CN114004268A - Online diagnosis method and device for traction system fault - Google Patents

Online diagnosis method and device for traction system fault Download PDF

Info

Publication number
CN114004268A
CN114004268A CN202010738088.2A CN202010738088A CN114004268A CN 114004268 A CN114004268 A CN 114004268A CN 202010738088 A CN202010738088 A CN 202010738088A CN 114004268 A CN114004268 A CN 114004268A
Authority
CN
China
Prior art keywords
fault
traction system
gmm
probability
hmm model
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.)
Pending
Application number
CN202010738088.2A
Other languages
Chinese (zh)
Inventor
李学明
徐绍龙
甘韦韦
郭维
袁靖
彭辉
黄明明
廖亮
谭永光
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.)
Zhuzhou CRRC Times Electric Co Ltd
Original Assignee
Zhuzhou CRRC Times Electric Co Ltd
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 Zhuzhou CRRC Times Electric Co Ltd filed Critical Zhuzhou CRRC Times Electric Co Ltd
Priority to CN202010738088.2A priority Critical patent/CN114004268A/en
Publication of CN114004268A publication Critical patent/CN114004268A/en
Pending legal-status Critical Current

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
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/29—Graphical models, e.g. Bayesian networks
    • G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks

Abstract

The invention relates to a method and a device for online diagnosis of traction system faults and a computer readable storage medium. The online diagnosis method comprises the following steps: monitoring a fault characteristic variable of the traction system; outputting a corresponding fault classification enabling flag in response to the fault characteristic variable being abnormal; collecting relevant time domain statistics of the traction system according to the fault classification enabling mark; extracting a correlation feature indicator from the correlation time domain statistic to generate a sequence of observation vectors; and loading the observation vector sequence into a pre-trained related fault model to determine the fault model corresponding to the maximum probability value as the fault diagnosis result of the traction system. The invention can carry out on-line positioning on the faults of the traction system and realize the accurate classification of various fault sources.

