Full-life rolling bearing fault diagnosis method based on mechanism data migration

Abstract

The invention relates to the technical field of rolling bearing fault diagnosis, and discloses a full-life rolling bearing fault diagnosis method based on mechanism data migration. The method comprises the steps that a rolling body radial displacement excitation function, the time-varying contact rigidity of a rolling body and a bearing raceway and the impact force at the contact defect position of the rolling body in a mass-spring-damping model are corrected based on the Hertz contact theory, and a nonlinear kinetic equation of the rolling bearing is obtained; solving a nonlinear kinetic equation of the rolling bearing through a fourth-order Runge-Kutta method, generating source domain data consistent with test parameters of the full-life rolling bearing, and constructing a source domain data set with a complete fault tag; according to the method, the dynamic mechanism model consistent with the full-life test bearing parameters is constructed, the feature distribution difference between the generated source domain data and the target domain data is small, a foundation is laid for efficient knowledge migration, and the problem that traditional source domain data is not matched with actual data is solved.

Description

Technical Field

[0001] This invention relates to the field of rolling bearing fault diagnosis technology, and in particular to a full-life rolling bearing fault diagnosis method based on mechanism data migration. Background Technology

[0002] Rolling bearings are core components of rotating machinery, and their operating condition directly determines equipment reliability. Approximately 30% of rotating machinery failures are caused by rolling bearing failure. In full-life testing scenarios, the data from the bearing's normal operation to failure contains key fault evolution information. However, in practical applications, there are two major problems: First, fault data throughout the entire life cycle is scarce or even missing, especially early fault samples, resulting in poor training performance of data-driven models. Second, data collected in the field is affected by noise and fluctuations in operating conditions, resulting in weak characteristic signals, making it difficult for traditional diagnostic methods to effectively extract fault information.

[0003] In existing technologies, fault diagnosis methods based on transfer learning attempt to address the problem of missing samples by transferring knowledge from source domain data. However, they have significant drawbacks: Firstly, source domain data often relies on laboratory simulations, which differ greatly in feature distribution from actual life-cycle data, limiting the transfer effect. Secondly, the loss function design is often simplistic, failing to fully utilize the complementary advantages of multiple loss functions and neglecting targeted feature optimization for noisy target domain data, resulting in diagnostic accuracy that falls short of engineering requirements. Furthermore, the parameters of traditional VMD (Variational Mode Decomposition) algorithms are often set empirically, leading to mode aliasing and affecting the accuracy of fault feature extraction.

[0004] Therefore, there is an urgent need in this field for a fault diagnosis method that can construct a source domain model that accurately matches the entire lifecycle data, optimize the feature extraction process, and improve the transfer learning effect. Summary of the Invention

[0005] The purpose of this invention is to provide a method for diagnosing full-life rolling bearing faults based on mechanism data migration, so as to solve or at least partially solve the technical problems mentioned in the background art.

[0006] To achieve this objective, the present invention adopts the following technical solution: This invention provides a method for full-life rolling bearing fault diagnosis based on mechanistic data migration, comprising: The rolling bearing is constructed as a mass-spring-damping model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator; Based on Hertz contact theory, the nonlinear dynamic equation of the rolling bearing is obtained by modifying the radial displacement excitation function of the rolling element, the time-varying contact stiffness between the rolling element and the bearing raceway, and the impact force at the contact defect of the rolling element in the mass-spring-damping model. The nonlinear dynamic equations of rolling bearings are solved by the fourth-order Runge-Kutta method, generating source domain data consistent with the test parameters of full-life rolling bearings, and constructing a source domain dataset with complete fault labels. Based on the source domain dataset and the pre-built target domain dataset, train a pre-built improved alternating transfer learning model; A trained improved alternating transfer learning model is used to perform full-life rolling bearing fault diagnosis, and the diagnosis results are output. The target domain dataset covers data samples from rolling bearing tests throughout the entire lifespan of the rolling bearing, from normal operation to failure, including sample data for four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault. The source domain dataset includes simulation data for the four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault.

[0007] Optionally, the nonlinear dynamic equation of the rolling bearing is: ; The first two lines are the equations of motion for the inner circle, where, For the quality of the inner ring, This is the damping coefficient between the inner ring and the bearing housing. This refers to the support stiffness between the inner ring and the bearing housing. , These represent the displacements of the inner ring in the x and y directions, respectively. , For speed, , For acceleration, , F represents the x and y components of the contact force between the inner ring and the rolling element, respectively. r Where is the bearing load, and e is the eccentricity of the inner ring. ω is the angular velocity of the inner ring. The third and fourth lines are the equations of motion for the outer circle, where... For the quality of the outer ring, , These represent the support stiffness and damping coefficient between the outer ring and the bearing housing, respectively. , These represent the horizontal and vertical displacements of the shaft and the outer ring, respectively. , These represent the displacements of the center of the rolling element in the x and y directions, respectively. , These are the outer ring fault excitation forces in the x and y directions, respectively; The fifth and sixth lines are the equations of motion for the rolling element, in which... For the mass of a single rolling element; The last line is the equation of motion for the cage, where, To maintain rack quality, Let J be the cage angular displacement corresponding to the j-th rolling element. This represents the extreme diameter of the j-th rolling element. To maintain the angular velocity of the frame, This indicates the angular position of the j-th rolling element. , The Hertzian contact force between the inner and outer rings of the bearing on the rolling elements in the radial direction.

[0008] Optionally, the radial displacement excitation function of the rolling element in the mass-spring-damped model is modified based on Hertz contact theory, specifically including: cage speed Rotational angular velocity of the inner ring The relationship between them is: d is the rolling element diameter; D is the bearing pitch diameter; β is the contact angle; ; In the above formula, Z represents the number of rolling elements, and j is a natural number greater than or equal to 1. The angular position of the first rolling element relative to the positive y-axis at time zero; The expression for the radial displacement excitation function s of the rolling element abruptly entering the defect region is: , ; In the above formula, R is the radius of the bearing outer ring raceway; r is the radius of the rolling element. The angle at which the rolling element enters the defect area; , These are the angular positions of the front and rear edges of the defect area, respectively. The radial displacement of the contact deformation between the j-th rolling element and the inner ring raceway of the bearing is: ; This refers to the radial clearance of the bearing; Radial displacement of the contact deformation between the j-th rolling element and the outer ring raceway of the bearing Revised to: .

[0009] Optionally, the time-varying contact stiffness between the rolling element and the bearing raceway is modified based on Hertz contact theory, specifically including: The formula for the Hertzian contact force between the inner and outer rings of the bearing on the rolling element in the radial direction is as follows: , ; In the formula: , , , , In the formula, , These are the stiffness coefficients of the inner and outer raceways of the bearing, respectively; n is the load-deformation coefficient. , These represent the relative displacement vectors of the j-th rolling element and the raceways of the inner and outer rings of the bearing, respectively. The formula for calculating time-varying contact stiffness is as follows: , ; In the formula, This is the corrected time-varying contact stiffness coefficient. S is the reduction factor, and S is the shape factor; It refers to the range of banding.

[0010] Optionally, the impact force at the rolling element contact defect is corrected based on Hertz contact theory, specifically including: Impact force generated at the edge of the rolling element after impacting the defect The calculation formula is: L is the defect width: m b For the mass of the rolling element; Will Orthogonal decomposition yields horizontal and vertical components: , : The angle between the impact force and the radial velocity of the rolling element: ; In the formula, The discriminant function for determining the presence of impact force is defined as follows: ; The magnitudes of the total Hertzian contact forces acting on the inner and outer rings of the rolling bearing are as follows: , .

