A seamless line longitudinal action monitoring method based on digital twinning

Abstract

The application discloses a seamless line longitudinal action monitoring method based on digital twinning, which comprises the following steps: S1, establishing a special section type of a field seamless line; S2, collecting characteristic point data by using a temperature sensor and a displacement sensor; S3, establishing a finite element simulation model, and calculating the temperature, displacement and rail longitudinal additional force of the special section; S4, establishing a mapping model to obtain a mapping relationship from a physical database to a virtual database; S5, comparing the mapping data with data not participating in the establishment of the mapping model, and verifying the model; and S6, constructing a digital twinning virtual body completely equal to the special section of the seamless line, and interacting with the field seamless line. The application can judge the service state of the field seamless line in real time based on the digital twinning technology, and can simulate various maintenance measures of the seamless line, so that a maintenance and repair plan can be made more efficiently.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of railway operation and maintenance simulation, and in particular to a seamless track longitudinal action monitoring method based on digital twinning. BACKGROUND

[0002] The permafrost area of the plateau railway has the characteristics of high-cold hypoxia, variable climate, unclear seasons, large daily temperature difference, long negative temperature period, strong solar radiation, common large slopes, more thick bed sections, and common roadbed thermal subsidence diseases. These special climatic and geological conditions and track structure directly cause the difference in the value and variation of the state index of the seamless track from other areas.

[0003] And when the train runs on a long and large slope, there is a problem of large longitudinal force caused by frequent braking, adhesion traction, and gravity (referred to as braking load), which is more complex than in flat slope sections. The rail is affected by the repeated braking load of the train, and the seamless track starts to climb on one side. The temperature rise caused by frequent braking of the train will also cause the seamless track to climb on one side.

[0004] In the prior art, the additional force calculation of beam-rail interaction generally uses large finite element method to analyze and calculate through the establishment of a three-dimensional space coupled solid model. However, the bridge and track in the actual environment are very complex engineering structures, and it is very difficult to accurately calculate and analyze through simulation. In addition, it is time-consuming and laborious to arrange sensors on site to monitor many points.

[0005] Therefore, in the prior art, under the premise of meeting engineering application, the track, bridge and pier are generally simplified into beam and plate units, and a seamless track longitudinal force calculation model is established based on the simplified beam and plate units. The calculation model can basically reflect the mutual interaction of beam and rail, but can only meet the needs of simple engineering and cannot monitor the horizontal and longitudinal deformation characteristics of the seamless track in complex environments, as well as the stress and deformation characteristics of each part of special sections such as high-pier large-span bridges, bridge-tunnel transition sections and seamless turnouts. SUMMARY

[0006] In order to solve the problem that the prior art cannot monitor the horizontal and longitudinal deformation characteristics of the seamless track in complex environments and the stress and deformation characteristics of each part of special sections, based on digital twinning technology, a simulation seamless track matching the special section of the field seamless track is constructed to solve the above problems.

[0007] The present application discloses a seamless track longitudinal action monitoring method based on digital twinning, comprising the following steps:

[0008] S1, determining the type of the special section of the field seamless track;

[0009] S2, collecting feature point data by using temperature sensors and displacement sensors;

[0010] S3, establishing a finite element simulation model to calculate the temperature, displacement and longitudinal additional force of the special section of the rail;

[0011] S4, establishing a mapping model to obtain the mapping relationship from the physical database to the virtual database;

[0012] S5, comparing the mapping data with the data not participating in the establishment of the mapping model to verify the model;

[0013] S6, constructing a digital twin virtual body completely equivalent to the special section of the seamless line through 3D rendering technology, and interacting with the field seamless line.

[0014] Preferably, the special section of the seamless line in S1 includes a high-pier large-span bridge, a bridge-tunnel transition section and a seamless turnout.

[0015] Preferably, in S2, the feature point data includes:

[0016] The displacement and temperature of the rail at the end of the high-pier large-span bridge, the displacement and temperature of the beam joint;

[0017] The displacement and temperature of the rail at the end of the simply supported beam of the first span at the tunnel portal of the bridge-tunnel transition section, the displacement and temperature of the rail at each interval of the temperature transition zone;

[0018] The displacement change of the rail at the tip of the point rail relative to the basic rail, the displacement change of the rail at the tip of the point rail relative to the wing rail, the displacement at the front of the turnout and the displacement at the position of the position limiter, and the displacement of the basic rail near the guard rail.