Description

Online diagnosis method and device for traction system fault
Technical Field
The invention relates to a fault diagnosis technology of a rail transit traction system, in particular to an online diagnosis method of a traction system fault and an online diagnosis device of the traction system fault.
Background
In the running process of trains such as locomotives, motor train units and the like, any tiny or potential faults and hidden dangers can cause chain reaction to cause accidents and even cause disastrous results if the tiny or potential faults and hidden dangers cannot be diagnosed and effective in time. The traction system is used as the heart of a high-speed train, and the running state of the traction system is influenced by factors such as complex running environment, corrosion, temperature, humidity, power supply surge, static electricity and the like, so that the traction system is easy to break down and cannot be eliminated in a periodic maintenance mode. If the train has faults in the running process, the online accurate fault source positioning can be preferably realized so as to timely eliminate the faults or execute a proper isolation protection strategy. If the failure causes are not diagnosed in time and the failure is eliminated, the driving accident will be caused, the normal operation of the train is delayed, and the transportation order of the whole line and the whole line is affected.
At present, the fault diagnosis of the train traction system is still mainly based on the acquisition of sensor signals, and simple fault detection methods such as over-threshold alarm are adopted, for example: detecting overvoltage and overcurrent at the network side of the traction system; and the input/output overcurrent detection, the intermediate direct-current overvoltage/undervoltage detection, the detection and protection functions of the cooling system, the overhigh/overlow water pressure and the like of the traction converter. However, such detection methods all belong to the detection of fault characterization, and the real reasons for the occurrence of such characterization cannot be diagnosed, and a temporary stop is generally required to be checked by a driver or system maintenance personnel.
In order to overcome the above defects in the prior art, there is a need in the art for an online fault diagnosis technique for a traction system, which is used for online positioning of faults of the traction system and accurate classification of various fault sources.
Disclosure of Invention
The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.
In order to overcome the defects in the prior art, the invention provides an online diagnosis method for the faults of the traction system, an online diagnosis device for the faults of the traction system and a computer readable storage medium, which are used for carrying out online positioning on the faults of the traction system and realizing accurate classification of various fault sources.
The online diagnosis method for the fault of the traction system provided by the invention comprises the following steps: monitoring a fault characteristic variable of the traction system; outputting a corresponding fault classification enabling flag in response to the fault characteristic variable being abnormal; collecting relevant time domain statistics of the traction system according to the fault classification enabling mark; extracting a correlation feature indicator from the correlation time domain statistic to generate a sequence of observation vectors; and loading the observation vector sequence into a pre-trained related fault model to determine the fault model corresponding to the maximum probability value as the fault diagnosis result of the traction system.
Optionally, in some embodiments of the present invention, the online diagnosis method may further include: judging whether the fault characteristic variable exceeds a given threshold value or not; and determining that the fault signature variable is abnormal in response to the fault signature variable exceeding the threshold.
Optionally, in some embodiments of the present invention, the step of collecting the relevant time domain statistics according to the fault classification enabling flag may include: the fault signature variables are obtained in response to the fault classification enable flags, and other relevant time domain statistics are obtained from the corresponding sensors according to the fault classification enable flags.
Preferably, in some embodiments of the present invention, the online diagnosis method may further include: performing fault feature analysis on historical fault data of the traction system to extract relevant feature variables; calculating a fault characteristic index according to the relevant characteristic variable to obtain a sample library of an observation vector sequence of the fault characteristic index; and training various types of fault models of the traction system according to the sample library.
Preferably, in some embodiments of the present invention, the step of obtaining the sample library of observation vector sequences may further comprise: defining a sample library having K of the observation vector sequences as O ═ { O ═ O(1),O(2),…,O(K)And (c) the step of (c) in which,
Figure BDA0002605697650000021
for the k-th sequence of observation vectors,
Figure BDA0002605697650000022
is O(k)D-dimensional observation vector at time t.
Preferably, in some embodiments of the present invention, the step of training the various types of fault models may include: establishing an HMM model describing statistical characteristics of the observation vector sequence; establishing a GMM-HMM model according to the HMM model; determining a probability density function of the observation vector sequence according to the GMM-HMM model; and initializing parameters according to the sample library of the historical fault data, and performing parameter estimation on the GMM-HMM model.
Preferably, in some embodiments of the present invention, the step of establishing the HMM model may further comprise: selecting N states S ═ S1,s2,…,sNAnd M Gaussian mixture elements
Figure BDA0002605697650000031
To build the HMM model, wherein xnmIs a state snThe mth mixing element of (1).
Preferably, in some embodiments of the present invention, the step of establishing the GMM-HMM model may further comprise: describing the GMM-HMM model as λ ═ (π, A, C, μ, U), where π is the initial state probability distribution, A is the state transition probability distribution, C is the mixed right weight, μ is the mean vector, and U is the covariance matrix; defining the initial state probability distribution pi as pi ═ pi1,π1,…,πN]TWherein q istFor the state of the observation vector sequence at time t, P [ ·]Is the probability, pi, of the observation vector sequencen＝P[q1＝sn]Is not less than 0 and
Figure BDA0002605697650000032