[0011] Optionally, constructing the rolling bearing as a mass-spring-damped model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator specifically includes: For rolling bearings used in full-life testing, the rolling bearing system is simplified into a mass-spring-damped model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator, and the model is assumed to satisfy the following conditions: The outer ring is fixed to a rigid structure, while the inner ring maintains a constant rotational speed. The rolling elements are evenly distributed with no relative slippage; The effect of the rolling element's rotation is not considered; The inner and outer rings and the rolling elements move in the same plane.

[0012] Optionally, the step of solving the nonlinear dynamic equations of the rolling bearing using the fourth-order Runge-Kutta method to generate source domain data consistent with the test parameters of the full-life rolling bearing, and constructing a source domain dataset with complete fault labels, specifically includes: The nonlinear dynamic equations of rolling bearings are solved using the fourth-order Runge-Kutta method, generating simulation data for four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault. Fault simulation is performed based on the generated simulation data, and a source domain dataset with complete fault labels is constructed.

[0013] Optionally, before training the pre-built improved alternating transfer learning model based on the source domain dataset and the pre-built target domain dataset, the method further includes: Construct the target domain dataset; the method is as follows: Collect full-life rolling bearing test data; wherein, the full-life rolling bearing test data covers the entire process of the bearing from normal operation to failure, including axial and radial vibration signals; The collected full-life rolling bearing test data were preprocessed to construct an original dataset containing four types of fault samples: normal state, inner ring fault, outer ring fault, and rolling element fault. The amplitude spectrum of the fault samples is selected as the fitness function of the CS-GWO algorithm, and the minimum value of the amplitude spectrum entropy is used as the fitness value to optimize the mode number K and the quadratic penalty factor in the VMD algorithm. ; Based on the optimized mode number K and the quadratic penalty factor K modal components are obtained by decomposing using the VMD algorithm; Multiple modal components with the highest correlation to the fault samples were selected as effective modes, summed and reconstructed, and envelope spectrum analysis was performed on the reconstructed modal components to extract feature information related to the rolling bearing faults, forming sample data.

[0014] Optionally, the loss function of the improved alternating transfer learning model is set as follows: The distance between the second-order statistics of the features in the source and target domains is defined as the CORAL loss function L1: , is the Frobenius norm of the mean square matrix; in, Let D be the number of dimensions of this sample, and D be the source domain dataset. S , and the target domain dataset D T The characteristic covariance matrices are C S With C T ; MMD is used to measure the feature set D. S and D T The distributional differences in the reproducing kernel Hilbert space are used as the sum of the MMD loss and the classification loss as the L2 loss function: , , , For the regenerating nucleus Hilbert space; Among them, L C The classification loss is the difference between the true label and the predicted label. , Mapping function , n is the balance coefficient. S Let y be the number of samples, and y be the true label of the sample data. The label predicted by the classifier.

[0015] Optionally, the formula for calculating the amplitude spectral entropy is: ; In the above formula, L i For modal components u i Amplitude spectrum; H i For modal components u i The amplitude spectral entropy; N is the length of the modal component.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This application constructs a dynamic mechanism model consistent with the bearing parameters in a full-life test. The generated source domain data and target domain data have small differences in feature distribution, laying the foundation for efficient knowledge transfer and solving the problem of mismatch between traditional source domain data and actual data. It does not rely on fault samples in a full-life scenario and can achieve accurate diagnosis only through mechanism simulation data. It is applicable to full-life condition monitoring and fault diagnosis of rolling bearings in various rotating machinery. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The flowchart illustrates a full-life rolling bearing fault diagnosis method based on mechanism data migration, as provided in this embodiment of the invention.

[0019] Figure 2 This is a schematic diagram of a nonlinear dynamic model of a rolling bearing provided in an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram illustrating the dimensional relationship of a rolling element entering a defect region, provided as an embodiment of the present invention.

[0021] Figure 4 This is a schematic diagram of a rolling body overcoming an obstacle, provided in an embodiment of the present invention.

[0022] Figure 5 This is a diagram showing the change in the contact surface of the elastic deformation of the contact edge of a rolling element, provided as an embodiment of the present invention.

[0023] Figure 6 This is a schematic diagram of the impact force at a contact defect of a rolling element, provided as an embodiment of the present invention.

[0024] Figure 7 The present invention provides a time-domain waveform diagram of simulation data for four states of a rolling bearing.

[0025] Figure 8a A spectrum diagram of simulation data of inner ring fault state of a rolling bearing provided for an embodiment of the present invention.

[0026] Figure 8b A spectrum diagram of simulation data of outer ring fault state of a rolling bearing provided for an embodiment of the present invention.

[0027] Figure 8c A spectrum diagram of simulation data of rolling element failure state of a rolling bearing provided for an embodiment of the present invention.

[0028] Figure 9 The time-domain waveform diagram of simulation data for four states of a rolling bearing provided in an embodiment of the present invention.

[0029] Figure 10a A spectrum diagram of simulation data for the inner ring fault state of another rolling bearing provided in an embodiment of the present invention.

[0030] Figure 10b A spectrum diagram of simulation data for the outer ring fault state of another rolling bearing provided in an embodiment of the present invention.

[0031] Figure 10c A spectrum diagram of simulation data of rolling element failure state of another rolling bearing provided for an embodiment of the present invention.

[0032] Figure 11 This is a schematic diagram of a rolling bearing full-life testing device provided in an embodiment of the present invention.

[0033] Figure 12 A flowchart of a CS-GWO optimized VMD algorithm provided for an embodiment of the present invention.

[0034] Figure 13a This is a time-domain waveform diagram of two-channel signals under an inner ring fault state during an IMS test, provided as an embodiment of the present invention.

[0035] Figure 13b The spectrum diagram of the inner ring fault signal corresponding to Figure 13-a is shown.

[0036] Figure 14 This is a schematic diagram illustrating the iterative process of optimizing VMD fitness values ​​using the CS-GWO algorithm, as provided in an embodiment of the present invention.

[0037] Figure 15 This is a time-domain waveform diagram after VMD processing, provided as an embodiment of the present invention.

[0038] Figure 16 This is a spectrum diagram after VMD processing provided in an embodiment of the present invention.

[0039] Figure 17 This invention provides an envelope spectrum reconstructed after VMD processing optimized by the CS-GWO algorithm, as an embodiment of the present invention.

[0040] Figure 18 This invention provides an envelope spectrum reconstructed after BA-optimized VMD decomposition.

[0041] Figure 19 This is an envelope spectrum reconstructed after SSA-optimized VMD decomposition, provided as an embodiment of the present invention.

[0042] Figure 20 This invention provides an envelope spectrum reconstructed after WOA-optimized VMD decomposition.

[0043] Figure 21a This invention provides a source domain time-domain waveform and grayscale image.

[0044] Figure 21b This invention provides a time-domain waveform and grayscale image of a target domain.

[0045] Figure 22 This is a migration task result confusion matrix provided in an embodiment of the present invention.

[0046] Figure 23 A t-SNE visualization result of a migration task provided in an embodiment of the present invention. Detailed Implementation To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0047] Example 1: Please refer to Figure 1 , Figure 1 A flowchart of a full-life rolling bearing fault diagnosis method based on mechanism data migration provided in this embodiment of the invention is shown. The method includes: Step 110: Construct the rolling bearing as a mass-spring-damping model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator.

