Smooth downslope connection polishing method for steel rail low step defect

Abstract

The invention relates to a smooth downslope connection polishing method for a low-step defect of a steel rail, and belongs to the technical field of engineering. A laser array scanner of the grinding equipment scans the steel rail, scanning data are analyzed, and a low treading point of the steel rail is found; the low treading point is positioned, a grinding head of grinding equipment grinds from the low treading point to the two sides, and the low treading point is ground to meet the specification; the grinding equipment moves to one side for grinding, and single-side down slope detection is completed; and the two sides are polished along the slope. By means of the method, the downslope connection problem of the interface and the original rail surface after steel rail low step defect grinding can be solved, and the excessive abrupt problem of the connection position of the grinding area and the original rail surface is avoided.

Description

Technical Field

[0001] The invention belongs to the field of engineering technology, and in particular relates to a method for smoothly connecting and grinding low-stepping defects of rails. Background Art

[0002] The most common damage to high-speed rails is between 0.2mm and 0.5mm in depth. Large grinding trains cannot accurately repair small defects. Small grinding trains are mainly operated manually, which is inefficient, quality cannot be guaranteed, and it is easy to cause over-grinding.

[0003] At present, there are two main types of rail grinding equipment. One is a large grinding vehicle, which mainly realizes the transition between the ground and unground surfaces of the rail by combining multiple grinding heads. Due to the high speed of large grinding vehicles (12-18km per hour), the grinding wheel adjustment accuracy is limited, resulting in the need for 150m of rail grinding along the slope. This method is only suitable for long-distance rail grinding along the slope, and cannot complete the grinding of low-stepping defects of rails less than 0.4mm.

[0004] The other is a small manual grinding car, which mainly completes the grinding of the rail by controlling the contact angle between the grinding wheel and the rail. Because there is no stable reference surface, the slope operation of manual grinding mainly depends on the experience of the grinding operator. When approaching the interface connection position, passive grinding is adopted, that is, relying on the gravity of the grinding car to give a certain pressure to the grinding wheel, dragging from the lowest point to the highest point to complete the slope connection of the grinding interface. This method cannot guarantee the consistency of the slope operation and the stability of the grinding quality. In actual operation, over-grinding is prone to occur, thereby reducing the service life of the rail.

[0005] CN202410493237.1 A method for grinding the uneven rail surface within 3m of a high-speed railway rail welding joint. The method for grinding the uneven rail welding joint mainly solves the problem of grinding the convex defects of the rail weld after welding, but cannot solve the problem of grinding the concave defects of the rail.

[0006] CN202411334649.7 large machine precision grinding process method, the grinding of this large grinding vehicle is only suitable for solving the straightness and profile grinding of long-distance rails, and cannot solve the local rail depression, especially the problem of the slope of the grinding interface in short-distance (less than 4 meters) local areas.

[0007] Under the background of the rapid development of high-speed railway, the two existing types of grinding equipment cannot meet the growing demand for the slope connection of the rail interface after short-distance local grinding. A method that can automatically complete the slope connection between the ground surface and the unground surface is particularly important. Summary of the invention

[0008] In order to solve the above technical problems, the present invention provides a method for smooth slope connection and grinding of low-stepping defects of rails. The present invention aims to automatically complete the slope connection problem of the grinding surface of high-speed rails by accurately controlling the action of the grinding equipment. According to regulations, the slope rate of the rail body straightness in the vertical direction of the rail straightness allowable deviation is less than or equal to 0.2mm / 1m, that is, a slope rate of 0.2 thousandths.

[0009] The specific technical solution is: a method for smoothly grinding the low-stepping defect of a rail along the slope, comprising:

[0010] initialization;

[0011] The laser array scanner of the grinding equipment scans the rails, and the scan data is analyzed to find the low stepping point of the rails;

[0012] Locate the low stepping point, and the grinding head of the grinding equipment grinds from the low stepping point to both sides, and the low stepping point grinding complies with the regulations;

[0013] The grinding equipment moves to one side to grind, and the inspection is completed on one side along the slope;

[0014] Move to the other side and grind, and complete the double-sided grinding along the slope.

