A rotor-bearing system fault diagnosis method and system based on virtual-real fusion

Abstract

This invention provides a rotor-bearing system fault diagnosis method and system based on virtual-real fusion, relating to the field of signal detection technology. The method comprises: deriving an expression for the magnetic pull of mixed eccentricity and imbalance, and establishing a rotor-bearing system twin model based on Hertzian contact theory; introducing rotor eccentricity and bearing inner / outer race faults to construct a twin model for multiple fault conditions; identifying multiple model parameters using an improved genetic algorithm (introducing feature sensitivity evaluation factors, adaptive crossover mutation probability, and simulated annealing) to obtain a modified twin model; generating twin fault samples based on the modified model, and training a one-dimensional cyclic generative adversarial network with a classifier (designed with a comprehensive loss function) to generate samples that closely resemble the true distribution; and finally, forming a balanced dataset from the virtual-real fusion samples, which is then trained using a one-dimensional convolutional neural network to achieve fault classification. This invention addresses the problems of insufficient fault data and significant discrepancies between the twin model and the real data, thereby improving the accuracy and reliability of fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the field of signal detection technology, and in particular to a rotor-bearing system fault diagnosis method and system based on virtual-real fusion. Background Art

[0002] In modern industrial systems, rotating machinery, as a critical foundational component for national economic development, is widely used in a wide range of sectors, including energy, manufacturing, and transportation. The stability of its rotor-bearing system is directly related to the safe and reliable operation of the entire machine. Failure in the rotor-bearing system can not only cause equipment downtime, resulting in production interruptions and economic losses, but can also lead to safety accidents and threaten personnel safety.

[0003] However, in actual operation, most rotating machinery operates in a healthy state for extended periods of time for safety reasons, making it extremely difficult to obtain rich fault data. This lack of data severely restricts fault analysis of rotor-bearing systems. Traditional fault diagnosis methods often rely on a large number of fault samples for model training and feature analysis. This lack of data makes it difficult for these methods to accurately identify and diagnose faults, failing to meet the real-world demand for early warning and precise diagnosis of equipment failures.

[0004] To address the lack of fault data, a digital twin model of the rotor-bearing system based on analytical methods has emerged. By introducing fault excitations tailored to the actual fault type, this model can capture twin data under different conditions, providing a new data source for fault diagnosis. However, due to factors such as the simplified numerical model and varying environmental noise, the resulting twin data differs from the real data, preventing direct virtual-real fusion. This results in biased fault diagnosis results based on this model, making it difficult to ensure diagnostic accuracy and reliability.

[0005] While genetic algorithms can identify multiple parameters in twin models and improve their quality to a certain extent, they still require further optimization to better adapt to complex and changing real-world operating conditions. Generative adversarial networks (GANs), based on twin data, can generate fault data that more closely matches the true distribution. However, their application in rotor-bearing system fault diagnosis is currently imperfect, with issues such as low generated data quality and incomplete classification information.

[0006] In summary, current rotor-bearing system fault diagnosis faces challenges such as difficulty in acquiring fault data, significant discrepancies in existing model data, and imperfect algorithms and network applications. Therefore, optimizing genetic algorithms and generative adversarial networks to acquire high-quality fault data and improve rotor-bearing system fault diagnosis capabilities has become an urgent issue. Summary of the Invention

[0007] To this end, the present invention provides a rotor-bearing system fault diagnosis method and system based on virtual-real fusion, which is used to solve the problems in the existing technology of imbalance of real sample data and difficulty in obtaining fault data during intelligent diagnosis of rotor-bearing systems, and the twin model being affected by multiple factors, resulting in large differences between data and real data and insufficient fault diagnosis capabilities.

[0008] In order to solve the above problems, an embodiment of the present invention provides a rotor-bearing system fault diagnosis method based on virtual-real fusion, which includes:

[0009] S1: Derive a unified expression for the mixed eccentric unbalanced magnetic pull, introduce the nonlinear restoring force of the bearing based on the Hertz contact theory, and establish a twin model of the rotor-bearing system;

[0010] S2: introducing a rotor eccentricity fault, a bearing inner race fault, and a bearing outer race fault into the rotor-bearing system twin model to obtain a rotor-bearing system twin model under different fault conditions;

[0011] S3: By introducing a feature sensitivity evaluation factor into the genetic algorithm to establish a fitness function, and combining adaptive crossover, mutation probability, and simulated annealing, multi-parameter identification is performed on the rotor-bearing system twin model under the different fault conditions to obtain a revised twin model;

[0012] S4: Obtain twin fault samples based on the modified twin model, add a classifier to the one-dimensional cyclic generative adversarial network and design a comprehensive loss function to train and generate generated samples that are close to the distribution of real samples;

[0013] S5: The generated samples are fused with the real samples to obtain a balanced data set, and a one-dimensional convolutional neural network is trained to implement fault classification.

[0014] Preferably, the process of establishing the rotor-bearing system twin model includes:

[0015] First, based on the theory of electrical machinery, the air gap length, magnetic permeance and magnetomotive force distribution under hybrid eccentricity are derived, and the Maxwell stress integral is used to obtain 、 Unified expression for directional unbalanced magnetic pull;

[0016] Then, the bearing rolling element position and contact deformation are derived based on Hertz contact theory, and the cumulative value is obtained. 、 Directional bearing nonlinear contact force;

[0017] Finally, the unbalanced magnetic pull and the nonlinear contact force of the bearing are combined to establish the dynamic equation of the rotor-bearing system.