[0048] For rolling bearings used in full-life testing (such as Rexnord ZA-2115 bearings), the rolling bearing system is simplified into a mass-spring-damped model consisting of an outer ring, inner ring, rolling elements, and a high-frequency resonator, and the model is assumed to satisfy the following conditions: The outer ring is fixed to a rigid structure, while the inner ring maintains a constant rotational speed. The rolling elements are evenly distributed with no relative slippage; The effect of the rolling element's rotation is not considered; The inner and outer rings and the rolling elements move in the same plane.

[0049] like Figure 2 As shown, Figure 2 A schematic diagram of a nonlinear dynamic model of a rolling bearing provided in an embodiment of the present invention; in the figure, For the quality of the inner ring, This is the damping coefficient between the inner ring and the bearing housing. This refers to the support stiffness between the inner ring and the bearing housing. For the quality of the outer ring, , These represent the support stiffness and damping coefficient between the outer ring and the bearing housing, respectively. For the mass of a single rolling element, This indicates the angular position of the j-th rolling element.

[0050] When a rolling element enters a defect region, it rotates around the leading edge of the defect, causing a sudden change in its radial displacement. This application calculates the relative position of the rolling element in the defect region using geometric relationships, thus accurately reconstructing the obstacle-crossing process of the rolling element. Figure 3 As shown.

[0051] according to Figure 3 The dimensional relationships shown indicate the angle at which the rolling element enters the defect region. The expressions for sine and cosine are: , ; cage speed Rotational angular velocity of the inner ring The relationship between them is: ; In the above formula, d is the rolling element diameter; D is the bearing pitch diameter; β is the contact angle; ; In the above formula, Z represents the number of rolling elements, j = 1, 2, 3, ...; The angular position of the first rolling element relative to the positive y-axis at time zero.

[0052] Step 120: Based on Hertz contact theory, modify the radial displacement excitation function of the rolling element, the time-varying contact stiffness between the rolling element and the bearing raceway, and the impact force at the contact defect of the rolling element in the mass-spring-damping model to obtain the nonlinear dynamic equation of the rolling bearing.

[0053] This step includes: optimizing the displacement excitation function based on the real-time position of the rolling element, correcting the impact force between the rolling element and the fault area by combining the size (width, depth) of the fault defect and the bearing speed, and correcting the equivalent contact stiffness by considering the elastic deformation characteristics of the contact edge. Specifically, according to Hertzian contact theory, the contact between the rolling elements and the raceway leads to elastic deformation, thus creating Hertzian contact force. This contact force is one of the important forces during bearing operation. When a bearing has defects, the contact deformation is divided into two cases: one is contact deformation in the non-defective area, and the other is contact deformation in the defective area. These deformations have a significant impact on the normal operation and fault diagnosis of the bearing.

[0054] like Figure 4 As shown, the bearing inner and outer rings are given two degrees of freedom in the horizontal and vertical directions, respectively. When the inner and outer rings of the bearing undergo relative displacement, , Let represent the relative displacement vectors of the j-th rolling element with respect to the inner and outer raceways of the bearing, respectively. Their specific coordinate relationships are as follows: , , In the formula, Indicates the extreme diameter of the j-th rolling element; , These represent the displacements of the shaft and the inner ring in the x (horizontal displacement) and y (vertical displacement) directions, respectively. , These represent the horizontal and vertical displacements of the shaft and the outer ring, respectively.

[0055] Under fault-free conditions, the contact deformations of the j-th rolling element with the inner and outer raceways of the bearing are as follows: ;in, This refers to the radial clearance of the bearing; ; When a local defect occurs in the outer raceway of a bearing, this application simplifies the defect area into a rectangle with a sufficiently large depth h, satisfying 4dh ≥ 4h. 2 +L 2 Where d is the diameter of the rolling element; L is the width of the defect. The situation where the rolling element touches the bottom of the defect during obstacle crossing is not discussed.

[0056] like Figure 3 As shown, when the rolling element enters the defect region, it rotates around the leading edge of the defect. The radial displacement of the rolling element undergoes a sudden change. When the rolling element simultaneously contacts the front and rear edges of the defect, that is... At that time, the radial displacement reaches its maximum value due to the sudden change. The value of the radial displacement during this abrupt change is a process of "from small to large" followed by "from large to small".

[0057] To address the above situation, traditional research uses a half-sine function to describe this process. However, the instantaneous displacement excitation of the rolling element over an obstacle is related to factors such as the relative position of the rolling element and the size of the fault. A single half-sine function is insufficient to accurately describe the changes in displacement excitation, especially when describing the relative position of the rolling element in the defect area. Therefore, to more accurately reconstruct the rolling element obstacle-crossing process, this method uses geometric relationships to calculate the relative position of the rolling element in the defect area. This method can more accurately capture the motion of the rolling element in the defect area.

[0058] Figure 4 In this context, R represents the outer ring raceway radius of the bearing; r represents the rolling element radius. The angle at which the rolling element enters the defect area; , These are the angular positions of the front and rear edges of the defect area, respectively. The expression is: ; The radial displacement excitation function s for the abrupt change in the rolling element entering the defect region is expressed as: ; Then the radial displacement of the contact deformation between the j-th rolling element and the outer ring raceway of the bearing Revised to: .

[0059] Furthermore, based on Hertzian contact theory, the equivalent contact stiffness between the rigid rolling element and the raceways of the inner and outer rings of the bearing is... It can be represented as: ; In the above formula, , These are the stiffness coefficients of the inner and outer raceways of the bearing, respectively; n is the load-deformation coefficient. In this embodiment, n is taken as 1.5 for the rolling element bearing. After obtaining the contact deformation of the rolling elements and the inner and outer raceways of the bearing, the nonlinear contact force F generated in the directions perpendicular to the inner and outer raceways is calculated, and its magnitude is: , ; and The correction angle for the Hertzian contact force on the outer and inner rings is calculated using the following formula. , , The formula for the Hertzian contact force between the inner and outer rings of the bearing on the rolling element in the radial direction is as follows: , ; Figure 5 The diagram shows the change in the elastic deformation contact surface at the contact edge of the rolling element. In a rolling bearing, when the rolling element moves to the edge of the defect area, the elastic deformation contact surface becomes half of its original size. The angular range of this contact surface is defined as the band range. At this point, the above assumption does not hold, so the equivalent contact stiffness between the rolling element and the raceways of the inner and outer rings of the bearing needs to be corrected.

[0060] The formula for calculating time-varying contact stiffness is as follows: ; In the formula, This is the corrected time-varying contact stiffness coefficient. This is a reduction factor, with a range of values. >0.5; S is the shape factor, which is set to 1 here; to simplify the calculation, the band range is used here. Take a constant value, that is: .

[0061] Furthermore, such as Figure 6 As shown, during the process of the rolling element over the obstacle, in addition to the sudden change in the radial displacement of the rolling element mentioned above, an instantaneous impact force will also be generated at the moment when the rolling element contacts the edge line of the defect. ; In the above formula, The radial velocity of the rolling element; The value is related to the angular position of the rolling element, and the expression is as follows: ; The formula for the radial velocity of the rolling element can be simplified to: ; In the formula, Let be the angle between the impact force and the radial velocity of the rolling element. Based on geometric relationships, its expression is: ; Impact force duration The time from when the rolling element impacts the edge of the defect to when it leaves the defect area is expressed as: ; The impact force generated at the edge of the rolling element after impacting the defect can be calculated using the impulse theorem. Substituting the above parameters into the equation, we get: ; In the above formula, L is the defect width: m b For the mass of the rolling element; Will Orthogonal decomposition yields horizontal and vertical components: ; When the rolling element passes through the defect area of ​​the outer ring, the magnitude of the total Hertzian contact force on the inner and outer rings of the bearing can be calculated using the following formula: , ; In the formula, The discriminant function for determining the presence of impact force is defined as follows: ; After correction, the total Hertzian contact force on the inner and outer rings of the bearing model with outer ring defects is shown below: , .