[0015] The present invention has the following beneficial effects:

[0016] The method can solve the problem of the smooth connection between the interface after the low-stepping defect of the rail and the original rail surface, and avoid the excessive abruptness of the connection between the grinding area and the original rail surface. The curve after grinding by this method has better connection smoothness, reduces the impact when the wheel passes, and improves the comfort of vehicle driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the structure of the grinding equipment of the present invention;

[0018] Figure 2 A flow chart of a method for smoothly joining and grinding a low-stepping defect of a rail according to the present invention;

[0019] Figure 3a This is the side view of the polished surface after the first polishing;

[0020] Figure 3b This is the side view of the polished surface after the second polishing;

[0021] Figure 3c This is the side view of the polished surface after polishing is completed;

[0022] Figure 4 This is a comparison diagram between the polished surface of the present invention and the polished surface of the prior art.

[0023] The reference numerals in the figure are: traveling wheel 1, grinding wheel 2, grinding head 3, frame 4, guide rail 5, laser array scanner 6. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above-mentioned purpose, the present invention adopts the following technical scheme.

[0025] The present invention relates to a method for smoothly grinding a low-step defect of a rail, wherein the low-step defect is a depression of the rail, and the low-step point is the lowest area of ​​the low-step defect. Figure 1 As shown, the grinding equipment includes a running wheel 1, a grinding wheel 2, a grinding head 3, a frame 4, a guide rail 5, and a laser array scanner 6, realizing the functions of automatic grinding, automatic running, and automatic measurement. The grinding equipment scans the rail through the laser array scanner 6 to obtain the low stepping point; the grinding wheel 2 is controlled by a servo motor and a lead screw, and the grinding wheel 2 moves up and down along the guide rail 5. The grinding wheel 2 can move left and right and forward and backward along the frame 4 through the grinding head 3. The control accuracy of the grinding wheel is 0.01mm, and the grinding equipment can use a driving motor to achieve self-running.

[0026] like Figure 2 As shown, a flow chart of a method for smoothly grinding a low-stepping defect of a rail according to the present invention is provided, and the method specifically comprises:

[0027] initialization;

[0028] The laser array scanner 6 scans the rails, and the scanned data is analyzed to find the low stepping point of the rails;

[0029] Locate the low stepping point, and grind with grinding head 3 from the low stepping point to both sides, and the low stepping point is ground in compliance with regulations;

[0030] The grinding equipment moves to one side to grind, and the inspection is completed on one side along the slope;

[0031] Move to the other side and grind, and complete the double-sided grinding along the slope.

[0032] The initialization includes: placing a grinding device on the rail, and adjusting the grinding plane of the grinding wheel of the grinding device to be parallel to the rail surface tangent plane of the rail.

[0033] Among them, the laser array scanner 6 scans the rails, analyzes the scanning data, and finds the low stepping point of the rails; including: starting the grinding equipment, driving the belt through the servo motor to make the grinding equipment move on the rails to be polished, the laser array scanner 6 on the grinding equipment scans the rails, collects the surface data of the rails, and finds the low stepping point of the rails to be polished after analysis.

[0034] Among them, the low stepping point is located, and the grinding head 3 grinds from the low stepping point to both sides, and the low stepping point grinding complies with the regulations; including:

[0035] Control the grinding wheel 2 on the grinding head 3 to move to the center of the low stepping point, and control the grinding wheel 2 to grind from the middle to both sides at a slope of 0.2 per thousand; repeat several times until the low stepping point is ground to compliance; that is, the vertical deviation of the rail body straightness of the allowable deviation of the rail straightness is less than or equal to 0.2mm / 1m.

[0036] The grinding equipment moves to one side to grind and detect that the single side is completed along the slope; including:

[0037] Step 1) The left running wheel of the grinding device is driven by a motor to the lowest point of the low step point, at which time the right running wheel of the grinding device will be on the original rail surface; Figure 3a shown.

[0038] Step 2) Start the grinding head and move it back and forth on the grinding equipment until Figure 3c The area above the middle red line A1B1 is polished, and the first polishing is completed;

[0039] Step 3) After the first grinding is completed through laser scanning detection, the grinding equipment starts to move to the right, calculated by the moving distance of the right end running wheel rail surface, and moves 1 cm each time, repeating step 2) until the left end running wheel moves to the original rail surface height; Figure 3b shown.

