A method and device for online diagnosis of traction system faults

Abstract

The present invention relates to an online diagnosis method and device for traction system faults, and a computer-readable storage medium. The online diagnosis method comprises: monitoring the fault characteristic variables of the traction system; outputting a corresponding fault classification enable flag in response to the abnormality of the fault characteristic variables; collecting relevant time domain statistics of the traction system according to the fault classification enable flag; extracting relevant characteristic indicators from the relevant time domain statistics to generate an observation vector sequence; and loading the observation vector sequence into a pre-trained relevant fault model to determine that the fault model corresponding to the maximum probability value is the fault diagnosis result of the traction system. The present invention can locate the traction system fault online and realize the accurate classification of various fault sources.

Description

Technical Field

[0001] The invention relates to a fault diagnosis technology for a rail transit traction system, and in particular to an online diagnosis method for a traction system fault, and an online diagnosis device for a traction system fault. Background Art

[0002] During the operation of locomotives, EMUs and other trains, any minor or potential faults and hidden dangers may trigger a chain reaction and cause accidents, or even lead to catastrophic consequences if they cannot be diagnosed and effectively dealt with in a timely manner. As the "heart" of high-speed trains, the traction system is prone to failures due to its complex operating environment, corrosion, temperature, humidity, power surges, static electricity and other factors that affect its operating status, and cannot be eliminated through regular maintenance. If a train fails during operation, it is best to accurately locate the fault source online so that the fault can be eliminated in time or an appropriate isolation protection strategy can be implemented. If the cause of the fault is not diagnosed and eliminated in time, it will cause traffic accidents, delay the normal operation of the train, and affect the transportation order of the entire line and even the entire railway.

[0003] At present, the fault diagnosis of train traction systems is still mainly based on the acquisition of sensor signals, using simple over-threshold alarm and other fault detection methods, such as: overvoltage and overcurrent detection on the grid side of the traction system; input / output overcurrent detection of the traction converter, intermediate DC overvoltage / undervoltage detection, cooling system temperature, water pressure too high / too low and other detection and protection functions. However, such detection methods are all fault symptom detection, and cannot diagnose the real cause of such symptoms. Generally, it is necessary to temporarily stop the train for the driver or system maintenance personnel to check.

[0004] In order to overcome the above-mentioned defects of the prior art, there is an urgent need in the art for an online fault diagnosis technology of a traction system, which is used to locate the traction system fault online and realize the accurate classification of various fault sources. Summary of the invention

[0005] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or decisive elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0006] In order to overcome the above-mentioned defects in the prior art, the present invention provides an online diagnosis method for traction system faults, an online diagnosis device for traction system faults, and a computer-readable storage medium for online positioning of traction system faults and accurate classification of various fault sources.

[0007] The online diagnosis method for the above-mentioned traction system fault provided by the present invention includes: monitoring the fault characteristic variables of the traction system; outputting a corresponding fault classification enable flag in response to the abnormality of the fault characteristic variables; collecting relevant time domain statistics of the traction system according to the fault classification enable flag; extracting relevant characteristic indicators from the relevant time domain statistics to generate an observation vector sequence; and loading the observation vector sequence into a pre-trained relevant fault model to determine that the fault model corresponding to the maximum probability value is the fault diagnosis result of the traction system.

[0008] Optionally, in some embodiments of the present invention, the online diagnosis method may further include: determining whether the fault characteristic variable exceeds a given threshold; and determining that the fault characteristic variable is abnormal in response to the fault characteristic variable exceeding the threshold.

[0009] Optionally, in some embodiments of the present invention, the step of collecting the relevant time domain statistics according to the fault classification enable flag may include: obtaining the fault characteristic variable in response to the fault classification enable flag, and obtaining other relevant time domain statistics from the corresponding sensor according to the fault classification enable flag.

[0010] Preferably, in some embodiments of the present invention, the online diagnosis method may further include: performing fault feature analysis on the historical fault data of the traction system to extract relevant feature variables; calculating fault feature indicators based on the relevant feature variables to obtain a sample library of observation vector sequences of the fault feature indicators; and training various types of fault models of the traction system based on the sample library.

