Heavy vehicle axle load identification method based on bridge deflection influence line distribution

Abstract

The invention discloses a heavy vehicle axle load identification method based on bridge deflection influence line distribution, and the method comprises the following steps: erecting a monitoring device for vehicle parameter identification on a bridge floor, and arranging a structure displacement detection device at the midspan of a bridge to identify the midspan deflection of the bridge; carrying out on-site vehicle running test to obtain influence lines of small heavy vehicles with different vehicle types and vehicle weights on the midspan deflection of the bridge; performing simulation amplification according to the influence line data of the small heavy vehicle; constructing a heavy vehicle axle load identification model, and performing model training through the obtained influence line data corresponding to the heavy vehicle; and inputting the data of the target heavy vehicle during bridge crossing into the heavy vehicle axle load identification model to obtain axle load data of the target heavy vehicle. The method has the advantages that the vehicle axle load is estimated through on-bridge visual recognition and under-bridge deflection calculation, and the influence of device installation on bridge floor traffic can be avoided; and meanwhile, the recyclable structural displacement detection device can effectively reduce the investment cost of equipment.

Description

Technical Field

[0001] The present application relates to the field of bridge technology, and in particular to a method for identifying axle weight of a heavy vehicle based on the distribution of bridge deflection influence lines. Background Art

[0002] The dynamic weighing system of vehicle load can be divided into the road dynamic weighing system (PWIM) and the bridge dynamic weighing system (BWIM). The bridge dynamic weighing system usually installs sensors at the bottom of the bridge to collect structural responses (such as strain gauges, accelerometers, fiber optic sensors, etc.), and uses relevant identification algorithms (such as Moses algorithm) to calculate the axle weight of the moving vehicle.

[0003] In order to ensure the accuracy and coverage of the measurement, it is usually necessary to install multiple sensors at multiple locations on the bridge, which increases the complexity and cost of the system. Especially on large bridges, a large number of sensors and cable connections are required, which may cause wiring and power management problems. Over time, the performance of traditional contact sensors will degrade due to long-term exposure to the external environment, resulting in measurement errors or complete failure; and the construction of the system will affect the normal operation of traffic.

[0004] At the same time, at the algorithm level, the vehicle load information that can be obtained only from the bridge deflection response is very limited, which makes the overweight vehicle load and the corresponding bridge deflection data in actual situations less, and cannot meet the requirements of deep learning model reasoning applications. In addition, deep learning model reasoning is easily affected by noise and abnormal data in the input data, causing large errors in the accuracy of the results. Summary of the invention

[0005] One of the objectives of the present application is to provide a method for identifying the axle weight of a heavy vehicle based on the distribution of bridge deflection influence lines, which can solve at least one of the defects in the above-mentioned background technology.

[0006] In order to achieve at least one of the above purposes, the technical solution adopted by the present application is: a method for identifying the axle weight of heavy vehicles based on the distribution of bridge deflection influence lines, comprising the following steps: S100: a monitoring device for identifying vehicle parameters is set up on the bridge deck, and a structural displacement detection device is arranged at the mid-span of the bridge to identify the mid-span deflection of the bridge; S200: Use small and heavy vehicles of different models and weights to conduct on-site running test on the bridge to obtain the influence lines of small and heavy vehicles of different models and weights on the mid-span deflection of the bridge; S300: performing simulation amplification according to the influence line data of the small heavy vehicle to obtain influence line simulation data of the mid-span deflection of the bridge corresponding to the large heavy vehicle; S400: constructing a heavy vehicle axle weight recognition model and performing model training through the influence line data corresponding to all heavy vehicles obtained; S500: Substitute the model of the target heavy vehicle when crossing the bridge and the influence line data of the generated bridge mid-span deflection into the trained heavy vehicle axle weight identification model, and then invert to obtain the axle weight data of the target heavy vehicle.

[0007] Preferably, the specific process of performing vehicle parameter identification by the monitoring device in step S100 is as follows: S110: Tracking and photographing images of vehicles traveling on the bridge deck by using a monitoring device; S120: Acquire the type of the target vehicle according to the captured image, and calculate the speed of the target vehicle according to the initial position and the calibrated position defined on the bridge deck; S130: According to the estimated speed of the target vehicle, intercepting an image of the target vehicle at at least one position in the original image data to identify the number of axles and wheelbase of the target vehicle; S140: Construct a parameter set of the target vehicle regarding the license plate, vehicle model, number of axles, and wheelbase.

