Machine learning-based locomotive vibration pattern recognition method and system

Abstract

The present invention relates to the technical field of locomotive vibration mode recognition, and a locomotive vibration mode recognition method and system based on machine learning, including: obtaining the real-time vibration acceleration of a target locomotive, calculating the real-time deviation angle according to the induction tilt angle and the real-time vibration acceleration, determining whether the real-time deviation angle is greater than a deviation threshold, if it is greater, performing vibration adjustment on the real-time vibration acceleration to obtain the target vibration acceleration, if it is not greater, using the real-time vibration acceleration as the target vibration acceleration, performing machine learning on the original vibration recognition model through a historical vibration data set to obtain the target vibration recognition model, and performing locomotive vibration mode recognition on the target locomotive according to the target vibration acceleration by using the target vibration recognition model. The present invention can reduce the human resource consumption in the process of locomotive vibration mode recognition and improve the accuracy of locomotive vibration mode recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of locomotive vibration mode recognition, and particularly to a locomotive vibration mode recognition method, system, electronic device and computer-readable storage medium based on machine learning. Background Art

[0002] The locomotive vibration generated during the operation of the locomotive is one of the main factors affecting the safety performance of the locomotive. The locomotive vibration not only accelerates the wear of parts and shortens the maintenance cycle, but excessive vibration may also lead to train operation safety accidents. Therefore, quickly and accurately identifying the locomotive vibration mode during the operation of the locomotive plays an important role in ensuring the safety of passengers and staff and improving the locomotive maintenance efficiency.

[0003] Currently, the method for identifying the locomotive vibration mode usually first obtains the vibration data during the locomotive running, and then experts in related fields manually identify the locomotive vibration mode according to the vibration data. Although this method can achieve the purpose of identifying the locomotive vibration mode, it requires a large amount of human resources, and the manual judgment is prone to subjective deviation, reducing the accuracy of the locomotive vibration mode recognition. Summary of the Invention

[0004] The present invention provides a locomotive vibration mode recognition method and a computer-readable storage medium based on machine learning, and its main purpose is to reduce the consumption of human resources in the process of locomotive vibration mode recognition and improve the accuracy of locomotive vibration mode recognition.

[0005] To achieve the above object, a locomotive vibration mode recognition method based on machine learning provided by the present invention includes:

[0006] Collecting real-time vibration signals of a target locomotive by using a pre-built acceleration sensor, wherein the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals and vertical vibration signals;

[0007] Performing analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration acceleration, wherein the real-time vibration acceleration includes: real-time lateral acceleration, real-time longitudinal acceleration and real-time vertical acceleration;

[0008] Obtaining the induced tilt angle of the target locomotive by using a pre-built gravity sensing device, calculating the vibration tilt angle according to the real-time vibration acceleration, and calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle;

[0009] Judging whether the real-time deviation angle is greater than a preset deviation threshold;

[0010] If the real-time deviation angle is greater than the deviation threshold, the real-time vibration acceleration is adjusted according to the real-time deviation angle to obtain the target vibration acceleration;

[0011] If the real-time deviation angle is not greater than the deviation threshold, the real-time vibration acceleration is used as the target vibration acceleration;

[0012] Obtain the historical vibration data set of the target locomotive, and set the original vibration value range of the target locomotive according to the historical vibration data set, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range;

[0013] According to the original vibration value range and the pre-constructed convolutional neural network, construct an original vibration recognition model, and perform machine learning on the original vibration recognition model using the historical vibration data set to obtain a target vibration recognition model;

[0014] According to the target vibration acceleration, use the target vibration recognition model to identify the locomotive vibration mode of the target locomotive.

[0015] Optionally, the performing analog-to-digital conversion and error processing on the real-time vibration signal to obtain the real-time vibration acceleration includes:

[0016] Perform analog-to-digital conversion on the real-time vibration signal to obtain the original vibration acceleration;

[0017] Obtain the lateral zero offset, longitudinal zero offset, and vertical zero offset of the acceleration sensor, and perform error processing on the original vibration acceleration according to the lateral zero offset, longitudinal zero offset, and vertical zero offset to obtain the real-time vibration acceleration, where the real-time vibration acceleration is expressed as:

[0018] ;

[0019] Where represents the real-time lateral acceleration, represents the original lateral acceleration, represents the lateral zero offset, represents the real-time longitudinal acceleration, represents the original longitudinal acceleration, represents the longitudinal zero offset, represents the real-time vertical acceleration, represents the original vertical acceleration, represents the vertical zero offset.

[0020] Optionally, the calculating the vibration tilt angle according to the real-time vibration acceleration, and calculating the real-time deviation angle of the target locomotive according to the induction tilt angle and the vibration tilt angle includes:

[0021] Calculate the vibration tilt angle according to the real-time vibration acceleration by using the following formula:

[0022] ;

[0023] wherein, represents the vibration tilt angle, represents the arctangent function, represents the lateral acceleration, represents the longitudinal acceleration, represents the vertical acceleration;

[0024] Calculate the real-time deviation angle according to the induced tilt angle and the vibration tilt angle by using the following formula:

[0025] ;

[0026] wherein, represents the real-time deviation angle, represents the induced tilt angle.

[0027] Optionally, the vibration adjustment of the real-time vibration acceleration according to the real-time deviation angle to obtain the target vibration acceleration includes:

[0028] Construct the following real-time rotation matrix according to the real-time deviation angle:

[0029] ;

[0030] wherein, represents the real-time rotation formula, represents the cosine function, represents the sine function;

[0031] According to the real-time rotation matrix, perform rotation adjustment on the real-time vibration acceleration by using the following formula to obtain the target vibration acceleration, wherein the target vibration acceleration includes: target lateral acceleration, target longitudinal acceleration and target vertical acceleration:

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] wherein, represents the target lateral acceleration, represents the target longitudinal acceleration, represents the target vertical acceleration.

[0037] Optionally, setting the original vibration value range of the target locomotive according to the historical vibration data set includes:

[0038] Obtain the historical vibration peaks of the historical vibration data set, where the historical vibration data set is multiple groups of acceleration data with vibration mode labels, and the historical vibration peaks include: lateral vibration peak, longitudinal vibration peak, and vertical vibration peak;

[0039] Respectively use the lateral vibration peak, longitudinal vibration peak, and vertical vibration peak as the original maximum lateral value, original maximum longitudinal value, and original maximum vertical value to obtain the original maximum vibration value;

[0040] Select the original minimum vibration value from the historical vibration data set according to the preset selection range, where the original minimum vibration value includes: original minimum lateral value, original minimum longitudinal value, and original minimum vertical value;

[0041] Construct the original vibration value range according to the original maximum vibration value and the original minimum vibration value, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range.

[0042] Optionally, performing machine learning on the original vibration recognition model using the historical vibration data set to obtain the target vibration recognition model includes:

[0043] Randomly divide the historical vibration data set into a vibration training data set and a vibration verification data set;

[0044] Successively extract vibration training data from the vibration training data set;

[0045] Input the vibration training data into the original vibration recognition model to obtain a model recognition result;

[0046] Identify the vibration mode label of the vibration training data, and determine whether the model recognition result is consistent with the vibration mode label of the vibration training data;

[0047] If the model recognition result is consistent with the vibration mode label of the vibration training data, then use the original vibration value range as the target vibration value range;

[0048] If the model recognition result is inconsistent with the vibration mode label of the vibration training data, then adjust the original vibration value range according to the vibration training data to obtain the target vibration value range;

[0049] Construct a verification vibration recognition model according to the convolutional neural network and the target vibration value range;

[0050] Extract vibration verification data from the vibration verification dataset in sequence, input the vibration verification data into the verification vibration recognition model, and obtain a model verification result;

[0051] Identify the vibration mode label of the vibration verification data, and determine whether the model verification result is consistent with the vibration mode label of the vibration verification data:

[0052] If the model verification result is inconsistent with the vibration mode label of the vibration verification data, obtain the inconsistent data in the vibration verification data, use the inconsistent data as vibration training data, use the verification vibration recognition model as the original vibration recognition model, and return to the step of inputting the vibration training data into the original vibration recognition model;

[0053] If the model verification result is consistent with the vibration mode label of the vibration verification data, use the verification vibration recognition model as the target vibration recognition model.

[0054] Optionally, the adjusting the original vibration value range according to the vibration training data to obtain a target vibration value range includes:

[0055] The vibration training data is expressed as:

[0056] ;

[0057] Wherein, represents the vibration training data, represents the training lateral acceleration, represents the training longitudinal acceleration, represents the training vertical acceleration;

[0058] The original vibration value range is expressed as:

[0059] ;

[0060] Wherein, represents the original lateral value range, represents the original minimum lateral value, represents the original maximum lateral value, represents the original longitudinal value range, represents the original minimum longitudinal value, represents the original maximum longitudinal value, represents the original vertical value range, represents the original minimum vertical value, represents the original maximum vertical value;

[0061] Judge the vibration difference type of the model verification result and the vibration mode label, wherein the vibration difference type includes: label abnormal vibration and label normal vibration;

[0062] If the vibration difference type is label abnormal vibration, the original vibration value range is processed to expand the value range according to the vibration training data to obtain the target vibration value range;

[0063] If the vibration difference type is label normal vibration, the original vibration value range is processed to shrink according to the vibration training data to obtain the target vibration value range.

[0064] Optionally, the vibration difference type of the judgment model verification result and the vibration mode label includes:

[0065] Obtain the training vibration mode of the vibration mode label and the recognized vibration mode of the model recognition result, where both the training vibration mode and the recognized vibration mode include: lateral vibration mode, longitudinal vibration mode, vertical vibration mode, and normal vibration mode;

[0066] Judge whether the training vibration mode appears in the recognized vibration mode;

[0067] If the training vibration mode does not appear in the recognized vibration mode, set the vibration difference type as label abnormal vibration;

[0068] If the training vibration mode appears in the recognized vibration mode, set the vibration difference type as label normal vibration.

[0069] Optionally, the locomotive vibration mode recognition of the target locomotive by using the target vibration recognition model according to the target vibration acceleration includes:

[0070] Input the target vibration acceleration into the target vibration recognition model to obtain a predicted vibration value;

[0071] Compare the predicted vibration value with the target vibration value range to obtain the locomotive vibration mode, where the locomotive vibration mode includes: lateral vibration mode, longitudinal vibration mode, vertical vibration mode, and normal vibration mode.

[0072] To achieve the above object, the present invention also provides a locomotive vibration mode recognition system based on machine learning, including:

[0073] A vibration data acquisition module, configured to collect real-time vibration signals of a target locomotive by using a pre-built acceleration sensor, where the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals, perform analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration acceleration, where the real-time vibration acceleration includes: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration;

[0074] An angle error calculation module, configured to obtain the sensed tilt angle of the target locomotive by using a pre-built gravity sensing device, calculate the vibration tilt angle according to the real-time vibration acceleration, and calculate the real-time deviation angle of the target locomotive according to the sensed tilt angle and the vibration tilt angle;

[0075] An angle error elimination module, configured to determine whether the real-time deviation angle is greater than a preset deviation threshold. If the real-time deviation angle is greater than the deviation threshold, perform vibration adjustment on the real-time vibration acceleration according to the real-time deviation angle to obtain the target vibration acceleration. If the real-time deviation angle is not greater than the deviation threshold, use the real-time vibration acceleration as the target vibration acceleration, obtain the historical vibration data set of the target locomotive, and set the original vibration value range of the target locomotive according to the historical vibration data set, where the original vibration value range includes: a lateral vibration value range, a longitudinal vibration value range, and a vertical vibration value range;

[0076] A vibration mode recognition module, configured to construct an original vibration recognition model according to the original vibration value range and a pre-built convolutional neural network, perform machine learning on the original vibration recognition model by using the historical vibration data set to obtain a target vibration recognition model, and perform locomotive vibration mode recognition on the target locomotive by using the target vibration recognition model according to the target vibration acceleration.

[0077] To solve the above problems, the present invention further provides an electronic device, where the electronic device includes:

[0078] A memory, storing at least one instruction; and

[0079] A processor, configured to execute the instruction stored in the memory to implement the above-mentioned locomotive vibration mode recognition method based on machine learning.

[0080] To solve the above problems, the present invention further provides a computer-readable storage medium, where at least one instruction is stored in the computer-readable storage medium, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned locomotive vibration mode recognition method based on machine learning.

[0081] To solve the problems described in the background art, the present invention utilizes an acceleration sensor to collect the real-time vibration signals of the target locomotive, achieving the intelligent real-time acquisition of the locomotive vibration acceleration. The real-time vibration signals are subjected to analog-to-digital conversion and error processing to obtain the real-time vibration acceleration, which converts the electrical signals of the locomotive vibration into real-time vibration acceleration values. A gravity sensing device is used to obtain the induced tilt angle of the target locomotive, completing the contactless acquisition of the induced tilt angle of the target locomotive, providing a basis for the subsequent calculation of the direction offset error of the acceleration sensor. The vibration tilt angle is calculated based on the real-time vibration acceleration, obtaining the vibration tilt angle indirectly calculated based on the real-time vibration acceleration, providing conditions for the subsequent calculation of the direction offset error of the acceleration sensor. By calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle, the numerical representation of the direction offset error of the acceleration sensor is realized. Since there is an acceptable range for the real-time deviation angle, that is, it is not necessary to eliminate all real-time deviation angles, it is possible to determine whether the real-time deviation angle is greater than the deviation threshold, and screen out the real-time deviation angles within the deviation threshold. If the real-time deviation angle is greater than the deviation threshold, it indicates that the real-time deviation angle is not within the acceptable range, so the real-time vibration acceleration needs to be vibrationally adjusted according to the real-time deviation angle to obtain the target vibration acceleration, completing the elimination of the real-time deviation angle of the target locomotive. If the real-time deviation angle is not greater than the deviation threshold, it indicates that the real-time deviation angle is within the acceptable range. By taking the real-time vibration acceleration as the target vibration acceleration, unnecessary real-time deviation angle elimination can be skipped, accelerating the process of the entire locomotive vibration mode recognition. By obtaining the historical vibration data set of the target locomotive and setting the original vibration value range of the target locomotive according to the historical vibration data set, the basis for locomotive vibration mode recognition is obtained. Further, according to the original vibration value range and the convolutional neural network, an original vibration recognition model is constructed, and the historical vibration data set is used to perform machine learning on the original vibration recognition model to obtain the target vibration recognition model. This step utilizes the deep learning characteristics of the convolutional neural network to construct a model that can intelligently recognize the locomotive vibration mode. The target vibration recognition model is not interfered by human subjective factors and has high accuracy. Finally, by inputting the target vibration acceleration into the target vibration recognition model, the locomotive vibration mode recognition of the target locomotive is completed. Since this process does not require a large amount of manual participation, the consumption of human resources is reduced. Therefore, the present invention can reduce the consumption of human resources in the locomotive vibration mode recognition process and improve the accuracy of locomotive vibration mode recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 FIG. is a schematic flow chart of a locomotive vibration mode recognition method based on machine learning provided by an embodiment of the present invention;

[0083] Figure 2Functional module diagram of a locomotive vibration pattern recognition system based on machine learning provided by an embodiment of the present invention;

[0084] Figure 3 Schematic structural diagram of an electronic device for implementing the locomotive vibration pattern recognition method based on machine learning provided by an embodiment of the present invention.

[0085] The realization, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Specific embodiments

[0086] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0087] The embodiments of the present application provide a locomotive vibration pattern recognition method based on machine learning. The execution subject of the locomotive vibration pattern recognition method based on machine learning includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiments of the present application. In other words, the locomotive vibration pattern recognition method based on machine learning can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0088] Refer to Figure 1 As shown, it is a flowchart of a locomotive vibration pattern recognition method based on machine learning provided by an embodiment of the present invention. In this embodiment, the locomotive vibration pattern recognition method based on machine learning includes:

[0089] S1. Use a pre-built acceleration sensor to collect the real-time vibration signals of the target locomotive, where the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals.

[0090] It should be explained that the acceleration sensor refers to a three-axis acceleration sensor that can measure the accelerations in three directions during the locomotive's driving. When setting this acceleration sensor, the driving direction of the locomotive is taken as the lateral direction, the direction perpendicular to the lateral direction but in the same plane as the driving direction is taken as the longitudinal direction, and the direction perpendicular to the plane formed by the lateral direction and the longitudinal direction is taken as the vertical direction. The target locomotive refers to the locomotive for which locomotive vibration pattern recognition is to be performed, such as: a heavy-haul electric locomotive.

[0091] It is understandable that the locomotive engine is the main power source of the vehicle. The piston movement and crankshaft rotation inside it will cause the vehicle to vibrate. Therefore, the vibration intensity of the vehicle generated inside the locomotive engine is relatively high. When the vehicle vibration is transmitted to the body, it will become less obvious due to the action of the shock absorber. Therefore, to fully identify and understand the vehicle vibration, the acceleration sensor should be installed at the locomotive engine.

[0092] It is understandable that the lateral vibration signal refers to the lateral electrical signal generated when the vehicle vibrates in the lateral direction of the vehicle. Among them, the lateral electrical signal represents the magnitude of the acceleration in the lateral direction of the vehicle. The longitudinal electrical signal refers to the longitudinal electrical signal generated when the vehicle vibrates in the longitudinal direction of the vehicle. Among them, the longitudinal electrical signal represents the magnitude of the acceleration in the longitudinal direction of the vehicle. The vertical vibration signal refers to the vertical electrical signal generated when the vehicle vibrates in the vertical direction of the vehicle. Among them, the vertical electrical signal represents the magnitude of the acceleration in the vertical direction of the vehicle.

[0093] It should be noted that the real-time vibration signal collected by the acceleration sensor should be based on the whole vehicle as the reference system, that is, the driving acceleration of the vehicle itself will not affect the real-time vibration signal.

[0094] S2 performs analog-to-digital conversion and error processing on the real-time vibration signal to obtain the real-time vibration acceleration, where the real-time vibration acceleration includes: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration.

[0095] It should be explained that the real-time vibration signal is a continuously changing voltage signal, not a specific acceleration value. Therefore, the real-time vibration signal needs to be subjected to analog-to-digital conversion to convert the voltage signal into an acceleration value. The analog-to-digital conversion refers to a technology that can convert a voltage signal into an acceleration value.

[0096] It is understandable that during the working process of the acceleration sensor, zero-point offset will occur due to vehicle vibration. The zero-point offset refers to the phenomenon that the reading of the acceleration sensor is not zero when the vehicle does not vibrate. This phenomenon will cause errors when the acceleration sensor collects the real-time vibration signal. Therefore, error processing needs to be done for this zero-point offset.

[0097] It is understandable that the real-time vibration acceleration refers to the acceleration values in the lateral, longitudinal, and vertical directions obtained after the original vibration acceleration undergoes analog-to-digital conversion and error processing.

[0098] Specifically, the performing analog-to-digital conversion and error processing on the real-time vibration signal to obtain the real-time vibration acceleration includes:

[0099] Performing analog-to-digital conversion on the real-time vibration signal to obtain the original vibration acceleration;

[0100] Obtain the lateral zero offset, longitudinal zero offset, and vertical zero offset of the acceleration sensor, and perform error processing on the original vibration acceleration according to the lateral zero offset, longitudinal zero offset, and vertical zero offset to obtain the real-time vibration acceleration, where the real-time vibration acceleration is expressed as:

[0101] ;

[0102] where, represents the real-time lateral acceleration, represents the original lateral acceleration, represents the lateral zero offset, represents the real-time longitudinal acceleration, represents the original longitudinal acceleration, represents the longitudinal zero offset, represents the real-time vertical acceleration, represents the original vertical acceleration, represents the vertical zero offset.

[0103] It can be understood that the original vibration acceleration refers to the acceleration with specific values obtained through analog-to-digital conversion, but this original vibration acceleration has not yet undergone error processing. The lateral zero offset refers to the zero offset that occurs in the lateral direction. The longitudinal zero offset refers to the zero offset that occurs in the longitudinal direction. The vertical zero offset refers to the zero offset that occurs in the vertical direction.

[0104] S3. Use the pre-built gravity sensing device to obtain the sensed tilt angle of the target locomotive, calculate the vibration tilt angle according to the real-time vibration acceleration, and calculate the real-time deviation angle of the target locomotive according to the sensed tilt angle and the vibration tilt angle.

[0105] It should be explained that the gravity sensing device refers to a device that measures the sensed tilt angle of the locomotive through the self-gravity of the locomotive, such as: optical sensors, gyroscopes. The sensed tilt angle refers to the angle between the driving direction of the locomotive and the horizontal plane directly measured by the gravity sensing device. The vibration tilt angle refers to the angle between the driving direction of the locomotive and the horizontal plane calculated indirectly. The real-time deviation angle refers to the difference between the sensed tilt angle and the vibration tilt angle.

[0106] It should be noted that the acceleration sensor is completely attached to the locomotive engine. Therefore, when the locomotive vibrates, the acceleration sensor may become loose due to the locomotive vibration, resulting in a slight change in the measurement direction of the acceleration sensor, which will cause a large angular error in subsequent calculations. Therefore, a gravity sensing device is introduced to calculate the angular error.

[0107] It is understandable that the gravity sensing device does not need to have direct contact with the locomotive. Therefore, the induced tilt angle measured by the gravity sensing device will be minimally affected by the vibration of the locomotive. Thus, the induced tilt angle can be considered as the relatively accurate angle between the traveling direction of the locomotive and the horizontal plane. Therefore, the real-time deviation angle can be used to represent the measurement error of the acceleration sensor.

[0108] Specifically, calculating the vibration tilt angle according to the real-time vibration acceleration, and calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle includes:

[0109] Calculating the vibration tilt angle according to the real-time vibration acceleration by using the following formula:

[0110] ;

[0111] Wherein, represents the vibration tilt angle, represents the arctangent function, represents the lateral acceleration, represents the longitudinal acceleration, represents the vertical acceleration;

[0112] Calculating the real-time deviation angle according to the induced tilt angle and the vibration tilt angle by using the following formula:

[0113] ;

[0114] Wherein, represents the real-time deviation angle, represents the induced tilt angle.

[0115] S4. Determine whether the real-time deviation angle is greater than a preset deviation threshold.

[0116] It is understandable that the deviation threshold refers to the deviation angle set artificially, and this deviation threshold can be obtained by querying relevant literature.

[0117] If the real-time deviation angle is greater than the deviation threshold, then execute S5. Perform vibration adjustment on the real-time vibration acceleration according to the real-time deviation angle to obtain the target vibration acceleration.

[0118] It is understandable that when the real-time deviation angle is greater than the deviation threshold, it indicates that there is a large deviation between the real-time vibration acceleration at this time and the actual vibration acceleration of the target locomotive. Therefore, it is necessary to adjust the real-time vibration acceleration of the target locomotive.

[0119] Specifically, performing vibration adjustment on the real-time vibration acceleration according to the real-time deviation angle to obtain the target vibration acceleration includes:

[0120] Construct the following real-time rotation matrix based on the real-time deviation angle:

[0121] ;

[0122] where, represents the real-time rotation formula, represents the cosine function, represents the sine function;

[0123] According to the real-time rotation matrix, use the following formula to perform rotation adjustment on the real-time vibration acceleration to obtain the target vibration acceleration, where the target vibration acceleration includes: target lateral acceleration, target longitudinal acceleration, and target vertical acceleration:

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] where, represents the target lateral acceleration, represents the target longitudinal acceleration, represents the target vertical acceleration.

[0129] It should be explained that the real-time rotation matrix refers to converting the original coordinate axes measured by the acceleration sensor into the actual coordinate axes representing the actual situation. Among them, the original coordinate axes refer to the three measurement directions initially set by the acceleration sensor, namely the lateral direction, the longitudinal direction, and the vertical direction, and the actual coordinate axes refer to the three directions actually measured when the acceleration sensor has a direction offset.

[0130] If the real-time deviation angle is not greater than the deviation threshold, execute S6, and use the real-time vibration acceleration as the target vibration acceleration.

[0131] It can be understood that when the real-time deviation angle is not greater than the deviation threshold, it means that there is only a small deviation between the real-time vibration acceleration and the actual vibration acceleration of the target locomotive at this time, and this deviation has little impact on subsequent calculations. In order to speed up the process of identifying the vibration mode of the entire locomotive, the real-time vibration acceleration is not adjusted.

[0132] S7. Obtain the historical vibration data set of the target locomotive, and set the original vibration value range of the target locomotive according to the historical vibration data set, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range.

[0133] It should be explained that the historical vibration data set refers to multiple sets of historical vibration accelerations of the target locomotive collected in the past. The historical vibration accelerations include: historical lateral acceleration, historical longitudinal acceleration, and historical vertical acceleration. Among them, each set of historical vibration accelerations is manually detected to obtain the locomotive vibration mode corresponding to the vibration acceleration, and this locomotive vibration mode is pasted as a vibration label on the corresponding historical vibration data. Among them, the locomotive vibration modes include: lateral vibration, longitudinal vibration, vertical vibration, and normal vibration.

[0134] It should be noted that the vibration label does not only contain one vibration mode. It may contain multiple locomotive vibration modes. If it is manually determined that a set of historical vibration accelerations does not have any vibration, then this set of historical vibration accelerations will be labeled with the vibration label of normal vibration.

[0135] It can be understood that the historical vibration data set will contain multiple historical lateral accelerations, multiple historical longitudinal accelerations, and multiple historical vertical accelerations. The lateral vibration value range refers to the range of lateral accelerations selected from multiple historical lateral accelerations. The longitudinal vibration value range refers to the range of longitudinal accelerations selected from multiple historical longitudinal accelerations. The vertical vibration value range refers to the range of vertical accelerations selected from multiple historical vertical accelerations.

[0136] Exemplarily, the historical lateral accelerations in a certain historical data set are respectively: 1 、7 、4 , the preset selection range is 2, then the original maximum lateral value is the maximum value in the historical lateral accelerations, which is 7 , starting from the historical lateral acceleration 7 as the starting point, select an original minimum lateral value of 4 at a distance of 2 in the direction of decreasing numerical value , so the original lateral value range is (4 7 ).

[0137] Specifically, setting the original vibration value range of the target locomotive according to the historical vibration data set includes:

[0138] Obtain the historical vibration peaks of the historical vibration data set. Among them, the historical vibration data set is multiple sets of acceleration data with vibration mode labels. The historical vibration peaks include: lateral vibration peak, longitudinal vibration peak, and vertical vibration peak;

[0139] Respectively take the lateral vibration peak, longitudinal vibration peak, and vertical vibration peak as the original maximum lateral value, original maximum longitudinal value, and original maximum vertical value to obtain the original maximum vibration value;

[0140] Selecting an original minimum vibration value from the historical vibration data set according to a preset selection range, wherein the original minimum vibration value includes: an original minimum lateral value, an original minimum longitudinal value, and an original minimum vertical value;

[0141] An original vibration value range is constructed according to the original maximum vibration value and the original minimum vibration value, wherein the original vibration value range includes: a lateral vibration value range, a longitudinal vibration value range, and a vertical vibration value range.

[0142] It is understood that the historical vibration data set may include multiple historical lateral accelerations, multiple historical longitudinal accelerations, and multiple historical vertical accelerations. The lateral vibration peak value refers to the historical lateral acceleration with the largest value among the multiple historical lateral accelerations. The longitudinal vibration peak value refers to the historical longitudinal acceleration with the largest value among the multiple historical longitudinal accelerations. The vertical vibration peak value refers to the historical vertical acceleration with the largest value among the multiple historical vertical accelerations. Selecting the minimum vibration value refers to using the original maximum vibration value as the starting point and determining a minimum vibration value in the direction of decreasing values according to the selection range.

[0143] For example, a historical vibration data set includes five groups of historical vibration data, wherein the first group of historical vibration data is: , the second set of historical vibration data is , the third set of historical vibration data is , where the subscript x represents the lateral direction, the subscript y represents the longitudinal direction, and the subscript z represents the vertical direction. The multiple historical lateral accelerations are: 、 and , from which the lateral vibration peak is selected as , multiple historical longitudinal accelerations are: 、 and , from which the longitudinal vibration peak is selected , multiple historical vertical accelerations are: 、 and , from which the vertical vibration peak is selected as .

[0144] For example, the historical lateral accelerations in a historical data set are: 1 , 7 , 4 , the preset selection range is 2, then the original maximum lateral value is the maximum value of the historical lateral acceleration, which is 7 , with a historical lateral acceleration of 7 As the starting point, select an original minimum horizontal value of 4 with a distance of 2 in the direction of decreasing value , so the original horizontal range is (4 7 ).

[0145] S8. Construct an original vibration recognition model according to the original vibration value range and the pre - constructed convolutional neural network, and perform machine learning on the original vibration recognition model using the historical vibration data set to obtain a target vibration recognition model.

[0146] It should be explained that the convolutional neural network refers to a deep learning model, which can be used to extract the data features of real - time vibration acceleration, perform deep learning on the real - time vibration acceleration according to the data features, and output the predicted vibration values obtained from the real - time vibration acceleration. Among them, the predicted vibration values include: predicted lateral acceleration, predicted longitudinal acceleration, and predicted vertical acceleration. The original vibration recognition model refers to a recognition model that uses the deep learning ability of the convolutional neural network to recognize the locomotive vibration mode. Machine learning refers to the process of training the original vibration recognition model through the historical vibration data set to make it have stronger locomotive vibration recognition ability. The target vibration recognition model refers to a recognition model with strong locomotive vibration recognition ability.

[0147] It can be understood that since the vibration labels in the historical vibration data set are obtained by manual subjective judgment, the vibration labels of the historical vibration data set have strong subjectivity. These vibration labels of the historical vibration data set can represent the human perception of locomotive vibration. However, due to the lack of a unified standard, there will be a large number of vibration differences in actual locomotive mode recognition. Therefore, machine learning can be performed on the convolutional neural network through these historical vibration data sets to obtain a target vibration recognition model that can uniformly describe locomotive vibration.

[0148] Specifically, performing machine learning on the original vibration recognition model using the historical vibration data set to obtain a target vibration recognition model includes:

[0149] Randomly divide the historical vibration data set into a vibration training data set and a vibration verification data set;

[0150] Sequentially extract vibration training data from the vibration training data set;

[0151] Input the vibration training data into the original vibration recognition model to obtain a model recognition result;

[0152] Identify the vibration mode label of the vibration training data, and determine whether the model recognition result is consistent with the vibration mode label of the vibration training data;

[0153] If the model recognition result is consistent with the vibration mode label of the vibration training data, then use the original vibration value range as the target vibration value range;

[0154] If the vibration mode label of the model recognition result is inconsistent with that of the vibration training data, adjust the original vibration value range according to the vibration training data to obtain the target vibration value range;

[0155] Construct a verification vibration recognition model according to the convolutional neural network and the target vibration value range;

[0156] Sequentially extract vibration verification data from the vibration verification dataset, input the vibration verification data into the verification vibration recognition model, and obtain a model verification result;

[0157] Identify the vibration mode label of the vibration verification data, and determine whether the model verification result is consistent with the vibration mode label of the vibration verification data:

[0158] If the model verification result is inconsistent with the vibration mode label of the vibration verification data, obtain the inconsistent data in the vibration verification data, use the inconsistent data as vibration training data, use the verification vibration recognition model as the original vibration recognition model, and return to the step of inputting the vibration training data into the original vibration recognition model;

[0159] If the model verification result is consistent with the vibration mode label of the vibration verification data, use the verification vibration recognition model as the target vibration recognition model.

[0160] It should be explained that the vibration training dataset refers to the dataset for model training of the original vibration recognition model. The vibration verification dataset refers to the dataset for accuracy verification of the verification vibration recognition model. The model recognition result refers to the locomotive vibration mode output after being recognized by the original vibration recognition model. The verification vibration recognition model refers to the recognition model obtained by training the original vibration recognition model with the vibration training dataset. The model verification result refers to the locomotive vibration mode output after being recognized by the verification vibration recognition model. The inconsistent data refers to the vibration verification data whose model verification result is inconsistent with the vibration mode label.

[0161] Specifically, adjusting the original vibration value range according to the vibration training data to obtain the target vibration value range includes:

[0162] The vibration training data is expressed as:

[0163] ;

[0164] Wherein, represents the vibration training data, represents the training lateral acceleration, represents the training longitudinal acceleration, represents the training vertical acceleration;

[0165] The original vibration value range is expressed as:

[0166] ;

[0167] Wherein, represents the original lateral value range, represents the original minimum lateral value, represents the original maximum lateral value, represents the original longitudinal value range, represents the original minimum longitudinal value, represents the original maximum longitudinal value, represents the original vertical value range, represents the original minimum vertical value, represents the original maximum vertical value;

[0168] Judge the vibration difference type of the model verification result and the vibration mode label. Among them, the vibration difference type includes: label abnormal vibration and label normal vibration;

[0169] If the vibration difference type is label abnormal vibration, the original vibration value range is processed to expand the value range according to the vibration training data to obtain the target vibration value range;

[0170] If the vibration difference type is label normal vibration, the original vibration value range is processed to shrink according to the vibration training data to obtain the target vibration value range.

[0171] It should be explained that the vibration difference type refers to the type where the vibration recognition result is inconsistent with the vibration mode label. The label abnormal vibration refers to the situation where the abnormal vibration label in the vibration mode label is inconsistent with the vibration recognition result, that is, the vibration mode label shows that a certain abnormal vibration has occurred in the target locomotive, but the vibration recognition result does not show this abnormal vibration. Among them, the abnormal vibration can include: lateral vibration, longitudinal vibration, and vertical vibration. The label normal vibration refers to the situation where the normal vibration label in the vibration mode label is consistent with the vibration recognition result, that is, the vibration mode label does not show that a certain abnormal vibration has occurred in the target locomotive, but the model recognition result shows that the target locomotive has produced this abnormal vibration.

[0172] It is understandable that when the vibration difference type is label abnormal vibration, it indicates that the vibration prediction value output by the original vibration recognition model exceeds the range of the original vibration value domain. Then, the original vibration value domain needs to be expanded. The expansion process means using the vibration prediction value as the target vibration maximum value or replacing the original vibration maximum value or the original vibration minimum value with the target vibration maximum value. When the vibration difference type is label normal vibration, it indicates that the vibration prediction value output by the original vibration recognition model is within the original vibration value domain. Then, the original vibration value domain needs to be shrunk. The shrinking process means inserting the vibration prediction value into the original vibration value domain and using the vibration prediction value as the target vibration maximum value or replacing the original vibration maximum value or the original vibration minimum value with the target vibration maximum value. Among them, the replacement result depends on the distance between the vibration prediction value, the original vibration maximum value, and the original vibration minimum value.

[0173] Specifically, the judgment of the model verification result and the vibration difference type of the vibration mode label includes:

[0174] Obtain the training vibration mode of the vibration mode label and the recognized vibration mode of the model recognition result. Among them, both the training vibration mode and the recognized vibration mode include: lateral vibration mode, longitudinal vibration mode, vertical vibration mode, and normal vibration mode;

[0175] Judge whether the training vibration mode appears in the recognized vibration mode;

[0176] If the training vibration mode does not appear in the recognized vibration mode, set the vibration difference type as label abnormal vibration;

[0177] If the training vibration mode appears in the recognized vibration mode, set the vibration difference type as label normal vibration.

[0178] Exemplarily, the original vibration value domains of a certain original vibration recognition model are respectively: the original lateral value domain is (4 7 ), the original longitudinal value domain is (1 6 ), the original vertical value domain is (2 8 ) Zhang input the following set of vibration training data into the original vibration recognition model: the historical lateral acceleration is 3 , the historical longitudinal acceleration is 5 , the historical vertical acceleration is 9 , the vibration mode label of this vibration training data is: lateral vibration and vertical vibration. The original vibration recognition model returns a set of vibration prediction values: the predicted lateral acceleration is 3 , the predicted longitudinal acceleration is 4 , the predicted vertical acceleration is 9 , compare the vibration prediction value with the original vibration value range, and the model recognition result is: longitudinal vibration. Among them, the vibration mode labels include lateral vibration and vertical vibration, but lateral vibration and vertical vibration do not appear in the model recognition result. Therefore, the original lateral value range and the original vertical value range need to be enlarged. Use the predicted lateral acceleration as the target lateral minimum value to replace the original lateral minimum value, and the target lateral value range is (3 7 ). Use the predicted vertical acceleration as the target vertical maximum value to replace the original vertical maximum value, and the target vertical value range is (2 9 ). The vibration mode label does not include longitudinal vibration, but longitudinal vibration appears in the model recognition result. Therefore, the original longitudinal value range needs to be reduced. Insert the predicted longitudinal acceleration into the original longitudinal value range and replace the original longitudinal maximum value as the target longitudinal maximum value, and the target longitudinal value range is (1 4 ).

[0179] S9. According to the target vibration acceleration, use the target vibration recognition model to perform locomotive vibration mode recognition on the target locomotive.

[0180] Specifically, performing locomotive vibration mode recognition on the target locomotive according to the target vibration acceleration by using the target vibration recognition model includes:[[]]

[0181] Input the target vibration acceleration into the target vibration recognition model to obtain a predicted vibration value;

[0182] Compare the predicted vibration value with the target vibration value range numerically to obtain the locomotive vibration mode. Among them, the locomotive vibration mode includes: lateral vibration mode, longitudinal vibration mode, vertical vibration mode and normal vibration mode.

[0183] It can be understood that the numerical comparison refers to the action of judging whether the predicted vibration value is within the target vibration value range. Among them, the locomotive vibration modes obtained through numerical comparison can be multiple abnormal vibration modes. Among them, the abnormal vibration modes include: lateral vibration, longitudinal vibration and vertical vibration.

[0184] To solve the problems described in the background art, the present invention utilizes an acceleration sensor to collect the real-time vibration signals of the target locomotive, achieving the intelligent real-time acquisition of the locomotive vibration acceleration. The real-time vibration signals are subjected to analog-to-digital conversion and error processing to obtain the real-time vibration acceleration, which converts the electrical signals of the locomotive vibration into real-time vibration acceleration values. A gravity sensing device is used to obtain the induced tilt angle of the target locomotive, completing the acquisition of the non-contact induced tilt angle of the target locomotive, providing a basis for calculating the direction offset error of the subsequent acceleration sensor. The vibration tilt angle is calculated based on the real-time vibration acceleration, obtaining the vibration tilt angle indirectly calculated based on the real-time vibration acceleration, providing conditions for calculating the direction offset error of the subsequent acceleration sensor. By calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle, the numerical representation of the direction offset error of the acceleration sensor is realized. Since there is an acceptable range for the real-time deviation angle, that is, it is not necessary to eliminate the deviation for all real-time deviation angles, it is possible to screen the real-time deviation angles within the deviation threshold by determining whether the real-time deviation angle is greater than the deviation threshold. If the real-time deviation angle is greater than the deviation threshold, it indicates that the real-time deviation angle is not within the acceptable range, so the real-time vibration acceleration needs to be vibrationally adjusted according to the real-time deviation angle to obtain the target vibration acceleration, completing the elimination of the real-time deviation angle of the target locomotive. If the real-time deviation angle is not greater than the deviation threshold, it indicates that the real-time deviation angle is within the acceptable range. By using the real-time vibration acceleration as the target vibration acceleration, unnecessary real-time deviation angle elimination can be skipped, accelerating the process of the entire locomotive vibration pattern recognition. By obtaining the historical vibration data set of the target locomotive and setting the original vibration value range of the target locomotive according to the historical vibration data set, the basis for locomotive vibration pattern recognition is obtained. Further, based on the original vibration value range and the convolutional neural network, an original vibration recognition model is constructed, and the historical vibration data set is used for machine learning of the original vibration recognition model to obtain the target vibration recognition model. This step utilizes the deep learning characteristics of the convolutional neural network to construct a model that can intelligently recognize the locomotive vibration pattern. The target vibration recognition model is not interfered by human subjective factors and has high accuracy. Finally, by inputting the target vibration acceleration into the target vibration recognition model, the locomotive vibration pattern recognition of the target locomotive is completed. Since this process does not require a large amount of manual participation, the consumption of human resources is reduced. Therefore, the present invention can reduce the consumption of human resources in the locomotive vibration pattern recognition process and improve the accuracy of locomotive vibration pattern recognition.

[0185] As Figure 2 shown, it is a functional module diagram of a locomotive vibration pattern recognition system based on machine learning provided by an embodiment of the present invention.

[0186] The locomotive vibration pattern recognition system 100 based on machine learning according to the present invention can be installed in an electronic device. According to the functions achieved, the locomotive vibration pattern recognition system 100 based on machine learning can include a vibration data acquisition module 101, an angle error calculation module 102, an angle error elimination module 103, and a vibration pattern recognition module 104. The modules in the present invention can also be referred to as units, which refer to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device.

[0187] The vibration data acquisition module 101 is used to collect real-time vibration signals of a target locomotive by using a pre-built acceleration sensor. The real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals. The real-time vibration signals are subjected to analog-to-digital conversion and error processing to obtain real-time vibration acceleration, where the real-time vibration acceleration includes: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration;

[0188] The angle error calculation module 102 is used to obtain the induced tilt angle of the target locomotive by using a pre-built gravity sensing device, calculate the vibration tilt angle according to the real-time vibration acceleration, and calculate the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle;

[0189] The angle error elimination module 103 is used to determine whether the real-time deviation angle is greater than a preset deviation threshold. If the real-time deviation angle is greater than the deviation threshold, the real-time vibration acceleration is vibration-adjusted according to the real-time deviation angle to obtain the target vibration acceleration. If the real-time deviation angle is not greater than the deviation threshold, the real-time vibration acceleration is used as the target vibration acceleration, and a historical vibration data set of the target locomotive is obtained. The original vibration value range of the target locomotive is set according to the historical vibration data set, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range;

[0190] The vibration pattern recognition module 104 is used to construct an original vibration recognition model according to the original vibration value range and a pre-built convolutional neural network, perform machine learning on the original vibration recognition model by using the historical vibration data set to obtain a target vibration recognition model, and perform locomotive vibration pattern recognition on the target locomotive by using the target vibration recognition model according to the target vibration acceleration.

[0191] Specifically, each module in the locomotive vibration pattern recognition system 100 based on machine learning in the embodiment of the present invention adopts the same technical means as those Figure 1 described in the locomotive vibration pattern recognition method based on machine learning described above, and can produce the same technical effects, which will not be elaborated here.

[0192] As shown Figure 3 in the figure, it is a schematic structural diagram of an electronic device for implementing a locomotive vibration mode recognition method based on machine learning provided by an embodiment of the present invention.

[0193] The electronic device 1 may include a processor 10, a memory 11, and a bus 12, and may further include a computer program stored in the memory 11 and executable on the processor 10, such as a locomotive vibration mode recognition method program based on machine learning.

[0194] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as the mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 1. Further, the memory 11 further includes an internal storage unit of the electronic device 1 and also includes an external storage device. The memory 11 can be used not only to store application software installed in the electronic device 1 and various types of data, such as the code of the locomotive vibration mode recognition method program, etc., but also to temporarily store data that has been output or will be output.

[0195] The processor 10 may be composed of integrated circuits in some embodiments. For example, it may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same or different functions, including a combination of one or more Central Processing Units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting various components of the entire electronic device through various interfaces and lines, and by running or executing programs or modules stored in the memory 11 (such as the locomotive vibration mode recognition method program, etc.), and calling data stored in the memory 11, to execute various functions of the electronic device 1 and process data.

[0196] The bus 12 can be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus 12 can be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to implement connection communication between the memory 11 and at least one processor 10, etc.

[0197] Figure 3 Only an electronic device with components is shown. Those skilled in the art can understand that Figure 3 the shown structure does not constitute a limitation on the electronic device 1, and it may include fewer or more components than shown, or combine certain components, or have a different component arrangement.

[0198] For example, although not shown, the electronic device 1 may further include a power source (such as a battery) for powering each component. Preferably, the power source can be logically connected to the at least one processor 10 through a power management device, so as to implement functions such as charging management, discharging management, and power consumption management through the power management device. The power source may also include any components such as one or more DC or AC power sources, a recharge device, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device 1 may also include various sensors, a Bluetooth module, a Wi-Fi module, etc., which will not be elaborated here.

[0199] Furthermore, the electronic device 1 may further include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0200] Optionally, the electronic device 1 may further include a user interface. The user interface may be a display, an input unit (such as a keyboard), and optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visual user interface.

[0201] The program of the locomotive vibration pattern recognition method based on machine learning stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0202] Utilize a pre-built acceleration sensor to collect the real-time vibration signals of the target locomotive. Among them, the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals;

[0203] Perform analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration accelerations. Among them, the real-time vibration accelerations include: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration;

[0204] Utilize a pre-built gravity sensing device to obtain the sensed tilt angle of the target locomotive, calculate the vibration tilt angle according to the real-time vibration acceleration, and calculate the real-time deviation angle of the target locomotive according to the sensed tilt angle and the vibration tilt angle;

[0205] Judge whether the real-time deviation angle is greater than a preset deviation threshold;

[0206] If the real-time deviation angle is greater than the deviation threshold, perform vibration adjustment on the real-time vibration acceleration according to the real-time deviation angle to obtain the target vibration acceleration;

[0207] If the real-time deviation angle is not greater than the deviation threshold, use the real-time vibration acceleration as the target vibration acceleration;

[0208] Obtain the historical vibration data set of the target locomotive, and set the original vibration value range of the target locomotive according to the historical vibration data set. Among them, the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range;

[0209] Construct an original vibration recognition model according to the original vibration value range and a pre-built convolutional neural network, and perform machine learning on the original vibration recognition model using the historical vibration data set to obtain the target vibration recognition model;

[0210] Perform locomotive vibration pattern recognition on the target locomotive according to the target vibration acceleration using the target vibration recognition model.

[0211] Specifically, the specific implementation method of the above instructions by the processor 10 can refer to Figures 1 to 3 The description of the relevant steps in the corresponding embodiment, which will not be elaborated here.

[0212] Furthermore, if the modules / units integrated in the electronic device 1 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM, Read-Only Memory).

[0213] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor of an electronic device, it can implement:

[0214] Collecting real-time vibration signals of a target locomotive by using a pre-built acceleration sensor, where the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals;

[0215] Performing analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration accelerations, where the real-time vibration accelerations include: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration;

[0216] Obtaining the induced tilt angle of the target locomotive by using a pre-built gravity sensing device, calculating the vibration tilt angle according to the real-time vibration accelerations, and calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle;

[0217] Judging whether the real-time deviation angle is greater than a preset deviation threshold;

[0218] If the real-time deviation angle is greater than the deviation threshold, performing vibration adjustment on the real-time vibration accelerations according to the real-time deviation angle to obtain target vibration accelerations;

[0219] If the real-time deviation angle is not greater than the deviation threshold, taking the real-time vibration accelerations as target vibration accelerations;

[0220] Obtaining the historical vibration data set of the target locomotive, and setting the original vibration value range of the target locomotive according to the historical vibration data set, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range;

[0221] Constructing an original vibration recognition model according to the original vibration value range and a pre-built convolutional neural network, and performing machine learning on the original vibration recognition model by using the historical vibration data set to obtain a target vibration recognition model;

[0222] Perform locomotive vibration mode recognition on the target locomotive by using the target vibration recognition model according to the target vibration acceleration.

[0223] In several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and there may be other partitioning methods in actual implementation.

[0224] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0225] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a hardware plus software functional module.

[0226] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention.

[0227] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A locomotive vibration pattern recognition method based on machine learning, characterized in that The method includes: Collecting real-time vibration signals of a target locomotive by using a pre-built acceleration sensor, where the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals; Performing analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration accelerations, where the real-time vibration accelerations include: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration; Obtaining the induced tilt angle of the target locomotive by using a pre-built gravity sensing device, calculating the vibration tilt angle according to the real-time vibration accelerations, and calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle; Judging whether the real-time deviation angle is greater than a preset deviation threshold; If the real-time deviation angle is greater than the deviation threshold, performing vibration adjustment on the real-time vibration accelerations according to the real-time deviation angle to obtain target vibration accelerations; If the real-time deviation angle is not greater than the deviation threshold, using the real-time vibration accelerations as the target vibration accelerations; Obtaining the historical vibration data set of the target locomotive, and setting the original vibration value range of the target locomotive according to the historical vibration data set, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range, and vertical vibration value range; Constructing an original vibration recognition model according to the original vibration value range and a pre-built convolutional neural network, and performing machine learning on the original vibration recognition model by using the historical vibration data set to obtain a target vibration recognition model, where the target vibration recognition model includes a target vibration value range, and constructing the target vibration value range includes: Obtaining vibration training data, inputting the vibration training data into the original vibration recognition model to obtain vibration mode labels and vibration prediction values, where the vibration prediction values include: predicted lateral acceleration, predicted longitudinal acceleration, and predicted vertical acceleration; Comparing the vibration prediction values with the original vibration value range to obtain a model recognition result, where the model recognition result is a mode label, determining a difference label according to the model recognition result and the vibration mode labels, and determining a difference vibration value range in the original vibration value range based on the difference label; Determining the difference acceleration corresponding to the difference vibration value range in the vibration prediction values, if the difference acceleration is less than the minimum value in the difference vibration value range, using the difference acceleration to replace the minimum value in the difference vibration value range to obtain the target vibration value range, if the difference acceleration is greater than the maximum value in the difference vibration value range, using the difference acceleration to replace the maximum value in the difference vibration value range to obtain the target vibration value range; Performing locomotive vibration mode recognition on the target locomotive by using the target vibration accelerations and the target vibration recognition model.

2. The locomotive vibration pattern recognition method based on machine learning according to claim 1, wherein The performing analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration accelerations includes: Performing analog-to-digital conversion on the real-time vibration signals to obtain original vibration accelerations; Obtaining the lateral zero offset, longitudinal zero offset, and vertical zero offset of the acceleration sensor, and performing error processing on the original vibration accelerations according to the lateral zero offset, longitudinal zero offset, and vertical zero offset to obtain real-time vibration accelerations, where the real-time vibration accelerations are expressed as: ； Among them, represents the real-time lateral acceleration, represents the original lateral acceleration, represents the lateral zero offset, represents the real-time longitudinal acceleration, represents the original longitudinal acceleration, represents the longitudinal zero offset, represents the real-time vertical acceleration, represents the original vertical acceleration, represents the vertical zero offset.

3. The method for identifying locomotive vibration patterns based on machine learning according to claim 2, characterized in that Calculating the vibration tilt angle based on the real-time vibration acceleration, and calculating the real-time deviation angle of the target locomotive according to the induced tilt angle and the vibration tilt angle, including: Calculating the vibration tilt angle based on the real-time vibration acceleration by using the following formula: ； Among them, represents the vibration tilt angle, represents the arctangent function, represents the lateral acceleration, represents the longitudinal acceleration, represents the vertical acceleration; Calculating the real-time deviation angle based on the induced tilt angle and the vibration tilt angle by using the following formula: ； Among them, represents the real-time deviation angle, represents the induced tilt angle.

4. The locomotive vibration pattern recognition method based on machine learning according to claim 3, characterized in that, Adjusting the real-time vibration acceleration according to the real-time deviation angle to obtain the target vibration acceleration, including: Constructing the following real-time rotation matrix according to the real-time deviation angle: ； Among them, represents the real-time rotation formula, represents the cosine function, represents the sine function; According to the real-time rotation matrix, adjusting the real-time vibration acceleration by using the following formula to obtain the target vibration acceleration, where the target vibration acceleration includes: target lateral acceleration, target longitudinal acceleration and target vertical acceleration: ； ； ； ； Among them, represents the target lateral acceleration, represents the target longitudinal acceleration, represents the target vertical acceleration.

5. The machine learning-based locomotive vibration pattern recognition method according to claim 1, wherein Setting the original vibration value range of the target locomotive according to the historical vibration data set, including: Obtaining the historical vibration peak values of the historical vibration data set, where the historical vibration data set is a set of acceleration data with vibration mode labels, and the historical vibration peak values include: lateral vibration peak value, longitudinal vibration peak value and vertical vibration peak value; Taking the lateral vibration peak value, the longitudinal vibration peak value and the vertical vibration peak value as the original maximum lateral value, the original maximum longitudinal value and the original maximum vertical value respectively to obtain the original maximum vibration value; Selecting the original minimum vibration value from the historical vibration data set according to the preset selection range, where the original minimum vibration value includes: original minimum lateral value, original minimum longitudinal value and original minimum vertical value; Constructing the original vibration value range according to the original maximum vibration value and the original minimum vibration value, where the original vibration value range includes: lateral vibration value range, longitudinal vibration value range and vertical vibration value range.

6. The method for locomotive vibration pattern recognition based on machine learning according to claim 5, characterized in that Performing machine learning on the original vibration recognition model by using the historical vibration data set to obtain the target vibration recognition model, including: Randomly dividing the historical vibration data set into a vibration training data set and a vibration verification data set; Successively extracting vibration training data from the vibration training data set; Inputting the vibration training data into the original vibration recognition model to obtain a model recognition result; Identifying the vibration mode label of the vibration training data and judging whether the model recognition result is consistent with the vibration mode label of the vibration training data; If the model recognition result is consistent with the vibration mode label of the vibration training data, taking the original vibration value range as the target vibration value range; If the model recognition result is inconsistent with the vibration mode label of the vibration training data, adjusting the original vibration value range according to the vibration training data to obtain the target vibration value range; Constructing a verification vibration recognition model according to the convolutional neural network and the target vibration value range; Successively extracting vibration verification data from the vibration verification data set, inputting the vibration verification data into the verification vibration recognition model to obtain a model verification result; Identifying the vibration mode label of the vibration verification data and judging whether the model verification result is consistent with the vibration mode label of the vibration verification data: If the vibration mode label of the model verification result is inconsistent with that of the vibration verification data, obtain the inconsistent data in the vibration verification data, use the inconsistent data as vibration training data, use the verified vibration recognition model as the original vibration recognition model, and return the step of inputting the vibration training data into the original vibration recognition model; [[ID=]. If the vibration mode label of the model verification result is consistent with that of the vibration verification data, use the verified vibration recognition model as the target vibration recognition model.

7. The method for identifying locomotive vibration patterns based on machine learning according to claim 6, characterized in that, The adjusting the original vibration value range according to the vibration training data to obtain a target vibration value range includes: The vibration training data is expressed as: ； Among them, represents vibration training data, represents the training lateral acceleration, represents the training longitudinal acceleration, represents the training vertical acceleration; The original vibration value range is expressed as: ； Among them, represents the original horizontal value range, represents the original minimum horizontal value, represents the original maximum horizontal value, represents the original vertical value range, represents the original minimum vertical value, represents the original maximum vertical value, represents the original vertical value range, represents the original minimum vertical value, represents the original maximum vertical value; Judge the vibration difference type between the model verification result and the vibration mode label, where the vibration difference type includes: label abnormal vibration and label normal vibration; If the vibration difference type is label abnormal vibration, perform a value range expansion process on the original vibration value range according to the vibration training data to obtain a target vibration value range; If the vibration difference type is label normal vibration, perform a reduction process on the original vibration value range according to the vibration training data to obtain a target vibration value range.

8. The machine learning-based locomotive vibration pattern recognition method according to claim 7, characterized in that, The judging the vibration difference type between the model verification result and the vibration mode label includes: Obtain the training vibration mode of the vibration mode label and the recognized vibration mode of the model recognition result, where both the training vibration mode and the recognized vibration mode include: lateral vibration mode, longitudinal vibration mode, vertical vibration mode, and normal vibration mode; Judge whether the training vibration mode appears in the recognized vibration mode; If the training vibration mode does not appear in the recognized vibration mode, set the vibration difference type as label abnormal vibration; If the training vibration mode appears in the recognized vibration mode, set the vibration difference type as label normal vibration.

9. The method for identifying locomotive vibration patterns based on machine learning according to claim 7, characterized in that, The performing locomotive vibration mode recognition on the target locomotive by using the target vibration recognition model according to the target vibration acceleration includes: Input the target vibration acceleration into the target vibration recognition model to obtain a predicted vibration value; Compare the predicted vibration value with the target vibration value range numerically to obtain a locomotive vibration mode, where the locomotive vibration mode includes: lateral vibration mode, longitudinal vibration mode, vertical vibration mode, and normal vibration mode.

10. A locomotive vibration pattern recognition system based on machine learning, characterized in that, The system includes: A vibration data acquisition module, configured to collect real-time vibration signals of a target locomotive by using a pre-built acceleration sensor, where the real-time vibration signals include: lateral vibration signals, longitudinal vibration signals, and vertical vibration signals, perform analog-to-digital conversion and error processing on the real-time vibration signals to obtain real-time vibration acceleration, where the real-time vibration acceleration includes: real-time lateral acceleration, real-time longitudinal acceleration, and real-time vertical acceleration; An angle error calculation module, configured to obtain the sensed tilt angle of the target locomotive by using a pre-built gravity sensing device, calculate the vibration tilt angle according to the real-time vibration acceleration, and calculate the real-time deviation angle of the target locomotive according to the sensed tilt angle and the vibration tilt angle; An angle error elimination module is used to determine whether the real-time deviation angle is greater than a preset deviation threshold. If the real-time deviation angle is greater than the deviation threshold, the real-time vibration acceleration is vibration-adjusted according to the real-time deviation angle to obtain a target vibration acceleration. If the real-time deviation angle is not greater than the deviation threshold, the real-time vibration acceleration is used as the target vibration acceleration. A historical vibration data set of the target locomotive is obtained, and an original vibration value range of the target locomotive is set according to the historical vibration data set. The original vibration value range includes: a lateral vibration value range, a longitudinal vibration value range, and a vertical vibration value range; A vibration mode recognition module is used to construct an original vibration recognition model according to the original vibration value range and a pre-constructed convolutional neural network, and perform machine learning on the original vibration recognition model by using the historical vibration data set to obtain a target vibration recognition model. The target vibration recognition model includes a target vibration value range, and constructing the target vibration value range includes: Obtaining vibration training data, inputting the vibration training data into the original vibration recognition model to obtain a vibration mode label and vibration prediction values. The vibration prediction values include: a predicted lateral acceleration, a predicted longitudinal acceleration, and a predicted vertical acceleration; Comparing the vibration prediction values with the original vibration value range to obtain a model recognition result. The model recognition result is a mode label. According to the model recognition result and the vibration mode label, a difference label is determined, and a difference vibration value range is determined in the original vibration value range based on the difference label; Determining a difference acceleration corresponding to the difference vibration value range in the vibration prediction values. If the difference acceleration is less than the minimum value in the difference vibration value range, the difference acceleration is used to replace the minimum value in the difference vibration value range to obtain the target vibration value range. If the difference acceleration is greater than the maximum value in the difference vibration value range, the difference acceleration is used to replace the maximum value in the difference vibration value range to obtain the target vibration value range. According to the target vibration acceleration, the target locomotive is subjected to locomotive vibration mode recognition by using the target vibration recognition model.

Images