[0018] Preferably, the 、 The unified expression of directional unbalanced magnetic pull is:

[0019] ;

[0020] ;

[0021] Where, for Unbalanced magnetic pull in the direction; for Unbalanced magnetic pull in the direction; is the rotor radius; is the axial length; is the amplitude of the air gap magnetomotive force; is the vacuum permeability; is the constant component of uniform air gap permeability when there is no eccentricity; is the permeability component caused by static eccentricity; is the permeability component caused by dynamic eccentricity; is the rotor mechanical angular frequency; is the initial phase; For time; is the dynamic eccentric angle; is the number of magnetic pole pairs.

[0022] Preferably, the 、 The nonlinear contact force of the directional bearing is:

[0023] ;

[0024] ;

[0025] Where, For Directional bearing nonlinear contact force; For Directional bearing nonlinear contact force; is the number of rolling elements; is the equivalent contact stiffness obtained from the local deformation of the contact area; For the The total contact deformation of the rolling elements; is the bearing clearance; For the The position of the rolling element.

[0026] Preferably, the rotor-bearing system twin model is expressed as:

[0027] ;

[0028] Where, is the rotor mass; is the damping coefficient; is the mechanical stiffness of the rotor shaft; and Respectively in 、 Exciting force in direction; and Respectively in 、 Nonlinear contact forces on bearings in the direction; 、 、 The rotor is Directional displacement, velocity, acceleration; 、 、 The rotor is Directional displacement, velocity, and acceleration.

[0029] Preferably, the method of introducing the rotor eccentricity fault, the bearing inner ring fault and the bearing outer ring fault is:

[0030] The rotor eccentricity fault is determined by setting the relative static eccentricity and relative dynamic eccentricity ,when When is the static eccentricity condition, When is the dynamic eccentricity condition, and When both are not 0, it is a mixed eccentricity condition;

[0031] The bearing inner ring fault and the bearing outer ring fault respectively set rectangular defects on the inner ring and outer ring surfaces, and the total contact deformation of the rolling body is characterized by the depth of the rolling body sinking into the defect, wherein the inner ring defect depth is related to the inner ring radius, the rolling body radius and the angle difference, and the outer ring defect depth is related to the outer ring radius and the angle difference.

[0032] Preferably, the inner ring defect depth is expressed as:

[0033] ;

[0034] Where, is the inner ring defect depth, 、 are the bearing inner ring radius and rolling element radius respectively; is the initial angle when the rolling element enters the defect area; It is the angle at which the rolling element sinks into the defect area to the maximum depth, i.e. the center angle; is the ending angle when leaving the defect area; It is the angular difference between the bearing inner ring and the rolling elements when they rotate.

[0035] Preferably, the outer ring defect depth is expressed as:

[0036] ;

[0037] Where, is the inner ring defect depth, 、 are the bearing inner ring radius and rolling element radius respectively; is the initial angle when the rolling element enters the defect area; It is the angle at which the rolling element sinks into the defect area to the maximum depth, i.e. the center angle; is the ending angle when leaving the defect area; It is the angular difference between the bearing inner ring and the rolling elements when they rotate.

[0038] Preferably, the method for establishing the characteristic sensitivity evaluation factor is:

[0039] The ratio of between-class variance to within-class variance is used as the feature sensitivity evaluation factor The inter-class variance is calculated by the total mean of the eigenvalues ​​of the real samples and the simulated samples and the number of samples, and the intra-class variance is calculated by the eigenvalue variance of each sample state, where the feature sensitivity evaluation factor is Expressed as:

[0040] ;

[0041] Where, and The sample status and sample status the intra-class variance in ; is the between-class variance.

[0042] Preferably, the expressions of the adaptive crossover and mutation probabilities are:

[0043] ;

[0044] ;

[0045] Where, 、 are the adaptive crossover probability and the adaptive mutation probability respectively; 、 are the maximum and minimum values ​​of the predefined crossover probability respectively; 、 are the maximum and minimum values ​​of the predefined mutation probability respectively; The fitness value of the individual with higher fitness value among the two individuals participating in the crossover operation; is the fitness value of the mutant individual; is the average fitness value in the current population; is the maximum fitness value in the current population.

[0046] Preferably, the implementation of the simulated annealing concept includes:

[0047] Set the outer loop to control temperature decay, and the initial temperature gradually decreases to the final temperature according to the cooling coefficient α; the inner loop iterates multiple times at each temperature and accepts new individuals according to the Metropolis criterion, where the acceptance probability is:

[0048] ;

[0049] Where, is the probability of acceptance, is the current temperature, and are the fitness values ​​of the new individual and the original individual respectively.

[0050] Preferably, the one-dimensional cyclic generative adversarial network includes two generators and , two discriminators and And two classifiers and , the comprehensive loss function of the one-dimensional cyclic generative adversarial network includes:

[0051] Generator and the discriminator The adversarial loss function between Expressed as:

[0052] ;

[0053] Generator and the discriminator The adversarial loss function between Expressed as:

[0054] ;

[0055] Where, and The sample space and One-dimensional signal in ; express Obey the sample space Expected distribution of data; express Obey the sample space Expected distribution of data;

[0056] Cycle consistency loss function Expressed as:

[0057] ;

[0058] Where, represents the L1 norm;

[0059] Classifier The cross entropy loss function Expressed as:

[0060] ;

[0061] Classifier The cross entropy loss function Expressed as:

[0062] ;

[0063] Where, is the sample space One-dimensional signal in ; is the sample space One-dimensional signal in ; is the sample space The total number of samples for parameter loss calculation; is the sample space The total number of samples for parameter loss calculation;

[0064] Final loss function Expressed as:

[0065] ;

[0066] Where, , , are weight coefficients, which are used to control the importance of cycle consistency loss and two classification losses respectively.

