the external drag force on the jth roller, α oj is the external contact angle of the jth roller, F cj is the centrifugal force on the jth roller, f ej is the outer raceway friction force on the jth roller, f ij is the inner raceway friction force on the jth roller, M gj is the gyroscopic torque acting on the jth roller, and D is the roller diameter.

9. The bearing fault measurement method combining a bearing dynamics model and a machine learning model according to claim 1, characterized in that: In the bearing dynamics numerical solution model, the cage dynamics differential equation is: Where F yj is the inertia force between the jth roller and the cage pocket along the y direction, F zj is the inertia force between the jth roller and the cage pocket along the z direction, M xj is the moment of inertia of the jth roller and cage pocket around the x direction, G cage is the weight of the cage, F cy and F cz M is the component of the resultant force on the cage generated by the oil film pressure in the y-axis and z-axis. cx is the friction torque of the fluid dynamic oil film on the cage, and Z is the number of rollers; The geometric deformation relationship between the roller and raceway in the bearing dynamics model is: Where A 1j and A 2j is the axial and radial distance between the inner and outer raceway curvature centers of the jth roller after loading, BD is the initial distance between the inner and outer raceway curvature centers, α 0 is the initial contact angle, δ a is the axial displacement of the inner raceway curvature center after loading, δ r is the radial displacement of the center of curvature of the inner raceway after loading, and θ is the angular displacement of the inner ring after loading; The control equation of the roller attitude angle is introduced into the bearing trackless control model: Where β is the roller attitude angle, M bi M is the spin friction torque of the roller in the inner raceway, bo is the spin friction torque of the roller on the outer raceway, γ o is the dimensionless geometric parameter of the roller and outer raceway, γ i is the dimensionless geometric parameter of the roller and inner raceway, γ′ is a dimensionless parameter, α i is the internal contact angle, α o is the external contact angle; The bearing grease lubrication model in the bearing dynamics model is: Where τ is the shear stress on the grease in the contact ellipse, τ y is the yield shear stress of the grease, is the plastic viscosity of the grease, n is the rheological index, η b is the dynamic viscosity of the grease at the current temperature, is the shear strain rate of the grease between the roller and the raceway, η ref is the dynamic viscosity of the grease at the reference temperature, T ref is the reference temperature, f ref is the viscosity-temperature parameter of the grease, T b is the current temperature.

Citation Information

Patent Citations

  • A method for vibration simulation of aero-engine intermediate bear with local damage fault

    CN109145501A

  • High-rotating speed rolling bearing-rotor system local damage fault simulation method

    CN110674585A