[0011] Preferably, in some embodiments of the present invention, the step of obtaining the sample library of the observation vector sequence may further include: defining the sample library having K observation vector sequences as O={O (1) ,O (2) ,…,O (K)},in, is the kth observation vector sequence, O (k) The D-dimensional observation vector at the tth moment.

[0012] Preferably, in some embodiments of the present invention, the step of training each type of fault model may include: establishing an HMM model that describes the statistical characteristics of the observation vector sequence; establishing a GMM-HMM model based on the HMM model; determining the probability density function of the observation vector sequence based on the GMM-HMM model; and initializing parameters based on a sample library of historical fault data and estimating parameters of the GMM-HMM model.

[0013] Preferably, in some embodiments of the present invention, the step of establishing the HMM model may further include: selecting N states S = {s 1 ,s 2 ,…,s N} and M Gaussian mixture elements To establish the HMM model, where x nm For state s n The mth mixing element of .

[0014] Preferably, in some embodiments of the present invention, the step of establishing the GMM-HMM model may further include: describing the GMM-HMM model as λ=(π,A,C,μ,U), wherein π is the initial state probability distribution, A is the state transition probability distribution, C is the mixed weight, μ is the mean vector, and U is the covariance matrix; defining the initial state probability distribution π as π=[π 1 ,π 1 ,…,π N ] T , where q t is the state of the observation vector sequence at time t, P[·] is the probability of the observation vector sequence, π n =P[q 1 =s n ]≥0 and The state transition probability distribution A is defined as Among them, a ij =P[q t+1 =s j |q t =s i ]≥0 and all i satisfy The mixed weight C is defined as Among them, c nm ≥0 and all n satisfy The mean vector μ is defined as in, Represents mixed element x nm The D-dimensional mean vector of Among them, U nm Represents mixed element x nm The D×D dimensional covariance matrix of .

[0015] Preferably, in some embodiments of the present invention, the step of determining the probability density function of the observation vector sequence may further include: transforming the state s n The observation vector The probability density function of Described as in, is the observation vector The corresponding mixed element x nm The probability density function of .

[0016] 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 pre-trained related GMM-HMM model λ=(π, A, C, μ, U) to calculate the probability of the observation vector sequence under the model; determining whether there is a next related GMM-HMM model; in response to there being a next related GMM-HMM model, loading the observation vector sequence into the next model to calculate the probability of the observation vector sequence under the next model; and in response to there being no next related GMM-HMM model, outputting the serial number of the GMM-HMM model corresponding to the maximum probability value.

[0017] 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 include: taking the logarithm of the probability P[O|λ] of the observation vector sequence to calculate the likelihood probability log of the observation vector sequence 10 (P[O|λ i ]); and the likelihood probability log 10 (P[O|λ i ]) as the probability of the observation vector sequence under the GMM-HMM model.

[0018] Optionally, in some embodiments of the present invention, the step of outputting the serial number of the GMM-HMM model corresponding to the maximum probability value may further include: determining whether the maximum probability value is greater than a preset probability threshold; and in response to the maximum probability value being greater than the probability threshold, outputting the serial number of the GMM-HMM model corresponding to the maximum probability value.

[0019] Optionally, in some embodiments of the present invention, the step of initializing parameters may further include: initializing the parameters using a segmented K-means algorithm based on a sample library of historical fault data. The step of estimating parameters of the GMM-HMM model may further include: completing parameter estimation of the GMM-HMM model based on a Baum-Welch algorithm.

[0020] According to another aspect of the present invention, an online diagnosis device for traction system faults is also provided.

[0021] The online diagnosis device for traction system faults provided by the present invention comprises a memory and a processor. The processor is connected to the memory and configured to implement the online diagnosis method for traction system faults provided by any one of the above embodiments, thereby locating the traction system faults online and achieving accurate classification of various fault sources.

