Intelligent identification device for typical blocky abrasive dust of steel rail grinding train

Abstract

The invention relates to an intelligent recognition device for typical blocky abrasive dust of a steel rail grinding train, the steel rail grinding train is provided with N sections and comprises two train heads and a grinding operation vehicle, and the two train heads are located at the two ends of the train respectively; the intelligent recognition device comprises visible light cameras, a light curtain sensor group and an integrated controller, and the visible light cameras are located on the two vehicle heads; the N light curtain sensor groups are arranged below the fireproof curtain; the integrated controller is located in a cab of the two vehicle heads; in the train running process, the train head at the front end in the train running direction serves as a train head, and the integrated controller on the train head is started; the vehicle head located at the tail end serves as a polishing operation vehicle, and the integrated controller on the vehicle head is closed; when the train is in a first operation working condition, the visible light camera sends a shot picture to the integrated controller, and the picture is identified by blocky abrasive dust; and in the second operation working condition, the light curtain sensor group on the polishing operation vehicle sends the obtained signal to the integrated controller, and the blocky abrasive dust is identified.

Description

Technical Field

[0001] The invention relates to the technical field of rail grinding train wear debris identification, and in particular to a typical block wear debris intelligent identification device for rail grinding trains. Background Art

[0002] With the increase in the operating speed and transportation capacity of high-speed railways, long-term and high-intensity wheel-rail contact will lead to various damages on the rail surface, such as crushing, side wear, corrugation, peeling and fat edge damage and defects. If these damages cannot be handled in time, the rail life will be reduced, the wheel-rail relationship will deteriorate, and the economy and safety of railway transportation will be seriously affected.

[0003] In the railway transportation system, the main factors affecting the service life of rails include rolling contact fatigue and wear. Therefore, rail maintenance mainly considers the production of high-quality rails and rail grinding. Due to the low efficiency of rail replacement, rail grinding has become the most economical and effective maintenance measure. In the early days, rail grinding technology mainly used restorative grinding and surface grinding, and later preventive grinding was developed as the focus.

[0004] At present, rail maintenance and repair are mainly carried out through rail grinding vehicles. Practice has proved that rail grinding is an effective means to prevent and repair rail diseases. Rail grinding can reduce contact fatigue, repair rail corrugation, cracks and other rail diseases, thereby extending the service life of railway rails, improving vehicle running stability and reducing noise.

[0005] However, during the operation of the rail grinding vehicle, the high-temperature grinding chips generated by the rapid friction between the high-speed rotating grinding wheel and the rail will accumulate and form blocks. Some of them will fall to the track section and nearby areas. After the grinding operation is completed, some of the blocky grinding chips fail to fall off after being knocked by manual handheld equipment and are still adsorbed on the inner surface of the fire curtain, which will cause a series of safety hazards: (1) Flying to the area near the line during the grinding process may cause a fire and affect the safety of the line cables and equipment; (2) If it falls onto the line during operation, it may cause damage to the communication and signal equipment on the line; (3) The massive grinding debris that falls on the track will be picked up by the high-speed train, damaging the equipment and body of the train; (4) If it falls into the turnout area, it may cause control circuit failure and lead to traffic accidents.

[0006] Therefore, there is an urgent need to provide an intelligent identification device for typical blocky grinding chips of rail grinding trains to solve the safety hazards caused by high-temperature blocky grinding chips flying during the grinding operation and agglomerated grinding chips falling during self-operation. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide an intelligent identification device for typical block-shaped grinding chips of rail grinding trains.

