CN110991471A - Fault diagnosis method for high-speed train traction system - Google Patents

Fault diagnosis method for high-speed train traction system Download PDF

Info

Publication number
CN110991471A
CN110991471A CN201910708383.0A CN201910708383A CN110991471A CN 110991471 A CN110991471 A CN 110991471A CN 201910708383 A CN201910708383 A CN 201910708383A CN 110991471 A CN110991471 A CN 110991471A
Authority
CN
China
Prior art keywords
fault
encoder
lstm
self
sequence data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910708383.0A
Other languages
Chinese (zh)
Other versions
CN110991471B (en
Inventor
冒泽慧
闫宇
姜斌
严星刚
吕迅竑
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Aeronautics and Astronautics
Original Assignee
Nanjing University of Aeronautics and Astronautics
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Aeronautics and Astronautics filed Critical Nanjing University of Aeronautics and Astronautics
Priority to CN201910708383.0A priority Critical patent/CN110991471B/en
Publication of CN110991471A publication Critical patent/CN110991471A/en
Application granted granted Critical
Publication of CN110991471B publication Critical patent/CN110991471B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M17/00—Testing of vehicles
    • G01M17/08—Railway vehicles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
    • Y04S10/52—Outage or fault management, e.g. fault detection or location

Abstract

The invention discloses a fault diagnosis method for a traction system of a high-speed train, which relates to the field of fault diagnosis of the high-speed train and comprises the following steps: acquiring sequence data from a semi-physical simulation platform and preprocessing the sequence data to obtain a data set, wherein the data set comprises a training set and a test set; improving an LSTM self-encoder by utilizing state differential feedback control to obtain an LSTM self-encoder I, wherein the LSTM self-encoder I consists of L LSTM units; training the LSTM self-encoder I by using the data set to obtain an LSTM self-encoder II; extracting original feature vectors from the two pairs of test sets by using the LSTM self-encoder; performing feature dimension reduction on the original feature vector by using a t-SNE algorithm; and carrying out fault diagnosis on the original feature vector subjected to dimension reduction by using a DBSCAN clustering method to obtain a diagnosis result. The invention can effectively solve the problems of difficult diagnosis and low diagnosis accuracy of the tiny gradual change fault of the traction system of the high-speed train.

