CN112506043A - Control method and control system for rail vehicle and vertical shock absorber - Google Patents
Control method and control system for rail vehicle and vertical shock absorber Download PDFInfo
- Publication number
- CN112506043A CN112506043A CN202011254231.7A CN202011254231A CN112506043A CN 112506043 A CN112506043 A CN 112506043A CN 202011254231 A CN202011254231 A CN 202011254231A CN 112506043 A CN112506043 A CN 112506043A
- Authority
- CN
- China
- Prior art keywords
- shock absorber
- module
- attitude information
- current
- vertical shock
- 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
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0205—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system
- G05B13/024—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61F—RAIL VEHICLE SUSPENSIONS, e.g. UNDERFRAMES, BOGIES OR ARRANGEMENTS OF WHEEL AXLES; RAIL VEHICLES FOR USE ON TRACKS OF DIFFERENT WIDTH; PREVENTING DERAILING OF RAIL VEHICLES; WHEEL GUARDS, OBSTRUCTION REMOVERS OR THE LIKE FOR RAIL VEHICLES
- B61F5/00—Constructional details of bogies; Connections between bogies and vehicle underframes; Arrangements or devices for adjusting or allowing self-adjustment of wheel axles or bogies when rounding curves
- B61F5/26—Mounting or securing axle-boxes in vehicle or bogie underframes
- B61F5/30—Axle-boxes mounted for movement under spring control in vehicle or bogie underframes
- B61F5/308—Axle-boxes mounted for movement under spring control in vehicle or bogie underframes incorporating damping devices
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Mechanical Engineering (AREA)
- Vehicle Body Suspensions (AREA)
Abstract
The invention discloses a control method and a control system for a rail vehicle and a vertical shock absorber, wherein the control method comprises the following steps: acquiring current attitude information of the rail vehicle; and determining the current optimal damping of the vertical shock absorber according to the current attitude information and preset ideal attitude information of the railway vehicle, and controlling the vertical shock absorber to generate the current optimal damping. The control method can enable the vertical shock absorber to change in a self-adaptive mode according to the state of the vehicle so as to adapt to the current vehicle state, and the dynamic performance of the rail vehicle is effectively improved.
Description
Technical Field
The invention relates to the technical field of railway vehicles, in particular to a railway vehicle and a control method and a control system of a vertical shock absorber.
Background
The vertical shock absorber is a key component for attenuating the vibration of the train body, and the parameter requirements on the shock absorber are different when the train runs in different states. The traditional vertical shock absorber is a passive shock absorber, the damping characteristic curve of the traditional vertical shock absorber is fixed, and the traditional vertical shock absorber cannot be adjusted according to the running state of a train.
In order to solve the above problems, active shock absorbers (including semi-active shock absorbers) such as magnetorheological shock absorbers are currently applied to rail vehicles to adjust damping forces of the active shock absorbers to adapt to different operating states of the vehicles.
However, the control logic of the existing control system for the active damper is relatively simple, the control algorithm cannot be intelligently optimized along with the change of the vehicle running state, and the improvement on the dynamic performance of the rail vehicle is not particularly obvious.
In view of this, how to improve the control logic of the vertical shock absorber of the railway vehicle so that the control logic can be adjusted in real time according to the current state of the vehicle to effectively improve the dynamic performance of the vehicle is a technical problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The invention aims to provide a control method and a control system for a vertical shock absorber of a railway vehicle, which can enable the vertical shock absorber to automatically adjust parameters in a controller and optimize the control effect through intelligent learning according to the state of the vehicle, thereby effectively improving the dynamic performance of the railway vehicle. It is another object of the present invention to provide a rail vehicle comprising the above control system.
In order to solve the technical problem, the invention provides a control method of a vertical shock absorber of a railway vehicle, wherein the vertical shock absorber is a semi-active shock absorber, two ends of the vertical shock absorber are respectively connected with a framework and an axle box of the railway vehicle, and the control method comprises the following steps:
acquiring current attitude information of the rail vehicle;
and determining the current optimal damping of the vertical shock absorber according to the current attitude information and preset ideal attitude information of the railway vehicle, and controlling the vertical shock absorber to generate the current optimal damping.
The method for controlling the vertical shock absorber of the railway vehicle comprises the following steps of: and determining the current optimal damping of the vertical shock absorber by adopting a depth determinacy strategic gradient algorithm.