[0008] To solve the above problems, the technical solution adopted by the present invention is: An intelligent identification device for typical block-shaped wear debris of a rail grinding train, wherein the length and width of the typical block-shaped wear debris are greater than 20 mm×20 mm; the rail grinding train is provided with N sections, which include a locomotive and a grinding operation vehicle, and the locomotive is provided with two sections, which are respectively located at the two ends of the train; The intelligent recognition device includes a visible light camera, a light curtain sensor group and an integrated controller, wherein two visible light cameras are provided, respectively located on the two vehicle heads; N light curtain sensor groups are provided, respectively located on each of the vehicle heads and the grinding operation vehicle, and are provided below the fire curtain; two integrated controllers are provided, respectively located in the cabs of the two vehicle heads; each integrated controller is connected to all visible light cameras and light curtain sensor groups; During the operation of the train, the front locomotive along the train running direction is used as the train locomotive, and the integrated controller on it is turned on; the locomotive at the rear end is used as the grinding operation locomotive, and the integrated controller on it is turned off; When the train is in the first operating condition, the visible light cameras located on the two locomotive heads will send the captured pictures to the integrated controller in the turned-on state to identify the block-like wear chips; when it is in the second operating condition, the light curtain sensor groups on all grinding work vehicles will send the obtained signals to the integrated controller in the turned-on state to identify the block-like wear chips.

[0009] As an embodiment of the invention, each of the light curtain sensor groups includes two laser opposing light curtain sensors, and each of the laser opposing sensors includes a projector and a receiver that are arranged oppositely; Each of the light projectors is composed of a plurality of infrared light emitting tubes arranged at intervals, and each of the receivers is composed of a plurality of infrared receiving tubes arranged at intervals; Each of the laser beam sensors uses a synchronous scanning method to identify the blocked channel, so that the size of the object under the fire curtain can be obtained through two laser beam light curtain sensors to realize the identification of the blocky wear debris.

[0010] As an embodiment of the invention, vertical support hangers are installed on the four sides of each car, and each of the support hangers extends downward to below the fire curtain; The bottom of the two support and suspension frames at the front side is provided with a first cross beam, and the bottom of the two support and suspension frames at the rear side is provided with a second cross beam; the bottom of the two support and suspension frames at the front side is provided with a first longitudinal beam, and the bottom of the two support and suspension frames at the rear side is provided with a second longitudinal beam; The first horizontal beam is provided with a light projector I or a receiver I, and the second horizontal beam is provided with a receiver I or a light projector I; the first longitudinal beam is provided with a light projector II or a receiver II, and the second longitudinal beam is provided with a receiver II or a light projector II.

[0011] As an implementation mode of the invention, the integrated controller in the turned-on state identifies the block-shaped wear debris based on the YOLOv5 algorithm after obtaining the picture taken by the visible light camera; The YOLOv5 network structure includes: input end, backbone network, neck and output end. The input end is used to perform data enhancement and adaptive anchor box calculation operations on the input image; the backbone network includes Foucs, Conv, C3, and SPP modules for image feature extraction; the neck combines the feature pyramid network and the path aggregation network to achieve the enhancement of the fusion capability of feature layers carrying different information; the output end is the prediction part of the network, which uses the extracted features to predict the type and location of the target.

[0012] As an implementation mode of the invention, in the backbone network, the Focus module is used to reduce the amount of calculation; the Conv module performs convolution, batch normalization and Silu function activation operations on the input feature map; In the C3 module, the BottleneckCSP module is replaced by the C3 module; the convolution after the residual output is removed, and the activation function in the convolution module is changed to the Silu function; the image is divided into two branches, one is passed through the convolution module and multiple BottleNeck stacks, and the other is passed through a convolution module, and finally the two branches are channel fused; The SPP module uses 5×5, 9×9, and 13×13 pooling filters to perform maximum pooling on the input image, then performs channel fusion of the pooled image with the original input image, and finally performs a convolution operation.

[0013] As an embodiment of the invention, each of the integrated controllers comprises a shell, an integrated control module and a power supply module are arranged inside the shell, and a touch display screen and an alarm buzzer are arranged outside the shell; The control module is connected to the power supply module, the touch display screen and the alarm buzzer, the power supply module is connected to the touch display screen and the alarm buzzer, and the control module is also connected to all visible light cameras and the light curtain sensor group.