[0067] Preferably, the network structure of the one-dimensional convolutional neural network is the same as the classifier structure of the one-dimensional cyclic generative adversarial network.

[0068] An embodiment of the present invention further provides a rotor-bearing system fault diagnosis system based on virtual-real fusion, which is used to implement the above-mentioned rotor-bearing system fault diagnosis method based on virtual-real fusion, and specifically includes:

[0069] A model building module is used to derive a unified expression for the mixed eccentric unbalanced magnetic pull, introduce the nonlinear restoring force of the bearing based on Hertz contact theory, and establish a twin model of the rotor-bearing system;

[0070] a fault introduction module, configured to introduce a rotor eccentricity fault, a bearing inner race fault, and a bearing outer race fault into the rotor-bearing system twin model, so as to obtain the rotor-bearing system twin models under different fault conditions;

[0071] A model correction module is used to establish a fitness function by introducing a feature sensitivity evaluation factor into a genetic algorithm, and to perform multi-parameter identification on the rotor-bearing system twin model under the different fault conditions by combining adaptive crossover, mutation probability, and simulated annealing to obtain a corrected twin model;

[0072] A sample generation module is used to obtain twin fault samples based on the modified twin model, add a classifier to the one-dimensional cyclic generative adversarial network, and design a comprehensive loss function to train and generate generated samples that are close to the distribution of real samples;

[0073] The fault classification module is used to fuse the generated samples with the real samples to obtain a balanced data set, and implement fault classification by training a one-dimensional convolutional neural network.

[0074] It can be seen from the above technical solutions that the present invention has the following beneficial effects:

[0075] The embodiment of the present invention provides a rotor-bearing system fault diagnosis method and system based on virtual-real fusion. The present invention combines twin models, parameter identification and cyclic generative adversarial networks to achieve virtual-real fusion of real samples and simulated samples. By introducing fault excitation into the rotor-bearing system twin model, simulated fault data under different working conditions are obtained; by introducing feature sensitivity evaluation factors and adaptive crossover and mutation probabilities in parameter identification, the influence of simplified numerical models, environmental noise and other factors on the twin model is reduced; by adding a classifier into the cyclic generative adversarial network and designing a new loss function, fault simulation samples that are close to the real distribution are further obtained. The present invention realizes the fusion of simulated samples and real samples, effectively makes up for the problem of insufficient rotor-bearing system fault samples in fault diagnosis, and improves the generalization performance of the classification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. Those skilled in the art can derive other drawings based on these drawings without inventive effort. Among them:

[0077] Figure 1 A flow chart of a rotor-bearing system fault diagnosis method based on virtual-real fusion provided by the present invention;

[0078] Figure 2 1 is a simplified structural diagram of the rotor-bearing system in the present invention, wherein (a) is a schematic diagram of the rotor-bearing system, and (b) is a coordinate system of the rotor-bearing system;

[0079] Figure 3 Schematic diagram of the fault excitation introduced in the present invention, where (a) is a rotor eccentricity fault, (b) is a bearing inner race fault, and (c) is a bearing outer race fault;

[0080] Figure 4 Spectrum diagram of the simulated sample and the real sample under the rotor eccentricity fault condition in the present invention, where (a) is the real rotor eccentricity fault spectrum and (b) is the simulated rotor eccentricity fault spectrum;

[0081] Figure 5 Spectrum diagram of the simulated sample and the real sample under the bearing inner ring fault condition in the present invention, where (a) is the real bearing inner ring fault spectrum, and (b) is the simulated bearing inner ring fault spectrum;

[0082] Figure 6 Spectrum diagram of the simulated sample and the real sample under the bearing outer ring fault condition in the present invention, where (a) is the real bearing outer ring fault spectrum, and (b) is the simulated bearing outer ring fault spectrum;

[0083] Figure 7 Flowchart of the improved genetic algorithm in the present invention;

[0084] Figure 8 This is a diagram showing the structure of a one-dimensional cyclic generative adversarial network in the present invention;

[0085] Figure 9 A block diagram of a rotor-bearing system fault diagnosis system based on virtual-real fusion provided by the present invention. DETAILED DESCRIPTION

[0086] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0087] Example 1: In order to solve the problems in the prior art of unbalanced real sample data, difficulty in obtaining fault data, and the large difference between the data and the real data and insufficient fault diagnosis ability caused by the influence of various factors on the twin model during intelligent diagnosis of the rotor-bearing system, Figure 1As shown, the present invention proposes a rotor-bearing system fault diagnosis method based on virtual-real fusion, which includes:

[0088] S1: Derive a unified expression for the mixed eccentric unbalanced magnetic pull, introduce the nonlinear restoring force of the bearing based on the Hertz contact theory, and establish a twin model of the rotor-bearing system;

[0089] S2: Introduce rotor eccentricity fault, bearing inner race fault, and bearing outer race fault into the rotor-bearing system twin model to obtain the rotor-bearing system twin model under different fault conditions;

[0090] S3: By introducing a feature sensitivity evaluation factor into the genetic algorithm to establish a fitness function, and combining adaptive crossover, mutation probability, and simulated annealing, multi-parameter identification is performed on the rotor-bearing system twin model under different fault conditions to obtain a revised twin model;

[0091] S4: Based on the modified twin model, twin fault samples are obtained. A classifier is added to the one-dimensional recurrent generative adversarial network and a comprehensive loss function is designed to train and generate generated samples that are close to the distribution of real samples.