According to the control method of the vertical shock absorber of the railway vehicle, the depth certainty strategic gradient algorithm comprises a strategic network module, a Q network module and a memory base module;
the method for determining the current optimal damping of the vertical shock absorber by adopting the depth certainty strategic gradient algorithm comprises the following steps:
determining a current error of the ideal attitude information and the current attitude information;
the strategy network module utilizes the Q network module to seek a calculation function of an optimal control instruction according to the current error, determines the current optimal damping corresponding to the current error according to the calculation function sought by the Q network, and sends a corresponding control instruction to the vertical shock absorber according to the current optimal damping;
the strategy network module and the Q network module carry out training and learning according to the data stored in the memory base module;
and the memory library module is used for storing the current error and the control instruction.
According to the control method of the vertical shock absorber of the railway vehicle, the memory base module further stores the error at the next moment, wherein the error at the next moment is the difference value between the ideal attitude information and the actual attitude information at the next moment;
the depth certainty strategic gradient algorithm further comprises a strategy evaluation module, wherein the strategy evaluation module is used for calculating rewards according to the ideal posture information and the actual posture information at the next moment, and storing the rewards to the memory base module so as to participate in training and learning of the strategy network module and the Q network module.
In the method for controlling a vertical vibration absorber of a railway vehicle, the reward calculation function in the strategy evaluation module is as follows:
rt＝(axt-ax't+1)2；
wherein, axtFor the ideal attitude information, ax't+1And the attitude information of the next moment.
According to the control method of the vertical shock absorber of the railway vehicle, the attitude information of the railway vehicle comprises the vibration parameters of the vehicle body and/or the vibration parameters of the framework.
The method for controlling the vertical vibration absorber of the railway vehicle comprises the step of controlling the vibration parameter of the vehicle body to comprise the vibration acceleration and/or the pitching vibration angular acceleration, and/or the vibration parameter of the framework to comprise the vibration acceleration and/or the pitching vibration angular acceleration.
According to the control method of the vertical shock absorber of the railway vehicle, when the attitude information of the railway vehicle is represented by a single parameter, the corresponding control damping quantity is determined according to the single parameter, and the control damping quantity is the current optimal damping;
when the attitude information of the rail vehicle is represented by more than two parameters, damping control quantities corresponding to the parameters are determined according to the parameters respectively, and then the current optimal damping is determined according to the damping control quantities.
The invention also provides a control system of a vertical shock absorber of a railway vehicle, wherein the vertical shock absorber is a semi-active shock absorber, two ends of the vertical shock absorber are respectively connected with a framework and an axle box of the railway vehicle, and the control system comprises:
the detection module is used for acquiring current attitude information of the rail vehicle;
and the control module is used for determining the current optimal damping of the vertical shock absorber according to the current attitude information acquired by the detection module and the pre-stored ideal attitude information of the rail vehicle, and sending a control command corresponding to the current optimal damping to the vertical shock absorber.
The control system of the vertical shock absorber of the railway vehicle comprises a driver and an algorithm unit, wherein the algorithm unit adopts a depth certainty strategic gradient algorithm and comprises a strategic network module, a Q network module and a memory base module;
the strategy network module is used for determining the current optimal damping according to the current attitude information, the ideal attitude information and a calculation function sought by the Q network module, and sending a control instruction corresponding to the current optimal damping to the driver;
the Q network module is used for seeking a calculation function of an optimal control instruction;
the strategy network module and the Q network module are also used for training and learning according to the data stored in the memory base module;
the memory library module is used for storing a current error and the control instruction, wherein the current error is a difference value between the ideal attitude information and the current attitude information.
In the control system of the vertical shock absorber of the railway vehicle, the memory bank module is further configured to store an error at a next moment, where the error at the next moment is a difference between the ideal attitude information and actual attitude information at the next moment;
the control module further comprises a strategy evaluation module, and the strategy evaluation module is used for calculating rewards according to the ideal posture information and the actual posture information at the next moment and storing the rewards to the memory base module so as to participate in training and learning of the strategy network module and the Q network module.
The control system of the vertical shock absorber of the railway vehicle comprises a control system, a control system and a control system, wherein the control system comprises a control system and a control system, and the control system comprises a control system and a control system, wherein the control system comprises a; the detection module comprises a vibration sensor and/or a gyroscope which are used for being installed on the vehicle body, and/or a vibration sensor and/or a gyroscope which are installed on the framework.
The invention also provides a railway vehicle which comprises the vertical shock absorber and a control system of any one of the vertical shock absorbers.