[0014] As an embodiment of the invention, the integrated controller performs the following steps: (1) Start the block wear debris monitoring program After the system is powered on, the rail grinding vehicle block-shaped wear debris shedding monitoring system program is automatically started and displayed on the touch screen. The staff can set the operating parameters and system parameters through the touch screen, wherein the operating parameters include the mobile phone number for receiving alarm SMS; (2) Operation interface display In response to the staff clicking the run button, the system enters the run state and the touch screen jumps to the run interface. The green light flashing on the run interface indicates that the system is running normally, and the red light flashing indicates that the system has a fault and needs to contact a professional for repair. At this time, the integrated control module identifies the blocky wear debris based on the picture taken by the visible light camera or the signal sent by all the light curtain sensor groups. (3) Falling object information alarm When block-shaped wear debris is confirmed to be identified, the alarm buzzer sounds an alarm, the touch display pops up an alarm interface, and sends an alarm text message to the set mobile phone number. The alarm text message includes: date, time, vehicle number, vehicle number, kilometer mark, and a reminder of falling objects.

[0015] As an embodiment of the invention, the integrated controller will also execute the SMS module exception handling steps, which include: performing self-inspection on the SMS module in the running state, and when it is detected that the 4G module has no phone card installed or the phone card is abnormal, the SMS module abnormal information is highlighted and marked in the running interface.

[0016] As an implementation mode of the invention, in the running state, the data query button clicked by the staff can jump to the log display interface; Among them, the log display interface displays: a display button for the day's log, a display button for the latest log, a display button for the month's log, and a button for downloading all logs to a USB flash drive; when a click instruction for the button for downloading all logs to a USB flash drive is received, all logs are downloaded to the inserted USB flash drive.

[0017] The beneficial effects of adopting the above technical solution are: An embodiment of the present invention provides an intelligent identification device for typical blocky wear chips of rail grinding trains. Taking into account the situation that blocky wear chips adsorbed on the inner surface of the fire curtain may fall and affect the safety of equipment, the present invention sets a light curtain sensor group under the fire curtain to detect the area under the waterproof curtain to realize the identification of blocky wear chips, so as to timely alarm and process when blocky wear chips are identified, which can avoid affecting the safety of line cables and equipment, causing damage to communication and signal equipment on the line, damaging the bottom equipment and body of the train, and causing traffic accidents.

[0018] In addition, the present invention can realize the identification of blocky wear chips under low-speed conditions by setting up a visible light camera and an integrated controller; it can realize the identification of blocky wear chips under high-speed conditions by setting up a light curtain sensor group and an integrated controller, thereby realizing running and operation within the full speed range, being suitable for all working conditions, and improving the scope of application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the technical principle of a light curtain sensor provided by the present invention.

[0020] Figure 2 It is a structural schematic diagram of the installation position of a light curtain sensor provided by the present invention.

[0021] Figure 3 It is a schematic diagram of a YOLOv5 network framework provided by the present invention.

[0022] Figure 4 It is a schematic diagram of an input image provided by the present invention.

[0023] Figure 5 It is a backbone network characteristic graph provided by the present invention.

[0024] Figure 6 It is a schematic diagram of a final feature map output provided by the present invention.

[0025] Figure 7 This is a hardware framework diagram provided by the present invention.

[0026] Figure 8 It is a software framework diagram provided by the present invention.

[0027] Fig. 9 The present invention provides a startup interface diagram.

[0028] Fig.10 The present invention provides a system setting password input interface diagram.

[0029] Fig.11 This is a system setting interface diagram provided by the present invention.

[0030] Fig.12 This is an operation interface diagram provided by the present invention.

[0031] Fig.13 The present invention provides a drop alarm interface diagram.

[0032] Fig.14 This is another drop alarm interface diagram provided by the present invention.

[0033] Fig.15The present invention provides a schematic diagram for explaining the format of an alarm text message content.

[0034] Fig.16 This is a schematic diagram of an abnormal prompt provided by the present invention.

[0035] Fig.17 This is a schematic diagram of falling object information provided by the present invention.

[0036] Fig.18 It is a schematic diagram of experimental information provided by the present invention. DETAILED DESCRIPTION

[0037] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is clearly and completely described below in conjunction with specific embodiments.

[0038] The embodiment of the present invention provides an intelligent identification device for typical block-shaped wear debris of a rail grinding train, wherein the length and width of the typical block-shaped wear debris are greater than 20 mm×20 mm; the rail grinding train is provided with N sections, which include a locomotive and a grinding operation vehicle, and the locomotive is provided with two sections, which are respectively located at the two ends of the train; The intelligent recognition device includes a visible light camera, a light curtain sensor group and an integrated controller, wherein two visible light cameras are provided, respectively located on two vehicle heads; N light curtain sensor groups are provided, respectively located on each of the vehicle heads and the grinding operation vehicle, and are provided below the fire curtain; two integrated controllers are provided, respectively located in the cabs of the two vehicle heads; each integrated controller is connected to all visible light cameras and light curtain sensor groups; During the operation of the train, the front locomotive along the train running direction is used as the train locomotive, and the integrated controller on it is turned on; the locomotive at the rear end is used as the grinding operation locomotive, and the integrated controller on it is turned off; When the train is in the first operating condition (low-speed condition, speed is 10~15km / h), the visible light cameras located on the two locomotive heads will send the taken pictures to the integrated controller in the turned-on state to identify the block-like wear chips; when in the second operating condition (high-speed condition, speed is greater than 15km / h), the light curtain sensor groups on all grinding operation vehicles will send the obtained signals to the integrated controller in the turned-on state to identify the block-like wear chips.

[0039] Taking the GMC-96B grinding car as an example, it includes locomotives (B1 and B2) at both ends. When B1 is the lead car, the arrangement order is: B1, C1, A, C2, C3, C4, B2; when B2 is the lead car, the arrangement order is: B2, C4, C3, C2, A, C1, B1; among them, A is the power car.

[0040] (1) Install visible light cameras on the B1 and B2 vehicles.

[0041] (2) A light curtain sensor group is arranged on each vehicle. However, when B1 is the vehicle head, C1, A, C2, C3, C4, and B2 are all used as grinding operation vehicles; when B2 is the vehicle head, C4, C3, C2, A, C1, and B1 are all used as grinding operation vehicles.

[0042] Therefore, no matter B1 or B2 is the vehicle head, the rest are grinding operation vehicles. In this example, there are 6 grinding operation vehicles in total, and each vehicle is equipped with a light curtain sensor group. When an object falls on the vehicle, the light curtain sensor group gives a switch signal (high level).

[0043] An embodiment of the present invention provides an intelligent identification device for typical blocky wear chips of rail grinding trains. Taking into account the situation that blocky wear chips adsorbed on the inner surface of the fire curtain may fall and affect the safety of equipment, the present invention sets a light curtain sensor group under the fire curtain to detect the area under the waterproof curtain to realize the identification of blocky wear chips, so as to timely alarm and process when blocky wear chips are identified, which can avoid affecting the safety of line cables and equipment, causing damage to communication and signal equipment on the line, damaging the bottom equipment and body of the train, and causing traffic accidents.

[0044] In addition, the present invention can realize the identification of blocky wear chips under low-speed conditions by setting up a visible light camera and an integrated controller; it can realize the identification of blocky wear chips under high-speed conditions by setting up a light curtain sensor group and an integrated controller, thereby realizing running and operation within the full speed range, being suitable for all working conditions, and improving the scope of application.

[0045] Hereinafter, the present invention specifically describes how to identify blocky wear debris through pictures taken by a light curtain sensor group and a visible light camera.

[0046] 1. Light curtain sensor group In a possible implementation, each of the light curtain sensor groups includes two laser opposing light curtain sensors, and each of the laser opposing sensors includes a light projector and a receiver that are arranged opposite to each other; Each of the light projectors is composed of a plurality of infrared light emitting tubes arranged at intervals, and each of the receivers is composed of a plurality of infrared receiving tubes arranged at intervals; Each of the laser beam sensors uses a synchronous scanning method to identify the blocked channel, so that the size of the object under the fire curtain can be obtained through two laser beam light curtain sensors to realize the identification of the blocky wear debris.

[0047] Light curtain sensor is a new type of sensor that has emerged in recent years. It consists of a projector and a receiver. Figure 1As shown. Multiple infrared light-emitting tubes are arranged in a straight line with equal intervals on the projector circuit board to emit infrared light, and the same receiving tubes are arranged on the receiver to receive infrared light; the light emitted by the projector is directly emitted to the receiver, forming an invisible light curtain system. The target object placed between the projector and the receiver will block part of the light so that it cannot reach the corresponding receiver device.

