Zero sample fault online diagnosis method based on multi-resolution diffusion model

Abstract

The embodiment of the invention provides a zero sample fault online diagnosis method based on a multi-resolution diffusion model. Firstly, data collected by a traction transmission control simulation platform is used for constructing a source domain data set US, and data collected during actual operation of a train is used for constructing a target domain data set UT; the fault category sets of the source domain and the target domain are not crossed, and data are processed through a multi-resolution diffusion model, so that abnormal value removal, missing data filling and standardized processing are realized. And constructing a category-attribute description matrix A ', and extracting a mapping relationship among the fault feature fis, the feature Bs and the attribute description ai of the source domain data through depth increment supervision principal component analysis. After target domain data Dt is obtained online, attribute description of fault categories is updated, a new sparse matrix A * is constructed, and model parameters are finely adjusted. And finally, performing real-time fault diagnosis on the zero sample data of the target domain through Bayesian reasoning, and inferring an optimally matched fault category according to the maximum posterior probability to realize accurate diagnosis and prediction of the data of the target domain.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of fault diagnosis of electric traction systems, and in particular to a zero-sample fault online diagnosis method based on a multi-resolution diffusion model. Background Art

[0002] With the widespread application of electric traction systems in modern transportation systems, ensuring the stability and safety of their operation has become an important technical challenge. As a core component of transportation tools such as railways and subways, the electric traction system mainly adjusts the operating status of the train by controlling the speed and torque of the motor. However, there are many types of faults that may occur in the actual operation of the electric traction system, which not only affect the operating efficiency of the train, but may also cause serious safety accidents.

[0003] At present, existing fault diagnosis methods usually rely on a large amount of labeled data for training, which is often difficult to achieve in practical applications, especially when new or rare fault types appear. Existing fault diagnosis methods rely on a large number of fault samples and are difficult to adapt to the fault detection needs in zero-sample scenarios in actual operation. In addition, when facing the complex and diverse working conditions of electric traction systems, existing fault diagnosis methods are difficult to balance real-time, accuracy and versatility. Therefore, a new method is urgently needed to make up for the shortcomings of existing technologies. Summary of the invention

[0004] An embodiment of the present invention provides a zero-sample fault online diagnosis method based on a multi-resolution diffusion model, which realizes real-time detection and diagnosis of zero-sample faults in electric traction systems by processing source domain data and target domain data in stages and combining deep incremental supervised learning.

[0005] In a first aspect, an embodiment of the present invention provides a zero-sample fault online diagnosis method based on a multi-resolution diffusion model, comprising the following steps:

[0006] S1), define the data collected by the traction transmission control simulation platform as source domain data, and define the data collected during the actual operation of the train as target domain data; construct a source domain dataset U S and the target domain dataset U T : Source domain dataset where Y s represents the source domain data sample space, E s represents the fault category space of source domain data samples, N s Represents the number of source domain data samples; target domain data where Y t represents the target domain data sample space, E t represents the fault category space of the target domain data samples, N tRepresents the number of target domain data samples;

[0007] Construct the fault category set F of the source domain and the target domain respectively s and F t , the fault categories of the source domain include common single faults and compound faults, and the fault categories of the target domain include compound faults that occur under different working conditions; where h is the number of fault categories in the source domain, q is the number of fault categories in the target domain, and the total number of fault categories is r = h + k. The set of fault categories of the source domain and the target domain is F s and F t do not intersect each other, that is, F s ∩F t =φ;

[0008] S2) In establishing a multi-resolution diffusion model, the source domain U obtained by data collection is S and the target domain U t Input the data and get the reconstructed data and , realizing functions such as removal of outliers, filling of missing fault data and standardization processing.

[0009] The multi-resolution diffusion model divides the time resolution into multiple stages from fine to coarse, and performs stage-by-stage analysis and reconstruction of the operating data of the traction transmission control system. Through the decomposition and reconstruction of different time resolutions, the fault signal is effectively restored, and the denoising performance and signal quality are improved.