Description

Fault diagnosis method for high-speed train traction system
Technical Field
The invention relates to the field of high-speed train fault diagnosis, in particular to a method for diagnosing a tiny gradual change fault of a high-speed train traction system based on data learning.
Background
At present, high-speed trains play more and more important roles in the aspects of passenger transport and freight transport in China, and a traction system is used as a power core system of the high-speed trains, so that accidents such as train stopping, delay and the like are caused when the traction system breaks down, and huge loss is caused. The micro gradual change fault occurs in the initial stage of the significant fault, and has the characteristics of unobvious fault characteristics and easy coverage by unknown disturbance and noise. Therefore, the detection and diagnosis of the slight gradual change fault of the traction system are difficult. The micro gradual change fault of the traction system is effectively detected and diagnosed, so that effective measures are taken in time, the safety of the system can be improved, and the maintenance cost and loss are reduced.
The effective fault diagnosis for the tiny gradual change fault can improve the safety of the system, reduce the maintenance cost and loss, and has important significance in the aspects of safe maintenance, equipment health management and the like. Due to the fact that the complexity of a traction system and the fault characteristics of the tiny gradual-change fault are not obvious, and the change characteristics are not obvious, detection and diagnosis of the tiny gradual-change fault are difficult to achieve effectively based on a data driving algorithm at present.
Disclosure of Invention
The invention aims to provide a fault diagnosis method for a traction system of a high-speed train, which aims to solve the technical problems of difficulty in diagnosing tiny gradual-change faults and low diagnosis accuracy in the prior art.
The invention provides a fault diagnosis method for a traction system of a high-speed train, which comprises the following steps:
(a) acquiring sequence data from a semi-physical simulation platform and preprocessing the sequence data to obtain a data set, wherein the data set comprises a training set and a test set;
(b) improving an LSTM self-encoder by utilizing state differential feedback control to obtain an LSTM self-encoder I, wherein the LSTM self-encoder I consists of L LSTM units;
(c) training the LSTM self-encoder I by using the data set to obtain an LSTM self-encoder II;
(d) extracting original feature vectors from the two pairs of test sets by using the LSTM self-encoder;
(e) performing feature dimension reduction on the original feature vector by using a t-SNE algorithm;
(f) and carrying out fault diagnosis on the original feature vectors subjected to the dimension reduction by using a DBSCAN clustering method to obtain a diagnosis result.
Compared with the prior art, the method for diagnosing the tiny gradual change fault of the high-speed train traction system based on data learning has the following beneficial effects that:
the method is characterized in that feature extraction is carried out on the basis of sequence data of the micro gradual change fault which is carried out by an improved LSTM unit self-encoder, and the method can better extract the original features of the micro gradual change fault;
the original features are subjected to dimension reduction processing through a t-SNE algorithm, so that the redundancy of original feature information is reduced, the diagnostic performance is improved, and the calculation complexity is reduced;
the non-supervision diagnosis of the tiny gradual-change fault is realized through the DBSCAN clustering method, the marking in advance is not needed, the diagnosis accuracy is very high, the unknown fault has certain diagnosis capability, and the diagnosis requirement of the tiny gradual-change fault is met.
Aiming at a traction motor system of a high-speed train, the method diagnoses tiny gradual-change faults such as degradation faults of a middle capacitor, a middle resistor and a speed sensor of the traction system, has wider application range and has engineering application value.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a diagram of an improved LSTM cell structure provided by an embodiment of the present invention;
FIG. 2(a) is a codec model overall framework and training flow provided by an embodiment of the present invention;
FIG. 2(b) provides an overall framework and flow for a fault diagnosis model for an embodiment of the present invention;
FIG. 3(a) is a diagram illustrating the effect of different dimension reduction methods for providing known fault types according to an embodiment of the present invention;
FIG. 3(b) is an effect diagram of the t-SNE dimension reduction method for providing unknown fault types according to the embodiment of the present invention;
FIG. 4(a) is a diagram illustrating the clustering effect of known fault categories according to an embodiment of the present invention;
FIG. 4(b) is a diagram illustrating the clustering effect of unknown fault categories according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of the architecture of a high-speed train traction system according to an embodiment of the present invention;
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and obviously, the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplification of description, but do not indicate or imply that the device or element referred to must have a specific orientation, be constructed in a specific orientation, and be operated, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. Specific meanings of the above terms in the present invention can be specifically understood by those of ordinary skill in the art.
The following description will explain embodiments of the present invention in further detail with reference to the accompanying drawings.
As shown in fig. 2(b), the method for diagnosing a micro gradual change fault of a high-speed train traction system based on data learning mainly comprises the following steps:
1. acquiring sequence data from a semi-physical simulation platform and preprocessing the sequence data to obtain a data set, wherein the data set comprises a training set and a test set
Preprocessing the acquired data: supplementing missing values, modifying abnormal values, smoothing and filtering, reducing noise and carrying out normalization processing.
Specifically, filling missing values in data by adopting a 2-nearest neighbor method for the missing values of the training set and the testing set; analyzing the abnormal values of the sequence data through a box diagram, and modifying the numerical values judged as the abnormal values in the data by adopting a 2-nearest distance neighbor method; smoothing filtering is carried out on the sequence data in a wavelet filtering mode, and noise in the sequence data is reduced; and (5) performing Z-normalization on the sequence data to finish the normalization of the sequence data.
2. Improving an LSTM self-encoder by utilizing state differential feedback control to obtain an LSTM self-encoder I, wherein the LSTM self-encoder I consists of L LSTM units
The basic LSTM unit is modified to apply state differential feedback control to the LSTM unit: state differential feedback control is applied in the modified LSTM unit.