[0022] Preferably, in some embodiments of the present invention, the online diagnostic device may include a plurality of the processors. The first processor is a fault detection unit, which may be configured to monitor the fault characteristic variables of the traction system online, and output a corresponding fault classification enable flag in response to an abnormality of the fault characteristic variables. The second processor is a fault classification unit, which may be configured to determine the fault diagnosis result of the traction system according to the fault classification enable flag, the fault characteristic variables and the relevant time domain statistics of the traction system.

[0023] Preferably, in some embodiments of the present invention, the online diagnosis device may further include a third processor. The third processor may be configured to train various fault models of the traction system according to historical fault data of the traction system.

[0024] According to another aspect of the present invention, a computer-readable storage medium is also provided herein.

[0025] The computer-readable storage medium provided by the present invention stores computer instructions thereon. When the computer instructions are executed by a processor, the online diagnosis method for traction system faults provided by any of the above embodiments can be implemented, thereby locating the traction system faults online and achieving accurate classification of various fault sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0027] Figure 1 A principle block diagram of an online diagnosis method for traction system faults provided according to some embodiments of the present invention is shown.

[0028] Figure 2 A schematic flow chart of an online diagnosis method for traction system faults provided according to one aspect of the present invention is shown.

[0029] Figure 3 A schematic flow chart of an online diagnosis method for traction system faults provided according to some embodiments of the present invention is shown. DETAILED DESCRIPTION

[0030] The following specific embodiments illustrate the implementation of the present invention, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to this implementation. On the contrary, the purpose of introducing the invention in conjunction with the implementation is to cover other options or modifications that may be extended based on the claims of the present invention. In order to provide a deep understanding of the present invention, the following description will include many specific details. The present invention can also be implemented without using these details. In addition, in order to avoid confusion or blurring the focus of the present invention, some specific details will be omitted in the description.

[0031] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0032] In addition, the terms "upper", "lower", "left", "right", "top", "bottom", "horizontal" and "vertical" used in the following description should be understood as the directions shown in the paragraph and the related drawings. Such relative terms are only used for the convenience of description and do not mean that the device described therein must be manufactured or operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0033] It is understood that although the terms "first", "second", "third", etc. may be used herein to describe various components, regions, layers and / or parts, these components, regions, layers and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or parts. Therefore, the first component, region, layer and / or part discussed below may be referred to as a second component, region, layer and / or part without departing from some embodiments of the present invention.

[0034] As mentioned above, the current fault diagnosis of train traction systems is still mainly based on collecting sensor signals and using simple over-threshold alarm and other fault detection methods. This fault detection method belongs to the detection of fault symptoms, and cannot diagnose the real cause of such symptoms. Generally, it is necessary to temporarily stop the train for the driver or system maintenance personnel to check.

[0035] In order to overcome the above-mentioned defects in the prior art, the present invention provides an online diagnosis method for traction system faults, an online diagnosis device for traction system faults, and a computer-readable storage medium for online positioning of traction system faults and accurate classification of various fault sources.

[0036] In some embodiments of the present invention, the online diagnosis method of the traction system fault can be implemented by a processor of an online diagnosis device of the traction system fault. The online diagnosis device includes but is not limited to a controller of the traction system and other independent devices with a fault diagnosis function. Specifically, the online diagnosis device of the traction system fault can include a memory and a processor. The memory can be a computer-readable storage medium on which computer instructions are stored. The processor can execute the computer instructions stored in the memory to implement the online diagnosis method of the traction system fault.

[0037] Please refer to Figure 1 , Figure 1 A principle block diagram of an online diagnosis method for traction system faults provided according to some embodiments of the present invention is shown.

[0038] like Figure 1 As shown, in some non-limiting embodiments, the online diagnosis method of traction system faults can be implemented in two parts: offline design and online implementation. Before performing online diagnosis of traction system faults, the processor of the online diagnosis device can first perform fault feature analysis on the historical fault case data of the traction system in the offline design stage, extract relevant feature variables and calculate feature indicators to obtain a sample library of observation vector sequences of fault feature indicators. Afterwards, the processor can train GMM-HMM model parameters of various faults based on this sample library to locate traction system faults online and achieve accurate classification of various fault sources.