[0062] In summary, the nonlinear dynamic equation of the rolling bearing is: ; The first two lines are the equations of motion for the inner circle, where, For the quality of the inner ring, This is the damping coefficient between the inner ring and the bearing housing. This refers to the support stiffness between the inner ring and the bearing housing. , These represent the displacements of the inner ring in the x and y directions, respectively. , For speed, , For acceleration, , F represents the x and y components of the contact force between the inner ring and the rolling element, respectively. r Where is the bearing load, and e is the eccentricity of the inner ring. The rotational angular velocity of the inner ring; The third and fourth lines are the equations of motion for the outer circle, where... For the quality of the outer ring, , These represent the support stiffness and damping coefficient between the outer ring and the bearing housing, respectively. , These represent the horizontal and vertical displacements of the shaft and the outer ring, respectively. , These represent the displacements of the center of the rolling element in the x and y directions, respectively. , These are the outer ring fault excitation forces in the x and y directions, respectively; The fifth and sixth lines are the equations of motion for the rolling element, in which... For the mass of a single rolling element; The last line is the equation of motion for the cage, where, To maintain rack quality, Let J be the cage angular displacement corresponding to the j-th rolling element. This represents the extreme diameter of the j-th rolling element. To maintain the angular velocity of the frame, This indicates the angular position of the j-th rolling element. , The Hertzian contact force between the inner and outer rings of the bearing on the rolling elements in the radial direction.

[0063] This set of nonlinear dynamic equations has no analytical solution and needs to be solved numerically to obtain the vibration response of the system in the time / frequency domain.

[0064] Step 130: Solve the nonlinear dynamic equations of the rolling bearing using the fourth-order Runge-Kutta method to generate source domain data consistent with the test parameters of the full-life rolling bearing, and construct a source domain dataset with complete fault labels.

[0065] The nonlinear dynamic equations of rolling bearings are solved using the fourth-order Runge-Kutta method, generating simulation data for four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault. Fault simulation is performed based on the generated simulation data, and a source domain dataset with complete fault labels is constructed.

[0066] It should be noted that the fourth-order Runge-Kutta method (RK4) is a commonly used high-precision single-step numerical solution method in engineering and numerical computation. It is used to solve initial value problems of first-order ordinary differential equations. Its core advantages are high accuracy and good stability. It does not require the calculation of higher-order derivatives and can complete the solution in one step through only four function value iterations. Since it is a commonly used mathematical method, its solution process will not be described in detail in this embodiment.

[0067] Step 140: Train the pre-built improved alternating transfer learning model based on the source domain dataset and the pre-built target domain dataset.

[0068] It should be noted that before step 140, the following steps are also included: Pre-set step 141: Construct the target domain dataset.

[0069] The specific method for constructing the target domain dataset is as follows: Collect full-life rolling bearing test data; wherein, the full-life rolling bearing test data covers the entire process of the bearing from normal operation to failure, including axial and radial vibration signals; The collected full-life rolling bearing test data were preprocessed to construct an original dataset containing four types of fault samples: normal state, inner ring fault, outer ring fault, and rolling element fault. The amplitude spectrum of the fault samples is selected as the fitness function of the CS-GWO algorithm, and the minimum value of the amplitude spectrum entropy is used as the fitness value to optimize the mode number K and the quadratic penalty factor in the VMD algorithm. ; Based on the optimized mode number K and the quadratic penalty factor K modal components are obtained by decomposing using the VMD algorithm; Multiple modal components with the highest correlation to the fault samples were selected as effective modes, summed and reconstructed, and envelope spectrum analysis was performed on the reconstructed modal components to extract feature information related to the rolling bearing faults, forming sample data.

[0070] More specifically, the amplitude spectrum of the selected fault sample is used as the fitness function of the CS-GWO algorithm, and the minimum value of the amplitude spectrum entropy is used as the fitness value to optimize the mode number K and the quadratic penalty factor in the VMD algorithm. The specific method is as follows: The Grey Wolf (GWO) algorithm is widely used in function optimization and other problems due to its strong search capabilities and high efficiency. However, its position update method makes its global search ability relatively weak, and it is prone to getting trapped in local minima. The Cuckoo Search (CS) algorithm, on the other hand, requires fewer parameters and can easily jump out of the current region, achieving a global search. The CS algorithm is inspired by Levi's flight and the parasitic feeding mechanism of the cuckoo. The location of the host's nest, parasitized by the cuckoo, is iteratively updated according to a formula based on the cuckoo's unique search method: ; In the above formula, i = 1, 2, ..., N, where N represents the number of host bird nests; This is an inner product operation; This represents the coordinates of the i-th host nest during the t-th iteration; Let these be the coordinates of the i-th host nest after iteration; =1 is the step size factor; It follows a Lévy distribution with a step size of .

[0071] Combining the advantages of the CS algorithm, such as its ability to easily jump from the computational region to the uncomputed region during global search, the GWO algorithm, which is improved upon, effectively overcomes the problem of the GWO algorithm easily getting trapped in local optima during the optimization process. This improvement allows the algorithm to better explore the search space, enhances its global search capability, and thus makes it more effective in solving various optimization problems.

[0072] Furthermore, using the minimum value of the amplitude spectral entropy of the modal components as the fitness function, the optimal combination of VMD parameters is obtained through iterative optimization using the CS-GWO algorithm, thus avoiding modal aliasing.

[0073] Among the many parameters of VMD, the number of decomposition modes K and the second-order penalty factor have the greatest impact on the results. Therefore, the choice of VMD parameter combination is crucial. Using it as the fitness function optimization objective of the CS-GWO algorithm can effectively avoid obtaining suboptimal solutions due to improper parameter combinations.

[0074] The amplitude spectrum is obtained by performing an FFT transform on the modal components obtained from VMD decomposition. The amplitude spectrum is then combined with the information entropy to obtain the amplitude spectrum entropy; the calculation formula is as follows: ; In the above formula, L i For modal components u i Amplitude spectrum; H i For modal components u i The amplitude spectral entropy; N is the length of the modal component; It can be seen that the change in amplitude spectral entropy can reflect whether the signal contains a lot of fault characteristic information. When the modal components obtained by VMD decomposition contain a lot of noise components and the fault impact signal components are few, the amplitude spectral entropy of the modal component will be large, as can be seen from the calculation results of the formula. Conversely, when the modal component contains less noise signal, the regular periodic fault impact signal response is obvious, that is, the modal component is rich in fault characteristic information and has a small amplitude spectral entropy.