In particular, the state differential of the state vector of the dynamic system at time t-1 and time t-2Signal pt-1Is updated at the time t-1 and is fed back to the system at the time t to participate in controlling the LSTM hidden unit state vector ctForgetting and updating. The improved LSTM unit is described by the following formula:
Figure RE-GDA0002383766780000051
wherein f istIndicating a forgotten gate level output, rtRepresenting refresh gate level output, ctThe state of the hidden unit is represented,
Figure BDA0002152907010000052
representing the new state quantity to be selected, ptDifferential vector, o, representing the hidden state vector of the celltRepresents the output of the output gate layer, htRepresenting the LSTM unit output.
If LSTM control vector
Figure BDA0002152907010000053
And if the LSTM unit hidden unit state c is stable under the action of the stable vector to be updated, the LSTM network is stable. Equation (4) shows the control vector
Figure BDA0002152907010000054
Related to the state differential vector p for the LSTM cell state c. When the sequence sample of the micro gradual change fault sequence data is short, the difference vector of the fault information variable is linearly changed when Taylor expands to the first order approximation. When the LSTM unit learns that the hidden unit state c has higher correlation with the tiny gradual change fault information variable, the difference vector p of the hidden unit state c of the LSTM unit is stable along with the iterative input of the sequence data. While
Figure BDA0002152907010000061
Has strong correlation with the difference vector p, so that a stable control vector is generated
Figure BDA0002152907010000062
At this time, the LSTM unit is hiddenThe meta-state variables are not easily forgotten and updated. When the relation between the hidden unit state c learned by the LSTM unit and the tiny gradual change fault information variable is weak, the hidden unit state c of the LSTM unit is controlled by the unstable control vector due to the unstable difference vector p
Figure BDA0002152907010000063
Can be forgotten and updated frequently. Secondly, the trained LSTM self-encoder can acquire a differential trend signal representing a fault information variable, and can diagnose a tiny gradual-change fault early.
3. Training the LSTM transcoder I with the data set to obtain an LSTM transcoder II
And obtaining the LSTM self-encoder by unsupervised training of the codec model by using the feature vectors of the training set. After the traction system sequence data is processed based on the improved LSTM self-encoder, the fault information variable is embedded into the original feature vector with fixed length. During the training of the codec model, the LSTM reads the sequential samples X of the pull system from the encoder, each sample
Figure RE-GDA0002383766780000061
Is vector sequence data, where X is X1,x2...,xL-1,xL，NvRepresenting the vector dimension and L the sequence data sample length. After the reading process of the sequence data sample X is completed, the LSTM self-encoder has the last layer of hidden unit state
Figure RE-GDA0002383766780000062
Will be retained and used as the original feature vector. GRU decoder initializes hidden unit state of first layer as original characteristic vector
Figure RE-GDA00023837667800000610
And outputs the target sequence
Figure RE-GDA0002383766780000063
Wherein
Figure RE-GDA0002383766780000064
The target of the training is the codec output sequence
Figure RE-GDA0002383766780000065
Remains similar to the input sequence X under the metric, but in reverse order. In the present invention, through minimization
Figure RE-GDA0002383766780000066
Square error J between X andedtraining is carried out:
Figure RE-GDA0002383766780000067
wherein DNRepresenting the entire training set, in the present invention the optimizer of the codec model training is the Adam optimizer.
4. Extracting original feature vectors from the test set by using the LSTM self-encoder
Using a trained LSTM self-encoder model, sequence samples X of a test set are input at an input section, each sample
Figure RE-GDA0002383766780000068
Is vector sequence data, where X is X1,x2...,xL-1,xL， NvRepresenting the vector dimension and L the sequence data sample length. After the reading process of the sequence data sample X is completed, the last layer of hidden unit state in the LSTM self-encoder
Figure RE-GDA0002383766780000069
Will be retained and taken as the original feature vector, thereby completing the task of LSTM autoencoder extraction of the original feature vector of the test set data.
And extracting an original feature vector from the sequence data of the tiny gradual change fault of the traction system by the trained LSTM self-encoder.
5. Performing feature dimension reduction on the original feature vector by using a t-SNE algorithm
By t-SNE algorithm to original feature vector
Figure RE-GDA0002383766780000071
And dimension reduction is carried out, and the problems of high dimension and information redundancy of the original characteristic vector are solved. In the t-SNE algorithm, the loss function JSNEFor the joint probability P of the data distribution in the original feature vector space and the target dimension reduction spaceDAnd QDThe difference in (a). The probability distribution difference is the Kullback-Leibler divergence. Loss function J of T-SNE algorithmSNEThe following are shown:
Figure RE-GDA0002383766780000072
wherein p isijAnd q isijIs defined as:
Figure RE-GDA0002383766780000073
the indices i, j, k, l denote the i, j, k, l samples,
Figure RE-GDA0002383766780000075
representing the original feature vector in the high-dimensional space, z representing the feature vector after dimensionality reduction, sigma representing the sum of the original feature vector and the sigma
Figure RE-GDA0002383766780000076
Bandwidth of the associated gaussian kernel function. During optimization, KL divergence is minimized by a stochastic gradient descent method. Loss function JSNEThe gradient information of (a) is given by:
the indices i, j, k, l denote the i, j, k, l samples,
Figure BDA0002152907010000077
representing the original feature vector in the high-dimensional space, z representing the feature vector after dimensionality reduction, sigma representing the sum of the original feature vector and the sigma
Figure BDA0002152907010000078
Bandwidth of the associated gaussian kernel function. In the optimization process, the random ladder is usedThe degree reduction method is used to minimize the KL divergence. Loss function JSNEThe gradient information of (a) is given by:
Figure BDA0002152907010000081
in the t-SNE algorithm, the similarity of the feature vector distribution before and after dimensionality reduction is maintained by minimizing the KL distance from the high-dimensional original feature vector space to the low-dimensional feature vector space.
6. Carrying out fault diagnosis on the original feature vectors subjected to dimension reduction by a DBSCAN clustering method to obtain a diagnosis result
Unsupervised training of the codec model using the training set sequence data yields an LSTM autoencoder model. Raw feature vector of sequence data to be diagnosed
Figure BDA0002152907010000082
The LSTM extracted from the trained self-encoder. Original feature vector
Figure BDA0002152907010000083
The dimension is high, information redundancy exists, and in order to reduce the information redundancy and reduce the complexity of the operation of the algorithm, the original characteristic vector is subjected to t-SNE algorithm
Figure BDA0002152907010000084