[0010] S3), construct the fault category set F of the source domain and the target domain respectively s and F t , summarize the source domain fault samples Y s Property description of a i (i) Establish a category-attribute description matrix A' and transform the matrix A' into a sparse category-attribute description matrix A through a one-hot encoder;

[0011] S4) The fault category space E s Associate the attribute label with the category attribute description matrix A to obtain the attribute label L s , and use deep incremental supervised principal component analysis (DI-SPCA) to extract the fault features f of the source domain data samples i s , construct the feature matrix B of the source domain samples s ; By training the learner model α i (i=1,2,3...,C) performs attribute learning to obtain the feature B of the source domain sample s With attribute description a i The mapping relationship between them;

[0012] S5) Operation data D obtained online t (target domain data), update the attribute description of the fault category, and construct a new sparse category-attribute description matrix A * ; and based on the updated matrix, the attribute learner α i Fine-tune parameters;

[0013] S6), for the input target domain sample , according to the output a of the attribute learner t , use Bayesian Inference to infer samples from attribute values The fault category of the fault category is, and the posterior probability of the fault category is, and the category with the largest posterior probability is selected as the optimal matching fault type to achieve real-time inference diagnosis of the fault type of zero-sample data in the target domain.

[0014] In some embodiments, the operating parameters of the source domain data set and the target domain data set in step S1) include: the inverter three-phase current {I sa , I sb , I sc}, train speed SP, motor speed Rs, motor electromagnetic torque Te, intermediate DC circuit upper and lower support capacitor voltage {U cd1 U cd2}、Transformer secondary side AC voltage U n And current data U n .

[0015] In some embodiments, the multi-resolution diffusion model described in step S2) performs multi-stage processing on the source domain data and the target domain data through stages with gradually increasing temporal resolution, and each stage gradually restores the fault characteristics by performing staged analysis and reconstruction on the data at different resolutions;

[0016] S21), divide the preprocessed data into training set S train and the test set S test ; The initial data is segmented into time series through the historical window LW and the current window CW, where the historical window LW covers data from 1 second to 5 seconds, and the current window CW covers data from 1.5 seconds to 5.5 seconds; the segmented data is decomposed from fine to coarse in four different time resolutions of 10μs, 100μs, 1ms, and 10ms to generate multi-resolution data;

[0017] S22), forward diffusion process, adding noise to the source data to generate noisy data Noise using diffusion coefficient and Gaussian noise: where Z is Gaussian noise sampled from a standard normal distribution N(0,1), β Sis the variance coefficient of Z, γ S =1-β S is the forward diffusion coefficient, is the cumulative diffusion coefficient;

[0018] S23) The reverse reconstruction process uses a Markov process. The state of each time step t depends only on the reconstructed state of the previous step. The reverse reconstruction includes a conditional network and a reconstruction network. In the S stage, the signal to be reconstructed is used. Signal X with the same time resolution ls As a reconstruction condition, obtaining reconstructed fault data;

[0019] S24), the model uses the segmentation error SmoothL1 as the loss function, and the loss function is:

[0020]

[0021] Among them, θ S are all the parameters of the conditional network and the reconstructed network in the S stage, yes In the reconstruction condition c S The output of the denoising network is passed through the denoising network; this loss measures the difference between the noise generated by the network and the real noise, and is optimized by iterative training adjustment θ S , minimize L SmoothL1 (θ S ) to improve the performance of the reconstructed network.

[0022] In some embodiments, in step S3), the source domain fault samples Y are summarized s Property description of a i (i) It consists of C attributes. The description of h types of fault data and categories is represented as a category-attribute description matrix A'. The one-hot encoder is used to transform the matrix of source domain fault description A' into a sparse category-attribute description matrix A: A∈R h×c = one-hot(A'), all elements in the matrix A are 1 or 0, indicating whether the attribute exists or not.

[0023] In some embodiments, in step S3) and step S4), DI-SPCA extracts the main features in the high-dimensional data by layer-by-layer dimensionality reduction, and maps the original data to a low-dimensional space. The feature extraction mapping function ζ of DI-SPCA is used to convert the source domain data sample Y s and the corresponding attribute tags As input, extract the fault feature vector f i s It is expressed as:

[0024]

[0025] In some embodiments, in step S5), during the online update process, the updated fault description matrix A is converted into *′ ∈R r*c ; Transformed into a new sparse category-attribute description matrix A * , to reflect new fault types and changes in operating conditions.