[0092] S5: The generated samples are fused with the real samples to obtain a balanced dataset, and a one-dimensional convolutional neural network is used for fault classification.

[0093] As can be seen from the above technical solution, the present invention proposes a rotor-bearing system fault diagnosis method based on virtual-real fusion. By deriving a unified expression for the mixed eccentric unbalanced magnetic pull and combining it with the nonlinear restoring force of the bearing, the method aims to establish a twin model that can reflect the fault characteristics. In the process of twin model parameter identification, a corresponding fitness function is established by introducing a feature sensitivity evaluation factor to maximize the reflection of the difference between the simulated sample and the real sample. In combination with adaptive crossover and mutation probability, the ability of the genetic algorithm to find the optimal solution is improved. By adding a classifier to the cyclic generative adversarial network, it is beneficial to retain the category information in the original signal. This helps to solve the problem of imbalanced rotor-bearing system fault samples and achieve more accurate rotor-bearing system fault diagnosis.

[0094] In step S1, the present invention derives a unified expression for the mixed eccentric unbalanced magnetic pull, introduces the nonlinear restoring force of the bearing based on the Hertz contact theory, and establishes a twin model of the rotor-bearing system.

[0095] First, based on the theory of electrical machinery, the air gap length, magnetic permeance and magnetomotive force distribution under hybrid eccentricity are derived, and the Maxwell stress integral is used to obtain 、 Unified expression for directional unbalanced magnetic pull.

[0096] Specifically, based on the theory of electrical machinery, static eccentricity and dynamic eccentricity are comprehensively considered, and any dynamic eccentricity angle under mixed eccentricity between the rotor and stator is The air gap length is a function related to time and angle, which can be approximately expressed as:

[0097] ;

[0098] Where, is the average air gap length when there is no eccentricity; is the relative static eccentricity, is the relative dynamic eccentricity, and are the ratios of static deviation and dynamic deviation to the average air gap respectively; is the angle at any position during the operation of the rotor; is the rotor mechanical angular frequency; For time; is the dynamic eccentric angle.

[0099] Furthermore, when the eccentricity is small, the air gap permeance can be expanded into a Fourier series, ignoring the high-order components, and can be expressed as:

[0100] ;

[0101] Where, is the vacuum permeability; is the constant component of uniform air gap permeability when there is no eccentricity; is the permeability component caused by static eccentricity; is the permeability component caused by dynamic eccentricity.

[0102] Furthermore, the fundamental magnetomotive force of the air gap is:

[0103] ;

[0104] Where, is the amplitude of the air gap magnetomotive force; is the initial phase; is the amplitude of the rotor fundamental magnetomotive force; is the amplitude of the fundamental magnetomotive force of the stator armature winding; is the internal power factor angle.

[0105] Furthermore, the air gap magnetic potential and magnetic permeance are used to establish the air gap magnetic flux density distribution, which can be expressed as:

[0106] .

[0107] Since the radial air gap flux density is much larger than the tangential air gap flux density, the tangential component can be ignored. According to Maxwell's electromagnetic field law, the electromagnetic force acting on the rotor is proportional to the square of its radial flux density, which can be expressed as:

[0108] .

[0109] Furthermore, by integrating the Maxwell stress on the rotor surface, a unified expression for the unbalanced magnetic pull under different pole pairs is proposed. Unbalanced magnetic pull in the direction and Unbalanced magnetic pull in the direction It can be expressed as:

[0110] ;

[0111] ;

[0112] Where, is the rotor radius; is the axial length; is the number of magnetic pole pairs.

[0113] Then, the bearing rolling element position and contact deformation are derived based on Hertz contact theory, and the cumulative value is obtained. 、 Directional bearing nonlinear contact forces.

[0114] Specifically, set The rolling element on the positive side of the shaft is numbered 1, then The position of the rolling elements It can be expressed by the following formula:

[0115] ;

[0116] Where, is the initial angular position of the rolling element numbered 1; is the angular velocity of the rolling body; is the number of rolling elements; The random sliding deviation is introduced, which is caused by the existence of lubricating oil film and reserved gap, resulting in a certain amount of slippage of the rolling element during movement; For Random probability function of uniform distribution on ; is the rotor rotation angular velocity; and are the inner and outer diameters of the bearing respectively.

[0117] Under the action of rotor eccentricity, the bearing vibrates, and the total contact deformation of the rolling element is Is the horizontal displacement with the bearing inner ring , vertical displacement of the bearing inner ring , No. The position of the rolling elements and bearing clearance The relevant functions are expressed as follows:

[0118] .