And obtaining a final feature vector z after dimension reduction. And finally, clustering the characteristic vector z by using a DBSCAN clustering method to realize the diagnosis of the tiny gradual change fault of the traction system.
The method for diagnosing the tiny gradual change fault of the high-speed train traction system based on data learning carries out simulation verification:
step 1, firstly, confirming the types of the tiny gradual-change faults, wherein the tiny gradual-change fault types comprise a middle capacitance degradation fault (fault I), a middle resistance degradation fault (fault II) and a speed sensor degradation fault (fault III) in the verification. Also included in this experimental validation is a speed sensor bias fault (fault IV), which is a minor abrupt fault. And the fault I and the fault II are used in the unsupervised fault diagnosis model training and final fault diagnosis model testing process. The untrained fault classes (fault III and fault IV in the experimental verification) are only used for verifying the trained fault diagnosis model property test, and the fault III and the fault IV are used for testing the generalization ability of the model to unknown faults. The details of the four faults in the experimental verification are shown in table 1.
Table 1 fault type description
Figure BDA0002152907010000091
The data for fault I and fault II are divided into training and test sets, with the data for each sample being in the form of a vector of sequence data. There are a total of about 14000 samples in the form of sequence data vectors in the entire data. 60% of the data in the entire data set was randomly selected to make up the training set (8500 samples), and the remaining samples made up the test set (5455 samples). Wherein, the number of the healthy samples is 1180, the number of the fault I samples is 990, and the number of the fault II samples is 935. Similarly, the number of samples for fault III is 1175, and the number of samples for fault IV is 1175, each series data sample ranging in length from 20 to 40.
And 2, extracting original characteristic vectors of the fault I and fault II test sets through the trained improved LSTM self-encoder. And the t-SNE is applied to the feature vector dimension reduction. FIG. 3(a) shows the dimensionality reduction effect of different dimensionality reduction methods (including t-SNE, kernel PCA and Isomap) on feature vectors of known fault types. Where · represents a healthy sample point, + represents a sample point with a fault I, and x represents a sample point with a fault II. According to the PCA dimension reduction result and the Isomap dimension reduction result, the feature vectors of the healthy sample are distributed in a stable circular ring shape in the low-dimensional space, two types of micro gradual change faults of a fault I and a fault II are distributed in a similar circular cone shape starting from the healthy circular ring in the low-dimensional space, and the distribution is consistent with the gradual time change characteristic of fault information variables in the micro gradual change faults, so that the LSTM self-encoder is proved to successfully extract the feature vectors related to the micro gradual change fault information variables. In the present invention, experimental verification tests are performed on the types of failures that are not used for codec training (unknown failure classes). The test set of faults III and IV is used to test the generalization ability of fault diagnosis algorithms to unknown fault types. Fault III is a minor gradual fault in the traction system and fault IV is a minor abrupt fault occurring in the traction system. FIG. 3(b) shows the dimension reduction effect of the feature vector of the unknown fault by using the t-SNE algorithm, and the dimension reduction effect of the feature vector has good inter-class distance and intra-class distance for the unknown fault class.
And 3, applying the DBSCAN clustering algorithm to the feature vector samples subjected to dimensionality reduction for clustering to finish fault diagnosis of the test set. The DBSCAN clustering algorithm is a density-based clustering algorithm, which generally assumes that classes can be determined by how closely the samples are distributed. Samples of the same class are closely related, i.e., samples of the same class must exist a short distance around any sample of the class. By categorizing closely connected samples into a class, a cluster category is obtained. By classifying all groups of closely connected samples into different categories, we obtain the final results of all the clustering categories. The method describes how close a sample set is based on a set of neighborhoods. The method used by the DBSCAN is simple, a core object without a category is selected as a seed at will, and then a sample set with the density reaching all the core objects is found, namely a cluster. And then continuously selecting another core object without categories to search a sample set with the reachable density, so as to obtain another cluster. Run until all core objects have a category.
Fig. 4(a) clusters data of known fault categories, and the clustering result shows that the fault diagnosis algorithm based on the improved LSTM self-encoder can successfully implement degradation fault diagnosis of the intermediate capacitance and the intermediate resistance of the traction system, i.e., implement micro gradual change fault diagnosis of the traction system. Fig. 4(b) clusters data of unknown fault categories, and the clustering result shows that the algorithm provided by the invention has a good diagnosis effect on some unknown faults.
As shown in the attached figure 4, the method can effectively diagnose the tiny gradual change fault of the traction motor of the high-speed train. For some unmarked faults, the method also has good diagnosis effect and strong generalization capability.
As shown in fig. 5, an embodiment of the present invention further provides a schematic structural diagram of a high-speed train traction system, where the traction system 120 includes: at least one processor 121, such as a CPU, at least one network interface 124 or other user interface 123, memory 125, at least one communication bus 122. A communication bus 122 is used to enable connection communication between these components. Optionally, a user interface 123 is also included, including a display, a keyboard or a pointing device (e.g., a mouse, trackball, touch pad or touch sensitive display screen). Memory 125 may comprise high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory. The memory 125 may optionally include at least one memory device located remotely from the processor 121.
In some embodiments, memory 125 stores elements, executable modules or data structures, or a subset thereof, or an expanded set thereof as follows:
an operating system 1251, containing various system programs, for implementing various basic services and for processing hardware-based tasks;
the application programs 1252 include various application programs for implementing various application services.
Specifically, the processor 121 is configured to execute the flows corresponding to steps (1) - (4); in this process, the processor 121 needs to receive the signal processing sent by the sensor group 126 through the network interface 124.
Meanwhile, the processor 121 is further configured to: the fault diagnosis result is output to the visualized interface through the user interface 123.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims (6)