[0075] Therefore, this method selects the amplitude spectrum (fault signal) as the fitness function of the CS-GWO algorithm, and uses the minimum value of the amplitude spectrum entropy as the fitness value to optimize the mode number K and the quadratic penalty factor in the VMD algorithm. The main steps for optimizing VMD based on the CS-GWO algorithm are as follows: Initialize the parameters for the CS-GWO algorithm and VMD, including the wolf population size, the number of algorithm iterations, the number of modes K, and the quadratic penalty factor. wait; Set the number of VMD modes K and the quadratic penalty factor. Training is performed, and the number of modes K and the quadratic penalty factor are calculated simultaneously. Find the comprehensive evaluation index H of the modal components under the given conditions. min The value is assigned to fitness. The position of the entire wolf pack is updated based on the fitness value obtained by each individual, so as to ensure that the algorithm continues to move in a better direction in the search space, thereby improving search efficiency and the final optimization result. Based on the updated position of the wolf pack, a random decision-making process is performed. Specifically, the algorithm selects a random number in the range [0,1] and compares it with the probability of finding bird eggs. If the random number is greater than the probability of finding bird eggs, the algorithm will update the current position of the wolf pack; otherwise, it will not update it. By calculating the fitness value of the wolf pack's location, we can evaluate the merits of the new location, determine which locations are closer to the optimal solution, and thus provide guidance for the next location update. The current iteration count is compared with the set maximum allowed iteration count. If the current iteration count has not reached the maximum allowed iteration count, the process jumps to step 3 to continue optimizing and finding the optimal solution. This process continues until the maximum allowed iteration count is reached, or other stopping conditions are met, such as convergence of the objective function, ultimately obtaining the globally optimal position of the food, which serves as the modality number K and the secondary penalty factor in the VMD algorithm. Output the optimal solution.

[0076] The rolling bearing fault feature extraction process of VMD is optimized using CS-GWO, as follows: Figure 12 As shown. The specific steps are as follows: Define the number of modes K and the quadratic penalty term. Initialize parameters such as the number of wolves and the maximum number of iterations; The CS-GWO algorithm is used to optimize the number of VMD parameter modes K and the quadratic penalty factor of the parameters. Determine the optimization results; By combining the parameters of the optimization results with the VMD algorithm, the vibration signal of the target channel can be effectively decomposed to obtain K modal components; Calculate the correlation between each modal component and the original signal, and select some effective modes with high correlation to sum and reconstruct the signal; Envelope spectrum analysis is performed on the modal components obtained through reconstruction to extract feature information related to bearing faults. The features are then compared with the actual fault feature frequencies to extract fault features.

[0077] This method uses optimal parameters to perform VMD decomposition on the target domain data, calculates the correlation between each modal component and the original signal, selects the modal components with high correlation to sum and reconstruct, and achieves noise suppression and fault feature enhancement.

[0078] Specifically, in step 140, before training the model, the data format is converted, that is, the source domain simulation data and the reconstructed target domain data are converted into two-dimensional grayscale images to give full play to the two-dimensional feature extraction capability of the convolutional neural network (CNN). The improved alternating transfer learning model constructed in this embodiment comprises a CNN network with 5 convolutional layers, 2 pooling layers, and 2 fully connected layers. A batch normalization layer is added after each convolutional layer. The ReLU function is used as the activation function for the hidden layers, and the Softmax function is used as the activation function for the output layer. The first pooling layer is located between the second and third convolutional layers, and the second pooling layer is located after the fifth convolutional layer and before the first fully connected layer. The CORAL loss function is calculated after the first convolutional layer of the CNN model to reduce the difference in second-order statistics between the source and target domain datasets. The sum of the MMD loss function and the classification loss is calculated in the fully connected layer, and the network weights and bias parameters are updated through alternating backpropagation.

[0079] The target domain dataset covers data samples from rolling bearing tests throughout the entire lifespan of the rolling bearing, from normal operation to failure, including sample data for four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault. The source domain dataset includes simulation data for the four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault.

[0080] More specifically, the method for setting the loss function of the improved alternating transfer learning model is as follows: The distance between the second-order statistics of the features in the source and target domains is defined as the CORAL loss function L1: , is the Frobenius norm of the mean square matrix; in, Let D be the number of dimensions of this sample, and D be the source domain dataset. S , and the target domain dataset D T The characteristic covariance matrices are C S With C T ; MMD is used to measure the feature set D. S and D T The distributional differences in the reproducing kernel Hilbert space are used as the sum of the MMD loss and the classification loss as the L2 loss function: , , , For the regenerating nucleus Hilbert space; Among them, L C The classification loss is the difference between the true label and the predicted label. , Mapping function , n is the balance coefficient. SLet y be the number of samples, and y be the true label of the sample data. The label predicted by the classifier.

[0081] Step 150: Use the trained improved alternating transfer learning model to perform full-life rolling bearing fault diagnosis and output the diagnosis results.

[0082] To verify the effectiveness of this method, specific application examples are given below.

[0083] The first step is to analyze the dynamic model of rolling bearing failure: The simulation model uses the SKF6205-2RS deep groove ball bearing, whose main parameters are shown in Table 1-1.

[0084] Table 1-1 The parameters in the model are set as follows: the total mass m of the inner ring and the shaft. i =0.5kg, system damping coefficient c=350N·s / m, the forces applied to the shaft are F x =F y =0N,F r =1000N, initial displacements x0 and y0 are both 10. -6 mm, initial velocity and All values ​​are 0 m / s. The damage diameter L is set to 0.1778 mm, the damage depth h to 0.2794 mm, the sampling frequency to 12 kHz, and the inner ring speed to 1797 r / min. Due to the strong nonlinearity of the bearing dynamic equations, the fourth-order Runge-Kutta method is used to solve the established dynamic model, with a calculation step size of [missing value]. =5×10 -6 s, the time-domain waveforms of the vibration acceleration signals collected under four states are obtained as follows Figure 7 As shown, vibration acceleration signals are obtained using a simulation model, and a source domain dataset is constructed.

[0085] To further verify the accuracy of the model results obtained by this method, an FFT is performed on the time-domain signal waveform for frequency domain analysis. Based on the currently used method, the formulas for calculating the characteristic frequency and fault frequency of the rolling bearing are as follows: Rotation frequency of rolling bearings for: where n is the bearing speed; The cosine value of the bearing contact angle is: Inner ring fault characteristic frequency for: ; Outer ring fault characteristic frequency for: ; Rolling element failure characteristic frequency for: .

[0086] The bearing speed is set here. =1797 r / min. Based on the main parameters of bearing SKF6205-2RS in Table 1-1 and the formula for rolling bearing failure frequency, the following can be calculated: Inner ring fault characteristic frequency =162.56Hz, outer ring fault frequency =106.99Hz, rolling element failure frequency =140.67Hz. FFT analysis was performed on the three sets of fault signals, and Figure 8 shows their spectrum. The calculated errors between the three simulated fault frequencies and the theoretically calculated fault frequencies are 0.19%, 0.29%, and 0.64%, respectively.

[0087] For example, the target domain data uses the Intelligent Maintenance System (IMS) bearing dataset from the University of Cincinnati, USA. The experiment uses Rexnord ZA-2115 double row bearings, and their bearing parameters are shown in Table 2-1 below.

[0088] Table 2-1 A radial load of 2721.55 kg is applied to the bearing via a spring mechanism, with a sampling frequency of 20 kHz and a motor speed of 2000 r / min. Due to the strong nonlinearity of the bearing dynamic equations, the fourth-order Runge-Kutta method is used to solve the established dynamic model, with a calculation step size of [missing information]. =5×10 -6 s, the time-domain waveforms of the vibration acceleration signals collected under four states are obtained as follows Figure 9 As shown, vibration acceleration signals were obtained using a simulation model, and a source domain dataset was constructed. Other parameters were set consistent with the University of Cincinnati bearing dataset.

[0089] To further verify the accuracy of our model results, we performed an FFT on the time-domain signal waveform and conducted frequency domain analysis. Based on the previous formulas for calculating the characteristic frequency and fault frequency of rolling bearings, we can obtain the inner race fault characteristic frequency. =296.68Hz, outer ring fault frequency =236.29Hz, rolling element failure frequency =139.72Hz. Figure 10 shows its spectrum. Calculations show that the errors between the three simulated fault frequencies and the theoretically calculated fault frequencies are 0.08%, 0.06%, and 0.11%, respectively.

