CN118037011A - Whole-car falling system of railway carriage - Google Patents
Whole-car falling system of railway carriage Download PDFInfo
- Publication number
- CN118037011A CN118037011A CN202410435115.7A CN202410435115A CN118037011A CN 118037011 A CN118037011 A CN 118037011A CN 202410435115 A CN202410435115 A CN 202410435115A CN 118037011 A CN118037011 A CN 118037011A
- Authority
- CN
- China
- Prior art keywords
- parameters
- module
- railway carriage
- whole
- falling
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 claims abstract description 46
- 239000000463 material Substances 0.000 claims abstract description 38
- 238000012417 linear regression Methods 0.000 claims abstract description 24
- 238000000034 method Methods 0.000 claims description 42
- 238000012549 training Methods 0.000 claims description 37
- 238000006073 displacement reaction Methods 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 16
- 230000008439 repair process Effects 0.000 claims description 14
- 238000013016 damping Methods 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 11
- 238000013507 mapping Methods 0.000 claims description 10
- 230000004044 response Effects 0.000 claims description 9
- 230000010354 integration Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000009826 distribution Methods 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000033001 locomotion Effects 0.000 description 17
- 230000008569 process Effects 0.000 description 13
- 238000004364 calculation method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000005096 rolling process Methods 0.000 description 6
- 239000000725 suspension Substances 0.000 description 6
- 230000007774 longterm Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000005299 abrasion Methods 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004886 head movement Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Landscapes
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention discloses a whole railway carriage falling system, which comprises: the auxiliary material parameter acquisition module is configured to acquire auxiliary material parameters of the whole railway passenger car to be dropped into; the terminal server is respectively connected with the cushioning guide module and the attached material parameter acquisition module, and a statically indeterminate balance module, a neural network module and a linear regression module are arranged in the terminal server; the cushioning guide module is arranged beside each cushioning station of the railway carriage to be formed and is configured to be connected with a terminal server, and parameters to be padded are obtained and displayed; the driving machine is configured to lift the railway carriage body to be dropped into, and after an operator finishes the cushioning according to the parameters to be cushioning displayed by the cushioning guiding module, the railway carriage body to be dropped into is lowered to finish the drop. The invention can accurately calculate the required padding parameters, greatly improve the precision of falling work and reduce the repeated times of falling and lifting. The high-efficiency flow obviously shortens the falling time and improves the efficiency of the whole falling work.
Description
Technical Field
The invention relates to the technical field of railway carriages, in particular to a whole railway carriage falling system.
Background
The passenger train plays a vital role in ensuring the operation performance, the safety and the reliability as core technical equipment for passenger transportation. To maintain its optimal performance, a railway carriage must undergo periodic maintenance and repair operations. In this process, ensuring the accuracy of the passenger car body is critical, as the balance of the car body directly affects the safety of the car and the riding comfort. According to the regulations for repairing railway passenger train section issued by China railway head office, strict regulations are provided for the heights of primary springs and secondary springs, car couplers and the like in the whole process of train falling. Because of numerous factors such as bogie manufacturers and different production batches, and performance attenuation of parts (such as springs and wheels) generated by material fatigue, stress relaxation or abrasion in the long-term service process of the passenger car, the quality distribution, spring height, coupler height, side bearing clearance and the like of the passenger car body are varied, so that the efficiency and quality of the process of repairing and landing the passenger car are affected.
In the conventional whole railway carriage falling process, after the carriage is fallen, a plurality of parts are required to be manually measured, so that the thickness of the gasket is adjusted, and the process often needs repeated lifting and carriage falling for a plurality of times (usually 3 to 5 times) to reach the standard requirement. In view of the size and weight of the railway carriage, repeated lifting and falling operations are time-consuming and labor-consuming, production efficiency is greatly reduced, and potential safety hazards of operation are increased.
In order to solve the problems, a numerical calculation method based on a statically indeterminate balance model of the whole bus is provided in a publication of key technical research on one-time falling of a passenger train section repair body, and aims to determine the thickness and the number of padding required in the process of falling in advance so as to realize one-time falling of the passenger train body. The technical scheme provides a solution to the hyperstatic problem by constructing and solving a kinetic equation. However, the method has a certain limitation in practical application, especially when the method is applied to a bus body and a framework with fatigue and micro deformation after long-term operation, because the model assumes the physical state of the bus body and the framework as an ideal state, larger deviation can occur when numerical calculation of falling parameters is performed. While such bias can be reduced by building a more accurate equation model, this refinement process can be quite time consuming and require significant computational resources.
Disclosure of Invention
In order to solve the defects in the prior art, the invention aims to provide a whole railway passenger train falling system which can intelligently collect the existing material parameters related to the falling of the railway passenger train and calculate the cushioning thickness and the number of each suspension element according to the collected parameters so as to guide the on-site completion of one-time falling operation of the railway passenger train body.