[0119] Furthermore, according to Hertz contact theory, the contact force of each rolling element is accumulated to obtain Directional bearing contact force and Directional bearing contact force , which can be expressed as:

[0120] ;

[0121] ;

[0122] Where, is the number of rolling elements; is the equivalent contact stiffness obtained from the local deformation of the contact area.

[0123] Finally, the simplified structural diagram of the rotor-bearing system is as follows Figure 2 As shown, the dynamic equation of the rotor-bearing system is established by combining the unbalanced magnetic pull and the nonlinear contact force of the bearing, which can be expressed as:

[0124] ;

[0125] Where, is the rotor mass; is the damping coefficient; is the mechanical stiffness of the rotor shaft; and Respectively in 、 Exciting force in direction; 、 、 The rotor is Directional displacement, velocity, acceleration; 、 、 The rotor is Directional displacement, velocity, and acceleration.

[0126] In step S2, the present invention obtains the rotor-bearing system twin model under different fault conditions by introducing rotor eccentricity fault, bearing inner ring fault and bearing outer ring fault into the rotor-bearing system twin model.

[0127] Specifically, rotor eccentricity faults such as Figure 3 As shown in (a), by setting the relative static eccentricity and relative dynamic eccentricity ,when When is the static eccentricity condition, When is the dynamic eccentricity condition, and When both are not 0, it is a mixed eccentricity condition.

[0128] Bearing inner ring fault and bearing outer ring fault are respectively set with rectangular defects on the inner ring and outer ring surfaces, and the total contact deformation of the rolling element is characterized by the depth of the rolling element sinking into the defect. The inner ring defect depth is related to the inner ring radius, the rolling element radius and the angle difference, and the outer ring defect depth is related to the outer ring radius and the angle difference.

[0129] Furthermore, the total contact deformation of the rolling element Expressed as:

[0130] ;

[0131] Where, It is the depth to which the rolling element sinks when passing through the defect.

[0132] Bearing inner ring defects such as Figure 3 As shown in (b), when a bearing inner ring defect is introduced, the inner ring defect depth It can be expressed as:

[0133] ;

[0134] Where, is the inner ring defect depth, 、 are the bearing inner ring radius and rolling element radius respectively; is the initial angle when the rolling element enters the defect area; It is the angle at which the rolling element sinks into the defect area to the maximum depth, i.e. the center angle; is the ending angle when leaving the defect area; is the angular difference between the bearing inner ring and the rolling element when they rotate, which can be expressed as:

[0135] .

[0136] Bearing outer ring defects such as Figure 3 As shown in (c), when a bearing outer ring defect is introduced, the outer ring defect depth It can be expressed as:

[0137] ;

[0138] Where, is the inner ring defect depth, 、 are the bearing inner ring radius and rolling element radius respectively; is the initial angle when the rolling element enters the defect area; It is the angle at which the rolling element sinks into the defect area to the maximum depth, i.e. the center angle; is the ending angle when leaving the defect area; It is the angular difference between the bearing inner ring and the rolling elements when they rotate.

[0139] The signals simulated by the twin model of the rotor-bearing system under different fault conditions are normalized and Fourier transformed. The spectrum diagrams of the simulated samples and the real samples under the three fault conditions are shown as follows: Figure 4 、 Figure 5 、 Figure 6 As shown in FIG, at this time, the simulated signal spectrum diagram basically includes the real signal transfer frequency and fault characteristic frequency.

[0140] In step S3, the present invention establishes a fitness function by introducing a feature sensitivity evaluation factor into the genetic algorithm, and combines the ideas of adaptive crossover, mutation probability and simulated annealing to perform multi-parameter identification on the rotor-bearing system twin model under different fault conditions to obtain a corrected twin model. Figure 7 This is the flowchart of the improved genetic algorithm.

[0141] Specifically, first, the ratio of between-class variance to within-class variance is used as the feature sensitivity evaluation factor The inter-class variance is calculated by the total mean of the eigenvalues ​​of the real samples and the simulated samples and the number of samples, and the intra-class variance is calculated by the eigenvalue variance of each sample state, where the feature sensitivity evaluation factor is Expressed as:

[0142] ;

[0143] Where, and The sample status and sample status the intra-class variance in ; is the between-class variance.

[0144] The between-class variance can be expressed as:

[0145] ;

[0146] Where, and are the number of real samples and simulated samples respectively; and The feature is in the sample state Middle samples and sample status Middle The value of the samples; is the overall mean of the feature in both samples.

[0147] The present invention selects the peak value, root mean square, kurtosis, skewness, peak factor, shape factor, center of gravity frequency and mean square frequency, and calculates the characteristic sensitivity evaluation factor respectively. ,choose The largest feature establishes the fitness function.

[0148] Then, adding adaptive crossover and mutation probabilities to the crossover and mutation operations can be expressed as:

[0149] ;

[0150] ;

[0151] Where, 、 are the adaptive crossover probability and the adaptive mutation probability respectively; 、 are the maximum and minimum values ​​of the predefined crossover probability respectively; 、 are the maximum and minimum values ​​of the predefined mutation probability respectively; The fitness value of the individual with higher fitness value among the two individuals participating in the crossover operation; is the fitness value of the mutant individual; is the average fitness value in the current population; is the maximum fitness value in the current population.

[0152] Finally, we add an outer loop and an inner loop to the genetic algorithm, combining the idea of ​​simulated annealing. The outer loop controls temperature decay, gradually decreasing the initial temperature by a cooling coefficient α to the final temperature. The inner loop iterates multiple times at each temperature, accepting new individuals using the Metropolis criterion, where the acceptance probability is:

[0153] ;

[0154] Where, is the probability of acceptance, is the current temperature, and are the fitness values ​​of the new individual and the original individual respectively.

[0155] In step S4, the present invention obtains twin fault samples based on the modified twin model, adds a classifier to the one-dimensional cyclic generative adversarial network and designs a comprehensive loss function to train and generate generated samples that are close to the distribution of real samples.

[0156] Specifically, the cyclic generative adversarial network consists of two generators ( and ) and 2 discriminators ( and ). The generator generates a modified false signal based on the input simulation signal. Its purpose is to make the generated false signal as close to the real signal as possible so that the discriminator cannot distinguish it. The discriminator needs to distinguish the generated false signal from the real signal as much as possible. In order to further make the categories of the generated signal and the simulation signal similar, two classifiers are introduced on this basis ( and ) in order to retain the category information in the original signal, the improved cyclic generative adversarial network structure is shown in the figure below Figure 8 shown.

[0157] In the cyclic generative adversarial network game confrontation process, there are two sample spaces and , the recurrent generative adversarial network aims to learn and The general mapping between and converts the simulated signal into a signal that approximates the real signal. The network training process is implemented by three types of loss functions, which are described as follows:

[0158] (1) Adversarial loss function

[0159] Generator and the discriminator The adversarial loss function between Expressed as:

[0160] ;

[0161] Generator and the discriminator The adversarial loss function between Expressed as:

[0162] ;

[0163] Where, and The sample space and One-dimensional signal in ; express Obey the sample space Expected distribution of data; express Obey the sample space The expectation of the data distribution.

[0164] (2) Cycle consistency loss function

[0165] The cyclic generative adversarial network uses a data conversion mechanism to introduce a cycle consistency loss, so that the signal can retain the original style characteristics after passing through two generators continuously. The cycle consistency loss uses the L1 norm and can be expressed as:

[0166] ;

[0167] Where, represents the L1 norm.

[0168] (3) Cross-entropy loss function

[0169] Two classifiers are introduced into the basic cycle generative adversarial network to retain category information, and the classifier is trained using the standard cross entropy loss function. and .

[0170] Classifier The cross entropy loss function Expressed as:

[0171] ;

[0172] Classifier The cross entropy loss function Expressed as:

[0173] .

[0174] Combining the above three types of loss functions, the final loss function for:

[0175] ;

[0176] Where, , , are weight coefficients, which are used to control the importance of cycle consistency loss and two classification losses respectively.

[0177] Therefore, the optimization goal of the one-dimensional cyclic generative adversarial network of the present invention is:

[0178] ;

[0179] ;

[0180] ;

[0181] Where, 、 、 、 、 、 To train to reach Nash equilibrium 、 、 、 、 、 status.

[0182] Specifically, 2048 points in the real samples and simulation samples are selected as a group, and 160 training samples and 40 test samples are set for each type of fault. The specific network parameters are shown in Table 1. The number of network iterations is 800, the learning rate is 0.0002, and the batch size is 32.

[0183] Table 1 Parameters of one-dimensional cyclic generative adversarial network

[0184]

[0185] In step S5, the present invention fuses the generated samples with the real samples to obtain a balanced data set, and uses a one-dimensional convolutional neural network training to achieve fault classification.

[0186] Specifically, the generated samples are fused with real samples to obtain a dataset with the same number of samples for each operating condition. Based on this dataset, a 1D CNN is trained to classify rotor-bearing system faults. The 1D CNN uses the same network structure as the classifier in the one-dimensional recurrent generative adversarial network.

[0187] Example 2: Figure 9 As shown, the present invention provides a rotor-bearing system fault diagnosis system based on virtual-real fusion, which is used to implement the rotor-bearing system fault diagnosis method based on virtual-real fusion of the above embodiment 1, specifically comprising:

[0188] A model building module 100 is used to derive a unified expression for the mixed eccentric unbalanced magnetic pull, introduce the nonlinear restoring force of the bearing based on the Hertz contact theory, and establish a twin model of the rotor-bearing system;

[0189] A fault introduction module 200 is used to introduce a rotor eccentricity fault, a bearing inner race fault, and a bearing outer race fault into the rotor-bearing system twin model to obtain the rotor-bearing system twin model under different fault conditions;

[0190] The model correction module 300 is used to establish a fitness function by introducing a feature sensitivity evaluation factor into the genetic algorithm. It then combines adaptive crossover, mutation probability, and simulated annealing to perform multi-parameter identification on the rotor-bearing system twin model under different fault conditions to obtain a corrected twin model.

[0191] The sample generation module 400 is used to obtain twin fault samples based on the modified twin model, add a classifier to the one-dimensional recurrent generative adversarial network, and design a comprehensive loss function to train and generate generated samples that are close to the distribution of real samples;

[0192] The fault classification module 500 is used to fuse the generated samples with the real samples to obtain a balanced data set, and implement fault classification using one-dimensional convolutional neural network training.

