Bridge dynamic weighing identification method based on influence line driving

Abstract

The invention discloses a bridge dynamic weighing identification method based on influence line driving, and the method comprises the steps: carrying out a moving vehicle load test on a bridge, and identifying a bridge influence line; generating random traffic flow information based on a local vehicle load spectrum; according to the vehicle information in the random traffic flow, generating a bridge response under the random traffic flow by using a bridge influence line virtual loading method, and establishing a data set of mapping the vehicle information to the bridge response; building a neural network framework, and training a neural network based on the vehicle information mapping bridge response data set to obtain a vehicle information identification agent model; and identifying vehicle information from the actually measured vehicle-induced bridge response by using the proxy model. According to the method, a data set of vehicle information mapping bridge response is established by using a random traffic flow and actual measurement influence line virtual loading method, a neural network framework training agent model is established, and the lane where the vehicle is located, the vehicle speed, the number of axles, the axle distance and the axle weight can be adaptively identified from the bridge response by using the agent model.

Description

Technical Field

[0001] The invention belongs to the field of bridge health monitoring, relates to dynamic weighing of bridges, and specifically relates to a dynamic weighing identification method for bridges based on influence line drive. Background Art

[0002] Bridge dynamic weighing technology uses the response of the bridge to identify vehicle information on the bridge deck, such as the lane, speed, number of axles, wheelbase, and axle weight. Compared with traditional pavement weighing systems, bridge dynamic weighing technology has the following advantages: easy installation and maintenance, no need to close the bridge; good sensor durability; monitoring data can be used for bridge health monitoring. However, traditional bridge dynamic weighing technology based on influence lines is limited to numerical simulation and laboratory verification, and is still difficult to implement in actual engineering applications. The main problems are: ① It is difficult to identify the axle weight of multiple vehicles; ② The accuracy of the identified axle weight is not high.

[0003] In fact, the vehicle load on the bridge deck has significant randomness and uncertainty. It is difficult to accurately solve the vehicle load using traditional, deterministic methods. Compared with the traditional influence line-based method, the neural network-based method can more accurately identify the axle weight and has the ability to adaptively perceive the vehicle speed, number of axles, wheelbase and lanes. However, the neural network-based method has the following obvious shortcomings: lack of effective data set sources, and it is difficult to train an effective proxy model to identify vehicle information.

[0004] Therefore, a new technical solution is needed to solve this problem. Summary of the invention

[0005] Purpose of the invention: In order to overcome the deficiencies in the prior art, a method for dynamic weighing and identification of bridges based on influence line drive is provided, which can overcome the shortcomings of traditional methods inaccurately identifying axle weights and difficulty in handling multiple vehicles. A data set of vehicle information mapping bridge response is established by using random traffic flow and measured influence line virtual loading methods, and a neural network framework training agent model is built. The agent model can be used to adaptively identify the vehicle's lane, speed, number of axles, axle spacing and axle weight from the bridge response.

[0006] Technical solution: To achieve the above purpose, the present invention provides a bridge dynamic weighing identification method based on influence line drive, comprising the following steps:

[0007] S1: Conduct moving vehicle load tests on bridges to identify bridge influence lines;

[0008] S2: Generate random traffic flow information based on the local vehicle load spectrum;

[0009] S3: Based on the vehicle information in the random traffic flow, the bridge response under the random traffic flow is generated by using the bridge influence line virtual loading method, and a data set of vehicle information mapping bridge response is established;

[0010] S4: Build a neural network framework, train the neural network based on the vehicle information mapping bridge response dataset, and obtain a proxy model for vehicle information recognition;

[0011] S5: Using a surrogate model to identify vehicle information from measured vehicle-induced bridge responses.

[0012] Furthermore, the method for identifying the influence line of the bridge in step S1 is: performing a moving vehicle load test, a heavy truck passes through each lane of the bridge at a slow speed, collecting the quasi-static response of the bridge (bridge strain and displacement), and then using a regularization method to identify the influence line of the bridge.

[0013] Furthermore, in step S2, the dynamic weighing system data of the local bridges of the same type are used for statistics, and the statistical content includes lanes, vehicle types, speeds, axle loads, wheelbases and vehicle headway.

[0014] Furthermore, in step S3, according to the vehicle information in the random traffic flow, the bridge response under the random traffic flow is generated by using the bridge influence line virtual loading method, specifically:

[0015] According to the structural reciprocity theorem, the virtual work done by the internal force of the structure in the first state on the displacement in the second state is equal to the virtual work done by the internal force in the second state on the displacement in the first state; the response of the bridge measuring point is equal to the weighted sum of the values ​​of each load and influence line, and the formula is as follows:

[0016]

[0017] Where y is the bridge response simulated using virtual loading, F i represents the i-th load value, l i Indicates that the influence line of the measuring point is at F i The value of the influence line at the position.