[0039] It can be understood that the above HMM model is a hidden Markov model, which is suitable for describing the timing characteristics of continuous signals that last for a certain period. The above GMM-HMM model is a Gaussian-hidden Markov mixed model, which is suitable for describing the probability of timing characteristics. In some embodiments, the above offline design part can be implemented by the main processor of the online diagnostic device. Optionally, in other embodiments, the online diagnostic device can also be configured with a dedicated sub-processor module to implement the above offline design part.

[0040] Specifically, in some embodiments, the processor may use the GMM-HMM algorithm to model the time series of the fault characteristic index. Assuming a sample library with K time series characteristic observation vector sequences, the definition is as shown in formula (1):

[0041] O={O (1) ,O (2),…,O (K)} (1)

[0042] Where: is the kth observation vector sequence, O (k) The D-dimensional observation vector at the tth moment.

[0043] In order to establish an HMM model that describes the statistical characteristics of the fault feature indicator vector sequence, the processor can select N states and M Gaussian mixture elements as follows:

[0044] S={s 1 ,s 2 ,…,s N} (2)

[0045]

[0046] Among them, x nm For state s n The mth mixing element of .

[0047] Thus, the GMM-HMM model can be described as:

[0048] λ=(π,A,C,μ,U) (4)

[0049] Among them, π is the initial state probability distribution, A is the state transition probability distribution, C is the mixing weight, μ is the mean vector, and U is the covariance matrix.

[0050] Assume q t is the state of the observation vector sequence at time t and P[·] is the probability of the observation vector sequence, then the parameters of the GMM-HMM model can be defined as:

[0051] π=[π 1 ,π 1 ,…,π N ] T (5)

[0052] Among them, π n =P[q 1 =s n ]≥0 and

[0053]

[0054] Among them, a ij =P[q t+1 =s j |q t =s i ]≥0 and all i satisfy

[0055]

[0056] Among them, c nm ≥0 and all n satisfy

[0057]

[0058]

[0059] in, and U nm Represents the mixed element x nm The D-dimensional mean vector and D×D-dimensional covariance matrix of state s n The observation vector Probability Density Function (PDF) It can be described as:

[0060]

[0061] In the formula, is the observation vector The corresponding mixed element x nm The probability density function of .

[0062] In some embodiments, after the fault characteristic index is determined, the observation vector dimension D is determined accordingly. Given the number of states N and the number of mixed elements M of the GMM-HMM model, the processor can use the segmented K-means algorithm to initialize parameters based on the sample library of historical case data, and complete the GMM-HMM model parameter estimation based on the Baum-Welch algorithm.

[0063] After completing the above-mentioned offline design part, the online diagnostic device can be applied to the online diagnosis of traction system faults, for online positioning of traction system faults, and accurate classification of various fault sources. In some embodiments, the online implementation part of the online diagnostic method can be realized by the cooperation between the fault detection unit and the fault classification unit of the online diagnostic device. In some embodiments, the fault detection unit and the fault classification unit can be realized by two processors that are connected to each other in communication. Optionally, in other embodiments, the fault detection unit and the fault classification unit can also be realized by the same processor executing two software modules respectively.

[0064] Please refer to Figure 2 , Figure 2 A schematic flow chart of an online diagnosis method for traction system faults provided according to one aspect of the present invention is shown.

[0065] like Figure 2 As shown, the online diagnosis method for the traction system fault provided by the present invention may include step 201: monitoring the fault characteristic variables of the traction system.

[0066] In some embodiments of the present invention, the online diagnostic device for traction system faults can further locate the fault based on the fault detection result of the threshold over-limit alarm. Specifically, when the traction system is abnormal, its related sensor acquisition signal will be transiently abnormal. The online diagnostic device can determine whether the traction system is abnormal based on the sensor signal collected by the sensor.

[0067] In some embodiments, the present invention can select the time domain statistics of the relevant sensor signals that are strongly correlated with the detection results of various types of threshold over-limit alarm faults as fault characteristic indicators based on engineering application experience. Taking the common primary overcurrent fault of the traction transformer in the traction system as an example, the primary voltage signal and the primary current signal are the most correlated with various types of primary overcurrent fault sources. Therefore, the present invention can select the primary voltage signal and the primary current signal as the fault characteristic variables of the primary overcurrent fault of the traction transformer. The processor can monitor the primary voltage signal and the primary current signal of the traction system to determine whether an abnormality occurs in the traction system.

