CN115758208A

CN115758208A - Traction converter fault diagnosis method and device, computer equipment and storage medium

Info

Publication number: CN115758208A
Application number: CN202211395520.8A
Authority: CN
Inventors: 孟苓辉; 周振威; 卢冠兰; 刘俊斌; 何世烈; 俞鹏飞; 余陆斌
Original assignee: China Electronic Product Reliability and Environmental Testing Research Institute
Current assignee: China Electronic Product Reliability and Environmental Testing Research Institute
Priority date: 2022-11-09
Filing date: 2022-11-09
Publication date: 2023-03-07

Abstract

The application relates to a traction converter fault diagnosis method, a traction converter fault diagnosis device, computer equipment, a storage medium and a computer program product. The method comprises the steps of obtaining working condition data of a traction converter; extracting time-frequency characteristics of the working condition data to obtain target characteristic data; inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; and performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result. In the whole scheme, the network model is trained more accurately through the multi-working-condition digital twin state sample data, fault prediction is carried out based on the two network models, and weighted fusion is carried out based on the multi-dimensional fault prediction result, so that a more accurate fault diagnosis result is obtained.

Description

Traction converter fault diagnosis method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of railway technology, and in particular, to a method, an apparatus, a computer device, a storage medium, and a computer program product for diagnosing a fault of a traction converter.

Background

With the construction of infrastructure and the rapid development of science and technology, the fields of electric power, intelligent automobiles, high-speed rails, urban rails, subways and the like bring rapid growth for the development of economy, however, the safety and the reliability of equipment should be guaranteed while high speed is pursued, and the safety problem is always the most important research subject. Therefore, PHM (fault prediction and Health Management) technology for different fields is important for the Health development of each field. The traction converter is a power conversion and traction device and plays an important role in the fields of electric power, intelligent automobiles, high-speed rails, urban rails, subways and the like.

Taking the application of the traction converter in a high-speed rail as an example, the high-speed railway electric traction system is under the action of frequent internal and external impacts, such as external factors of power supply network oscillation, distortion, harmonic waves, artificial misoperation and the like, and internal factors of system aging, accumulated damage and the like, and system faults are caused even the system is damaged when the system is serious. The traction converter is a key subsystem of the electric locomotive and guarantees safe and stable running of the train. The traction converter is used as a key subsystem of a high-speed rail, an urban rail, a subway and a hybrid power train, and once the system fails, huge loss and even disastrous results are often caused.

At present, a fault diagnosis method for a traction converter cannot realize accurate fault diagnosis of the traction converter.

Disclosure of Invention

In view of the above, there is a need to provide an accurate fault diagnosis method, apparatus, computer device, computer readable storage medium and computer program product for a traction converter.

In a first aspect, the application provides a traction converter fault diagnosis method. The method comprises the following steps:

acquiring working condition data of a traction converter;

extracting time-frequency characteristics of the working condition data to obtain target characteristic data;

inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long and short term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions;

and performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result.

In one embodiment, the deep residual shrinkage network model is obtained by training in the following way: acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting digital twin state sample data into an initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result; performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value; and updating parameters of the initial depth residual error shrinkage network model based on the first loss value until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual error shrinkage network model.

In one embodiment, the deep long-short term memory network model is trained in the following way: acquiring digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting the digital twin state sample data into an initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result; performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value; and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is smaller than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

In one embodiment, the digital twin state sample data comprises sample working condition data and a fault type corresponding to the working condition data; the method for acquiring digital twin state sample data of the traction converter under different working conditions comprises the following steps: and acquiring working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the obtaining of the working condition data of the traction converter under different working conditions and the fault type corresponding to each working condition data based on the digital twin model comprises: acquiring state data and an initial digital twin model of the traction converter under different working conditions; inputting the state data into an initial digital twin model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data to obtain an entity fault result; error calculation is carried out according to the entity fault result and the fault simulation result to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, so as to obtain a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the digital twin model comprises a virtual simulation model and a semi-physical simulation model; inputting the state data into an initial digital twin model for fault simulation, and obtaining a fault simulation result, wherein the fault simulation result comprises the following steps: inputting the state data into a virtual simulation model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data, and obtaining an entity fault result comprises the following steps: and inputting the state data into a semi-physical simulation model for fault simulation to obtain an entity fault result.

In one embodiment, performing weighted fusion according to the first failure prediction result and the second failure prediction result, and obtaining the failure diagnosis result includes: acquiring the prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and acquiring the prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight; and performing weighted calculation according to the first fault weight and the first fault prediction result, and the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

In a second aspect, the application further provides a traction converter fault diagnosis device. The device includes:

the acquisition module is used for acquiring working condition data of the traction converter;

the extraction module is used for extracting the time-frequency characteristics of the working condition data to obtain target characteristic data;

the prediction module is used for inputting the target characteristic data into the deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into the deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long-term and short-term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions;

and the fusion module is used for performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result.

