Train bearing fault diagnosis method, system and device based on data fusion and medium

Abstract

The application discloses a train bearing fault diagnosis method, system and device based on data fusion and a medium, and comprises the following steps: a bearing fault dynamics model is constructed, bearing simulation vibration signals are further acquired, and the fault characteristic frequency of the bearing fault dynamics model is acquired; the fault characteristic frequency of the bearing fault dynamics model is fused with the fault characteristic frequency of a real fault bearing test piece, simulation fusion fault data are acquired; the simulation fusion fault data frequency domain signals are input into a convolutional neural network for pre-training, migration learning is performed on real test fault data based on a parameter migration strategy, a bearing fault diagnosis model is constructed, and the fault position of a train bearing is acquired based on the optimized bearing fault diagnosis model. The application achieves the purpose of reducing the feature distribution difference between the source domain and the target domain by fusing simulation data and real data, and the mechanism simulation data are used for assisting in training the model, so that the generalization ability, stability and accuracy of the convolutional neural network model are improved.

Description

Technical Field

[0001] This invention belongs to the field of bogie bearing fault technology, and relates to a train bearing fault diagnosis method, system, device and medium based on data fusion. Background Technology

[0002] Bearings, as indispensable key components in mechanical equipment, are receiving increasing attention from researchers for their condition monitoring and fault diagnosis. In the context of the big data era and the rapid development of artificial intelligence technology, deep learning has been applied to intelligent fault diagnosis fields such as signal processing, feature extraction, fault classification, and life prediction. However, the excellent diagnostic performance of many deep learning models relies on a large number of labeled fault data samples. In actual train operation, the preventative and conservative maintenance strategy makes it difficult to obtain sufficient bearing fault data to train diagnostic models, limiting the application of deep learning methods in high-speed train fault diagnosis. Therefore, improving the diagnostic performance of bearing fault diagnosis with small sample sizes in high-speed trains has a promising future.

[0003] In recent years, researchers have mainly used methods to solve small sample problems, including optimizing network structure, data augmentation, and transfer learning. Currently, most of the fault data used in bearing fault diagnosis comes from high-fidelity fault diagnosis test benches. Building these test benches requires significant manpower, material resources, and financial investment. Due to safety limitations, some fault types and operating conditions cannot be simulated. Using generative adversarial neural networks to generate samples with the same distribution as real data lacks interpretability. The development of industrial big data necessitates the use of mechanistic knowledge to assist in training fault diagnosis models. Simulation data differs significantly from real data. Due to the influence of the external environment and different equipment, real data contains noise, and there are differences in the data distribution between the source and target domains. Summary of the Invention

[0004] The purpose of this invention is to solve the problems in the existing technology where setting up a test bench consumes a lot of resources and some fault types and working conditions cannot be simulated; the use of generative adversarial neural networks to generate samples with the same distribution as real data has poor interpretability and the simulation data differs greatly from the real data. The invention provides a train bearing fault diagnosis method, system, device and medium based on data fusion.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] A data fusion-based method for diagnosing train bearing faults includes:

[0007] A bearing failure dynamics model was constructed, and parameter information of the failed bearing specimens was collected.

[0008] Based on the bearing fault dynamics model and the parameter information of the faulty bearing specimen, the bearing simulation vibration signal is obtained, thereby obtaining the fault characteristic frequency of the bearing fault dynamics model.

[0009] Vibration response information of real faulty bearing specimens was collected, and the fault characteristic frequencies of the real faulty bearing specimens were obtained.

[0010] The simulation and experimental data fusion method is used to fuse the fault characteristic frequencies of the bearing fault dynamics model with the fault characteristic frequencies of the real fault bearing specimen to obtain the simulation fused fault data.

[0011] By comparing the similarity between the frequency domain envelope diagram of the simulated fused data and the frequency domain envelope diagram of the vibration response information of the real faulty bearing specimen, the optimal data fusion strategy is selected.

[0012] The frequency domain signal of the simulated fused fault data is input into the convolutional neural network for pre-training. Based on the parameter transfer strategy, the fault data of the real test is transferred to learn and construct a bearing fault diagnosis model.

[0013] Based on the optimized bearing fault diagnosis model, the fault location of the train bearing is obtained.

[0014] A further improvement of the present invention is that:

[0015] Furthermore, a bearing failure dynamic model is constructed, specifically: the bearing failure dynamic model is established as a 4-DOF bearing dynamic model, which includes an outer ring, an inner ring, rolling elements, and an input shaft; several rolling elements are arranged between the outer ring and the inner ring, and corresponding raceways are arranged on the inner side of the outer ring and the outer side of the inner ring; the rolling elements are arranged between the raceways, and the inner ring rotates with the input shaft; at the same time, the following assumptions are made: (1) the outer ring of the bearing is fixed, and the inner ring rotates with the input shaft; (2) the contact between the rolling elements and the raceways is Hertzian based; (3) the nonlinearity of the bearing is caused by the nonlinear contact force between different components and the gap between the rolling elements and the inner and outer rings; (4) the rolling elements are uniformly distributed, and there is no slippage during the rolling process.

