Small sample life prediction model training method based on multi-channel source domain transfer learning

Abstract

The application discloses a small sample life prediction model training method based on multi-channel source domain transfer learning, and relates to the technical field of predictive maintenance of rotating machinery. The method comprises the following steps: constructing a five-degree-of-freedom differential equation of the rotating machinery; based on the five-degree-of-freedom differential equation, constructing a target nonlinear dynamics model of the rotating machinery gradually from a healthy stage, a degradation stage to a damage stage; performing simulation analysis on the target nonlinear dynamics model to obtain source domain data of multiple channels; and training a preset life prediction model of the rotating machinery according to the source domain data of the multiple channels and target domain data of the rotating machinery, wherein the preset life prediction model comprises a public feature extractor, a domain-specific feature extractor, a domain-specific regressor and a domain-specific discriminator. The application can provide rich source domain data for the model when the training samples are scarce, improve the prediction accuracy of the model, and process the differences between the source domain data of different channels, thereby guaranteeing the generalization ability of the model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of predictive maintenance of rotating machinery, and particularly to a small-sample life prediction model training method based on multi-channel source domain transfer learning. BACKGROUND

[0002] Industrial equipment is increasingly large, automated, and intelligent. Bearings, gears, and other rotating machinery are widely used in aerospace, transportation, and other fields, and their operating conditions are directly related to the life of various mechanical equipment. Therefore, life prediction of bearings and other rotating machinery has positive practical significance.

[0003] Currently, data measurement is usually performed when the rotating machinery is working, and the actually measured data is used as training samples. However, due to the complex working environment of the rotating machinery, it is difficult to obtain a large number of training samples, which results in only a small number of samples or even no samples available for training, and the prediction accuracy of the model cannot be guaranteed. SUMMARY

[0004] Therefore, the present application provides a small-sample life prediction model training method based on multi-channel source domain transfer learning, which can provide rich source domain data for the model when the training samples are scarce, improve the prediction accuracy of the model, and handle the differences between source domain data of different channels to ensure the generalization ability of the model.

[0005] According to a first aspect of the present application, a small-sample life prediction model training method based on multi-channel source domain transfer learning is provided, which comprises:

[0006] constructing a five-degree-of-freedom differential equation of the rotating machinery;

[0007] based on the five-degree-of-freedom differential equation, constructing a target nonlinear dynamics model of the rotating machinery gradually from a healthy stage, a degradation stage to a damage stage;

[0008] performing simulation analysis on the target nonlinear dynamics model to obtain source domain data of multiple channels;

[0009] training a preset life prediction model of the rotating machinery according to the source domain data of the multiple channels and target domain data of the rotating machinery, wherein the preset life prediction model comprises a common feature extractor corresponding to the source domain data of the multiple channels, and domain-specific feature extractors, domain-specific regressors, and domain-specific discriminators corresponding to the source domain data of the multiple channels, respectively.

[0010] According to a second aspect of the present application, a small-sample life prediction model training device based on multi-channel source domain transfer learning is provided, which comprises:

[0011] constructing unit, configured to construct a five-degree-of-freedom differential equation of the rotating machinery;

[0012] The constructing unit is further configured to construct, based on the five-degree-of-freedom differential equation, a target nonlinear dynamics model of the rotating machinery gradually from a healthy stage, a degradation stage to a damage stage.

[0013] The simulation unit is configured to perform simulation analysis on the target nonlinear dynamics model to obtain source domain data of multiple channels.

[0014] The training unit is configured to train a preset life prediction model of the rotating machinery according to the source domain data of the multiple channels and target domain data of the rotating machinery, wherein the preset life prediction model comprises a common feature extractor corresponding to the source domain data of the multiple channels in common, and domain-specific feature extractors, domain-specific regressors and domain-specific discriminators corresponding to the source domain data of the multiple channels respectively.

[0015] According to a third aspect of the present application, a storage medium having a computer program stored thereon is provided, the program being executed by a processor to implement the above-mentioned small-sample life prediction model training method based on multi-channel source domain transfer learning.

[0016] According to a fourth aspect of the present application, an electronic device is provided, comprising a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, the processor executing the program to implement the above-mentioned small-sample life prediction model training method based on multi-channel source domain transfer learning.

[0017] Through the above technical solutions, the small-sample life prediction model training method based on multi-channel source domain transfer learning provided by the present application can construct a target nonlinear dynamics model of the rotating machinery gradually from a healthy stage, a degradation stage to a damage stage, and perform simulation analysis on the target nonlinear dynamics model, so as to generate a large amount of simulation data covering the whole cycle and different channel modes, thereby obtaining a variety of fault types and a large amount of source domain data, and further providing rich source domain data for the model when the training samples are scarce, and improving the prediction accuracy of the model. In addition, in the process of transfer learning, the common feature extractor corresponding to the source domain data of the multiple channels in common, and the domain-specific feature extractors, domain-specific regressors and domain-specific discriminators corresponding to the source domain data of the multiple channels respectively are trained, so that the source domain data and the target domain data of different channels can be aligned to different feature spaces, thereby processing the differences between the source domain data of different channels, and improving the generalization ability of the life prediction model.