In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the following steps when executing the computer program:

acquiring working condition data of a traction converter;

In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

acquiring working condition data of a traction converter;

In a fifth aspect, the present application further provides a computer program product. The computer program product comprising a computer program which when executed by a processor performs the steps of:

acquiring working condition data of a traction converter;

According to the method, the device, the computer equipment, the storage medium and the computer program product for diagnosing the fault of the traction converter, the working condition data of the traction converter are obtained; extracting time-frequency characteristics of the working condition data to obtain target characteristic data; inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long and short term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions; and performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result. According to the whole scheme, a deep residual shrinkage network model and a deep long-term and short-term memory network model are obtained on the basis of digital twin state sample data training, a fault diagnosis model obtained through multi-working-condition digital twin state sample data training is more accurate, fault prediction is conducted on state data of the traction converter on the basis of the two fault diagnosis models, and a multi-dimensional fault prediction result is subjected to weighted fusion to obtain a more accurate fault diagnosis result.

Drawings

FIG. 1 is an environmental diagram illustrating an exemplary method for fault diagnosis of a traction converter;

FIG. 2 is a schematic flow diagram of a traction converter fault diagnosis method in one embodiment;

FIG. 3 is a flowchart illustrating a deep residual shrinkage network model training process in one embodiment;

FIG. 4 is a diagram illustrating a structure of a depth residual shrinking network model in one embodiment;

FIG. 5 is a schematic flow chart of a fault diagnosis method for a traction converter in another embodiment;

FIG. 6 is a block diagram of a traction converter fault diagnosis apparatus according to an embodiment;

FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.

The PHM (fault prediction and Health Management) is proposed to meet the requirements of autonomous security and autonomous diagnosis, and is an upgrade development of a condition-based maintenance CBM (condition based maintenance). The method emphasizes state perception in asset equipment management, monitors equipment health condition and frequent fault areas and periods, and predicts the occurrence of faults through data monitoring and analysis, thereby greatly improving operation and maintenance efficiency.

At present, the research of the domestic PHM technology is in the primary exploration stage, and particularly, the research of the phoenix feather unicorn in the railway field is carried out by the PHM technology. With the construction and rapid development of railway engineering in China in recent years, high-speed rails, urban rails, subways and the like have already played a great part in national economy and development of China, the safety and reliability of trains should be guaranteed while high speed is pursued, and safety problems are the most important research subjects forever, so that the PHM technology is urgently required to be technically explored and researched in the field of rail transit.

The high-speed railway electric traction system is under the action of frequent internal and external impacts, such as external factors of power supply network oscillation, distortion, harmonic waves, artificial misoperation and the like and internal factors of system aging, accumulated damage and the like, and system faults and even system damage are caused in serious cases. If the factors influencing the system damage can be analyzed, a system health assessment or fault prediction model is established, the occurrence of potential faults of the system is avoided through model analysis and reasoning by combining with the real-time information of the key characteristic states of the train and then through external safety control measures, and the safe and stable operation of the system is guaranteed to the greatest extent. The traction converter is a key subsystem of the electric locomotive and guarantees safe and stable running of the train. The traction converter is used as a key subsystem of a high-speed rail, an urban rail, a subway and a hybrid power train, and generally comprises a plurality of parts, has multiple structural levels, and is complex in relation between different parts and high in coupling. The system has the characteristics of complex structure, difficult fault diagnosis and equipment maintenance, nonlinearity, coupling, randomness and the like, so that the system is easy to break down, and high reliability and safety are required. Such systems often suffer significant damage and even catastrophic results once they fail.

Under the condition that the converter works in a complex and severe environment for a long time, the performance of a key power device is degraded, even a fault is generated, the fault or the performance degradation can generate certain influence on the converter, and a learner models the converter based on certain reliability theories such as a Markov model, bayesian and the like, but the state transition and inference probability of a probabilistic statistics and inference model are difficult to accurately determine, so that the evaluation on the reliability or the service life of the system cannot be accurately realized in real time. The traditional fault diagnosis or prediction method can not fuse various state information of the system, the accuracy of single information or means is poor, with the development of digital twin and artificial intelligence technologies, the PHM based on the digital twin technology can realize the fusion of multi-source state information and the fusion of virtual and real models by combining artificial intelligence, and the like, and can correct twin bodies and feed back the twin bodies in real time and accurately to continuously approach the real physical state of the system, so that the application proposes a thinking of realizing the accurate fault prediction and health management of the traction converter based on the fusion of the digital twin model and a deep learning algorithm.