[0040] Move to the other side and grind, double-sided grinding is completed along the slope; including: Step 1) The grinding equipment moves to the left until the right running wheel reaches the lowest point of the low step point, and the left running wheel will now be on the original unrepaired rail surface; Step 2) Start the grinding equipment, and grind the grinding head from the low point to the high point until the grinding is completed; Step 3) After the first grinding is completed through laser scanning detection, the grinding equipment starts to move to the left, calculated by the moving distance of the left end running wheel rail surface, moving 1 cm each time, and repeating step 2) until the right end running wheel moves to the original rail surface height.

[0041] After the double-side grinding is completed, the straightness curve of the rail will be as follows Figure 4Curve shown: As can be seen from the figure, the curve 2 polished by the polishing equipment of the present invention has better smoothness characteristics than the curve 1 polished in the prior art.

[0042] In order to ensure smooth connection between the ground and unground surfaces of the rail, passive grinding is adopted, that is, the grinding wheel presses on the rail only by its own weight, and the control part does not apply pressure to the grinding wheel. The connection along the slope is completed by moving the grinding equipment first and then the grinding wheel.

[0043] The present invention automatically completes the positioning of the lower cutter position of the grinding head through laser scanning data analysis; completes the automatic grinding of the low step point through program control; completes the slope connection between interfaces through a passive grinding method in which the grinding head presses on the rail with its own weight; and completes the grinding of the slope curve through the overall movement of the equipment.

[0044] By moving the grinding equipment to one side, e.g. Figure 3c The grinding surface is shown as follows: Figure 3c As shown in the figure, the side sight line after the first grinding is A1B1; the side sight line after the second moving grinding is A2B2; the third is A3B3; the fourth is A4B4, and so on, until An reaches the rail plane. At this time, the curve formed by A1A2A3A4…An is the slope smooth curve of the ground surface and the unground surface of the rail low stepping defect.

[0045] During the grinding process, an adaptive control model based on machine learning is introduced. Through the machine learning algorithm, an adaptive control model is established to dynamically adjust the motion trajectory of the grinding head according to the real-time scanning data of the rail surface grinding. The grinding methods using the adaptive control model include:

[0046] Data collection: During the grinding process, the laser array scanner collects data on the rail surface in real time, including the depth and length of the rail's low stepping point and the actual amount of grinding completed each time.

[0047] Model training: Use historical grinding data to train the machine learning model so that it can automatically generate the optimal grinding path and force based on different rail surface characteristics.

[0048] Real-time adjustment: During the grinding process, the model dynamically adjusts the motion trajectory of the grinding head based on the real-time collected data to ensure the optimal grinding effect.

[0049] The polishing algorithm of the adaptive control model includes:

[0050] Select the reinforcement learning model Q-learning. In each control cycle, select the optimal action based on the current state to generate the motion trajectory of the grinding head. After executing the action, observe the next state and reward and update the model. Through continuous learning and optimization, the optimal grinding path is generated.

[0051] Q-learning is a reinforcement learning algorithm based on the Q-value function. The Q-value function is used to evaluate the long-term benefits of taking a certain action in a certain state. The update formula of the Q-value function is:

[0052] (1)

[0053] Among them: s: current state; : current action; : next state; : next action; R: immediate reward; : Learning rate, used to control the weight of new information, the initial value is set to 0.1; : Discount factor, controlling the importance of future rewards, with an initial value of 0.95.

[0054] In each control cycle, the optimal action is selected according to the current state s The process is as follows:

[0055] The first step is Q value query: query the Q value of all possible actions under the current state s ;

[0056] The second step is action selection: select the current action that maximizes the Q value under the current state s. :

[0057] ;

[0058] Step 3: Execute the action: Execute the current action selected , and observe the next state s′ and the immediate reward R.

[0059] The current state s contains all the information related to the grinding operation, which is used to describe the current environment and the state of the grinding head. The state is defined as:

[0060] ;

[0061] in:

[0062] D: Depth of low step point (unit: mm).

[0063] W: Width of the low step point (unit: mm).

[0064] x: lateral position of the grinding head on the rail surface (unit: mm).

[0065] y: longitudinal position of the grinding head on the rail surface (unit: mm).

[0066] V: Current moving speed of the grinding head (unit: mm / s).

[0067] R is the immediate feedback obtained after executing the action, which is used to evaluate the quality of the action. In the process of rail grinding, the design of the reward function encourages the grinding head to complete the grinding task efficiently and accurately. The reward function is defined as:

[0068] (2)

[0069] in: : Roughness of the rail surface after grinding (unit: mm).

[0070] E: Grinding efficiency (grinding length per unit time, unit: mm² / s).

[0071] F: Grinding force of the grinding head (unit: N).

[0072] The roughness of the rail surface after grinding The weight coefficient, β is the grinding efficiency The weight coefficient, γ is the grinding strength of the grinding head The weight coefficients of . (The initial values ​​are set to 0.4, 0.3, and 0.3 respectively).

[0073] After executing the action, the next state s′ and the immediate reward R are observed, and the Q-value function is updated. Through Q-learning, by defining the state, calculating the reward, updating the model, continuous learning and optimization, the optimal grinding path is generated. This method can significantly improve the accuracy and efficiency of rail grinding, reduce manual intervention, and realize intelligent and automated grinding process.

[0074] Model training includes:

[0075] Divide the data into training and test sets; use the training set to train the machine learning model; use the test set to evaluate the performance of the model.

[0076] This method uses grid search. Through grid search, all combinations are traversed, and model training and evaluation are performed for each hyperparameter combination. Finally, the hyperparameter combination with the best performance on the validation set is selected through experimental verification. For example, based on a test of a polishing equipment, the specific results are as follows:

[0077] In formula (1) : The learning rate is adjusted to 0.05; : The discount factor is adjusted to 0.95, the weight of the roughness α in formula (2) is adjusted to 0.5; the weight of the grinding efficiency β is adjusted to 0.3; the weight of the grinding intensity γ is adjusted to 0.2.

Claims

1. A method for smoothly grinding rail low-stepping defects along the slope, characterized in that: The following steps are involved: initialization; The laser array scanner of the grinding equipment scans the rails, and the scan data is analyzed to find the low stepping point of the rails; Locate the low stepping point, and the grinding head of the grinding equipment grinds from the low stepping point to both sides, and the low stepping point grinding complies with the regulations; The grinding equipment moves to one side to grind, and the inspection is completed on one side along the slope; Move to the other side and grind, and complete the double-sided grinding along the slope.

2. A method for smoothly grinding rail low-stepping defects according to claim 1, characterized in that: Initialization includes: Place the grinding device on the rail so that the grinding plane of the grinding wheel on the grinding head is parallel to the rail surface tangent plane of the rail.

3. A method for smoothly grinding rail low-stepping defects along the slope according to claim 2, characterized in that: in, The laser array scanner of the grinding equipment scans the rails, and the scanning data is analyzed to find the low stepping point of the rails; including: starting the grinding equipment to move the grinding equipment on the rails to be ground, the laser array scanner on the grinding equipment scans the rails, collects the surface data of the rails, and finds the low stepping point of the rails to be ground after analysis.

4. A method for smoothly grinding rail low-stepping defects according to claim 3, characterized in that: in, Locate the low stepping point, and use the grinding head of the grinding equipment to grind from the low stepping point to both sides to make the low stepping point grinding comply with the regulations; including: Control the grinding wheel on the grinding head to move to the center of the low stepping point, and control the grinding wheel to grind from the middle to both sides at a slope of 0.2 thousandths; repeat several times until the low stepping point is ground to compliance.

5. A method for smoothly grinding rail low-stepping defects along the slope according to claim 4, characterized in that: Grinding compliance specifically means: the vertical deviation of the rail body straightness of the allowable deviation of rail straightness is less than or equal to 0.2mm / 1m.