The control method and the system provided by the invention are used for controlling the vertical shock absorber of the rail vehicle, the vertical shock absorber is a semi-active shock absorber, particularly, ideal attitude information is taken as a control target, the current optimal damping of the vertical shock absorber is determined according to the current attitude information of the rail vehicle, so that the damping force of the vertical shock absorber acting on the train is adjusted, the damping of the vertical shock absorber is adjusted in real time according to the running state of the rail vehicle, and the vertical shock absorber can be adjusted in a self-adaptive mode according to the running condition of the vehicle, so that the train suspension system is always in a better matching requirement, and the dynamic performance of the rail vehicle is effectively improved.
The rail vehicle with the control system provided by the invention also has the same technical effects.
Drawings
FIG. 1 is a schematic illustration in partial schematic form of one embodiment of a control system for a vertical shock absorber of a railway vehicle according to the present invention;
FIG. 2 is a schematic diagram of a control system for a vertical shock absorber of a railway vehicle according to the present invention;
FIG. 3 is a control schematic of the control module shown in FIG. 2.
Description of reference numerals:
a vehicle body 11, a frame 12, a vertical shock absorber 13;
a control module 21, a detection module 22;
an algorithm unit 211, a policy network module 2111, a Q network module 2112, a policy evaluation module 2113, a repository module 2114, and a driver 212.
Detailed Description
In order that those skilled in the art will better understand the disclosure, the invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
For ease of understanding and clarity of description, the following description is provided in conjunction with a method and system for controlling a vertical shock absorber of a railway vehicle, and some of the advantages will not be repeated.
Referring to fig. 1 and 2, fig. 1 is a partial schematic structural view of an embodiment of a control system for a vertical shock absorber of a railway vehicle according to the present invention; FIG. 2 is a schematic diagram of a control system for a rail vehicle vertical shock absorber according to the present invention.
The railway vehicle comprises a vehicle body 11 and a frame 12 carrying the vehicle body 11, one frame 12 being provided at each end for a section of the railway vehicle, and a vertical damper 13 being provided generally between the frame 12 and an axle box mounted to the frame 12 for damping vibrations of the vehicle body 11.
Generally, at least one vertical damper 13 is disposed between the axle housing and the frame 12 at a location corresponding to one axle housing location, i.e., at a location corresponding to one wheel, although two may be used as desired.
In this embodiment, the vertical shock absorber 13 is a semi-active shock absorber, and the control method and control system are primarily used to control the damping of the semi-active shock absorber.
Fig. 1 shows a side view of a frame 12 in a position where two vertical dampers 13 are provided for each axle box position. It will be appreciated that the number of vertical dampers 13 at each pedestal housing location is set as desired in practice and is not limited to the embodiment shown in the drawings.
The focus here is on how to control the damping of the vertical shock absorber 13.
In this embodiment, the method of controlling the vertical damper 13 includes:
acquiring current attitude information of the rail vehicle;
and determining the current optimal damping of the vertical shock absorber according to the current attitude information and the preset ideal attitude information of the railway vehicle, and controlling the vertical shock absorber to generate the current optimal damping so as to apply the corresponding damping force to the vehicle system.
In this embodiment, the control system for vertical shock absorber 13 includes:
the detection module 22 is used for acquiring current attitude information of the rail vehicle;
the control module 21 is configured to determine a current optimal damping of the vertical shock absorber according to the current attitude information acquired by the detection module 22 and preset ideal attitude information of the rail vehicle, and send a control instruction corresponding to the current optimal damping to the vertical shock absorber, so that the vertical shock absorber generates the current optimal damping.
With reference to fig. 2, specifically, the current attitude information of the rail vehicle may be represented by an operating state parameter, the operating state parameter of the train is obtained by the detection module 22, converted into a corresponding current attitude parameter by attitude sensing, and input to the control module 21 together with the ideal attitude parameter, and the intelligent control strategy of the control module 21 may determine the damping control amount of the vertical shock absorber 13 according to the input related parameter, so as to send a control instruction to the vertical shock absorber 13 to adjust the damping.
According to the control method and the control system, the ideal attitude information is taken as a control target, the current optimal damping of the vertical shock absorber is determined according to the current attitude information of the railway vehicle, so that the damping force of the vertical shock absorber acting on the train is adjusted, the damping of the vertical shock absorber is adjusted in real time according to the running state of the railway vehicle, the vertical shock absorber can be adjusted in a self-adaptive mode according to the running condition of the vehicle, the train suspension system is always in a better matching requirement, and the dynamic performance of the railway vehicle is effectively improved.
In this embodiment, control module 21 specifically employs a depth-determining strategic gradient algorithm to determine the current optimal damping for vertical shock absorber 13.