DT (Digital Twin), digital Twin, is a simulation process integrating multidisciplinary, multi-physical quantity, multi-scale, multi-probability by fully utilizing data such as physical models, sensor updates, operation histories and the like, and completing mapping in a virtual space so as to reflect the whole life cycle process of corresponding entity equipment. Digital twinning is an beyond-realistic concept that can be viewed as a digital mapping system of one or more important, interdependent equipment systems.

The fault diagnosis method for the traction converter provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104, or may be located on the cloud or other network server. The terminal 102 sends a fault detection instruction to the server, and the server 104 acquires working condition data of the traction converter according to the fault detection instruction; extracting time-frequency characteristics of the working condition data to obtain target characteristic data; inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long and short term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions; and performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart car-mounted devices, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The server 104 may be implemented as a stand-alone server or as a server cluster comprised of multiple servers.

In one embodiment, as shown in fig. 2, a traction converter fault diagnosis method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:

step 202, obtaining working condition data of the traction converter.

The traction converter is one of key parts of a train, is arranged at the bottom of a motor train of the train, and has the main functions of converting electric energy between a direct current system and an alternating current system, converting 1500V direct current from a contact network into 0-1150V three-phase alternating current, and realizing starting, braking and speed regulation control of an alternating current traction motor through voltage regulation and frequency regulation control. The working condition data of the traction converter comprises three-phase side electric data and direct current side electric data, further the three-phase side electric data comprises three-phase voltage data, three-phase current data, motor data and the like, the direct current side electric data comprises direct current bus voltage and direct current bus current, and the motor data comprises motor rotating speed, motor torque and the like. The working conditions of the traction converter include different working conditions such as no-load, light load and heavy load, and each working condition also includes different fault information, such as degradation fault of a key part capacitor, aging or open/short circuit of a power device, and the like, and the faults can cause the voltage or current output by the electric traction system of the high-speed railway to change in different degrees.

Specifically, the server receives a fault detection instruction sent by the terminal, and acquires working condition data of the traction converter acquired by the terminal according to the fault detection instruction, wherein the working condition data is latest current working condition data acquired in real time. The server can also obtain the latest current working condition data from the terminal at regular time so as to detect the fault in real time. The terminal can acquire real-time working condition data of the traction converter through the data acquisition sensor and transmit the real-time working condition data to the server.

And step 204, extracting the time-frequency characteristics of the working condition data to obtain target characteristic data.

The time-frequency characteristics may be waveform characteristics of the working condition data, and the time-frequency characteristics may describe frequency information of the working condition data within a period of time. Given a discrete time signal x (n), a two-dimensional representation x (n, f) can be obtained by a time-frequency feature extraction method, which is a complex function of time and frequency. Thus, it is possible to provide how the amplitude and phase of the different frequency components of the condition data vary with time. The time-frequency representation of the working condition data, no matter amplitude or phase, can be a group of characteristics, and 2D images are directly presented through novel classifiers based on deep learning, such as a convolutional neural network.

One limitation of time and frequency domain feature extraction methods in general is that important features with high resolution are discarded because the information is computed from only one domain. For example, the time domain characteristics do not provide oscillation information, while in the case of frequency analysis, detailed information of the time variation of the spectral signal is not provided, but these are important concerns for signal analysis research. This limitation can be addressed by time-frequency analysis, such as short-time Fourier transform (STFT), morlet Wavelet transform (MW), and filter-based hilbert transform (FHT).

Specifically, after the server acquires the current working condition data of the traction converter, noise reduction is performed on the working condition data, noise interference is reduced, the working condition data after noise reduction is obtained, time-frequency feature analysis and extraction are performed on the working condition data after noise reduction, and target feature data are obtained. Further, the server performs time-frequency feature analysis and extraction on the working condition data subjected to noise reduction based on short-time Fourier transform to obtain target feature data; or the server performs time-frequency feature analysis and extraction on the denoised working condition data based on Morlet wavelet transformation to obtain target feature data; or the server performs time-frequency feature analysis and extraction on the working condition data subjected to noise reduction based on the Hilbert transform of the filter to obtain target feature data.

And step 206, inputting the target characteristic data into the deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into the deep long-short term memory network model for fault prediction to obtain a second fault prediction result.

The deep residual error shrinkage network model and the deep long-term and short-term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions. The DRSN (Deep Residual shronkage Network) is a novel improvement of a Deep Residual error Network, soft thresholding is used as a nonlinear layer and introduced into a Network structure of the ResNet, and the characteristic learning effect of a Deep learning method on noise-containing data or complex data can be improved. The LSTM (Long Short-Term Memory), a deep Long Short-Term Memory Neural Network, is a time-cycle Neural Network, and is designed specifically to solve the Long-Term dependence problem of a general RNN (Recurrent Neural Network), and all RNNs have a chain form of a repetitive Neural Network module.

Specifically, the server inputs the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputs the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result. And performing parallel processing on the target characteristic data through a deep residual shrinkage network model and a deep long-short term memory network model to obtain results predicted by the two models, namely a first fault prediction result and a second fault prediction result.