[0019] Preferably, S3 includes the following steps:

[0020] Establish a seamless line model of a high-pier large-span bridge, according to the principle of beam-rail interaction, establish a lightweight calculation model of the seamless line-bridge-pier for calculating the longitudinal force of the seamless line on the high-pier large-span bridge, solve the model by using the nonlinear finite element method, and calculate the mechanical properties of the seamless line of the high-pier large-span bridge under different working conditions;

[0021] Establish a seamless line model of a bridge-tunnel transition section, according to the principle of beam-rail interaction, establish a lightweight analysis model of the track-substructure in the connected area of the bridge and the tunnel, solve the model by using the nonlinear finite element method, and calculate the mechanical properties of the ballastless track seamless line of the bridge-tunnel transition section under different working conditions;

[0022] Establish a seamless turnout model, according to the principle of beam-rail interaction, establish a lightweight analysis model of the track-substructure in the turnout area, and solve the model by using the nonlinear finite element method.

[0023] Preferably, the data of the physical database in S4 is physical data collected by the sensors in S2, and the data in the virtual database is virtual data calculated in S3.

[0024] Preferably, the establishment of the mapping model comprises the following steps:

[0025] With the physical database as input and the virtual database as output, a neural network model of the special section seamless line under the action of multiple working conditions is constructed and trained, and the training data of the neural network model is a physical data set of N sensor collection points in the physical database and a virtual data set of M corresponding virtual data calculated by the finite element simulation model in the virtual database, and a mapping model is trained based on the physical data set and the virtual data set.

[0026] Preferably, the neural network model is a CNN-Bi-GRU-GWO neural network model with adaptive hyperparameters, and the neural network model comprises four layers of networks.

[0027] The first layer of network is an input layer, and one-dimensional time series data is extracted by a CNN network;

[0028] The second layer of network is a Bi-GRU time series network layer, and the CNN feature extraction data of the previous layer is taken as input to further extract features in the time scale;

[0029] The third layer of network is a neural network hyperparameter optimization layer, which adaptively optimizes the hyperparameters of the CNN network and the Bi-GRU time series network;

[0030] The fourth layer of network is an output layer, which is constructed to be trained and optimized, and the mean square error is taken as the objective function of the model, and the model is trained by minimizing the mean square error.

[0031] Preferably, S5 comprises the following steps:

[0032] S51. In the construction of the mapping model, a part of the data in the physical database and the virtual database that has the same test point and the same data type is removed, and it is proposed that part of the data does not participate in the construction of the mapping model, and is only used for the verification of the mapping model;

[0033] S52. The physical data of the removed point in the physical database is calculated to obtain the virtual data corresponding to the removed point by the mapping model;

[0034] S53. The calculated virtual data is compared with the physical data of the point;

[0035] S54, judging the virtual data output by the mapping model, if the error of the virtual data and the physical data corresponding to the point is within the controllable range, the verification of the mapping model is completed, if the error is outside the controllable range, the parameters of the mapping model are modified and optimized.

[0036] Preferably, the S6 comprises the following steps:

[0037] By controlling the parameters of the digital twin virtual body including fastener resistance, ballast resistance, environmental temperature, locking rail temperature and train live load;

[0038] Through the mapping model, the indexes of the seamless line including rail temperature, climbing amount and rail longitudinal additional force are calculated, and a maintenance plan meeting the specification requirements is formulated;

[0039] After the maintenance and repair of the seamless line, the data collected from the seamless line is fed back and interactively verified with the seamless line digital twin virtual body.

[0040] The beneficial effects of the present application are:

[0041] (1) Based on the digital twin technology, a digital twin virtual body matched with the special section seamless line is constructed, the service state of the on-site seamless line can be judged in real time through the virtual body, and the change situation and historical change law of each index of the special section seamless line are more intuitively mastered.