[0090] Furthermore, the target domain dataset uses the IMS rolling bearing dataset, and the schematic diagram of the bearing test bench and sensor placement is shown below. Figure 11 As shown, the experiment involved mounting four bearings on the shaft. An AC motor connected to the shaft via a friction belt maintained a constant rotational speed of 2000 r / min. A radial load of 2721.55 kg was applied to the shaft and bearings via a spring mechanism. All bearings were forcibly lubricated. Rexnord ZA-2115 double-row bearings were mounted on the shaft. The PCB 353B33 high-sensitivity quartz ICP accelerometers were mounted on the bearing housings (two accelerometers per bearing (x-axis and y-axis) in dataset 1, and one accelerometer per bearing in datasets 2 and 3). The datasets of experimental results are shown in Table 2-2.

[0091] Table 2-2 As can be seen from the dataset in Table 2-2, all experimental bearings operated in good condition at the beginning of the full life test. As time progressed, the bearings continued to work and eventually reached their service life limit, resulting in inner ring failure, outer ring failure, and rolling element failure in the experimental test results. Therefore, the entire process of the full life test can be regarded as the operation process of the Rexnord ZA-2115 double row bearing in the full life test scenario. Thus, the IMS dataset will be used here to simulate the dataset collected in the full life test scenario. We selected bearings 2, 3, and 4 from dataset 1 and bearing 1 from dataset 2 as the normal state (N), inner ring (I), rolling element failure (R), and outer ring failure (O) in the target domain dataset, respectively.

[0092] The first dataset in the IMS test contained vibration signals from both axial and radial channels of the bearing. The vibration signals from this first dataset were selected for analysis. At the end of the experiment, bearing 3 in the test bench developed an inner ring fault. The time-domain waveforms of the two channels are shown below. Figure 13a As shown, from Figure 13a As can be seen, the original vibration signals of the inner rings in both channels contain a small amount of background noise and impact signal components, but no obvious periodic signal was observed. The signal from channel 5 was selected for analysis, combined with... Figure 13b The spectrum diagram of the inner ring fault signal shows that there is a lot of noise interference in the signal. The fault characteristic signal is not prominent in the spectrum diagram. Furthermore, due to the different directions of the bearing vibration signal measured by the sensor, there are certain differences in the time and spectrum diagrams of the two channels.

[0093] Taking the vibration signal from channel one as an example, the CS-GWO algorithm is used to optimize the VMD mode number and the quadratic penalty term. Their search spaces are set to [2,10] and [1000,20000], respectively, the number of wolf packs is 10, and the maximum number of iterations is 30. The CS-GWO algorithm's fitness value optimization process is as follows: Figure 14 As shown.

[0094] It can be seen that the fitness value decreases continuously with the increase of the number of iterations, and converges to the optimal value after 9 iterations. The fitness value determined by the CS-GWO algorithm ultimately determines the number of modes in VMD to be 4, and the quadratic penalty factor to be 8640. Modal decomposition is then performed in the VMD algorithm, and the time-domain waveforms of the modal components obtained from the channel decomposition are shown below. Figure 15 As shown, the spectrum is as follows Figure 16 As shown; To select the optimal modal components, correlation analysis was performed on the modal components and the original vibration signal to assess their correlation strength. Modal components with weak correlation were discarded, while those with strong correlation were retained and summed to reconstruct the original signal. The modal components after VMD decomposition were selected, and the strongly correlated modal component signals were summed and reconstructed, as shown in the envelope spectrum. Figure 17 As shown. By Figure 17 It can be seen that the characteristic frequency of the bearing inner ring fault has a large amplitude spectral peak at its second harmonic, which is basically consistent with the theoretically calculated inner ring fault frequency, with an error rate of only 0.03%. This indicates that the bearing inner ring has local damage, which is consistent with the actual situation.

[0095] To demonstrate the superiority of our proposed method, the parameters of VMD were optimized using the Bat Algorithm (BA), Sparrow Search Algorithm (SSA), and Whale Algorithm (WOA), respectively. The results were: 5 modalities with a penalty factor of 429; 4 modalities with a penalty factor of 12860; and 5 modalities with a penalty factor of 880.

[0096] Correlation analysis was performed between the modal components obtained from signal decomposition using different algorithms and the original signal. The modal component signals with strong correlations were summed and reconstructed, and their envelope spectra are shown below. Figures 18-20 As shown. Comparison Figure 17It can be seen that the inner fault characteristic frequencies and their harmonics are relatively obvious in the envelope spectrum obtained by BA-optimized VMD, but other prominent interference frequencies are still present. The envelope spectrum obtained by SSA-optimized VMD also shows that the inner fault characteristic frequencies and their harmonics are relatively obvious, and the interference frequencies caused by noise are relatively small in the high-frequency region, indicating that the SSA-optimized VMD has a good suppression effect on interference frequencies in the high-frequency region. However, there are large interference frequencies in the low-frequency region, and the decomposition result has little effect on suppressing interference frequencies in the low-frequency region. The envelope spectrum obtained by WOA-optimized VMD shows the inner fault characteristic frequencies and their harmonics, but its ability to remove interference signals across the entire frequency band is relatively average, and some interference frequencies are still quite prominent. The parameter-optimized VMD... Figure 17 In the middle, the fault characteristic frequency and its harmonic components in the inner circle become more obvious, indicating that after processing by the parameter-optimized variational mode algorithm, the fault-related characteristic components are further extracted from the background noise, and the interference signal in the whole frequency band is also well suppressed. This proves that the method used in this paper is superior to the results of VMD optimization by the BA algorithm, SSA algorithm and WOA algorithm.

[0097] Using the fitness value determined by the above algorithms, the original signal of the channel is then subjected to mode decomposition according to the returned parameter values. To quantitatively evaluate the performance of different algorithms, fault characteristic coefficients (FFC) and signal-to-noise ratios (SNR) are introduced as quantitative indicators for the modal components decomposed by the four algorithms to quantitatively analyze the bearing fault frequency components. The fault characteristic coefficients (FFC) obtained by optimizing the VMD parameters using different algorithms are compared, as shown in Table 3-1.

[0098] Table 3-1 Comparative analysis shows that the CS-GWO algorithm optimizes the VMD fault feature coefficient FFC to the largest, proving that it has the strongest feature extraction capability. In addition, the obtained signal contains more fault components, which reduces the weakening of fault feature energy. The largest signal-to-noise ratio indicates that this method can more effectively reduce the influence of noise in the signal and has stronger noise robustness.

[0099] Furthermore, the dynamic simulation model uses the Rexnord ZA-2115 double-row bearing dataset, divided into a group (dataset I), which includes normal conditions and three fault conditions, to construct the source domain dataset. The target domain dataset uses the University of Cincinnati bearing dataset, divided into a group (dataset II), which includes normal conditions and three fault conditions, to construct part of the target domain data.

[0100] Table 4-1 To verify the effectiveness of the proposed migration algorithm, migration tasks with different fault types were set up on different datasets, i.e., cross-operating condition migration fault diagnosis experiments. Taking migration task I→II as an example, I is the source domain dataset and II is the target domain dataset. The specific composition of the experimental data for this migration task is shown in Table 4-1.

[0101] Based on the simulation results and the data length of the CWRU bearing samples, the classification labels in each task were divided into 300 data samples, for a total of 1200 data samples, with a sample length of 1024. Each sample underwent data preprocessing, converting it into a two-dimensional grayscale image. The source domain dataset and 80% of the bearing data from the target domain were randomly assigned to the training set, while the remaining 20% ​​were assigned to the test set.