And step 208, performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result.

Specifically, the server obtains the weight coefficients of the deep residual shrinkage network model and the deep long and short term memory network model, and performs weighted calculation on the first fault prediction result and the second fault prediction result based on a weighted average algorithm and the weight coefficients of the deep residual shrinkage network model and the deep long and short term memory network model to obtain a fault diagnosis result.

In the fault diagnosis method for the traction converter, working condition data of the traction converter are obtained; extracting time-frequency characteristics of the working condition data to obtain target characteristic data; inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long and short term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions; and performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result. According to the whole scheme, a deep residual shrinkage network model and a deep long-term and short-term memory network model are obtained on the basis of digital twin state sample data training, a fault diagnosis model obtained through multi-working-condition digital twin state sample data training is more accurate, fault prediction is conducted on state data of the traction converter on the basis of the two fault diagnosis models, and a multi-dimensional fault prediction result is subjected to weighted fusion to obtain a more accurate fault diagnosis result.

In an alternative embodiment, as shown in fig. 3, the deep residual shrinkage network model is trained in the following manner:

and 302, acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions.

And the digital twin state sample data carries a fault type mark. Digital twin state sample data is generated by a digital twin model. The digital twin state sample data comprises multiple fault data and fault types of the traction converter under multiple working conditions, and the fault types are accurate fault types obtained by experts based on different fault data analysis.

Specifically, the server obtains digital twin state sample data of the traction converter under different working conditions and an untrained initial depth residual shrinkage network model based on the digital twin model.

And 304, inputting the digital twin state sample data into the initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result.

On the basis of the depth residual error network, a small sub-network is introduced into the depth residual error shrinkage network, a group of threshold values are obtained by learning of the sub-network, and soft thresholding is carried out on each channel of the feature map. This process may be viewed as a trainable feature selection process. Specifically, the convolutional layer converts important features into values with larger absolute values, and converts features corresponding to redundant information into values with smaller absolute values; the boundary between the two is learned by the sub-network and the redundant features are set to zero by soft thresholding while the important features have non-zero output.

As shown in fig. 4, the deep residual shrinkage network model includes a convolutional layer, two residual modules, batch normalization, global mean pooling, and a fully-connected output layer. Further, the residual module includes two Batch Normalization (BN), two rectifying Linear Unit activation functions (ReLU), two convolution layers (Convolutional layer), a residual contraction module, and an Identity short table.

Specifically, the server inputs the digital twin state sample data into the initial depth residual shrinkage network model, and performs fault diagnosis on the digital twin state sample data through the initial depth residual shrinkage network to obtain a first sample fault diagnosis result.

And step 306, performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value.

Specifically, the server obtains the fault type of the digital twin state sample data, and loss calculation is carried out according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value.

And 308, updating parameters of the initial depth residual error shrinkage network model based on the first loss value until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual error shrinkage network model.

Specifically, the server performs back propagation based on the first loss value, and continuously updates parameters of the initial depth residual shrinkage network model until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual shrinkage network model.

In the embodiment, the characteristics related to noise are eliminated based on the soft thresholding of the deep residual shrinkage network, and more importantly, the threshold is obtained by automatic learning in the residual shrinkage module, so that robust analysis, identification and extraction of weak characteristics of the traction converter fault under strong noise interference can be realized. The deep residual shrinkage network can be used for data containing noise and data without noise. This is because the threshold in the deep residual shrinkage network is adaptively determined according to the sample situation. In other words, if the samples do not contain redundant information and do not require soft thresholding, the threshold can be trained very close to zero, so that the soft thresholding is equivalent to not being present. Meanwhile, the soft thresholding sets the features in the range of [ - τ, τ ] to 0, and allows other features farther from 0 to shrink toward 0. This interval of zero becomes [ - τ + b, τ + b ] if viewed together with the offset b of the previous convolutional layer. Because τ and b are parameters that can be obtained through automatic learning, from this viewpoint, soft thresholding is actually to set the feature of any interval to zero, which is a more flexible way to delete a feature of a certain value range, and can also be understood as a more flexible nonlinear mapping.

In an alternative embodiment, the deep long short term memory network model is trained in the following way: acquiring digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; inputting the digital twin state sample data into an initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result; performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value; and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is smaller than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

And carrying a fault type mark by the digital twin state sample data. RNN models suffer from long-term dependence and are not able to efficiently learn features in longer time series. Long-term memory networks (LSTMs) can effectively alleviate the long-term dependence problem. LSTM introduces an input gate i, a forget gate f, an output gate o, and an internal memory unit c, compared to RNN.

Specifically, the server obtains digital twin state sample data of the traction converter under different working conditions and an untrained initial depth long-term and short-term memory network model based on the digital twin model. And the server inputs the digital twin state sample data into the initial depth long-short term memory network model, and fault diagnosis is carried out on the digital twin state sample data through the initial depth long-short term memory network model to obtain a second sample fault diagnosis result. And the server acquires the fault type of the digital twin state sample data, and performs loss calculation according to the fault diagnosis result of the second sample and the fault type mark to obtain a second loss value. And the server performs back propagation based on the second loss value, continuously updates the parameters of the initial deep long-short term memory network model until the latest second loss value is smaller than a preset second loss threshold value, and obtains the deep long-short term memory network model.

In the embodiment, the memory storage and the deep characteristic analysis and extraction of the historical fault state information of the converter are realized by the advantages of the deep long-term and short-term memory network, the characteristic information of the converter fault under the multi-mode condition is extracted and enriched by a dynamic difference method, and the problem that the static characteristic is insufficient to represent the fault evolution trend characteristic of the multi-mode process can be solved.

In an optional embodiment, the step of obtaining digital twin state sample data of the traction converter under different working conditions comprises the following steps: and acquiring working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

The digital twin state sample data comprises sample working condition data and fault types corresponding to the working condition data.

Specifically, the traction converter is modeled based on virtual simulation based on the angle of an electrical model and a physical model to obtain virtual data, meanwhile actual test data are obtained based on a semi-physical simulation test platform and are fused with each other to form a digital twin model, and working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data are generated based on the digital twin model.

In the implementation, aiming at the defects of insufficient fault data information, single diagnosis type, poor visualization degree and the like of the existing traction converter, the digital twin state sample data generated based on the digital twin model comprises data of different fault types under various working conditions, and the deep residual shrinkage network model and the deep long-term and short-term memory network model are trained based on the digital twin state sample data, so that the fault diagnosis performance of the model is improved, and the online intelligent state monitoring and fault diagnosis of the traction converter are realized.

In an optional embodiment, the obtaining of the working condition data of the traction converter under different working conditions and the fault type corresponding to each working condition data based on the digital twin model comprises: acquiring state data and an initial digital twin model of the traction converter under different working conditions; inputting the state data into an initial digital twin model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data to obtain an entity fault result; error calculation is carried out according to the entity fault result and the fault simulation result to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, and obtaining a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

The state data of the traction converter under different working conditions can be obtained by analyzing historical state data of the traction converter. The digital twin model is a semi-virtual simulation model of the traction converter, and the structure of the traction converter in the digital twin model is consistent with that of a real traction converter.

Specifically, the server obtains state data and an initial digital twin model of the traction converter under different working conditions. The server inputs the state data into the initial digital twin model for fault simulation to obtain simulation waveform data in the state; carrying out entity simulation according to the state data to obtain entity waveform data; error calculation is carried out according to the entity waveform data and the simulation waveform data to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, and obtaining a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In the embodiment, in order to obtain fault characteristics of the high-speed train under various different faults under the traction working condition, waveforms of current, voltage, motor rotating speed, torque and the like under different faults are contrastively analyzed, and then fault identification and diagnosis of different faults are completed.

In an alternative embodiment, inputting the state data into the initial digital twin model for fault simulation, and obtaining the fault simulation result includes: inputting the state data into a virtual simulation model for fault simulation to obtain a fault simulation result; performing entity simulation according to the state data, and obtaining an entity fault result comprises: and inputting the state data into a semi-physical simulation model for fault simulation to obtain an entity fault result.

The digital twin model comprises a virtual simulation model and a semi-physical simulation model. The controller and the controlled object of the virtual simulation model are both virtual, and the virtual simulation model is based on a pure digital virtual MATLAB electrical simulation model built by a SIMULINK module of the MATLAB, so that the simulation of various working conditions and fault modes of the traction converter can be realized. Further, the virtual simulation model simulates data such as output three-phase voltage current, output three-phase direct-current side voltage current and the like under different conditions such as open circuit, short circuit, degradation and the like of key parts such as a power switch device, a capacitor and the like of the traction converter.

The semi-physical simulation model is an improvement of a virtual simulation model, the controller still adopts the virtual simulation model, the main circuit of the controlled object adopts a physical test bed, fault injection or simulation is input from the virtual model of the controller, and output waveforms collected from the test bed are fed back to a virtual control end to form a closed loop.

Specifically, the server inputs the state data into a virtual simulation model for fault simulation to obtain simulation waveform data; and inputting the state data into a semi-physical simulation model for fault simulation to obtain entity waveform data. Performing error calculation according to the entity waveform data and the simulation waveform data to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, and obtaining a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In the embodiment, virtual modeling and fault data acquisition are realized based on pure virtual simulation, test data testing and verification are realized based on semi-physical simulation, a fault twin data set driven by simulation and test data together is realized through multiple means, and then deep fault feature mining, analysis and fault diagnosis of the traction converter can be realized more accurately and efficiently based on deep learning algorithm diagnosis and fusion.

In an optional embodiment, performing weighted fusion according to the first failure prediction result and the second failure prediction result, and obtaining the failure diagnosis result includes: acquiring the prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and acquiring the prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight; and performing weighted calculation according to the first fault weight and the first fault prediction result, and the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

Specifically, in the training process of the deep residual shrinkage network model and the deep long and short term memory network model, when the error of the two models is lower than the corresponding preset model loss threshold, the server calculates the prediction accuracy of the models to obtain the prediction accuracy of the deep residual shrinkage network model and the prediction accuracy of the deep long and short term memory network model. The server obtains the prediction accuracy of the deep residual shrinkage network model, the prediction accuracy of the deep residual shrinkage network model is used as a first fault weight, and the first fault weight is multiplied by a first fault prediction result to obtain a first diagnosis result; the server obtains the prediction accuracy of the deep long-short term memory network model, the prediction accuracy of the deep long-short term memory network model is used as a second fault weight, the second fault weight and a second fault prediction result are multiplied to obtain a second diagnosis result, and the first diagnosis result are added to obtain a fault diagnosis result.

In the embodiment, the prediction results based on the two models are subjected to weighted fusion, so that the error of single model prediction can be eliminated, the prediction results of the two models are integrated, the weight of the model result with high prediction accuracy is increased according to the model prediction accuracy, and the accuracy of the fault diagnosis result is improved.

In order to easily understand the technical solution provided by the embodiment of the present application, as shown in fig. 5, a complete fault diagnosis process of the traction converter is used to briefly describe the fault diagnosis method of the traction converter provided by the embodiment of the present application:

(1) Acquiring state data and an initial digital twin model of the traction converter under different working conditions; inputting the state data into a virtual simulation model for fault simulation to obtain a fault simulation result; inputting the state data into a semi-physical simulation model for fault simulation to obtain an entity fault result; performing error calculation according to the entity fault result and the fault simulation result to obtain an error value; and adjusting the parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, so as to obtain the digital twin model.

(2) And generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

(3) Acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting digital twin state sample data into an initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result; performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value; and updating parameters of the initial depth residual error shrinkage network model based on the first loss value until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual error shrinkage network model.

(4) Acquiring digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting the digital twin state sample data into an initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result; performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value; and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is smaller than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

(5) And acquiring working condition data of the traction converter.

(6) And extracting the time-frequency characteristics of the working condition data to obtain target characteristic data.

(7) And inputting the target characteristic data into the deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into the deep long-short term memory network model for fault prediction to obtain a second fault prediction result.

(8) And performing weighted fusion according to the first fault prediction result and the second fault prediction result to obtain a fault diagnosis result.

It should be understood that, although the steps in the flowcharts related to the embodiments are shown in sequence as indicated by the arrows, the steps are not necessarily executed in sequence as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least a part of the steps in the flowcharts related to the above embodiments may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the application also provides a traction converter fault diagnosis device for realizing the traction converter fault diagnosis method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the method, so that specific limitations in one or more embodiments of the traction converter fault diagnosis device provided below can be referred to the limitations on the traction converter fault diagnosis method in the above, and details are not described herein again.

In one embodiment, as shown in fig. 6, there is provided a traction converter fault diagnosis apparatus including: an acquisition module 602, an extraction module 604, a prediction module 606, and a fusion module 608, wherein:

an obtaining module 602 is configured to obtain operating condition data of the traction converter.

And an extraction module 604, configured to extract time-frequency characteristics of the operating condition data to obtain target characteristic data.

The prediction module 606 is used for inputting the target characteristic data into the deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into the deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long and short term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions.

And the fusion module 608 is configured to perform weighted fusion according to the first failure prediction result and the second failure prediction result to obtain a failure diagnosis result.

In one embodiment, the traction converter fault diagnosis device further comprises a training module, a fault diagnosis module and a fault diagnosis module, wherein the training module is used for acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting digital twin state sample data into an initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result; performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value; and updating parameters of the initial depth residual error shrinkage network model based on the first loss value until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual error shrinkage network model.

In one embodiment, the training module is further configured to obtain digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting the digital twin state sample data into an initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result; performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value; and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is smaller than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

In one embodiment, the digital twin state sample data comprises sample working condition data and a fault type corresponding to the working condition data; the training module is further used for obtaining working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the training module is further configured to obtain state data of the traction converter and an initial digital twin model under different working conditions; inputting the state data into an initial digital twin model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data to obtain an entity fault result; performing error calculation according to the entity fault result and the fault simulation result to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, so as to obtain a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the digital twin model includes a virtual simulation model and a semi-physical simulation model; the training module is also used for inputting the state data into the virtual simulation model for fault simulation to obtain a fault simulation result; and inputting the state data into a semi-physical simulation model for fault simulation to obtain an entity fault result.