[0018] The above description is only a summary of the technical solutions of the present application, in order to enable the technical means of the present application to be implemented according to the content of the description, and in order to enable the above and other purposes, characteristics and advantages of the present application to be more apparent and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0019] The drawings described herein are used to provide further understanding of the present application, and form a part of the present application, the schematic embodiments of the present application and the description thereof are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A flowchart of a small sample life prediction model training method based on multi-channel source domain transfer learning provided by an embodiment of the present application is shown;

[0021] Figure 2 A schematic diagram of a five-degree-of-freedom dynamic model provided by an embodiment of the present application is shown;

[0022] Figure 3 A flowchart of a nonlinear dynamic model construction method provided by an embodiment of the present application is shown;

[0023] Figure 4 A flowchart of a preset life prediction model training method provided by an embodiment of the present application is shown;

[0024] Figure 5 A flowchart of source domain data generation, model training and model testing provided by an embodiment of the present application is shown;

[0025] Figure 6 A structural diagram of a small sample life prediction model training device based on multi-channel source domain transfer learning provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] Hereinafter, the present application will be described in detail with reference to the drawings and in conjunction with embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0027] Because the working environment of the rotating machinery is relatively complex, it is difficult to obtain a large number of training samples, so that only a small number of samples or even no samples can be used for training, and the prediction accuracy of the model cannot be guaranteed.

[0028] In order to solve the above problems, an embodiment of the present application provides a small sample life prediction model training method based on multi-channel source domain transfer learning, as shown in the figure, the method comprises the following steps: Figure 1

[0029] Step 10, constructing a five-degree-of-freedom differential equation of the rotating machinery.​

[0030] Wherein, the rotating machinery can specifically include gears, engines, bearings, etc., and the bearing includes a plurality of rolling elements, a bearing inner ring and a bearing outer ring.

[0031] The embodiment of the present application is mainly suitable for health monitoring of rotating machinery, and mainly relates to key equipment such as aircraft engines, railway vehicles and wind driven generators.

[0032] The embodiment of the present application takes bearings as an example to illustrate the construction process of the five-degree-of-freedom differential equation and the target nonlinear dynamics model in detail, but is not limited to bearings.

[0033] Specifically, the constructed five-degree-of-freedom differential equation of the bearing is as follows,

[0034]

[0035] Wherein, m, k and c respectively represent mass, stiffness and damping, x and y respectively represent horizontal vibration and vertical vibration, the subscripts s, p and r respectively represent the bearing inner ring, the bearing outer ring and the unit resonator, F s represents an external radial load, f x and f y are horizontal and vertical nonlinear contact forces applied on the bearing inner ring and the bearing outer ring, and g is the acceleration of gravity.

[0036] Step 20, based on the five-degree-of-freedom differential equation, a target nonlinear dynamics model of the rotating machinery gradually from a healthy stage, a degradation stage to a damage stage is constructed.

[0037] Wherein, the target nonlinear dynamics model refers to the nonlinear dynamics model of the final damage stage.

[0038] In order to simulate the vibration behavior of the bearing in its entire service life, the embodiment of the present application constructs a five-degree-of-freedom dynamics model for different degradation modes, including the bearing inner ring, the bearing outer ring and the rolling element, as shown in Figure 2 The five-degree-of-freedom dynamics model includes a nonlinear dynamics model of a healthy stage, a nonlinear dynamics model of a degradation stage and a nonlinear dynamics model of a damage stage.

[0039] For the specific process of constructing the nonlinear dynamics model, as shown in Figure 3 , it includes:

[0040] Step 21, a first contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the healthy stage is constructed.

[0041] For the embodiment of the application, when the bearing is running in the healthy stage, there is no damage on the bearing inner ring, the bearing outer ring and the rolling element, and taking any one rolling element and the bearing inner ring and the bearing outer ring in contact with it as the research object, the first contact deformation formula for representing the corresponding deformation amount is as follows:

[0042] delta j =(x s -x p )cos theta j +(y s -y p )sin theta j +c

[0043] Wherein, theta j is the angle position of the jth rolling element, delta j is the deformation amount in the healthy stage, and c is the radial clearance of the bearing.

[0044] Step 22, substituting the first contact deformation formula into the five-degree-of-freedom differential equation to construct the nonlinear dynamic model of the bearing in the healthy stage.

[0045] For the embodiment of the application, substituting the first contact deformation formula into the above five-degree-of-freedom differential equation can obtain the nonlinear dynamic model in the healthy stage (normal working condition).

[0046] Step 23, adjusting the deformation of the nonlinear dynamic model in the healthy stage to obtain the nonlinear dynamic model of the bearing in the degradation stage.

[0047] For the embodiment of the application, for the degradation stage, based on the first contact deformation formula, the second contact deformation formula between the arbitrary rolling element and the bearing inner ring and the bearing outer ring in the case of bearing outer ring defect, the third contact deformation formula between the arbitrary rolling element and the bearing inner ring and the bearing outer ring in the case of bearing inner ring defect, and the fourth contact deformation formula between the arbitrary rolling element and the bearing inner ring and the bearing outer ring in the case of rolling element defect are respectively constructed; based on the second contact deformation formula, the third contact deformation formula and the fourth contact deformation formula, the deformation of the nonlinear dynamic model in the healthy stage is adjusted to obtain the nonlinear dynamic model of the bearing in the degradation stage.

[0048] Specifically, in the degradation stage, in order to simulate the influence of damage in different parts on the whole cycle degradation of the bearing, the embodiment of the application simulates three kinds of fault environment, and divides the degradation of the bearing into the degradation process of the bearing inner ring, the bearing outer ring and the rolling element respectively. The defect in the degradation stage is derived from the surface indentation, and when the rolling element passes through the defect, the surface topography change will generate excitation, which is reflected on the corresponding vibration signal.

[0049] For the inner ring defect, a second contact deformation formula for indicating the contact deformation between the rolling body and the raceway is as follows:

[0050] δ j = (x s - x p ) cos θ j + (y s - y p ) sin θ j + c + δ0

[0051]

[0052] wherein δ j is a deformation amount for the inner ring defect in the degradation stage, δ0 is a deformation adjustment amount for the inner ring defect of the bearing in the degradation stage, d is the diameter of the rolling body, L in is the width of the inner ring defect.

[0053] For the outer ring defect, a third contact deformation formula for indicating the contact deformation between the rolling body and the raceway is as follows:

[0054] δ j = (x s - x p ) cos θ j + (y s - y p ) sin θ j + c - δ0

[0055]

[0056] wherein δ j is a deformation amount for the outer ring defect in the degradation stage, δ0 is a deformation adjustment amount for the outer ring defect of the bearing in the degradation stage, d is the diameter of the rolling body, L out is the width of the outer ring defect.

[0057] For the rolling body defect, a fourth contact deformation formula for indicating the contact deformation between the rolling body and the raceway is as follows:

[0058] δ j = (x s - x p ) cos θ j + (y s - y p ) sin θ j + c + δ0

[0059] δ0 = h ball

[0060] wherein δ jis the deformation variable for the rolling element defect in the degradation stage, δ0 is the deformation adjustment variable for the rolling element defect in the degradation stage, d is the diameter of the rolling element, h ball is the width of the rolling element defect.

[0061] After obtaining the second contact deformation formula, the third contact deformation formula and the fourth contact deformation formula, the deformation variable in the nonlinear dynamic model of the healthy stage is replaced or adjusted by using the second contact deformation formula, the third contact deformation formula and the fourth contact deformation formula, to obtain the nonlinear dynamic model of the degradation stage.

[0062] Step 24, performing deformation adjustment on the nonlinear dynamic model of the degradation stage to obtain the nonlinear dynamic model of the bearing in the damage stage.

[0063] For the embodiment of the present application, for the damage stage, based on the second contact deformation formula, the third contact deformation formula and the fourth contact deformation formula, a fifth contact deformation formula between the arbitrary one of the rolling elements and the bearing inner ring and the bearing outer ring in the case of bearing outer ring defect, a sixth contact deformation formula between the arbitrary one of the rolling elements and the bearing inner ring and the bearing outer ring in the case of bearing inner ring defect, and a seventh contact deformation formula between the arbitrary one of the rolling elements and the bearing inner ring and the bearing outer ring in the case of rolling element defect are respectively constructed; based on the fifth contact deformation formula, the sixth contact deformation formula and the seventh contact deformation formula, deformation adjustment is performed on the nonlinear dynamic model of the degradation stage to obtain the nonlinear dynamic model of the bearing in the damage stage.

[0064] Specifically, in the damage stage, the defect size increases, the generated vibration excitation is more intense, and the defect can be simplified as a rectangular defect. Since the defect size increases, the rolling element can contact the bottom of the defect.

[0065] For the inner ring damage, the fifth contact deformation formula for representing the contact deformation between the rolling element and the raceway is as follows:

[0066] δ j = (x s -x p )cosθ j +(y s -y p )sinθ j +c-δ0

[0067]

[0068] wherein δ j is the deformation variable for the inner ring damage in the damage stage, δ0 is the deformation adjustment variable for the bearing inner ring damage in the damage stage, ΔΦ4 is the edge angle of the rolling element contacting the defect edge falling completely into the bottom of the defect, hin a width of a rectangular defect generated in the outer ring raceway.

[0069] For the outer ring damage, a sixth contact deformation formula for indicating a contact deformation between the rolling element and the raceway is as follows:

[0070] δ j = (x s - x p )cosθ j + (y s - y p )sinθ j + c + δ0

[0071]

[0072] wherein δ j is a deformation variable for the outer ring damage at the damage stage, δ0is a deformation adjustment variable for the bearing outer ring damage at the damage stage, h out is a width of a rectangular defect generated in the outer ring raceway.

[0073] For the rolling element damage, a seventh contact deformation formula for indicating a contact deformation between the rolling element and the raceway is as follows:

[0074] δ j = (x s - x p )cosθ j + (y s - y p )sinθ j + c + δ0

[0075] δ0= h ball

[0076] wherein δ j is a deformation variable for the rolling element damage at the damage stage, δ0is a deformation adjustment variable for the rolling element damage at the damage stage, h ball is a width of a defect of the rolling element.

[0077] After the fifth contact deformation formula, the sixth contact deformation formula and the seventh contact deformation formula are obtained, the deformation variable in the nonlinear dynamics model at the degradation stage is replaced or adjusted by using the same, to obtain a nonlinear dynamics model at the damage stage.

[0078] Step 25, determining the nonlinear dynamics model at the damage stage as the target nonlinear dynamics model.

[0079] The embodiment of the present application simulates a plurality of complex bearing degradation processes by establishing nonlinear dynamic models under different degradation modes (such as inner ring failure, outer ring failure and rolling element failure). Compared with the simplified assumption of only considering a single degradation mode, the embodiment of the present application can generate a large amount of simulation data covering the whole cycle and different channel modes, thereby providing rich source domain data for the model when the training sample is scarce.

[0080] Step 30, simulation analysis is performed on the target nonlinear dynamic model to obtain source domain data of multiple channels.

[0081] The source domain data of multiple channels includes source domain data of bearing inner ring failure, source domain data of bearing outer ring failure and source domain data of rolling element failure.

[0082] For the embodiment of the present application, in order to obtain the source domain data of multiple channels, step 103 specifically includes: solving the equation corresponding to the target nonlinear dynamic model to obtain time domain waveforms and frequency domain waveforms of the multiple channels; and determining the source domain data of multiple channels based on the time domain waveforms and the frequency domain waveforms.

[0083] Specifically, when solving the equation, the key characteristic frequencies such as BPFO, BPFI and BSF can be calculated, and by removing the direct current component of the frequency domain signal, the signals of the bearing inner ring, the bearing outer ring and the rolling element damage stage are further analyzed in time domain and frequency domain. In the time domain graph, obvious pulse signals can be observed, and especially when the rolling element passes through the defect part, the vibration signal presents regular pulses. In the frequency domain analysis, especially in the observation of the envelope spectrum, the simulation signal of the bearing inner ring damage shows a wave peak highly consistent with the theoretical calculation frequency, and the error is maintained within an acceptable range.

[0084] Step 40, training a preset life prediction model of the rotating machinery according to the source domain data of the multiple channels and the target domain data of the rotating machinery.

[0085] The preset life prediction model includes a common feature extractor corresponding to the source domain data of the multiple channels, and a domain-specific feature extractor, a domain-specific regressor and a domain-specific discriminator corresponding to the source domain data of the multiple channels respectively. The common feature extractor is composed of three convolutional layers, which are used to extract common representations from the source domain data and the target domain data to map the original data to a common feature space; the domain-specific feature extractor includes three convolutional layers, each of which is connected with an average pooling layer and an activation layer; the domain-specific regressor and the domain-specific discriminator are composed of fully connected layers.

[0086] It should be noted that the number of layers in the model in the embodiment of the present application is not limited to the above, and can also be other numbers of layers.

[0087] The training process for the preset life prediction model specifically includes the following steps, as shown in Figure 4

[0088] Step 41, constructing an initial common feature extractor corresponding to the source domain data of the plurality of channels, and initial domain-specific feature extractors, initial domain-specific regressors and initial domain-specific discriminators corresponding to the source domain data of the plurality of channels respectively.

[0089] The number of the initial common feature extractor is one, and the number of the initial domain-specific feature extractors, the initial domain-specific regressors and the initial domain-specific discriminators is the same as the number of the channels.

[0090] For the embodiment of the present application, the initial common feature extractor, the initial domain-specific feature extractors, the initial domain-specific regressors and the initial domain-specific discriminators are respectively constructed by initializing network parameters.

[0091] Step 42, inputting the source domain data of the plurality of channels and the target domain data into the initial common feature extractor respectively for feature extraction, to obtain common features between the source domain data of the plurality of channels and the target domain data respectively.

[0092] For example, the source domain data and the target domain data of channel A are input into the initial common feature extractor to extract the common features between the source domain data and the target domain data of channel A, and at the same time, the source domain data and the target domain data of channel B are input into the initial common feature extractor to extract the common features between the source domain data and the target domain data of channel B.

[0093] Step 43, inputting the source domain data of the plurality of channels and the corresponding common features into the corresponding initial domain-specific feature extractors for feature extraction, to obtain domain-specific features of the source domain data of the plurality of channels.

[0094] For the embodiment of the present application, since there are unique differences between different source domain data, an independent feature extractor needs to be designed for each specific domain to capture the uniqueness of each domain. The structure of each initial domain-specific feature extractor is uniform, including three convolutional layers, and an average pooling layer and an activation layer are connected after each convolutional layer. It should be noted that the number of convolutional layers is not limited to the above, and can also be other numbers.

[0095] For example, the source domain data of channel A and the corresponding common features are input into the initial domain-specific feature extractor corresponding to channel A for feature extraction. Similarly, the source domain data of channel B and the corresponding common features are input into the initial domain-specific feature extractor corresponding to channel B for feature extraction.

[0096] ​The embodiment of the application considers that the source domain data under different channel degradation modes have certain differences at the feature level, and it is difficult to align the source domain data directly with the target domain data to a feature space, therefore, the embodiment of the application adopts a multi-channel source domain transfer model domain nonlinear dynamics knowledge fusion mode, which aligns the source domain data and the target domain data of each channel to different feature spaces and optimizes the differences.

[0097] Step 44: inputting the domain-specific features of the source domain data of the plurality of channels into corresponding initial domain-specific regressors and initial domain-specific discriminators for life prediction to obtain predicted lives output by the plurality of initial domain-specific regressors.

[0098] wherein the number of the initial domain-specific regressors and the initial domain-specific discriminators is the same as the number of the initial domain-specific feature extractors.

[0099] The embodiment of the application can ensure that the source domain data and the target domain data from different channels do not interfere with each other through the respective feature extractors and regressors by adopting the plurality of initial domain-specific feature extractors and the plurality of initial domain-specific regressors.

[0100] Step 45: constructing a loss function according to the common features, the domain-specific features and the predicted lives output by the plurality of initial domain-specific regressors.

[0101] For the embodiment of the application, in constructing the loss function, step 45 specifically comprises: calculating a regression error loss according to the predicted lives output by the plurality of initial domain-specific regressors and the real life of the rotating machinery; calculating a transfer domain adaptation loss according to the source domain data of the plurality of channels and the target domain data; calculating an output difference loss between different initial domain-specific regressors according to the predicted lives output by the plurality of initial domain-specific regressors; calculating a total adversarial loss of the plurality of initial domain-specific discriminators according to the common features and the domain-specific features; and constructing the loss function according to the regression error loss, the transfer domain adaptation loss, the output difference loss and the total adversarial loss.

[0102] Specifically, the specific formula of the finally constructed loss function is as follows:

[0103]

[0104] wherein, is a training parameter of the common feature extractor, is a training parameter of the i th domain-specific feature extractor, is a training parameter of the i th domain-specific discriminator, is a training parameter of the i th domain-specific regressor, L reg is a regression error loss, L MMDL alig is the output difference loss between different domain-specific regressors. adv is the total adversarial loss of multiple domain-specific discriminators, and α, β, γ are weight factors corresponding to different loss functions in the training process.

[0105] In calculating the regression error loss L reg , the difference between the predicted lifetime and the real lifetime output by the initial domain-specific regressor can be used to determine the regression error loss.

[0106] In calculating the transfer domain adaptation loss L MMD , the difference between the source domain data and the target domain data can be evaluated by calculating the MMD distance, and the difference is used to represent the transfer domain adaptation loss L MMD , and the specific calculation formula is as follows:

[0107]

[0108] where X S is the source domain data, n S is the number of samples corresponding to the source domain data, X T is the target domain data, n T is the number of samples corresponding to the target domain data, H is the reproducing Hilbert space, and φ represents the mapping of X S and X T in H.

[0109] In calculating the output difference loss between different domain-specific regressors, the specific formula is as follows:

[0110]

[0111] wherein, and represent the output of different domain-specific regressors, i.e. the predicted lifetime, and N represents the number of domain-specific regressors.

[0112] In calculating the total adversarial of multiple domain-specific discriminators, the specific formula is as follows:

[0113]

[0114] wherein, E S,i and E T,i represent the i-th group of source domain data and target domain data sequentially passing through the common feature extractor and the domain-specific feature extractor for feature extraction, L adv represents the total adversarial loss of multiple domain-specific discriminators.

[0115] Thus, the regression error loss, the transfer domain adaptation loss, the output difference loss and the total adversarial loss can be calculated according to the above formula, and the total loss function can be obtained by adding the regression error loss, the transfer domain adaptation loss, the output difference loss and the total adversarial loss.

[0116] Step 46, based on the loss function, iteratively training the initial common feature extractor, the initial domain-specific feature extractor, the initial domain-specific regressor and the initial domain-specific discriminator until a preset number of iterations is reached, stopping training, and outputting the common feature extractor, the domain-specific feature extractor, the domain-specific regressor and the domain-specific discriminator.

[0117] For the embodiment of the application, after the total loss function is constructed, the parameters in the initial common feature extractor, the initial domain-specific feature extractor, the initial domain-specific regressor and the initial domain-specific discriminator are continuously updated based on the total loss function until a preset number of iterations is reached, the iteration is stopped, and finally the trained common feature extractor, domain-specific feature extractor, domain-specific regressor and domain-specific discriminator are output.

[0118] The multi-channel source domain fusion transfer learning model provided by the embodiment of the application aligns the source domain data of each channel with the target domain data to different feature spaces, and designs different adapters for feature extraction and modeling. Through the feature optimization strategy, the feature difference between multiple source domains can be effectively reduced, and the generalization ability and prediction accuracy of the model can be improved. In addition, the embodiment of the application combines the simulation data generated by the dynamic model with the multi-source domain transfer learning model to realize the effective fusion of data-driven and knowledge-driven, so as to realize more accurate and reliable remaining life prediction in a small sample scene, and thus to solve the dual challenges of data scarcity and distribution difference.

[0119] Step 47, determining the preset life prediction model according to the common feature extractor, the domain-specific feature extractor, the domain-specific regressor and the domain-specific discriminator.

[0120] For the embodiment of the application, after the preset life prediction model is constructed, the target domain test data of the rotating machinery is input into the preset life prediction model for life prediction, and the life prediction result of the rotating machinery is obtained. The overall process of the embodiment of the application is shown in Figure 5 , which includes simulation original domain generation, model training and model testing.

[0121] The embodiment of the application provides a small sample life prediction model training method based on multi-channel source domain transfer learning, a target nonlinear dynamics model of a rotating machine gradually from a healthy stage, a degradation stage to a damage stage is constructed, simulation analysis is performed on the target nonlinear dynamics model, a large amount of simulation data under different channel modes covering a whole cycle can be generated, so that source domain data of various fault types and in large quantity can be obtained, and then the source domain data can be provided for the model when training samples are scarce, and the prediction accuracy of the model is improved. In addition, in the process of transfer learning, the common feature extractor corresponding to the source domain data of the plurality of channels, the domain-specific feature extractor, the domain-specific regressor and the domain-specific discriminator corresponding to the source domain data of the plurality of channels can align the source domain data and the target domain data of different channels to different feature spaces, so that the difference between the source domain data of different channels can be processed, and the generalization ability of the life prediction model is improved.

[0122] Further, as a specific implementation of the method shown in Figure 1 、 Figure 3 and Figure 4 , the embodiment provides a small sample life prediction model training device based on multi-channel source domain transfer learning, as shown in Figure 6 , the device comprises a construction unit 101, a simulation unit 102 and a training unit 103.

[0123] The construction unit 101 can be used to construct a five-degree-of-freedom differential equation of a rotating machine.

[0124] The construction unit 101 can also be used to construct a target nonlinear dynamics model of the rotating machine gradually from a healthy stage, a degradation stage to a damage stage based on the five-degree-of-freedom differential equation.

[0125] The simulation unit 102 can be used to perform simulation analysis on the target nonlinear dynamics model to obtain source domain data of a plurality of channels.

[0126] The training unit 103 can be used to train a preset life prediction model of the rotating machine according to the source domain data of the plurality of channels and the target domain data of the rotating machine, wherein the preset life prediction model comprises a common feature extractor corresponding to the source domain data of the plurality of channels, and a domain-specific feature extractor, a domain-specific regressor and a domain-specific discriminator corresponding to the source domain data of the plurality of channels.

[0127] In some embodiments, the rotating machine comprises a bearing, the bearing comprises a plurality of rolling bodies, a bearing inner ring and a bearing outer ring, the construction unit 101 comprises a first construction module, an adjustment module and a first determination module.

[0128] The first construction module can be configured to construct a first contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring at the health stage.

[0129] The first construction module can be further configured to substitute the first contact deformation formula into the five-degree-of-freedom differential equation to construct a nonlinear dynamic model of the bearing at the health stage.

[0130] The adjustment module can be configured to perform deformation adjustment on the nonlinear dynamic model at the health stage to obtain a nonlinear dynamic model of the bearing at the degradation stage.

[0131] The adjustment module can be further configured to perform deformation adjustment on the nonlinear dynamic model at the degradation stage to obtain a nonlinear dynamic model of the bearing at the damage stage.

[0132] The first determination module can be configured to determine the nonlinear dynamic model at the damage stage as the target nonlinear dynamic model.

[0133] In some embodiments, the adjustment module can be specifically configured to, for the degradation stage, construct, based on the first contact deformation formula, a second contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the case of a bearing outer ring defect, a third contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the case of a bearing inner ring defect, and a fourth contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the case of a rolling element defect; and perform deformation adjustment on the nonlinear dynamic model at the health stage based on the second contact deformation formula, the third contact deformation formula, and the fourth contact deformation formula to obtain the nonlinear dynamic model of the bearing at the degradation stage.

[0134] In some embodiments, the adjustment module can be further specifically configured to, for the damage stage, construct, based on the second contact deformation formula, the third contact deformation formula, and the fourth contact deformation formula, a fifth contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the case of a bearing outer ring defect, a sixth contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the case of a bearing inner ring defect, and a seventh contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the case of a rolling element defect; and perform deformation adjustment on the nonlinear dynamic model at the degradation stage based on the fifth contact deformation formula, the sixth contact deformation formula, and the seventh contact deformation formula to obtain the nonlinear dynamic model of the bearing at the damage stage.

[0135] In some embodiments, the simulation unit 102 can be specifically configured to solve equations corresponding to the target nonlinear dynamic model to obtain time-domain waveforms and frequency-domain waveforms of the plurality of channels; and determine source domain data of the plurality of channels based on the time-domain waveforms and the frequency-domain waveforms.

[0136] In some embodiments, the training unit 103 includes a second construction module, an extraction module and a prediction module, a training module, and a second determination module.

[0137] The second construction module can be configured to construct an initial common feature extractor corresponding to the source domain data of the plurality of channels, and initial domain-specific feature extractors, initial domain-specific regressors, and initial domain-specific discriminators corresponding to the source domain data of the plurality of channels, respectively.

[0138] The extraction module can be configured to input the source domain data of the plurality of channels and the target domain data into the initial common feature extractor for feature extraction, respectively, to obtain common features between the source domain data of the plurality of channels and the target domain data.

[0139] The extraction module can also be configured to input the source domain data of the plurality of channels and the corresponding common features into the corresponding initial domain-specific feature extractors for feature extraction, to obtain domain-specific features of the source domain data of the plurality of channels.

[0140] The prediction module can be configured to input the domain-specific features of the source domain data of the plurality of channels into the corresponding initial domain-specific regressors and initial domain-specific discriminators for life prediction, to obtain predicted lives output by the plurality of initial domain-specific regressors.

[0141] The second construction module can also be configured to construct a loss function according to the common features, the domain-specific features, and the predicted lives output by the plurality of initial domain-specific regressors.

[0142] The training module can be configured to perform iterative training on the initial common feature extractor, the initial domain-specific feature extractors, the initial domain-specific regressors, and the initial domain-specific discriminators based on the loss function, and stop training when a preset number of iterations is reached, and output the common feature extractor, the domain-specific feature extractors, the domain-specific regressors, and the domain-specific discriminators.

[0143] The second determination module can be configured to determine the preset life prediction model according to the common feature extractor, the domain-specific feature extractors, the domain-specific regressors, and the domain-specific discriminators.

[0144] In some embodiments, the second construction module can be specifically configured to calculate a regression error loss according to the predicted life span output by the plurality of initial domain-specific regressors and the actual life span of the rotating machinery, calculate a transfer domain adaptation loss according to the source domain data and the target domain data of the plurality of channels, calculate an output difference loss between different domain-specific regressors according to the predicted life span output by the plurality of initial domain-specific regressors, calculate a total adversarial loss of a plurality of domain-specific discriminators according to the common features and the domain-specific features, and construct the loss function according to the regression error loss, the transfer domain adaptation loss, the output difference loss and the total adversarial loss.

[0145] In some embodiments, the device further comprises an acquisition unit and a prediction unit.

[0146] The acquisition unit can be configured to acquire the target domain test data of the rotating machinery.

[0147] The prediction unit can be configured to input the target domain test data into the preset life span prediction model for life span prediction, to obtain a life span prediction result of the rotating machinery.

[0148] It should be noted that other corresponding descriptions of the various functional units involved in the device for training a small-sample life span prediction model based on multi-channel source domain transfer learning provided in the embodiments of the present application can be referred to the corresponding descriptions in Figure 1 、 Figure 3 and Figure 4 , which will not be repeated here.

[0149] Based on the above methods as shown in Figure 1 、 Figure 3 and Figure 4 , correspondingly, the present embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to implement the above-mentioned life span prediction model training method based on multi-channel source domain transfer learning as shown in Figure 1 、 Figure 3 and Figure 4 .

[0150] Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions for causing an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various implementation scenarios of the present application.

[0151] Based on the above methods as shown in Figure 1 、 Figure 3 and Figure 4 , and Figure 6In order to achieve the above-mentioned purposes, the electronic device provided by the embodiment of the present application can be a personal computer, a tablet computer, a server, or other network devices, etc., which comprises a storage medium and a processor; the storage medium is used for storing a computer program; and the processor is used for executing the computer program to realize the above-mentioned multi-channel source domain transfer learning based small sample life prediction model training method. Figure 1 、 Figure 3 and Figure 4 the multi-channel source domain transfer learning based small sample life prediction model training method.

[0152] Optionally, the above-mentioned entity device can further comprise a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface can comprise a display screen, an input unit such as a keyboard, etc. The optional user interface can further comprise a USB interface, a card reader interface, etc. The network interface can optionally comprise a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0153] Those skilled in the art can understand that the above-mentioned entity device structure provided by the embodiment does not constitute a limitation on the entity device, and can comprise more or fewer components, or combine certain components, or different component arrangements.

[0154] The storage medium can further comprise an operating system and a network communication module. The operating system is a program for managing the hardware and software resources of the above-mentioned entity device, and supports the running of the information processing program and other software and / or programs. The network communication module is used for realizing the communication between the components in the storage medium, and the communication with other hardware and software in the information processing entity device.

[0155] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary general hardware platforms, or by hardware.

[0156] The embodiment of the application can generate a large amount of simulation data under different channel modes by constructing a target nonlinear dynamic model of a rotating machine gradually from a healthy stage, a degradation stage to a damage stage and performing simulation analysis on the target nonlinear dynamic model, thereby obtaining a plurality of fault types and a large amount of source domain data, and further providing rich source domain data for the model when training samples are scarce, and improving the prediction accuracy of the model. In addition, in the process of transfer learning, by training a common feature extractor corresponding to the source domain data of a plurality of channels, and a domain-specific feature extractor, a domain-specific regressor and a domain-specific discriminator corresponding to the source domain data of a plurality of channels, the source domain data and the target domain data of different channels can be aligned to different feature spaces, thereby processing the difference between the source domain data of different channels, and improving the generalization ability of the life prediction model.

[0157] Those skilled in the art can understand that the drawings are only schematic diagrams of a preferred implementation scenario, and the modules or flows in the drawings are not necessarily required for implementing the present application. Those skilled in the art can understand that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the description of the implementation scenario, or can be changed and located in one or more devices different from the implementation scenario. The modules of the above implementation scenario can be combined into one module, or can be further split into a plurality of sub-modules.

[0158] The above application serial numbers are only for description, and do not represent the advantages and disadvantages of the implementation scenario. The above disclosure is only a few specific implementation scenarios of the present application, but the present application is not limited thereto, and any changes that can be thought of by those skilled in the art should fall within the protection scope of the present application.

Claims

1. A method for training a small sample life prediction model based on multi-channel source domain transfer learning, characterized in that, The method comprises the following steps: constructing a five-degree-of-freedom differential equation of a rotating machine; based on the five-degree-of-freedom differential equation, constructing a target nonlinear dynamic model of the rotating machine gradually from a healthy stage, a degradation stage to a damage stage; performing simulation analysis on the target nonlinear dynamic model to obtain source domain data of multiple channels; training a preset life prediction model of the rotating machine according to the source domain data of the multiple channels and target domain data of the rotating machine, wherein the preset life prediction model comprises a common feature extractor corresponding to the source domain data of the multiple channels, and domain-specific feature extractors, domain-specific regressors and domain-specific discriminators corresponding to the source domain data of the multiple channels respectively; wherein the rotating machine comprises a bearing, the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring, and the bearing ​ ​ ​ ​ ​ ​ ​ ​ 2. The method of claim 1, wherein, ​ For the damage stage, based on the second contact deformation formula, the third contact deformation formula and the fourth contact deformation formula, a fifth contact deformation formula between the arbitrary rolling body and the bearing inner ring and the bearing outer ring in the case of bearing outer ring defect, a sixth contact deformation formula between the arbitrary rolling body and the bearing inner ring and the bearing outer ring in the case of bearing inner ring defect, and a seventh contact deformation formula between the arbitrary rolling body and the bearing inner ring and the bearing outer ring in the case of rolling body defect are respectively constructed. Based on the fifth contact deformation formula, the sixth contact deformation formula and the seventh contact deformation formula, the nonlinear dynamic model of the degradation stage is deformed and adjusted to obtain the nonlinear dynamic model of the bearing in the damage stage.

3. The method of claim 1, wherein, The simulation analysis on the target nonlinear dynamic model obtains source domain data of multiple channels, including: Solving the equation corresponding to the target nonlinear dynamic model obtains time domain waveforms and frequency domain waveforms of the multiple channels; Based on the time domain waveforms and the frequency domain waveforms, source domain data of multiple channels are determined.

4. The method according to any one of claims 1 to 3, characterized in that, The training of the preset life prediction model of the rotating machinery according to the source domain data of the multiple channels and the target domain data of the rotating machinery includes: An initial common feature extractor corresponding to the source domain data of the multiple channels, and initial domain-specific feature extractors, initial domain-specific regressors and initial domain-specific discriminators corresponding to the source domain data of the multiple channels are constructed; The source domain data of the multiple channels and the target domain data are respectively input into the initial common feature extractor for feature extraction to obtain common features between the source domain data of the multiple channels and the target domain data; The source domain data of the multiple channels and the corresponding common features are input into the corresponding initial domain-specific feature extractors for feature extraction to obtain domain-specific features of the source domain data of the multiple channels; The domain-specific features of the source domain data of the multiple channels are input into the corresponding initial domain-specific regressors and initial domain-specific discriminators for life prediction to obtain predicted lives output by the multiple initial domain-specific regressors; A loss function is constructed according to the common features, the domain-specific features and the predicted lives output by the multiple initial domain-specific regressors; Based on the loss function, the initial common feature extractor, the initial domain-specific feature extractors, the initial domain-specific regressors and the initial domain-specific discriminators are iteratively trained until a preset iteration number is reached, and the training is stopped, and the common feature extractor, the domain-specific feature extractor, the domain-specific regressor and the domain-specific discriminator are output. The preset life prediction model is determined according to the common feature extractor, the domain-specific feature extractor, the domain-specific regressor and the domain-specific discriminator.

5. The method of claim 4, wherein, The construction of the loss function according to the common features, the domain-specific features and the predicted lives output by the multiple initial domain-specific regressors includes: According to the predicted life output by the plurality of initial domain-specific regressors and the real life of the rotating machinery, a regression error loss is calculated; According to the source domain data and the target domain data of the plurality of channels, a transfer domain adaptation loss is calculated; According to the predicted life output by the plurality of initial domain-specific regressors, an output difference loss between different domain-specific regressors is calculated; According to the common features and the domain-specific features, a total adversarial loss of a plurality of domain-specific discriminators is calculated; According to the regression error loss, the transfer domain adaptation loss, the output difference loss and the total adversarial loss, the loss function is constructed.

6. The method of claim 1, wherein, The method further comprises: acquiring target domain test data of the rotating machinery; inputting the target domain test data into the preset life prediction model for life prediction to obtain a life prediction result of the rotating machinery.

7. A small sample life prediction model training device based on multi-channel source domain transfer learning, characterized in that, comprise: a construction unit configured to construct a five-degree-of-freedom differential equation of a rotating machinery, wherein the rotating machinery comprises a bearing, and the bearing comprises a plurality of rolling elements, a bearing inner ring and a bearing outer ring; the construction unit is further configured to construct, based on the five-degree-of-freedom differential equation, a target nonlinear dynamics model of the rotating machinery gradually changing from a healthy stage to a degradation stage and then to a damage stage; a simulation unit configured to perform simulation analysis on the target nonlinear dynamics model to obtain source domain data of a plurality of channels; a training unit configured to train a preset life prediction model of the rotating machinery according to the source domain data of the plurality of channels and target domain data of the rotating machinery, wherein the preset life prediction model comprises a common feature extractor corresponding to the source domain data of the plurality of channels in common, and domain-specific feature extractors, domain-specific regressors and domain-specific discriminators corresponding to the source domain data of the plurality of channels respectively; the construction unit comprises a first construction module, an adjustment module and a first determination module, the first construction module is configured to construct a first contact deformation formula between any one of the plurality of rolling elements and the bearing inner ring and the bearing outer ring in the healthy stage; the first construction module is further configured to substitute the first contact deformation formula into the five-degree-of-freedom differential equation to construct a nonlinear dynamics model of the bearing in the healthy stage; the adjustment module is configured to perform deformation adjustment on the nonlinear dynamics model of the healthy stage to obtain a nonlinear dynamics model of the bearing in the degradation stage; the adjustment module is further configured to perform deformation adjustment on the nonlinear dynamics model of the degradation stage to obtain a nonlinear dynamics model of the bearing in the damage stage; the first determination module is configured to determine the nonlinear dynamics model of the damage stage as the target nonlinear dynamics model; The adjustment module is specifically configured to, for the degradation stage, based on the first contact deformation formula, respectively construct a second contact deformation formula between the arbitrary one of the rolling bodies and the bearing inner ring and the bearing outer ring in the case of bearing outer ring defect, a third contact deformation formula between the arbitrary one of the rolling bodies and the bearing inner ring and the bearing outer ring in the case of bearing inner ring defect, and a fourth contact deformation formula between the arbitrary one of the rolling bodies and the bearing inner ring and the bearing outer ring in the case of rolling body defect; and based on the second contact deformation formula, the third contact deformation formula, and the fourth contact deformation formula, perform deformation adjustment on the nonlinear dynamics model of the healthy stage to obtain the nonlinear dynamics model of the bearing in the degradation stage.

8. A storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, implements the method of any one of claims 1 to 6.

9. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor, when executing the computer program, implements the method of any one of claims 1 to 6.

Images