[0102] To verify the effectiveness of the proposed algorithm, it is compared with Transfer Component Analysis (TCA), Convolutional Neural Network (CNN), CNN-MMD, CNN-CORAL, and our algorithm.

[0103] Table 4-2 This method selected various combinations of convolutional and pooling layers for comparative experiments. The highest accuracy, approaching 100%, was achieved with 5 convolutional layers and 2 pooling layers. Therefore, 5 convolutional layers and 2 pooling layers were chosen. The final structural parameters of the deep convolutional neural network feature extraction module selected for the experiment are shown in Table 5-5. The batch size for each iteration was 64, and the learning rate was 0.001.

[0104] Based on the cross-operating condition migration fault diagnosis experiment, a migration task was set up. The proposed algorithm was tested with five other algorithms on the model proposed by this method. The average value of the five results was taken. The fault diagnosis results are shown in Table 4-3.

[0105] Table 4-3 Based on the data in Table 4-3, the following conclusions can be drawn: (1) For the migration task I→II, the highest average fault diagnosis accuracy of this algorithm reached 99.89%, which is better than all the comparison algorithms.

[0106] (2) In this experiment, TCA lost useful distribution information due to its weak ability to measure distribution differences, resulting in the lowest accuracy of the algorithm. CNN can learn distribution features from two-dimensional signals, with an average diagnostic rate of 78.16%. The reason for this is that the distribution of the training set data and the test set data are quite different, indicating that CNN needs to rely on transfer learning to obtain higher accuracy.

[0107] (3) CNN-MMD only calculates the distance loss between the source domain data and the target domain data, and only obtains a mean diagnostic accuracy of 90.14%. CNN-CORAL uses the CORAL loss function in the convolutional layer, which reduces the feature distribution difference between the data in different domains, and further reduces the feature size and feature range learned from the data in different domains. The diagnostic results are better than CNN-MMD, and the mean diagnostic accuracy is 90.43%.

[0108] (4) Although LATL has a complex network structure and alternately calculates two loss functions, its training time is too long and cannot meet the real-time requirements of the monitoring system in the application scenario. In the experiment set by this method, its fault diagnosis accuracy was 92.93%, which is 6.96% lower than this algorithm.

[0109] (5) The fault diagnosis accuracy of LATL without the application of feature extraction algorithm for noise reduction is only 95.86%, which is lower than that of LATL with the application of feature extraction algorithm. This is because the target domain data is denoised, which reduces the feature differences between domain data, reduces the recognition difficulty of transfer learning, and thus improves the accuracy of fault diagnosis.

[0110] The process and results of the transfer task I→II based on this algorithm are visualized. The time-domain waveforms and grayscale images of the source and target domains of transfer task I→II are shown in Figure 21.

[0111] As can be seen from Figure 21, since the source domain dataset I is generated by constructing dynamic equations, its grayscale image also shows regular stripes. There are certain differences between the time domain waveforms and grayscale images of the source domain dataset I and the target domain dataset II, indicating that the data in the different domains have different distribution characteristics.

[0112] The accuracy of the transfer results is shown in Table 4-3, and the confusion matrix of the algorithm transfer results is as follows: Figure 22 As shown in the figure, the classification accuracy of this algorithm is high, basically reaching 100%, effectively reducing the number of confused samples and correctly classifying samples that were misclassified in other algorithms.

[0113] Figure 23 The figure shows the t-SNE visualization results of a single trial for transfer task I→II, with a classification accuracy of 100%. As can be seen from the figure, the features extracted using this algorithm exhibit significant discriminative power across the four fault states. This demonstrates that the algorithm effectively ensures feature separability between different fault categories while reducing the differences in feature distribution between data from different domains, i.e., reducing intra-class distance, thereby improving the model's fault diagnosis and recognition rate.

[0114] In summary, the full-life rolling bearing fault diagnosis method based on mechanism data migration proposed in this embodiment has at least the following technical effects: High matching degree of source domain data: By constructing a dynamic mechanism model consistent with the bearing parameters of the full life test, the generated source domain data and target domain data feature distribution have small differences, which lays the foundation for efficient knowledge transfer and solves the problem of mismatch between traditional source domain data and actual data.

[0115] Excellent feature extraction accuracy: The CS-GWO algorithm is used to optimize VMD parameters, effectively avoiding mode aliasing. Combined with correlation analysis to screen effective modes, the accuracy of fault feature extraction and noise resistance are significantly improved. The fault feature coefficient (FFC) reaches 5.28% and the signal-to-noise ratio (SNR) reaches 12.94.

[0116] Excellent transfer learning results: By alternating between the CORAL loss function and the MMD loss function, the complementary advantages of the two are fully utilized, reducing the differences in distribution between domains and the intra-class distance, and the diagnostic accuracy is as high as 99.89%, which is far superior to traditional TCA, CNN and single loss function transfer learning algorithms.

[0117] Highly practical for engineering applications: It does not require failure samples in the entire life cycle scenario, and can achieve accurate diagnosis through mechanism simulation data alone. It is suitable for the full life cycle condition monitoring and fault diagnosis of rolling bearings in various rotating machinery.

[0118] Based on the same concept, embodiments of the present invention also provide a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement a mechanism-based data migration-based full-life rolling bearing fault diagnosis method provided in embodiments of the present invention.

[0119] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0120] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0121] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0122] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0123] Based on the same concept, embodiments of the present invention also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements a full-life rolling bearing fault diagnosis method based on mechanism data migration provided in embodiments of the present invention.

[0124] Computer program products may be loaded onto computer devices, and the components of computer devices may include, but are not limited to: one or more processors or processing units, system memory, and buses connecting different system components (including system memory and processing units).

[0125] Computer devices typically include a variety of computer system-readable media. These media can be any available media that can be accessed by a computer device, including volatile and non-volatile media, and removable and non-removable media.

[0126] System memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The computer device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media. The computer program product has a set (e.g., at least one) of program modules configured to perform the functions of the various embodiments of the present invention.

[0127] A program / utility having a set (at least one) of program modules can be stored, for example, in memory. Such program modules include, but are not limited to, an operating system, one or more applications, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules typically perform the functions and / or methods described in the embodiments of this invention.

[0128] Computer devices can also communicate with one or more external devices (such as keyboards, pointing devices, monitors, etc.), one or more devices that enable users to interact with the computer device, and / or any device that enables the computer device to communicate with one or more other computing devices (such as network interface cards, modems, etc.). This communication can be performed through input / output (I / O) interfaces. Furthermore, computer devices can communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapters. As shown in the figure, the network adapter communicates with other modules of the computer device via a bus. It should be understood that other hardware and / or software modules can be used in conjunction with the computer device, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0129] The processing unit executes various functional applications and data processing by running programs stored in the system memory, such as implementing a mechanism-based data migration-based full-life rolling bearing fault diagnosis method provided in this embodiment of the invention.