defining the state transition probability distribution A as
Figure BDA0002605697650000033
Wherein, aij＝P[qt+1＝sj|qt＝si]Not less than 0 and all i satisfy
Figure BDA0002605697650000034
Defining the mixing right C as
Figure BDA0002605697650000035
Wherein, cnmNot less than 0 and all n satisfy
Figure BDA0002605697650000036
Defining the mean vector μ as
Figure BDA0002605697650000037
Wherein the content of the first and second substances,
Figure BDA0002605697650000038
represents a mixing element xnmD-dimensional mean vector of(ii) a And defining the covariance matrix U as
Figure BDA0002605697650000041
Wherein, UnmRepresents a mixing element xnmD x D dimensional covariance matrix.
Preferably, in some embodiments of the present invention, the step of determining the probability density function of the observation vector sequence may further comprise: will state snIs observed vector
Figure BDA0002605697650000042
Probability density function of
Figure BDA0002605697650000043
Is described as
Figure BDA0002605697650000044
Wherein the content of the first and second substances,
Figure BDA0002605697650000045
as observation vectors
Figure BDA0002605697650000046
Corresponding mixing element xnmIs determined.
Optionally, in some embodiments of the present invention, the step of determining the fault model corresponding to the maximum probability value may include: loading the observation vector sequence into a relevant GMM-HMM model which is trained in advance, wherein lambda is (pi, A, C, mu, U) so as to calculate the probability of the observation vector sequence under the model; judging whether a next relevant GMM-HMM model exists or not; loading the sequence of observation vectors into a next said model in response to there being a next relevant GMM-HMM model to calculate a probability of the sequence of observation vectors being under the next said model; and responding to the GMM-HMM model without the next correlation, and outputting the sequence number of the GMM-HMM model corresponding to the maximum probability value.
Preferably, in some embodiments of the present invention, the step of calculating the probability of the observation vector sequence under a GMM-HMM model may further comprise: probability P [ O | λ for the observation vector sequence]Taking a logarithm to calculate a likelihood probability log of the observation vector sequence10(P[O|λi]) (ii) a And determining the likelihood probability log10(P[O|λi]) As the probability of the observation vector sequence under the GMM-HMM model.
Optionally, in some embodiments of the present invention, the step of outputting the sequence number of the GMM-HMM model corresponding to the maximum probability value may further include: judging whether the maximum probability value is greater than a preset probability threshold value or not; and responding to the maximum probability value being larger than the probability threshold value, and outputting the sequence number of the GMM-HMM model corresponding to the maximum probability value.
Optionally, in some embodiments of the present invention, the step of performing parameter initialization may further include: and initializing the parameters by utilizing a segmented K-means algorithm according to the sample library of the historical fault data. The step of parameter estimating the GMM-HMM model may further comprise: parameter estimation of the GMM-HMM model is accomplished based on the Baum-Welch algorithm.
According to another aspect of the present invention, there is also provided an online diagnosis device for a traction system fault.
The online diagnosis device for the traction system fault comprises a memory and a processor. The processor is connected with the memory and is configured to implement the online diagnosis method for the traction system fault provided by any one of the above embodiments, so as to perform online positioning on the traction system fault and realize accurate classification of various fault sources.
Preferably, in some embodiments of the present invention, the online diagnosis apparatus may include a plurality of the processors. The first processor is a fault detection unit and can be configured to monitor a fault characteristic variable of the traction system on line and output a corresponding fault classification enabling mark in response to an abnormality of the fault characteristic variable. The second processor is a fault classification unit and can be configured to determine a fault diagnosis result of the traction system according to the fault classification enabling mark, the fault characteristic variable and relevant time domain statistics of the traction system.
Preferably, in some embodiments of the present invention, the online diagnosis apparatus may further include a third processor. The third processor may be configured to train various types of fault models for the traction system based on historical fault data for the traction system.
According to another aspect of the present invention, a computer-readable storage medium is also provided herein.
The present invention provides the above computer readable storage medium having stored thereon computer instructions. When the computer instruction is executed by the processor, the method for diagnosing the traction system fault on line provided by any one of the embodiments can be implemented, so that the traction system fault is positioned on line, and accurate classification of various fault sources is realized.
Drawings
The above features and advantages of the present disclosure will be better understood upon reading the detailed description of embodiments of the disclosure in conjunction with the following drawings. In the drawings, components are not necessarily drawn to scale, and components having similar relative characteristics or features may have the same or similar reference numerals.
FIG. 1 illustrates a functional block diagram of a method for online diagnosis of traction system faults provided in accordance with some embodiments of the present invention.
FIG. 2 illustrates a flow diagram of a method for online diagnosis of a traction system fault provided in accordance with an aspect of the present invention.
FIG. 3 illustrates a flow diagram of a method for online diagnosis of traction system faults provided in accordance with some embodiments of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided for illustrative purposes, and other advantages and effects of the present invention will become apparent to those skilled in the art from the present disclosure. While the invention will be described in connection with the preferred embodiments, there is no intent to limit its features to those embodiments. On the contrary, the invention is described in connection with the embodiments for the purpose of covering alternatives or modifications that may be extended based on the claims of the present invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The invention may be practiced without these particulars. Moreover, some of the specific details have been left out of the description in order to avoid obscuring or obscuring the focus of the present invention.