1. A fault diagnosis method for a traction system of a high-speed train is characterized by comprising the following steps:
(a) acquiring sequence data from a semi-physical simulation platform and preprocessing the sequence data to obtain a data set, wherein the data set comprises a training set and a test set;
(b) improving an LSTM self-encoder by utilizing state differential feedback control to obtain an LSTM self-encoder I, wherein the LSTM self-encoder I consists of L LSTM units;
(c) training the LSTM self-encoder I by using the data set to obtain an LSTM self-encoder II;
(d) extracting original feature vectors from the two pairs of test sets by using the LSTM self-encoder;
(e) performing feature dimension reduction on the original feature vector by using a t-SNE algorithm;
(f) and carrying out fault diagnosis on the original feature vector subjected to dimension reduction by using a DBSCAN clustering method to obtain a diagnosis result.
2. The method for diagnosing the fault of the traction system of the high-speed train according to claim 1, wherein: the pretreatment in (a) comprises:
performing missing value filling on the sequence data by using a k-nearest distance neighbor method;
analyzing the abnormal values of the sequence data through a box diagram, and modifying the sequence data judged as the abnormal values by using a k-nearest distance neighbor method;
and smoothing and filtering the sequence data by using wavelet filtering, and normalizing the sequence data by using a Z-score normalization method.
3. The method for diagnosing the fault of the traction system of the high-speed train according to claim 1, wherein: the (c) includes:
training the LSTM self-encoder I by using the data set to obtain an LSTM self-encoder II;
the LSTM self-encoder reads the data set as soon as the data set is read
Figure RE-FDA0002383766770000011
Is vector sequence data, X ═ X1,x2...,xL-1,xL，NvRepresenting the vector dimension, L representing the sequence data sample length;
a last layer of hidden unit state of the LSTM autoencoder
Figure RE-FDA0002383766770000021
As a feature vector, taking the feature vector as a hidden unit state of a first layer of a GRU decoder
Figure RE-FDA0002383766770000022
The GRU decoder output sequence
Figure RE-FDA0002383766770000023
Wherein
Figure RE-FDA0002383766770000024
To the above
Figure RE-FDA0002383766770000025
Minimizing the square error J with XedThe training is carried out by the user,
Figure RE-FDA0002383766770000026
wherein DNRepresenting the entire training set, the training remains similar under the metric, in reverse order.
4. The method for diagnosing the fault of the traction system of the high-speed train according to claim 1, wherein: in the (d), the method further comprises:
and the LSTM self-encoder II extracts an original feature vector, the original feature vector contains fault information, and the fault information variable is embedded into the original feature vector with a fixed length.
5. The method for diagnosing the fault of the traction system of the high-speed train according to claim 1, wherein: the (e) comprises:
defining a loss function JSNESaid loss function JSNEThe joint probability P of the original characteristic vector and the data distribution in the target dimension reductionDAnd QDThe difference of (2), the probability PDAnd QDThe difference of the difference probability distribution adopts Kullback-Leibler divergence;
the Kullback-Leibler divergence was minimized by the random gradient descent method.
6. The method for diagnosing the fault of the traction system of the high-speed train according to claim 1, wherein: in the (f) above, the first step of the process,
and carrying out fault diagnosis on the feature vectors subjected to the dimension reduction by using a DBSCAN clustering method to obtain a diagnosis result.
CN201910708383.0A 2019-08-01 2019-08-01 Fault diagnosis method for high-speed train traction system Active CN110991471B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910708383.0A CN110991471B (en) 2019-08-01 2019-08-01 Fault diagnosis method for high-speed train traction system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910708383.0A CN110991471B (en) 2019-08-01 2019-08-01 Fault diagnosis method for high-speed train traction system