[0042] (2) Through the seamless line digital twin virtual body, various seamless line maintenance measures can be simulated, and the change situation of each index of the line after maintenance and repair is calculated, so that the maintenance and repair plan can be more efficiently formulated. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The flow chart of the seamless line longitudinal action monitoring method based on digital twinning of the embodiment of the present application;

[0044] Figure 2 The characteristic point monitoring schematic diagram of the high-pier large-span bridge of the embodiment of the present application;

[0045] Figure 3 The characteristic point monitoring schematic diagram of the temperature and longitudinal displacement at the bridge-tunnel transition section of the embodiment of the present application;

[0046] Figure 4 The characteristic point monitoring schematic diagram of the lateral displacement at the bridge-tunnel transition section of the embodiment of the present application;

[0047] Figure 5 The schematic diagram of the temperature equivalent load application of the bridge-tunnel transition section area model of the embodiment of the present application;

[0048] Figure 6 The characteristic point monitoring schematic diagram of the seamless turnout of the embodiment of the present application;

[0049] Figure 7 The schematic diagram of the data acquisition module sensor acquisition point of the embodiment of the application.

[0050] The reference signs are as follows:

[0051] 11 - rail; 12 - continuous beam; 13 - simply supported beam; 14 - lower foundation; 15 - roof temperature measuring point; 16 - web temperature measuring point; 17 - beam end displacement measuring point; 18 - beam joint; 21 - beam end; 22 - fastener; 31 - basic rail; 32 - switch rail; 33 - switch; 34 - limiter; 35 - guard rail; 36 - wing rail; 37 - center rail. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below with reference to the drawings and examples.

[0053] The embodiment of the application discloses a seamless line longitudinal action monitoring method based on digital twinning, and the flow is as shown in Figure 1 The method comprises the following steps:

[0054] S1, the type of the special section of the field seamless line is established, so that the monitoring of the characteristic points and the establishment of the simulation model are carried out according to the actual situation of the field. The special section of the seamless line includes a high-pier large-span bridge, a bridge-tunnel transition section and a seamless turnout.

[0055] S2, temperature sensors and displacement sensors are used to collect characteristic point data, including:

[0056] The displacement and temperature of the rail at the beam end of the high-pier large-span bridge, the displacement and temperature of the beam joint, the displacement and temperature of the rail at the first span simply supported beam end of the tunnel portal of the bridge-tunnel transition section, the displacement and temperature of the rail at each interval of the temperature transition zone, the displacement change of the switch rail tip relative to the basic rail, the displacement change of the center rail tip relative to the wing rail, the displacement at the front of the turnout and the limiter, and the displacement of the basic rail near the guard rail.

[0057] In a specific embodiment, by arranging temperature and displacement sensors on the seamless line for all-weather testing, the temperature and displacement changes of the seamless line in the special sections such as the high-pier large-span bridge, the bridge-tunnel transition section and the seamless turnout are collected, as well as the environmental parameters such as the air temperature and solar radiation in the special sections. The temperature acquisition instrument used is an infrared thermal imager, and the displacement acquisition instrument is a high-definition camera using the Moire fringe micro-displacement measurement technology, which can realize non-destructive monitoring of the seamless line. The sensor test position is selected as the part with larger displacement in the special section of the seamless line.

[0058] As shown in Figure 2As shown, the displacement and temperature of the rail 11 at the end of the long-pier large-span bridge, the continuous beam 11 and the simply supported beam 13 are monitored and collected. The temperature at the top of the beam body 15, the web of the beam body 16 and the rail waist of the rail 11 is monitored and collected. As shown Figure 3 As shown, the temperature at the 50m, 40m, 30m, 20m, 10m, 8m, 6m, 4m, 2m, 0m and the end of the beam of the bridge-tunnel transition section is monitored and collected. The longitudinal displacement of the rail at the end of the beam 21, the hole opening, 25m inside the hole and 50m inside the hole of the bridge-tunnel transition section is monitored and collected. As shown Figure 4 As shown, the transverse displacement of the rail 11 within the range of 6m in the tunnel of the bridge-tunnel transition section is monitored. As shown Figure 6 As shown, the displacement change of the switch rail 32 tip relative to the basic rail 31 and the tip of the heart rail 37 relative to the wing rail 36 is monitored and collected, the displacement of the basic rail before the switch and the displacement of the positioner 34 are monitored and collected, and the displacement of the basic rail 31 near the guard rail 35 is monitored and collected. About 10 sleeper points are arranged before the switch, 2 points are arranged near the switch 33, and a total of 3 points are arranged. A total of 2 points are arranged on the switch heart rail 37, and a total of 5 temperature force points are arranged, and the rail waist temperature of the monitoring points is monitored. In this embodiment, only part of the key feature points of the seamless line need to be monitored and collected, the site point arrangement is reduced, the site monitoring cost is saved, and remote non-destructive monitoring is adopted, which reduces the safety hidden danger of site monitoring.

[0059] S3, a finite element simulation model is established, and the temperature, displacement and longitudinal additional force of the rail of the special section are calculated.

[0060] A seamless line model of a high-pier large-span bridge is established. According to the principle of beam-rail bridge interaction, a lightweight calculation model of the seamless line-bridge-pier of the high-pier large-span bridge is established, the model is solved by using a nonlinear finite element method, and the mechanical properties of the seamless line of the high-pier large-span bridge under different working conditions are calculated. The rail is simulated by using a beam element; the fastener is simulated by using a nonlinear spring element; the parameters are valued according to the current specification or the measured results; the beam body and the pier are simulated by using a beam element, the beam body-pier top shares a node, the foundation at the bottom of the pier is simulated by using a node elastic support, and the remaining part is simulated by using a solid element.

[0061] A seamless line model of a bridge-tunnel transition section is established. According to the principle of beam-rail interaction, a track-underground foundation lightweight analysis model of the bridge-tunnel connecting area is established, the model is solved by using a nonlinear finite element method, and the mechanical properties of the ballastless track seamless line of the bridge-tunnel transition section under different working conditions are calculated. The rail is simulated by using a beam element, the nonlinear constraint parameters of the longitudinal resistance of the fastener and the longitudinal resistance of the ballast are valued according to the current specification or the measured results; the tunnel solid element only simulates the tunnel lining, the simulation length is 20-50m according to the field situation, and the remaining part is simulated by using a solid element; as shown Figure 5As shown, the temperature load of the bridge-tunnel transition section can be equivalent to a distributed force q on the center axis of the rail, considering the alternating change and temperature gradient during the calculation:

[0062]

[0063] wherein E is the elastic modulus of the rail, A is the cross-sectional area of the rail, a is the linear expansion coefficient of the rail, ΔT is the rail temperature difference inside and outside the tunnel, i.e., T1-T0, L is the length of the rail temperature transition zone, and t is time.

[0064] A seamless turnout model is established. According to the principle of beam-rail interaction, a lightweight analysis model of the rail and the underlying foundation in the turnout area is established, and the model is solved by using the nonlinear finite element method. In the model, the rail is simulated by using a beam element, the fasteners in the turnout area are simulated by using a nonlinear element, the longitudinal and lateral resistance and the vertical stiffness of the fasteners are comprehensively considered, and the parameters are valued according to the current specifications or measured results; the longitudinal and lateral resistance of the track bed is simulated by using a nonlinear spring element, and the vertical stiffness of the track bed is simulated by using a vertical spring. The track slab and the base slab of the ballastless track are simulated by using a spatial plate element; the structure of the spacer and the spacer iron is simulated by using a nonlinear spring element, and the remaining parts are simulated by using a solid element.

[0065] S4, a mapping model is established to obtain the mapping relationship from the physical database to the virtual database. The data of the physical database is the physical data collected by the sensors in S2, and the data in the virtual database is the virtual data calculated in S3.

[0066] A neural network model of the physical data and the virtual data of the special section seamless line under the action of multiple working conditions is constructed and trained by using the physical database as the input and the virtual database as the output, the training data of the neural network model is the physical data set of the N sensor collection points in the physical database and the virtual data set of the M corresponding virtual data calculated by the finite element simulation model, and the mapping model is trained based on the physical data set and the virtual data set. As shown in the following formula: Figure 7 In a specific embodiment, N is 4 and M is 7.

[0067] A CNN-Bi-GRU-GWO neural network model based on adaptive hyperparameters is constructed to connect the physical database and the virtual database. The neural network model includes four layers: the first layer is an input layer, one-dimensional time series data is extracted by a CNN network; the second layer is a Bi-GRU time series network layer, the CNN feature extraction data of the previous layer is taken as the input to further extract the features in the time scale; the third layer is a neural network hyperparameter optimization layer, which adaptively optimizes the hyperparameters of the CNN network and the Bi-GRU time series network; the fourth layer is an output layer, which is trained and optimized, and the mean square error is taken as the objective function of the model. The model is trained by minimizing the mean square error, and the expression of the mean square error function is as follows:

[0068]

[0069] wherein n is the total number of data points, P i is the ith predicted value, A i is the ith actual value.

[0070] S5, comparing the mapping data with the data not participating in the establishment of the mapping model, verifying the model.

[0071] S51, when constructing the mapping model, the test points in the physical database and the virtual database that have common data types are excluded, and a part of the data is proposed to not participate in the construction of the mapping model, and is only used for verifying the mapping model. As shown in the figure, in a specific embodiment, 20% of the data does not participate in the construction of the mapping model, and is only used for model verification. The physical database test point coordinates W1, W2, W3, and W5 are actually used for constructing the mapping model, the test point coordinates X1, X2, X3, and X5 of the virtual database actually used for constructing the mapping model, and W4 and X4 are only used for model verification. Figure 7

[0072] S52, the physical data of the excluded point in the physical database is calculated through the mapping model to obtain the virtual data corresponding to the excluded point. That is Figure 7 the temperature, displacement data of the physical data coordinate points W1, W2, W3, and W5 are calculated through the mapping model to obtain the temperature, displacement, and rail longitudinal additional force of the virtual coordinate points X1, X2, X3, and X5.

[0073] S53, comparing the calculated virtual data with the physical data of the point. That is Figure 7 the temperature and displacement data of the virtual coordinate point X4 are calculated through the mapping model by using the temperature and displacement data of the physical data coordinate points W1, W2, W3, and W5, and are compared with the measured temperature and displacement data of the physical data coordinate point W4.

[0074] S54, judging the virtual data output by the mapping model, if the error between the virtual data and the physical data corresponding to the point is within a controllable range (not more than 5 in this embodiment), the verification of the mapping model is completed, and if the error is outside the controllable range, the parameters of the mapping model are modified and optimized.

[0075] ​S6, through the Unity3D platform rendering, build a digital twin virtual body completely equal to the special section of the seamless line, and interact with the field seamless line through the digital twin virtual body. By controlling the digital twin virtual body to change the parameters including fastener resistance, ballast resistance, environmental temperature, locking rail temperature and train live load; through the mapping model to calculate the indexes of the seamless line including rail temperature, climbing amount, rail longitudinal additional force, and formulate a maintenance plan that meets the specification requirements. After the maintenance and repair of the seamless line, the data collected by the seamless line is fed back to interact with the digital twin virtual body of the seamless line for verification.

[0076] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application, and various changes and improvements can be made without departing from the spirit and scope of the present application. These changes and improvements all fall within the scope of the claimed application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the longitudinal effects of a seamless railway line based on digital twins, characterized in that: The following steps are involved: S1. Determine the special section type of the seamless line on site; S2, using temperature sensors and displacement sensors to collect feature point data; S3. Establish a finite element simulation model to calculate the temperature, displacement and additional longitudinal force of the rail in special sections; S4. Establish a mapping model to obtain a mapping relationship between the physical database and the virtual database; The data in the physical database are physical data collected by the sensor in S2, and the data in the virtual database are virtual data calculated by S3; The mapping model establishment comprises the following steps: Using a physical database as input and a virtual database as output, a neural network model of physical and virtual data of a special section of seamless railway under multiple operating conditions is constructed and trained. The training data of the neural network model are physical data sets of N sensor acquisition points in the physical database and M corresponding virtual data sets in the virtual database calculated by the finite element simulation model. The mapping model is trained based on the physical and virtual data sets. The neural network model includes four layers of networks: The first layer of the network is the input layer, which extracts features from the one-dimensional time series data through the CNN network; The second layer is the Bi-GRU temporal network layer, which takes the feature extraction data of the previous layer of CNN as input and further extracts features on the time scale; The third network layer is the neural network hyperparameter optimization layer, which adaptively optimizes the hyperparameters of the CNN network and Bi-GRU timing network; The fourth network layer is the output layer, which is constructed for training and optimization, and the mean square error is used as the objective function of the model. The model is trained by minimizing the mean square error. S5. Compare the mapped data with the data that did not participate in the establishment of the mapping model to verify the model; S6. Through 3D rendering, a digital twin virtual body that is exactly the same as the special section of the seamless line is constructed, and it interacts with the on-site seamless line.

2. The method for monitoring the longitudinal effect of a seamless railway line based on digital twins according to claim 1 is characterized in that: The special sections of the seamless line in S1 include high-pier long-span bridges, bridge-tunnel transition sections and seamless turnouts.

3. The method for monitoring the longitudinal effect of a seamless railway line based on digital twins according to claim 2 is characterized in that: The feature point data in S2 includes: Displacement and temperature of rails at the ends of high-pier and long-span bridges, and displacement and temperature of beam joints; Rail displacement and rail temperature at the end of the simply supported beam at the tunnel entrance of the first span of the bridge-tunnel transition section, and rail displacement and rail temperature at each section of the temperature transition zone; The displacement changes of the tip of the seamless turnout point rail relative to the stock rail, the displacement changes of the tip of the heart rail relative to the wing rail, the displacement before the switch and at the limiter, and the stock rail displacement near the guard rail.

4. The method for monitoring the longitudinal effect of a seamless railway line based on digital twins according to claim 3 is characterized in that: The S3 includes the following steps: A CWR model for high-pier, long-span bridges was established. Based on the principle of beam-rail-bridge interaction, a lightweight calculation model for CWR, one bridge and one pier, was developed to calculate the longitudinal forces on high-pier, long-span bridges. The model was solved using the nonlinear finite element method to calculate the mechanical properties of CWR for high-pier, long-span bridges under different working conditions. A seamless track model for the bridge-tunnel transition section was established. Based on the principle of beam-track interaction, a lightweight analysis model of the track-substructure in the bridge-tunnel area was established. The model was solved using the nonlinear finite element method to calculate the mechanical characteristics of the seamless track in the bridge-tunnel transition section under different working conditions. A seamless turnout model is established. Based on the principle of beam-rail interaction, a lightweight analysis model of the track-substructure in the turnout area is established, and the model is solved using the nonlinear finite element method.

5. The method for monitoring the longitudinal effect of a seamless railway based on digital twins according to claim 4 is characterized in that: The S5 comprises the following steps: S51. When constructing the mapping model, remove the test points that are common to the physical database and the virtual database, and a portion of the data with a common data type. This portion is not used in the construction of the mapping model and is only used for verification of the mapping model. S52, calculating virtual data corresponding to the removed points by using a mapping model based on the physical data of the removed points in the physical database; S53, comparing the calculated virtual data with the physical data of the point; S54, judging the virtual data output by the mapping model. If the error between the virtual data and the physical data corresponding to the point is within a controllable range, the mapping model is verified. If the error is outside the controllable range, the parameters of the mapping model are corrected and optimized.

6. The method for monitoring the longitudinal effect of a seamless railway line based on digital twins according to claim 5 is characterized in that: The S6 comprises the following steps: By controlling the digital twin virtual body, parameters including fastener resistance, roadbed resistance, ambient temperature, locked rail temperature and train live load can be changed; Through the mapping model, the seamless railway is checked and calculated, including rail temperature, creep, and rail longitudinal additional force indicators, to formulate a maintenance plan that meets regulatory requirements; After maintenance and repair of the seamless line, data feedback collected from the seamless line is interactively verified with the seamless line digital twin virtual body.

Images