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 either 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. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Additionally, the terms "upper," "lower," "left," "right," "top," "bottom," "horizontal," "vertical" and the like as used in the following description are to be understood as referring to the segment and the associated drawings in the illustrated orientation. The relative terms are used for convenience of description only and do not imply that the described apparatus should be constructed or operated in a particular orientation and therefore should not be construed as limiting the invention.
It will be understood that, although the terms first, second, third, etc. may be used herein to describe various elements, regions, layers and/or sections, these elements, regions, layers and/or sections should not be limited by these terms, but rather are used to distinguish one element, region, layer and/or section from another element, region, layer and/or section. Thus, a first component, region, layer or section discussed below could be termed a second component, region, layer or section without departing from some embodiments of the present invention.
As described above, the fault diagnosis of the train traction system is still mainly based on collecting sensor signals and is performed by adopting fault detection methods such as simple over-threshold alarm. The fault detection method belongs to the detection of fault characteristics, can not diagnose the real reasons of the characteristics, and generally needs to stop temporarily and be checked by a driver or system maintenance personnel.
In order to overcome the defects in the prior art, the invention provides an online diagnosis method for the faults of the traction system, an online diagnosis device for the faults of the traction system and a computer readable storage medium, which are used for carrying out online positioning on the faults of the traction system and realizing accurate classification of various fault sources.
In some embodiments of the present invention, the above-mentioned online diagnosis method for the fault of the traction system may be implemented by a processor of an online diagnosis device for the fault of the traction system. The online diagnostic device includes, but is not limited to, a controller of the traction system and other independent devices having a fault diagnosis function. Specifically, the online diagnosis device for the traction system fault can comprise a memory and a processor. The memory may be a computer-readable storage medium having stored thereon computer instructions. The processor may execute computer instructions stored on the memory to implement the above-described method of online diagnosis of traction system faults.
Referring to fig. 1, fig. 1 is a schematic block diagram illustrating a method for online diagnosis of traction system faults according to some embodiments of the present invention.
As shown in FIG. 1, in some non-limiting embodiments, the method for online diagnosis of traction system faults can be implemented by dividing into an offline design and an online implementation. Before the online diagnosis of the traction system fault is carried out, the processor of the online diagnosis device can firstly carry out fault feature analysis on historical fault case data of the traction system in an offline design stage, extract relevant feature variables and calculate feature indexes to obtain a sample library of observation vector sequences of the fault feature indexes. Then, the processor can train GMM-HMM model parameters of various faults based on the sample library so as to be used for carrying out online positioning on the traction system faults and realizing accurate classification of various fault sources.
It will be appreciated that the HMM model described above is a hidden markov model, suitable for describing the temporal characteristics of a continuous signal that lasts for a certain period. The GMM-HMM model is a Gaussian-hidden Markov mixed model and is suitable for describing the probability of time sequence characteristics. In some embodiments, the offline design portion may be implemented by an overall processor of the online diagnostic device. Alternatively, in other embodiments, the online diagnostic apparatus may also be configured with a dedicated sub-processor module to implement the offline design portion.
Specifically, in some embodiments, the processor may employ a GMM-HMM algorithm to model a time series of indicators of a fault characteristic. Assuming a sample library with K time series characteristic observation vector sequences, the definition is shown in formula (1):
O＝{O(1),O(2),…,O(K)} (1)
in the formula:
Figure BDA0002605697650000071
for the k-th sequence of observation vectors,
Figure BDA0002605697650000072
is O(k)D-dimensional observation vector at time t.
To build an HMM model describing the statistical features of a sequence of fault feature indicator vectors, the processor may select N states and M gaussian mixture elements as follows:
S＝{s1,s2,…,sN} (2)
Figure BDA0002605697650000081
wherein x isnmIs a state snThe mth mixing element of (1).
As such, the GMM-HMM model may be described as:
λ＝(π,A,C,μ,U) (4)
wherein, pi is the probability distribution of the initial state, A is the probability distribution of the state transition, C is the weight of the mixed right, mu is the mean vector, and U is the covariance matrix.
Suppose qtFor the state of the observation vector sequence at time t and P [ ·]The probability of the observation vector sequence is used as the parameter of the GMM-HMM modelMeaning as follows:
π＝[π1,π1,…,πN]T (5)
wherein, pin＝P[q1＝sn]Is not less than 0 and
Figure BDA0002605697650000082
Figure BDA0002605697650000083
wherein, aij＝P[qt+1＝sj|qt＝si]Not less than 0 and all i satisfy
Figure BDA0002605697650000084
Figure BDA0002605697650000085
Wherein, cnmNot less than 0 and all n satisfy
Figure BDA0002605697650000086
Figure BDA0002605697650000091
Figure BDA0002605697650000092
Wherein the content of the first and second substances,
Figure BDA0002605697650000093
and UnmRespectively represent mixed elements xnmD-dimensional mean vector and D x D-dimensional covariance matrix. State snIs observed vector
Figure BDA0002605697650000094
Probability Density Function (PDF)
Figure BDA0002605697650000095
Can be described as:
Figure BDA0002605697650000096
in the formula (I), the compound is shown in the specification,
Figure BDA0002605697650000097
as observation vectors
Figure BDA0002605697650000098
Corresponding mixing element xnmIs determined.