[0193] A rotor-bearing system fault diagnosis system based on virtual-real fusion in this embodiment is used to implement the aforementioned rotor-bearing system fault diagnosis method based on virtual-real fusion. Therefore, the specific implementation method of the rotor-bearing system fault diagnosis system based on virtual-real fusion can be seen in the embodiment part of the rotor-bearing system fault diagnosis method based on virtual-real fusion in the previous text. For example, the model establishment module 100, the fault introduction module 200, the model correction module 300, the sample generation module 400, and the fault classification module 500 are respectively used to implement steps S1, S2, S3, S4, and S5 in the above-mentioned rotor-bearing system fault diagnosis method based on virtual-real fusion. Therefore, its specific implementation method can refer to the description of the corresponding embodiments of each part. In order to avoid redundancy, it will not be repeated here.

[0194] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0195] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0196] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0197] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications derived therefrom remain within the scope of protection of the present invention.

Claims

1. A rotor-bearing system fault diagnosis method based on virtual-real fusion, characterized in that: include: S1: Derive a unified expression for the mixed eccentric unbalanced magnetic pull, introduce the nonlinear restoring force of the bearing based on the Hertz contact theory, and establish a twin model of the rotor-bearing system; The process of establishing the rotor-bearing system twin model includes: Firstly, the air gap length, magnetic permeance and magnetomotive force distribution under hybrid eccentricity are derived based on the electromechanical theory, and the unified expression of the unbalanced magnetic pull in the x and y directions is obtained through Maxwell stress integration. Then, the bearing rolling element position and contact deformation are derived based on Hertz contact theory, and the nonlinear contact force of the bearing in the x and y directions is accumulated. Finally, the unbalanced magnetic pull and the nonlinear contact force of the bearing are combined to establish the dynamic equation of the rotor-bearing system; The unified expression of the unbalanced magnetic pull in the x and y directions is: Where, F xUMP is the unbalanced magnetic pull in the x direction; F yUMP is the unbalanced magnetic pull in the y direction; R is the rotor radius; L is the axial length; F m is the amplitude of the air gap magnetomotive force; μ0 is the vacuum permeability; Λ0 is the constant component of the uniform air gap permeability when there is no eccentricity; Λ s is the permeability component caused by static eccentricity; Λ d is the permeability component caused by dynamic eccentricity; ω d is the rotor mechanical angular frequency; is the initial phase; t is the time; θ is the dynamic eccentricity angle; p is the number of magnetic pole pairs; S2: introducing a rotor eccentricity fault, a bearing inner race fault, and a bearing outer race fault into the rotor-bearing system twin model to obtain a rotor-bearing system twin model under different fault conditions; S3: By introducing a feature sensitivity evaluation factor into the genetic algorithm to establish a fitness function, and combining adaptive crossover, mutation probability, and simulated annealing, multi-parameter identification is performed on the rotor-bearing system twin model under the different fault conditions to obtain a revised twin model; S4: Obtain twin fault samples based on the modified twin model, add a classifier to the one-dimensional cyclic generative adversarial network and design a comprehensive loss function to train and generate generated samples that are close to the distribution of real samples; S5: The generated samples are fused with the real samples to obtain a balanced data set, and a one-dimensional convolutional neural network is trained to implement fault classification.

2. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1 is characterized in that: The nonlinear contact force of the bearing in the x and y directions is: Where, F bx is the nonlinear contact force of the bearing in the x direction; F by is the nonlinear contact force of the bearing in the y direction; N b is the number of rolling elements; k b is the equivalent contact stiffness obtained by local deformation of the contact area; δ j is the total contact deformation of the jth rolling element; c r is the bearing clearance; θ j is the position of the jth rolling element.

3. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1 or 2, characterized in that: The expression of the rotor-bearing system twin model is: Where m is the rotor mass; c is the damping coefficient; k is the mechanical stiffness of the rotor shaft; F ax and F ay are the exciting forces in the x and y directions respectively; F bx and F by The nonlinear contact forces of the bearings in the x and y directions are respectively; are the displacement, velocity and acceleration of the rotor in the x direction; y, are the displacement, velocity and acceleration of the rotor in the y direction respectively.

4. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1 is characterized in that: The method of introducing the rotor eccentricity fault, the bearing inner ring fault and the bearing outer ring fault is: The rotor eccentricity fault is determined by setting the relative static eccentricity ε s and relative dynamic eccentricity ε d , when ε d =0 is the static eccentricity condition, ε s =0 is the dynamic eccentricity condition, ε s and ε d When both are not 0, it is a mixed eccentricity condition; The bearing inner ring fault and the bearing outer ring fault respectively set rectangular defects on the inner ring and outer ring surfaces, and the total contact deformation of the rolling body is characterized by the depth of the rolling body sinking into the defect, wherein the inner ring defect depth is related to the inner ring radius, the rolling body radius and the angle difference, and the outer ring defect depth is related to the outer ring radius and the angle difference.

5. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 4 is characterized in that: The inner race defect depth is expressed as: Where H 内 is the inner ring defect depth, r i 、r b are the inner ring radius and rolling element radius of the bearing respectively; θ1 is the initial angle when the rolling element enters the defect area; θ2 is the angle when the rolling element sinks into the defect area to the maximum depth, that is, the center angle; θ3 is the end angle when leaving the defect area; θ′ j It is the angular difference between the bearing inner ring and the rolling elements when they rotate.

6. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 4 is characterized in that: The outer ring defect depth is expressed as: Where H 外 is the inner ring defect depth, r o 、r b are the inner ring radius and rolling element radius of the bearing respectively; θ1 is the initial angle when the rolling element enters the defect area; θ2 is the angle when the rolling element sinks into the defect area to the maximum depth, that is, the center angle; θ3 is the end angle when leaving the defect area; θ′ j It is the angular difference between the bearing inner ring and the rolling elements when they rotate.

7. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1 is characterized in that: The method for establishing the characteristic sensitivity evaluation factor is: The ratio of between-class variance to within-class variance is used as the feature sensitivity evaluation factor f e The inter-class variance is calculated by the total mean of the eigenvalues ​​of the real samples and the simulated samples and the number of samples, and the intra-class variance is calculated by the eigenvalue variance of each sample state, where the feature sensitivity evaluation factor f e Expressed as: Where, and are the intra-class variances in sample state a and sample state b respectively; is the between-class variance.

8. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1 is characterized in that: The expressions of the adaptive crossover and mutation probabilities are: Where, P c 、P m are adaptive crossover probability and adaptive mutation probability respectively; P cmax 、P cmin are the maximum and minimum values ​​of the predefined crossover probability; P mmax 、P mmin are the maximum and minimum values ​​of the predefined mutation probability respectively; f′ is the fitness value of the individual with higher fitness value among the two individuals participating in the crossover operation; f is the fitness value of the mutant individual; f avg is the average fitness value in the current population; f max is the maximum fitness value in the current population.

9. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1, characterized in that: The implementation of the simulated annealing idea includes: Set the outer loop to control temperature decay, and the initial temperature gradually decreases to the final temperature according to the cooling coefficient α; the inner loop iterates multiple times at each temperature and accepts new individuals according to the Metropolis criterion, where the acceptance probability is: Where P is the acceptance probability, T is the current temperature, and f new and f are the fitness values ​​of the new individual and the original individual respectively.

10. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1, characterized in that: The one-dimensional cyclic generative adversarial network includes two generators G and F, two discriminators D X and D Y And two classifiers C X and C Y , the comprehensive loss function of the one-dimensional cycle generative adversarial network include: Generator G and Discriminator D Y The adversarial loss function L between GAN (G,D Y ,X,Y) is expressed as: L GAN (G,D Y ,X,Y)=E y～Pdata(y) [log D Y (y)]+E x～Pdata(x) [log(1-D Y (G(x)))]; Generator F and Discriminator D X The adversarial loss function L between GAN (F,D X ,Y,X) is expressed as: L GAN (F,D X ,Y,X)=E x～Pdata(x) [log D X (x)]+E y～Pdata(y) [log(1-D X (F(y)))]; Where x and y are one-dimensional signals in sample space X and Y respectively; E x～Pdata(x) Indicates that x obeys the expectation of the data distribution of the sample space X; E y～Pdata(y) Indicates that y follows the expectation of the data distribution of the sample space Y; Cycle consistency loss function L cyc (G,F) is expressed as: L cyc (G,F)=E x～Pdata(x) [‖F(G(x))-x‖1]+E y～Pdata(y) [‖G(F(y))-y‖1]; Where, ‖·‖1 represents the L1 norm; Classifier C X The cross entropy loss function L c (C X ) is expressed as: Classifier C Y The cross entropy loss function L c (C Y ) is expressed as: Where x i is a one-dimensional signal in the sample space X; y j is a one-dimensional signal in the sample space Y; N X The total number of samples for which parameter loss is calculated in the sample space X; N Y The total number of samples for which parameter loss is calculated in the sample space Y; The final loss function L(G,F,D X ,D Y ,C X ,C Y ) is expressed as: L(G,F,D X ,D Y ,C X ,C Y )=L GAN (G,D Y ,X,Y)+L GAN (F,D X ,Y,X)+λ1L cyc (G,F)+λ2L c (C X )+λ3L c (C Y ); Where λ1, λ2, and λ3 are weight coefficients, which are used to control the importance of cycle consistency loss and the two classification losses, respectively.

11. The rotor-bearing system fault diagnosis method based on virtual-real fusion according to claim 1, characterized in that: The network structure of the one-dimensional convolutional neural network is the same as the classifier structure of the one-dimensional cyclic generative adversarial network.

12. A rotor-bearing system fault diagnosis system based on virtual-real fusion, characterized in that: The system is used to implement the rotor-bearing system fault diagnosis method based on virtual-real fusion according to any one of claims 1 to 11, specifically comprising: A model building module is used to derive a unified expression for the mixed eccentric unbalanced magnetic pull, introduce the nonlinear restoring force of the bearing based on Hertz contact theory, and establish a twin model of the rotor-bearing system; a fault introduction module, configured to introduce a rotor eccentricity fault, a bearing inner race fault, and a bearing outer race fault into the rotor-bearing system twin model, so as to obtain the rotor-bearing system twin models under different fault conditions; A model correction module is used to establish a fitness function by introducing a feature sensitivity evaluation factor into a genetic algorithm, and to perform multi-parameter identification on the rotor-bearing system twin model under the different fault conditions by combining adaptive crossover, mutation probability, and simulated annealing to obtain a corrected twin model; A sample generation module is used to obtain twin fault samples based on the modified twin model, add a classifier to the one-dimensional cyclic generative adversarial network, and design a comprehensive loss function to train and generate generated samples that are close to the distribution of real samples; The fault classification module is used to fuse the generated samples with the real samples to obtain a balanced data set, and implement fault classification by training a one-dimensional convolutional neural network.

Images