[0068] like Figure 1 As shown, the fault detection unit can first use the voltage sensor provided at the primary side of the traction transformer to monitor the voltage signal of the primary side of the traction transformer in real time, and use the current sensor provided at the primary side of the traction transformer to monitor the current signal of the primary side of the traction transformer in real time. Afterwards, the fault detection unit can compare the acquired voltage signal and current signal with the given voltage threshold and current threshold, respectively.

[0069] If the acquired voltage signal exceeds a given voltage threshold, the fault detection unit can determine that the voltage fault characteristic variable is abnormal. Conversely, if the acquired voltage signal does not exceed the given voltage threshold, the fault detection unit can determine that the voltage fault characteristic variable is not abnormal.

[0070] 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.

[0071] like Figure 2 As shown, the online diagnosis method for the traction system fault provided by the present invention may further include step 202: outputting a corresponding fault classification enable flag in response to an abnormal fault characteristic variable.

[0072] 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 that lasts for a certain period to the fault classification unit. The fault classification enable flag may last for multiple power frequency periods to identify the abnormal fault characteristic variable.

[0073] like Figure 2 As shown, the online diagnosis method for the traction system fault provided by the present invention may further include step 203: collecting relevant time domain statistics of the traction system according to the fault classification enabling flag.

[0074] In the above embodiment, in response to the fault classification enable flag, the fault classification unit can obtain the fault characteristic 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 above-mentioned traction transformer primary overcurrent fault as an example, the fault classification unit can collect the maximum peak value, minimum peak value, high amplitude duration, low amplitude duration, minimum peak change rate and other time domain statistics of the primary voltage and primary current in the periodic sliding window as other time domain statistics related to the traction transformer primary overcurrent fault, so as to realize the accurate classification of different fault sources of the traction transformer primary overcurrent fault.

[0075] like Figure 2 As shown, the online diagnosis method for the traction system fault provided by the present invention may further include step 204: extracting relevant characteristic indicators from relevant time domain statistics to generate an observation vector sequence.

[0076] In some embodiments of the present invention, the fault classification unit can define the maximum peak value, minimum peak value, high amplitude duration, low amplitude duration, minimum peak change rate and other time domain statistics of the primary voltage and primary current in the periodic sliding window as the fault characteristic indicators of the primary overcurrent fault of the traction transformer. Afterwards, the fault classification unit can perform normalization and other pre-processing on the above-mentioned fault characteristic variables such as the primary voltage and primary current and other related time domain statistics related to the abnormality of the fault characteristic variables, so as to comprehensively calculate the related characteristic indicators and generate the corresponding observation vector sequence

[0077] like Figure 2 As shown, the online diagnosis method for the traction system fault provided by the present invention may further include step 205: loading the observation vector sequence into a pre-trained related fault model to determine that the fault model corresponding to the maximum probability value is the fault diagnosis result of the traction system.

[0078] In some embodiments of the present invention, the fault classification unit can determine multiple fault models related to the detected fault feature variable abnormality according to the fault classification enable flag. These fault models can be GMM-HMM models pre-trained in the above-mentioned offline design part. Afterwards, the fault classification unit can generate the observation vector sequence generated by the online implementation part. These pre-trained related fault models λ = (π, A, C, μ, U) are loaded one by one to determine the probability P[O|λ] of the observation vector sequence under each fault model, so as to comprehensively decide the specific fault type of the traction system.

[0079] Please refer to Figure 3 , Figure 3 A schematic flow chart of an online diagnosis method for traction system faults provided according to some embodiments of the present invention is shown.

[0080] like Figure 3 As shown, in some embodiments of the present invention, in response to the fault classification enable flag associated with a plurality of pre-trained GMM-HMM models. The fault classification unit may first transform the observation vector sequence The first pre-trained related GMM-HMM model λ = (π, A, C, μ, U) is loaded to calculate the probability P[O|λ] of the observation vector sequence under the first model.

