Intelligent Diagnosis Method for Train Traction Motor Defects Based on Multimodal Fusion

Abstract

This invention discloses an intelligent diagnostic method for train traction motor defects based on multimodal fusion, relating to the fields of pre-diagnosis and health management. It addresses the problem of insufficient accuracy in diagnosing train traction motor defects, and includes the following steps: Step S1: Acquire the motor's operating status data; Step S2: Acquire a baseline data list and a defect data list; calculate the defect data difference between the baseline data list and the defect data list, and map the defect data difference to obtain a mapped value; Step S3: Perform feature calculation to obtain the feature values ​​of the motor defects, and sort the various operating data of the motor according to the feature values; obtain the mapping interval of the mapped value, and construct a polar coordinate image by combining the mapped value and the sorting result; Step S4: Acquire the motor's real-time data; perform a similarity test between the real-time data image and the feature image of the defective motor; This invention can effectively improve the accuracy of motor defect diagnosis.

Description

Technical Field

[0001] This invention pertains to fault prediction and health management, specifically to the diagnosis of defects in train traction motors, and more specifically to a multimodal fusion-based intelligent diagnostic method for train traction motor defects. Background Technology

[0002] Intelligent diagnostic methods for train traction motor defects fall under the category of fault prediction and health management. By intelligently diagnosing defects in train traction motors, they aim to ensure train safety. However, existing intelligent diagnostic methods for train traction motor defects have the following specific shortcomings during the diagnostic process:

[0003] 1. Most existing methods for diagnosing defects in train traction motors rely on a single data source, resulting in insufficient diagnostic accuracy and difficulty in effectively inspecting motor defects affected by multiple factors. Furthermore, the method of diagnosing motors by maintenance personnel makes the diagnostic results highly susceptible to human error, easily leading to misdiagnosis or missed diagnosis.

[0004] 2. The existing train traction motor defect diagnosis mainly relies on real-time monitoring of the train traction motor. When an abnormality occurs, the real-time monitoring data is checked for abnormality and the train traction motor is diagnosed. However, the diagnosis is severely delayed and it is difficult to effectively prevent and judge the problem before it occurs.

[0005] 3. Existing methods for diagnosing defects in train traction motors lack intelligent processing capabilities. When dealing with large amounts of data, existing methods struggle to achieve rapid and accurate analysis and processing, resulting in relatively low diagnostic efficiency. This inefficient diagnostic process not only consumes a significant amount of human and time resources but also affects the normal operation of the train.

[0006] To address this, we propose a multimodal fusion-based intelligent diagnostic method for train traction motor defects. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide an intelligent diagnostic method for train traction motor defects based on multimodal fusion, aiming to improve the accuracy of intelligent diagnosis of train traction motor defects.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: a multimodal fusion-based intelligent diagnostic method for train traction motor defects, the specific working process of each step of which is as follows:

[0009] Step S1: Monitor the motor operation data under different conditions to obtain the motor's operating status data;

[0010] Step S2: Based on the motor's operating status data, obtain the baseline data list and the defect data list; calculate the defect data difference between the baseline data list and the defect data list, and perform data mapping on the defect data difference to obtain the mapped value;

[0011] Step S3: Perform feature calculation on the motor operation data based on the mapping value to obtain the feature value of the motor defect, and sort the various motor operation data according to the feature value; obtain the mapping interval of the mapping value, and combine the mapping value and the sorting result to construct a polar coordinate image to obtain the feature image of the defective motor;

[0012] Step S4: Obtain real-time data of the motor, map the real-time data of the motor to construct a real-time data image; perform similarity test between the real-time data image and the feature image of the defective motor to diagnose the defect of the motor.

[0013] Furthermore, the specific steps of step S1 are as follows:

[0014] Step S11: Acquire the normal operating data of the motor as the baseline data; obtain the number of motor operating data items in the baseline data, denoted as as; based on the number of motor operating data items, obtain each motor operating data item, denoted as sj. a , a∈[1,as]; for sj a Statistical analysis was performed to obtain a baseline data list (jsl).

[0015] Step S12: Obtain the type of motor defect, denoted as bs. Based on the type of motor defect, obtain the abnormal operation data. Combine the number of motor operation data items in the baseline data and denote the abnormal operation data as qx. a (b), b∈[1, bs]; Statistical analysis of abnormal operation data yields a defect data list qsl(b);

[0016] The motor's operating status data consists of a list of baseline data for the motor and a list of defect data.

[0017] Furthermore, the specific steps of step S2 are as follows:

[0018] Step S21: Based on the baseline data list, obtain each operating data item sj of the normal motor. a Based on the defect data list, obtain each operational data item (qx) of the defective motor. a (b); Run the data qx a (b) with sj a The difference is calculated to obtain the defect data difference value qcz. a (b);

[0019] Step S22: Map the defect data differences to extract the defect data differences for different types of motor defects; map the defect data differences according to their data distribution to obtain the mapped values; specifically as follows:

[0020] Extract the defect data differences for different types of motor defects to obtain a difference list clb. a ,clb a =[qcz a (1), qcz a (2), ..., qcz a (bs)]; Get the maximum value in the difference list, denoted as mx a Find the minimum value in the difference list, denoted as mi. a ;

[0021] Based on the maximum and minimum values ​​in the difference list, calculate the difference qcz for all defect data in the difference list. a (b) Perform data mapping to obtain the mapped value ysz;

[0022] ;

[0023] Among them: ysz a (b) represents the mapping value of the b-th data in the a-th difference list;

[0024] The mapped values ​​are statistically analyzed to obtain a list of mapped values, ylb.

[0025] Furthermore, the specific steps of step S3 are as follows:

[0026] Step S31: Obtain the mapping value list, sum the absolute values ​​of each row of mapping values ​​according to the mapping value list to obtain the feature values ​​of the running data; sort the feature values ​​in descending order; obtain the feature value sorting list; obtain the sorting number in the feature value sorting list and associate it with each item of running data;

[0027] Step S32: Obtain the mapping range of the mapping value; based on the mapping range and the sorting result, set the initial value; obtain the number of items of the motor operation data; set the initial angle of the mapping value; and represent the reference data in coordinates based on the initial value and the initial angle.

[0028] Step S33: Obtain the mapping value, and calculate the coordinates of various data of different types of defective motors based on the mapping value and the initial value; connect the coordinates of various data of different types of defective motors to obtain the feature image of the defective motor.

[0029] Furthermore, the specific steps of step S31 are as follows:

[0030] Step S311: Extract the mapping values ​​from the mapping value list to obtain the mapping value ysz. a (1) to ysz a (bs); for the mapping value ysz a (1) to ysz a (bs) performs absolute value summation to obtain the characteristic value of each running data item, denoted as the characteristic value tzz. a ;

[0031] Step S312: Perform statistical analysis on the eigenvalues ​​to obtain tzz1 to tzz as For eigenvalues ​​tzz1 to tzz as Sort the data in descending order to obtain a feature value sorting list. Based on the feature value sorting list, obtain the sorting number of the feature value of each data item in the sorting list, denoted as px(a).

[0032] Furthermore, the specific steps of step S32 are as follows:

[0033] Step S321: Denote the mapping range of the mapped values ​​as qjf; obtain the sorting index px(a) of the feature value of each running data in the sorting list; set the initial value according to the mapping range gjf and the sorting index px(a); denote the initial value as csz;

[0034] ;

[0035] Among them: csz a This represents the initial value of the data in the a-th item.

[0036] Step S322: Based on the number of items in the motor operation data as, set the initial angle of the mapping value. Divide the circumference angle equally according to the number of items in the motor operation data to obtain as equidistant lines. Combine the mapping value with the sorting number px(a) of the feature value of each operation data item in the sorting list to make the mapping value lie on different equidistant lines, and obtain the initial angle of the mapping value. Record the initial angle as cjd.

[0037] ;

[0038] Among them: cjd a This represents the initial angle of the data in item a;

[0039] Step S323: Perform statistical analysis on the initial values ​​and initial angles to obtain the initial coordinates jzb of the reference data. a jzb a = (csz) a cjs a ).

[0040] Furthermore, the specific steps of step S33 are as follows:

[0041] Step S331: Extract the column data from the mapping value list to obtain the mapping values ​​of various operating data of the defective motor; obtain the initial values ​​of various operating data, and deflect the initial values ​​according to the mapping values ​​to obtain the coordinate length of the defective motor;

[0042] The angle of the defective motor is obtained based on the initial angle, thus obtaining the coordinate angle of the defective motor;

[0043] The defect coordinates of the defective motor are obtained from its coordinate length and coordinate angle.

[0044] Step S332: Statistically analyze the defect coordinates of the defective motors, connect the defect coordinates of the defective motors to obtain the feature image of the defective motors; store the feature images of all defective motors; extract features from the feature images of the defective motors.

[0045] Furthermore, the specific steps of step S332 are as follows:

[0046] Obtain the defect coordinates qzb of the defective motor a (b) = (qcd) a (b), zbj a (b) The defect area qmj(b) is obtained by calculating the area enclosed by the defective motor based on the defect coordinates of the defective motor.

[0047] ;

[0048] Among them: qcd as qcd1 represents the coordinate length of the 'as'th coordinate in the real-time data image, and qcd1 represents the coordinate length of the '1'th coordinate in the real-time data image; 2Π / as is a constant value for the angle change between adjacent coordinates.

[0049] Based on the defect coordinates of the defective motor, the coordinate length qcd of the defect coordinates is... a (b) with coordinate angle zbj a (b) Integrate the coordinate characteristic values ​​qzt(b) of the defective motor;

[0050] ;

[0051] The coordinate length qcd of the defect coordinates a (b) with coordinate angle zbj a (b) Integrate and summarize the characteristics of defective motors.

[0052] Furthermore, the specific steps of step S4 are as follows:

[0053] Step S41: Obtain the mapping ratio of the data mapping, map the real-time data according to the mapping ratio, transform the coordinates of the mapped data to obtain the coordinate points of the real-time data, connect the coordinate points of the real-time data, and construct a real-time data image.

[0054] Step S42: Calculate the similarity between the real-time data image and the feature image of the defective motor, perform defect diagnosis on the motor based on the real-time data image and the feature image of the defective motor, store the real-time data image and the diagnosis results, and optimize the feature image of the defective motor.

[0055] Furthermore, the specific steps of step S42 are as follows:

[0056] Step S421: Obtain the coordinates of the real-time data image based on the real-time data image, denoted as sst. a = (scd) a sjd a ), where scd a For coordinate length, sjd a The coordinate angle is used; the area enclosed by the image is calculated based on the coordinates of the real-time data image to obtain the real-time area smj;

[0057] The coordinates of the real-time data image, and the coordinate length scd of the real-time image. a With coordinate angle sjd a The coordinate feature values ​​stz of the real-time image are then integrated.

[0058] Step S422: Obtain the defect area qmj(b) and coordinate feature value qzt(b) of the defective motor. Combine the real-time area smj and coordinate feature value sts of the real-time image to calculate the similarity between the real-time image and the feature image, and obtain the similarity value xsz.

[0059] ;

[0060] Where: xsz(b) represents the similarity value between the real-time image and the defect image of the b-th defective motor;

[0061] The defective motor corresponding to the largest similarity value is obtained to obtain the motor's pre-diagnosis result;

[0062] Based on the motor's preliminary diagnosis results, the motor is inspected. If the inspection results are consistent with the preliminary diagnosis results, the motor diagnosis is completed. If the inspection results are inconsistent with the preliminary diagnosis results, the inspection results are used as new motor defects to supplement the existing motor defects, and the real-time data is stored as the characteristic data of the defective motor.

[0063] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0064] 1. The intelligent diagnostic method for train traction motor defects based on multimodal fusion disclosed in this invention acquires the motor's operating status data, combines a benchmark data list with a defect data list, performs data mapping and feature calculation, and constructs a polar coordinate image, thereby realizing intelligent diagnosis of motor defects. This not only improves the accuracy of diagnosis but also reduces the risk of misdiagnosis and missed diagnosis.

[0065] 2. This invention monitors the motor's operating status in real time, compares the motor's operating status with polar coordinate images, and uses similarity analysis to proactively prevent motor malfunctions, promptly identify potential problems, effectively prevent motor failures, and ensure train safety.

[0066] 3. By extracting image features in polar coordinates, the processing speed of similarity testing is optimized, the testing efficiency is improved, the processing needs of large amounts of data are met, and the time and computing power costs required for data processing are reduced; this provides strong technical support for the health management and maintenance of train traction motors. Attached Figure Description

[0067] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0068] Figure 1 This is a schematic diagram of the method of the present invention;

[0069] Figure 2 This is a schematic diagram of the data processing of the present invention;

[0070] Figure 3 This is a schematic diagram of image construction for the present invention;

[0071] Figure 4 This is a schematic diagram of image comparison for the present invention; Detailed Implementation

[0072] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0073] Example 1

[0074] Please see Figure 1 The train traction motor defect diagnosis method of the present invention belongs to pre-diagnosis and health management, and provides an intelligent diagnosis method for train traction motor defects based on multimodal fusion, including:

[0075] Step S1: Monitor the motor operation data under different conditions to obtain the motor's operating status data;

[0076] It should be noted that the operating data of the motor includes: the effective value of the voltage of each phase, the effective value of the current of each phase, the three-phase active power, the harmonics and distortion rate of the current of each phase, the motor speed and torque, the voltage frequency, the voltage zero-negative sequence unbalance, and the harmonics and distortion rate of the voltage of each phase.

[0077] Step S11: Acquire the normal operating data of the motor during normal operation as the baseline data; obtain the number of motor operating data items in the baseline data, denoted as as; based on the number of motor operating data items, obtain each motor operating data item, denoted as sj. a ; where: sj a This represents the motor data for the a-th item, where a ∈ [1, as].

[0078] For SJ a Statistical analysis was performed to obtain a baseline data list (jsl).

[0079] Step S12: Obtain the type of motor defect, denoted as bs. Based on the type of motor defect, obtain the abnormal operating data for each type of motor defect. Combine this with the number of motor operating data items in the baseline data, and denote the abnormal operating data as qx. a (b); where: qx a (b) represents the a-th operational data of the b-th type of motor defect, where b∈[1, bs];

[0080] Statistical analysis of abnormal operation data yields a defect data list qsl(b);

[0081] The motor's baseline data list and defect data list together constitute the motor's operating status data.

[0082] Step S2: Based on the motor's operating status data, obtain the baseline data list and the defect data list; calculate the defect data difference between the baseline data list and the defect data list, and perform data mapping on the defect data difference to obtain the mapped value;

[0083] Step S21: Based on the baseline data list, obtain each operating data item sj of the motor during normal operation. a Based on the defect data list, when obtaining motor defects, each piece of motor operating data qx is obtained. a (b); Run the data qx a (b) with sj a The difference is calculated to obtain the defect data difference value qcz. a (b);

[0084] ;

[0085] It should be noted that comparing abnormal operating data with normal operating data through the difference can better reflect the data characteristics when the motor has defects.

[0086] Step S22: Map the defect data difference by extracting the defect data difference of the same operating data of different types of motor defects; map the defect data difference according to the data distribution of the defect data difference to obtain the mapped value of the defect data difference;

[0087] Step S221: Extract the defect data differences of the same operating data for different types of motor defects to obtain the difference list clb. a ,clb a =[qcz a (1), qcz a (2), ..., qcz a (bs)]; Get the maximum value in the difference list, denoted as mx a Find the minimum value in the difference list, denoted as mi. a ;

[0088] Step S222: Based on the maximum and minimum values ​​in the difference list, calculate the difference qcz for all defect data in the difference list. a (b) Perform data mapping to obtain the mapped value ysz;

[0089] ;

[0090] Among them: ysz a (b) represents the mapping value of the b-th data in the a-th difference list;

[0091] The mapped values ​​are statistically analyzed to obtain a list of mapped values, ylb.

[0092] ;

[0093] It should be noted that: each column in the mapping value list represents the mapping value of various operating data for the same type of motor defect, and each row represents the mapping value of the same operating data in different motor defects;

[0094] It should be noted that by mapping the differences, the correlation between various operating data and defective motors can be displayed intuitively, thereby enhancing the accuracy of motor diagnosis.

[0095] Step S3: Perform feature calculation on the motor operation data based on the mapping value to obtain the feature value of the motor defect, and sort the various motor operation data according to the feature value; obtain the mapping interval range of the mapping value, and combine the mapping value and the sorting result to construct a polar coordinate image to obtain the feature image of the defective motor;

[0096] Please see Figure 2 Step S31: Obtain the mapping value list, sum the absolute values ​​of each row of mapping values ​​according to the mapping value list to obtain the feature values ​​of all motor defects for each running data; sort the feature values ​​in descending order; obtain the feature value sorting list; obtain the sorting number in the feature value sorting list and associate it with each running data.

[0097] Step S311: Extract the mapping values ​​from the mapping value list to obtain the mapping value ysz. a (1) to ysz a (bs); for the mapping value ysz a (1) to ysz a (bs) performs absolute value summation to obtain the characteristic value of each running data item, denoted as the characteristic value tzz. a ;

[0098] ;

[0099] Among them: ysz a (b) represents the mapping value of the operating data item a for the b-th defective motor;

[0100] It should be noted that by extracting the mapping values ​​of each row and statistically analyzing different types of defective motors, the accuracy of the feature value calculation is ensured.

[0101] Step S312: Perform statistical analysis on the eigenvalues ​​to obtain tzz1 to tzz as For eigenvalues ​​tzz1 to tzz as Sort the data in descending order to obtain a feature value sorting list. Based on the feature value sorting list, obtain the sorting number of the feature value of each running data item in the sorting list, denoted as px(a); for example, if the feature value of the first running data item has a sorting number of 1 in the feature value sorting list, then px(1) = 1.

[0102] Step S32: Obtain the mapping range of the mapping value; based on the mapping range and the sorting result, set the initial value; obtain the number of items of the motor operation data; set the initial angle of the mapping value; and represent the reference data in coordinates based on the initial value and the initial angle.

[0103] Step S321: Denote the mapping range of the mapped values ​​as qjf; obtain the sorting index px(a) of the feature value of each running data in the sorting list; set the initial value according to the mapping range gjf and the sorting index px(a); denote the initial value as csz;

[0104] ;

[0105] Among them: csza This represents the initial value of the data for the a-th item.

[0106] Step S322: Based on the number of items as of the motor operation data, set the initial angle of the mapping value. Divide 2Π into as equal parts according to the number of items as of the motor operation data to obtain as equal dividing lines. Combine the mapping value with the sorting number px(a) of the feature value of each operation data in the sorting list to make the mapping value lie on different equal dividing lines to obtain the initial angle of the mapping value. Record the initial angle as cjd.

[0107] ;

[0108] Among them: cjd a This represents the initial angle of the data in item a.

[0109] Step S323: Perform statistical analysis on the initial values ​​and initial angles to obtain the initial coordinates jzb of the reference data. a jzb a = (csz) a cjs a ).

[0110] Please see Figure 3 Step S33: Obtain the mapping value, and calculate the coordinates of various data of different types of defective motors based on the mapping value and the initial value; connect the coordinates of various data of different types of defective motors to obtain the feature image of the defective motor.

[0111] Step S331: Extract the column data from the mapping value list to obtain the mapping values ​​of various operating data of the defective motor; obtain the initial values ​​of various operating data, and deflect the initial values ​​according to the mapping values ​​to obtain the coordinate length qcd of the defective motor;

[0112] ;

[0113] Among them: qcd a (b) represents the coordinate length of the a-th item of the operating data for the motor with defect type b; ysz a (b) represents the mapping value of the a-th item of the operating data of the motor with the b-th defect.

[0114] The angle of the defective motor is obtained based on the initial angle, and the coordinate angle zbj of the defective motor is obtained.

[0115] The defect coordinates of the defective motor are obtained from its coordinate length and coordinate angle, resulting in qzb. a (b) = (qcd) a (b), zbj a (b)).

[0116] Step S332: Statistically analyze the defect coordinates of the defective motor, and assign the defect coordinates qzb1(b) to qzb... as (b) Connect the features to obtain the feature image of the defective motor of type b; store the feature images of all defective motors; extract features from the feature images of the defective motors; the specific feature extraction of the defective motors is as follows:

[0117] Step S3321: Obtain the defect coordinates qzb of the defective motor. a (b) = (qcd) a (b), zbj a (b) The defect area qmj(b) is obtained by calculating the area enclosed by the defective motor based on the defect coordinates of the defective motor.

[0118] ;

[0119] Among them: qcd as qcd1 represents the coordinate length of the 'as'th coordinate in the real-time data image, and 2Π / as represents the coordinate length of the '1'th coordinate in the real-time data image; 2Π / as is a constant value for the angle change between adjacent coordinates.

[0120] Step S3322: Based on the defect coordinates of the defective motor, determine the coordinate length qcd of the defect coordinates. a (b) with coordinate angle zbj a (b) Integrate the coordinate characteristic values ​​qzt(b) of the defective motor;

[0121] ;

[0122] By changing the coordinate length qcd of the defect coordinates a (b) with coordinate angle zbj a (b) The integration method transforms two-dimensional data into one-dimensional data and summarizes the characteristics of defective motors.

[0123] Step S4: Obtain real-time data of the motor, map the real-time data of the motor to construct a real-time data image; perform similarity test between the real-time data image and the feature image of the defective motor to diagnose the defect of the motor.

[0124] Step S41: Obtain the mapping ratio of the data mapping, map the real-time data according to the mapping ratio, transform the coordinates of the mapped data to obtain the coordinate points of the real-time data, connect the coordinate points of the real-time data, and construct a real-time data image.

[0125] Please see Figure 4Step S42: Calculate the similarity between the real-time data image and the feature image of the defective motor, perform defect diagnosis on the motor based on the real-time data image and the feature image of the defective motor, store the real-time data image and the diagnosis result, and optimize the feature image of the defective motor.

[0126] Step S421: Obtain the coordinates of the real-time data image based on the real-time data image, denoted as sst. a = (scd) a sjd a ), where scd a For coordinate length, sjd a The coordinate angle is used; the area enclosed by the image is calculated based on the coordinates of the real-time data image to obtain the real-time area smj;

[0127] ;

[0128] Among them: scd as scd1 represents the coordinate length of the 'as'th coordinate in the real-time data image, and scd1 represents the coordinate length of the '1'th coordinate in the real-time data image; 2Π / as is a constant value for the angle change between adjacent coordinates.

[0129] The coordinates of the real-time data image, and the coordinate length scd of the real-time image. a With coordinate angle sjd a The coordinate feature values ​​stz of the real-time image are then integrated.

[0130] ;

[0131] Step S422: Obtain the defect area qmj(b) and coordinate feature value qzt(b) of the defective motor. Combine the real-time area smj and coordinate feature value sts of the real-time image to calculate the similarity between the real-time image and the feature image, and obtain the similarity value xsz.

[0132] ;

[0133] Where: xsz(b) represents the similarity value between the real-time image and the defect image of the b-th defective motor;

[0134] The defective motor corresponding to the largest similarity value is obtained to obtain the motor's pre-diagnosis result;

[0135] Based on the motor's preliminary diagnosis results, the motor is inspected. If the inspection results are consistent with the preliminary diagnosis results, the motor diagnosis is completed. If the inspection results are inconsistent with the preliminary diagnosis results, the inspection results are used as new motor defects to supplement the existing motor defects, and the real-time data is stored as the characteristic data of the defective motor.

[0136] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multimodal fusion-based intelligent diagnostic method for train traction motor defects, characterized in that, The intelligent diagnostic method includes: Step S1: Monitor the motor operating data under different conditions to obtain motor operating status data; Step S2: Based on the motor's operating status data, obtain the baseline data list and the defect data list; calculate the defect data difference between the baseline data list and the defect data list, and perform data mapping on the defect data difference to obtain the mapped value; Step S3: Perform feature calculation on the motor operation data based on the mapping value to obtain the feature value of the motor defect, and sort the various motor operation data according to the feature value; obtain the mapping interval of the mapping value, and combine the mapping value and the sorting result to construct a polar coordinate image to obtain the feature image of the defective motor; Step S4: Obtain real-time data of the motor, map the real-time data of the motor to construct a real-time data image; perform similarity test between the real-time data image and the feature image of the defective motor to diagnose the defect of the motor.

2. The intelligent diagnostic method for train traction motor defects based on multimodal fusion as described in claim 1, characterized in that, The specific steps of step S1 are as follows: Step S11: Acquire the normal operating data of the motor as the baseline data; obtain the number of motor operating data items in the baseline data, denoted as as; based on the number of motor operating data items, obtain each motor operating data item sj. a ; for sj a Statistical analysis was performed to obtain a baseline data list (jsl). Step S12: Obtain the type of motor defect, denoted as bs. Based on the type of motor defect, obtain the abnormal operation data. Combine the number of motor operation data items in the baseline data and denote the abnormal operation data as qx. a (b); Statistical analysis of abnormal operation data yields a defect data list qsl(b); The motor's operating status data consists of a list of baseline data for the motor and a list of defect data.

3. The intelligent diagnostic method for train traction motor defects based on multimodal fusion as described in claim 1, characterized in that, The specific steps of step S2 are as follows: Step S21: Based on the baseline data list, obtain each operating data item sj of the normal motor. a Based on the defect data list, obtain each operational data item (qx) of the defective motor. a (b); Run the data qx a (b) with sj a The difference was calculated to obtain the defect data difference value qcz. a (b); Step S22: Map the defect data differences to extract the defect data differences for different types of motor defects; map the defect data differences according to their data distribution to obtain the mapped values; specifically as follows: Extract the defect data differences for different types of motor defects to obtain a difference list clb. a ,clb a =[qcz a (1), qcz a (2), ..., qcz a (bs)]; Get the maximum value in the difference list, denoted as mx a Find the minimum value in the difference list, denoted as mi. a ; Based on the maximum and minimum values ​​in the difference list, calculate the difference qcz for all defect data in the difference list. a (b) Perform data mapping to obtain the mapped value ysz; ； Among them: ysz a (b) represents the mapping value of the b-th data in the a-th difference list; The mapped values ​​are statistically analyzed to obtain a list of mapped values, ylb.

4. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 1, characterized in that, The specific steps of step S3 are as follows: Step S31: Obtain the mapping value list, sum the absolute values ​​of each row of mapping values ​​according to the mapping value list to obtain the feature values ​​of the running data; sort the feature values ​​in descending order; obtain the feature value sorting list; obtain the sorting number in the feature value sorting list and associate it with each item of running data; Step S32: Obtain the mapping range of the mapping value; based on the mapping range and the sorting result, set the initial value; obtain the number of items of the motor operation data; set the initial angle of the mapping value; and represent the reference data in coordinates based on the initial value and the initial angle. Step S33: Obtain the mapping value, and calculate the coordinates of various data of different types of defective motors based on the mapping value and the initial value; connect the coordinates of various data of different types of defective motors to obtain the feature image of the defective motor.

5. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 4, characterized in that, The specific steps of step S31 are as follows: Extract the mapping values ​​from the mapping value list to obtain the mapping value ysz. a (1) to ysz a (bs); for the mapping value ysz a (1) to ysz a (bs) performs absolute value summation to obtain the characteristic value of each running data item, denoted as the characteristic value tzz. a ; Statistical analysis of the eigenvalues ​​yields tzz1 to tzz as For eigenvalues ​​tzz1 to tzz as Sort the data in descending order to obtain a feature value sorting list. Based on the feature value sorting list, obtain the sorting number of the feature value of each data item in the sorting list, denoted as px(a).

6. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 4, characterized in that, The specific steps of step S32 are as follows: Let qjf denote the mapping range of the mapped values; obtain the sorting index px(a) of the feature value of each running data item in the sorting list; set the initial value according to the mapping range gjf and the sorting index px(a); and denote the initial value as csz. ； Among them: csz a This represents the initial value of the data in item a. Based on the number of items in the motor operation data as, the initial angle of the mapping value is set. The circumference angle is divided equally according to the number of items in the motor operation data to obtain as equidistant lines. The mapping value is combined with the sorting number px(a) of the feature value of each operation data item in the sorting list to make the mapping value lie on different equidistant lines, and the initial angle of the mapping value is obtained. The initial angle is denoted as cjd. ； Among them: cjd a This represents the initial angle of the data in item a; By statistically analyzing the initial values ​​and initial angles, the initial coordinates jzb of the reference data are obtained. a jzb a = (csz) a cjs a ).

7. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 4, characterized in that, The specific steps of step S33 are as follows: Step S331: Extract the column data from the mapping value list to obtain the mapping values ​​of various operating data of the defective motor; obtain the initial values ​​of various operating data, and deflect the initial values ​​according to the mapping values ​​to obtain the coordinate length of the defective motor; The angle of the defective motor is obtained based on the initial angle, thus obtaining the coordinate angle of the defective motor; The defect coordinates of the defective motor are obtained from its coordinate length and coordinate angle. Step S332: Statistically analyze the defect coordinates of the defective motors, connect the defect coordinates of the defective motors to obtain the feature images of the defective motors; store the feature images of all defective motors. Feature extraction is performed on the defective motor from its feature image.

8. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 7, characterized in that, The specific steps of step S332 are as follows: Obtain the defect coordinates qzb of the defective motor a (b) = (qcd) a (b), zbj a (b) The defect area qmj(b) is obtained by calculating the area enclosed by the defective motor based on the defect coordinates of the defective motor. ； Among them: qcd as qcd1 represents the coordinate length of the 'as'th coordinate in the real-time data image, and qcd1 represents the coordinate length of the '1'th coordinate in the real-time data image; 2Π / as is a constant value for the angle change between adjacent coordinates. Based on the defect coordinates of the defective motor, the coordinate length qcd of the defect coordinates is... a (b) with coordinate angle zbj a (b) Integrate the coordinate characteristic values ​​qzt(b) of the defective motor; ； The coordinate length qcd of the defect coordinates a (b) with coordinate angle zbj a (b) Integrate and summarize the characteristics of defective motors.

9. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 1, characterized in that, The specific steps of step S4 are as follows: Step S41: Obtain the mapping ratio of the data mapping, map the real-time data according to the mapping ratio, transform the coordinates of the mapped data to obtain the coordinate points of the real-time data, connect the coordinate points of the real-time data, and construct a real-time data image. Step S42: Calculate the similarity between the real-time data image and the feature image of the defective motor, perform defect diagnosis on the motor based on the real-time data image and the feature image of the defective motor, store the real-time data image and the diagnosis results, and optimize the feature image of the defective motor.

10. The intelligent diagnostic method for train traction motor defects based on multimodal fusion according to claim 9, characterized in that, The specific steps of step S42 are as follows: Step S421: Obtain the coordinates of the real-time data image based on the real-time data image, denoted as sst. a = (scd) a sjd a ), where scd a For coordinate length, sjd a The coordinate angle is used; the area enclosed by the image is calculated based on the coordinates of the real-time data image to obtain the real-time area smj; The coordinates of the real-time data image, and the coordinate length scd of the real-time image. a With coordinate angle sjd a The coordinate feature values ​​stz of the real-time image are then integrated. Step S422: Obtain the defect area qmj(b) and coordinate feature value qzt(b) of the defective motor. Combine the real-time area smj and coordinate feature value sts of the real-time image to calculate the similarity between the real-time image and the feature image, and obtain the similarity value xsz. ； Where: xsz(b) represents the similarity value between the real-time image and the defect image of the b-th defective motor; The defective motor corresponding to the largest similarity value is obtained to obtain the motor's pre-diagnosis result; Based on the motor's preliminary diagnosis results, the motor is inspected. If the inspection results are consistent with the preliminary diagnosis results, the motor diagnosis is completed. If the inspection results are inconsistent with the preliminary diagnosis results, the inspection results are used as new motor defects to supplement the existing motor defects, and the real-time data is stored as the characteristic data of the defective motor.

Images