In one embodiment, the fusion module 608 is further configured to obtain a prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and obtain a prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight; and performing weighted calculation according to the first fault weight and the first fault prediction result as well as the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

All or part of each module in the traction converter fault diagnosis device can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 7. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer equipment is used for storing working condition data of the traction converter under different working conditions, fault types corresponding to the working condition data, a digital twin model, a deep long-term and short-term memory network model and a deep residual shrinkage network model. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a traction converter fault diagnosis method.

Those skilled in the art will appreciate that the architecture shown in fig. 7 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:

acquiring working condition data of a traction converter;

In one embodiment, the processor, when executing the computer program, further performs the steps of: the depth residual shrinkage network model is obtained by training in the following way: acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting digital twin state sample data into an initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result; performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value; and updating parameters of the initial depth residual error shrinkage network model based on the first loss value until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual error shrinkage network model.

In one embodiment, the processor, when executing the computer program, further performs the steps of: the deep long-short term memory network model is obtained by training in the following way: acquiring digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting the digital twin state sample data into an initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result; performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value; and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is smaller than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

In one embodiment, the processor when executing the computer program further performs the steps of: the digital twin state sample data comprises sample working condition data and a fault type corresponding to the working condition data; the method for acquiring digital twin state sample data of the traction converter under different working conditions comprises the following steps: and acquiring working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the processor when executing the computer program further performs the steps of: based on the digital twin model, the method for acquiring the working condition data of the traction converter under different working conditions and the fault types corresponding to the working condition data comprises the following steps: acquiring state data and an initial digital twin model of the traction converter under different working conditions; inputting the state data into an initial digital twin model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data to obtain an entity fault result; error calculation is carried out according to the entity fault result and the fault simulation result to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, and obtaining a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the processor when executing the computer program further performs the steps of: the digital twin model comprises a virtual simulation model and a semi-physical simulation model; inputting the state data into an initial digital twin model for fault simulation, and obtaining a fault simulation result, wherein the fault simulation result comprises the following steps: inputting the state data into a virtual simulation model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data, and obtaining an entity fault result comprises the following steps: and inputting the state data into a semi-physical simulation model for fault simulation to obtain an entity fault result.

In one embodiment, the processor, when executing the computer program, further performs the steps of: performing weighted fusion according to the first fault prediction result and the second fault prediction result, and obtaining a fault diagnosis result comprises the following steps: acquiring the prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and acquiring the prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight; and performing weighted calculation according to the first fault weight and the first fault prediction result as well as the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

acquiring working condition data of a traction converter;

In one embodiment, the computer program when executed by the processor further performs the steps of: the depth residual shrinkage network model is obtained by training in the following way: acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting digital twin state sample data into an initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result; performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value; and updating parameters of the initial depth residual error shrinkage network model based on the first loss value until the latest first loss value is smaller than a preset first loss threshold value, so as to obtain the depth residual error shrinkage network model.

In one embodiment, the computer program when executed by the processor further performs the steps of: the deep long-short term memory network model is obtained by training in the following way: acquiring digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark; inputting the digital twin state sample data into an initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result; performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value; and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is smaller than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

In one embodiment, the computer program when executed by the processor further performs the steps of: the digital twin state sample data comprises sample working condition data and a fault type corresponding to the working condition data; the method for acquiring digital twin state sample data of the traction converter under different working conditions comprises the following steps: and acquiring working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the computer program when executed by the processor further performs the steps of: based on the digital twin model, acquiring working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data comprises the following steps: acquiring state data and an initial digital twin model of the traction converter under different working conditions; inputting the state data into an initial digital twin model for fault simulation to obtain a fault simulation result; carrying out entity simulation according to the state data to obtain an entity fault result; error calculation is carried out according to the entity fault result and the fault simulation result to obtain an error value; adjusting parameters of the initial digital twin model based on the error values until the latest error value is smaller than a preset loss threshold value, and obtaining a digital twin model; and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

In one embodiment, the computer program when executed by the processor further performs the steps of: the digital twin model comprises a virtual simulation model and a semi-physical simulation model; inputting the state data into an initial digital twin model for fault simulation, and obtaining a fault simulation result, wherein the fault simulation result comprises the following steps: inputting the state data into a virtual simulation model for fault simulation to obtain a fault simulation result; performing entity simulation according to the state data, and obtaining an entity fault result comprises: and inputting the state data into a semi-physical simulation model for fault simulation to obtain an entity fault result.