[0026] In some embodiments, in step S5) and step S6), the posterior probability of the fault category is expressed as:

[0027]

[0028] in, Since there is a category-attribute description matrix A * Provides the association information between the target fault category and attribute, with p(a t |t)=1, comprehensive sample Fault characteristics and predicted attribute label a t , the posterior probability is expressed as:

[0029]

[0030] in, is the feature f i t Under the condition, the property The probability of belonging to fault category t; Is an attribute The prior probability among all fault categories; p(t) is the prior probability of each category.

[0031] In some embodiments, in step S6), in the calculation of the posterior probability of the fault category, it is assumed that the prior probabilities of all categories are equal, that is, p(t) = constant, and the maximum a posteriori calculation in the next step has no effect on the result; , select the class with the largest posterior probability as the sample The best matching fault type:

[0032]

[0033] in, Represents the attribute description vector of the e-th fault type of the target domain sample, yes The i-th element of ; thus, the category-attribute description matrix A *It plays the role of knowledge transfer. Through the attribute information of known fault categories, the model applies this knowledge to fault samples of unknown categories to achieve real-time reasoning and diagnosis of fault types of zero-sample data in the target domain. After the reasoning and diagnosis are completed, the fault category prediction results generated by the model will be used to guide system maintenance and fault handling.

[0034] In a second aspect, an embodiment of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the zero-sample fault online diagnosis method based on a multi-resolution diffusion model as described in the first aspect is implemented.

[0035] In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the zero-sample fault online diagnosis method based on a multi-resolution diffusion model as described in the first aspect.

[0036] According to the zero-sample fault online diagnosis method based on the multi-resolution diffusion model provided by the embodiment of the present invention, firstly, the data collected by the traction transmission control simulation platform is used to construct a source domain data set U S , the data collected during the actual operation of the train is used to construct the target domain dataset U T The fault category sets of the source domain and the target domain do not overlap. The data is processed by a multi-resolution diffusion model to achieve outlier removal, missing data filling and standardization. Next, the category-attribute description matrix A' is constructed, and the fault features f of the source domain data are extracted by deep incremental supervised principal component analysis (DI-SPCA). i s , Feature B s With attribute description a i The mapping relationship between them. Get the target domain data D online t After that, update the attribute description of the fault category and construct a new sparse matrix A * , and fine-tune the model parameters. Finally, Bayesian reasoning is used to perform real-time fault diagnosis on zero-sample data in the target domain, and the optimal matching fault category is inferred based on the maximum a posteriori probability, so as to achieve accurate diagnosis and prediction of target domain data, thereby guiding system maintenance and fault handling. Based on this, the embodiment of the present invention uses the attributes of known fault categories to diagnose new fault types and compound faults. By using fault categories and attributes as learning objects, it is possible to achieve attribute migration from source domain fault categories to target domain fault categories in the absence of target domain fault category data samples, thereby achieving zero-sample fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1It is a flowchart of a zero-sample fault online diagnosis method based on a multi-resolution diffusion model provided by an embodiment of the present invention;

[0038] Figure 2 It is a periodic interference fault waveform of the inverter A phase current sensor at a resolution of [10e-6] provided by an embodiment of the present invention;

[0039] Figure 3 This is a waveform diagram of a periodic interference fault of a current sensor of phase A of an inverter at a resolution of [5e-3] provided by an embodiment of the present invention;

[0040] Figure 4 This is a waveform diagram of a periodic interference fault of a current sensor of phase A of an inverter at a resolution of [20e-3] provided by an embodiment of the present invention;

[0041] Figure 5 This is a denoising network flow chart of a multi-resolution diffusion model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0043] In the embodiments of the present invention, the words "further", "exemplarily" or "preferably" are used to indicate examples, illustrations or descriptions, and should not be interpreted as being more preferred or more advantageous than other embodiments or designs. The use of the words "further", "exemplarily" or "preferably" is intended to present the related concepts in a specific way.

[0044] At present, existing fault diagnosis methods usually rely on a large amount of labeled data for training, which is often difficult to achieve in practical applications, especially when new or rare fault types appear. Existing fault diagnosis methods rely on a large number of fault samples and are difficult to adapt to the fault detection needs in zero-sample scenarios in actual operation. In addition, when facing the complex and diverse working conditions of electric traction systems, existing fault diagnosis methods are difficult to balance real-time, accuracy and versatility. Therefore, a new method is urgently needed to make up for the shortcomings of existing technologies.