[0008] Preferably, the monitoring device includes a traffic camera and a laser radar; the monitoring device uses the traffic camera to track and shoot the vehicle with camera images during the day; and the monitoring device uses the traffic camera and the laser radar to coordinately track and shoot the vehicle with camera images and point cloud images at night.

[0009] Preferably, the structural displacement detection device includes a target and a corner reflector installed at the mid-span position of the bridge main beam, and an industrial camera and a millimeter-wave radar installed at the pier position; the industrial camera monitors the displacement of the target to obtain the mid-span deflection of the bridge, and the millimeter-wave radar monitors the phase difference of the electromagnetic wave signal of the corner reflector to obtain the mid-span deflection of the bridge.

[0010] Preferably, when performing step S200, a plurality of structural displacement detection devices for monitoring the mid-span deflection of the bridge are provided and are arranged in sequence according to the main beams corresponding to the bridge; the lateral proportional coefficient η of the deflection under the action of different lanes and vehicle weights is calculated according to the mid-span deflection of the main beams of each lane obtained from an on-site sports car test; after completing step S200, some of the structural displacement detection devices are removed, and based on the mid-span deflection of the corresponding main beams obtained by the remaining structural displacement detection devices, the mid-span deflection of the remaining main beams is calculated by using the obtained lateral proportional coefficient η of the deflection.

[0011] Preferably, the transverse proportional coefficient η of the deflection corresponding to the i-th main beam is i The calculation formula is: ; Among them, ω i represents the mid-span deflection of the ith main beam, Represents the sum of the mid-span deflections of the n main beams corresponding to the bridge.

[0012] Preferably, when the structural displacement detection device is disassembled after completing step S200, at least two structural displacement detection devices remain. The process of determining the transverse connection stiffness between the main beams by using the transverse proportional coefficients corresponding to the remaining structural displacement detection devices is as follows: Set the transverse connection stiffness determination formula: (ω x / ω y ) = (η x / η y ); if the lateral connection stiffness judgment formula is established, the lateral connection stiffness between the main beams corresponding to the bridge meets the requirements, otherwise it will be determined that the lateral connection stiffness between the main beams corresponding to the bridge does not meet the requirements, and the lateral proportional coefficient of the deflection of each main beam will be recalibrated through a single vehicle model; where, ω x and ω y Respectively represent the mid-span deflection of the main beam corresponding to the remaining two structural displacement detection devices, η x and η y They respectively represent the lateral proportional coefficients of the deflection of the main beams corresponding to the remaining two structural displacement detection devices.

[0013] Preferably, step S300 includes the following process: S310: Construct a finite element model of the bridge; S320: Using the measured deflection influence line data of the small heavy vehicle running test and the corresponding vehicle information, the finite element model of the bridge is corrected for stiffness; S330: Apply the load corresponding to the heavy vehicle to the modified bridge finite element model, and then obtain the influence line simulation data of the mid-span deflection of the bridge corresponding to the heavy vehicle loading.

[0014] Preferably, step S400 includes the following specific processes: S410: Build a deep learning model architecture for the heavy vehicle axle weight recognition model and select the root mean square error as the model data-driven loss function L data ; S420: embedding the bridge finite element model with completed stiffness correction into the deep learning model training process; S430: Input given bridge deflection influence line data into the deep learning model to generate corresponding axle weight vector and wheelbase information; S440: converting the predicted axle weight vector and wheelbase information into loads of a bridge finite element model and performing structural simulation to obtain corresponding bridge deflection influence line simulation data; S450: The error between the given bridge deflection influence line data and the bridge deflection influence line simulation data is used as the physical information loss function L of the deep learning model phy ; S460: Loss function L data and L phy As the optimization object of the deep learning model, parameter backpropagation is performed to iterate the model parameters.

[0015] Preferably, step S500 includes the following specific processes: S510: When the monitoring device detects that a target heavy vehicle is on the bridge, the displacement detection device of the underbridge structure is triggered to calculate the mid-span deflection of the bridge; S520: using signal processing technology to eliminate structural dynamic response and environmental noise components in the measured deflection influence line data, and obtaining the bridge deflection influence line data of the measured point; S530: estimating the bridge deflection influence lines of other points in the mid-span of the bridge according to the calculated deflection lateral proportionality coefficient and the bridge deflection influence line data of the measured points; S540: After the target heavy vehicle passes the bridge, the bridge deflection influence line data of all points are input into the vehicle axle weight identification model to estimate the axle weight of each axle.

[0016] Compared with the prior art, the beneficial effects of this application are: Through the heavy vehicle axle weight identification method based on the distribution of bridge deflection influence lines, the vehicle axle weight is estimated by using visual recognition on the bridge and deflection calculation under the bridge; compared with the traditional method, the impact of device installation on bridge traffic can be avoided, and the recyclable structural displacement detection device can effectively reduce the investment cost of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The figure is a schematic diagram of the overall workflow of this application.