Publications (2)

Publication Number Publication Date
CN110991471A true CN110991471A (en) 2020-04-10
CN110991471B CN110991471B (en) 2023-07-11

Family

ID=70081714

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910708383.0A Active CN110991471B (en) 2019-08-01 2019-08-01 Fault diagnosis method for high-speed train traction system

Country Status (1)

Country Link
CN (1) CN110991471B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111860686A (en) * 2020-07-31 2020-10-30 北京航空航天大学 Causal relationship adjacency matrix characteristic extraction method for fault diagnosis
CN112861071A (en) * 2021-02-05 2021-05-28 哈尔滨工程大学 High-speed rail traction system anomaly detection method based on deep self-coding
CN113139590A (en) * 2021-04-13 2021-07-20 索信达(北京)数据技术有限公司 Dimension reduction method and device for time series data, computer equipment and storage medium
CN113507331A (en) * 2021-07-01 2021-10-15 西北工业大学 Sight distance and non-sight distance transmission identification method and device based on channel state information

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108197648A (en) * 2017-12-28 2018-06-22 华中科技大学 A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models
CN108536123A (en) * 2018-03-26 2018-09-14 北京交通大学 The method for diagnosing faults of the train control on board equipment of the long neural network of memory network combination in short-term
KR20190021812A (en) * 2017-08-24 2019-03-06 한국과학기술원 Method and apparatus for quantitative evaluation assessment of vr content perceptual quality using deep running analysis of vr sickness factors
CN109561084A (en) * 2018-11-20 2019-04-02 四川长虹电器股份有限公司 URL parameter rejecting outliers method based on LSTM autoencoder network
CN109919364A (en) * 2019-02-18 2019-06-21 华南理工大学 Multivariate Time Series prediction technique based on adaptive noise reduction and integrated LSTM
CN110021165A (en) * 2019-03-18 2019-07-16 浙江工业大学 A kind of traffic flow forecasting method based on Autoencoder-LSTM Fusion Model