In order to achieve the above object, the present invention provides a technical solution comprising:
whole car system of falling into of railway carriage includes:
the auxiliary material parameter acquisition module is configured to acquire auxiliary material parameters of the whole railway passenger car to be dropped into; the attached material parameters comprise: the mass of each part of the vehicle body, the vehicle body distribution load, the spring damping parameter and the wheel parameter;
The terminal server is respectively connected with the padding guide module and the attached material parameter acquisition module, and a statically indeterminate balance module, a neural network module and a linear regression module are arranged in the terminal server;
The statically indeterminate balance model to be fallen into the whole railway carriage is built in, and displacement response of the statically indeterminate balance model is obtained based on an explicit dynamics integration method and the attached material parameter solution; calculating a first theoretical cushioning parameter according to the displacement response and the whole railway carriage falling requirement, and transmitting the first theoretical cushioning parameter to the linear regression module;
The neural network module is internally provided with a trained full-connection feedforward neural network, and the parameters of the attached materials are input into the full-connection feedforward neural network to obtain second theoretical padding parameters and transmitted to the linear regression module;
the linear regression module is internally provided with a linear regression model According to the input first theoretical padding parameter/>And the second theoretical padding parameter/>The combined output is the parameter theta to be padded, wherein,And/>Weights of the first theoretical padding parameter and the second theoretical padding parameter respectively and/>B is a bias term;
the cushioning guide module is arranged beside each cushioning station of the railway carriage to be formed and is configured to be connected with the terminal server, and the parameters to be cushioned are obtained and displayed;
and the driving machine is configured to lift the passenger train body to be dropped into the passenger train, and after the operator finishes the cushioning according to the parameters to be cushioning displayed by the cushioning guiding module, the passenger train body to be dropped into the passenger train is dropped into the passenger train body.
In some preferred embodiments, the attached material parameter obtaining module includes:
The load detection unit is arranged at the joint of the driving machine and the passenger train body to be dropped into the passenger train and is used for acquiring the mass and the distributed load of each part of the passenger train body to be dropped into the passenger train;
and the section repair terminal is arranged in the section repair spring test workshop and the wheel turning workshop and is configured to acquire spring damping parameters and wheel parameters of the railway carriage to be formed.
In some preferred embodiments, the method for solving the statically indeterminate balance model to be dropped into the whole railway carriage based on the explicit dynamics integration method and the attached material parameters by the statically indeterminate balance module comprises the following steps:
constructing a kinetic equation of the whole railway carriage to be fallen into:
; wherein/> 、/>、/>Respectively a mass matrix, a spring damping matrix and a rigidity matrix of each part of the whole train of the railway train to be dropped、/>、Acceleration, velocity and displacement vectors, respectively,/>Distributing a load vector for the vehicle body;
Velocity based on linear integration and integral control parameters alpha and beta, respectively, is built Displacement/>The differential expression of (2) is as follows:
；
；
substituting the differential expression into a dynamics equation of the whole railway carriage to be dropped to obtain a displacement response of the statically indeterminate balance model, namely that the whole railway carriage to be dropped is in Speed of time/>Displacement of。
In some preferred embodiments, the neural network module comprises a railway carriage history whole car fall-into debugging database, a training module and a fully-connected feedforward neural network which are connected in sequence;
The railway carriage history whole car falling debugging database stores the railway carriage history whole car falling debugging data of a plurality of types; the history falling debugging data comprise attached material parameters, the number of falling times and actual padding parameters of each falling time when the history falling of the whole railway carriage is debugged;
the training module calls the historical whole railway passenger car falling debugging data of the same type as the railway passenger car to be fallen from the historical whole railway passenger car falling debugging database, and the historical whole railway passenger car falling debugging data is divided into a training set and a testing set to finish training of the fully-connected feedforward neural network.
In some preferred embodiments, the neural network module further comprises a preprocessing unit connected to the training module and the fully-connected feedforward neural network, respectively; the preprocessing unit is configured to:
The method comprises the steps that cleaning histories fall into debugging data, actual padding parameters which are obtained by last trial fall are reserved after each fall into debugging, and the cleaned histories fall into debugging data to be associated with corresponding attached material parameters and serve as a first sample;
Carrying out preset parameter disturbance on the attached material parameters of the samples with the same actual padding parameters so as to form a new second sample;
the first and second samples are combined as a sample set for training the fully connected feedforward neural network.
In some preferred embodiments, the linear regression modelThe training method of (1) comprises:
Initialization of ，/>；
Constructing a loss function by means of the mean square error: where N is the number of samples in the training sample set,/> For training the first theoretical padding parameters of the sample,/>Actual padding parameters for training samples;
And carrying out iterative optimization on the linear regression model by using a gradient descent method until the value of the loss function is not reduced any more or reaches the preset iteration times.
In some preferred embodiments, the padding parameters include: the thickness of the center plate pad, the thickness of the spring pad and the thickness of the side bearing pad.
In some preferred embodiments, the fully connected feedforward neural network further includes a mapping layer after the output layer;
The mapping layer is internally provided with a thickness list of the actual core disc pad thickness, the spring pad thickness and the side bearing pad thickness in the field, and the mapping layer is used for matching the core disc pad thickness, the spring pad thickness and the side bearing pad thickness with the smallest absolute value of the difference value with the predicted value in the thickness list according to the predicted value output by the output layer, and outputting a matching result as a second theoretical padding parameter.