[0018] Figure 2 This is a schematic diagram of the construction process of the vehicle axle weight identification model of the present application. DETAILED DESCRIPTION

[0019] Below, the present application is further described in conjunction with specific implementation methods. It should be noted that in the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms should not be understood as necessarily being directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.

[0020] In the description of the present application, it should be noted that directional words, such as the terms "center", "lateral", "longitudinal", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", etc., indicating directions and positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of narrating the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and cannot be understood as limiting the specific scope of protection of the present application.

[0021] It should be noted that the terms "first", "second", etc. in the description and claims of the present application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0022] In this application, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be connected, detachably connected, or integrated; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0023] In the present application, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0024] The terms "including" and "having" and any variations thereof in the specification and claims of the present application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0025] One of the preferred embodiments of the present application is as follows: Figure 1 As shown, a method for identifying axle weight of heavy vehicles based on the distribution of bridge deflection influence lines includes the following steps: S100: A monitoring device for identifying vehicle parameters is set up on the bridge deck, and a structural displacement detection device is arranged at the mid-span of the bridge to identify the mid-span deflection of the bridge.

[0026] S200: On-site running test was conducted on the bridge using small and heavy vehicles of different models and weights to obtain the influence lines of small and heavy vehicles of different models and weights on the mid-span deflection of the bridge.

[0027] S300: Perform simulation amplification based on the influence line data of the small heavy vehicle to obtain the influence line simulation data of the mid-span deflection of the bridge corresponding to the large heavy vehicle.

[0028] S400: construct a heavy vehicle axle weight recognition model and perform model training using the influence line data corresponding to all heavy vehicles obtained.

[0029] S500: Substitute the model of the target heavy vehicle when crossing the bridge and the influence line data of the generated bridge mid-span deflection into the trained heavy vehicle axle weight identification model, and then invert to obtain the axle weight data of the target heavy vehicle.

[0030] It can be understood that, firstly, in the technical solution of the present application, the monitoring device is generally installed on the side of the bridge, and the structural displacement detection device is installed below the bridge; therefore, the technical solution of the present application will basically not affect the normal traffic of the bridge during the installation and construction of the equipment.

[0031] Secondly, the traditional method generally uses the finite element simulation method for the mid-span deflection of the bridge for different vehicles. Although this method is efficient, the authenticity of the data obtained is poor, especially the simulation data of large and heavy vehicles. In the technical solution of the present application, the measured data of the mid-span deflection of the bridge is obtained by the sports car test on the bridge site, which can effectively improve the data accuracy compared with the traditional method. Of course, considering that the number of large and heavy vehicles passing through the bridge per unit time is small, the obtained data is difficult to meet the subsequent training of the axle weight identification model of large and heavy vehicles; and large and heavy vehicles may be special vehicles, which require professional drivers and corresponding escort personnel, etc., which do not meet the actual sports car test requirements. Therefore, the technical solution of the present application obtains the mid-span deflection data of the bridge corresponding to the small heavy vehicle, and then simulates and amplifies the obtained measured data to obtain the influence line simulation data of the mid-span deflection of the bridge corresponding to the large heavy vehicle, and finally uses the measured deflection influence line data of the small heavy vehicle and the deflection influence line simulation data of the large heavy vehicle for the training of the axle weight identification model of the heavy vehicle. Compared with the traditional method, while ensuring the accuracy of the model training, the speed of data acquisition can also be effectively improved.

[0032] It should be known that vehicles weighing more than 15 tons can generally be defined as heavy vehicles, and the classification of large heavy vehicles and small heavy vehicles can generally be defined according to the load capacity of the bridge; for example, heavy vehicles weighing less than 40 tons can be defined as small heavy vehicles, and vehicles weighing more than 40 tons can be defined as large heavy vehicles.

[0033] In this embodiment, Figure 2 As shown, the vehicle parameters identified by the monitoring device in step S100 include the vehicle model, license plate, wheels and speed. The number of axles and wheelbase corresponding to the vehicle can be calculated according to the position and number of the vehicle wheels. There are many specific structures of the monitoring device for identifying vehicle parameters. For the sake of easy understanding, one of the structures will be described in detail below.

[0034] Specifically, Figure 2As shown, the monitoring device includes a traffic camera. The specific structure and working principle of the traffic camera are well-known technologies to those skilled in the art, so they will not be elaborated in detail here. For example, a 25-frame-rate 8-megapixel high-definition zoom network camera can be used. The traffic camera can be set up on the side of the bridge by a pole or a vertical pole and separated from the upper bridge position of the vehicle by a certain distance. The traffic camera can take pictures of the vehicle, and compress the image data and transmit it to the edge computing terminal through a wired network for analysis and processing to obtain data such as the vehicle model, license plate, number and position of wheels corresponding to the vehicle. Considering that the light at night is dark, the detection accuracy of the traffic camera will be reduced; therefore, the monitoring device of this embodiment also includes a laser radar, and the specific structure and working principle of the laser radar are well-known technologies to those skilled in the art, so they will not be elaborated in detail here. The monitoring device tracks and shoots the vehicle with camera images through the traffic camera during the day; the monitoring device tracks and shoots the vehicle with camera images and point cloud images in coordination with the traffic camera and the laser radar at night.

[0035] In this embodiment, the specific process of vehicle parameter identification based on the above monitoring device is as follows: S110: Tracking and photographing images of vehicles traveling on the bridge deck by means of a monitoring device.

[0036] S120: Acquire the type of the target vehicle according to the captured image, and infer the speed of the target vehicle according to the initial position and the calibrated position defined on the bridge deck.

[0037] S130: According to the estimated speed of the target vehicle, an image of the target vehicle at least one position in the original image data is intercepted to identify the number of axles and the wheelbase of the target vehicle.

[0038] S140: Construct a parameter set of the target vehicle regarding the license plate, vehicle model, number of axles, and wheelbase.

[0039] Specifically, during the day, the type of target vehicle is first detected through the vehicle type recognition algorithm for the image data taken by the traffic camera, and the target vehicle trajectory is tracked to the calibration position for speed estimation; that is, the distance between the initial monitoring position and the calibration position marked on the bridge deck in the image data is known, and the speed of the target vehicle is estimated based on the time it takes for the target vehicle to move from the initial position to the calibration position. Then, the target vehicle image is intercepted in the original captured image using the inferred position boundary box, and then the wheel detection is performed based on the intercepted target vehicle image to identify the number of axles and wheelbase of the target vehicle, and the license plate position and characters of the target vehicle are also identified. At night, the traffic camera and the laser radar collect data at the same time, and the data collected by the laser radar and the data collected by the traffic camera are fused and superimposed, and then the above recognition process is performed through the fused data.

[0040] It should be known that, during the process of tracking and photographing the target vehicle by the traffic camera, the target vehicle may be blocked by other vehicles, which may interfere with the calculation of the wheelbase and number of axles of the target vehicle. Therefore, when identifying the wheels of the target vehicle, it is often necessary to capture image information at different positions during the vehicle's travel. If the entire travel process of the target vehicle is directly positioned and identified, the data recognition process may be relatively long. Therefore, in this embodiment, after estimating the speed of the target vehicle, the position of the target vehicle can be estimated according to the selected time, and then the image can be directly captured at the corresponding estimated position in the original image data, which can achieve rapid positioning and identification of the target vehicle to improve the calculation process of the wheelbase and number of axles.

[0041] In this embodiment, Figure 2 As shown, the structural displacement detection device for monitoring the mid-span deflection of a bridge includes a target installed at the mid-span position of the bridge, and an industrial camera installed on the pier cap beam on one side of the bridge. The specific structure and working principle of the industrial camera are well-known technologies to those skilled in the art, so they will not be elaborated in detail here; the selection of industrial cameras is based on right-hand accuracy, for example, the integer pixel accuracy is 0.1mm / pixel. When monitoring the mid-span deflection of a bridge, the industrial camera can monitor the target to perform displacement calculations, and then the mid-span deflection of the bridge can be obtained and the influence line can be drawn. The displacement calculation of the mid-span deflection of the bridge is also carried out by transmitting the image data collected by the industrial camera to the edge computing terminal through a wired network for analysis and processing.

[0042] Considering that industrial camera imaging may have uneven lighting or overexposure, and foggy weather will cause the image clarity of the target to decrease. Therefore, the structural displacement detection device of this embodiment also includes a corner reflector installed in the middle of the bridge span and a millimeter-wave radar installed on the pier cap beam on the other side of the bridge on the basis of the above structure. The specific structure and working principle of the millimeter-wave radar are well-known technologies to those skilled in the art, so they will not be elaborated in detail here. The millimeter-wave radar can emit electromagnetic waves to the corner reflector, and measure the deflection of the bridge span according to the phase difference of the received electromagnetic wave signal. That is, in bad weather, the data collected by the industrial camera can be corrected by the data collected by the millimeter-wave radar to maintain the measurement accuracy of the deflection of the bridge span.

[0043] It should be known that the deflection change of the bridge is mainly reflected in the deflection deformation of the main beam, and a single bridge generally corresponds to multiple main beams. Therefore, the detection of the mid-span deflection of the bridge is the detection of the mid-span deflection of the main beam; then in order to ensure the measurement accuracy, a target can be installed in the span of each main beam, and each target corresponds to an industrial camera. Due to the high cost of millimeter-wave radar, it is difficult to achieve full coverage, so in this embodiment, it is preferred to install a corner reflector and a corresponding millimeter-wave radar on the middle main beam.

[0044] It is understandable that the deck of a bridge is generally divided into multiple lanes. When a vehicle is driving in different lanes, the deflection response of each main beam is different; therefore, when performing step S200, it is necessary to perform a sports car test on all lanes. The sports car test can use the natural driving data of social vehicles, or send a designated vehicle to perform a sports car test during a period of less traffic; considering the efficiency of the test and the stability of data collection, this embodiment preferably uses a designated vehicle to perform a sports car test. The selection of the designated vehicle can be set according to actual needs. For example, the model of the designated vehicle can be selected from 2-axle vehicles, 3-axle vehicles, 4-axle vehicles, 5-axle vehicles and 6-axle vehicles, the vehicle weight range can be selected from 10 tons, 20 tons, 30 tons and 40 tons, the vehicle speed can be selected from 30km / h, 40km / h and 50km / h, and the actual driving lane position of the vehicle when going on the bridge is evenly distributed.

[0045] After completing the selection of the designated vehicle, the vehicle information of the above-mentioned different working conditions and the measured bridge mid-span deflection data are obtained according to the on-site sports car test; the deflection and vehicle axle weight information are matched, and the bridge deflection influence line is extracted using signal processing technology (such as variational mode decomposition and wavelet transformation, etc.), and then the lateral proportional coefficient η of the deflection under different lanes and different vehicle weights is calculated based on the deflection curve data of different main beams; that is, the change of the mid-span deflection of the bridge is characterized by the lateral proportional coefficient η of the deflection.

[0046] Specifically, the lateral proportional coefficient η of the deflection corresponding to the i-th main beam is i The calculation formula is: ; Among them, ω i represents the mid-span deflection of the ith main beam, represents the sum of the mid-span deflections of the n main beams corresponding to the bridge. Then the lateral proportional coefficient η of the deflection caused by vehicles on any lane to each main beam is x It can be expressed in the form of a matrix, namely η x =[η x1 , η x2 , ..., η xn ]. After obtaining the corresponding deflection lateral proportional coefficient matrix, the signal is filtered and then the smooth curve data is output for the training of the heavy vehicle recognition model.

[0047] For ease of understanding, the following will be explained in detail through a specific example. Assuming that the target bridge has 5 main beams, corresponding to 3 lanes, the deflection lateral proportional coefficient matrix of each main beam corresponding to the 3 lanes is η1=[η 11 , η 12 , η 13 , η 14 , η 15 ],η2=[η 21 , η 22, η 23 , η 24 , η 25 ],η3=[η 31 , η 32 , η 33 , η 34 , η 35 ]. 11 For example, η 11 It represents a column vector, which represents the lateral proportional coefficient of the deflection of the vehicle at different driving positions in the first lane.

[0048] It is understandable that in the actual use of the bridge, within a short service life of the bridge, such as within 3-5 years, and under normal traffic conditions, the decrease in the transverse connection stiffness between the main beams in the multi-beam system is very small and can be ignored. After the bridge has been in service for a long time, such as 5-10 years, the decrease in the transverse connection stiffness between the main beams is large. At this time, in order to ensure the accurate prediction of the heavy vehicle identification model, the transverse proportional coefficient of the deflection of the main beam needs to be recalibrated. The specific calibration method can be to recalibrate the transverse proportional coefficient of the deflection of each main beam through a single vehicle model.

[0049] It should be known that in order to ensure the accuracy of the on-site sports car test, the present embodiment sets a corresponding industrial camera and target for each main beam; and after completing the on-site sports car test, for the subsequent monitoring of the bridge, it is only necessary to identify the deflection of some main beams to determine the reduction of the transverse connection stiffness of the main beams. Therefore, from the perspective of cost saving, after completing step S200, some structural displacement detection devices are removed, and the mid-span deflection of the corresponding main beam obtained by the remaining structural displacement detection devices is calculated by the obtained deflection transverse proportional coefficient η. For ease of understanding, a detailed description will be given below through specific examples.

[0050] Specifically, for example, a bridge has 5 main beams. After completing the on-site sports car test, 3 of the industrial cameras and the corresponding targets can be recovered, and the remaining 2 industrial cameras can be used for subsequent long-term bridge deflection measurements, and the deflections of the other 3 main beams can be calculated based on the lateral proportional coefficient of the deflection of each main beam. The 5 main beams are numbered from #1 to #5 in sequence. Assuming that the vehicle is traveling in the first lane, the remaining two industrial cameras are located at the positions of main beam #2 and main beam #4, respectively. The deflection corresponding to main beam #2 is ω2, and the deflection corresponding to main beam #4 is ω4. According to the lateral deflection proportional coefficient, the deflection of the other main beams can be obtained as follows: ω1=(η 11 ×ω2) / η 12 .

[0051] ω3=[(η 13 ×ω2) / η12 +(η 13 ×ω4) / η 14 ] / 2.

[0052] ω5=(η 15 ×ω4) / η 14 .

[0053] It should be noted that in the above calculation formula, η 11 , η 12 , η 13 , η 14 and η 15 The values ​​are all obtained from the on-site sports car test; that is, after completing the on-site sports car test, the lateral proportional coefficient of the deflection of each main beam corresponding to each lane can be regarded as a constant.

[0054] In this embodiment, when the structural displacement detection device is disassembled after completing step S200, at least two structural displacement detection devices remain. The process of determining the transverse connection stiffness between the main beams by using the transverse proportional coefficients corresponding to the remaining structural displacement detection devices is as follows: Set the transverse connection stiffness determination formula: (ω x / ω y ) = (η x / η y ); if the transverse connection stiffness judgment formula is established, the transverse connection stiffness between the main beams corresponding to the bridge meets the requirements, otherwise it will be determined that the transverse connection stiffness between the main beams corresponding to the bridge does not meet the requirements, where ω x and ω y Respectively represent the mid-span deflection of the main beam corresponding to the remaining two structural displacement detection devices, η x and η y They respectively represent the lateral proportional coefficients of the deflection of the main beams corresponding to the remaining two structural displacement detection devices.

[0055] For ease of understanding, let's take the example of retaining the industrial camera on the main beam #2 and main beam #4 of the first lane mentioned above; when the transverse connection stiffness of the main beam has not decreased in magnitude, the deflection ratio between any two main beams is equal to the ratio of the transverse proportional coefficients of the deflection of the two. From the above content, it can be seen that after completing the sports car test, the transverse proportional coefficient of the deflection of each main beam corresponding to each lane can be regarded as a constant, and the ratio of the transverse proportional coefficient of the deflection corresponding to main beam #2 and main beam #4 can be regarded as a constant k. It is only necessary to continuously monitor the deflection of main beam #2 and main beam #4 and calculate whether the ratio of ω2 and ω4 is equal to k or the difference is within the set error range during the subsequent use of the bridge, so as to monitor the decrease in the transverse connection stiffness of the main beam.

[0056] In this embodiment, step S300 includes the following process: S310: Construct a finite element model of the bridge based on the construction plan of the bridge.

[0057] S320: Use the measured deflection influence line data from the small heavy vehicle runaway test and the corresponding vehicle information to perform stiffness correction on the finite element model of the bridge.

[0058] S330: applying a load corresponding to a large heavy vehicle to the modified bridge finite element model, thereby obtaining influence line simulation data of the mid-span deflection of the bridge corresponding to the large heavy vehicle loading.

[0059] For ease of understanding, the following is a detailed description through specific examples. Assume that the heavy vehicle model is selected as 4-axle, 5-axle and 6-axle, and the vehicle weight is 50 tons, 60 tons, 70 tons and 80 tons and above. Step S300 mainly forms a training data set for the vehicle axle weight identification model through the mid-span deflection of the bridge obtained by the on-site sports car test and the corresponding driving vehicle information (number of vehicle axles and axle weight). And use the measured data to correct and optimize the main beam stiffness and the lateral connection stiffness between the main beams of the bridge finite element model, and expand the deflection data of the bridge sports car working condition with a vehicle weight of more than 40 tons.

[0060] Specifically, first, the measured deflection data and vehicle information of the sports car test below 40 tons are used to correct the stiffness of the finite element model. For example, the known vehicle load information is applied to the established finite element model to obtain the simulated bridge deflection data, and the error analysis is performed with the measured deflection data of the corresponding vehicle weight. At the same time, the main beam stiffness and the lateral connection stiffness between the main beams are continuously adjusted until the error between the two is minimized. Then, vehicle loads of 50 tons, 60 tons, 70 tons and 80 tons or more are applied to the above-mentioned stiffness-corrected bridge finite element model, and the bridge deflection influence line data of the vehicle load of more than 40 tons is calculated, and the bridge deflection data of the actual sports car test condition cannot be expanded. Finally, the simulated bridge deflection influence line data of heavy vehicles above 40 tons is used as the input of the vehicle axle weight identification model, and the corresponding axle weight is used as the output of the vehicle axle weight identification model; the measured and simulated data samples are mixed to construct the training data set of the model.

[0061] In this implementation, step S400 includes the following specific processes: S410: Build a deep learning model architecture for the heavy vehicle axle weight recognition model and select the root mean square error as the model data-driven loss function L data .

[0062] S420: Embed the bridge finite element model with completed stiffness correction into the deep learning model training process.

[0063] S430: Input given bridge deflection influence line data into the deep learning model to generate corresponding axle weight vector and wheelbase information.

[0064] S440: Convert the predicted axle weight vector and wheelbase information into the load of the bridge finite element model and perform structural simulation to obtain corresponding bridge deflection influence line simulation data.

[0065] S450: The error between the given bridge deflection influence line data and the bridge deflection influence line simulation data is used as the physical information loss function L of the deep learning model phy .

[0066] S460: Loss function L data and L phy As the optimization object of the deep learning model, parameter backpropagation is performed to iterate the model parameters.

[0067] It can be understood that the loss function of the heavy vehicle axle weight identification model is L = L data +λL phy ; Among them, λ is the penalty term coefficient, the specific value needs to be adjusted according to the actual training situation, and the value range is [0, 1].

[0068] In this embodiment, step S500 includes the following specific processes: S510: When the monitoring device detects that a target heavy vehicle is on the bridge, the displacement detection device of the underbridge structure is triggered to calculate the mid-span deflection of the bridge.

[0069] S520: Using signal processing technology to eliminate structural dynamic response and environmental noise components in the measured deflection influence line data, and obtain the bridge deflection influence line data of the measured point.

[0070] S530: Estimate the bridge deflection influence lines at other points in the mid-span of the bridge according to the calculated deflection lateral proportionality coefficient and the bridge deflection influence line data at the measured points.

[0071] S540: After the target heavy vehicle passes the bridge, the bridge deflection influence line data of all points are input into the vehicle axle weight identification model to estimate the axle weight of each axle.

[0072] The above describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited by the above embodiments, and the above embodiments and the specification only describe the principles of the present application. The present application may have various changes and improvements without departing from the spirit and scope of the present application, and these changes and improvements fall within the scope of the present application for which protection is sought. The scope of protection claimed by the present application is defined by the attached claims and their equivalents.

Claims

1. A method for identifying axle weight of heavy vehicles based on the distribution of bridge deflection influence lines, characterized in that: The steps include: S100: a monitoring device for identifying vehicle parameters is set up on the bridge deck, and a structural displacement detection device is arranged at the mid-span of the bridge to identify the mid-span deflection of the bridge; S200: Use small and heavy vehicles of different models and weights to conduct on-site running test on the bridge to obtain the influence lines of small and heavy vehicles of different models and weights on the mid-span deflection of the bridge; S300: performing simulation amplification according to the influence line data of the small heavy vehicle to obtain influence line simulation data of the mid-span deflection of the bridge corresponding to the large heavy vehicle; S400: constructing a heavy vehicle axle weight recognition model and performing model training through the influence line data corresponding to all heavy vehicles obtained; S500: Substitute the model of the target heavy vehicle when crossing the bridge and the influence line data of the generated bridge mid-span deflection into the trained heavy vehicle axle weight identification model, and then invert to obtain the axle weight data of the target heavy vehicle.

2. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 1 is characterized in that: The specific process of performing vehicle parameter identification by the monitoring device in step S100 is as follows: S110: Tracking and photographing images of vehicles traveling on the bridge deck by using a monitoring device; S120: Acquire the type of the target vehicle according to the captured image, and calculate the speed of the target vehicle according to the initial position and the calibrated position defined on the bridge deck; S130: According to the estimated speed of the target vehicle, intercepting an image of the target vehicle at at least one position in the original image data to identify the number of axles and wheelbase of the target vehicle; S140: Construct a parameter set of the target vehicle regarding the license plate, vehicle model, number of axles, and wheelbase.

3. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 2 is characterized in that: The monitoring device includes a traffic camera and a laser radar; the monitoring device uses the traffic camera to track and shoot the vehicle with camera images during the day; and the monitoring device uses the traffic camera and the laser radar to coordinately track and shoot the vehicle with camera images and point cloud images at night.

4. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 2 is characterized in that: The structural displacement detection device includes a target and a corner reflector installed at the mid-span position of the bridge main beam, and an industrial camera and a millimeter-wave radar installed at the bridge pier; the industrial camera monitors the displacement of the target to obtain the mid-span deflection of the bridge, and the millimeter-wave radar monitors the phase difference of the electromagnetic wave signal of the corner reflector to obtain the mid-span deflection of the bridge.

5. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 1 is characterized in that: When performing step S200, a plurality of structural displacement detection devices for monitoring the mid-span deflection of the bridge are provided and are arranged in sequence according to the main beams corresponding to the bridge; According to the mid-span deflection of the main beam of each lane obtained from the on-site sports car test, the lateral proportional coefficient η of the deflection under the action of different lanes and vehicle weights is calculated; After completing step S200, some of the structural displacement detection devices are removed, and based on the mid-span deflections of the corresponding main beams obtained by the remaining structural displacement detection devices, the mid-span deflections of the remaining main beams are calculated using the obtained deflection lateral proportional coefficient η.

6. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 5 is characterized in that: The transverse proportional coefficient η of the deflection corresponding to the i-th main beam i The calculation formula is: ; Among them, ω i represents the mid-span deflection of the ith main beam, Represents the sum of the mid-span deflections of the n main beams corresponding to the bridge.

7. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 5 is characterized in that: When the structural displacement detection device is disassembled after step S200, at least two structural displacement detection devices remain. The process of determining the transverse connection stiffness between the main beams by using the transverse proportional coefficients corresponding to the remaining structural displacement detection devices is as follows: Set the lateral connection stiffness judgment formula: (ω x / ω y ) = (η x / η y ); if the transverse connection stiffness judgment formula is established, the transverse connection stiffness between the main beams corresponding to the bridge meets the requirements, otherwise it will be determined that the transverse connection stiffness between the main beams corresponding to the bridge does not meet the requirements, and the transverse proportional coefficient of the deflection of each main beam is recalibrated through a single vehicle model; Among them, ω x and ω y Respectively represent the mid-span deflection of the main beam corresponding to the remaining two structural displacement detection devices, η x and η y They respectively represent the lateral proportional coefficients of the deflection of the main beams corresponding to the remaining two structural displacement detection devices.

8. The method for identifying axle weight of heavy vehicles based on the distribution of bridge deflection influence lines according to any one of claims 1 to 7, characterized in that: Step S300 includes the following process: S310: Construct a finite element model of the bridge; S320: Using the measured deflection influence line data of the small heavy vehicle running test and the corresponding vehicle information, the finite element model of the bridge is corrected for stiffness; S330: applying a load corresponding to a large heavy vehicle to the modified bridge finite element model, thereby obtaining influence line simulation data of the mid-span deflection of the bridge corresponding to the large heavy vehicle loading.

9. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 8, characterized in that: Step S400 includes the following specific processes: S410: Build a deep learning model architecture for the heavy vehicle axle weight recognition model and select the root mean square error as the model data-driven loss function L data ; S420: embedding the bridge finite element model with completed stiffness correction into the deep learning model training process; S430: Input given bridge deflection influence line data into the deep learning model to generate corresponding axle weight vector and wheelbase information; S440: converting the predicted axle weight vector and wheelbase information into loads of a bridge finite element model and performing structural simulation to obtain corresponding bridge deflection influence line simulation data; S450: The error between the given bridge deflection influence line data and the bridge deflection influence line simulation data is used as the physical information loss function L of the deep learning model phy ; S460: Loss function L data and L phy As the optimization object of the deep learning model, parameter backpropagation is performed to iterate the model parameters.

10. The heavy vehicle axle weight identification method based on the bridge deflection influence line distribution according to claim 9, characterized in that: Step S500 includes the following specific processes: S510: When the monitoring device detects that a target heavy vehicle is on the bridge, the displacement detection device of the underbridge structure is triggered to calculate the mid-span deflection of the bridge; S520: using signal processing technology to eliminate structural dynamic response and environmental noise components in the measured deflection influence line data, and obtaining the bridge deflection influence line data of the measured point; S530: estimating the bridge deflection influence lines of other points in the mid-span of the bridge according to the calculated deflection lateral proportionality coefficient and the bridge deflection influence line data of the measured points; S540: After the target heavy vehicle passes the bridge, the bridge deflection influence line data of all points are input into the vehicle axle weight identification model to estimate the axle weight of each axle.

Images