[0130] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0131] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for full-life rolling bearing fault diagnosis based on mechanism data migration, characterized in that, include: The rolling bearing is constructed as a mass-spring-damping model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator; Based on Hertz contact theory, the nonlinear dynamic equation of the rolling bearing is obtained by modifying the radial displacement excitation function of the rolling element, the time-varying contact stiffness between the rolling element and the bearing raceway, and the impact force at the contact defect of the rolling element in the mass-spring-damping model. The nonlinear dynamic equations of rolling bearings are solved by the fourth-order Runge-Kutta method, generating source domain data consistent with the test parameters of full-life rolling bearings, and constructing a source domain dataset with complete fault labels. Based on the source domain dataset and the pre-built target domain dataset, train a pre-built improved alternating transfer learning model; A trained improved alternating transfer learning model is used to perform full-life rolling bearing fault diagnosis, and the diagnosis results are output. The target domain dataset covers data samples from rolling bearing tests throughout the entire lifespan of the rolling bearing, from normal operation to failure, including sample data for four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault. The source domain dataset includes simulation data for the four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault.

2. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 1, characterized in that, The nonlinear dynamic equation of the rolling bearing is: ； The first two lines are the equations of motion for the inner circle, where, For the quality of the inner ring, This is the damping coefficient between the inner ring and the bearing housing. This refers to the support stiffness between the inner ring and the bearing housing. , These represent the displacements of the inner ring in the x and y directions, respectively. , For speed, , For acceleration, , F represents the x and y components of the contact force between the inner ring and the rolling element, respectively. r Where is the bearing load, and e is the eccentricity of the inner ring. The rotational angular velocity of the inner ring; The third and fourth lines are the equations of motion for the outer circle, where... For the quality of the outer ring, , These represent the support stiffness and damping coefficient between the outer ring and the bearing housing, respectively. , These represent the horizontal and vertical displacements of the shaft and the outer ring, respectively. , These represent the displacements of the center of the rolling element in the x and y directions, respectively. , These are the outer ring fault excitation forces in the x and y directions, respectively; The fifth and sixth lines are the equations of motion for the rolling element, in which... For the mass of a single rolling element; The last line is the equation of motion for the cage, where, To maintain rack quality, Let J be the cage angular displacement corresponding to the j-th rolling element. This represents the extreme diameter of the j-th rolling element. To maintain the angular velocity of the frame, This indicates the angular position of the j-th rolling element. , The Hertzian contact force between the inner and outer rings of the bearing on the rolling elements in the radial direction.

3. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 2, characterized in that, The radial displacement excitation function of the rolling element in the mass-spring-damped model is modified based on Hertz contact theory, specifically including: Cage speed Rotational angular velocity of the inner ring The relationship between them is: d is the rolling element diameter; D is the bearing pitch diameter; β is the contact angle; ； In the above formula, Z represents the number of rolling elements, and j is a natural number greater than or equal to 1. The angular position of the first rolling element relative to the positive y-axis at time zero; The expression for the radial displacement excitation function s of the rolling element abruptly entering the defect region is: ， ； In the above formula, R is the radius of the bearing outer ring raceway; r is the radius of the rolling element. The angle at which the rolling element enters the defect area; , These are the angular positions of the front and rear edges of the defect area, respectively. The radial displacement of the contact deformation between the j-th rolling element and the inner ring raceway of the bearing is: ; This refers to the radial clearance of the bearing; Radial displacement of the contact deformation between the j-th rolling element and the outer ring raceway of the bearing Revised to: 。 4. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 3, characterized in that, The time-varying contact stiffness between the rolling elements and the bearing raceway is modified based on Hertz contact theory, specifically including: The formula for the Hertzian contact force between the inner and outer rings of the bearing on the rolling element in the radial direction is as follows: ， ； In the formula: ， ， ， ， In the formula, , These are the stiffness coefficients of the inner and outer raceways of the bearing, respectively; n is the load-deformation coefficient. , These represent the relative displacement vectors of the j-th rolling element and the raceways of the inner and outer rings of the bearing, respectively. The formula for calculating time-varying contact stiffness is as follows: ， ； In the formula, This is the corrected time-varying contact stiffness coefficient. S is the reduction factor, and S is the shape factor; It refers to the range of banding.

5. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 4, characterized in that, The impact force at the rolling element contact defect is corrected based on Hertz contact theory, specifically including: Impact force generated at the edge of the rolling element after impacting the defect The calculation formula is: L is the defect width: m b For the mass of the rolling element; Will Orthogonal decomposition yields horizontal and vertical components: , : The angle between the impact force and the radial velocity of the rolling element: ； In the formula, The discriminant function for determining the presence of impact force is defined as follows: ； The magnitudes of the total Hertzian contact forces acting on the inner and outer rings of the rolling bearing are as follows: ， 。 6. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 1, characterized in that, The construction of the rolling bearing as a mass-spring-damped model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator specifically includes: For rolling bearings used in full-life testing, the rolling bearing system is simplified into a mass-spring-damped model consisting of an outer ring, an inner ring, rolling elements, and a high-frequency resonator, and the model is assumed to satisfy the following conditions: The outer ring is fixed to a rigid structure, while the inner ring maintains a constant rotational speed. The rolling elements are evenly distributed with no relative slippage; The effect of the rolling element's rotation is not considered; The inner and outer rings and the rolling elements move in the same plane.

7. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 2, characterized in that, The process involves solving the nonlinear dynamic equations of rolling bearings using the fourth-order Runge-Kutta method, generating source domain data consistent with full-life rolling bearing test parameters, and constructing a source domain dataset with complete fault labels. Specifically, this includes: The nonlinear dynamic equations of rolling bearings are solved using the fourth-order Runge-Kutta method, generating simulation data for four types of faults: normal state, inner ring fault, outer ring fault, and rolling element fault. Fault simulation is performed based on the generated simulation data, and a source domain dataset with complete fault labels is constructed.

8. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 1, characterized in that, Before training the pre-built improved alternating transfer learning model based on the source domain dataset and the pre-built target domain dataset, the process also includes: Construct the target domain dataset; the method is as follows: Collect full-life rolling bearing test data; wherein, the full-life rolling bearing test data covers the entire process of the bearing from normal operation to failure, including axial and radial vibration signals; The collected full-life rolling bearing test data were preprocessed to construct an original dataset containing four types of fault samples: normal state, inner ring fault, outer ring fault, and rolling element fault. The amplitude spectrum of the fault samples is selected as the fitness function of the CS-GWO algorithm, and the minimum value of the amplitude spectrum entropy is used as the fitness value to optimize the mode number K and the quadratic penalty factor in the VMD algorithm. ; Based on the optimized mode number K and the quadratic penalty factor K modal components are obtained by decomposing using the VMD algorithm; Multiple modal components with the highest correlation to the fault samples were selected as effective modes, summed and reconstructed, and envelope spectrum analysis was performed on the reconstructed modal components to extract feature information related to the rolling bearing faults, forming sample data.

9. The method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 8, characterized in that, The method for setting the loss function of the improved alternating transfer learning model is as follows: The distance between the second-order statistics of the features in the source and target domains is defined as the CORAL loss function L1: , is the Frobenius norm of the mean square matrix; in, Let D be the number of dimensions of this sample, and D be the source domain dataset. S , and the target domain dataset D T The characteristic covariance matrices are C S With C T ; MMD is used to measure the feature set D. S and D T The distributional differences in the reproducing kernel Hilbert space are used as the sum of the MMD loss and the classification loss as the L2 loss function: ， ， , For the regenerating nucleus Hilbert space; Among them, L C The classification loss is the difference between the true label and the predicted label. , Mapping function , n is the balance coefficient. S Let y be the number of samples, and y be the true label of the sample data. The label predicted by the classifier.

10. A method for full-life rolling bearing fault diagnosis based on mechanism data migration according to claim 8, characterized in that, The formula for calculating amplitude spectral entropy is: ； In the above formula, L i For modal components u i Amplitude spectrum; H i For modal components u i The amplitude spectral entropy; N is the length of the modal component.

Images