[0018] Furthermore, in step S5, a convolutional neural network is used to build a dynamic weighing technology framework for bridges. The network framework accepts a vehicle-induced bridge response signal as input, performs normalization on each sample, extracts signal features through convolutional layers, residual blocks, pooling layers and other layers, and distinguishes vehicle information from the bridge response.

[0019] Furthermore, in step S5, a neural network is trained based on the vehicle information mapping bridge response data set, and the training method is: when building a framework, the bridge displacement is used as input, and the vehicle information (number of axles, axle dead weight, vehicle speed and lane where the vehicle is located) is used as output to establish a mapping relationship between the vehicle information and the bridge response.

[0020] Furthermore, the structure of the neural network in step S5 includes

[0021] Furthermore, the operation of the neural network in step S5 includes:

[0022] (a) Embed the original sequence into the variable;

[0023] (b) Self-attention applied to embedded variable tokens with enhanced interpretability, revealing multivariate correlations;

[0024] (c) Extracting the level representation of each token through a shared feed-forward network;

[0025] (d) Layer normalization is used to reduce the differences between variables.

[0026] Beneficial effects: Compared with the prior art, the present invention embeds a neural network algorithm into the dynamic weighing system of the bridge, replacing the traditional mathematical model method to solve the vehicle load. The random traffic flow and measured influence line virtual loading method are used to establish a data set for mapping the bridge response of vehicle information, and a neural network framework training agent model is built. The agent model can be used to adaptively identify the vehicle lane, speed, number of axles, axle spacing and axle weight from the bridge response. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the method of the present invention;

[0028] Figure 2 This is the result diagram of bridge influence line identification;

[0029] Figure 3 is the vehicle load spectrum;

[0030] Figure 4 This is a schematic diagram of virtual loading;

[0031] Figure 5 Virtual loading result diagram for influence lines;

[0032] Figure 6 It is a neural network framework diagram;

[0033] Figure 7 This is the displacement diagram of the bridge under random traffic flow. DETAILED DESCRIPTION

[0034] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0035] like Figure 1 As shown, the present invention provides a bridge dynamic weighing identification method based on influence line drive, comprising the following steps:

[0036] S1: Conduct moving vehicle load tests on bridges to identify bridge influence lines;

[0037] Conduct a moving vehicle load test, where a heavy truck passes through each lane of the bridge at a slow speed, collect the quasi-static response of the bridge (bridge strain and displacement), and then use the regularization method to identify the influence line of the bridge;

[0038] In this embodiment, a 30-ton four-axle test vehicle is selected to confirm the axle weight and axle spacing of the vehicle in advance. The test vehicle passes through each lane of the bridge at a speed of 2m / s to measure the displacement and strain of the mid-span and 1 / 4 span of the bridge. The bridge displacement under random traffic flow obtained in this embodiment is as follows: Figure 7 The influence line of the bridge is identified by using the regularization method. The identification method is as follows:

[0039] P·L=R (1)

[0040] Where P is the position matrix of the axle, L is the bridge influence line vector, and R is the bridge response;

[0041] In this embodiment, the results of identifying the bridge influence line based on the bridge response generated by the test vehicle are as follows: Figure 2 shown.

[0042] S2: Generate random traffic flow information based on the local vehicle load spectrum;

[0043] Traffic load has great regional characteristics and is also related to the traffic capacity of the tested bridge. In order to make the statistical results more accurately express the vehicle load distribution of the tested bridge, the dynamic weighing system data of the same type of local bridges should be used. The statistical content includes lanes, vehicle types, speeds, axle loads, wheelbases and vehicle headway.

[0044] The vehicle load spectrum obtained in this embodiment is as follows: Figure 3 According to the vehicle load spectrum, the Monte Carlo random sampling technique is used to extract vehicle information samples to form random traffic flow information.

[0045] S3: Based on the vehicle information in the random traffic flow, the bridge response under the random traffic flow is generated by using the bridge influence line virtual loading method, and a data set of vehicle information mapping bridge response is established;

[0046] According to the structural reciprocity theorem, the virtual work done by the internal forces of the structure in the first state on the displacement in the second state is equal to the virtual work done by the internal forces of the structure in the second state on the displacement in the first state; Figure 4 , the response of the bridge measuring point is equal to the weighted sum of the loads and the influence line values, and the formula is as follows:

[0047]

[0048] Where y is the bridge response simulated using virtual loading, F i represents the i-th load value, l i Indicates that the influence line of the measuring point is at F i The value of the influence line at the position.

[0049] Based on the random traffic flow load information and the influence line virtual loading method, the static response of the bridge under the random traffic flow can be effectively simulated. Taking the partial displacement response as an example, Figure 5 The measured response of the bridge under the action of traffic flow is shown. The measured response after filtering is almost consistent with the response of virtual loading, which shows that the influence line virtual loading method can effectively simulate the static response of the bridge.

[0050] S4: Build a neural network framework, train the neural network based on the vehicle information mapping bridge response dataset, and obtain a proxy model for vehicle information recognition;

[0051] The neural network is trained based on the vehicle information mapping bridge response dataset. The training method is as follows: when building the framework, the bridge displacement is used as input, and the vehicle information (number of axles, axle dead weight, vehicle speed, and vehicle lane) is used as output to establish a mapping relationship between vehicle information and bridge response.

[0052] like Figure 6 The structure and operation architecture of the neural network are shown. The operation of the neural network includes:

[0053] (a) Embed the original sequence into the variable;

[0054] (b) Self-attention applied to embedded variable tokens with enhanced interpretability, revealing multivariate correlations;

[0055] (c) Extracting the level representation of each token through a shared feed-forward network;

[0056] (d) Layer normalization is used to reduce the differences between variables.

[0057] S5: Measure the vehicle-induced bridge response and filter it to extract the measured static response of the bridge. Use the proxy model to estimate vehicle information such as the number of axles, axle dead weight, vehicle speed, and vehicle lane.

Claims

1. A bridge dynamic weighing identification method based on influence line drive, characterized in that: The steps include: S1: Conduct moving vehicle load tests on bridges to identify bridge influence lines; S2: Generate random traffic flow information based on the local vehicle load spectrum; S3: Based on the vehicle information in the random traffic flow, the bridge response under the random traffic flow is generated by using the bridge influence line virtual loading method, and a data set of vehicle information mapping bridge response is established; S4: Build a neural network framework, train the neural network based on the vehicle information mapping bridge response dataset, and obtain a proxy model for vehicle information recognition; S5: Using a surrogate model to identify vehicle information from measured vehicle-induced bridge responses.

2. The bridge dynamic weighing identification method based on influence line drive according to claim 1 is characterized in that: The method for identifying the influence line of the bridge in step S1 is: performing a moving vehicle load test, where a heavy truck passes through each lane of the bridge at a slow speed, collecting the quasi-static response of the bridge, and then using a regularization method to identify the influence line of the bridge.

3. The bridge dynamic weighing identification method based on influence line drive according to claim 2 is characterized in that: In step S1, the influence line of the bridge is identified by using a regularization method, and the identification method is as follows: P·L=R (1) Among them, P is the position matrix of the axle, L is the bridge influence line vector, and R is the bridge response.

4. The bridge dynamic weighing identification method based on influence line drive according to claim 1 is characterized in that: In step S2, the dynamic weighing system data of the local bridges of the same type are used for statistics, and the statistical content includes lanes, vehicle types, vehicle speeds, axle loads, wheelbases and vehicle headway.

5. The bridge dynamic weighing identification method based on influence line drive according to claim 1 is characterized in that: In step S3, the bridge response under random traffic flow is generated by using the bridge influence line virtual loading method according to the vehicle information in the random traffic flow, specifically: According to the structural reciprocity theorem, the virtual work done by the internal force of the structure in the first state on the displacement in the second state is equal to the virtual work done by the internal force in the second state on the displacement in the first state; the response of the bridge measuring point is equal to the weighted sum of the values ​​of each load and influence line, and the formula is as follows: Where y is the bridge response simulated using virtual loading, F i represents the i-th load value, l i Indicates that the influence line of the measuring point is at F i The value of the influence line at the position.

6. The bridge dynamic weighing identification method based on influence line drive according to claim 1 is characterized in that: In step S5, a convolutional neural network is used to build a dynamic weighing technology framework for bridges. The network framework accepts the vehicle-induced bridge response signal as input, performs normalization on each sample, extracts signal features through layers, and distinguishes vehicle information from the bridge response.

7. The bridge dynamic weighing identification method based on influence line drive according to claim 6 is characterized in that: In the step S5, signal features are extracted through layer blocks, and the layer blocks include convolutional layers, residual blocks, and pooling layers.

8. The bridge dynamic weighing identification method based on influence line drive according to claim 1 is characterized in that: In step S5, a neural network is trained based on the vehicle information mapping bridge response data set. The training method is: when building a framework, the bridge displacement is used as input and the vehicle information is used as output to establish a mapping relationship between the vehicle information and the bridge response.

9. The bridge dynamic weighing identification method based on influence line drive according to claim 8 is characterized in that: The operation of the neural network in step S5 includes: (a) Embed the original sequence into the variable; (b) Self-attention applied to embedded variable tokens; (c) Extracting the level representation of each token through a shared feed-forward network; (d) Layer normalization is used to reduce the differences between variables.

Images