[0048] The light curtain system uses synchronous scanning to identify the blocked channel. First, a projector channel sends out a light pulse and the corresponding receiver searches for the pulse at the same time. When it finds it, it immediately completes the scan of one channel; then it moves to the next channel until all optical channels have completed the first scan. The cycle from the first optical channel entering the scan to the completion of a scan of all optical channels is called a scan cycle.

[0049] The length of the scanning cycle depends on the number of optical axes of the light curtain device. During a scanning cycle when the light curtain system is working normally, the internal processor of the system records which optical channels are light-transmitted and which optical channels are light-shielded, and further decodes them. It responds to the repeater with some processing signals required by the user, such as analog voltage, current, serial communication data flow, transistor NPN / PNP signal changes, etc.; then the length, width, height, diameter, position, coordinates, offset and other information of the measured object can be calculated.

[0050] Therefore, for any light curtain sensor group, the area below the fire curtain of the grinding vehicle corresponding to the light curtain sensor group can be detected to determine whether there are blocky grinding chips falling from the fire curtain.

[0051] In addition, in order to detect objects larger than 20 mm×20 mm, the present invention sets the distance between the incident light axes to 10 mm (that is, the distance between any two adjacent infrared light-emitting tubes and the infrared receiving tube is 10 mm) so that the detection accuracy reaches 10 mm. In addition, in order to achieve complete coverage of the area below the fire curtain of the work vehicle, such as Figure 2 As shown, in the present invention, when the light curtain sensor is arranged longitudinally (along the direction of train operation, i.e., the left-right direction in the present invention), the sensor detection width is not less than the lateral width of the fire curtain of the grinding trolley; when the light curtain sensor is arranged transversely (i.e., the front-to-back direction in the present invention), the sensor detection width is not less than the longitudinal width of the fire curtain of the grinding trolley. By way of example, taking the aforementioned GMC-96B grinding trolley as an example, the longitudinal (left-right) detection range in the present invention is 3800~4200 mm, and the lateral (front-to-back) detection range is 2800~3200 mm.

[0052] In addition, the falling time of a 20 mm object is calculated based on the vertical falling gravity acceleration. In the present invention, the system response speed is designed to be 6.5 ms.

[0053] Therefore, the present invention realizes a high-precision, fast detection sensor with a longitudinal detection range of about 4000 mm and a lateral detection range of about 3000 mm.

[0054] Furthermore, in order to realize the installation of the light curtain sensor, vertical support hangers are installed on the four sides of each car, and each of the support hangers extends downward to the bottom of the fire curtain; The bottom of the two support and suspension frames at the front side is provided with a first cross beam, and the bottom of the two support and suspension frames at the rear side is provided with a second cross beam; the bottom of the two support and suspension frames at the front side is provided with a first longitudinal beam, and the bottom of the two support and suspension frames at the rear side is provided with a second longitudinal beam; The first horizontal beam is provided with a light projector I or a receiver I, and the second horizontal beam is provided with a receiver I or a light projector I; the first longitudinal beam is provided with a light projector II or a receiver II, and the second longitudinal beam is provided with a receiver II or a light projector II.

[0055] That is, when the first horizontal beam is provided with a projector I, the second horizontal beam is provided with a receiver I; when the first horizontal beam is provided with a receiver I, the second horizontal beam is provided with a projector I; similarly, when the first longitudinal beam is provided with a projector II, the second longitudinal beam is provided with a receiver II; when the first longitudinal beam is provided with a receiver II, the second longitudinal beam is provided with a projector II.

[0056] Therefore, the integrated controller can identify the block-shaped wear debris through the switch signal (high level) sent by the receiver I and / or the receiver II.

[0057] 2. Visible light camera In a possible implementation, the integrated controller in the turned-on state identifies the block-shaped wear debris based on the YOLOv5 algorithm after obtaining the picture taken by the visible light camera; The YOLOv5 network structure includes: input end, backbone network, neck and output end. The input end is used to perform data enhancement and adaptive anchor box calculation operations on the input image; the backbone network includes Foucs, Conv, C3, SPP modules for image feature extraction; the neck combines the feature pyramid network and the path aggregation network to enhance the fusion ability of feature layers carrying different information; the output end is the prediction part of the network, which uses the extracted features to predict the type and location of the target. Its network framework is as follows Figure 3 shown.

[0058] In the backbone network of this model, the Focus module is used to reduce the amount of calculation; the Conv module performs convolution, batch normalization and Silu function activation operations on the input feature map; In the C3 module, the BottleneckCSP module is replaced by the C3 module; the convolution after the residual output is removed, and the activation function in the convolution module is changed to the Silu function; the image is divided into two branches, one is passed through the convolution module and multiple BottleNeck stacks, and the other is passed through a convolution module, and finally the two branches are channel fused; The SPP module uses 5×5, 9×9, and 13×13 pooling filters to perform maximum pooling on the input image, then performs channel fusion of the pooled image with the original input image, and finally performs a convolution operation.

[0059] To ensure the accuracy of the model, the present invention verifies the model as follows: Feature extraction is a very important part of the detection process of blocky wear debris, and has a direct impact on the subsequent detection accuracy. Therefore, this section visualizes the feature graphs of the convolution and pooling processes in the YOLOv5 network training process to explore what useful feature information the network model has learned in each layer. It also lays a good foundation for improving and optimizing the network after discovering problems later. The input image is as follows: Figure 4 shown.

[0060] The output feature map is divided into two parts: the feature map in the backbone network and the feature map in the detection stage. Among them, the feature map output in the backbone network of YOLOv5 is as follows: Figure 5 shown.

[0061] As can be seen from the figure, the feature map output by Focus will lose some original information, and only the rough edges and contours of blocky wear chips can be seen. The feature maps output by Conv, C3-1, Conv, C3-3 and other modules show that the image features of blocky wear chips are becoming increasingly blurred, indicating that the features extracted by the model have gradually transformed from shallow information rich in geometric information to deep features richer in semantic information.

[0062] The characteristic graphs output by the three output terminals in the detection stage are as follows: Figure 6 shown.

[0063] The Detect detection end outputs feature maps with three resolutions: 80×80, 40×40, and 20×20. The feature map with a resolution of 80×80 has more edge and contour information of shallow networks and is suitable for small target detection. The feature map with a resolution of 20×20 contains richer semantic information and is suitable for large target detection. Figure 3 It can be seen from the feature map that the 20×20 feature map is closer to the edge contour of the block wear debris in the original input image. Therefore, it can be known that the detection result output by YOLOv5 through the large target detection window. Through the above feature map visualization experiment, we can have a more intuitive understanding of the detection process of YOLOv5.

[0064] To sum up, it is reasonable and feasible to use the YOLOv5 algorithm for image recognition and detection.

[0065] 3. Integrated controller In a possible implementation, each of the integrated controllers includes a shell, an integrated control module and a power supply module are arranged inside the shell, and a touch display screen and an alarm buzzer are arranged outside the shell; The control module is connected to the power supply module, the touch screen and the alarm buzzer, the power supply module is connected to the touch screen and the alarm buzzer, and the control module is also connected to all visible light cameras and light curtain sensor groups. The hardware framework diagram of the integrated control system is as follows: Figure 7 As shown in the software framework diagram Figure 8 shown.

[0066] Depend on Figure 8 It can be seen that the integrated controller performs the following steps: (1) Start the block wear debris monitoring program: After the system is powered on, the rail grinding vehicle block-shaped wear debris shedding monitoring system program is automatically started and displayed on the touch screen. The staff can set the operating parameters and system parameters through the touch screen, wherein the operating parameters include the mobile phone number for receiving alarm SMS; Specifically, a. Operation parameter settings: In the startup interface, the staff can set the operating parameters of "vehicle scheduling", "uplink and downlink", and "mobile phone number for receiving alarm SMS". After the settings are completed, click the <Run> button to enter the running state; Fig. 9 As shown; Note: After the running parameters are set, the system automatically saves them as the default parameters for the next startup; b. System parameter settings: In the startup interface, click the <System Settings> button, and the system pops up a password dialog box (such as Fig.10 After entering the password, enter the system parameter setting interface (as shown in Fig.11 As shown), you can set the "vehicle number", "working device-GYK distance", "speed" and "kilometer mark"; Note: The system parameter setting password is only open to the system administrator; after the system parameters are set, the system automatically saves them as the default parameters for the next startup.

