A traction motor overcurrent fault identification method and system

Abstract

The present invention proposes a traction motor overcurrent fault identification method and system, which relates to the field of overcurrent fault diagnosis technology. The method involves acquiring historical signals and performing offline preprocessing to extract a first feature identifier. The method then segments the first feature identifier within a sliding window, generates a first feature index, performs a first event sequence conversion, constructs a traction motor overcurrent fault location and identification template, and divides it into a training set and a test set. The training set is used to train the traction motor overcurrent fault identification model, and the test set is used to test the effectiveness of the traction motor overcurrent fault identification model. The method then performs online fault detection on real-time signals to extract a second feature identifier. The method then segments the second feature identifier within a sliding window, generates a second feature index, performs a second event sequence conversion, obtains an online template vector, and inputs it into the trained traction motor overcurrent fault identification model, ultimately outputting a fault identification result. The method effectively reduces algorithm complexity and enables rapid, accurate, and real-time identification of fault types.

Description

Technical Field

[0001] The present invention relates to the technical field of overcurrent fault diagnosis, and in particular to a traction motor overcurrent fault identification method and system. Background Art

[0002] The high-speed rail traction system is a core component that ensures the normal operation of the train and is known as the heart of the train. It consists primarily of three major components: the traction transformer, the traction converter, and the traction motor. Its operating principle is as follows: 25kV AC power is received through the pantograph, stepped down by the traction transformer, converted to DC by a four-quadrant rectifier, filtered by an intermediate DC link, and finally converted by an inverter into three-phase AC with adjustable frequency and amplitude to drive the traction motor, thereby achieving train speed control. The electric locomotive traction drive system is the core power system for train operation. When operating under high speed, high power, and complex operating conditions, the traction motor is prone to overcurrent faults. Once an overcurrent fault occurs, if it is not diagnosed and effectively addressed in a timely manner, it may cause traction system failure, resulting in train delays and even major safety accidents. Therefore, real-time and accurate diagnosis of traction motor overcurrent faults is crucial to ensuring train operation safety.

[0003] Currently, the following methods are primarily used to diagnose traction motor overcurrent faults: First, a simple threshold-based alarm method generates an alarm when the collected motor current signal exceeds the set threshold. This method can only detect the fault phenomenon but cannot accurately locate and determine the cause of the fault. Second, fault diagnosis and classification are performed based on artificial intelligence methods such as neural networks and decision trees. While these methods can identify faults to a certain extent, their high algorithm complexity and poor real-time performance make them difficult to widely apply in practical engineering. Traction motor overcurrent faults have complex causes and can be caused by a variety of factors, including speed sensor failure, inverter IGBT module failure, and motor failure. After a fault occurs, the system often undergoes multiple operating condition changes. Traditional diagnostic methods based on single-moment features or static patterns make it difficult to accurately determine the cause of the fault. Summary of the Invention

[0004] In order to solve the problems of high algorithm complexity, poor real-time performance and difficulty in quickly and accurately identifying fault types in the above-mentioned existing technologies, the present invention proposes a traction motor overcurrent fault identification method and system, which can effectively reduce the algorithm complexity and quickly and accurately identify the fault type in real time.

[0005] In order to achieve the above technical effects, the technical solutions of the present invention are as follows:

[0006] A method for identifying an overcurrent fault in a traction motor comprises the following steps:

[0007] S1. Obtain historical signals of the traction drive system, perform offline preprocessing on the historical signals, and obtain a first characteristic identifier with fault location discrimination;

[0008] S2. Segmenting the first feature identifier within the sliding window to obtain a first feature index, generating a corresponding first event using the first feature index, and performing sequence conversion on the first event to obtain a traction motor overcurrent fault location identification template;

[0009] S3. Divide the traction motor overcurrent fault location identification template into a training set and a test set, use the training set to train the constructed traction motor overcurrent fault identification model, and use the test set to test the effectiveness of the traction motor overcurrent fault identification model to obtain a trained traction motor overcurrent fault identification model;

[0010] S4 obtains the real-time signal of the traction drive system, performs online fault detection on the real-time signal, and obtains a second characteristic identifier having a fault location distinction;

[0011] S5. Segmenting the second feature identifier within the sliding window to obtain a second feature index, generating a corresponding second event using the second feature index, and performing sequence conversion on the second event to obtain an online template vector;

[0012] S6. Input the online template vector into the trained traction motor overcurrent fault recognition model and output the fault recognition result.

[0013] Preferably, the historical signal includes the first A phase current signal , the first B phase current signal , the first intermediate voltage sensor signal and the first speed signal , the first feature identifier includes a first mean , the first intermediate voltage sensor historical signal The first variance , the first intermediate voltage sensor signal Minimum value of , first speed signal The second variance , the first current minimum , the first current maximum value , the first current effective value and the first current normalized value ;

[0014] The offline preprocessing of the historical signal to obtain a first feature identifier with fault location discrimination includes:

[0015] According to the historical signal of the first intermediate voltage sensor and the first speed signal , get the first mean The calculation expression is:

[0016]

[0017] in, k Indicates the running time, Indicates the k The first mean of running time , represents the intermediate variable, = , represents the sliding window size, Indicates the value is and , ;

[0018] According to the historical signal of the first intermediate voltage sensor , get the first variance The calculation expression is:

[0019]

[0020] in, Indicates the k First variance of running time , Indicates in k Runtime The value is The first mean ;

[0021] According to the historical signal of the first intermediate voltage sensor , get the first intermediate voltage sensor signal Minimum value of The calculation expression is:

[0022]

[0023] in, Indicates the k The first intermediate voltage sensor signal at the time of operation Minimum value of , Indicates taking the minimum value, No. k Historical signal of the first intermediate voltage sensor at the time of operation , express Historical signal of the first intermediate voltage sensor at the time of operation , Indicates the Historical signal of the first intermediate voltage sensor at the time of operation ;

[0024] According to the first speed signal , and get the second variance The calculation expression is:

[0025]

[0026] in, Indicates the k Run-time second variance , Indicates in k Runtime The value is The first mean ;

[0027] According to the first A-phase current signal and the first B-phase current signal , calculate the first current minimum value The calculation expression is:

[0028]

[0029] in, Indicates the k The first minimum current value during operation , y Indicates the value is a and b The subscript of Indicates in k The first A-phase current signal at the time of operation and the first B-phase current signal , Indicates in The first A-phase current signal at the time of operation and the first B-phase current signal , exist The first A-phase current signal at the time of operation and the first B-phase current signal ;

[0030] According to the first A-phase current signal and the first B-phase current signal , calculate the maximum current The calculation expression is:

[0031]

[0032] in, Indicates taking the maximum value;

[0033] According to the first A-phase current signal and the first B-phase current signal , calculate the first current effective value The calculation expression is:

[0034]

[0035] in, Indicates the k The first effective current value at the time of operation ;

[0036] According to the first A-phase current signal and the first B-phase current signal , calculate the first current normalized value The calculation expression is:

[0037]

[0038] in, Indicates the k The first normalized current value at the time of operation .

[0039] Preferably, the sliding window characteristic data is obtained within the sliding window, wherein the sliding window characteristic data includes the current frequency of the traction motor , sliding window size , step length , Number of sliding windows , the number of units distributed on the left side of the traction control unit's action point , the number of units distributed on the right side of the traction control unit's action point and data length ;

[0040] The sliding window size The calculation expression is:

[0041]

[0042] in, Indicates the sampling period;

[0043] The step length The calculation expression is:

[0044]

[0045] The number of the distribution on the left side of the traction control unit action point The calculation expression is:

[0046]

[0047] The number of the distribution on the right side of the traction control unit action point The calculation expression is:

[0048]

[0049] The data length The calculation expression is:

[0050]

[0051] The first feature identifier is segmented according to the sliding window feature data to obtain the first feature indexes, which are calculated as follows:

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] in, 、 、 、 、 represents the first characteristic index, 、 、 、 、 Indicates the k The first characteristic indicator of runtime.

[0058] Preferably, the step of generating a corresponding first event by using the first characteristic indicator and performing sequence conversion on the first event to obtain a traction motor overcurrent fault location and identification template includes:

[0059] S21. The first characteristic index 、 、 、 、 are input into the first hysteresis comparator respectively, and the first hysteresis comparator outputs the corresponding first event The calculation expression is as follows:

[0060]

[0061] in, Indicates that the values ​​are The subscript of Indicates the first hysteresis comparator start point, Indicates the first hysteresis comparator shutdown point, , ;like Not less than the first hysteresis comparator opening point , then the first event The output is 1; if Less than the first hysteresis comparator turn-on point and is greater than the first hysteresis comparator shutdown point , then the first event Keep the output state consistent with the previous moment event; if Not greater than the first hysteresis comparator shutdown point , then the first event The output is 0;

[0062] S22. The first event Press the X The order forms a binary time series vector , the binary time series vector Convert to decimal The following mathematical expression:

[0063]

[0064] in, Indicates the current calculated decimal number The sliding window position, (1, n ), represents the length of the first vector, ;

[0065] S23. Based on the decimal number , get the traction motor overcurrent fault location identification vector The calculation expression is as follows:

[0066]

[0067] in, Indicates that the number of sliding windows is the length of the first vector The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of

[0068] The calculation expression is:

[0069]

[0070] The calculation expression is:

[0071]

[0072] S24. The traction motor overcurrent fault location identification vector obtained under different fault types A collection is formed to obtain the traction motor overcurrent fault location and identification template.

[0073] Preferably, the traction motor overcurrent fault recognition model includes a sequence input layer, a bidirectional long short-term memory network layer, a self-attention layer, a first fully connected layer, a first activation function layer, a dropout layer, a second fully connected layer, a second activation function layer and a classification layer connected in sequence;

[0074] The traction motor overcurrent fault location and identification template is normalized into a 16-dimensional template vector and input into the sequence input layer. The sequence input layer outputs time series data. The bidirectional long short-term memory network layer receives the time series data and outputs the bidirectional dependency features of the time series. The self-attention layer receives the bidirectional dependency features and outputs the time step feature data weighted by attention. The first fully connected layer receives the time step feature data and outputs the first linear feature combination result. The first activation function layer receives the first linear feature combination result and outputs the first activation feature result. The discard layer receives the first activation feature result and outputs the feature result after discarding. The second fully connected layer receives the feature result after discarding and outputs the second linear feature combination result. The second activation function layer receives the second linear feature combination result and outputs the fault type prediction probability. The classification layer receives the fault type prediction probability and outputs the fault type identification result.

[0075] Preferably, the training of the traction motor overcurrent fault identification model includes:

[0076] S31. Set the training parameters of the traction motor overcurrent fault identification model, the training parameters including: maximum rounds of 150, batch size of 32, initial learning rate of 0.005, learning rate scheduling of 0.2 times every 50 rounds, L2 regularization coefficient of 0.0001, and gradient clipping threshold of 1;

[0077] S32. Preliminarily optimize the training parameters using the Adam optimizer to obtain preliminarily optimized hyperparameters;

[0078] S33. Use a genetic algorithm to globally optimize the hyperparameters after preliminary optimization until the hyperparameter combination corresponding to the individual with the highest fitness value is output, thereby completing the training of the traction motor overcurrent fault identification model.

[0079] Preferably, the global optimization of the preliminarily optimized hyperparameters using a genetic algorithm comprises:

[0080] S331. Set the hyperparameter configuration of each individual in the population size;

[0081] S332. Use the hyperparameter configuration of each individual to build a traction motor overcurrent fault recognition model. Train the traction motor overcurrent fault recognition model using the training set, and calculate the classification accuracy α and Macro-F1 score β on the test set. Calculate the fitness value X based on the classification accuracy α and Macro-F1 score β as follows:

[0082] X 0.7α+0.3β

[0083] S333. Selecting a superior individual from the current population size as a parent using tournament selection or roulette wheel selection based on the fitness value X;

[0084] S334. Perform a crossover operation on the parent individuals, using single-point crossover or uniform crossover, to swap the hyperparameter values ​​of the two parent individuals to generate a new crossover individual;

[0085] S335. Randomly perturb and fine-tune the hyperparameter values ​​of the new individuals in the crossover with a mutation probability of 5% to 10% to obtain mutated new individuals.

[0086] S336. Add the new individuals from the crossover and mutation to the population to form the next generation and then perform iterations.

[0087] S337. When the number of update iterations reaches the iteration threshold or the change in the fitness value X is less than 0.001 within 5 consecutive generations, the global optimization is terminated and the hyperparameter combination corresponding to the individual with the highest fitness value is output.

[0088] Preferably, the real-time signal includes a second A-phase current signal , the second B phase current signal , the second intermediate voltage sensor signal and the second speed signal The second characteristic identifier includes a second intermediate voltage sensor signal Minimum value of , the second mean , the second intermediate voltage sensor signal Third-party difference , second speed signal The fourth variance , the second current minimum , the second current maximum value , the second current normalized value , the second current effective value The online fault detection of the real-time signal to obtain a second characteristic identifier with fault location discrimination includes:

[0089] S41. Determine the collected second A-phase current signal and the second B-phase current signal Is the current greater than the current protection threshold in several consecutive sampling cycles? If yes, then enter the fault diagnosis phase and execute S42; if no, then continue to collect the second A phase current signal and the second B-phase current signal Perform online fault detection;

[0090] S42. According to the second intermediate voltage sensor signal and the second speed signal , and obtain the second mean The calculation expression is:

[0091]

[0092] in, k Indicates the running time, Indicates the k Second mean of running time , represents the intermediate variable, = , represents the sliding window size, Indicates the value is and , ;

[0093] According to the second intermediate voltage sensor signal , get the third party difference The calculation expression is:

[0094]

[0095] in, Indicates the k Third-party differences at runtime , Indicates in k Runtime The value is The second mean ;

[0096] According to the second intermediate voltage sensor signal , get the second intermediate voltage sensor signal Minimum value of The calculation expression is:

[0097]

[0098] in, Indicates the k Second intermediate voltage sensor signal at run time Minimum value of , Indicates taking the minimum value, No. k Second intermediate voltage sensor signal at run time , express Second intermediate voltage sensor signal at run time , Indicates the Second intermediate voltage sensor signal at run time ;

[0099] According to the second speed signal , and get the fourth variance The calculation expression is:

[0100]

[0101] in, Indicates the k Runtime fourth variance , Indicates in k Runtime The value is The second mean ;

[0102] According to the second A-phase current signal and the second B-phase current signal , calculate the second current minimum The calculation expression is:

[0103]

[0104] in, Indicates the k The second minimum current during operation , y Indicates the value is a and b The subscript of Indicates in k The second A phase current signal at the time of operation and the second B-phase current signal , Indicates in The second A phase current signal at the time of operation and the second B-phase current signal , exist The second A phase current signal at the time of operation and the second B-phase current signal ;

[0105] According to the second A-phase current signal and the second B-phase current signal , calculate the second current maximum value The calculation expression is:

[0106]

[0107] in, Indicates taking the maximum value;

[0108] According to the second A-phase current signal and the second B-phase current signal , calculate the second current effective value The calculation expression is:

[0109]

[0110] in, Indicates the k The second current effective value during operation ;

[0111] According to the second A-phase current signal and the second B-phase current signal , calculate the second current normalized value The calculation expression is:

[0112]

[0113] in, Indicates the k The second current normalized value at the time of operation .

[0114] Preferably, the second feature identifier is segmented within the sliding window to obtain the second feature indexes, which are respectively calculated as follows:

[0115]

[0116]

[0117]

[0118]

[0119]

[0120] in, 、 、 、 、 represents the second characteristic index, 、 、 、 、 Indicates the k The second characteristic indicator of the running time;

[0121] Generating a corresponding second event using the second characteristic indicator, performing sequence conversion on the second event to obtain an online template vector, including:

[0122] S51. The second characteristic index 、 、 、 、 are input into the second hysteresis comparator, and the second hysteresis comparator outputs the corresponding second event ;

[0123] S52. The second event Press the X The order forms a binary time series event , the binary timing events Convert to decimal The following mathematical expression:

[0124]

[0125] in, Indicates the current calculated decimal number The sliding window position, (1, m ), m represents the length of the second vector, ;

[0126] S53. Based on the decimal number , get the template vector The calculation expression is as follows:

[0127]

[0128] in, Indicates that the number of sliding windows is the length of the second vector m The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of

[0129] The calculation expression is:

[0130]

[0131] The calculation expression is:

[0132] .

[0133] The present invention also provides a traction motor overcurrent fault identification system, comprising:

[0134] an offline preprocessing module, configured to obtain historical signals of the traction drive system, perform offline preprocessing on the historical signals, and obtain a first feature identifier with fault location discrimination capability;

[0135] an offline segmentation and conversion module, configured to segment the first feature identifier within a sliding window to obtain a first feature index, generate a corresponding first event using the first feature index, and perform sequence conversion on the first event to obtain a traction motor overcurrent fault location and recognition template;

[0136] a traction motor overcurrent fault identification model construction module, configured to construct a traction motor overcurrent fault identification model, divide the traction motor overcurrent fault location identification template into a training set and a test set, train the traction motor overcurrent fault identification model using the training set, and test the effectiveness of the traction motor overcurrent fault identification model using the test set, thereby obtaining a trained traction motor overcurrent fault identification model;

[0137] An online preprocessing module, configured to obtain a real-time signal of the traction drive system, perform online fault detection on the real-time signal, and obtain a second characteristic identifier with fault location discrimination;

[0138] an online segmentation and conversion module, configured to segment the second feature identifier within a sliding window to obtain a second feature index, generate a corresponding second event using the second feature index, and perform sequence conversion on the second event to obtain an online template vector;

[0139] The output module is used to input the online template vector into the trained traction motor overcurrent fault recognition model and output the fault recognition result.

[0140] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0141] The present invention proposes a traction motor overcurrent fault identification method and system. First, in the offline stage, the historical signal is preprocessed offline to obtain a first feature identifier, the first feature identifier is segmented according to the sliding window feature data to obtain a first feature index, the first feature index is used to generate a corresponding first event, and the first event is sequenced, the purpose is to reduce the offline data dimension to one dimension, thereby effectively reducing the complexity of the calculation, then a traction motor overcurrent fault identification model is constructed and trained to obtain a trained traction motor overcurrent fault identification model; then, online fault detection is performed on the real-time signal to obtain a second feature identifier with fault location discrimination; the second feature identifier is segmented within the sliding window to obtain a second feature index, the second feature index is used to generate a corresponding second event, and the second event is sequenced, the purpose is to reduce the online data dimension to one dimension, thereby further effectively reducing the complexity of the calculation; finally, the online template vector is input into the trained traction motor overcurrent fault identification model to achieve fast and accurate judgment of the fault type, significantly improving the efficiency and accuracy of real-time fault identification. Under the premise of ensuring the accuracy of fault identification, the present invention can improve the speed and real-time performance of data matching through the traction motor overcurrent fault identification model, thereby better meeting the needs of real-time diagnosis of motor overcurrent faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0142] Figure 1 A schematic diagram showing a flow chart of a method for identifying an overcurrent fault of a traction motor proposed in an embodiment of the present invention;

[0143] Figure 2 Another schematic flow chart showing a method for identifying an overcurrent fault of a traction motor according to an embodiment of the present invention;

[0144] Figure 3 A diagram showing a topological structure model of a power generation circuit of a traction drive system proposed in an embodiment of the present invention;

[0145] Figure 4 The figure shows a structural block diagram of a traction motor overcurrent fault identification system proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0146] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting the present invention;

[0147] In order to better illustrate the present embodiment, some parts of the drawings may be omitted, enlarged or reduced, and do not represent actual sizes. The description of the directions of parts such as "upper" and "lower" does not limit the present invention.

[0148] It is understandable to those skilled in the art that some well-known contents may be omitted in the drawings;

[0149] To facilitate understanding of this embodiment, first, the prior art information of this embodiment is introduced as follows:

[0150] Existing patent documents disclose a traction converter fault diagnosis method, device, medium, and product. In this application, intermediate DC voltage data is obtained as a dataset and decomposed using variational mode decomposition (VMD) to obtain a multi-channel standard fault signature set. VMD has high frequency domain resolution, enabling separation of different frequency features in the presence of noise. Based on this characteristic, the PE-Spearman rank correlation coefficient was chosen as the supporting channel weighting layer. Based on this, a one-dimensional deep separable convolutional neural network model with this supporting channel weighting layer was established. The supporting channel weighting layer serves as a bridge between VMD and the neural network model. Finally, the multi-channel standard fault signature set is input into the model to obtain a fault classification. This method is computationally complex, making it difficult to quickly output results. It also requires high hardware performance, otherwise it cannot meet the requirements for rapid fault diagnosis. Furthermore, the model parameter structure established by this method is relatively fixed, resulting in poor adaptability.

[0151] Existing patent literature also discloses a method for real-time identification and diagnosis of inverter overcurrent based on a set of sequential operating condition events. This application comprises two parts: offline design and online diagnosis. In the offline design phase, historical fault data is preprocessed to determine the window size. Within the window, relevant characteristic variables are calculated to obtain characteristic indices. Events are generated using a hysteresis comparator, and a template library of sequential operating condition event sets for different fault types is established. The online phase includes fault detection and fault decision modules. The fault detection module extracts sensor signals related to inverter overcurrent in real time, compares them with detection thresholds, and generates a diagnostic enable flag that persists for a certain period of time for the fault decision module. After receiving the flag and the adaptive window size, the fault decision module generates an operating condition event set based on the analog signals collected by the detection module. This is matched in real time against the fault diagnosis template library of the sequential operating condition event set to output the fault type. This method requires a large data set to be read, and the calculation of characteristic indices and event sets is time-consuming. Furthermore, this method places high demands on the quality of the acquired data, requiring a perfect match for output.

[0152] An existing paper discloses an analysis of motor overcurrent faults during braking of a mining electric wheel dump truck. This method uses the measured voltage output by the inverter to infer the real-time values ​​of the amplitude and phase of the spatial voltage reference vector in the control program for overcurrent fault detection, but it fails to trace the source of the overcurrent fault.

[0153] Existing papers also disclose research on high-speed train traction motor fault diagnosis methods based on the TS fuzzy model. This method establishes a TS fuzzy model of the traction motor. Based on the TS fuzzy model, the robust fault diagnosis problem of the traction motor affected by uncertain factors such as parameter changes, interference, and noise is studied. The main research focuses on current sensor faults and traction motor stator inter-turn short-circuit faults, but some other types of faults cannot be accurately diagnosed.

[0154] Existing papers also disclose real-time diagnosis of traction motor overcurrent based on timing feature pattern recognition. This method constructs an event set through timing feature indicators, establishes a diagnostic template library in an offline state, and realizes online diagnosis of overcurrent faults. However, the paper can only distinguish overcurrent faults of the speed signal type, and does not conduct in-depth research on other fault causes.

[0155] In summary, the current diagnosis of motor overcurrent in traction drive systems mostly adopts a model-based construction method, and most of them can only diagnose inverter overcurrent of a single fault source type. However, the construction of the traction drive system model is complex and difficult to express with a single mathematical model. A small interference will have a great impact on the diagnostic results. The current diagnostic method for calculating characteristic indicators mostly adopts a fixed sliding window size mode. The calculation time for large amounts of data will be longer, the hardware performance requirements are high, and the maintenance efficiency is affected. The current matching method in the diagnostic process has high requirements for the accuracy of reading data and the setting of event trigger thresholds, and has certain requirements for the length of data time series, which makes the application scope small. Therefore, in order to solve the problems of high algorithm complexity, poor real-time performance, and difficulty in quickly and accurately identifying fault types in the existing technology, the present invention proposes a traction motor overcurrent fault identification method and system.

[0156] The terms used in the drawings to describe positional relationships are for illustrative purposes only and are not to be construed as limiting the present invention.

[0157] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0158] Example 1

[0159] like Figure 1 and Figure 2 As shown, this embodiment proposes a method for identifying a traction motor overcurrent fault, comprising the following steps:

[0160] S1. Obtain historical signals of the traction drive system, perform offline preprocessing on the historical signals, and obtain a first characteristic identifier with fault location discrimination;

[0161] See also Figure 3 , is a topological structure model of the traction drive system, consisting of three parts: the traction transformer, the traction converter, and the traction motor. Single-phase AC 25kV AC current flows into the train body through the pantograph, the main circuit breaker (VCB), and the primary winding of the traction transformer. After being stepped down by the traction transformer, it provides single-phase AC power to the train traction converter through the secondary winding. The electric energy undergoes AC-DC-AC conversion in the converter to power the train traction motor. The historical signal includes the first A-phase current signal , the first B phase current signal , the first intermediate voltage sensor signal and the first speed signal By analyzing the mechanism of motor overcurrent fault, the historical signal is preprocessed offline to obtain the first feature identifier including the first mean , the first intermediate voltage sensor historical signal The first variance , the first intermediate voltage sensor signal Minimum value of , first speed signal The second variance , the first current minimum , the first current maximum value , the first current effective value and the first current normalized value The analysis of the motor overcurrent fault mechanism includes:

[0162] S11: When the train is operating normally, the motor overcurrent output current waveform is periodic sine, and the absolute value of each sampling cycle value is less than the protection threshold, and the enable protection will not be triggered. The first intermediate voltage sensor signal and the first speed signal The disturbances are all within a reasonable range and there will be no large deviations. At this time, it can be judged that the train is in normal operation.

[0163] S12: When a speed signal type motor overcurrent fault occurs in the traction drive system, the first A phase current signal , the first B phase current signal The waveform will be obviously distorted. After the traction control unit TCU works, All dropped to 0, the first speed signal A positive-negative jump occurs. The first intermediate voltage sensor signal There were no noticeable changes throughout the process.

[0164] S13: When a traction motor type motor overcurrent fault occurs in the traction drive system, before the traction control unit TCU works The waveform is not distorted, but it shows a divergent trend. There is obvious burr disturbance. After the traction control unit TCU works Both drop to zero, the first intermediate voltage sensor signal The burr disturbance phenomenon disappears.

[0165] S14: When the traction drive system has an intermediate voltage sensor signal type motor overcurrent fault, the first intermediate voltage sensor signal is normally operated before the traction control unit TCU works. Suddenly fell, The current quickly diverges and exceeds the protection threshold. After the traction control unit TCU works, the current returns to zero and the first intermediate voltage sensor signal Restore normal operation. First speed signal There were no significant changes during the entire process.

[0166] S15: When the traction drive system has an inverter type motor overcurrent fault, the inverter suddenly fails during normal operation before the traction control unit TCU works, and the first intermediate voltage sensor signal Instantly dropped to near 0, and at the same time Rapidly diverges and exceeds the protection threshold, the first speed signal After the traction control unit TCU works, the current drops to 0 and the first intermediate voltage sensor signal After returning to normal levels, it dropped rapidly again.

[0167] S16: From S11 to S15, the traction drive system motor overcurrent fault mechanism is summarized, and it can be clearly obtained that the collection 、 and It has good discrimination, so the first feature identifier is constructed including 、 、 、 、 、 、 .

[0168] S17: Use MATLAB-Simulink to simulate the motor overcurrent fault of the train traction drive system, and obtain the data of the normal system and the motor overcurrent fault as the original data, including the normal state and four types of motor overcurrent fault states, a total of 5 groups of data, each group of 17,500 sample data.

[0169] The offline preprocessing of the historical signal to obtain a first feature identifier with fault location discrimination includes:

[0170] According to the historical signal of the first intermediate voltage sensor and the first speed signal , get the first mean The calculation expression is:

[0171]

[0172] in, k Indicates the running time, Indicates the k The first mean of running time , represents the intermediate variable, = , represents the sliding window size, Indicates the value is and , ;

[0173] According to the historical signal of the first intermediate voltage sensor , get the first variance The calculation expression is:

[0174]

[0175] in, Indicates the k First variance of running time , Indicates in k Runtime The value is The first mean ;

[0176] According to the historical signal of the first intermediate voltage sensor , get the first intermediate voltage sensor signal Minimum value of The calculation expression is:

[0177]

[0178] in, Indicates the k The first intermediate voltage sensor signal at the time of operation Minimum value of , Indicates taking the minimum value, No. k Historical signal of the first intermediate voltage sensor at the time of operation , express Historical signal of the first intermediate voltage sensor at the time of operation , Indicates the Historical signal of the first intermediate voltage sensor at the time of operation ;

[0179] According to the first speed signal , and get the second variance The calculation expression is:

[0180]

[0181] in, Indicates the k Run-time second variance , Indicates in k Runtime The value is The first mean ;

[0182] According to the first A-phase current signal and the first B-phase current signal , calculate the first current minimum value The calculation expression is:

[0183]

[0184] in, Indicates the k The first minimum current value during operation , y Indicates the value is a and b The subscript of Indicates in k The first A-phase current signal at the time of operation and the first B-phase current signal , Indicates in The first A-phase current signal at the time of operation and the first B-phase current signal , exist The first A-phase current signal at the time of operation and the first B-phase current signal ;

[0185] According to the first A-phase current signal and the first B-phase current signal , calculate the maximum current The calculation expression is:

[0186]

[0187] in, Indicates taking the maximum value;

[0188] According to the first A-phase current signal and the first B-phase current signal , calculate the first current effective value The calculation expression is:

[0189]

[0190] in, Indicates the k The first effective current value at the time of operation ;

[0191] According to the first A-phase current signal and the first B-phase current signal , calculate the first current normalized value The calculation expression is:

[0192]

[0193] in, Indicates the k The first normalized current value at the time of operation .

[0194] S2. Segmenting the first feature identifier within the sliding window to obtain a first feature index, generating a corresponding first event using the first feature index, and performing sequence conversion on the first event to obtain a traction motor overcurrent fault location identification template;

[0195] In S2, a sliding window characteristic data is obtained within the sliding window, wherein the sliding window characteristic data includes the current frequency of the traction motor obtained by fast Fourier transform (FFT). , calculate the extracted data length according to the current frequency, and calculate the sliding window size , step length , Number of sliding windows , the number of units distributed on the left side of the traction control unit's action point , the number of units distributed on the right side of the traction control unit's action point and data length ;

[0196] The sliding window size Current frequency The size is dynamically selected. The processing analysis of the S2 sliding window is as follows:

[0197] S21: When analyzing and processing the mark, the sliding window idea is adopted to process the data. The sliding window size and step size are obtained by fast Fourier transform through the collected current data. , according to the current frequency To determine the size and step size of the sliding window. Current frequency used for fast Fourier transform To enable diagnosis, the first 5000 data points out of the 17500 data points are extracted, which enables the calculation of the current frequency. It is not affected by the jump after the fault, ensuring the accuracy of the calculation. In addition, it can also ensure that there are at least two complete current cycles in the calculation.

[0198] S22: The number of sliding windows required for different frequencies is obtained by offline data training The number of sliding windows can be obtained according to the S21 current frequency Taking the action time of the traction control unit TCU as the base point, the sliding window is distributed on both sides of the base point.

[0199] The sliding window size The calculation expression is:

[0200]

[0201] in, Indicates the sampling period;

[0202] The step length The calculation expression is:

[0203]

[0204] Different current frequencies are obtained by offline data training Number of sliding windows required ;

[0205] The number of the distribution on the left side of the traction control unit action point The calculation expression is:

[0206]

[0207] The number of the distribution on the right side of the traction control unit action point The calculation expression is:

[0208]

[0209] The data length The calculation expression is:

[0210] .

[0211] The first feature identifier is segmented according to the sliding window feature data to obtain the first feature indexes, which are respectively calculated as follows:

[0212]

[0213]

[0214]

[0215]

[0216]

[0217] in, 、 、 、 、 represents the first characteristic index, 、 、 、 、 Indicates the k The first characteristic indicator of runtime.

[0218] The method of generating a corresponding first event by using the first characteristic indicator and performing sequence conversion on the first event to obtain a traction motor overcurrent fault location and identification template includes:

[0219] S21. The first characteristic index 、 、 、 、 are input into the first hysteresis comparator respectively, and the first hysteresis comparator outputs the corresponding first event The calculation expression is as follows:

[0220]

[0221] in, Indicates that the values ​​are The first hysteresis comparator is 5, Indicates the first hysteresis comparator start point, Indicates the first hysteresis comparator shutdown point, , ;like Not less than the first hysteresis comparator opening point , then the first event The output is 1; if Less than the first hysteresis comparator turn-on point and is greater than the first hysteresis comparator shutdown point , then the first event Keep the output state consistent with the previous moment event; if Not greater than the first hysteresis comparator shutdown point , then the first event The output is 0;

[0222] S22. The first event Press the X The order forms a binary time series vector , the binary time series vector Convert to decimal The following mathematical expression:

[0223]

[0224] in, Indicates the current calculated decimal number The sliding window position, (1, n ), represents the length of the first vector, ;

[0225] S23. Based on the decimal number , get the traction motor overcurrent fault location identification vector The calculation expression is as follows:

[0226]

[0227] in, Indicates that the number of sliding windows is the length of the first vector The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of

[0228] The calculation expression is:

[0229]

[0230] The calculation expression is:

[0231]

[0232] S24. The traction motor overcurrent fault location identification vector obtained under different fault types A collection is formed to obtain the traction motor overcurrent fault location and identification template.

[0233] Each first event has two states, 0 and 1. Write the first event out as a binary time series vector and write the binary time series vector Convert to normalized decimal number , statistical sequence characteristics, generating the number of sliding windows The traction motor overcurrent fault location identification vector is the length .

[0234] S3. Construct a traction motor overcurrent fault recognition model, divide the traction motor overcurrent fault location recognition template into a training set and a test set, use the training set to train the traction motor overcurrent fault recognition model, and use the test set to test the effectiveness of the traction motor overcurrent fault recognition model, to obtain a trained traction motor overcurrent fault recognition model;

[0235] S4 obtains the real-time signal of the traction drive system, performs online fault detection on the real-time signal, and obtains a second characteristic identifier having a fault location distinction;

[0236] S5. Segmenting the second feature identifier within the sliding window to obtain a second feature index, generating a corresponding second event using the second feature index, and performing sequence conversion on the second event to obtain an online template vector;

[0237] S6. Input the online template vector into the trained traction motor overcurrent fault recognition model and output the fault recognition result.

[0238] In this embodiment, first, in the offline phase, the historical signal is preprocessed offline to obtain a first feature identifier. The first feature identifier is segmented based on the sliding window feature data to obtain a first feature index. The first feature index is used to generate a corresponding first event, and the first event is converted into a sequence. The purpose is to reduce the offline data dimension to one dimension, thereby effectively reducing the computational complexity. Then, a traction motor overcurrent fault recognition model is constructed and trained to obtain a trained traction motor overcurrent fault recognition model. Then, online fault detection is performed on the real-time signal to obtain a second feature identifier with fault location discrimination. The second feature identifier is segmented within the sliding window to obtain a second feature index. The second feature index is used to generate a corresponding second event, and the second event is converted into a sequence. The purpose is to reduce the online data dimension to one dimension, thereby further effectively reducing the computational complexity. Finally, the online template vector is input into the trained traction motor overcurrent fault recognition model to achieve rapid and accurate fault type discrimination, significantly improving the efficiency and accuracy of real-time fault recognition. The present invention can improve the speed and real-time performance of data matching through the traction motor overcurrent fault recognition model while ensuring fault recognition accuracy, thereby better meeting the needs of real-time diagnosis of motor overcurrent faults.

[0239] Example 2

[0240] This embodiment further explains S3 in the above-mentioned method for identifying overcurrent faults of traction motors.

[0241] S3, the traction motor overcurrent fault recognition model includes a sequentially connected sequence input layer, a bidirectional long short-term memory network layer including 50 hidden units, a self-attention layer provided with 2 attention heads and 8 hidden units, a first fully connected layer, a first activation function layer, a dropout layer, a second fully connected layer, a second activation function layer, and a classification layer;

[0242] The traction motor overcurrent fault location and identification template is normalized into a 16-dimensional template vector and input into the sequence input layer. The sequence input layer outputs time series data. The bidirectional long short-term memory network layer receives the time series data and outputs the bidirectional dependency features of the time series. The self-attention layer receives the bidirectional dependency features and outputs the time step feature data weighted by attention. The first fully connected layer receives the time step feature data and outputs the 30-dimensional first linear feature combination result. The first activation function layer receives the first linear feature combination result and outputs the first activation feature result. The discard layer receives the first activation feature result with a discard rate of 0.2 and outputs the feature result after discarding. The second fully connected layer receives the feature result after discarding and outputs the second linear feature combination result. The second activation function layer receives the second linear feature combination result and outputs the fault type prediction probability. The classification layer receives the fault type prediction probability and outputs the fault type identification result.

[0243] The traction motor overcurrent fault location recognition template is feature extracted and normalized (0–1), and divided into a training set and a test set (ratio 7:3). Each sample is reconstructed into an LSTM sequence format, and the label data is converted into a categorical variable. The training of the traction motor overcurrent fault recognition model includes:

[0244] S31. Set the training parameters of the traction motor overcurrent fault identification model, the training parameters including: maximum rounds of 150, batch size of 32, initial learning rate of 0.005, learning rate scheduling of 0.2 times every 50 rounds, L2 regularization coefficient of 0.0001, and gradient clipping threshold of 1;

[0245] S32. Preliminarily optimize the training parameters using the Adam optimizer to obtain preliminarily optimized hyperparameters;

[0246] S33. Use a genetic algorithm to globally optimize the hyperparameters after preliminary optimization until the hyperparameter combination corresponding to the individual with the highest fitness value is output, thereby completing the training of the traction motor overcurrent fault identification model.

[0247] The method of using a genetic algorithm to globally optimize the initially optimized hyperparameters includes:

[0248] S331. Set the hyperparameter configuration for each individual in the population size; preferably 20 individuals, with a range of 10 to 50 individuals. Randomly generate the initial population within the defined hyperparameter value range, with each individual corresponding to a set of hyperparameter configuration combinations, including the number of bidirectional LSTM hidden units (30–100), the number of self-attention heads (2–8), the initial learning rate (0.001–0.01), the batch size (16–64), and the L2 regularization coefficient (0.00001–0.001).

[0249] S332. Use the hyperparameter configuration of each individual to build a traction motor overcurrent fault recognition model. Train the traction motor overcurrent fault recognition model using the training set, and calculate the classification accuracy α and Macro-F1 score β on the test set. Calculate the fitness value X based on the classification accuracy α and Macro-F1 score β as follows:

[0250]

[0251] S333. Based on the fitness value X, tournament selection or roulette wheel selection is used to select excellent individuals from the current population size as parents to retain high fitness hyperparameter configuration;

[0252] S334. Perform a crossover operation on the parent individuals, using single-point crossover or uniform crossover, to swap the hyperparameter values ​​of the two parent individuals to generate new crossover individuals to enhance population diversity;

[0253] S335. Randomly perturb and fine-tune the hyperparameter values ​​of the new individuals in the crossover with a mutation probability of 5% to 10% to obtain mutated new individuals to avoid falling into local optimality.

[0254] S336. Add the new individuals from the crossover and mutation to the population to form the next generation and then perform iterations.

[0255] S337. When the number of update iterations reaches the iteration threshold or the change of the fitness value X within 5 consecutive generations is less than 0.001, the global optimization is terminated and the hyperparameter combination corresponding to the individual with the highest fitness value is output. The hyperparameter combination corresponding to the individual with the highest fitness value finally output is used to construct the final traction motor overcurrent fault identification model.

[0256] The traction motor overcurrent fault identification model is designed to extract deep features from complex time series and optimize model hyperparameters using a genetic algorithm (GA) to achieve high accuracy, robustness, and real-time performance. The model's performance is evaluated using classification accuracy, confusion matrices, and Macro-F1 scores on training and test sets. Combining a self-attention mechanism with masking analysis, the model determines the importance of each time step to the prediction result, providing effective support for analyzing the mechanism and localizing traction motor overcurrent faults.

[0257] Example 3

[0258] See also Figure 2 , S4 said real-time signal includes the second A phase current signal , the second B phase current signal , the second intermediate voltage sensor signal and the second speed signal The second characteristic identifier includes a second intermediate voltage sensor signal Minimum value of , the second mean , the second intermediate voltage sensor signal Third-party difference , second speed signal The fourth variance , the second current minimum , the second current maximum value , the second current normalized value , the second current effective value The online fault detection of the real-time signal to obtain a second characteristic identifier with fault location discrimination includes:

[0259] S41. Determine the collected second A-phase current signal and the second B-phase current signal Is the current greater than the current protection threshold within 5 consecutive sampling cycles? If yes, then enter the fault diagnosis phase and execute S42; if no, then continue to collect the second A phase current signal and the second B-phase current signal Perform online fault detection;

[0260] The condition for further judging whether the system has motor overcurrent fault in step S31 is: the current sampling period is If the absolute value of the current sampling data is greater than the current protection threshold in five consecutive sampling cycles, it means that a motor overcurrent fault has occurred in the system, and the system sends an enable diagnosis flag; if the absolute value of the current sampling data cannot be greater than the current protection threshold in five consecutive sampling cycles, it means that the system is in a normal state.

[0261] S42. According to the second intermediate voltage sensor signal and the second speed signal , and obtain the second mean The calculation expression is:

[0262]

[0263] in, k Indicates the running time, Indicates the k Second mean of running time , represents the intermediate variable, = , represents the sliding window size, Indicates the value is and , ;

[0264] According to the second intermediate voltage sensor signal , get the third party difference The calculation expression is:

[0265]

[0266] in, Indicates the k Third-party differences at runtime , Indicates in k Runtime The value is The second mean ;

[0267] According to the second intermediate voltage sensor signal , get the second intermediate voltage sensor signal Minimum value of The calculation expression is:

[0268]

[0269] in, Indicates the k Second intermediate voltage sensor signal at run time Minimum value of , Indicates taking the minimum value, No. k Second intermediate voltage sensor signal at run time , express Second intermediate voltage sensor signal at run time , Indicates the Second intermediate voltage sensor signal at run time ;

[0270] According to the second speed signal , and get the fourth variance The calculation expression is:

[0271]

[0272] in, Indicates the k Runtime fourth variance , Indicates ink Runtime The value is The second mean ;

[0273] According to the second A-phase current signal and the second B-phase current signal , calculate the second current minimum The calculation expression is:

[0274]

[0275] in, Indicates the k The second minimum current during operation , y Indicates the value is a and b The subscript of Indicates in k The second A phase current signal at the time of operation and the second B-phase current signal , Indicates in The second A phase current signal at the time of operation and the second B-phase current signal , exist The second A phase current signal at the time of operation and the second B-phase current signal ;

[0276] According to the second A-phase current signal and the second B-phase current signal , calculate the second current maximum value The calculation expression is:

[0277]

[0278] in, Indicates taking the maximum value;

[0279] According to the second A-phase current signal and the second B-phase current signal , calculate the second current effective value The calculation expression is:

[0280]

[0281] in, Indicates the k The second current effective value during operation ;

[0282] According to the second A-phase current signal and the second B-phase current signal , calculate the second current normalized value The calculation expression is:

[0283]

[0284] in, Indicates the k The second current normalized value at the time of operation .

[0285] Extract the fault data segment with a data length of 17500, first perform fast Fourier transform on the extracted current data to obtain the current frequency , determine the sliding window size according to the current frequency and step length , and finally get Sliding windows and their number on the right side of the traction control unit's action point and data length ; The obtained sliding window size and step length On this basis, the identifier is calculated, and the second feature identifier is calculated in each window to obtain the second feature index 、 、 、 、 .

[0286] In step S5, the second feature identifier is segmented within the sliding window to obtain the second feature indices, which are respectively calculated as follows:

[0287]

[0288]

[0289]

[0290]

[0291]

[0292] in, 、 、 、 、 represents the second characteristic index, 、 、 、 、 Indicates the k The second characteristic indicator of runtime.

[0293] S5, using the second characteristic indicator to generate a corresponding second event, performing sequence conversion on the second event to obtain a template vector, includes:

[0294] S51. The second characteristic index 、 、 、 、 are input into the second hysteresis comparator, and the second hysteresis comparator outputs the corresponding second event ;

[0295] S52. The second event Press the X The order forms a binary time series event , the binary timing events Convert to decimal The following mathematical expression:

[0296]

[0297] in, Indicates the current calculated decimal number The sliding window position, (1, m ), m represents the length of the second vector, ;

[0298] S53. Based on the decimal number , get the template vector The calculation expression is as follows:

[0299]

[0300] in, Indicates that the number of sliding windows is the length of the second vector m The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of

[0301] The calculation expression is:

[0302]

[0303] The calculation expression is:

[0304] .

[0305] See also Figure 2 , S6 converts the online template vector The normalized 16-dimensional template vector is input into the trained traction motor overcurrent fault recognition model, and the trained traction motor overcurrent fault recognition model outputs the fault recognition result.

[0306] The present invention seeks to protect a method for binary encoding the feature identifiers in each sliding window and converting the resulting binary sequence into a one-dimensional normalized decimal time series. This process effectively reduces the data dimension, significantly reduces the amount of computation, improves computational efficiency and real-time performance, and ensures data consistency across the numerical range, thereby significantly improving the recognition efficiency of the traction motor overcurrent fault identification model.

[0307] The present invention includes two parts: offline modeling and online diagnosis. In the offline stage, the fault mechanism analysis is first performed on the current signal, voltage signal and speed signal in the historical data, and characteristic indicators such as minimum value, mean value and variance are extracted; then the current frequency is calculated by fast Fourier transform to determine the sliding window; the characteristic indicators are calculated within the sliding window, and the hysteresis comparator is used to generate characteristic events, and finally a template library containing different fault types is constructed. In the online stage, the current signal is collected in real time and compared with the protection threshold. If a continuous overcurrent phenomenon is detected, the fault diagnosis process is triggered. Subsequently, the cached data segment is extracted, the sliding window size and step size are adjusted according to the current frequency, the real-time characteristic indicators are calculated within the window, and the corresponding event sequence is generated. The sequence is converted from binary to a normalized decimal sequence. Compared with the prior art, the advantages of the present invention are

[0308] 1. Traction drive systems are difficult to construct mathematical models for. Existing diagnostic methods based on these models struggle to accurately represent the system and are susceptible to interference that could affect diagnostic results. Furthermore, traditional fault diagnosis methods based on the dynamic time warping (DTW) algorithm are computationally intensive and time-consuming when processing multidimensional data, making them difficult to meet real-time requirements. This invention, however, does not rely on a single mathematical model. By converting multidimensional binary time series into a normalized one-dimensional decimal time series, it can more stably and accurately diagnose motor overcurrent faults, reducing the impact of interference on diagnostic results.

[0309] 2. The fixed sliding window size mode in existing diagnostic methods is inefficient when processing large amounts of data and requires high hardware performance. This invention dynamically adjusts the sliding window size and step size based on the current frequency, effectively reducing the amount of calculation, improving data processing efficiency, reducing dependence on hardware performance, and thus improving maintenance efficiency.

[0310] 3. The present invention's dynamic adjustment of the sliding window size and step size effectively reduces dependence on hardware performance and optimizes processing efficiency. Compared to traditional methods with fixed window sizes, it can more flexibly adapt to data characteristics under different working conditions, further improving computational efficiency and fault detection accuracy.

[0311] 4. The method proposed in this paper can accurately locate four types of motor overcurrent faults, including speed signal fault, traction motor fault, intermediate DC voltage signal fault, and converter module fault. Compared with some existing methods that can only diagnose a single fault source type or inaccurately diagnose some faults, the comprehensiveness and accuracy of fault diagnosis are significantly improved.

[0312] It's also worth noting that, in addition to the currently established feature identification and indicator calculation methods, Hidden Markov Models (HMMs) can be used for feature extraction. HMMs can describe the transition process of system states and the probability of observation in different states, making them useful for analyzing state changes during motor overcurrent faults. However, because HMMs require accurate estimation of model parameters, they place high demands on system state assumptions. Therefore, practical adjustments and optimizations may be necessary based on specific circumstances.

[0313] Example 4

[0314] See also Figure 4 This embodiment also proposes a traction motor overcurrent fault identification system, including:

[0315] an offline preprocessing module, configured to obtain historical signals of the traction drive system, perform offline preprocessing on the historical signals, and obtain a first feature identifier with fault location discrimination capability;

[0316] an offline segmentation and conversion module, configured to segment the first feature identifier within a sliding window to obtain a first feature index, generate a corresponding first event using the first feature index, and perform sequence conversion on the first event to obtain a traction motor overcurrent fault location and recognition template;

[0317] a traction motor overcurrent fault identification model construction module, configured to construct a traction motor overcurrent fault identification model, divide the traction motor overcurrent fault location identification template into a training set and a test set, train the traction motor overcurrent fault identification model using the training set, and test the effectiveness of the traction motor overcurrent fault identification model using the test set, thereby obtaining a trained traction motor overcurrent fault identification model;

[0318] An online preprocessing module, configured to obtain a real-time signal of the traction drive system, perform online fault detection on the real-time signal, and obtain a second characteristic identifier with fault location discrimination;

[0319] an online segmentation and conversion module, configured to segment the second feature identifier within a sliding window to obtain a second feature index, generate a corresponding second event using the second feature index, and perform sequence conversion on the second event to obtain an online template vector;

[0320] The output module is used to input the online template vector into the trained traction motor overcurrent fault recognition model and output the fault recognition result.

[0321] In this embodiment, first, in the offline phase, the historical signal is preprocessed offline to obtain a first feature identifier. The first feature identifier is segmented based on the sliding window feature data to obtain a first feature index. The first feature index is used to generate a corresponding first event, and the first event is converted into a sequence. The purpose is to reduce the offline data dimension to one dimension, thereby effectively reducing the computational complexity. Then, a traction motor overcurrent fault recognition model is constructed and trained to obtain a trained traction motor overcurrent fault recognition model. Then, online fault detection is performed on the real-time signal to obtain a second feature identifier with fault location discrimination. The second feature identifier is segmented within the sliding window to obtain a second feature index. The second feature index is used to generate a corresponding second event, and the second event is converted into a sequence. The purpose is to reduce the online data dimension to one dimension, thereby further effectively reducing the computational complexity. Finally, the online template vector is input into the trained traction motor overcurrent fault recognition model to achieve rapid and accurate fault type discrimination, significantly improving the efficiency and accuracy of real-time fault recognition. The present invention can improve the speed and real-time performance of data matching through the traction motor overcurrent fault recognition model while ensuring fault recognition accuracy, thereby better meeting the needs of real-time diagnosis of motor overcurrent faults.

[0322] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. A person skilled in the art would be able to make other variations or modifications based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying overcurrent faults in a traction motor, characterized in that: The following steps are involved: S1. Obtain historical signals of the traction drive system, perform offline preprocessing on the historical signals, and obtain a first characteristic identifier with fault location discrimination; S2. Segmenting the first feature identifier within a sliding window to obtain a first feature index, generating a corresponding first event using the first feature index, and performing sequence conversion on the first event to obtain a traction motor overcurrent fault location and identification template, including: S21. The first characteristic index 、 、 、 、 are input into the first hysteresis comparator respectively, and the first hysteresis comparator outputs the corresponding first event ; S22. The first event Press the X The order forms a binary time series vector , the binary time series vector Convert to decimal ; S23. Based on the decimal number , get the traction motor overcurrent fault location identification vector The calculation expression is as follows: in, Indicates that the number of sliding windows is the length of the first vector The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of The calculation expression is: The calculation expression is: S24. The traction motor overcurrent fault location identification vector obtained under different fault types Combining the collection to obtain the traction motor overcurrent fault location and identification template; S3. Construct a traction motor overcurrent fault recognition model, divide the traction motor overcurrent fault location recognition template into a training set and a test set, use the training set to train the traction motor overcurrent fault recognition model, and use the test set to test the effectiveness of the traction motor overcurrent fault recognition model, to obtain a trained traction motor overcurrent fault recognition model; S4 obtains a real-time signal of the traction drive system, performs online fault detection on the real-time signal, and obtains a second characteristic identifier having a fault location distinction; S5. Segmenting the second feature identifier within the sliding window to obtain a second feature index, generating a corresponding second event using the second feature index, and performing sequence conversion on the second event to obtain an online template vector; S6. Input the online template vector into the trained traction motor overcurrent fault recognition model and output the fault recognition result.

2. The traction motor overcurrent fault identification method according to claim 1, characterized in that: The historical signal includes the first A phase current signal , the first B phase current signal , the first intermediate voltage sensor signal and the first speed signal , the first feature identifier includes a first mean , the first intermediate voltage sensor historical signal The first variance , the first intermediate voltage sensor signal Minimum value of , first speed signal The second variance , the first current minimum , the first current maximum value , the first current effective value and the first current normalized value ; The offline preprocessing of the historical signal to obtain a first feature identifier with fault location discrimination includes: According to the historical signal of the first intermediate voltage sensor and the first speed signal , get the first mean The calculation expression is: in, k Indicates the running time, Indicates the k The first mean of running time , represents the intermediate variable, = , represents the sliding window size, Indicates the value is and , ; According to the historical signal of the first intermediate voltage sensor , get the first variance The calculation expression is: in, Indicates the k First variance of running time , Indicates in k Runtime The value is The first mean ; According to the historical signal of the first intermediate voltage sensor , get the first intermediate voltage sensor signal Minimum value of The calculation expression is: in, Indicates the k The first intermediate voltage sensor signal at the time of operation Minimum value of , Indicates taking the minimum value, No. k Historical signal of the first intermediate voltage sensor at the time of operation , express Historical signal of the first intermediate voltage sensor at the time of operation , Indicates the Historical signal of the first intermediate voltage sensor at the time of operation ; According to the first speed signal , and get the second variance The calculation expression is: in, Indicates the k Run-time second variance , Indicates in k Runtime The value is The first mean ; According to the first A-phase current signal and the first B-phase current signal , calculate the first current minimum value The calculation expression is: in, Indicates the k The first minimum current value during operation , y Indicates the value is a and b The subscript of Indicates in k The first A-phase current signal at the time of operation and the first B-phase current signal , Indicates in The first A-phase current signal at the time of operation and the first B-phase current signal , exist The first A-phase current signal at the time of operation and the first B-phase current signal ; According to the first A-phase current signal and the first B-phase current signal , calculate the maximum current The calculation expression is: in, Indicates taking the maximum value; According to the first A-phase current signal and the first B-phase current signal , calculate the first current effective value The calculation expression is: in, Indicates the k The first effective current value at the time of operation ; According to the first A-phase current signal and the first B-phase current signal , calculate the first current normalized value The calculation expression is: in, Indicates the k The first normalized current value at the time of operation .

3. The traction motor overcurrent fault identification method according to claim 2, characterized in that: Acquire sliding window characteristic data within the sliding window, wherein the sliding window characteristic data includes the current frequency of the traction motor , sliding window size , step length , Number of sliding windows , the number of units distributed on the left side of the traction control unit's action point , the number of units distributed on the right side of the traction control unit's action point and data length ; The sliding window size The calculation expression is: in, Indicates the sampling period; The step length The calculation expression is: The number of the distribution on the left side of the traction control unit action point The calculation expression is: The number of the distribution on the right side of the traction control unit action point The calculation expression is: The data length The calculation expression is: The first feature identifier is segmented according to the sliding window feature data to obtain the first feature indexes, which are calculated as follows: in, 、 、 、 、 represents the first characteristic index, 、 、 、 、 Indicates the k The first characteristic indicator of runtime.

4. The traction motor overcurrent fault identification method according to claim 3, characterized in that: The traction motor overcurrent fault recognition model includes a sequence input layer, a bidirectional long short-term memory network layer, a self-attention layer, a first fully connected layer, a first activation function layer, a dropout layer, a second fully connected layer, a second activation function layer and a classification layer connected in sequence; The traction motor overcurrent fault location and identification template is normalized into a 16-dimensional template vector and input into the sequence input layer. The sequence input layer outputs time series data. The bidirectional long short-term memory network layer receives the time series data and outputs the bidirectional dependency features of the time series. The self-attention layer receives the bidirectional dependency features and outputs the time step feature data weighted by attention. The first fully connected layer receives the time step feature data and outputs the first linear feature combination result. The first activation function layer receives the first linear feature combination result and outputs the first activation feature result. The discard layer receives the first activation feature result and outputs the feature result after discarding. The second fully connected layer receives the feature result after discarding and outputs the second linear feature combination result. The second activation function layer receives the second linear feature combination result and outputs the fault type prediction probability. The classification layer receives the fault type prediction probability and outputs the fault type identification result.

5. The traction motor overcurrent fault identification method according to claim 4, characterized in that: The training of the traction motor overcurrent fault identification model includes: S31. Set the training parameters of the traction motor overcurrent fault identification model, the training parameters including: maximum rounds of 150, batch size of 32, initial learning rate of 0.005, learning rate scheduling of 0.2 times every 50 rounds, L2 regularization coefficient of 0.0001, and gradient clipping threshold of 1; S32. Preliminarily optimize the training parameters using the Adam optimizer to obtain preliminarily optimized hyperparameters; S33. Use a genetic algorithm to globally optimize the hyperparameters after preliminary optimization until the hyperparameter combination corresponding to the individual with the highest fitness value is output, thereby completing the training of the traction motor overcurrent fault identification model.

6. The traction motor overcurrent fault identification method according to claim 5, characterized in that: The method of using a genetic algorithm to globally optimize the initially optimized hyperparameters includes: S331. Set the hyperparameter configuration of each individual in the population size; S332. Use the hyperparameter configuration of each individual to build a traction motor overcurrent fault recognition model. Train the traction motor overcurrent fault recognition model using the training set, and calculate the classification accuracy α and Macro-F1 score β on the test set. Calculate the fitness value X based on the classification accuracy α and Macro-F1 score β as follows: X 0.7α+0.3β S333. Selecting a superior individual from the current population size as a parent using tournament selection or roulette wheel selection based on the fitness value X; S334. Perform a crossover operation on the parent individuals, using single-point crossover or uniform crossover, to swap the hyperparameter values ​​of the two parent individuals to generate a new crossover individual; S335. Randomly perturb and fine-tune the hyperparameter values ​​of the new individuals in the crossover with a mutation probability of 5% to 10% to obtain mutated new individuals. S336. Add the new individuals from the crossover and mutation to the population to form the next generation and then perform iterations. S337. When the number of update iterations reaches the iteration threshold or the change in the fitness value X is less than 0.001 within 5 consecutive generations, the global optimization is terminated and the hyperparameter combination corresponding to the individual with the highest fitness value is output.

7. The traction motor overcurrent fault identification method according to claim 6, characterized in that: The real-time signal includes a second A-phase current signal , the second B phase current signal , the second intermediate voltage sensor signal and the second speed signal The second characteristic identifier includes a second intermediate voltage sensor signal Minimum value of , the second mean , the second intermediate voltage sensor signal Third-party difference , second speed signal The fourth variance , the second current minimum , the second current maximum value , the second current normalized value , the second current effective value ; The performing online fault detection on the real-time signal to obtain a second characteristic identifier with fault location discrimination includes: S41. Determine the collected second A-phase current signal and the second B-phase current signal Is the current greater than the current protection threshold in several consecutive sampling cycles? If yes, then enter the fault diagnosis phase and execute S42; if no, then continue to collect the second A phase current signal and the second B-phase current signal Perform online fault detection; S42. According to the second intermediate voltage sensor signal and the second speed signal , and obtain the second mean The calculation expression is: in, k Indicates the running time, Indicates the k Second mean of running time , represents the intermediate variable, = , represents the sliding window size, Indicates the value is and , ; According to the second intermediate voltage sensor signal , get the third party difference The calculation expression is: in, Indicates the k Third-party differences at runtime , Indicates in k Runtime The value is The second mean ; According to the second intermediate voltage sensor signal , get the second intermediate voltage sensor signal Minimum value of The calculation expression is: in, Indicates the k Second intermediate voltage sensor signal at run time Minimum value of , Indicates taking the minimum value, No. k Second intermediate voltage sensor signal at run time , express Second intermediate voltage sensor signal at run time , Indicates the Second intermediate voltage sensor signal at run time ; According to the second speed signal , and get the fourth variance The calculation expression is: in, Indicates the k Runtime fourth variance , Indicates in k Runtime The value is The second mean ; According to the second A-phase current signal and the second B-phase current signal , calculate the second current minimum The calculation expression is: in, Indicates the k The second minimum current during operation , y Indicates the value is a and b The subscript of Indicates in k The second A phase current signal at the time of operation and the second B-phase current signal , Indicates in The second A phase current signal at the time of operation and the second B-phase current signal , exist The second A phase current signal at the time of operation and the second B-phase current signal ; According to the second A-phase current signal and the second B-phase current signal , calculate the second current maximum value The calculation expression is: in, Indicates taking the maximum value; According to the second A-phase current signal and the second B-phase current signal , calculate the second current effective value The calculation expression is: in, Indicates the k The second current effective value during operation ; According to the second A-phase current signal and the second B-phase current signal , calculate the second current normalized value The calculation expression is: in, Indicates the k The second current normalized value at the time of operation .

8. The traction motor overcurrent fault identification method according to claim 7, characterized in that: The second feature identifier is segmented within the sliding window to obtain the second feature indexes, which are calculated as follows: in, 、 、 、 、 represents the second characteristic index, 、 、 、 、 Indicates the k The second characteristic indicator of the running time; Generating a corresponding second event using the second characteristic indicator, performing sequence conversion on the second event, and obtaining an online template vector, including: S51. The second characteristic index 、 、 、 、 are input into the second hysteresis comparator, and the second hysteresis comparator outputs the corresponding second event ; S52. The second event Press the X The order forms a binary time series event , the binary timing events Convert to decimal The following mathematical expression: in, Indicates the current calculated decimal number The sliding window position, (1, m ), m represents the length of the second vector, ; S53. Based on the decimal number , get the online template vector The calculation expression is as follows: in, Indicates that the number of sliding windows is the length of the second vector m The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of The calculation expression is: The calculation expression is: 。 9. A traction motor overcurrent fault identification system, characterized in that: include: an offline preprocessing module, configured to obtain historical signals of the traction drive system, perform offline preprocessing on the historical signals, and obtain a first feature identifier with fault location discrimination capability; An offline segmentation and conversion module is configured to segment the first feature identifier within a sliding window to obtain a first feature index, generate a corresponding first event using the first feature index, and perform sequence conversion on the first event to obtain a traction motor overcurrent fault location and recognition template, including: S21. The first characteristic index 、 、 、 、 are input into the first hysteresis comparator respectively, and the first hysteresis comparator outputs the corresponding first event ; S22. The first event Press the X The order forms a binary time series vector , the binary time series vector Convert to decimal ; S23. Based on the decimal number , get the traction motor overcurrent fault location identification vector The calculation expression is as follows: in, Indicates that the number of sliding windows is the length of the first vector The elements of the vector, Indicates a decimal number The minimum value of Indicates a decimal number The maximum value of The calculation expression is: The calculation expression is: S24. The traction motor overcurrent fault location identification vector obtained under different fault types Combining the collection to obtain the traction motor overcurrent fault location and identification template; a traction motor overcurrent fault identification model construction module, configured to construct a traction motor overcurrent fault identification model, divide the traction motor overcurrent fault location identification template into a training set and a test set, train the traction motor overcurrent fault identification model using the training set, and test the effectiveness of the traction motor overcurrent fault identification model using the test set, thereby obtaining a trained traction motor overcurrent fault identification model; An online preprocessing module, configured to obtain a real-time signal of the traction drive system, perform online fault detection on the real-time signal, and obtain a second characteristic identifier with fault location discrimination; an online segmentation and conversion module, configured to segment the second feature identifier within a sliding window to obtain a second feature index, generate a corresponding second event using the second feature index, and perform sequence conversion on the second event to obtain an online template vector; The output module is used to input the online template vector into the trained traction motor overcurrent fault recognition model and output the fault recognition result.

Images