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20190021812A (en) * 2017-08-24 2019-03-06 한국과학기술원 Method and apparatus for quantitative evaluation assessment of vr content perceptual quality using deep running analysis of vr sickness factors
CN108197648A (en) * 2017-12-28 2018-06-22 华中科技大学 A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models
CN108536123A (en) * 2018-03-26 2018-09-14 北京交通大学 The method for diagnosing faults of the train control on board equipment of the long neural network of memory network combination in short-term
CN109561084A (en) * 2018-11-20 2019-04-02 四川长虹电器股份有限公司 URL parameter rejecting outliers method based on LSTM autoencoder network
CN109919364A (en) * 2019-02-18 2019-06-21 华南理工大学 Multivariate Time Series prediction technique based on adaptive noise reduction and integrated LSTM
CN110021165A (en) * 2019-03-18 2019-07-16 浙江工业大学 A kind of traffic flow forecasting method based on Autoencoder-LSTM Fusion Model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
魏庆东等: "基于数据扩展的短时交通流量预测", vol. 32, no. 2 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111860686A (en) * 2020-07-31 2020-10-30 北京航空航天大学 Causal relationship adjacency matrix characteristic extraction method for fault diagnosis
CN111860686B (en) * 2020-07-31 2022-08-19 北京航空航天大学 Causal relationship adjacency matrix characteristic extraction method for fault diagnosis
CN112861071A (en) * 2021-02-05 2021-05-28 哈尔滨工程大学 High-speed rail traction system anomaly detection method based on deep self-coding
CN113139590A (en) * 2021-04-13 2021-07-20 索信达(北京)数据技术有限公司 Dimension reduction method and device for time series data, computer equipment and storage medium
CN113139590B (en) * 2021-04-13 2022-03-22 索信达(北京)数据技术有限公司 Dimension reduction method and device for time series data, computer equipment and storage medium
CN113507331A (en) * 2021-07-01 2021-10-15 西北工业大学 Sight distance and non-sight distance transmission identification method and device based on channel state information

Also Published As

Publication number Publication date
CN110991471B (en) 2023-07-11

Similar Documents

Publication Publication Date Title
CN110132598B (en) Fault noise diagnosis algorithm for rolling bearing of rotating equipment
CN106895975B (en) Method for Bearing Fault Diagnosis based on Stacked SAE deep neural network
CN111666982B (en) Electromechanical equipment fault diagnosis method based on deep neural network
CN110991471A (en) Fault diagnosis method for high-speed train traction system
CN102112933B (en) Error detection method and system
Yang et al. ART–KOHONEN neural network for fault diagnosis of rotating machinery
Yang et al. Bearing fault automatic classification based on deep learning
CN111914883B (en) Spindle bearing state evaluation method and device based on deep fusion network
He et al. Deep variational autoencoder classifier for intelligent fault diagnosis adaptive to unseen fault categories
CN113935406B (en) Mechanical equipment unsupervised fault diagnosis method based on countermeasure flow model
CN112414694B (en) Equipment multistage abnormal state identification method and device based on multivariate state estimation technology
CN105425150B (en) A kind of Method of Motor Fault Diagnosis based on RBF and PCA-SVDD
CN115828140B (en) Method, system and application for detecting fault by fusing neighborhood mutual information and random forest
Gao et al. A process fault diagnosis method using multi‐time scale dynamic feature extraction based on convolutional neural network
CN116226646A (en) Method, system, equipment and medium for predicting health state and residual life of bearing
Cheng et al. A Koopman operator approach for machinery health monitoring and prediction with noisy and low-dimensional industrial time series
CN114297918A (en) Aero-engine residual life prediction method based on full-attention depth network and dynamic ensemble learning
CN113792758A (en) Rolling bearing fault diagnosis method based on self-supervision learning and clustering
CN114742115A (en) Rolling bearing fault diagnosis model and method based on temperature and vibration characteristic fusion
Chou et al. SHM data anomaly classification using machine learning strategies: A comparative study
CN115290326A (en) Rolling bearing fault intelligent diagnosis method
CN114356897A (en) Electromechanical actuator fault diagnosis method based on data fusion
CN109255201A (en) A kind of ball screw assembly, health evaluating method based on SOM-MQE
Zhou et al. Degradation state recognition of rolling bearing based on K-means and CNN algorithm
CN114692694B (en) Equipment fault diagnosis method based on feature fusion and integrated clustering

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant