Train wheelset dimension measurement correction prediction method and system based on wheel-rail relationship

Abstract

The present invention provides a method and system for measuring, correcting and predicting the size of train wheelsets based on wheel-rail relationship, including obtaining real-time detection status data of train vehicles and corresponding lines; obtaining corrected measurement values ​​based on the trained error correction model of the corresponding line according to the real-time detection status data of the train vehicles; obtaining predicted values ​​based on the trained online prediction model of the train wheels of the corresponding line; the training steps are as follows: for any line, a pre-constructed error correction model is trained using a correction training set to obtain a trained error correction model; for the wheels of the train on any line, a pre-constructed online prediction model is trained using a prediction training set based on the trained error correction model to obtain a trained online prediction model. The reliability of wheelset size measurement is improved, especially the dynamic measurement of heavy-loaded freight cars on the main line, so that it can maintain a high measurement accuracy during long-term operation in complex and harsh environments.

Description

Technical Field

[0001] The invention relates to a method and system for measuring, correcting and predicting train wheel set dimensions based on wheel-rail relationship, and belongs to the technical field of rail transit safety detection and monitoring equipment. Background Art

[0002] The geometric dimension detection of train wheels is one of the key links in the safe operation of rail transit. The detection results directly reflect the wear and running status of train wheels. At present, most wheel dimension detection equipment adopts visual measurement technology and uses multiple sets of visual sensors for combined measurement. It is easily affected by the external environment, especially the strong vibration interference of heavy-duty freight railways, which will cause the accuracy of the visual measurement system to decrease. Among these environmental factors that affect the measurement accuracy and reliability of the wheel dimension detection system, the spatial position relationship between the wheel, rail and detection system is one of the important factors. The compactness of the ballast, the height and flatness of the track, etc., will directly affect the vibration intensity when the train passes and the position of the wheel and the detection system, causing errors in the measurement data.

[0003] When the measurement data obtained by the safety detection equipment has large errors, it will have a certain impact on the operation and maintenance of rail transit. False alarms will cause unnecessary manual review and re-inspection, resulting in a waste of manpower. Missed alarms will make it impossible to detect excessive wheel wear, posing a safety hazard.

[0004] Visual measurement systems usually use cameras as sensors to obtain images of the object being measured or images of feature targets (such as structured light strips) on the object being measured, and calculate the three-dimensional data of the features of the object being measured based on the forming model to achieve three-dimensional measurement. During the system measurement process, the positional relationship of the camera optical system relative to the measured object should be kept as constant as possible, especially when multiple camera optical systems are combined for measurement. Since the rails and wheels are constantly worn over time, the relative positional relationship between the camera and imaging modules such as the light source in the wheelset size detection system and the wheels of the object being measured is not fixed, resulting in changes in measurement errors. In addition, railway workers will perform screening, tamping and rail grinding operations during the engineering operations. Changes in ballast and rails will affect the relative position of the wheels of the object being measured and the vibration of the measurement environment, resulting in changes in measurement errors. At the same time, the weight of cargo loaded on trucks running on different lines is different. For example, the Daqin Railway has fully loaded and unloaded trucks on the up and down lines respectively. There are certain differences in the vibration of the trackside when the trucks pass, and the measurement results will also have different deviations.

[0005] The measurement accuracy of train wheel set dimensions is sub-millimeter, the numerical range of the measurement values ​​is usually large, and there are multiple levels of over-limit safety limits. However, there are many railway lines in my country, and the operating environments are different. The vehicle models and uses are also different. The wear of rails and wheels varies greatly, and the factors affecting data measurement errors are complex. Conventional big data analysis and data prediction methods make it difficult to establish a universal stable prediction model to achieve effective and reliable judgment and correction of measurement data. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method and system for measuring, correcting and predicting train wheelset dimensions based on the wheel-rail relationship. Based on a trained error correction model and an online prediction model, the reliability of wheelset dimension measurement is improved, especially dynamic measurement of heavy-loaded mainline trucks, so that the measurement accuracy can be maintained at a high level during long-term operation in complex and harsh environments, and the function of predicting future wheelset dimension change trends is provided.

[0007] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0008] On the one hand, the present invention discloses a method for measuring and correcting train wheel set dimensions based on wheel-rail relationship, comprising the following steps:

[0009] Obtain real-time detection status data of train vehicles and corresponding routes;

[0010] According to the real-time detection status data, based on a trained error correction model of the corresponding line, a corrected measurement value of the wheelset size of the train vehicle is obtained;

[0011] According to the state information in the real-time detection state data, based on the trained online prediction model of the train wheels of the corresponding line, a predicted value of the size of each wheel of the train vehicle is obtained;

[0012] The training steps of the error correction model and the online prediction model are as follows:

[0013] For any line, obtain the correction training set and pre-build the corresponding error correction model;

[0014] Using the correction training set to train the pre-built error correction model to obtain a trained error correction model;

[0015] For the wheels of trains on any line, based on the trained error correction model, obtain the prediction training set and build the corresponding online prediction model;

[0016] The pre-built online prediction model is trained using the prediction training set to obtain a trained online prediction model.

[0017] Furthermore, the steps for obtaining the correction training set are as follows:

[0018] A correction data set is obtained, wherein the correction data set includes multiple on-site wheelset size detection status data for the selected line. Each detection status data includes the measured value of the geometric size of the wheelset detection equipment, the corresponding manual review value and related status information. The expression of the detection status data Y is:

[0019]

[0020] Where y represents the measured value of the geometric dimensions of the wheelset testing equipment; y m Indicates the manual review value corresponding to this geometric dimension inspection; d y Indicates the mileage of the train wheelset during measurement; Indicates the time interval between the measurement time point and the last screening operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last tamping operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last rail grinding operation on the line where the system is installed;

[0021] In response to the deviation between the measured value in the detection status data Y and the corresponding verification value being less than a preset first threshold, the detection status data is placed in a correction training set; otherwise, the detection status data is deemed to be erroneous data, discarded, and not placed in the correction training set.

[0022] Further, the pre-built error correction model is trained using the correction training set, comprising the following steps:

[0023] Training a pre-built error correction model according to a correction training set to obtain a trained error correction model;

[0024] Based on the trained error correction model, the measured value of each detection state data of the correction training set is corrected to obtain a corrected measured value;

[0025] In response to the deviation between the corrected measurement value and the corresponding verification value being greater than a preset second threshold, the weight of the corresponding detection state data is increased, thereby obtaining a corrected training set with updated weights;

[0026] According to the correction training set after the weight is updated, the trained error correction model is optimized and trained to obtain a trained error correction model.

[0027] Furthermore, the training steps of the error correction model are as follows:

[0028] Input the specified parameters of the detection state data into the error correction model to obtain the error prediction value e of the wheelset dimension measurement value y ;

[0029] Calculate the true value e′ corresponding to the error prediction value y ;

[0030] According to the error prediction value e y and the corresponding truth value e′ y , train the error correction model;

[0031] The error correction model is expressed as follows:

[0032]

[0033] In the formula, e y represents the error prediction value of the wheelset dimension measurement value, and F(*) represents the error correction model;

[0034] The truth value e′ y The expression is: e′ y =y m -y;

[0035] The corrected measured value y j The expression of y is: j =e y +y.

[0036] Furthermore, the steps for obtaining the prediction training set are as follows:

[0037] Obtaining multiple geometric dimension measurement values ​​and corresponding status information of wheelset detection equipment for train wheels on the selected line;

[0038] Based on the trained error correction model, each measurement value is corrected to obtain a corrected measurement value;

[0039] According to the corrected measurement values ​​and the corresponding state information, a set of corrected data is obtained; the set of corrected data is the prediction training set.

[0040] Further, the pre-built online prediction model is trained using the prediction training set, including the following steps:

[0041] According to the prediction training set, the pre-built online prediction model is trained to obtain a trained online prediction model;

[0042] Based on the trained online prediction model, the corrected measurement value in any correction data is predicted to obtain a predicted value;

[0043] In response to the deviation between the predicted value and the corresponding corrected measured value being greater than a preset third threshold, the corresponding corrected data is added as the weight of the training data, thereby obtaining a predicted training set with updated weights;

[0044] According to the prediction training set after the weight is updated, the trained online prediction model is optimized and trained to obtain a trained online prediction model.

[0045] Furthermore, the training steps of the online prediction model are as follows:

[0046] For any calibration data, the state information in the calibration data is input into the online prediction model to obtain the predicted value of the wheel size;

[0047] The corrected measured value in the corrected data is taken as the true value of the predicted value;

[0048] Training an online prediction model according to the predicted values ​​and the corresponding true values;

[0049] The online prediction model is expressed as follows:

[0050]

[0051] In the formula, p y represents the predicted value of wheel size; cn represents vehicle number; wn represents wheel number; F cn,wn (*) indicates the online prediction model of train wheels for the selected line; d y Indicates the mileage of the vehicle wheelset; Indicates the time interval between the measurement time point and the last screening operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last tamping operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last rail grinding operation on the line where the system is installed.

[0052] Furthermore, the error correction model includes a time series prediction model or a regression prediction model, and the time series prediction model includes an ARMA linear model, an ARIMA linear model, an LSTM model, a prophet model, and an LSTNet deep learning network;

[0053] The online prediction model includes a time series prediction model or a regression prediction model, and the time series prediction model includes an ARMA linear model, an ARIMA linear model, an LSTM model, a prophet model and an LSTNet deep learning network.

[0054] Further, in response to the existence of a corresponding verification value for a measurement value in the real-time detection status data, and when a deviation between the measurement value and the corresponding verification value is less than a preset first threshold, the real-time detection status data is put into a correction training set to implement training updates for the trained error correction model;

[0055] In response to the predicted value obtained by processing the real-time detection status data through the online prediction model, when the deviation between the predicted value obtained through processing the error correction model and the corrected predicted value is less than a preset third threshold, the real-time detection status data is placed in the prediction training set to achieve training updates for the trained prediction model.

[0056] On the other hand, the present invention discloses a train wheelset dimension measurement correction prediction system based on wheel-rail relationship, comprising:

[0057] Wheelset dimension detection data acquisition and measurement unit, used to obtain real-time detection status data of train vehicles and corresponding lines;

[0058] A data analysis and correction unit, used to execute the aforementioned train wheel set size measurement correction prediction method based on wheel-rail relationship, and obtain the corrected measurement value and prediction value of the wheelset size of the train vehicle based on the real-time detection state data of the train vehicle and the trained error correction model and online prediction model of the corresponding line;

[0059] A data storage unit, used to store real-time detection status data of train vehicles, corrected measurement values ​​and predicted values;

[0060] Data display and business processing unit, used to display data, query data and manage data.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] First, according to the different factors that affect the measurement accuracy of the wheelset size visual measurement system, big data analysis technology is used to establish an error correction model for the wheelset size measurement system for each line, and the measurement data is corrected to improve the measurement accuracy; secondly, according to the different characteristics of each wheel, big data analysis technology is used to establish an online prediction model for each wheel size parameter based on the corrected measurement data to obtain the predicted value of the wheelset size. The present invention can improve the reliability of wheelset size measurement, especially the dynamic measurement of heavy-loaded trucks on the main line, so that it can maintain a high measurement accuracy during long-term operation in complex and harsh environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flow chart of the method for measuring, correcting and predicting the size of a train wheelset based on the wheel-rail relationship provided in Example 1;

[0064] Figure 2 It is a structural schematic diagram of the train wheelset dimension measurement, correction and prediction system based on wheel-rail relationship provided in Example 2. DETAILED DESCRIPTION

[0065] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0066] Example 1.

[0067] This embodiment 1 provides a method for measuring and correcting train wheelset dimensions based on a wheel-rail relationship, comprising the following steps:

[0068] Obtain real-time detection status data of train vehicles and corresponding routes;

[0069] According to the real-time detection status data of the train vehicle, based on the trained error correction model of the corresponding line, the corrected measurement value of the wheelset size of the train vehicle is obtained;

[0070] According to the status information in the real-time detection status data, based on the trained online prediction model of the train wheels of the corresponding line, the predicted value of the size of each wheel of the train vehicle is obtained;

[0071] The training steps of the error correction model and the online prediction model are as follows:

[0072] For any line, obtain the correction training set and pre-build the corresponding error correction model;

[0073] Using the correction training set to train the pre-built error correction model to obtain a trained error correction model;

[0074] For the wheels of trains on any line, based on the trained error correction model, obtain the prediction training set and build the corresponding online prediction model;

[0075] The pre-built online prediction model is trained using the prediction training set to obtain a trained online prediction model.

[0076] The technical concept of the present invention is as follows: first, according to the different factors that affect the measurement accuracy of the wheelset size visual measurement system, big data analysis technology is used to establish an error correction model for the wheelset size measurement system according to each line, and the measurement data is corrected to improve the measurement accuracy; secondly, according to the different characteristics of each wheel, big data analysis technology is used to establish an online prediction model for each wheel size parameter based on the corrected measurement data to obtain the future prediction value of the wheelset size. The present invention can improve the reliability of wheelset size measurement, especially the dynamic measurement of heavy-loaded trucks on the main line, so that it can maintain a high measurement accuracy during long-term operation in complex and harsh environments.

[0077] like Figure 1 The training steps of the error correction model and the online prediction model are as follows:

[0078] S1. For each line, obtain the correction training set and pre-build the corresponding error correction model.

[0079] S1.1. The steps for obtaining the calibration training set are as follows:

[0080] The calibration data set is obtained. The calibration data set includes multiple on-site wheel set size detection status data for the selected line. Each detection status data includes the measured value of the geometric size of the wheel set detection equipment, the corresponding manual review value and related status information. The expression of the detection status data Y is:

[0081]

[0082] Where y represents the measured value of the geometric dimensions of the wheelset testing equipment; y m Indicates the manual review value corresponding to this geometric dimension inspection; d y Indicates the mileage of the train wheelset during measurement; Indicates the time interval between the measurement time point and the last screening operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last tamping operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last rail grinding operation on the line where the system is installed.

[0083] Among them, d y , It can be obtained from railway information systems, such as railway freight car technical management system HMIS.

[0084] In response to the deviation between the measured value in the detection status data Y and the corresponding verification value being less than the preset first threshold T1, the detection status data Y is put into the correction training set; otherwise, the detection status data Y is determined to be erroneous data, discarded, and not put into the correction training set.

[0085] It should be noted that in this embodiment, the on-site wheelset size detection status data is accumulated and collected for each route respectively, and the accumulated detection status data are as different as possible on site and as large in quantity as possible, and the detection status data are as different as possible in vehicle types and acquisition times, including different time periods during the day and night, different seasons of the year, and different weather conditions.

[0086] S1.2, the expression of the error correction model is as follows:

[0087]

[0088] In the formula, e y represents the error prediction value of the wheelset dimension measurement value, and F(*) represents the error correction model;

[0089] Corrected measurement value y j The expression of y is: j =e y +y.

[0090] It should be noted that the error correction model in this embodiment can select a time series prediction model, such as an ARMA linear model, an ARIMA linear model, an LSTM model, a prophet model, and an LSTNet deep learning network; or a regression prediction model can be selected.

[0091] S2. Use the correction training set to train the pre-built error correction model to obtain a trained error correction model.

[0092] Training the pre-built error correction model using the correction training set includes the following steps:

[0093] Training a pre-built error correction model according to a correction training set to obtain a trained error correction model;

[0094] Based on the trained error correction model, the measured value of each detection state data of the correction training set is corrected to obtain a corrected measured value;

[0095] In response to the deviation between the corrected measurement value and the corresponding verification value being greater than a preset second threshold value T2, where T1>T2, the weight of the corresponding detection state data is increased, thereby obtaining a corrected training set with updated weights;

[0096] According to the correction training set after the weight is updated, the trained error correction model is optimized and trained to obtain a trained error correction model.

[0097] It should be noted that in this embodiment, whether to start optimization training is determined based on the sum of the weights of the correction training set after the weights are updated. For example, when the sum of the weights is greater than the set weight threshold, model optimization training is started.

[0098] In actual use, the correction training set will be expanded according to the actual detection status data. When the expanded data of the correction training set is greater than the set expansion threshold, the model optimization training will also be started.

[0099] Furthermore, the training steps of the error correction model are as follows:

[0100] Input the specified parameters of the detection state data into the error correction model to obtain the error prediction value e of the wheelset dimension measurement value y ;

[0101] Calculate the true value e′ corresponding to the error prediction value y , e′ y =y m -y;

[0102] According to the error prediction value e y and the corresponding truth value e′ y , the error correction model is trained. In this embodiment, conventional training methods, such as minimizing the loss function, are adopted, which is not limited here.

[0103] S3. For the wheels of a train on any line, based on the trained error correction model, obtain a prediction training set and build a corresponding online prediction model.

[0104] S3.1. The steps for obtaining the prediction training set are as follows:

[0105] Obtaining multiple geometric dimension measurement values ​​and corresponding status information of wheelset detection equipment for train wheels on the selected line;

[0106] Based on the trained error correction model, each measurement value is corrected to obtain a corrected measurement value;

[0107] According to the corrected measurement values ​​and the corresponding state information, a set of corrected data is obtained; the set of corrected data is the prediction training set.

[0108] In this embodiment, the on-site wheelset size detection status data is accumulated and collected for each wheelset on each line. The accumulated detection status data are as different as possible on site and as large in quantity as possible. The detection status data are as different as possible for different vehicle types and acquisition times, including different time periods during the day and night, different seasons of the year, and different weather conditions. Then, based on the error correction model trained in step S2, corresponding correction data is obtained.

[0109] S3.2, the expression of the online prediction model is as follows:

[0110]

[0111] In the formula, p y represents the predicted value of wheel size; cn represents vehicle number; wn represents wheel number; F cn,wn (*) indicates the online prediction model of train wheels for the selected line; d y Indicates the mileage of the vehicle wheelset; Indicates the time interval between the measurement time point and the last screening operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last tamping operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last rail grinding operation on the line where the system is installed.

[0112] The online prediction model in this embodiment can select a time series prediction model, such as an ARMA linear model, an ARIMA linear model, an LSTM model, a prophet model, and an LSTNet deep learning network; or a regression prediction model can be selected.

[0113] S4. Use the prediction training set to train the pre-built online prediction model to obtain a trained online prediction model.

[0114] Training the pre-built online prediction model using the prediction training set includes the following steps:

[0115] According to the prediction training set, the pre-built online prediction model is trained to obtain a trained online prediction model;

[0116] Based on the trained online prediction model, the corrected measurement value in any correction data is predicted to obtain a predicted value;

[0117] In response to the deviation between the predicted value and the corresponding corrected measured value being greater than a preset third threshold, the corresponding corrected data is added as the weight of the training data, thereby obtaining a predicted training set with updated weights;

[0118] According to the prediction training set after the weight is updated, the trained online prediction model is optimized and trained to obtain a trained online prediction model.

[0119] Furthermore, the training steps of the online prediction model are as follows:

[0120] For any calibration data, the status information in the calibration data, i.e., the mileage d of the vehicle wheelset, is y and track status information Input the online prediction model to obtain the predicted value of wheel size;

[0121] The corrected measured value in the corrected data is taken as the true value of the predicted value;

[0122] Train the online prediction model based on the predicted value and the corresponding true value;

[0123] It should be noted that in this embodiment, whether to start optimization training will be determined based on the sum of the weights of the prediction training set after the weights are updated. For example, when the sum of the weights is greater than the set weight threshold, model optimization training will be started.

[0124] In actual use, the prediction training set will be expanded according to the actual detection status data. When the expanded data of the prediction training set is greater than the set expansion threshold, the model optimization training will also be started.

[0125] S5. Actual use of the model.

[0126] Obtain real-time detection status data of train vehicles and corresponding routes;

[0127] According to the real-time detection status data of the train vehicle, based on the trained error correction model of the corresponding line, the corrected measurement value of the train vehicle is obtained;

[0128] According to the state information in the real-time detection state data, based on the trained online prediction model of the train wheels of the corresponding line, a prediction value of each wheel of the train vehicle is obtained;

[0129] In response to the existence of a corresponding verification value for the measurement value in the real-time detection state data, and when the deviation between the measurement value and the corresponding verification value is less than a preset first threshold, the real-time detection state data is put into a correction training set to implement training and updating of the trained error correction model. The training steps are detailed in S2;

[0130] It should be noted that not every measurement value will have a corresponding manual review value, that is, the user will selectively perform manual review based on some measurement values, and the detection status data in the correction training set all contain the manual review values ​​corresponding to the measurement values. Therefore, not all real-time detection status data can be included in the correction training set, and data screening is required.

[0131] In response to the predicted value obtained by the real-time detection status data through the online prediction model processing, when the deviation between the predicted value obtained through the error correction model processing and the corrected predicted value is less than a preset third threshold, the real-time detection status data is put into the prediction training set to realize the training update of the trained prediction model. The training steps are detailed in S4.

[0132] Example 2.

[0133] This embodiment 2 provides a train wheelset dimension measurement correction prediction system based on wheel-rail relationship, such as Figure 2 As shown, including:

[0134] Wheelset dimension detection data acquisition and measurement unit, used to obtain real-time detection status data of train vehicles and corresponding lines;

[0135] A data analysis and correction unit, used to execute the train wheelset size measurement correction prediction method based on wheel-rail relationship of embodiment 1, and obtain the corrected measurement value and prediction value of the wheelset size of the train vehicle based on the real-time detection state data of the train vehicle and the trained error correction model and online prediction model of the corresponding line;

[0136] A data storage unit, used for storing real-time measurement data of train vehicles, corrected measurement values ​​and predicted values;

[0137] Data display and business processing unit, used to display data, query data and manage data.

[0138] Among them, the wheelset size detection, collection and measurement unit, the data storage unit and the data analysis and correction unit are connected in sequence, the data storage unit communicates bidirectionally with the data analysis and correction unit, and the data storage unit communicates bidirectionally with the data display and business processing unit.

[0139] The specific working principle is as follows: the wheel set size detection status data acquisition and measurement unit collects image data, the vehicle number and wheel number information of the measured wheel set through the structured light three-dimensional measurement method, and calculates the three-dimensional information of the wheel according to the structured light three-dimensional reconstruction method, and then calculates the measurement value of each geometric parameter according to the definition of the wheel set geometric size index in the relevant railway standards, that is, the real-time detection status data of the train vehicle. The result is then uploaded to the data storage unit for storage and management, and the data in the data storage unit is extracted through the data analysis and correction unit, and error correction and online prediction model are performed. The processing result is returned to the data storage unit, and the measured data after analysis and correction in the data storage unit is obtained and displayed through the data display and business processing unit. The data display and business processing unit provides data query, statistics, and management functions to cooperate with and support the corresponding business processing of users, and also provides user business operation functions such as manual review of data records, user management, and scheduling management.

[0140] The data storage unit uses a database to store information in strips according to vehicle number, wheel number, measurement value, mileage, and operation time information such as cleaning, tamping, and rail grinding, and records the measurement results of the wheelset size detection equipment and the correction results processed by the data analysis and correction unit. The results can be provided to users for operation and maintenance business operations. At the same time, the current equipment training data recorded is used for optimization training of error correction models and online prediction models; the data storage unit also provides data docking functions and interfaces with other information systems of users (such as HMIS systems, etc.), transmits wheelset size detection status data to other information systems, and obtains data from other information systems, such as mileage and operation time information such as cleaning, tamping, and rail grinding.

[0141] The data analysis and correction unit takes data from the data storage unit, obtains correction values ​​and prediction values, and optimizes and trains the error correction model and the online prediction model. The data analysis and correction unit returns the processed results to the data storage unit.

[0142] It should be noted that the train wheelset size measurement, correction and prediction system based on wheel-rail relationship provided in this embodiment 2 is installed at the application site according to the line deployment, and is connected to at least one set of wheelset size detection systems on the line. Preferably, as many wheelset size detection systems of the line as possible are connected to collect the measurement data of each system and the wheel and wheel-rail relationship status data.

[0143] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0144] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0145] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0146] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0147] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A train wheelset dimension measurement correction prediction method based on wheel-rail relationship, characterized in that: The steps include: Obtain real-time detection status data of train vehicles and corresponding routes; According to the real-time detection status data, based on a trained error correction model of the corresponding line, a corrected measurement value of the wheelset size of the train vehicle is obtained; According to the state information in the real-time detection state data, based on the trained online prediction model of the train wheels of the corresponding line, a predicted value of the size of each wheel of the train vehicle is obtained; The training steps of the error correction model and the online prediction model are as follows: For any line, obtain the correction training set and pre-build the corresponding error correction model; Using the correction training set to train the pre-built error correction model to obtain a trained error correction model; For the wheels of trains on any line, based on the trained error correction model, obtain the prediction training set and build the corresponding online prediction model; Using the prediction training set to train the pre-built online prediction model to obtain a trained online prediction model; The steps for obtaining the prediction training set are as follows: Obtaining the measured values ​​of geometric dimensions and corresponding status information of multiple wheelset detection equipment for train wheels on the selected line; Based on the trained error correction model, each measurement value is corrected to obtain a corrected measurement value; According to the corrected measurement value and the corresponding state information, a correction data is obtained; the set of the correction data is the prediction training set; The pre-built online prediction model is trained using the prediction training set, including the following steps: According to the prediction training set, the pre-built online prediction model is trained to obtain a trained online prediction model; Based on the trained online prediction model, the corrected measurement value in any correction data is predicted to obtain a predicted value; In response to the deviation between the predicted value and the corresponding corrected measured value being greater than a preset third threshold, the corresponding corrected data is added as the weight of the training data, thereby obtaining a predicted training set with updated weights; According to the prediction training set after weight update, the trained online prediction model is optimized and trained to obtain a trained online prediction model; The training steps of the online prediction model are as follows: For any calibration data, the state information in the calibration data is input into the online prediction model to obtain the predicted value of the wheel size; The corrected measured value in the corrected data is taken as the true value of the predicted value; Training an online prediction model according to the predicted values ​​and the corresponding true values; The online prediction model is expressed as follows: In the formula, p y represents the predicted value of wheel size; cn represents vehicle number; wn represents wheel number; F cn,wn (*) indicates the online prediction model of train wheels for the selected line; d y Indicates the mileage of the vehicle wheelset; Indicates the time interval between the measurement time point and the last screening operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last tamping operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last rail grinding operation on the line where the system is installed.

2. The train wheelset dimension measurement correction prediction method based on wheel-rail relationship according to claim 1 is characterized in that: The steps for obtaining the correction training set are as follows: A correction data set is obtained, wherein the correction data set includes multiple on-site wheelset size detection status data for the selected line. Each detection status data includes the measured value of the geometric size measured by the wheelset detection equipment, the corresponding manual review value and related status information. The expression of the detection status data Y is: Where y represents the measured value of the geometric dimension measured by the wheelset testing equipment; y m Indicates the manual review value corresponding to this geometric dimension inspection; d y Indicates the mileage of the train wheelset during measurement; Indicates the time interval between the measurement time point and the last screening operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last tamping operation of the line where the system is installed; Indicates the time interval between the measurement time point and the last rail grinding operation on the line where the system is installed; In response to the deviation between the measured value in the detection status data Y and the corresponding verification value being less than a preset first threshold, the detection status data is placed in a correction training set; otherwise, the detection status data is deemed to be erroneous data, discarded, and not placed in the correction training set.

3. The train wheelset dimension measurement correction prediction method based on wheel-rail relationship according to claim 2 is characterized in that: The pre-built error correction model is trained using the correction training set, comprising the following steps: Training a pre-built error correction model according to a correction training set to obtain a trained error correction model; Based on the trained error correction model, the measured value of each detection state data of the correction training set is corrected to obtain a corrected measured value; In response to the deviation between the corrected measurement value and the corresponding verification value being greater than a preset second threshold, the weight of the corresponding detection state data is increased, thereby obtaining a corrected training set with updated weights; According to the correction training set after the weight is updated, the trained error correction model is optimized and trained to obtain a trained error correction model.

4. The train wheelset dimension measurement correction prediction method based on wheel-rail relationship according to claim 3 is characterized in that: The training steps of the error correction model are as follows: Input the specified parameters of the detection state data into the error correction model to obtain the error prediction value e of the wheelset dimension measurement value y ; Calculate the error prediction value e y The corresponding truth value e′ y ; According to the error prediction value e y and the corresponding truth value e′ y , train the error correction model; The error correction model is expressed as follows: In the formula, e y represents the error prediction value of the wheelset dimension measurement value, and F(*) represents the error correction model; The truth value e′ y The expression is: e′ y =y m -y; The corrected measured value y j The expression of y is: j =e y +y.

5. The train wheelset dimension measurement correction prediction method based on wheel-rail relationship according to claim 1 is characterized in that: The error correction model includes a time series prediction model or a regression prediction model, and the time series prediction model includes an ARMA linear model, an ARIMA linear model, an LSTM model, a prophet model, and an LSTNet deep learning network; The online prediction model includes a time series prediction model or a regression prediction model, and the time series prediction model includes an ARMA linear model, an ARIMA linear model, an LSTM model, a prophet model and an LSTNet deep learning network.

6. The train wheelset dimension measurement and correction method based on wheel-rail relationship according to claim 1 is characterized in that: In response to the existence of a corresponding verification value for a measurement value in the real-time detection status data, and when a deviation between the measurement value and the corresponding verification value is less than a preset first threshold, the real-time detection status data is placed in a correction training set to implement training updates for the trained error correction model; In response to the predicted value obtained by processing the real-time detection status data through the online prediction model, when the deviation between the predicted value obtained through processing the error correction model and the corrected predicted value is less than a preset third threshold, the real-time detection status data is placed in the prediction training set to achieve training updates for the trained prediction model.

7. A train wheelset dimension measurement, correction and prediction system based on wheel-rail relationship, characterized in that: include: Wheelset dimension detection data acquisition and measurement unit, used to obtain real-time detection status data of train vehicles and corresponding lines; A data analysis and correction unit, used to execute the train wheelset size measurement, correction and prediction method based on wheel-rail relationship according to any one of claims 1 to 6, and obtain the corrected measurement value and prediction value of the wheelset size of the train vehicle based on the real-time detection status data of the train vehicle and the trained error correction model and online prediction model of the corresponding line; A data storage unit, used to store real-time detection status data of train vehicles, corrected measurement values ​​and predicted values; Data display and business processing unit, used to display data, query data and manage data.

Images