[0081] Considering that the probability P[O|λ] is usually a very small number, data underflow is inevitable after the likelihood estimates are multiplied. In some preferred embodiments, the processor can perform logarithmic operations on the probability P[O|λ] to calculate the observation vector sequence The likelihood probability log 10 (P[O|λ i ]), and log the likelihood probability value 10 (PpO|λ i ]) as the observation vector sequence The probability value under the first GMM-HMM model can effectively prevent the problem of data underflow and improve the accuracy of fault diagnosis.

[0082] After completing the probability calculation of the first model, the fault classification unit can determine whether there are other related GMM-HMM models. In response to there being other related GMM-HMM models, the fault classification unit can further convert the observation vector sequence Load the second pre-trained related GMM-HMM model λ=(π,A,C,μ,U) to calculate the probability P[O|λ] of the observation vector sequence under the second model. And so on, until the probability calculation of all related GMM-HMM models is completed.

[0083] Afterwards, the fault classification unit can compare the observation vector sequences under each GMM-HMM model The likelihood probability log 10 (P[O|λ i ]) to determine the maximum probability value. In some preferred embodiments, the fault classification unit can further compare the maximum probability value with a given probability threshold P th Make a comparison.

[0084] If the maximum probability value is greater than the given probability threshold P th , the fault classification unit can output the serial number of the GMM-HMM model corresponding to the maximum probability value to indicate the type of fault in the traction system. Conversely, if the maximum probability value is less than the probability threshold P th , it means that the probability of various types of faults occurring in the traction system is very small, so there is no need to output fault diagnosis results indicating any fault type.

[0085] It can be understood from the above description that the present invention uses the method of fault characteristic index analysis and time series modeling to establish a detailed GMM-HMM model for various types of faults that may occur in the traction system. When diagnosing traction system faults online, the present invention can predict various types of faults that may be related based on a small number of sensor signal anomalies, and calculate characteristic indicators by collecting relevant sensor information to determine the probability values ​​of various types of faults, thereby achieving online precise positioning of various types of fault sources in the traction system. In addition, the present invention does not require changing the existing hardware of the train traction system, and can be implemented only by software means, so it is easy to implement in engineering.

[0086] Furthermore, by accurately locating various fault sources of the traction system online, the present invention can also perform differentiated protection isolation actions based on specific fault points, thereby improving train availability.

[0087] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0088] Although the processor, fault detection unit and fault classification unit described in the above embodiments can be implemented by a combination of software and hardware. However, it is understood that these processors, fault detection units and fault classification units can also be implemented in software or hardware alone. For hardware implementation, the processor, fault detection unit and fault classification unit can 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, microcontrollers, microprocessors, other electronic devices for performing the above functions, or a selection combination of the above devices. For software implementation, the processor, fault detection unit and fault classification unit can be implemented by independent software modules such as procedures and functions running on a general-purpose chip, each of which can perform one or more functions and operations described herein.

[0089] 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 apparent to those skilled in the art, and the general 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 should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An online diagnosis method for traction system faults, characterized in that: include: monitoring fault characteristic variables of the traction system; In response to the fault characteristic variable being abnormal, outputting a corresponding fault classification enable flag; collecting relevant time domain statistics of the traction system according to the fault classification enabling flag; Extracting relevant characteristic indicators from the relevant time-domain statistics to generate an observation vector sequence; Loading the observation vector sequence into a pre-trained related fault model to determine that the fault model corresponding to the maximum probability value is the fault diagnosis result of the traction system, wherein the fault model is a GMM-HMM model; Performing fault feature analysis on historical fault data of the traction system to extract relevant feature variables; The fault characteristic index is calculated according to the relevant characteristic variables, and the sample library having K observation vector sequences is defined as O={O (1) ,O (2) ,…,O (K) }, to obtain a sample library of the observation vector sequence of the fault characteristic index, where: is the kth observation vector sequence, O (k) The D-dimensional observation vector at the tth moment; Select N states S = {s1, s2, ..., s N } and M Gaussian mixture elements To establish an HMM model that describes the statistical characteristics of the observation vector sequence, where x nm For state s n The mth mixing element of ; The GMM-HMM model is described 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 weight, μ is the mean vector, and U is the covariance matrix; The initial state probability distribution π is defined as π=[π1,π1,…,π N ] T , where q t is the state of the observation vector sequence at time t, P[·] is the probability of the observation vector sequence, π n = P [q1=s n ]≥0 and The state transition probability distribution A is defined as Among them, a ij =P[q t+1 =s j |q t =s i ]≥0 and all i satisfy The mixed weight C is defined as Among them, c nm ≥0 and all n satisfy The mean vector μ is defined as in, Represents mixed element x nm The D-dimensional mean vector of ; The covariance matrix U is defined as Among them, U nm Represents mixed element x nm The D×D dimensional covariance matrix of The state s n The observation vector The probability density function of Described as The probability density function of the observation vector sequence is determined according to the GMM-HMM model, wherein: is the observation vector The corresponding mixed element x nm The probability density function of ; and Parameters are initialized according to the sample library of the historical fault data, and parameters of the GMM-HMM model are estimated.

2. The online diagnosis method according to claim 1, characterized in that: Also includes: Determining whether the fault characteristic variable exceeds a given threshold; as well as In response to the fault characteristic variable exceeding the threshold, it is determined that the fault characteristic variable is abnormal.

3. The online diagnosis method according to claim 1, characterized in that: The step of collecting the relevant time domain statistics according to the fault classification enabling flag comprises: The fault characteristic variable is acquired in response to the fault classification enable flag, and other relevant time domain statistics are acquired from the corresponding sensor according to the fault classification enable flag.

4. The online diagnosis method according to claim 1, characterized in that: The steps of determining the fault model corresponding to the maximum probability value include: Loading the observation vector sequence into a pre-trained related GMM-HMM model to calculate the probability of the observation vector sequence under the model; Determine whether there is a next related GMM-HMM model; In response to there being a next related GMM-HMM model, loading the observation vector sequence into the next model to calculate the probability of the observation vector sequence under the next model; and In response to there being no next related GMM-HMM model, the sequence number of the GMM-HMM model corresponding to the maximum probability value is output.

5. The online diagnosis method according to claim 4, characterized in that: The step of calculating the probability of the observation vector sequence under a GMM-HMM model further includes: Take the logarithm of the probability P[O|λ] of the observation vector sequence to calculate the likelihood probability log of the observation vector sequence 10 (P[O|λ i ]);as well as The likelihood probability log 10 (P[O|λ i ]) as the probability of the observation vector sequence under the GMM-HMM model.

6. The online diagnosis method according to claim 4, characterized in that: The step of outputting the serial number of the GMM-HMM model corresponding to the maximum probability value further includes: Determining whether the maximum probability value is greater than a preset probability threshold; and In response to the maximum probability value being greater than the probability threshold, the serial number of the GMM-HMM model corresponding to the maximum probability value is output.

7. The online diagnosis method according to claim 1, characterized in that: The step of initializing parameters further includes: initializing the parameters using a segmented K-means algorithm based on the sample library of historical fault data, The step of estimating parameters of the GMM-HMM model further includes: completing parameter estimation of the GMM-HMM model based on the Baum-Welch algorithm.

8. An online diagnostic device for traction system faults, characterized in that: The device comprises a memory and a processor, wherein the processor is connected to the memory and configured to implement the online diagnosis method according to any one of claims 1 to 7.

9. The online diagnostic device according to claim 8, characterized in that: comprising a plurality of said processors, wherein: The first processor is a fault detection unit configured to monitor the fault characteristic variables of the traction system online and output a corresponding fault classification enable flag in response to an abnormality of the fault characteristic variables. The second processor is a fault classification unit configured to determine a fault diagnosis result of the traction system according to the fault classification enable flag, the fault characteristic variable and relevant time domain statistics of the traction system.

10. The online diagnostic device according to claim 9, characterized in that: The system also includes a third processor configured to train various fault models of the traction system according to historical fault data of the traction system.

11. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the online diagnosis method according to any one of claims 1 to 7 is implemented.

Images