6. A method for smoothly grinding rail low-stepping defects according to claim 4, characterized in that: The grinding equipment moves to one side to grind and detect that the single side is completed along the slope; including: Step 1) The left end running wheel of the grinding equipment is moved to the lowest point of the low stepping point, at which time the right end running wheel of the grinding equipment will be on the original rail surface; Step 2) Start the grinding wheel and move it back and forth on the grinding device to grind until the area is polished, completing the first grinding; Step 3) After the first grinding is completed through laser scanning detection, the grinding equipment starts to move to the right, and the distance moved by the right end running wheel is calculated. Each time it moves a fixed distance, step 2) is repeated until the left end running wheel moves to the original rail surface height.

7. A method for smoothly grinding rail low-stepping defects according to claim 6, characterized in that: Move to the other side and grind, and complete the double-sided grinding along the slope; include: Step 1) The grinding equipment moves to the left until the right running wheel reaches the lowest point of the low step point, and the left running wheel will now be on the original unrepaired rail surface; Step 2) Start the grinding equipment, and grind the grinding wheel from the low point to the high point until the grinding is completed; Step 3) After the first grinding is completed through laser scanning detection, the grinding equipment starts to move to the left, and the distance moved by the left running wheel is calculated. Each time it moves a fixed distance, step 2) is repeated until the right running wheel moves to the original rail surface height.

8. The method for smoothly grinding the low-stepping defect of a rail according to claim 2 is characterized in that: Passive grinding is adopted, that is, the grinding wheel presses on the rail only by its own weight.

9. A method for smoothly grinding rail low-stepping defects along the slope according to claim 2, characterized in that: The down-slope connection is accomplished by moving the grinding equipment first and then the grinding wheel.

10. A method for smoothly grinding rail low-stepping defects along the slope according to claim 6 or 7, characterized in that: The fixed distance is 1cm.

11. A method for smoothly grinding rail low-stepping defects according to claim 1, characterized in that: During the grinding process, an adaptive control model based on machine learning is introduced to dynamically adjust the motion trajectory of the grinding head according to the real-time scanning data of the rail surface grinding. The grinding methods using the adaptive control model include: Data collection: During the grinding process, the laser array scanner collects data on the rail surface in real time, including the depth and length of the rail's low step point and the actual amount of grinding completed each time; Model training: Use historical grinding data to train the machine learning model so that it can automatically generate the optimal grinding path and force based on different rail surface characteristics; Real-time adjustment: During the grinding process, the model dynamically adjusts the motion trajectory of the grinding head based on the real-time collected data to ensure the optimal grinding effect.

12. A method for smoothly grinding rail low-stepping defects according to claim 11, characterized in that: The polishing algorithm of the adaptive control model includes: The Q-learning reinforcement learning algorithm is used to evaluate the long-term benefits of taking a certain action in a certain state through the Q-value function. The update formula of the Q-value function is: （1） Among them: s is the current state; For the current action; For the next state; is the next action; R is the immediate reward; is the learning rate, is the discount factor; In each control cycle, the optimal action is selected according to the current state s The process is as follows: The first step is Q value query: query the Q value of all possible actions under the current state s ; Step 2: Action selection: Select the current action that maximizes the Q value in the current state. : ； Step 3: Execute the action: Execute the current action selected , and observe the next state s′ and immediate reward R; The current state s contains all the information related to the grinding operation, which is used to describe the current environment and the state of the grinding head. The state is defined as: ； in: D: Depth of low stepping point; W: width of low step point; x: lateral position of the grinding head on the rail surface; y: longitudinal position of the grinding head on the rail surface; V: current moving speed of the grinding head; R is the immediate feedback obtained after executing the action, which is used to evaluate the quality of the action. In the process of rail grinding, the design of the reward function encourages the grinding head to complete the grinding task efficiently and accurately. The reward function is defined as: （2） in: : Roughness of the rail surface after grinding; E: grinding efficiency; F: grinding force of the grinding head; The roughness of the rail surface The weight coefficient, β is the grinding efficiency The weight coefficient, γ is the grinding intensity The weight coefficient of After executing the action, the next state is observed And the immediate reward R, and update the Q-value function, through Q-learning, by defining the state, calculating the reward, updating the model, continuous learning and optimization, to generate the optimal polishing path.

Images