In one embodiment, the computer program when executed by the processor further performs the steps of: performing weighted fusion according to the first fault prediction result and the second fault prediction result, and obtaining a fault diagnosis result comprises the following steps: acquiring the prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and acquiring the prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight; and performing weighted calculation according to the first fault weight and the first fault prediction result as well as the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, performs the steps of:

acquiring working condition data of a traction converter;

inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long-term and short-term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions;

In one embodiment, the computer program when executed by the processor further performs the steps of: performing weighted fusion according to the first fault prediction result and the second fault prediction result, and obtaining a fault diagnosis result comprises the following steps: acquiring the prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and acquiring the prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight; and performing weighted calculation according to the first fault weight and the first fault prediction result, and the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by hardware that is instructed by a computer program, and the computer program may be stored in a non-volatile computer-readable storage medium, and when executed, may include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include a Read-Only Memory (ROM), a magnetic tape, a floppy disk, a flash Memory, an optical Memory, a high-density embedded nonvolatile Memory, a resistive Random Access Memory (ReRAM), a Magnetic Random Access Memory (MRAM), a Ferroelectric Random Access Memory (FRAM), a Phase Change Memory (PCM), a graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), for example. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the various embodiments provided herein may be, without limitation, general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, or the like.

All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.

The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A traction converter fault diagnosis method, characterized in that the method comprises:

acquiring working condition data of a traction converter;

2. The method of claim 1, wherein the deep residual shrinkage network model is trained by:

acquiring digital twin state sample data and an initial depth residual shrinkage network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark;

inputting the digital twin state sample data into the initial depth residual shrinkage network model for fault diagnosis to obtain a first sample fault diagnosis result;

performing loss calculation according to the first sample fault diagnosis result and the fault type mark to obtain a first loss value;

and updating the parameters of the initial depth residual shrinkage network model based on the first loss value until the latest first loss value is less than a preset first loss threshold value, so as to obtain the depth residual shrinkage network model.

3. The method of claim 1, wherein the deep long short term memory network model is trained by:

acquiring digital twin state sample data and an initial depth long-term and short-term memory network model of the traction converter under different working conditions; the digital twin state sample data carries a fault type mark;

inputting the digital twin state sample data into the initial depth long-short term memory network model for fault diagnosis to obtain a second sample fault diagnosis result;

performing loss calculation according to the second sample fault diagnosis result and the fault type mark to obtain a second loss value;

and updating the parameters of the initial deep long-short term memory network model based on the second loss value until the latest second loss value is less than a preset second loss threshold value, so as to obtain the deep long-short term memory network model.

4. The method according to claim 2 or 3, wherein the digital twin state sample data comprises sample operating condition data and a fault type corresponding to the operating condition data;

the step of acquiring digital twin state sample data of the traction converter under different working conditions comprises the following steps:

and acquiring working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

5. The method of claim 4, wherein the obtaining operating condition data of the traction converter under different operating conditions and the fault type corresponding to each operating condition data based on the digital twin model comprises:

acquiring state data and an initial digital twin model of the traction converter under different working conditions;

inputting the state data into the initial digital twin model for fault simulation to obtain a fault simulation result;

carrying out entity simulation according to the state data to obtain an entity fault result;

performing error calculation according to the entity fault result and the fault simulation result to obtain an error value;

adjusting parameters of the initial digital twin model based on the error value until the latest error value is smaller than a preset loss threshold value, so as to obtain a digital twin model;

and generating working condition data of the traction converter under different working conditions and fault types corresponding to the working condition data based on the digital twin model.

6. The method of claim 5, wherein the digital twin model comprises a virtual simulation model and a semi-physical simulation model;

inputting the state data into the initial digital twin model for fault simulation, and obtaining a fault simulation result, wherein the fault simulation result comprises the following steps:

inputting the state data into the virtual simulation model for fault simulation to obtain a fault simulation result;

the performing entity simulation according to the state data to obtain an entity fault result comprises:

and inputting the state data into the semi-physical simulation model for fault simulation to obtain an entity fault result.

7. The method according to claim 1, wherein the performing weighted fusion according to the first failure prediction result and the second failure prediction result to obtain a failure diagnosis result comprises:

acquiring the prediction accuracy of the deep residual shrinkage network model to obtain a first fault weight, and acquiring the prediction accuracy of the deep long-term and short-term memory network model to obtain a second fault weight;

and performing weighted calculation according to the first fault weight and the first fault prediction result, and the second fault weight and the second fault prediction result to obtain a fault diagnosis result.

8. A traction converter fault diagnostic apparatus, characterized in that the apparatus comprises:

the prediction module is used for inputting the target characteristic data into a deep residual shrinkage network model for fault prediction to obtain a first fault prediction result, and inputting the target characteristic data into a deep long-short term memory network model for fault prediction to obtain a second fault prediction result; the deep residual shrinkage network model and the deep long-term and short-term memory network model are obtained by training digital twin state sample data of the traction converter under different working conditions;

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.