In some embodiments, after the fault signature indicators are determined, the observation vector dimension D is determined accordingly. Given the number of states N and the number of mixture elements M of the GMM-HMM model, the processor may perform parameter initialization using a segmented K-means algorithm based on a sample library of historical case data and perform GMM-HMM model parameter estimation based on a Baum-Welch algorithm.
After the off-line design part is completed, the on-line diagnosis device can be applied to on-line diagnosis of the traction system fault, and is used for carrying out on-line positioning on the traction system fault and realizing accurate classification of various fault sources. In some embodiments, the online implementation part of the online diagnosis method may be implemented by the mutual cooperation of the fault detection unit and the fault classification unit of the online diagnosis device. In some embodiments, the fault detection unit and the fault classification unit may be implemented by two processors communicatively connected to each other. Alternatively, in other embodiments, the fault detection unit and the fault classification unit may be implemented by the same processor executing two software modules respectively.
Referring to fig. 2, fig. 2 is a flow chart illustrating an online diagnosis method for a traction system fault according to an aspect of the present invention.
As shown in fig. 2, the method for online diagnosing a fault of a traction system according to the present invention may include step 201: fault signature variables of the traction system are monitored.
In some embodiments of the invention, the online diagnostic of traction system faults may further localize the fault based on the fault detection of the threshold overrun alarm. Specifically, when the traction system is abnormal, the related sensor acquisition signal of the traction system is instantaneously abnormal. The online diagnosis device can judge whether the traction system is abnormal or not according to the sensor signals acquired by the sensor.
In some embodiments, the invention can select the time domain statistics of the relevant sensor signals strongly related to various threshold over-limit alarm fault detection results as fault characteristic indexes according to engineering application experience. Taking a common primary side overcurrent fault of a traction transformer in a traction system as an example, the primary side voltage signal and the primary side current signal have the maximum relevance with various fault sources of the primary side overcurrent. Therefore, the invention can select the primary side voltage signal and the primary side current signal as the fault characteristic variable of the primary side overcurrent fault of the traction transformer. The processor can monitor the primary voltage signal and the primary current signal of the traction system to judge whether the traction system is abnormal.
As shown in fig. 1, the fault detection unit may first monitor the voltage signal of the primary side of the traction transformer in real time by using the voltage sensor disposed on the primary side of the traction transformer, and monitor the current signal of the primary side of the traction transformer in real time by using the current sensor disposed on the primary side of the traction transformer. Thereafter, the fault detection unit may compare the acquired voltage signal and current signal with a given voltage threshold and current threshold, respectively.
If the acquired voltage signal exceeds a given voltage threshold, the fault detection unit may determine that a fault characteristic variable of the voltage is abnormal. On the contrary, if the acquired voltage signal does not exceed the given voltage threshold, the fault detection unit may determine that the fault characteristic variable of the voltage is not abnormal.
Accordingly, if the acquired current signal exceeds a given current threshold, the fault detection unit may determine that the fault characteristic variable of the current is abnormal. Otherwise, the fault detection unit may determine that the fault characteristic variable of the current is not abnormal.
As shown in fig. 2, the method for online diagnosing a fault of the traction system according to the present invention may further include step 202: and outputting a corresponding fault classification enabling mark in response to the fault characteristic variable abnormity.
In the above embodiment, in response to determining that any fault characteristic variable is abnormal, the fault detection unit may generate a fault classification enable flag lasting for a certain period to the fault classification unit. The fault classification enabling mark can last for a plurality of power frequency cycles and is used for identifying abnormal fault characteristic variables.
As shown in fig. 2, the method for online diagnosing a fault of a traction system according to the present invention may further include step 203: and collecting relevant time domain statistics of the traction system according to the fault classification enabling mark.
In the above embodiment, in response to the fault classification enable flag, the fault classification unit may obtain the fault feature variable from the fault detection unit and obtain other relevant time domain statistics from the corresponding sensor according to the received fault classification enable flag. Specifically, taking the primary side overcurrent fault of the traction transformer as an example, the fault classification unit may collect time domain statistics such as the maximum peak value, the minimum peak value, the high amplitude duration, the low amplitude duration, the minimum peak change rate, and the like of the primary side voltage and the primary side current in the period sliding window, so as to serve as other time domain statistics related to the primary side overcurrent fault of the traction transformer, and to implement accurate classification of different fault sources of the primary side overcurrent fault of the traction transformer.
As shown in fig. 2, the method for online diagnosing a fault of the traction system according to the present invention may further include step 204: relevant feature indicators are extracted from the relevant time domain statistics to generate a sequence of observation vectors.
In some embodiments of the invention, the fault classification unit may classify the maximum peak value, the minimum peak value, the high amplitude duration, the low amplitude duration, the minimum peak rate of change, etc. of the primary voltage and the primary current within the periodically sliding window as a time domain statistic,the fault characteristic index of the primary side overcurrent fault of the traction transformer is defined. Then, the fault classification unit can carry out preprocessing such as normalization on the fault characteristic variables such as the primary voltage and the primary current and other related time domain statistics related to the abnormal fault characteristic variables, so as to comprehensively calculate the related characteristic indexes and generate a corresponding observation vector sequence
Figure BDA0002605697650000111
As shown in fig. 2, the method for online diagnosing a fault of a traction system according to the present invention may further include step 205: and loading the observation vector sequence into a pre-trained related fault model to determine the fault model corresponding to the maximum probability value as a fault diagnosis result of the traction system.
In some embodiments of the invention, the fault classification unit may determine a plurality of fault models associated with the detected fault signature variable anomaly from the fault classification enable flags. These fault models may be GMM-HMM models that were pre-trained in the offline design section described above. Thereafter, the fault classification unit may implement the partially generated observation vector sequence on-line
Figure BDA0002605697650000112
Loading the relevant fault models lambda (pi, A, C, mu and U) trained in advance one by one to determine the probability P [ O | lambda ] of the observation vector sequence under each fault model]Thereby comprehensively deciding the specific fault type of the traction system.
Referring to fig. 3, fig. 3 is a flow chart illustrating a method for online diagnosis of a traction system fault according to some embodiments of the present invention.
As shown in FIG. 3, in some embodiments of the invention, a plurality of pre-trained GMM-HMM models are associated in response to a fault classification enable flag. The fault classification unit can firstly observe the vector sequence
Figure BDA0002605697650000113
Load the first of these pre-trained related GMM-HMM models λTo calculate the probability P [ O | λ ] of the observation vector sequence under this first model]。
Taking into account the probability P [ O | λ]Usually some very small number, the likelihood estimates are multiplied to avoid underflow problems. In some preferred embodiments, the processor may compare the probability P [ O | λ [ ]]Performing a logarithmic operation to calculate a sequence of observation vectors
Figure BDA0002605697650000121
Likelihood probability value log of10(P[O|λi]) And the likelihood probability value log is determined10(PpO|λi]) As a sequence of observation vectors
Figure BDA0002605697650000122
And (3) a probability value under the first GMM-HMM model, so that the problem of data underflow is effectively prevented to improve the accuracy of fault diagnosis.
After the probability calculation of the first model is completed, the fault classification unit may determine whether there are any other relevant GMM-HMM models. The fault classification unit may further classify the sequence of observation vectors in response to still other related GMM-HMM models
Figure BDA0002605697650000123
Loading a second pre-trained correlated GMM-HMM model λ (π, A, C, μ, U) to compute the probability P [ O | λ ] of the observation vector sequence under the second model]. And so on until the probability calculation of all relevant GMM-HMM models is completed.
Thereafter, the fault classification unit may compare the observation vector sequences under the GMM-HMM models
Figure BDA0002605697650000124
Likelihood probability value log of10(P[O|λi]) To determine the maximum probability value therein. In some preferred embodiments, the fault classification unit may further associate the maximum probability value with a given probability threshold PthA comparison is made.
If the maximum probability value is greater than a given probability threshold value PthThe fault classification unit may output the sequence number of the GMM-HMM model corresponding to the maximum probability value to indicate the type of fault occurring in the traction system. Otherwise, if the maximum probability value is smaller than the probability threshold value PthIt means that the probability of various types of faults occurring in the traction system is small, and therefore it is not necessary to output a fault diagnosis result indicating any fault type.
As can be understood from the above description, the invention utilizes the fault characteristic index analysis and the time sequence modeling method to establish a detailed GMM-HMM model for various fault types which may occur in the traction system. When the traction system fault is diagnosed on line, the method can predict various possibly related fault types based on a small number of sensor signal abnormalities, and performs characteristic index calculation by collecting related sensor information to determine probability values of various fault types, so that various fault sources of the traction system are accurately positioned on line. In addition, the method does not need to change the existing hardware of the train traction system, and can be realized only by a software means, so the method is easy to realize in engineering.
Furthermore, through the online accurate positioning of various fault sources of the traction system, the differential protection isolation action can be executed based on specific fault points, so that the train availability is improved.
While, for purposes of simplicity of explanation, the methodologies are shown and described as a series of acts, it is to be understood and appreciated that the methodologies are not limited by the order of acts, as some acts may, in accordance with one or more embodiments, occur in different orders and/or concurrently with other acts from that shown and described herein or not shown and described herein, as would be understood by one skilled in the art.
Although the processor, the fault detection unit and the fault classification unit described in the above embodiments may be implemented by a combination of software and hardware. It is to be understood that these processors, fault detection unit and fault classification unit may also be implemented in software or hardware alone. For a hardware implementation, the processor, the fault detection unit, and the fault classification unit may be implemented in one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), digital signal processing devices (DAPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, other electronic devices configured to perform the functions described herein, or a selected combination thereof. For software implementations, the processor, fault detection unit, and fault classification unit may be implemented by separate software modules running on a common chip, such as program modules (processes) and function modules (functions), each of which may perform one or more of the functions and operations described herein.
The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (17)

1. An online diagnosis method for a traction system fault is characterized by comprising the following steps:
monitoring a fault characteristic variable of the traction system;
outputting a corresponding fault classification enabling flag in response to the fault characteristic variable being abnormal;
collecting relevant time domain statistics of the traction system according to the fault classification enabling mark;
extracting a correlation feature indicator from the correlation time domain statistic to generate a sequence of observation vectors; and
and loading the observation vector sequence into a pre-trained related fault model to determine the fault model corresponding to the maximum probability value as a fault diagnosis result of the traction system.
2. The online diagnostic method of claim 1, further comprising:
judging whether the fault characteristic variable exceeds a given threshold value or not; and
and judging that the fault characteristic variable is abnormal in response to the fault characteristic variable exceeding the threshold value.
3. The online diagnostic method of claim 1, wherein the step of collecting the relevant time domain statistics based on the fault classification enable flag comprises:
the fault signature variables are obtained in response to the fault classification enable flags, and other relevant time domain statistics are obtained from the corresponding sensors according to the fault classification enable flags.
4. The online diagnostic method of claim 1, further comprising:
performing fault feature analysis on historical fault data of the traction system to extract relevant feature variables;
calculating a fault characteristic index according to the relevant characteristic variable to obtain a sample library of an observation vector sequence of the fault characteristic index; and
and training various fault models of the traction system according to the sample library.
5. The online diagnostic method of claim 4, wherein the step of obtaining a sample library of the sequence of observation vectors further comprises:
defining a sample library having K of the observation vector sequences as O ═ { O ═ O(1),O(2),…,O(K)And (c) the step of (c) in which,
Figure FDA0002605697640000021
for the k-th sequence of observation vectors,
Figure FDA0002605697640000022
is O(k)D-dimensional observation vector at time t.
6. The online diagnostic method of claim 5, wherein the step of training the various types of fault models comprises:
establishing an HMM model describing statistical characteristics of the observation vector sequence;
establishing a GMM-HMM model according to the HMM model;
determining a probability density function of the observation vector sequence according to the GMM-HMM model; and
and initializing parameters according to the sample library of the historical fault data, and performing parameter estimation on the GMM-HMM model.
7. The online diagnostic method of claim 6, wherein the step of building the HMM model further comprises:
selecting N states S ═ S1,s2,…,sNAnd M Gaussian mixture elements
Figure FDA0002605697640000023
To build the HMM model, wherein xnmIs a state snThe mth mixing element of (1).
8. The online diagnostic method of claim 7, wherein the step of building the GMM-HMM model further comprises:
describing the GMM-HMM model as λ ═ (π, A, C, μ, U), where π is the initial state probability distribution, A is the state transition probability distribution, C is the mixed right weight, μ is the mean vector, and U is the covariance matrix;
defining the initial state probability distribution pi as pi ═ pi1,π1,…,πN]TWherein q istFor the state of the observation vector sequence at time t, P [ ·]Is the probability, pi, of the observation vector sequencen＝P[q1＝sn]Is not less than 0 and
Figure FDA0002605697640000024
defining the state transition probability distribution A as
Figure FDA0002605697640000031
Wherein, aij＝P[qt+1＝sj|qt＝si]Not less than 0 and all i satisfy
Figure FDA0002605697640000032
Defining the mixing right C as
Figure FDA0002605697640000033
Wherein, cnmNot less than 0 and all n satisfy
Figure FDA0002605697640000034
Defining the mean vector μ as
Figure FDA0002605697640000035
Wherein the content of the first and second substances,
Figure FDA0002605697640000036
represents a mixing element xnmD-dimensional mean vector of (1); and
defining the covariance matrix U as
Figure FDA0002605697640000037
Wherein, UnmRepresents a mixing element xnmD x D dimensional covariance matrix.
9. The online diagnostic method of claim 8, wherein the step of determining the probability density function of the sequence of observation vectors further comprises:
will state snIs observed vector
Figure FDA0002605697640000038
Probability density function of
Figure FDA0002605697640000039
Is described as
Figure FDA00026056976400000310
Wherein the content of the first and second substances,
Figure FDA00026056976400000311
as observation vectors
Figure FDA00026056976400000312
Corresponding mixing element xnmIs determined.
10. The online diagnostic method of claim 6, wherein the step of determining the fault model corresponding to the maximum probability value comprises:
loading the observation vector sequence into a relevant GMM-HMM model which is trained in advance so as to calculate the probability of the observation vector sequence under the model;
judging whether a next relevant GMM-HMM model exists or not;
loading the sequence of observation vectors into a next said model in response to there being a next relevant GMM-HMM model to calculate a probability of the sequence of observation vectors being under the next said model; and
and responding to the GMM-HMM model without the next correlation, and outputting the sequence number of the GMM-HMM model corresponding to the maximum probability value.
11. The online diagnostic method of claim 10, wherein the step of calculating the probability of the sequence of observation vectors under a GMM-HMM model further comprises:
probability P [ O | λ for the observation vector sequence]Taking a logarithm to calculate a likelihood probability log of the observation vector sequence10(P[O|λi]) (ii) a And
comparing the likelihood probability log10(P[O|λi]) As the probability of the observation vector sequence under the GMM-HMM model.
12. The online diagnosis method as claimed in claim 10, wherein the step of outputting the sequence number of the GMM-HMM model corresponding to the maximum probability value further comprises:
judging whether the maximum probability value is greater than a preset probability threshold value or not; and
and outputting the sequence number of the GMM-HMM model corresponding to the maximum probability value in response to the maximum probability value being larger than the probability threshold.
13. The online diagnostic method of claim 6, wherein the step of performing parameter initialization further comprises: initializing the parameters by utilizing a segmented K-means algorithm according to the sample library of the historical fault data,
the step of performing parameter estimation on the GMM-HMM model further comprises: parameter estimation of the GMM-HMM model is accomplished based on the Baum-Welch algorithm.
14. An online diagnosis device for a traction system fault, characterized by comprising a memory and a processor, wherein the processor is connected with the memory and is configured to implement the online diagnosis method of any one of claims 1 to 13.
15. The online diagnostic apparatus of claim 14, comprising a plurality of said processors, wherein,
the first processor is a fault detection unit and is configured to monitor a fault characteristic variable of the traction system on line and output a corresponding fault classification enabling mark in response to the fault characteristic variable being abnormal,
and the second processor is a fault classification unit and is configured to determine a fault diagnosis result of the traction system according to the fault classification enabling mark, the fault characteristic variable and the relevant time domain statistic of the traction system.
16. The online diagnostic apparatus of claim 15, further comprising a third processor configured to train classes of fault models for the traction system based on historical fault data for the traction system.
17. A computer readable storage medium having stored thereon computer instructions, which when executed by a processor, perform the online diagnostic method of any one of claims 1 to 13.
CN202010738088.2A 2020-07-28 2020-07-28 Online diagnosis method and device for traction system fault Pending CN114004268A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010738088.2A CN114004268A (en) 2020-07-28 2020-07-28 Online diagnosis method and device for traction system fault

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010738088.2A CN114004268A (en) 2020-07-28 2020-07-28 Online diagnosis method and device for traction system fault

Publications (1)

Publication Number Publication Date
CN114004268A true CN114004268A (en) 2022-02-01

Family

ID=79920388

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010738088.2A Pending CN114004268A (en) 2020-07-28 2020-07-28 Online diagnosis method and device for traction system fault

Country Status (1)

Country Link
CN (1) CN114004268A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116165473A (en) * 2023-04-26 2023-05-26 广东工业大学 Real-time tracing method for network side overcurrent faults of train traction transmission system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116165473A (en) * 2023-04-26 2023-05-26 广东工业大学 Real-time tracing method for network side overcurrent faults of train traction transmission system

Similar Documents

Publication Publication Date Title
CN109186813B (en) Temperature sensor self-checking device and method
CN108984893B (en) Gradient lifting method-based trend prediction method
CN105700518B (en) A kind of industrial process method for diagnosing faults
EP1416348B1 (en) Methodology for temporal fault event isolation and identification
EP2989705B1 (en) Method and apparatus for defect pre-warning of power device
CN109000940B (en) Abnormal axle temperature diagnosis method and system for rolling stock
CN110414154B (en) Fan component temperature abnormity detection and alarm method with double measuring points
US11669080B2 (en) Abnormality detection device, abnormality detection method, and program
CN108304348B (en) Bearing residual life prediction method based on binary wiener process
CN110704911B (en) Bridge damage cross positioning method based on cluster structure similarity
CN109917777B (en) Fault detection method based on mixed multi-sampling rate probability principal component analysis model
CN112036505A (en) Method and device for determining equipment state of turnout switch machine and electronic equipment
CN112581719B (en) Semiconductor packaging process early warning method and device based on time sequence generation countermeasure network
CN111796233B (en) Method for evaluating secondary errors of multiple voltage transformers in double-bus connection mode
CN110990788A (en) Bearing residual life prediction method based on ternary wiener process
CN112036581A (en) Performance detection method and device of vehicle air conditioning system, storage medium and terminal
CN110469461B (en) Fracture estimation method and device for fan toothed belt and readable storage medium
CN114004268A (en) Online diagnosis method and device for traction system fault
CN111046331A (en) Health monitoring method based on signal abnormity
JP2016045852A (en) Abnormality diagnostic device and abnormality diagnostic method
CN112016208B (en) Hidden fault diagnosis method and system considering disturbance
CN112733446A (en) Data-driven self-adaptive anomaly detection method
CN112231849B (en) Axle temperature fault detection method based on NEST and SPRT fusion algorithm
CN106462908B (en) Maintenance management index calculation device and maintenance management index calculation method
CN111474476A (en) Motor fault prediction method

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