[0045] Based on this, the present invention provides a zero-sample fault online diagnosis method based on a multi-resolution diffusion model. First, the data collected by the traction transmission control simulation platform are used to construct a source domain dataset U S , the data collected during the actual operation of the train is used to construct the target domain dataset U TThe fault category sets of the source domain and the target domain do not overlap. The data is processed by a multi-resolution diffusion model to achieve outlier removal, missing data filling and standardization. Next, the category-attribute description matrix A' is constructed, and the fault features f of the source domain data are extracted by deep incremental supervised principal component analysis (DI-SPCA). i s , Feature B s With attribute description a i The mapping relationship between them. Get the target domain data D online t After that, update the attribute description of the fault category and construct a new sparse matrix A * , and fine-tune the model parameters. Finally, Bayesian reasoning is used to perform real-time fault diagnosis on zero-sample data in the target domain, and the optimal matching fault category is inferred based on the maximum a posteriori probability, so as to achieve accurate diagnosis and prediction of target domain data, thereby guiding system maintenance and fault handling. Based on this, the embodiment of the present invention uses the attributes of known fault categories to diagnose new fault types and compound faults. By using fault categories and attributes as learning objects, it is possible to achieve attribute migration from source domain fault categories to target domain fault categories in the absence of target domain fault category data samples, thereby achieving zero-sample fault diagnosis.

[0046] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0047] like Figure 1 As shown, Figure 1 It is a flowchart of a zero-sample fault online diagnosis method based on a multi-resolution diffusion model provided by an embodiment of the present invention. The zero-sample fault online diagnosis method based on a multi-resolution diffusion model may include but is not limited to the following steps.

[0048] S1), define the data collected by the traction transmission control simulation platform as source domain data, and define the data collected during the actual operation of the train as target domain data; construct a source domain dataset U S and the target domain dataset U T : Source domain dataset where Y s represents the source domain data sample space, E s represents the fault category space of source domain data samples, N s Represents the number of source domain data samples; target domain data where Y t represents the target domain data sample space, E t represents the fault category space of the target domain data samples, N t Represents the number of target domain data samples;

[0049] Construct the fault category set F of the source domain and the target domain respectively s and Ft , the fault categories of the source domain include common single faults and compound faults, and the fault categories of the target domain include compound faults that occur under different working conditions; where h is the number of fault categories in the source domain, q is the number of fault categories in the target domain, and the total number of fault categories is r = h + k. The set of fault categories of the source domain and the target domain is F s and F t do not intersect each other, that is, F s ∩F t =φ;

[0050] S2) In establishing a multi-resolution diffusion model, the source domain U obtained by data collection is S and the target domain U t Input the data and get the reconstructed data and Y t * , realizing functions such as removal of outliers, filling of missing fault data and standardization processing.

[0051] S3), construct the fault category set F of the source domain and the target domain respectively s and F t , summarize the source domain fault samples Y s Property description of a i (i) Establish a category-attribute description matrix A' and transform the matrix A' into a sparse category-attribute description matrix A through a one-hot encoder;

[0052] S4) The fault category space E s Associate the attribute label with the category attribute description matrix A to obtain the attribute label L s , and use deep incremental supervised principal component analysis (DI-SPCA) to extract the fault features f of the source domain data samples i s , construct the feature matrix B of the source domain samples s ; By training the learner model α i (i=1,2,3...,C) performs attribute learning to obtain the feature B of the source domain sample s With attribute description a i The mapping relationship between them;

[0053] S5) Operation data D obtained online t (target domain data), update the attribute description of the fault category, and construct a new sparse category-attribute description matrix A * ; and based on the updated matrix, the attribute learner α i Fine-tune parameters;

[0054] S6), for the input target domain sample According to the output a of the attribute learner t , using Bayesian Inference to infer samples from attribute values The fault category of the fault category is, and the posterior probability of the fault category is, and the category with the largest posterior probability is selected as the optimal matching fault type to achieve real-time inference diagnosis of the fault type of zero-sample data in the target domain.

[0055] Preferably, in step S2), the operating parameters of the source domain and target domain data sets include: inverter three-phase current {I sa , I sb , I sc}, train speed SP, motor speed Rs, motor electromagnetic torque Te, intermediate DC circuit upper and lower support capacitor voltage {U cd1 U cd2}、Transformer secondary side AC voltage U n And current data U n .

[0056] Preferably, in step S2), the multi-resolution diffusion model performs multi-stage processing on the source domain and target domain data in stages with gradually increasing temporal resolution, and each stage gradually restores the fault characteristics by performing phased analysis and reconstruction on the data at different resolutions.

[0057] Preferably, in step S2), wherein:

[0058] S21), divide the preprocessed data into training set S train and the test set S test ;

[0059] The initial data is split into time series by the historical window (LW) and the current window (CW), where the LW covers data from 1 second to 5 seconds and the CW covers data from 1.5 seconds to 5.5 seconds;

[0060] like Figure 2 , Figure 3 and Figure 4 As shown in the figure, the segmented data is decomposed in stages from fine to coarse according to three different time resolutions of 10e-6, 5e-3, and 20e-3 to generate multi-resolution data X c1 , X c2 , X c3 , X s1 , X s2 , X s3 .

[0061] S22), forward diffusion process, adding noise to the source data to generate noisy data

[0062] In the forward diffusion network, the diffusion coefficient and Gaussian noise are used to add noise:

[0063]

[0064] where Z is Gaussian noise sampled from a standard normal distribution N(0,1), β S is the variance coefficient of Z, γ S =1-β S is the forward diffusion coefficient, is the cumulative diffusion coefficient;

[0065] The purpose is to allow the model to learn the data distribution under different noise levels through noise perturbation.

[0066] S23), the inverse reconstruction process adopts the Markov process, and the state of each time step t only depends on the reconstructed state of the previous step, making the reconstruction process simple, stable and easy to calculate;

[0067] Retrograde reconstruction includes the conditional network and the reconstructed network;

[0068] In the S phase, the conditional network uses a signal X with the same temporal resolution as the signal to be reconstructed. ls Generate reconstruction conditions, the reconstruction conditions are: c s =concat(z lookback ,X c(s+1) ),s=K,…,1;

[0069] Among them, z lookback Yes X ls The tensor obtained by linear mapping;

[0070] The reconstruction network uses segmentation error (SmoothL1) as the loss function, and the loss function is:

[0071]

[0072] Among them, θ S are all the parameters of the conditional network and the reconstructed network in the S stage, In the reconstruction condition c S The output of the denoising network is as follows: Figure 5 As shown:

[0073]

[0074] This loss measures the difference between the noise generated by the network and the real noise, and is optimized by iterative training adjustment. S , minimize L SmoothL1 (θ S) to improve the performance of the reconstruction network, continuously improve data reconstruction, ensure reconstruction quality, and obtain reconstructed fault data.

[0075] Preferably, in step S3), in order to identify and diagnose the target fault of zero sample, the source domain fault samples Y are summarized s Property description of a i (i).

[0076] It consists of C attributes, and the description of h types of fault data and categories can be expressed as the category-attribute description matrix A'.

[0077] The one-hot encoder is used to transform the matrix of source domain fault description A' into a sparse category-attribute description matrix A:

[0078] A∈R h×c =one-hot(A')

[0079] All elements in matrix A are either 1 or 0, indicating the presence or absence of an attribute.

[0080] Preferably, in step S3) and step S4), DI-SPCA extracts main features from high-dimensional data by reducing dimensionality layer by layer, maps the original data to a low-dimensional space, and reduces computational complexity.

[0081] The feature extraction mapping function ζ of DI-SPCA is used to transform the source domain data sample Y s and the corresponding attribute tags as input.

[0082] Extract fault feature vector f i s It is expressed as:

[0083]

[0084] Preferably, in step S5), during the online update process, the updated fault description matrix is ​​represented by a one-hot encoder as follows:

[0085] A *′ ∈R r*c

[0086] Transformed into a new sparse category-attribute description matrix A * , to reflect new fault types and changes in operating conditions.

[0087] Preferably, in step S3) and step S4), the posterior probability of the fault category is expressed as:

[0088]

[0089] in,

[0090] There is a category-attribute description matrix A * Provides the association information between the target fault category and attribute, with p(a t |t)=1;

[0091] Comprehensive sample Fault characteristics and predicted attribute label a t , the posterior probability is expressed as:

[0092]

[0093] in, is the feature f i t Under the condition, the property The probability of belonging to fault category t;

[0094] Is an attribute The prior probability among all fault categories; p(t) is the prior probability of each category;

[0095] Preferably, in step S5), in the calculation of the posterior probability of the fault category, it is assumed that the prior probabilities of all categories are equal, that is:

[0096] p(t) = constant

[0097] The next step of maximum a posteriori calculation has no effect on the results;

[0098] because Select the class with the largest posterior probability as the sample The best matching fault type:

[0099]

[0100] in, Represents the attribute description vector of the e-th fault type of the target domain sample, yes The ith element of .

[0101] Thus, the category-attribute description matrix A * It plays the role of knowledge transfer. Through the attribute information of known fault categories, the model can apply this knowledge to fault samples of unknown categories.

[0102] Realize real-time reasoning and diagnosis of fault types of zero-sample data in the target domain;

[0103] This process not only relies on the current input data, but also combines the category attribute information in the source domain historical data to ensure the accuracy of the diagnosis.

[0104] In addition, an embodiment of the present invention also discloses an electronic device, comprising: at least one processor; at least one memory for storing at least one program; when the at least one program is executed by the at least one processor, a zero-sample fault online diagnosis method based on a multi-resolution diffusion model as in any of the previous embodiments is implemented.

[0105] In addition, an embodiment of the present invention further discloses a computer-readable storage medium storing computer-executable instructions for executing the zero-sample fault online diagnosis method based on a multi-resolution diffusion model as in any of the previous embodiments.

[0106] The system architecture and application scenarios described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. Those skilled in the art will appreciate that with the evolution of the system architecture and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are equally applicable to similar technical problems.

[0107] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0108] In hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0109] The terms "component", "module", "system", etc. used in this specification are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, or a computer. By way of illustration, both applications running on a computing device and a computing device can be components. One or more components may reside in a process or an execution thread, and a component may be located on a computer or distributed between two or more computers. In addition, these components may be executed from various computer-readable media having various data structures stored thereon. Components may communicate, for example, through a local or remote process based on a signal having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, or a network, such as the Internet interacting with other systems through signals).

Claims

1. A zero-sample fault online diagnosis method based on a multi-resolution diffusion model, characterized in that: The following steps are involved: S1), define the data collected by the traction transmission control simulation platform as source domain data, and define the data collected during the actual operation of the train as target domain data; construct a source domain dataset U S and the target domain dataset U T : Source domain dataset Among them, Y s represents the source domain data sample space, E s represents the fault category space of source domain data samples, N s Represents the number of source domain data samples; target domain data set Among them, Y t represents the target domain data sample space, E t represents the fault category space of the target domain data samples, N t Represents the number of target domain data samples; Construct the fault category set F of the source domain and the target domain respectively s and F t , the fault categories of the source domain include single faults and compound faults, and the fault categories of the target domain include compound faults that occur under different working conditions, where h is the number of fault categories in the source domain, q is the number of fault categories in the target domain, and the total number of fault categories is r = h + k. The set of fault categories of the source domain and the target domain is F s and F t do not intersect each other, that is, F s ∩F t =φ; S2) In establishing a multi-resolution diffusion model, the source domain U obtained by data collection is S and the target domain U t Input the data and get the reconstructed data and S3), construct the fault category set F of the source domain and the target domain respectively s and F t , summarize the source domain fault samples Y s Property description of a i (i) Establish a category-attribute description matrix A' and transform the matrix A' into a sparse category-attribute description matrix A through a one-hot encoder; S4) The fault category space E s Associate the attribute label with the category attribute description matrix A to obtain the attribute label L s , and use deep incremental supervised principal component analysis DI-SPCA to extract the fault features of source domain data samples f i s , construct the feature matrix B of the source domain samples s ; By training the learner model α i (i=1,2,3...,C) performs attribute learning to obtain the feature B of the source domain sample s With attribute description a i The mapping relationship between them; S5) Operation data D obtained online t , update the attribute description of the fault category and construct a new sparse category-attribute description matrix A * ; and based on the updated matrix, the attribute learner α i Fine-tune parameters; S6), for the input target domain sample According to the output a of the attribute learner t , using Bayesian inference to infer samples from attribute values The fault categories are calculated, the posterior probability of each fault category is calculated, and the category with the largest posterior probability is selected as the optimal matching fault type.

2. The method according to claim 1, characterized in that: The operating parameters of the source domain data set and the target domain data set in step S1) include: inverter three-phase current {I sa , I sb , I sc }, train speed SP, motor speed Rs, motor electromagnetic torque Te, intermediate DC circuit upper and lower support capacitor voltage {U cd1 U cd2 }、Transformer secondary side AC voltage U n And current data U n .

3. The method according to claim 1, characterized in that: The multi-resolution diffusion model described in step S2) performs multi-stage processing on the source domain data and the target domain data in stages with gradually increasing temporal resolution. In each stage, the fault characteristics are gradually restored by performing phased analysis and reconstruction on the data at different resolutions. S21), divide the preprocessed data into training set S train and the test set S test ; The initial data is segmented into time series through the historical window LW and the current window CW, where the historical window LW covers data from 1 second to 5 seconds, and the current window CW covers data from 1.5 seconds to 5.5 seconds; the segmented data is decomposed from fine to coarse in four different time resolutions of 10μs, 100μs, 1ms, and 10ms to generate multi-resolution data; S22), forward diffusion process, adding noise to the source data to generate noisy data Noise using diffusion coefficient and Gaussian noise: where Z is Gaussian noise sampled from a standard normal distribution N(0,1), β S is the variance coefficient of Z, γ S =1-β S is the forward diffusion coefficient, is the cumulative diffusion coefficient; S23) The reverse reconstruction process uses a Markov process. The state of each time step t depends only on the reconstructed state of the previous step. The reverse reconstruction includes a conditional network and a reconstruction network. In the S stage, the signal to be reconstructed is used. Signal X with the same time resolution ls As a reconstruction condition, obtaining reconstructed fault data; S24), the model uses the segmentation error SmoothL1 as the loss function, and the loss function is: Among them, θ S are all the parameters of the conditional network and the reconstructed network in the S stage, yes In the reconstruction condition c S The output of the denoising network is passed through the denoising network; this loss measures the difference between the noise generated by the network and the real noise, and is optimized by iterative training adjustment θ S , minimize L SmoothL1 (θ S ) to improve the performance of the reconstructed network.

4. The method according to claim 1, characterized in that: In step S3), the source domain fault samples Y are summarized s Property description of a i (i) It consists of C attributes. The description of h types of fault data and categories is represented as a category-attribute description matrix A'. The one-hot encoder is used to transform the matrix of source domain fault description A' into a sparse category-attribute description matrix A: A∈R h×c = one-hot(A'), all elements in the matrix A are 1 or 0, indicating whether the attribute exists or not.

5. The method according to claim 1, characterized in that: In step S3) and step S4), DI-SPCA extracts the main features from the high-dimensional data by layer-by-layer dimensionality reduction and maps the original data to a low-dimensional space. The feature extraction mapping function ζ of DI-SPCA is used to transform the source domain data sample Y s and the corresponding attribute tags As input, extract the fault feature vector f i s It is expressed as:

6. The method according to claim 1, characterized in that: In step S5), during the online update process, the updated fault description matrix A is converted into *′ ∈R r*c ; Transformed into a new sparse category-attribute description matrix A * , to reflect new fault types and changes in operating conditions.

7. The method according to claim 1, characterized in that: In step S5) and step S6), the posterior probability of the fault category is expressed as: in, Since there is a category-attribute description matrix A * Provides the association information between the target fault category and attribute, with p(a t |t)=1, comprehensive sample Fault characteristics and predicted attribute label a t , the posterior probability is expressed as: in, is the feature f i t Under the condition, the property The probability of belonging to fault category t; Is an attribute The prior probability among all fault categories; p(t) is the prior probability of each category.

8. The method according to claim 1, characterized in that: In step S6), in the calculation of the posterior probability of the fault category, it is assumed that the prior probabilities of all categories are equal, that is, p(t) = constant, and the maximum a posteriori calculation in the next step has no effect on the result; Select the class with the largest posterior probability as the sample The best matching fault type: in, Represents the attribute description vector of the e-th fault type of the target domain sample, yes The i-th element of ; thus, the category-attribute description matrix A * It plays the role of knowledge transfer. Through the attribute information of known fault categories, the model applies this knowledge to fault samples of unknown categories to achieve real-time reasoning and diagnosis of fault types of zero-sample data in the target domain. After the reasoning and diagnosis are completed, the fault category prediction results generated by the model will be used to guide system maintenance and fault handling.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the zero-sample fault online diagnosis method based on a multi-resolution diffusion model as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the zero-sample fault online diagnosis method based on a multi-resolution diffusion model according to any one of claims 1 to 8.

Images