Wherein, the Deep Deterministic Policy Gradient algorithm (DDPG) is further refined and enriched based on the reinforced learning principle. The strategy network and the Q network of the DDPG are both formed by a deep neural network, and the action strategy and the approximation function are generated by utilizing the neural network. In order to stabilize the learning process, two sets of neural networks are arranged in the two parts, one is called a current network and the other is called a target network, and the initial parameters of the two sets of networks are the same. The current network parameters are completely updated after training, and the target network parameters are only slightly updated according to the old and new parameters of the current network.
In addition, the memory base is used for storing conversion data in iterative interaction, and after a certain interaction round, a part of data is randomly extracted from the base to train the current network. In order to increase exploration on the motion space and obtain richer training data, the DDPG gives random noise to the generated motion and then gives the motion to the environment for execution. After the training is finished to obtain a usable controller, the process of adding random noise should be removed.
Referring to fig. 3, fig. 3 is a control schematic diagram of the control module shown in fig. 2.
Specifically, the control module 21 includes an algorithm unit 211 and a driver 212, and the algorithm unit 211 is configured to implement a depth deterministic strategic gradient algorithm, including a strategic network module 2111, a Q network module 2112, and a repository module 2114.
A method for determining a current optimum damping for a vertical shock absorber 13 utilizing a depth-determining strategic gradient algorithm comprising:
determining current ideal attitude information axtWith current attitude information ax’tCurrent error e oft(ii) a The control target in fig. 3 represents the ideal attitude information;
the strategy network module 2111 according to the current ideal attitude information axtAnd current attitude information ax’tDetermined current error etSeeking a calculation function of an optimal control command by using the Q network module 2112, determining the current optimal damping corresponding to the current error by using the calculation function sought by the Q network module 2112, and sending a corresponding control command a according to the current optimal dampingtTo the vertical damper 13.
Specifically, the policy network module 2111 sends the control instruction atIs sent to driver 212, and driver 212 sends a drive signal that drives the change in damping of vertical shock absorber 13.
In actual operation, the control instruction of the policy network module 2111 is: a ist＝μ(et) Wherein the μ function is a control strategy generated according to the error, the current strategy network and the target strategy network in the strategy network module 2111 are obtained by fitting the μ function and the μ' function respectively by a convolutional neural network through a training method, and the parameters are θμAnd thetaμ’。
In practical operation, the policy network module 2111 specifically controls the command a according to the policy network moduletObtaining a current command corresponding to the damping force through a table stored in advanceThe current command is sent to the driver 212. The table can be obtained by calibration in advance, and a table corresponding to the damping force and the current is obtained. Of course, it is understood that in practical applications, besides controlling the damping change by the current, the damping change may also be controlled by other parameters, which are determined according to the requirements.
Wherein the Q network module 2112 is used to find a calculation function of the optimal control instruction.
The Q network module 2112 includes a current Q network and a target Q network, the current Q network based on the N sets of sampled data (e) in the memory bank module 2114i、ai、ri、ei+1) And the state e provided by the target policy networkt′、μ′(et+1) Calculating the current Q value R, and calculating the target Q value y by the target Q networktQ 'in R + γ Q', and copying the network parameters from the current Q network to the target Q network.
It can be understood that, in practical applications, the Q network module 2112 and the policy network module 2111 are mutually influenced.
In this scheme, the Q network uses a bp (back propagation) neural network. The basic principle of the BP neural network algorithm is that an input signal is propagated forwards through a layer-by-layer network to finally obtain output, then an error is propagated backwards from an output layer to an input layer, each layer of error is distributed to all neurons in the layer, neuron parameters are updated layer by utilizing a gradient descent method based on an error value, and the aim is to minimize the mean square error between a network output value and an expected output value. A minimum threshold is typically set and training is stopped when the mean square error is less than this value.
The output of each neuron can be expressed as:
wherein x isiIs the output of a neuron in the upper layer, omegaiThe weight connection between a certain neuron and the present neuron in the previous layer, b is a threshold value in the present neuron, and sigma is an activation function with various choices. The BP neural network can add a plurality of layers of neural networks between an input layer and an output layer, and the layers are not connected with the outsideA neural network that is directly connected is referred to as a "hidden layer," e.g., in one embodiment, a network with 3 hidden layers, 10 neurons per layer, may be selected.
The traditional BP neural network has low learning speed and is easy to fall into a local minimum value. Because the strategy Network directly generates the control signal, in order to make the strategy Network training convergence faster and result better, a Residual Network (Residual Network) is used in the scheme. In contrast to conventional BP neural networks, residual networks add a path from input to output. The output of the network can be expressed as:
the residual network corresponds to changing the learning objective, and instead of learning a complete output, the residual network accelerates the convergence rate by learning the difference between the target values h (x) and x, i.e., the so-called residual f (x) h (x) -x. In this scheme, a residual network with 2 hidden layers and 60 neurons in each layer can be selected.