[0016] Furthermore, the 4-DOF bearing dynamics model is considered as two subsystems with two degrees of freedom each in the x and y directions; according to Hertzian contact theory, the contact force F between the rolling element and the raceway... H for:

[0017] F H =Kδ 1.5 (1)

[0018] Where K is stiffness, δ is elastic approximation; for a certain rolling element, the elastic deformation δ i Represented as:

[0019] δ i =(x in -x out sinθ i+(y in -y out cosθ i -c r -H (2)

[0020] Where, x in and x out Represented as the displacement of the inner and outer rings in the x-direction, y-direction... in and y out c represents the displacement of the inner and outer rings in the y-direction. r The bearing clearance is represented by H, the bearing displacement under different fault types, and θ is the bearing displacement. i Let be the angular position of the i-th rolling element at time t;

[0021] The total contact force F of the z rolling elements H The components F in the x and y dimensions Hx and F Hy It can be represented as,

[0022]

[0023]

[0024] Based on the analysis of the bearing forces, and considering the centrifugal force caused by installation and manufacturing deviations, the simplified dynamic equations of the rolling bearing 4-DOF system are as follows:

[0025]

[0026]

[0027]

[0028]

[0029] Where, m in k is the mass of the inner ring and the shaft. in For the stiffness of the inner ring, c in For the damping of the inner ring, m out For the mass of the outer ring, k out For the stiffness of the outer ring, c out ε is the damping of the outer ring, and e is the eccentricity.

[0030] Furthermore, the bearing failure dynamics model also includes: introducing a fault controlled by parameter H in the dynamics model, simplifying the bearing failure into a local rectangular damage, and the rolling element experiencing time-varying displacement excitation and instantaneous impact force when passing through the local damage.

[0031] When the bearing is operating under normal conditions, H = 0;

[0032] When there is a fault in the outer ring, H is represented as:

[0033]

[0034] Where θ i Let be the angular position of the i-th rolling element at time t. For the location of the fault, ψ out The angle of the fault width. r out Let H0 be the radius of the outer ring, and H0 be the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. According to geometric relationships, this can be expressed as:

[0035]

[0036] When there is a fault in the inner ring, H represents...

[0037]

[0038] The location of the inner ring fault changes with the rotation of the shaft, φ in =ωt,ψ in The angle of the fault width. r in Let H0 be the radius of the inner ring, and H0 be the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. According to geometric relationships, this can be expressed as:

[0039]

[0040] When the rolling element is faulty, H is represented as:

[0041]

[0042] When the damaged rolling element is the k-th rolling element, the damaged area of ​​the rolling element contacts both the outer and inner rings twice each, and the location of the rolling element failure changes with the rotation of the rolling element. ψ bo The angular width of the rolling damage area in contact with the outer ring. ψ bi The angular width of the rolling damage area in contact with the inner ring. H0 represents the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. Ignoring the influence of the inner ring curvature on the displacement change based on geometric relationships, it is expressed as:

[0043]

[0044] Furthermore, based on the bearing fault dynamics model and the parameter information of the faulty bearing specimen, the bearing simulation vibration signal is obtained, thereby obtaining the fault characteristic frequency of the bearing fault dynamics model, specifically:

[0045] The parameter information of the faulty bearing specimen includes: the parameters of the faulty bearing specimen, the rotational speed, the fault type, and the sampling frequency.

[0046] Input the parameters, rotational speed, fault type, and sampling frequency of the faulty bearing specimen to obtain the bearing simulation vibration signal. Plot the time domain diagram of the bearing simulation vibration signal using MATLAB. Use the Hilbert envelope algorithm to obtain the frequency domain envelope diagram of the bearing simulation vibration signal to obtain the fault characteristic frequency of the bearing fault dynamic model. Calculate the theoretical fault characteristic frequency of the bearing specimen. Verify the accuracy of the dynamic model by comparing the fault characteristic frequency of the bearing dynamic model with the theoretical fault characteristic frequency.

[0047] The calculation of the theoretical fault characteristic frequency of the bearing specimen is specifically as follows:

[0048] The characteristic frequency of bearing outer ring failure is shown in formula (15):

[0049]

[0050] The characteristic frequency of bearing inner ring failure is shown in formula (16):

[0051]

[0052] The characteristic frequencies of bearing rolling element failures are shown in formula (17):

[0053]

[0054] Among them, f r The input shaft frequency is z, the number of rolling elements is d, and the diameter of the rolling elements is d. m Where α is the bearing pitch diameter and α is the bearing contact angle.