[0067] (2) The operation interface displays: In response to the staff clicking the run button, the running state is entered (such as Fig.12As shown in the figure), the touch screen jumps to the operation interface. The green light flashing on the operation interface indicates that the system is running normally, and the red light flashing indicates that the system is faulty and needs to contact professionals for repair. At this time, the blocky wear debris is identified through the pictures taken by the visible light camera or the signals sent by all light curtain sensor groups through the integrated control module; (3) Falling object information alarm When blocky wear debris is identified, the alarm buzzer sounds an alarm and the touch screen pops up an alarm interface (such as Fig.13 As shown, and send an alarm text message to the set mobile phone number, the alarm text message includes: date, time, vehicle number, vehicle number, kilometer mark, and a reminder message of falling objects (such as Fig.14 and Fig.15 Detailed description is as follows: The integrated controller also executes SMS module exception handling steps, which include: (4) Perform self-check on the SMS module during operation. When it is detected that the 4G module has no phone card installed or the phone card is abnormal, the abnormal information of the SMS module is highlighted and marked in the operation interface (such as Fig.16 shown). Furthermore, in the running state, the data query button clicked by the staff can jump to the log display interface; The log display interface includes: a button to display the current day's log, a button to display the latest log, a button to display the current month's log, and a button to download all logs from a USB flash drive (e.g. Fig.17 As shown); when receiving a click instruction to download all logs to the USB flash drive button, all logs are downloaded to the inserted USB flash drive.

[0068] In order to verify the accuracy of the intelligent identification device for typical block-shaped grinding debris of rail grinding trains provided by the present invention, the present invention uses a 20 mm×20 mm spherical object (which is more difficult to identify than a 20 mm×20 mm block in appearance) as a test object for simulation testing. The results show that: within the range of the light curtain sensor, its recognition accuracy can reach more than 98%. The text message sent to the ground personnel is as follows Fig.18 As shown, the SMS sending function is normal, and it can be seen from the time tag that rapid detection and identification of continuous falling objects can be completed.

Claims

1. An intelligent identification device for typical blocky grinding debris of rail grinding trains, characterized in that: The length and width of the typical block-shaped grinding chips are greater than 20 mm×20 mm; the rail grinding train is provided with N sections, which include a locomotive and a grinding operation vehicle, and the locomotive is provided with two, respectively located at the two ends of the train; The intelligent recognition device includes a visible light camera, a light curtain sensor group and an integrated controller, wherein two visible light cameras are provided, respectively located on the two vehicle heads; N light curtain sensor groups are provided, respectively located on each of the vehicle heads and the grinding operation vehicle, and are provided below the fire curtain; two integrated controllers are provided, respectively located in the cabs of the two vehicle heads; each integrated controller is connected to all visible light cameras and light curtain sensor groups; During the operation of the train, the front locomotive along the train running direction is used as the train locomotive, and the integrated controller on it is turned on; the locomotive at the rear end is used as the grinding operation locomotive, and the integrated controller on it is turned off; When the train is in the first operating condition, the visible light cameras located on the two locomotive heads will send the captured pictures to the integrated controller in the turned-on state to identify the block-like wear chips; when it is in the second operating condition, the light curtain sensor groups on all grinding work vehicles will send the obtained signals to the integrated controller in the turned-on state to identify the block-like wear chips.

2. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 1 is characterized in that: Each of the light curtain sensor groups includes two laser beam light curtain sensors, and each of the laser beam sensors includes a projector and a receiver that are arranged opposite to each other; Each of the light projectors is composed of a plurality of infrared light emitting tubes arranged at intervals, and each of the receivers is composed of a plurality of infrared receiving tubes arranged at intervals; Each of the laser beam sensors uses a synchronous scanning method to identify the blocked channel, so that the size of the object under the fire curtain can be obtained through two laser beam light curtain sensors to realize the identification of the blocky wear debris.

3. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 2 is characterized in that: Each car is provided with vertical support hangers on four sides, and each of the support hangers extends downward to below the fire curtain; The bottom of the two support and suspension frames at the front side is provided with a first cross beam, and the bottom of the two support and suspension frames at the rear side is provided with a second cross beam; the bottom of the two support and suspension frames at the front side is provided with a first longitudinal beam, and the bottom of the two support and suspension frames at the rear side is provided with a second longitudinal beam; The first horizontal beam is provided with a light projector I or a receiver I, and the second horizontal beam is provided with a receiver I or a light projector I; the first longitudinal beam is provided with a light projector II or a receiver II, and the second longitudinal beam is provided with a receiver II or a light projector II.

4. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 1 is characterized in that: The integrated controller in the turned-on state identifies the block-shaped wear debris based on the YOLOv5 algorithm after obtaining the picture taken by the visible light camera; The YOLOv5 network structure includes: input end, backbone network, neck and output end. The input end is used to perform data enhancement and adaptive anchor box calculation operations on the input image; The backbone network includes Foucs, Conv, C3, and SPP modules for image feature extraction; the neck combines the feature pyramid network and the path aggregation network to enhance the fusion capability of feature layers carrying different information; the output end serves as the prediction part of the network, which uses the extracted features to predict the type and location of the target.

5. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 4 is characterized in that: In the backbone network, the Focus module is used to reduce the amount of calculation; the Conv module performs convolution, batch normalization and Silu function activation operations on the input feature map; In the C3 module, the BottleneckCSP module is replaced by the C3 module; the convolution after the residual output is removed, and the activation function in the convolution module is changed to the Silu function; the image is divided into two branches, one is passed through the convolution module and multiple BottleNeck stacks, and the other is passed through a convolution module, and finally the two branches are channel fused; The SPP module uses 5×5, 9×9, and 13×13 pooling filters to perform maximum pooling on the input image, then performs channel fusion of the pooled image with the original input image, and finally performs a convolution operation.

6. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 1 is characterized in that: Each of the integrated controllers comprises a housing, an integrated control module and a power supply module are arranged inside the housing, and a touch display screen and an alarm buzzer are arranged outside the housing; The control module is connected to the power supply module, the touch display screen and the alarm buzzer, the power supply module is connected to the touch display screen and the alarm buzzer, and the control module is also connected to all visible light cameras and the light curtain sensor group.

7. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 6 is characterized in that: The integrated controller performs the following steps: (1) Start the block wear debris monitoring program After the system is powered on, the rail grinding vehicle block-shaped wear debris shedding monitoring system program is automatically started and displayed on the touch screen. The staff can set the operating parameters and system parameters through the touch screen, wherein the operating parameters include the mobile phone number for receiving alarm SMS; (2) Operation interface display In response to the staff clicking the run button, the system enters the run state and the touch screen jumps to the run interface. The green light flashing on the run interface indicates that the system is running normally, and the red light flashing indicates that the system has a fault and needs to contact a professional for repair. At this time, the integrated control module identifies the blocky wear debris based on the picture taken by the visible light camera or the signal sent by all the light curtain sensor groups. (3) Falling object information alarm When block-shaped wear debris is confirmed to be identified, the alarm buzzer sounds an alarm, the touch display pops up an alarm interface, and sends an alarm text message to the set mobile phone number. The alarm text message includes: date, time, vehicle number, vehicle number, kilometer mark, and a reminder of falling objects.

8. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 7 is characterized in that: The integrated controller will also execute the SMS module exception handling steps, which include: performing self-inspection on the SMS module in the running state, and when it is detected that the 4G module has no phone card installed or the phone card is abnormal, the SMS module abnormal information is highlighted and marked in the running interface.

9. The intelligent identification device for typical block-shaped wear debris of rail grinding train according to claim 7, characterized in that: In the running state, the data query button can be clicked by the staff to jump to the log display interface; Among them, the log display interface displays: a display button for the day's log, a display button for the latest log, a display button for the month's log, and a button for downloading all logs to a USB flash drive; when a click instruction for the button for downloading all logs to a USB flash drive is received, all logs are downloaded to the inserted USB flash drive.

Images