The bank memory module 2114 is used for storing the current error etAnd control instructions a generated by the policy network module 2111t。
In practical application, mini-batch data is randomly adopted from the memory bank module 2114, and the policy network module 2111 and the Q network module 2112 are trained and learned to obtain the optimal network parameter thetaμAnd thetaμ’And thetaQAnd thetaQ’。
In a specific embodiment, the bank module 2114 further stores the error e at the next timet+1Error e of the next timet+1Ideal attitude information a for the next momentx t+1Actual attitude information a with the next timex't+1The difference of (a).
In a specific embodiment, the algorithm unit 211 further includes a policy evaluation module 2113, and the policy evaluation module 2113 is configured to calculate the reward riAnd stored in the memory bank module 2114 to participate in training learning for the policy network module 2111 and the Q network module 2112.
rt＝(axt-ax't+1)2(ii) a Wherein, axtAs ideal attitude information, ax't+1Is the attitude information of the next moment.
As described above, the principle of each block of the control block 21 is explained, and the control flow of the control block 21 can be understood as follows with reference to fig. 3:
1) current ideal attitude axtWith the currently acquired actual attitude ax’tMaking difference to obtain current error et；
2) The current error etInputting the strategy network module 2111, generating a control command at；
3) Table lookup to determine the corresponding control command atCurrent, vertical shock absorber response atActing on the vehicle system, acquiring the actual posture a after the actionx’t+1(ii) a It can be understood that ax’t+1As attitude information of the next time
4) Calculating a reward function: r ist＝-(axt-ax't+1)2；
5) Ideal posture a at the next momentxt+1And the actual attitude a at the next momentx’t+1The difference is made to obtain the error e of the next timet+1；
6) The above (e)t、at、rt、et+1) Store in the memory module 2114;
7) randomly sampling mini-batch data from the memory bank module 2114, training the policy network module 2111 and the Q network module 2112, and updating the network parameter thetaμAnd thetaμ’And thetaQAnd thetaQ’；
8) Return to 1), at which time et+1Become et。
It should be noted here that, although the current ideal posture and the ideal posture at the next moment are mentioned in the foregoing control strategies, it can be understood that the ideal posture information is actually unchanged, and only in the control strategies, the ideal posture value needs to be assigned to each round of control, so that the above distinction is made.
In this embodiment, as described above, the attitude information of the rail vehicle may be characterized by the running state parameters of the train, and specifically, the running state parameters characterizing the attitude information include the vibration parameters of the vehicle body 11 and/or the vibration parameters of the frame 12.
The vibration parameters of the vehicle body 11 include vibration acceleration and/or pitch vibration angular acceleration, and the vibration parameters of the frame 12 include vibration acceleration and/or pitch vibration angular acceleration.
The aforementioned detection module 22 is used for detecting these vibration parameters, and is specifically selected and arranged according to actual control requirements, for example, the vibration acceleration may be obtained by a vibration sensor, the vibration angular acceleration may be obtained by a gyroscope, and the vibration angular acceleration may also be obtained by a vibration acceleration sensor.
Specifically, which parameter or parameters are selected to represent the attitude information of the rail vehicle is subject to actual control requirements.
As above, the attitude information of the rail vehicle may be characterized by one parameter or two or more parameters, specifically, the attitude of the rail vehicle to be adjusted is taken as a reference, that is, the corresponding attitude information is different according to different control targets. For example, when the control of the heave vibration, pitch vibration or roll vibration of the vehicle body 11 is required, the attitude information thereof needs to be represented by more than two parameters.
Specifically, when the posture information of the rail vehicle is represented by a single parameter, the algorithm unit 211 specifically uses the parameter representation as a control target, that is, the ideal posture information is an ideal value of the representation parameter, the current posture information is a current measured value of the representation parameter, and the current error is a difference between the ideal value and the current measured value of the representation parameter.
The algorithm unit 211 determines a control damping amount according to the related information of the characterization parameter, and the control damping amount is the current optimal damping.
Specifically, when the attitude information of the rail vehicle is represented by more than two parameters, the control module 21 has more than two control targets, and the algorithm unit 211 determines the damping control amount corresponding to each characterization parameter according to each characterization parameter.
For example, when the floating vibration and nodding vibration of the vehicle body are controlled simultaneously, the control target may be selected as the floating vibration acceleration and nodding vibration acceleration, theoretically, each acceleration is controlled to be 0, and then the reward function is as follows:
rt＝-[λ1(axt-ax't+1)2+λ2(aθt-aθ’t+1)2]；
aθtand aθ’t+1Respectively are ideal attitude information of nodding motion at the gravity center of the vehicle body and attitude information at the next moment. Lambda [ alpha ]1And λ2Is a control coefficient.
Generated control instruction atThe floating and sinking vibration and nodding vibration control are superimposed.
And (5) carrying out multi-target analogy in sequence.
It should be noted that, as mentioned above, at least four vertical vibration absorbers 13 are disposed on one frame 12, and the attitude information of the corresponding rail vehicle is different for the vertical vibration absorbers 13 at different positions, that is, the control amount of each vertical vibration absorber 13 may be different, in practical applications, a plurality of algorithm units 211 may be disposed in the same control module 21 to correspond to a plurality of vertical vibration absorbers 13 one by one, and of course, one control module 21 may be disposed for each vertical vibration absorber 13.
The corresponding attitude information of each vertical shock absorber 13 can be obtained by arranging a corresponding detection module 22 near each vertical shock absorber 13, or by arranging a detection module 22 at a proper position of the vehicle body 11 or the frame 12 according to actual requirements, and then determining the attitude information corresponding to each vertical shock absorber 13 according to the detection information of the detection module 22. That is, the attitude information of the rail vehicle indicated in the foregoing method or system refers to the attitude information for the vertical shock absorber 13 to be controlled.
In addition, the invention also provides a railway vehicle, which comprises the vertical shock absorber and a control system of any one of the vertical shock absorbers.
Since the rail vehicle has the control system, the technical effects of the rail vehicle are the same as those of the control system, and the details are not repeated here.
The control method and the control system for the rail vehicle and the vertical shock absorber provided by the invention are described in detail above. The principles and embodiments of the present invention are explained herein using specific examples, which are presented only to assist in understanding the method and its core concepts. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
Claims (13)
1. A control method for a vertical shock absorber of a railway vehicle, the vertical shock absorber being a semi-active shock absorber, both ends of which are connected to a frame and an axle box of the railway vehicle, respectively, characterized in that the control method comprises:
acquiring current attitude information of the rail vehicle;
and determining the current optimal damping of the vertical shock absorber according to the current attitude information and preset ideal attitude information of the railway vehicle, and controlling the vertical shock absorber to generate the current optimal damping.
2. The method of claim 1, wherein determining the current optimal damping of the vertical shock absorber based on the current attitude information and the predetermined desired attitude information of the rail vehicle comprises: and determining the current optimal damping of the vertical shock absorber by adopting a depth determinacy strategic gradient algorithm.
3. The method of claim 2, wherein the depth-determinative strategic gradient algorithm comprises a strategic network module, a Q-network module, and a memory library module;
the method for determining the current optimal damping of the vertical shock absorber by adopting the depth certainty strategic gradient algorithm comprises the following steps:
determining a current error of the ideal attitude information and the current attitude information;
the strategy network module utilizes the Q network module to seek a calculation function of an optimal control instruction according to the current error, determines the current optimal damping corresponding to the current error according to the calculation function sought by the Q network, and sends a corresponding control instruction to the vertical shock absorber according to the current optimal damping;
the strategy network module and the Q network module carry out training and learning according to the data stored in the memory base module;
and the memory library module is used for storing the current error and the control instruction.
4. The method of claim 3, wherein the memory bank module further stores an error at a next time, the error at the next time being a difference between the ideal attitude information and actual attitude information at the next time;
the depth certainty strategic gradient algorithm further comprises a strategy evaluation module, wherein the strategy evaluation module is used for calculating rewards according to the ideal posture information and the actual posture information at the next moment, and storing the rewards to the memory base module so as to participate in training and learning of the strategy network module and the Q network module.
5. The method of claim 4, wherein the reward calculation function in the policy evaluation module is:
rt＝(axt-ax't+1)2；
wherein, axtFor the ideal attitude information, ax't+1And the attitude information of the next moment.
6. The method for controlling a vertical shock absorber of a railway vehicle according to any one of claims 1 to 5, wherein the attitude information of the railway vehicle comprises a vibration parameter of a vehicle body and/or a vibration parameter of a frame.
7. The method of claim 6, wherein the vibration parameter of the body comprises a vibration acceleration and/or a pitch vibration angular acceleration, and/or wherein the vibration parameter of the frame comprises a vibration acceleration and/or a pitch vibration angular acceleration.
8. The method for controlling a vertical shock absorber of a railway vehicle as claimed in claim 6, wherein when the attitude information of the railway vehicle is characterized by a single parameter, a corresponding control damping amount is determined according to the single parameter, and the control damping amount is the current optimal damping;
when the attitude information of the rail vehicle is represented by more than two parameters, damping control quantities corresponding to the parameters are determined according to the parameters respectively, and then the current optimal damping is determined according to the damping control quantities.
9. Control system of rail vehicle's vertical shock absorber, vertical shock absorber is semi-active shock absorber, its both ends respectively with rail vehicle's framework and axle box connection, characterized in that, control system includes:
the detection module is used for acquiring current attitude information of the rail vehicle;
and the control module is used for determining the current optimal damping of the vertical shock absorber according to the current attitude information acquired by the detection module and the pre-stored ideal attitude information of the rail vehicle, and sending a control command corresponding to the current optimal damping to the vertical shock absorber.
10. The control system of a rail vehicle vertical shock absorber of claim 9, wherein the control module comprises a driver and an algorithm unit, the algorithm unit employing a depth-determinative strategic gradient algorithm comprising a strategic network module, a Q-network module, and a memory library module;
the strategy network module is used for determining the current optimal damping according to the current attitude information, the ideal attitude information and a calculation function sought by the Q network module, and sending a control instruction corresponding to the current optimal damping to the driver;
the Q network module is used for seeking a calculation function of an optimal control instruction;
the strategy network module and the Q network module are also used for training and learning according to the data stored in the memory base module;
the memory library module is used for storing a current error and the control instruction, wherein the current error is a difference value between the ideal attitude information and the current attitude information.
11. The control system of a rail vehicle vertical shock absorber according to claim 10, wherein the memory bank module is further configured to store an error at a next time, the error at the next time being a difference between the ideal attitude information and actual attitude information at the next time;
the control module further comprises a strategy evaluation module, and the strategy evaluation module is used for calculating rewards according to the ideal posture information and the actual posture information at the next moment and storing the rewards to the memory base module so as to participate in training and learning of the strategy network module and the Q network module.
12. The control system for a vertical shock absorber of a railway vehicle according to any one of claims 9 to 11, wherein the attitude information of the railway vehicle comprises a vibration parameter of a vehicle body and/or a vibration parameter of a frame; the detection module comprises a vibration sensor and/or a gyroscope which are used for being installed on the vehicle body, and/or a vibration sensor and/or a gyroscope which are installed on the framework.
13. A rail vehicle comprising a vertical shock absorber, further comprising a control system for the vertical shock absorber of any one of claims 9-12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011254231.7A CN112506043B (en) | 2020-11-11 | 2020-11-11 | Control method and control system for rail vehicle and vertical shock absorber |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011254231.7A CN112506043B (en) | 2020-11-11 | 2020-11-11 | Control method and control system for rail vehicle and vertical shock absorber |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112506043A true CN112506043A (en) | 2021-03-16 |
| CN112506043B CN112506043B (en) | 2022-10-11 |
Family
ID=74957905
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011254231.7A Active CN112506043B (en) | 2020-11-11 | 2020-11-11 | Control method and control system for rail vehicle and vertical shock absorber |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112506043B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113238487A (en) * | 2021-06-03 | 2021-08-10 | 中国第一汽车股份有限公司 | Fuzzy control method and control system for adjustable damping shock absorber |
| WO2023029762A1 (en) * | 2021-09-02 | 2023-03-09 | 杭州中车车辆有限公司 | Vehicle body attitude adjustment method and vehicle body attitude adjustment system |
| CN117807895A (en) * | 2024-02-28 | 2024-04-02 | 中国电建集团昆明勘测设计研究院有限公司 | Magnetorheological damper control method and device based on deep reinforcement learning |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120318166A1 (en) * | 2011-06-14 | 2012-12-20 | Amsted Rail Company, Inc. | Railway freight car truck |
| CN104249743A (en) * | 2013-06-28 | 2014-12-31 | 日立汽车系统株式会社 | Suspension control device |
| CN105069259A (en) * | 2015-09-06 | 2015-11-18 | 山东理工大学 | Analytic calculating method for optimal damping ratio of secondary vertical suspension of low-speed rail vehicle |
| CN110155101A (en) * | 2019-05-17 | 2019-08-23 | 中车青岛四方机车车辆股份有限公司 | Laterally full active control vibration insulating system and the wherein control method of controller |
| CN110989576A (en) * | 2019-11-14 | 2020-04-10 | 北京理工大学 | Target following and dynamic obstacle avoidance control method for differential slip steering vehicle |
| CN111775649A (en) * | 2019-04-04 | 2020-10-16 | 上海汽车集团股份有限公司 | Shock absorber control method, device and system |
| CN111781940A (en) * | 2020-05-19 | 2020-10-16 | 中车工业研究院有限公司 | Train attitude control method based on DQN reinforcement learning |
-
2020
- 2020-11-11 CN CN202011254231.7A patent/CN112506043B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120318166A1 (en) * | 2011-06-14 | 2012-12-20 | Amsted Rail Company, Inc. | Railway freight car truck |
| CN104249743A (en) * | 2013-06-28 | 2014-12-31 | 日立汽车系统株式会社 | Suspension control device |
| CN105069259A (en) * | 2015-09-06 | 2015-11-18 | 山东理工大学 | Analytic calculating method for optimal damping ratio of secondary vertical suspension of low-speed rail vehicle |
| CN111775649A (en) * | 2019-04-04 | 2020-10-16 | 上海汽车集团股份有限公司 | Shock absorber control method, device and system |
| CN110155101A (en) * | 2019-05-17 | 2019-08-23 | 中车青岛四方机车车辆股份有限公司 | Laterally full active control vibration insulating system and the wherein control method of controller |
| CN110989576A (en) * | 2019-11-14 | 2020-04-10 | 北京理工大学 | Target following and dynamic obstacle avoidance control method for differential slip steering vehicle |
| CN111781940A (en) * | 2020-05-19 | 2020-10-16 | 中车工业研究院有限公司 | Train attitude control method based on DQN reinforcement learning |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113238487A (en) * | 2021-06-03 | 2021-08-10 | 中国第一汽车股份有限公司 | Fuzzy control method and control system for adjustable damping shock absorber |
| CN113238487B (en) * | 2021-06-03 | 2022-08-09 | 中国第一汽车股份有限公司 | Fuzzy control method and control system for adjustable damping shock absorber |
| WO2023029762A1 (en) * | 2021-09-02 | 2023-03-09 | 杭州中车车辆有限公司 | Vehicle body attitude adjustment method and vehicle body attitude adjustment system |
| CN117807895A (en) * | 2024-02-28 | 2024-04-02 | 中国电建集团昆明勘测设计研究院有限公司 | Magnetorheological damper control method and device based on deep reinforcement learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112506043B (en) | 2022-10-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112506043B (en) | Control method and control system for rail vehicle and vertical shock absorber | |
| RU2568163C1 (en) | Vehicle control device and process | |
| RU2568048C1 (en) | Vehicle control device and process | |
| CN110347155B (en) | Intelligent vehicle automatic driving control method and system | |
| JP2005538886A (en) | Fuzzy controller using a reduced number of sensors | |
| CN104203609A (en) | Vehicle control device and vehicle control method | |
| CN109334378B (en) | Vehicle ISD suspension active control method based on single neuron PID control | |
| CN111487863B (en) | Active suspension reinforcement learning control method based on deep Q neural network | |
| CN110673480B (en) | Robust control method of time-varying displacement constrained active suspension system | |
| CN111781940B (en) | Train attitude control method based on DQN reinforcement learning | |
| CN108973578B (en) | AGV suspension system elasticity adjusting method based on deep reinforcement learning | |
| CN112158045A (en) | Active suspension control method based on depth certainty strategy gradient | |
| JP4609767B2 (en) | Optimal system control method | |
| Kaldas et al. | Rule optimized fuzzy logic controller for full vehicle semi-active suspension | |
| KR102654627B1 (en) | Vehicle control device, vehicle control method and vehicle control system | |
| CN113591360B (en) | Magneto-rheological damper structural parameter optimization method based on whole vehicle dynamics model | |
| CN115826402A (en) | Active suspension control strategy generation method based on deep reinforcement learning algorithm | |
| Tchamna et al. | Attitude control of full vehicle using variable stiffness suspension control | |
| CN108608822B (en) | AGV suspension system elasticity adjusting method and system | |
| Gustafsson et al. | Neural network controller for semi-active suspension systems with road preview | |
| KR20230138097A (en) | Method for controlling driving force of vehicle | |
| KR20230152918A (en) | Method for controlling regenerative braking of vehicle | |
| Li | Robust control design for vehicle active suspension systems with uncertainty | |
| CN113791539A (en) | Self-adaptive neural network control method based on nonlinear active suspension | |
| CN117841591B (en) | ISD suspension control method based on improved fuzzy neural network PID |
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 | ||
| GR01 | Patent grant |