Advantageous effects
1. Improving the falling precision: the invention can accurately calculate the required padding parameters. The accurate parameter calculation greatly improves the precision of falling work and ensures the running performance and riding comfort of the railway carriage.
2. Efficiency is improved: compared with the traditional repeated vehicle falling and lifting, the method can calculate accurate padding parameters at one time, and reduces repeated vehicle falling and lifting times. The high-efficiency flow obviously shortens the falling time and improves the efficiency of the whole falling work.
3. Cost saving: the invention is helpful to reduce the manpower cost and the time cost by reducing the manual measurement and adjustment times in the falling process. At the same time, the repeated use and abrasion of the equipment are reduced, and the maintenance cost is reduced.
4. The operation is simplified: the invention simplifies the falling operation flow through automatic data acquisition and processing. The operator only needs to operate according to the padding parameters provided by the system, so that the operation complexity and the dependence on professional skills are reduced.
5. Safety enhancement: the invention reduces repeated lifting and falling operations and reduces the safety risk in the falling process. Meanwhile, through accurate calculation and guidance, potential safety hazards possibly caused by improper falling are reduced.
6. The adaptability is strong: the invention can automatically adjust and optimize the falling parameters according to the railway carriages with different models and actual use conditions, and has strong adaptability and flexibility.
Drawings
FIG. 1 is a schematic diagram of a complete railway carriage system in accordance with a preferred embodiment of the present invention;
FIG. 2 is a schematic front view of a statically indeterminate balance model with built-in statically indeterminate balance model in a preferred embodiment of the invention;
Fig. 3 is a schematic rear view of a statically indeterminate balance model with a statically indeterminate balance module built in according to a preferred embodiment of the present invention.
Detailed Description
Examples
The present invention will be further described with reference to the accompanying drawings, in order to make the objects, technical solutions and advantages of the present invention more apparent. In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "top," "bottom," "inner," "outer," and the like indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the present invention and to simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention.
As shown in fig. 1, this embodiment provides a whole-train landing system of a railway carriage, including:
The auxiliary material parameter acquisition module is configured to acquire auxiliary material parameters of the whole railway passenger car to be dropped into; the attached material parameters comprise: the mass of each part of the vehicle body, the distribution load of the vehicle body, the spring damping parameters and the wheel parameters. In the conventional passenger car falling debugging operation, the influence of material parameters on the falling quality is not considered, which is a main reason for failing to pass once. However, the railway passenger car is designed with more materials and is subjected to operation management in different workshops, so that the difficulty is brought to parameter collection. The invention develops a multi-station collaborative operation system for vehicle falling operation, which mainly relates to a vehicle body distribution load parameter, a primary/secondary spring height parameter, a primary/secondary spring compression stiffness parameter and a wheel diameter parameter, wherein the designed workshops are a vehicle body decomposition workshop, a spring test workshop and a wheel turning repair workshop respectively. In some preferred embodiments, a load detection unit is arranged at the joint of a car lifting machine used for existing car falling and a car body of a railway carriage to be fallen into, and is used for acquiring the mass and the distributed load of each part of the car body of the railway carriage to be fallen into, specifically, a load cell is additionally arranged, and the synchronous lifting of four car lifting machines used for the same car body can be realized by matching with a synchronous lifting control system, so that the lifting error is controlled in the process, and the lifting height difference of the four car lifting machines is not more than 2mm after the car lifting is ensured. At the moment, the vehicle body distribution load obtained by the load detection unit is transmitted to the vehicle-falling station computer terminal. The parameters of the springs are already acquired in the test bed of the existing spring test plant, but the position of the springs and the associated vehicle are not correlated. In some preferred embodiments, the segment repair terminals may be configured in a segment repair spring test shop and a wheel turning repair shop, wherein the segment repair terminals configured in the segment repair spring test shop upload spring data acquired from the spring test machine, including spring damping parameters to be dropped into the railway carriage, to the terminal server in an assigned vehicle group. Similarly, the wheel parameter information, including the wheel diameter, is obtained from the wheel-turning machine computer by a segment-turning terminal disposed in the segment-turning wheel-turning shop and uploaded to the terminal server. The multi-station, multi-period, multi-workshop, multi-dimensional, multi-layer and highly staggered information matrix is realized.
And the terminal server is respectively connected with the padding guide module and the attached material parameter acquisition module, and a statically indeterminate balance module, a neural network module and a linear regression module are arranged in the terminal server. It should be understood that, on the one hand, the terminal server is equipped with an operation device and a storage device, which may be locally disposed or may be disposed at the cloud, and the specific manner of disposing the terminal server may be specifically disposed by those skilled in the art according to the needs and the actual situations in the field, which is not specifically required by the present invention. On the other hand, the connection modes between the modules or units not explicitly described in the present invention are all signal connections for transmitting data, and the specific implementation modes thereof may be any one of wired signal connections or wireless signal connections, which are not limited in any way.