[0055] Furthermore, vibration response information of real faulty bearing specimens was collected to obtain the fault characteristic frequencies of the real faulty bearing specimens. Specifically, vibration signals of real faulty bearing specimens from trains were collected, and a time-domain graph of the vibration response was constructed using Matlab software. A frequency-domain envelope graph was generated using the Hilbert envelope algorithm to obtain the fault characteristic frequencies of the real faulty bearings. These frequencies were then compared with the fault characteristic frequencies of the bearing dynamics model to verify the effectiveness of the dynamics model in guiding the diagnosis of real bearing faults.

[0056] Furthermore, based on the simulation and experimental data fusion method, the fault characteristic frequencies of the bearing fault dynamics model are fused with the fault characteristic frequencies of the real fault bearing specimens to obtain simulation fused fault data. Specifically, the simulation and experimental data fusion method includes weighted fusion and convolutional fusion.

[0057] The weight fusion involves normalizing the fault characteristic frequencies of the bearing fault dynamics model and the fault characteristic frequencies of the actual faulty bearing specimens, and then superimposing them with different weighting ratios in the time domain, as shown in formula (18):

[0058] D(t)=α*Nor(Sim(t))+β*Nor(u(t)) (18)

[0059] Where D(t) is the fused signal, α and β are weights and α+β=1, Nor(x) is the normalization process, Sim(t) is the bearing fault simulation data, and u(t) is the bearing normal data from the actual test.

[0060] The convolution fusion process involves normalizing the simulated fault data and the normal data from the actual experiment, followed by convolution. The convolved signal exhibits a similar distribution to both the simulation and the experiment in the time and frequency domains, and its duration becomes twice that of the original signal, as shown in formula (19).

[0061] C(t)=conv(Nor(Sim(t)),Nor(u(t))) (19)

[0062] Where C(t) is the convolutional fusion signal, Nor(x) is the normalization process, Sim(t) is the bearing fault simulation data, and u(t) is the bearing normal data from the actual test.

[0063] A train bearing fault diagnosis system based on data fusion includes:

[0064] The module constructs a bearing failure dynamics model and collects parameter information of the failed bearing specimen;

[0065] The first acquisition module acquires bearing simulation vibration signals based on the bearing fault dynamics model and parameter information of the faulty bearing specimen, thereby acquiring the fault characteristic frequency of the bearing fault dynamics model.

[0066] The second acquisition module collects vibration response information of real faulty bearing specimens and acquires the fault characteristic frequencies of real faulty bearing specimens.

[0067] The fusion module, based on the simulation and experimental data fusion method, fuses the fault characteristic frequencies of the bearing fault dynamics model with the fault characteristic frequencies of the real faulty bearing specimen to obtain simulated fused fault data.

[0068] The comparison module compares the similarity between the frequency domain envelope diagram of the simulated fused data and the frequency domain envelope diagram of the vibration response information of the real faulty bearing specimen, and selects the optimal data fusion strategy.

[0069] The pre-training module inputs the frequency domain signal of the simulated fused fault data into the convolutional neural network for pre-training, and performs transfer learning on the fault data of the real test based on the parameter transfer strategy to build a bearing fault diagnosis model.

[0070] The third acquisition module acquires the fault location of the train bearing based on the optimized bearing fault diagnosis model.

[0071] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0072] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] This invention constructs a bearing fault dynamics model to obtain sufficiently labeled simulated fault data, and verifies the accuracy of the constructed dynamics model using a bearing dataset. A method for fusing simulated data and normal experimental data is proposed to reduce the impact of feature distribution differences between simulated and experimental data on the transfer effect. The diagnostic model is pre-trained in the frequency domain using the fused data; the frequency domain signal contains bearing fault frequency features, which is more conducive to the neural network learning the different features between different fault data. A parameter transfer strategy is employed to train the diagnostic model with a small amount of experimental data. This invention can fuse simulated and real data to reduce the feature distribution differences between the source and target domains, and uses mechanistic simulation data to assist in model training, which is of great significance for improving the generalization ability, stability, and accuracy of convolutional neural network models; it also solves the problem of the scarcity of high-quality fault sample data in actual high-speed train operation. Attached Figure Description

[0075] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0076] Figure 1 This is a schematic diagram of the data fusion-based train bearing fault diagnosis method of the present invention.

[0077] Figure 2a This is a diagram of the 4-DOF bearing dynamics model of the present invention;

[0078] Figure 2b This is a force analysis diagram of the 4-DOF bearing dynamics model of the present invention;

[0079] Figure 2c This is a geometric diagram showing the fault relationships of the outer ring and rolling elements of the present invention;

[0080] Figure 3a This is a frequency domain diagram of the outer ring fault simulation data of the present invention;

[0081] Figure 3b This is a time-domain plot of the outer ring fault simulation data of the present invention;

[0082] Figure 4 This is a framework diagram of the high-speed train bearing fault diagnosis method driven by the fusion of simulation and experimental data of the present invention.

[0083] Figure 5a This is a schematic diagram of the structure and principle of the rolling test bench for vehicle bearing failure according to the present invention;

[0084] Figure 5b The bearing position is shown in the rolling test bench for vehicle bearing failure of the present invention.

[0085] Figure 6a This is a frequency domain diagram of the outer ring fault test data of the present invention;

[0086] Figure 6b This is a time-domain plot of the outer ring fault test data of the present invention;

[0087] Figure 7a This is the frequency domain diagram of the outer ring fault 25% weighted fused data of the present invention;

[0088] Figure 7b This is a frequency domain plot of experimental data from the outer ring fault 50% weighted fusion data of the present invention;

[0089] Figure 7c This is a frequency domain plot of experimental data from the outer ring fault 75% weighted fusion data of the present invention;

[0090] Figure 7d This is a frequency domain plot of the experimental data from the outer ring fault convolutional fusion data of the present invention;

[0091] Figure 8 This is the confusion matrix diagram of the present invention;

[0092] Figure 9 This is a schematic diagram of the train bearing fault diagnosis system based on data fusion according to the present invention. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0094] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0095] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0096] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0097] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0098] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0099] The present invention will now be described in further detail with reference to the accompanying drawings:

[0100] See Figure 1 This invention discloses a train bearing fault diagnosis method based on data fusion, comprising:

[0101] S101, Construct a bearing failure dynamics model and collect parameter information of the failed bearing specimen;

[0102] The bearing failure dynamics model is established as a 4-DOF bearing dynamics model, which includes an outer ring, an inner ring, rolling elements, and an input shaft. Several rolling elements are arranged between the outer ring and the inner ring, and corresponding raceways are arranged on the inner side of the outer ring and the outer side of the inner ring. The rolling elements are arranged between the raceways, and the inner ring rotates with the input shaft. The following assumptions are made: (1) The outer ring of the bearing is fixed, and the inner ring rotates with the input shaft; (2) The contact between the rolling elements and the raceways is Hertzian based; (3) The nonlinearity of the bearing is caused by the nonlinear contact force between different components and the gap between the rolling elements and the inner and outer rings; (4) The rolling elements are uniformly distributed, and there is no slippage during the rolling process.

[0103] The 4-DOF bearing dynamics model is considered as two subsystems with two degrees of freedom each in the x and y directions; according to Hertzian contact theory, the contact force F between the rolling element and the raceway... H for:

[0104] F H =Kδ 1.5 (1)

[0105] Where K is stiffness, δ is elastic approximation; for a certain rolling element, the elastic deformation δ i Represented as:

[0106] δ i =(x in -x out sinθ i +(y in -y out cosθ i -c r -H (2)

[0107] Where, x in and x out Represented as the displacement of the inner and outer rings in the x-direction, y-direction... in and y out c represents the displacement of the inner and outer rings in the y-direction. r The bearing clearance is represented by H, the bearing displacement under different fault types, and θ is the bearing displacement. i Let be the angular position of the i-th rolling element at time t;

[0108] The total contact force F of the z rolling elements H The components F in the x and y dimensions Hx and F Hy It can be represented as,

[0109]

[0110]

[0111] Based on the analysis of the bearing forces, and considering the centrifugal force caused by installation and manufacturing deviations, the simplified dynamic equations of the rolling bearing 4-DOF system are as follows:

[0112]

[0113]

[0114]

[0115]

[0116] Where, m in k is the mass of the inner ring and the shaft. in For the stiffness of the inner ring, c in For the damping of the inner ring, m out For the mass of the outer ring, k out For the stiffness of the outer ring, c out ε is the damping of the outer ring, and e is the eccentricity.

[0117] The bearing failure dynamics model also includes: introducing a fault controlled by parameter H in the dynamics model, simplifying the bearing failure into a local rectangular damage, and the rolling element experiencing time-varying displacement excitation and instantaneous impact force when passing through the local damage.

[0118] When the bearing is operating under normal conditions, H = 0;

[0119] When there is a fault in the outer ring, H is represented as:

[0120]

[0121] Where θ i Let be the angular position of the i-th rolling element at time t. For the location of the fault, ψ out The angle of the fault width. r out Let H0 be the radius of the outer ring, and H0 be the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. According to geometric relationships, this can be expressed as:

[0122]

[0123] When there is a fault in the inner ring, H represents...

[0124]

[0125] The location of the inner ring fault changes with the rotation of the shaft, φ in =ωt,ψ in The angle of the fault width. r in Let H0 be the radius of the inner ring, and H0 be the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. According to geometric relationships, this can be expressed as:

[0126]

[0127] When the rolling element is faulty, H is represented as:

[0128]

[0129] When the damaged rolling element is the k-th rolling element, the damaged area of ​​the rolling element contacts both the outer and inner rings twice each, and the location of the rolling element failure changes with the rotation of the rolling element. ψ bo The angular width of the rolling damage area in contact with the outer ring. ψ bi The angular width of the rolling damage area in contact with the inner ring. H0 represents the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. Ignoring the influence of the inner ring curvature on the displacement change based on geometric relationships, it is expressed as:

[0130]

[0131] S102, based on the bearing fault dynamics model and the parameter information of the faulty bearing specimen, obtain the bearing simulation vibration signal, thereby obtaining the fault characteristic frequency of the bearing fault dynamics model.

[0132] The parameter information of the faulty bearing specimen includes: the parameters of the faulty bearing specimen, the rotational speed, the fault type, and the sampling frequency;

[0133] Input the parameters, rotational speed, fault type, and sampling frequency of the faulty bearing specimen to obtain the bearing simulation vibration signal. Plot the time domain diagram of the bearing simulation vibration signal using MATLAB. Use the Hilbert envelope algorithm to obtain the frequency domain envelope diagram of the bearing simulation vibration signal to obtain the fault characteristic frequency of the bearing fault dynamic model. Calculate the theoretical fault characteristic frequency of the bearing specimen. Verify the accuracy of the dynamic model by comparing the fault characteristic frequency of the bearing dynamic model with the theoretical fault characteristic frequency.

[0134] The calculation of the theoretical fault characteristic frequency of the bearing specimen is specifically as follows:

[0135] The characteristic frequency of bearing outer ring failure is shown in formula (15):

[0136]

[0137] The characteristic frequency of bearing inner ring failure is shown in formula (16):

[0138]

[0139] The characteristic frequencies of bearing rolling element failures are shown in formula (17):

[0140]

[0141] Among them, f r The input shaft frequency is z, the number of rolling elements is d, and the diameter of the rolling elements is d. m Where α is the bearing pitch diameter and α is the bearing contact angle.

[0142] S103, collect vibration response information of real faulty bearing specimens and obtain the fault characteristic frequencies of real faulty bearing specimens.

[0143] Vibration signals from actual faulty bearing specimens on trains were collected, and the time-domain plot of the vibration response was constructed using Matlab software. The frequency-domain envelope plot was generated using the Hilbert envelope algorithm to obtain the fault characteristic frequencies of the actual faulty bearings. These frequencies were compared with the fault characteristic frequencies of the bearing dynamics model to verify the effectiveness of the dynamics model in guiding the diagnosis of actual bearing faults.

[0144] S104, Based on the simulation and experimental data fusion method, the fault characteristic frequency of the bearing fault dynamics model is fused with the fault characteristic frequency of the real fault bearing specimen to obtain the simulation fused fault data.

[0145] The simulation and experimental data fusion method includes weighted fusion and convolutional fusion; the weighted fusion involves normalizing the fault characteristic frequencies of the bearing fault dynamics model and the fault characteristic frequencies of the actual faulty bearing specimens, and then superimposing them with different weights in the time domain, as shown in formula (18).

[0146] D(t)=α*Nor(Sim(t))+β*Nor(u(t)) (18)

[0147] Where D(t) is the fused signal, α and β are weights and α+β=1, Nor(x) is the normalization process, Sim(t) is the bearing fault simulation data, and u(t) is the bearing normal data from the actual test.

[0148] The convolution fusion process involves normalizing the simulated fault data and the normal data from the actual experiment, followed by convolution. The convolved signal exhibits a similar distribution to both the simulation and the experiment in the time and frequency domains, and its duration becomes twice that of the original signal, as shown in formula (19).

[0149] C(t)=conv(Nor(Sim(t)),Nor(u(t))) (19)

[0150] Where C(t) is the convolutional fusion signal, Nor(x) is the normalization process, Sim(t) is the bearing fault simulation data, and u(t) is the bearing normal data from the actual test.

[0151] S105. Compare the similarity between the frequency domain envelope diagram of the simulated fused data and the frequency domain envelope diagram of the vibration response information of the real faulty bearing specimen, and select the optimal data fusion strategy.

[0152] S106, The frequency domain signal of the simulated fused fault data is input into the convolutional neural network for pre-training, and the fault data of the real test is transferred to learn based on the parameter transfer strategy to build a bearing fault diagnosis model.

[0153] CNN is a type of feedforward neural network that has been widely used in the field of bearing fault diagnosis. It includes a feature extraction stage and a classification stage. Convolutional layers and pooling layers acquire nonlinear features of the input signal in different forms, while fully connected layers can effectively classify different faults.

[0154] Transfer learning transfers knowledge from one source domain (fusion of simulation data) to another target domain (real bearing fault diagnosis), enabling the target domain to achieve better learning results. Parameter-based transfer learning involves sharing some parameters or prior distributions of hyperparameters between models of related tasks. Unlike multi-task learning that simultaneously learns both source and target tasks, transfer learning allows us to apply additional weights to the target domain to improve overall performance.

[0155] Experimental data from a subset of real faulty bearing specimens were selected, and the fault diagnosis model was trained based on the frequency domain of the selected real experimental data. The trained fault diagnosis model was then validated based on experimental data from a subset of unselected real faulty bearing specimens to obtain the optimal bearing fault diagnosis model.

[0156] S107, based on the optimized bearing fault diagnosis model, obtains the fault location of the train bearing.

[0157] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0158] Using the bogie axle box bearing as the test object, the bearing parameters were obtained from actual vehicle disassembly. Nineteen rolling elements were disassembled from the actual vehicle, each with a diameter of 26 mm, a pitch circle diameter of 183.93 mm, a contact angle of 0°, an input shaft frequency of 616 r / min, an actual vehicle speed of 100 km / h at a constant speed, and a sampling frequency of 48 kHz. The bearing fault types were B1 normal, B2 rolling element fault, B3 outer ring peeling, and B4 outer ring corrosion. A 4-DOF bearing dynamic model was established, and the model diagram and local fault set relationships are shown in Figures 2(a), 2(b), and 2(c). Sufficient simulation data for normal operation, rolling element failure, outer ring peeling, and outer ring corrosion were obtained. The time domain plots and frequency domain envelopes are shown in Figure 3(a) and Figure 3(b). The input shaft frequency was calculated to be 10.3 Hz, the outer ring failure frequency was 83.7 Hz, the inner ring failure frequency was 111.3 Hz, the rolling element failure frequency was 71.2 Hz, and the cage failure frequency was 4.4 Hz.

[0159] See Figure 4 This invention discloses a framework for a high-speed train bearing fault diagnosis method driven by the fusion of simulation and experimental data. When conducting high-speed train bearing fault tests, the test system mainly consists of a whole-vehicle bearing fault rolling test bench, a disassembled faulty bearing from a real vehicle, and an on-board temperature and vibration composite detection system. The whole-vehicle bearing fault rolling test bench and the disassembled faulty bearing from a real vehicle highly replicate the vibration response of the real vehicle under actual conditions. Vibration signals are collected through the on-board temperature and vibration composite detection system. The structure and principle of the whole-vehicle bearing fault rolling test bench are shown in Figure 5(a). The test bench uses a certain model's lead car as the test vehicle, and the test vehicle has the same structure and similar load type as the actual train on the line. The two axles of the bogie in front of the test vehicle are replaced with disassembled faulty bearings from a real vehicle. The bearing installation positions are shown in Figure 5(b), facing the front of the train, from left to right and front to back: left axle 1, left axle 2, right axle 1, and right axle 2. The bearing fault types are B1 normal, B2 rolling element fault, B3 outer ring peeling, and B4 outer ring corrosion, respectively. The vibration response of the actual test was obtained, and its time domain plot and frequency domain envelope plot are shown in Figure 6(a) and Figure 6(b).

[0160] Based on the simulation and experimental data fusion method, weighted fused data with weights of α=0.25 and β=0.75, α=0.5 and β=0.5, and α=0.75 and β=0.25 were generated, referred to as 25% weighted fused data, 50% weighted fused data, and 75% weighted fused data, respectively. Convolutional fused data was also generated. Their frequency domain envelope diagrams are shown in Figures 7(a), 7(b), 7(c), and 7(d). The fault characteristics of the 25% weighted fused data and the convolutional fused data were masked by noise, making it difficult for the convolutional neural network to learn the feature distribution of the fault data. Both the 50% weighted fused data and the 75% weighted fused data significantly reflected the fault characteristic frequencies. To reduce the difference between simulation and experimental data and to include more equipment environmental noise in the fused data, this paper selected the 50% weighted fused data as the optimal fused data.

[0161] In this invention, the convolutional neural network feature extractor consists of two convolutional layers and two max-pooling layers, while the classifier consists of three fully connected layers. During transfer learning, the feature extractor directly reuses the pre-trained structural parameters, and the structural parameters of the classifier are fine-tuned. Sufficient 50% weighted fused data in the frequency domain is used as the source domain and input into the convolutional neural network for pre-training. A parameter transfer strategy is employed. The diagnostic model is trained in the frequency domain using 96 sets of real experimental data and validated in the frequency domain using 384 sets of real experimental data, achieving an accuracy of 99.21%. Its confusion matrix is ​​shown below. Figure 8 As shown.

[0162] See Figure 9 This invention discloses a train bearing fault diagnosis system based on data fusion, comprising:

[0163] The module constructs a bearing failure dynamics model and collects parameter information of the failed bearing specimen;

[0164] The first acquisition module acquires bearing simulation vibration signals based on the bearing fault dynamics model and parameter information of the faulty bearing specimen, thereby acquiring the fault characteristic frequency of the bearing fault dynamics model.

[0165] The second acquisition module collects vibration response information of real faulty bearing specimens and acquires the fault characteristic frequencies of real faulty bearing specimens.

[0166] The fusion module, based on the simulation and experimental data fusion method, fuses the fault characteristic frequencies of the bearing fault dynamics model with the fault characteristic frequencies of the real faulty bearing specimen to obtain simulated fused fault data.

[0167] The comparison module compares the similarity between the frequency domain envelope diagram of the simulated fused data and the frequency domain envelope diagram of the vibration response information of the real faulty bearing specimen, and selects the optimal data fusion strategy.

[0168] The pre-training module inputs the frequency domain signal of the simulated fused fault data into the convolutional neural network for pre-training, and performs transfer learning on the fault data of the real test based on the parameter transfer strategy to build a bearing fault diagnosis model.

[0169] The third acquisition module acquires the fault location of the train bearing based on the optimized bearing fault diagnosis model.

[0170] An embodiment of the present invention provides a terminal device. This terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the various method embodiments described above. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the various device embodiments described above.

[0171] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention.

[0172] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0173] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0174] The memory can be used to store the computer program and / or module. The processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0175] If the modules / units integrated into the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0176] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A train bearing fault diagnosis method based on data fusion, characterized in that, include: A bearing failure dynamic model was constructed, and parameter information of the failed bearing specimen was collected. Specifically, the bearing failure dynamic model was established as a 4-degree-of-freedom bearing dynamic model, which includes an outer ring, an inner ring, rolling elements, and an input shaft. Several rolling elements were set between the outer ring and the inner ring, and corresponding raceways were set on the inner side of the outer ring and the outer side of the inner ring. The rolling elements were set between the raceways, and the inner ring rotated with the input shaft. At the same time, the following assumptions were made: (1) The outer ring of the bearing is fixed, and the inner ring rotates with the input shaft; (2) The contact between the rolling elements and the raceways is Hertzian contact; (3) The nonlinearity of the bearing is caused by the nonlinear contact force between different components and the gap between the rolling elements and the inner and outer rings; (4) The rolling elements are uniformly distributed, and there is no slippage during the rolling process. Based on the bearing fault dynamics model and the parameter information of the faulty bearing specimen, the bearing simulation vibration signal is obtained, thereby obtaining the fault characteristic frequency of the bearing fault dynamics model. Vibration response information of real faulty bearing specimens was collected, and the fault characteristic frequencies of the real faulty bearing specimens were obtained. The simulation and experimental data fusion method is used to fuse the fault characteristic frequencies of the bearing fault dynamics model with the fault characteristic frequencies of the real fault bearing specimen to obtain the simulation fused fault data. By comparing the similarity between the frequency domain envelope diagram of the simulated fused data and the frequency domain envelope diagram of the vibration response information of the real faulty bearing specimen, the optimal data fusion strategy is selected. The frequency domain signal of the simulated fused fault data is input into the convolutional neural network for pre-training. Based on the parameter transfer strategy, the fault data of the real test is transferred to learn and construct a bearing fault diagnosis model. Based on the optimized bearing fault diagnosis model, the fault location of the train bearing is obtained.

2. The train bearing fault diagnosis method based on data fusion according to claim 1, characterized in that, The 4-DOF bearing dynamics model is considered as two subsystems with 2 degrees of freedom each in the x and y directions; according to Hertzian contact theory, the contact force between the rolling element and the raceway... F H for: in K For stiffness, For elastic approximation; elastic deformation of a certain rolling element. Represented as: in, and Represented as inner and outer circles x Displacement in the direction, and Represented as inner and outer circles y Displacement in the direction, This is expressed as bearing clearance. H This represents the bearing displacement under different fault types. For the first i A rolling element in t The angular position at any given moment; z Total contact force of each rolling element F H exist x and y Components in dimension F Hx and F Hy Represented as, Based on the analysis of the bearing forces, and considering the centrifugal force caused by installation and manufacturing deviations, the simplified dynamic equations of the rolling bearing 4-DOF system are as follows: in, For the mass of the inner ring and shaft, For the stiffness of the inner ring, For the inner ring damping, For the quality of the outer ring, For the stiffness of the outer ring, For the damping of the outer ring, e It is the eccentricity.

3. The train bearing fault diagnosis method based on data fusion according to claim 2, characterized in that, The bearing failure dynamics model further includes: introducing fault parameters into the dynamics model. H The control simplifies the bearing failure into a local rectangular damage. When the rolling element passes through the local damage, there is a time-varying displacement excitation and an instantaneous impact force. When the bearing is working under normal conditions ; When there is a fault in the outer ring H Represented as: in For the first i A rolling element in t Angular position at time, Location of the fault. The angle of the fault width. , The radius of the outer ring, The displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage region is expressed as follows, based on geometric relationships: When there is a fault in the inner ring H Represented as The location of the inner ring fault changes with the rotation of the shaft, φ in =ωt, The angle of the fault width. , Let be the radius of the inner circle. The displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage region is expressed as follows, based on geometric relationships: When the rolling element is faulty H Represented as: When the damaged rolling element is the first k When there is one rolling element, the damaged area of ​​the rolling element contacts both the outer and inner rings twice each, and the location of the rolling element failure changes with the rotation of the rolling element. , , The angular width of the rolling damage area in contact with the outer ring. , The angular width of the rolling damage area in contact with the inner ring. , Let be the displacement change between the original contact position and the new contact position when the rolling element enters the rectangular damage area. Ignoring the influence of the inner ring curvature on the displacement change based on geometric relationships, it is expressed as:

4. The train bearing fault diagnosis method based on data fusion according to claim 3, characterized in that, Based on the bearing fault dynamics model and the parameter information of the faulty bearing specimen, the bearing simulation vibration signal is obtained, thereby obtaining the fault characteristic frequency of the bearing fault dynamics model, specifically: The parameter information of the faulty bearing specimen includes: the parameters of the faulty bearing specimen, the rotational speed, the fault type, and the sampling frequency. Input the parameters, rotational speed, fault type, and sampling frequency of the faulty bearing specimen to obtain the bearing simulation vibration signal. Plot the time domain diagram of the bearing simulation vibration signal using MATLAB. Use the Hilbert envelope algorithm to obtain the frequency domain envelope diagram of the bearing simulation vibration signal to obtain the fault characteristic frequency of the bearing fault dynamic model. Calculate the theoretical fault characteristic frequency of the bearing specimen. Verify the accuracy of the dynamic model by comparing the fault characteristic frequency of the bearing dynamic model with the theoretical fault characteristic frequency. The calculation of the theoretical fault characteristic frequency of the bearing specimen is specifically as follows: The characteristic frequency of bearing outer ring failure is shown in formula (15): The characteristic frequency of bearing inner ring failure is shown in formula (16): The characteristic frequencies of bearing rolling element failures are shown in formula (17): in, For input shaft frequency, For the number of rolling elements, The diameter of the rolling element, For bearing pitch diameter, This refers to the bearing contact angle.

5. The train bearing fault diagnosis method based on data fusion according to claim 4, characterized in that, The process of collecting vibration response information from real faulty bearing specimens and obtaining the fault characteristic frequencies of the real faulty bearing specimens involves: collecting vibration signals from real faulty bearing specimens from trains, constructing a time-domain graph of the vibration response using Matlab software, generating a frequency-domain envelope graph using the Hilbert envelope algorithm, obtaining the fault characteristic frequencies of the real faulty bearings, and comparing them with the fault characteristic frequencies of the bearing dynamics model to verify the effectiveness of the dynamics model in guiding the diagnosis of real bearing faults.

6. The train bearing fault diagnosis method based on data fusion according to claim 5, characterized in that, The simulation and experimental data fusion method integrates the fault characteristic frequencies of the bearing fault dynamics model with the fault characteristic frequencies of the real fault bearing specimens to obtain simulated fused fault data. Specifically, the simulation and experimental data fusion method includes weighted fusion and convolutional fusion. The weight fusion involves normalizing the fault characteristic frequencies of the bearing fault dynamics model and the fault characteristic frequencies of the actual faulty bearing specimens, and then superimposing them with different weighting ratios in the time domain, as shown in formula (18): in, For signal fusion, , As weight and , For normalization processing, For bearing fault simulation data, These are normal bearing data from actual tests; The convolution fusion process involves normalizing the simulated fault data and the normal data from the actual experiment, followed by convolution. The convolved signal exhibits a similar distribution to both the simulation and the experiment in the time and frequency domains, and its duration becomes twice that of the original signal, as shown in formula (19). in, For convolutional fusion signals, For normalization processing, For bearing fault simulation data, These are normal bearing data from actual tests.

7. A train bearing fault diagnosis system based on data fusion, characterized in that, The train bearing fault diagnosis method based on data fusion as described in claim 1 includes: The module constructs a bearing failure dynamics model and collects parameter information of the failed bearing specimen; The first acquisition module acquires bearing simulation vibration signals based on the bearing fault dynamics model and parameter information of the faulty bearing specimen, thereby acquiring the fault characteristic frequency of the bearing fault dynamics model. The second acquisition module collects vibration response information of real faulty bearing specimens and acquires the fault characteristic frequencies of real faulty bearing specimens. The fusion module, based on the simulation and experimental data fusion method, fuses the fault characteristic frequencies of the bearing fault dynamics model with the fault characteristic frequencies of the real faulty bearing specimen to obtain simulated fused fault data. The comparison module compares the similarity between the frequency domain envelope diagram of the simulated fused data and the frequency domain envelope diagram of the vibration response information of the real faulty bearing specimen, and selects the optimal data fusion strategy. The pre-training module inputs the frequency domain signal of the simulated fused fault data into the convolutional neural network for pre-training, and performs transfer learning on the fault data of the real test based on the parameter transfer strategy to build a bearing fault diagnosis model. The third acquisition module acquires the fault location of the train bearing based on the optimized bearing fault diagnosis model.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

Images