The statically indeterminate balance module is internally provided with a statically indeterminate balance model to be fallen into the whole railway carriage, and obtains displacement response of the statically indeterminate balance model based on an explicit dynamics integration method and the auxiliary material parameter solution; and calculating a first theoretical cushioning parameter according to the displacement response and the whole railway carriage falling requirement, and transmitting the first theoretical cushioning parameter to the linear regression module. The schematic diagrams of the statically indeterminate balance model to be dropped into the whole railway carriage are shown in fig. 2-3. It should be noted that the padding parameters in the present invention actually include: the thickness of the center plate pad, the thickness of the spring pad and the thickness of the side bearing pad.
A vehicle with a spring suspension is a multiple degree of freedom vibration system. The vehicle is generally composed of a vehicle body, a secondary vibration damping spring and damper, a bogie frame, a primary vibration damping spring and damper, and an axle box. In the prior art, when building mathematical models for dynamic characteristics of a vehicle or an entire train, components of a vehicle system such as a vehicle body, a bogie frame, an axle box, and the like are considered as rigid bodies. Each rigid body has 6 degrees of freedom, three displacements and three rotations, which are generally expressed in the vehicle coordinate system as head-shaking, roll and click motions. And constructing a statically indeterminate balance model of the whole vehicle aiming at the current vehicle, and carrying out fine balance solving on the model. In some preferred embodiments, because of the complex structure of the vehicle, the hyperstatic equation has multiple solutions and does not meet the requirements of vehicle falling, and the invention provides a method for solving the hyperstatic problem by adopting a dynamic equation. Considering the vehicle system as six degrees of freedom of the vehicle body floating and sinking motion, the vehicle body nodding motion, the vehicle body rolling motion, the framework floating and sinking motion, the framework rolling motion and the framework nodding motion, the motion equations of the six degrees of freedom are respectively as follows:
(1) Equation of motion of vehicle body
1. Floating and sinking movement
；
Wherein,The weight of the vehicle body; /(I)Is a secondary suspension damping; /(I)Is a diagonal matrix,/>The column vector is:
；
In the method, in the process of the invention, The parallel stiffness of the swing bolster springs is 1 and 3; /(I)The parallel stiffness of the 2-bit swing bolster spring and the 4-bit swing bolster spring is the same; the parallel stiffness of the 5-bit swing bolster spring and the 7-bit swing bolster spring is the same; /(I) The parallel stiffness of the 6-bit swing bolster spring and the 8-bit swing bolster spring is the same; /(I)Two are half of the lateral distance of the suspension; /(I)Is half the distance of the vehicle; /(I)、/>、/>Respectively performing sinking and floating, side rolling and nodding movement on the vehicle body; /(I)、The 1-position bogie frame and the 2-position bogie frame respectively do sinking and floating movements; /(I)、/>The side rolling motions of the bogie frames are respectively 1 and 2.
2. Side rolling movement
；
Wherein,Is a diagonal matrix,/>The column vector is:
；
3. Spot head movement
；
Wherein,Is a diagonal matrix,/>The column vector is:
；
(2) Frame equation of motion
1. Floating and sinking movement
；
Wherein,The weight of the vehicle body; /(I)Is a secondary suspension damping; /(I)Is a diagonal matrix,/>The column vector is:
；
Wherein, Parallel stiffness of the springs of the 1-position axle box and the 3-position axle box; /(I)Parallel stiffness of the 2-bit axle box spring and the 4-bit axle box spring; Parallel stiffness of 5-bit axle box spring and 7-bit axle box spring; /(I) Parallel stiffness of 6-bit axle box spring and 8-bit axle box spring; /(I)Is one half of the lateral distance of the suspension; /(I)Is half the distance between the bogie wheel pairs.
(2) Side rolling movement
；
Wherein,Is a diagonal matrix,/>The column vector is:
；
(3) Spot head movement
；
Wherein,Is a diagonal matrix,/>The column vector is:
。
Constructing a kinetic equation of the whole railway passenger car to be fallen, introducing the formula into the kinetic equation of the whole railway passenger car to be fallen, and solving the displacement response of the kinetic equation by using a numerical method, wherein the method specifically comprises the following steps:
; wherein/> 、/>、/>Respectively a mass matrix, a spring damping matrix and a rigidity matrix of each part of the whole train of the railway train to be dropped、/>、/>Acceleration, velocity and displacement vectors, respectively,/>Load vectors are distributed for the vehicle body.
Velocity based on linear integration and integral control parameters alpha and beta, respectively, is builtDisplacement/>The differential expression of (2) is as follows:
；
；
substituting the differential expression into a kinetic equation of the whole railway carriage to be dropped to obtain the following steps:
；
Wherein the effective mass Effective force vector/>The method comprises the following steps of:
；
；
solving the equation to obtain the whole passenger train in the railway Speed of time/>Displacement/>。
It should be understood that the first theoretical padding parameter calculated by the statically indeterminate balance module has higher accuracy, but when the built-in statically indeterminate balance model is used for processing the bus body and the frame with fatigue and micro deformation after long-term operation, the physical states of the bus body and the frame are assumed to be ideal by the model, so that larger deviation may occur when numerical calculation of falling parameters is performed. The present invention therefore also provides a neural network module to eliminate such bias, through data analysis and machine learning techniques, to be able to accommodate these changes, providing a customized fall-into solution for different states of railway carriages.
The neural network module is internally provided with a trained full-connection feedforward neural network, the parameters of the attached materials are input into the full-connection feedforward neural network, and second theoretical padding parameters are obtained and transmitted to the linear regression module. The fully connected feedforward neural network (Fully Connected Feedforward Neural Network), also known as a multi-layer perceptron (MLP), is made up of multiple fully connected layers, each layer being fully connected to the previous layer, with no feedback connection (i.e., the network does not form a loop). The method has strong nonlinear relation capturing capability when processing non-image and non-time sequence (namely table or structured data) data, and is particularly suitable for the technical problems related to the application. Those skilled in the art can know that the trained fully-connected feedforward neural network is completed by a specific method according to actual needs by those skilled in the art and combined with the data set of the application scene. The specific training method can be to complete training before the training is built in, so that a model with all parameters being debugged is built in, or to build the training module and the neural network module in the neural network module at the same time, so that the model is continuously improved and finely adjusted in continuous use to enhance the performance of the model. In some preferred embodiments, a preferred implementation of a neural network module is provided, comprising:
the neural network module comprises a railway carriage history whole car fall into debugging database, a training module and a full-connection feedforward neural network which are connected in sequence;
The railway carriage history whole car falling debugging database stores the railway carriage history whole car falling debugging data of a plurality of types; the history falling debugging data comprise attached material parameters, the number of falling times and actual padding parameters of each falling time when the history falling of the whole railway carriage is debugged;
the training module calls the historical whole railway passenger car falling debugging data of the same type as the railway passenger car to be fallen from the historical whole railway passenger car falling debugging database, and the historical whole railway passenger car falling debugging data is divided into a training set and a testing set to finish training of the fully-connected feedforward neural network.
It should be noted that, since the training of the fully-connected feedforward neural network is prone to over-fitting in the case of a smaller data volume, for this purpose, in some preferred embodiments, the neural network module is further configured with a preprocessing unit to provide a preferred training method so as to avoid the problem of over-fitting, which specifically includes:
the neural network module further comprises a preprocessing unit which is respectively connected with the training module and the fully-connected feedforward neural network; the preprocessing unit is configured to:
The method comprises the steps that cleaning histories fall into debugging data, actual padding parameters which are obtained by last trial fall are reserved after each fall into debugging, and the cleaned histories fall into debugging data to be associated with corresponding attached material parameters and serve as a first sample;
Carrying out preset parameter disturbance on the attached material parameters of the samples with the same actual padding parameters so as to form a new second sample; the parameter perturbation is to create new samples by small changes on the basis of the existing data set, so that the data volume is increased, the diversity of the data is also increased, and the model generalization capability is greatly improved. In operation, it is also necessary to determine a reasonable disturbance range based on an understanding of the influence of each parameter. For parameters that affect large amounts, the perturbation range may need to be relatively conservative to prevent impractical data points from being generated. While for parameters that affect less, a larger disturbance range may be considered. The specific arrangement thereof can be set by those skilled in the art according to actual needs of the field, and the present invention is not further limited.
The first and second samples are combined as a sample set for training the fully connected feedforward neural network.
It should be noted that, since the thickness of the core pad and the spring pad installed on the segment repair site are limited by specifications, for example, the spring pad has a specification of only 5mm, the installation requirement of the core pad is within 50mm, and the like, and the predicted result directly output by the fully connected feedforward neural network may not meet the specification requirement (for example, a numerical value with a decimal number or a numerical value exceeding the specification limit, and the like). Thus, in some preferred embodiments it is contemplated to solve this problem by mapping between actual specification requirements and predicted output, including in particular:
The fully-connected feedforward neural network further comprises a mapping layer after the output layer;
The mapping layer is internally provided with a thickness list of the actual core disc pad thickness, the spring pad thickness and the side bearing pad thickness in the field, and the mapping layer is used for matching the core disc pad thickness, the spring pad thickness and the side bearing pad thickness with the smallest absolute value of the difference value with the predicted value in the thickness list according to the predicted value output by the output layer, and outputting a matching result as a second theoretical padding parameter.
The linear regression module is internally provided with a linear regression modelAccording to the input first theoretical padding parameter/>And the second theoretical padding parameter/>The combined output is the parameter theta to be padded, wherein,And/>Weights of the first theoretical padding parameter and the second theoretical padding parameter respectively and/>B is the bias term.
As described above, the first theoretical padding parameter obtained by the statically indeterminate balancing module through calculation of the theoretical equation does not consider errors generated by fatigue of the structural member after long-time operation, while the second theoretical padding parameter obtained by the neural network module through the fully-connected feedforward neural network is obtained through a black box state, which does not consider the actual theoretical basis of the acquisition process, so that it is inevitable that there is a certain error between the two. In order to eliminate the error, the invention introduces a linear regression module, fully considers the respective advantages and limitations of two parameters based on a model fusion method, and combines the two parameters in a weighted mode. Further, when weighting is used, the assignment of weights is typically dependent on the credibility, historical performance, expert experience, or statistical evidence from data analysis of the individual parameters. There is no fixed method available for all cases, so a person skilled in the art can decide on the specific setting of weights according to the specific case and the available data. In some preferred embodiments, a training method of a linear regression model is provided to obtain more objective and accurate weight configuration, which specifically includes:
Initialization of ，/>; In this case, for simplicity, the/>, can beExpressed as/>Thereby only optimizing/>And b.
Constructing a loss function using a Mean Square Error (MSE): where N is the number of samples in the training sample set,/> For training the first theoretical padding parameters of the sample,/>Is the actual padding parameter of the training sample.
And carrying out iterative optimization on the linear regression model by using a gradient descent method until the value of the loss function is not reduced any more or reaches the preset iteration times. The nature of the gradient descent method is an optimizer of the linear regression model, and the purpose of the gradient descent method is to minimize a loss function, and specifically, the gradient descent method comprises the following specific steps:
For a pair of And b calculating the gradient of the loss function L:
；
these gradients are then used to update And b:
；
where α is the learning rate, and this parameter is set by those skilled in the art according to the actual needs, and too large may result in too fast convergence or divergence, and too small may result in too slow training.
The cushioning guide module is arranged beside each cushioning station of the railway carriage to be formed and is configured to be connected with the terminal server, and the parameters to be cushioned are obtained and displayed. Specifically, the cushioning guidance module includes a display device for displaying the parameter to be cushioning to an operator, or may be a printing device for printing the parameter to be cushioning, which is not limited in the present invention. As previously described, the shimming parameters include a center-plate-pad thickness, a spring-pad thickness, and a side-bearing-pad thickness, and thus the shimming guidance module may be disposed beside the shimming stations of the center-plate pad, spring-pad, and side-bearing pad, respectively.
And the driving machine is configured to lift the passenger train body to be dropped into the passenger train, and after the operator finishes the cushioning according to the parameters to be cushioning displayed by the cushioning guiding module, the passenger train body to be dropped into the passenger train is dropped into the passenger train body. It should be understood that, since the drop method is not the focus of the present invention, and the standard drop method is also available for reference in the passenger train section repair process, the present invention is not specifically limited to the specific manner in which the driver completes the drop operation.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (8)
1. The whole car system of falling into of railway passenger train, its characterized in that includes:
the auxiliary material parameter acquisition module is configured to acquire auxiliary material parameters of the whole railway passenger car to be dropped into; the attached material parameters comprise: the mass of each part of the vehicle body, the vehicle body distribution load, the spring damping parameter and the wheel parameter;
The terminal server is respectively connected with the padding guide module and the attached material parameter acquisition module, and a statically indeterminate balance module, a neural network module and a linear regression module are arranged in the terminal server;
The statically indeterminate balance model to be fallen into the whole railway carriage is built in, and displacement response of the statically indeterminate balance model is obtained based on an explicit dynamics integration method and the attached material parameter solution; calculating a first theoretical cushioning parameter according to the displacement response and the whole railway carriage falling requirement, and transmitting the first theoretical cushioning parameter to the linear regression module;
The neural network module is internally provided with a trained full-connection feedforward neural network, and the parameters of the attached materials are input into the full-connection feedforward neural network to obtain second theoretical padding parameters and transmitted to the linear regression module;
the linear regression module is internally provided with a linear regression model According to the input first theoretical padding parameter/>And the second theoretical padding parameter/>The combined output is the parameter theta of the pad to be added, wherein/>And/>Weights of the first theoretical padding parameter and the second theoretical padding parameter respectively and/>B is a bias term;
the cushioning guide module is arranged beside each cushioning station of the railway carriage to be formed and is configured to be connected with the terminal server, and the parameters to be cushioned are obtained and displayed;
and the driving machine is configured to lift the passenger train body to be dropped into the passenger train, and after the operator finishes the cushioning according to the parameters to be cushioning displayed by the cushioning guiding module, the passenger train body to be dropped into the passenger train is dropped into the passenger train body.
2. The whole railway carriage drop system as claimed in claim 1, wherein the attached material parameter obtaining module comprises:
The load detection unit is arranged at the joint of the driving machine and the passenger train body to be dropped into the passenger train and is used for acquiring the mass and the distributed load of each part of the passenger train body to be dropped into the passenger train;
and the section repair terminal is arranged in the section repair spring test workshop and the wheel turning workshop and is configured to acquire spring damping parameters and wheel parameters of the railway carriage to be formed.
3. The whole railway carriage falling system according to claim 2, wherein the method for solving the statically indeterminate balance model to be fallen into the whole railway carriage based on the explicit dynamics integration method and the attached material parameters comprises the following steps:
constructing a kinetic equation of the whole railway carriage to be fallen into:
; wherein/> 、/>、/>Respectively a mass matrix, a spring damping matrix and a rigidity matrix of each part of the whole train of the railway train to be dropped、/>、/>Acceleration, velocity and displacement vectors, respectively,/>Distributing a load vector for the vehicle body;
Velocity based on linear integration and integral control parameters alpha and beta, respectively, is built Displacement/>The differential expression of (2) is as follows:
；
；
substituting the differential expression into a dynamics equation of the whole railway carriage to be dropped to obtain a displacement response of the statically indeterminate balance model, namely that the whole railway carriage to be dropped is in Speed of time/>Displacement/>。
4. The whole railway carriage drop system as in claim 1 or 2, wherein the neural network module comprises a whole railway carriage history drop debugging database, a training module and a fully connected feedforward neural network which are connected in sequence;
The railway carriage history whole car falling debugging database stores the railway carriage history whole car falling debugging data of a plurality of types; the history falling debugging data comprise attached material parameters, the number of falling times and actual padding parameters of each falling time when the history falling of the whole railway carriage is debugged;
the training module calls the historical whole railway passenger car falling debugging data of the same type as the railway passenger car to be fallen from the historical whole railway passenger car falling debugging database, and the historical whole railway passenger car falling debugging data is divided into a training set and a testing set to finish training of the fully-connected feedforward neural network.
5. The whole passenger train drop-in system according to claim 4, wherein the neural network module further comprises a preprocessing unit respectively connected with the training module and the fully-connected feedforward neural network; the preprocessing unit is configured to:
The method comprises the steps that cleaning histories fall into debugging data, actual padding parameters which are obtained by last trial fall are reserved after each fall into debugging, and the cleaned histories fall into debugging data to be associated with corresponding attached material parameters and serve as a first sample;
Carrying out preset parameter disturbance on the attached material parameters of the samples with the same actual padding parameters so as to form a new second sample;
the first and second samples are combined as a sample set for training the fully connected feedforward neural network.
6. The whole railway carriage fall system as in claim 1 or 2, wherein the linear regression modelThe training method of (1) comprises:
Initialization of ，/>；
Constructing a loss function by means of the mean square error: where N is the number of samples in the training sample set,/> For training the first theoretical padding parameters of the sample,/>Actual padding parameters for training samples;
And carrying out iterative optimization on the linear regression model by using a gradient descent method until the value of the loss function is not reduced any more or reaches the preset iteration times.
7. The whole railway carriage fall system as in claim 1 or 2, wherein the shimming parameters include: the thickness of the center plate pad, the thickness of the spring pad and the thickness of the side bearing pad.
8. The whole body drop system of the railway carriage according to claim 7, wherein the fully connected feedforward neural network further comprises a mapping layer after the output layer;
The mapping layer is internally provided with a thickness list of the actual core disc pad thickness, the spring pad thickness and the side bearing pad thickness in the field, and the mapping layer is used for matching the core disc pad thickness, the spring pad thickness and the side bearing pad thickness with the smallest absolute value of the difference value with the predicted value in the thickness list according to the predicted value output by the output layer, and outputting a matching result as a second theoretical padding parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410435115.7A CN118037011B (en) | 2024-04-11 | 2024-04-11 | Whole-car falling system of railway carriage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410435115.7A CN118037011B (en) | 2024-04-11 | 2024-04-11 | Whole-car falling system of railway carriage |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118037011A true CN118037011A (en) | 2024-05-14 |
| CN118037011B CN118037011B (en) | 2024-06-18 |
Family
ID=90989925
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410435115.7A Active CN118037011B (en) | 2024-04-11 | 2024-04-11 | Whole-car falling system of railway carriage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118037011B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6422061B1 (en) * | 1999-03-03 | 2002-07-23 | Cyrano Sciences, Inc. | Apparatus, systems and methods for detecting and transmitting sensory data over a computer network |
| CN1730333A (en) * | 2005-08-06 | 2006-02-08 | 太原铁路局科学技术研究所 | Railway coach suspension spring matching method |
| WO2018042131A1 (en) * | 2016-08-31 | 2018-03-08 | Université Pierre et Marie Curie | Compliant autonomous vehicle and method for controlling same |
| CN110232452A (en) * | 2019-06-12 | 2019-09-13 | 中国神华能源股份有限公司 | Repair method and system based on track car team state of the art |
| CN114282398A (en) * | 2021-06-02 | 2022-04-05 | 四川轻化工大学 | Bridge health monitoring system and method based on big data |
| CN114330122A (en) * | 2021-12-28 | 2022-04-12 | 东北电力大学 | Hydroelectric generating set large shaft axis adjusting method based on machine learning |
| CN115391922A (en) * | 2022-08-30 | 2022-11-25 | 中车唐山机车车辆有限公司 | Method for calculating axle box spring cushioning amount of motor train unit, server and storage medium |
| US20230129779A1 (en) * | 2019-12-17 | 2023-04-27 | Wincor Nixdorf International Gmbh | Method, control device and registration terminal for computer-supported detection of the empty state of a transport container |
| CN116366369A (en) * | 2023-05-15 | 2023-06-30 | 成都工业职业技术学院 | Data communication method, communication device and communication terminal in rail transit |
| US20230419249A1 (en) * | 2022-06-27 | 2023-12-28 | Freeport Minerals Corporation | System and method for determining a dump location for a vehicle |
-
2024
- 2024-04-11 CN CN202410435115.7A patent/CN118037011B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6422061B1 (en) * | 1999-03-03 | 2002-07-23 | Cyrano Sciences, Inc. | Apparatus, systems and methods for detecting and transmitting sensory data over a computer network |
| CN1730333A (en) * | 2005-08-06 | 2006-02-08 | 太原铁路局科学技术研究所 | Railway coach suspension spring matching method |
| WO2018042131A1 (en) * | 2016-08-31 | 2018-03-08 | Université Pierre et Marie Curie | Compliant autonomous vehicle and method for controlling same |
| CN110232452A (en) * | 2019-06-12 | 2019-09-13 | 中国神华能源股份有限公司 | Repair method and system based on track car team state of the art |
| US20230129779A1 (en) * | 2019-12-17 | 2023-04-27 | Wincor Nixdorf International Gmbh | Method, control device and registration terminal for computer-supported detection of the empty state of a transport container |
| CN114282398A (en) * | 2021-06-02 | 2022-04-05 | 四川轻化工大学 | Bridge health monitoring system and method based on big data |
| CN114330122A (en) * | 2021-12-28 | 2022-04-12 | 东北电力大学 | Hydroelectric generating set large shaft axis adjusting method based on machine learning |
| US20230419249A1 (en) * | 2022-06-27 | 2023-12-28 | Freeport Minerals Corporation | System and method for determining a dump location for a vehicle |
| CN115391922A (en) * | 2022-08-30 | 2022-11-25 | 中车唐山机车车辆有限公司 | Method for calculating axle box spring cushioning amount of motor train unit, server and storage medium |
| CN116366369A (en) * | 2023-05-15 | 2023-06-30 | 成都工业职业技术学院 | Data communication method, communication device and communication terminal in rail transit |
Non-Patent Citations (3)
| Title |
|---|
| 杨振祥;黄晓蓉;韩锟;黄明高;: "机车重心位置的检测及对调簧的影响", 铁道机车车辆, no. 01, 28 February 2007 (2007-02-28), pages 35 - 38 * |
| 赵悦: "客车段修车体一次落成关键技术研究", 西南交通大学硕士学位论文, 30 September 2021 (2021-09-30), pages 1 - 90 * |
| 韩锟;潘迪夫;: "机车二系弹簧载荷均匀性分配调整的混合建模方法", 中南大学学报(自然科学版), 26 January 2012 (2012-01-26), pages 379 - 383 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118037011B (en) | 2024-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101216376B (en) | Four-column type rail vehicle bogie rigidity detection system | |
| CN102650563B (en) | Ground testing system for on-track micro vibration of spacecraft | |
| CN109835375B (en) | High-speed railway train automatic driving system based on artificial intelligence technology | |
| CN107161360B (en) | Replaceable free across the scale verifying device of pedestal motion reappearance of space tasks | |
| CN118037011B (en) | Whole-car falling system of railway carriage | |
| CN115374536A (en) | Method for constructing four-degree-of-freedom numerical model of cab suspension system | |
| Dižo et al. | Use of multibody system dynamics as a tool for rail vehicle behaviour diagnostics | |
| CN201145641Y (en) | System for detecting four-pillar type rail vehicle steering frame rigidity | |
| CN111721556B (en) | Impact speed control method and device, storage medium and electronic equipment | |
| Lin et al. | Robust active vibration control for rail vehicle pantograph | |
| CN116049989A (en) | Bogie digital twin model construction method, system, electronic equipment and medium | |
| CN113449376B (en) | Method, system and equipment for selecting shock absorber of suspension equipment under train | |
| CN109388814B (en) | Method for calculating axle load of floating car type 5-module low-floor urban rail vehicle | |
| CN115391922A (en) | Method for calculating axle box spring cushioning amount of motor train unit, server and storage medium | |
| CN108225248A (en) | Railway freight car wheel is to axle journal automatic measurement machine | |
| Kalivoda et al. | Mechatronic bogie for roller rig tests | |
| CN111950150B (en) | Modeling and simulation method and system for twisted orbit test | |
| Lin et al. | A hybrid evolutionary approach for robust active suspension design of light rail vehicles | |
| CN111914342A (en) | Locomotive axle readjusting method | |
| Enblom | Two-level numerical optimization of ride comfort in railway vehicles | |
| CN114104631A (en) | Urban rail vehicle bogie maintenance system | |
| CN110567736B (en) | Bogie uniform load testing method and device | |
| CN110823608A (en) | Fatigue test system and method for railway freight car cross rod assembly | |
| CN117184445B (en) | Model parameter equivalent correction method for aircraft tire rigidity test | |
| Bosso et al. | Study and Simulation of a Hydropneumatic Suspension for Telescopic Handler Vehicles |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |