Vehicle dynamic weighing method, device, equipment and medium based on neural network

Abstract

The embodiments of the present application provide a vehicle dynamic weighing method, device, equipment and medium based on a neural network, which relates to the field of deep learning. By acquiring the information of passing vehicles on the bridge and combining the finite element bridge vibration model with random traffic samples, a theoretical sample library for deep learning training of random traffic bridge response is constructed, and the actual dynamic response data of the bridge when there are passing vehicles and the noise data when there are no passing vehicles are collected. Through data fusion, the noise data and the theoretical sample library for deep learning training of random traffic bridge response are combined to obtain training samples. According to the training samples, a neural network model is trained to obtain a deep learning model for dynamic weighing of vehicles on the bridge. The model can accurately output the weight of vehicles passing the bridge, and the robustness and accuracy of the model are significantly improved by integrating actual noise data.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a vehicle dynamic weighing method, device, equipment and medium based on a neural network. Background Art

[0002] With the development of society and economy, the demand for transportation is increasing. Vehicle load is one of the main loads borne by bridge structures at the bridge deck. Excessive traffic and heavy vehicle weight can easily cause bridge damage, bridge deck damage, and accelerated bridge deterioration. Therefore, obtaining the weight of vehicles passing on the bridge is of great significance for the improvement of bridge structure design specifications, safety assessment, and maintenance decisions.

[0003] At present, dynamic weighing of bridge vehicles usually uses induction coils installed on the bridge deck to collect vehicle information. However, due to the large difference between the theoretical model and the actual bridge and the high noise of the bridge, the collected data is inaccurate and has poor robustness. Therefore, a method that can accurately obtain vehicle weight is urgently needed. Summary of the invention

[0004] In view of this, the purpose of the present invention is to overcome the deficiencies in the prior art and to provide a vehicle dynamic weighing method, device, equipment and medium based on a neural network.

[0005] The present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present disclosure provides a vehicle dynamic weighing method based on a neural network, the method comprising:

[0007] Obtaining information about vehicles passing through the bridge to be monitored, and generating random traffic samples according to the information about vehicles passing through;

[0008] A bridge coupling vibration analysis model of the bridge to be monitored is established based on the finite element method, and a random traffic flow bridge response deep learning training theory sample library is established according to the random traffic flow samples and the bridge coupling vibration analysis model;

[0009] According to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, the dynamic response data and noise data of the bridge to be monitored are collected in real time, wherein the sensing data collected by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there is no vehicle passing is the noise data;

[0010] The noise data and the random traffic flow bridge response deep learning training theory sample library are fused by a data fusion algorithm to obtain training samples;

[0011] Performing neural network model training through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network;

[0012] The dynamic response data is input into the vehicle dynamic weighing deep learning model, and the weight of the vehicle passing through the bridge to be monitored is output.

[0013] According to a specific implementation method disclosed in the present application, the step of obtaining the passing vehicle information of the bridge to be monitored and generating a random traffic sample according to the passing vehicle information includes:

[0014] The passing vehicle information includes vehicle type ratio distribution, vehicle weight distribution, vehicle speed distribution and vehicle distance distribution, and determining the parameter distribution of the passing vehicle information, wherein the vehicle type ratio distribution is represented by a polynomial distribution, the vehicle weight distribution is represented by a normal distribution or a uniform distribution, the vehicle speed distribution is represented by a normal distribution, and the vehicle distance distribution is represented by an exponential distribution;

[0015] A random traffic flow sample of the passing vehicle information is generated based on a random number generation method.

[0016] According to a specific implementation method disclosed in the present application, the steps of establishing a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, and establishing a random traffic flow bridge response deep learning training theory sample library according to the random traffic flow samples and the bridge coupling vibration analysis model include:

[0017] The bridge coupling vibration analysis model includes a vibration model of the bridge to be monitored, a vehicle vibration equation of the bridge to be monitored, and an interaction force between the bridge to be monitored and the vehicle;

[0018] The vibration model of the bridge to be monitored is expressed as:

[0019] ;

[0020] The vehicle vibration equation of the bridge to be monitored is expressed as:

[0021] ;

[0022] The interaction force between the bridge to be monitored and the vehicle is expressed as:

[0023] ;

[0024] Among them, V i is the vehicle speed, m i For vehicle quality, is the vehicle interval time, N is the total number of vehicles in the random traffic flow, w(x,t) is the displacement of the bridge at position x and time t, E is the Young's modulus of the beam material, I is the section moment of inertia of the beam, ρ is the density of the beam, A is the cross-sectional area of ​​the beam, q(x,t) is the distributed load along the beam length x and time t, y v (t) is the vertical displacement of the vehicle, C V is the damping coefficient of the vehicle, K V is the spring stiffness of the vehicle, g is the acceleration due to gravity, δ(xI t ) is the Dirac delta function;

[0025] The random traffic flow samples are input into the bridge coupled vibration analysis model to obtain a theoretical sample library for deep learning training of random traffic flow bridge response.

[0026] According to a specific implementation method disclosed in the present application, the step of collecting the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored includes:

[0027] The dynamic displacement sensor is used to measure the displacement change of the bridge to be monitored when a vehicle passes by;

[0028] The accelerometer is used to measure the acceleration response of the bridge to be monitored;

[0029] The dynamic strain gauge is used to measure the strain change of the bridge to be monitored;

[0030] The bridge type of the bridge to be monitored is obtained, and the minimum sampling frequency of the dynamic response data of the bridge to be monitored is set according to the bridge type.

[0031] According to a specific implementation method disclosed in the present application, the step of fusing the noise data with the random traffic bridge response deep learning training theory sample library through a data fusion algorithm to obtain training samples includes:

[0032] Performing spectrum analysis, noise source identification processing and preprocessing on the noise data to obtain final noise data;

[0033] The Kalman filtering algorithm is used to fuse the final noise data and the deep learning training theoretical sample library of random traffic bridge response, and the estimated value of the response of the bridge to be monitored is dynamically updated. The Bayesian estimation method is used to fuse the final noise data and the deep learning training theoretical sample library of random traffic bridge response to update and correct the probability model.

[0034] According to a specific implementation method disclosed in the present application, the step of training a neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network includes:

[0035] The loss function of the vehicle dynamic weighing deep learning model includes data error and physical error. The data error is expressed as:

[0036] , used to measure the error between the deflection of the vehicle dynamic weighing deep learning model and the measured deflection, where N is the number of data points, x i and t i are the position and time of the ith data point, respectively. is the output deflection, W measured is the measured deflection;

[0037] The physical error is expressed as:

[0038] , which is used to measure whether the output deflection of the vehicle dynamic weighing deep learning model meets the physical constraints, where N is the number of data points, E is the Young's modulus of the beam material, and I is the section moment of inertia of the beam.

[0039] According to a specific embodiment disclosed in the present application, after the step of inputting the dynamic response data into the vehicle dynamic weighing deep learning model and outputting the weight of the vehicle passing through the bridge to be monitored, the step further includes:

[0040] Collecting the periodic dynamic response data of the bridge to be monitored according to a preset period, fusing the periodic dynamic response data with the training samples to obtain periodic training samples;

[0041] Generate adversarial samples based on the periodic training samples;

[0042] The data error and adversarial error of the vehicle dynamic weighing deep learning model are optimized according to the adversarial sample.

[0043] In a second aspect, an embodiment of the present disclosure provides a vehicle dynamic weighing device based on a neural network, the device comprising:

[0044] An acquisition module, used to acquire the passing vehicle information of the bridge to be monitored, and generate a random traffic flow sample according to the passing vehicle information;

[0045] Establishing a module for establishing a bridge coupling vibration analysis model of the bridge to be monitored based on a finite element method, and establishing a random traffic flow bridge response deep learning training theory sample library according to the random traffic flow samples and the bridge coupling vibration analysis model;

[0046] A collection module, used for collecting dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, wherein the sensing data collected by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there is no vehicle passing is the noise data;

[0047] A fusion module, used to fuse the noise data with the random traffic bridge response deep learning training theory sample library through a data fusion algorithm to obtain a training sample;

[0048] A training module, used to train a neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network;

[0049] An output module is used to input the dynamic response data into the vehicle dynamic weighing deep learning model and output the weight of the vehicle passing through the bridge to be monitored.

[0050] In a third aspect, an electronic device is provided in an embodiment of the present disclosure, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the vehicle dynamic weighing method based on a neural network as described in any one of the first aspects when executing the computer program.

[0051] In a fourth aspect, a computer-readable storage medium is provided in an embodiment of the present disclosure, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the vehicle dynamic weighing method based on a neural network described in any one of the first aspects are implemented.

[0052] The vehicle dynamic weighing method based on neural network provided in the present application obtains the passing vehicle information of the bridge to be monitored, generates random traffic samples according to the passing vehicle information, establishes a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, establishes a random traffic bridge response deep learning training theoretical sample library according to the random traffic samples and the bridge coupling vibration analysis model, and collects the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored. The sensing data collected when there is no vehicle passing is the noise data. The noise data and the random traffic bridge response deep learning training theoretical sample library are fused through a data fusion algorithm to obtain training samples. The neural network model is trained through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network. The dynamic response data is input into the vehicle dynamic weighing deep learning model, and the weight of the vehicle passing through the bridge to be monitored is output. Samples with actual noise characteristics are established through data fusion. The neural network based on physical information can accurately obtain the vehicle weight, thereby improving the robustness and accuracy of the model.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work. In each of the drawings, similar components are numbered similarly.

[0055] Figure 1 A flowchart of a vehicle dynamic weighing method based on a neural network provided in an embodiment of the present application is shown;

[0056] Figure 2 A schematic structural diagram of a vehicle dynamic weighing device based on a neural network provided in an embodiment of the present application is shown;

[0057] Figure 3 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0058] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0059] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. In contrast, when an element is referred to as being "directly on" another element, there is no intermediate element. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0060] In the present invention, 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 a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection 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.

[0061] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of the template are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0063] Example 1

[0064] Through research, it was found that in the prior art, the response data of the bridge when the vehicle passes through the bridge is measured to indirectly calculate the vehicle weight. The finite element model of the bridge is theoretically calculated to obtain the mapping relationship between the bridge response data, the spatial position of the vehicle and the vehicle weight. The weight of the actual passing vehicles is then inferred based on the mapping relationship. This method cannot accurately measure the weight of vehicles passing on the bridge and has poor robustness due to practical reasons such as the large difference between the theoretical model and the actual bridge, the large difference between the experimental vehicle and the actual vehicle, and the large measurement noise of the bridge response. The present invention provides a technical solution that conforms to actual traffic conditions and can accurately measure the weight of vehicles passing on the bridge. Figure 1 As shown, it is a schematic diagram of a flow chart of a vehicle dynamic weighing method based on a neural network provided in an embodiment of the present application. The vehicle dynamic weighing method based on a neural network provided in an embodiment of the present application comprises the following steps:

[0065] Step S101, obtaining the passing vehicle information of the bridge to be monitored, and generating a random traffic flow sample according to the passing vehicle information.

[0066] The step of obtaining the passing vehicle information of the bridge to be monitored and generating a random traffic flow sample according to the passing vehicle information comprises:

[0067] The passing vehicle information includes vehicle type ratio distribution, vehicle weight distribution, vehicle speed distribution and vehicle distance distribution, and determining the parameter distribution of the passing vehicle information, wherein the vehicle type ratio distribution is represented by a polynomial distribution, the vehicle weight distribution is represented by a normal distribution or a uniform distribution, the vehicle speed distribution is represented by a normal distribution, and the vehicle distance distribution is represented by an exponential distribution;

[0068] Generate random traffic flow samples of the passing vehicle information based on a random number generation method

[0069] Specifically, the vehicle information on the bridge to be monitored can be obtained by collecting similar bridge traffic survey statistics, using the Monte Carlo method to write a random traffic flow simulation analysis program in a programming language, comprehensively considering the weight, speed, vehicle type and distance of the vehicles on the bridge and the distribution parameters of each parameter, and combining the various parameters through the random number generation method to generate a large number of typical random traffic flow samples with the characteristics of the bridge area to be monitored. The random traffic flow samples generated in this way not only have practical reference significance, but also can improve the accuracy of the data.

[0070] Step S102: establishing a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, and establishing a random traffic flow bridge response deep learning training theory sample library according to the random traffic flow samples and the bridge coupling vibration analysis model.

[0071] It can be understood that the bridge coupled vibration analysis model of the bridge to be monitored performs modal analysis to obtain vibration characteristics such as the natural frequency and modal vibration shape of the bridge, and analyzes the dynamic response of the bridge under different working conditions by considering vehicle loads.

[0072] The steps of establishing a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, and establishing a random traffic flow bridge response deep learning training theoretical sample library according to the random traffic flow samples and the bridge coupling vibration analysis model include:

[0073] The bridge coupling vibration analysis model includes a vibration model of the bridge to be monitored, a vehicle vibration equation of the bridge to be monitored, and an interaction force between the bridge to be monitored and the vehicle;

[0074] The vibration model of the bridge to be monitored is expressed as:

[0075] ;

[0076] The vehicle vibration equation of the bridge to be monitored is expressed as:

[0077] ;

[0078] The interaction force between the bridge to be monitored and the vehicle is expressed as:

[0079] ;

[0080] Among them, V i is the vehicle speed, m i For vehicle quality, is the vehicle interval time, N is the total number of vehicles in the random traffic flow, w(x,t) is the displacement of the bridge at position x and time t, E is the Young's modulus of the beam material, I is the section moment of inertia of the beam, ρ is the density of the beam, A is the cross-sectional area of ​​the beam, q(x,t) is the distributed load along the beam length x and time t, y v (t) is the vertical displacement of the vehicle, C V is the damping coefficient of the vehicle, K V is the spring stiffness of the vehicle, g is the acceleration due to gravity, δ(xI t ) is the Dirac delta function;

[0081] The random traffic flow samples are input into the bridge coupled vibration analysis model to obtain a theoretical sample library for deep learning training of random traffic flow bridge response.

[0082] Specifically, by inputting the simulated random traffic flow samples into the bridge coupled vibration analysis model, which includes the vibration model, the vehicle vibration equation of the bridge and the interaction force between the bridge and the vehicle, the response data of the bridge when vehicles pass through the bridge under different working conditions can be obtained. By establishing a mapping relationship between random traffic flow samples and bridge response data, a theoretical sample library for deep learning training of random traffic flow bridge responses can be obtained.

[0083] Step S103, based on the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, the dynamic response data and noise data of the bridge to be monitored are collected in real time, wherein the sensing data collected by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there are no vehicles passing is the noise data.

[0084] Specifically, the dynamic displacement sensor, accelerometer and dynamic strain gauge are installed at the locations where stress concentration or large deformation occurs under the action of bridge loads, such as the mid-span of the main beam, the crossbeam node and the end of the suspender. It can accurately collect response data when vehicles pass on the bridge and noise data when no vehicles pass on the bridge.

[0085] The step of collecting the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored comprises:

[0086] The dynamic displacement sensor is used to measure the displacement change of the bridge to be monitored when a vehicle passes by;

[0087] The accelerometer is used to measure the acceleration response of the bridge to be monitored;

[0088] The dynamic strain gauge is used to measure the strain change of the bridge to be monitored;

[0089] The bridge type of the bridge to be monitored is obtained, and the minimum sampling frequency of the dynamic response data of the bridge to be monitored is set according to the bridge type.

[0090] Specifically, the time required for vehicles to pass through different types of bridges is different. If the sensors do not collect enough information within the vehicle's passing time, the vehicle weight cannot be accurately measured. Therefore, by obtaining the bridge type, different sensor sampling frequencies are set for different bridge types to ensure data accuracy. Common bridge types include simply supported bridges, continuous beam bridges, and arch bridges. The frequency formula for simply supported bridges is: Calculate its minimum sampling frequency through the continuous beam bridge frequency formula Calculate its minimum sampling frequency through the arch bridge frequency formula , Calculate its minimum sampling frequency, where l is the calculated span of the structure, E is the elastic modulus of the structural material, and I c is the section moment of inertia of the mid-span section of the structure, m c is the mass per unit length at the mid-span of the structure, G is the structural gravity per meter at the mid-span of the structure, and k is the arch span ratio.

[0091] Step S104: The noise data and the random traffic flow bridge response deep learning training theory sample library are fused through a data fusion algorithm to obtain training samples.

[0092] Specifically, the training samples include the response data of the bridge when there are vehicles passing through, the noise data of the bridge when there are no vehicles passing through, and the vehicle information of the vehicles passing through the bridge. Through data fusion, training samples with actual noise data are obtained, which can provide data support for subsequent neural network model training and improve the prediction accuracy and generalization ability of the model.

[0093] The step of fusing the noise data with the random traffic flow bridge response deep learning training theory sample library through a data fusion algorithm to obtain training samples includes:

[0094] Performing spectrum analysis, noise source identification processing and preprocessing on the noise data to obtain final noise data;

[0095] The Kalman filtering algorithm is used to fuse the final noise data and the deep learning training theoretical sample library of random traffic bridge response, and the estimated value of the response of the bridge to be monitored is dynamically updated. The Bayesian estimation method is used to fuse the final noise data and the deep learning training theoretical sample library of random traffic bridge response to update and correct the probability model.

[0096] Furthermore, the noise data is spectrally analyzed using fast Fourier transform to identify the main frequency components of the noise signal and determine the amplitude of each frequency component. The main sources of noise data, such as wind load, environmental vibration, and temperature change, are identified based on the spectral characteristics and time domain characteristics of the noise data. The noise data is preprocessed by filtering and smoothing to eliminate outliers and interference in the data and ensure the accuracy and stability of the data. The Kalman filter algorithm can effectively eliminate noise interference and improve the precision and accuracy of the data. The Bayesian estimation method can provide a quantitative assessment of data uncertainty and improve the robustness and reliability of data fusion.

[0097] Step S105, performing neural network model training using the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network.

[0098] Specifically, the historical response data fused with noise data in the training samples is used as input, and the weight of vehicles on the bridge is used as output to train the dynamic weighing deep learning model on the bridge based on the physical information neural network. The response data is the historical response data of the bridge when vehicles pass through, and the weight of vehicles passing through the bridge is the historical vehicle data.

[0099] The steps of training a neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network include:

[0100] The loss function of the vehicle dynamic weighing deep learning model includes data error and physical error. The data error is expressed as:

[0101] , which is used to measure the error between the deflection of the vehicle dynamic weighing deep learning model and the measured deflection, where N is the number of data points, x i and t i are the position and time of the ith data point, respectively. is the output deflection, W measured is the measured deflection;

[0102] The physical error is expressed as:

[0103] , which is used to measure whether the output deflection of the vehicle dynamic weighing deep learning model meets the physical constraints, where N is the number of data points, E is the Young's modulus of the beam material, and I is the section moment of inertia of the beam.

[0104] Furthermore, the vehicle weight of the bridge to be monitored is calculated using the deep learning model of vehicle dynamic weighing based on physical information, which improves the interpretability and computational efficiency. Defining the physical constraint loss term based on the deep learning model of vehicle dynamic weighing based on physical information can ensure that the network output meets the constraints of the physical equations, making the model's prediction results more accurate and reliable, and able to meet the physical constraints required by the bridge in actual working conditions.

[0105] Step S106: input the dynamic response data into the vehicle dynamic weighing deep learning model, and output the weight of the vehicle passing through the bridge to be monitored.

[0106] Specifically, this embodiment establishes a deep learning model for vehicle dynamic weighing through a deep learning model, with the real-time dynamic response data of the bridge as the input coefficient and the vehicle weight as the output coefficient. The vehicle weight can be obtained by collecting the dynamic response data of the bridge when there are vehicles passing in real time and inputting it into the deep learning model for vehicle dynamic weighing. It avoids the technical difficulties of the finite element model correction in the prior art, combines deep learning and physically constrained machine learning models, and uses physical laws to guide the model training during the learning process, thereby improving the generalization ability of the model. It has excellent robustness when there is less data or more noise, and is more in line with the actual working conditions of the bridge.

[0107] After the step of inputting the dynamic response data into the vehicle dynamic weighing deep learning model and outputting the weight of the vehicle passing through the bridge to be monitored, the method further includes:

[0108] Collecting the periodic dynamic response data of the bridge to be monitored according to a preset period, fusing the periodic dynamic response data with the training samples to obtain periodic training samples;

[0109] Generate adversarial samples based on the periodic training samples;

[0110] The data error and adversarial error of the vehicle dynamic weighing deep learning model are optimized according to the adversarial sample.

[0111] Specifically, the preset period units include days, months and years. The collected new bridge dynamic response data are integrated into the training samples. Adversarial sample generation technology, such as FGSM (Fast Gradient Sign Method), is used to generate adversarial samples. A total loss function including an adversarial loss function is designed to minimize both data error and adversarial error during the optimization process.

[0112] Understandably, bridge monitoring often involves large sensor data noise and more abnormal data. To avoid affecting the calculation of vehicle weight, adversarial training is performed on the model using measured data over a certain period of time to improve the model's detection capability and robustness, thereby ensuring the model's stability and accuracy in various complex environments.

[0113] The vehicle dynamic weighing method based on neural network provided in the embodiment of the present application obtains the passing vehicle information of the bridge to be monitored, generates random traffic samples according to the passing vehicle information, establishes a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, establishes a random traffic bridge response deep learning training theoretical sample library according to the random traffic samples and the bridge coupling vibration analysis model, and collects the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored. The sensing data collected when there is no vehicle passing is the noise data. The noise data and the random traffic bridge response deep learning training theoretical sample library are fused through a data fusion algorithm to obtain training samples. The neural network model is trained through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network. The dynamic response data is input into the vehicle dynamic weighing deep learning model, and the weight of the vehicle passing through the bridge to be monitored is output. Samples with actual noise characteristics are established through data fusion. The neural network based on physical information can accurately obtain the vehicle weight, thereby improving the robustness and accuracy of the model.

[0114] Example 2

[0115] like Figure 2 FIG. 2 is a schematic diagram of a vehicle dynamic weighing device 200 based on a neural network provided in an embodiment of the present application, wherein the device comprises:

[0116] An acquisition module 201 is used to acquire the passing vehicle information of the bridge to be monitored, and generate a random traffic flow sample according to the passing vehicle information;

[0117] Establishing module 202, for establishing a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, and establishing a random traffic flow bridge response deep learning training theory sample library according to the random traffic flow samples and the bridge coupling vibration analysis model;

[0118] The acquisition module 203 is used to acquire the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, wherein the sensing data acquired by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there is no vehicle passing is the noise data;

[0119] A fusion module 204 is used to fuse the noise data with the random traffic bridge response deep learning training theory sample library through a data fusion algorithm to obtain a training sample;

[0120] A training module 205 is used to train a neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network;

[0121] The output module 206 is used to input the dynamic response data into the vehicle dynamic weighing deep learning model and output the weight of the vehicle passing through the bridge to be monitored.

[0122] The neural network-based vehicle dynamic weighing device 200 provided in this embodiment can implement the neural network-based vehicle dynamic weighing method shown in Example 1, which will not be described again here to avoid repetition.

[0123] The vehicle dynamic weighing device based on neural network provided in the embodiment of the present application obtains the passing vehicle information of the bridge to be monitored, generates random traffic samples according to the passing vehicle information, establishes a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, establishes a random traffic bridge response deep learning training theoretical sample library according to the random traffic samples and the bridge coupling vibration analysis model, and collects the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored. The sensing data collected by the strain gauge when there is no vehicle passing is the noise data. The noise data and the random traffic bridge response deep learning training theoretical sample library are fused through a data fusion algorithm to obtain training samples. The neural network model is trained through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network. The dynamic response data is input into the vehicle dynamic weighing deep learning model, and the weight of the vehicle passing through the bridge to be monitored is output. A sample library with actual noise characteristics is established through data fusion. The neural network based on physical information can accurately obtain the vehicle weight, thereby improving the robustness of the model.

[0124] Example 3

[0125] In addition, an embodiment of the present disclosure provides an electronic device 300, including a memory and a processor, wherein the memory stores a computer program.

[0126] For details, see Figure 3 The electronic device 300 includes: a receiver 301, a bus interface and a processor 302.

[0127] In the embodiment of the present invention, the electronic device 300 further includes a memory 303. Figure 3In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 302 and memory represented by memory 303. The bus architecture may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The receiver 301 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. The processor 302 is responsible for managing the bus architecture and general processing, and the memory 303 may store data used by the processor 302 when performing operations.

[0128] The electronic device 300 provided in the embodiment of the present invention can implement the vehicle dynamic weighing method based on neural network shown in Example 1, and will not be described again here to avoid repetition.

[0129] The electronic device provided in the embodiment of the present application obtains the information of passing vehicles on the bridge to be monitored, generates random traffic samples according to the passing vehicle information, establishes a bridge coupling vibration analysis model of the bridge to be monitored based on the finite element method, establishes a random traffic flow bridge response deep learning training theoretical sample library according to the random traffic flow samples and the bridge coupling vibration analysis model, collects the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, wherein the sensing data collected by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there is no vehicle passing is the noise data, fuses the noise data with the random traffic flow bridge response deep learning training theoretical sample library through a data fusion algorithm to obtain training samples, trains a neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network, inputs the dynamic response data into the vehicle dynamic weighing deep learning model, outputs the weight of the vehicle passing through the bridge to be monitored, and establishes a sample library with actual noise characteristics through data fusion. The neural network based on physical information can accurately obtain the vehicle weight, thereby improving the robustness of the model.

[0130] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in an alternative implementation, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the structure diagram and / or the flow chart, and the combination of boxes in the structure diagram and / or the flow chart, can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0131] In addition, the functional modules or units in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0132] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0133] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A vehicle dynamic weighing method based on neural network, characterized in that: The method comprises: Obtaining information about vehicles passing through the bridge to be monitored, and generating random traffic samples according to the information about vehicles passing through; The passing vehicle information includes vehicle type ratio distribution, vehicle weight distribution, vehicle speed distribution and vehicle distance distribution, determining the parameter distribution of the passing vehicle information, wherein the vehicle type ratio distribution is represented by a polynomial distribution, the vehicle weight distribution is represented by a normal distribution or a uniform distribution, the vehicle speed distribution is represented by a normal distribution, and the vehicle distance distribution is represented by an exponential distribution, and generating a random traffic flow sample of the passing vehicle information based on a random number generation method; A bridge coupling vibration analysis model of the bridge to be monitored is established based on the finite element method, and a random traffic flow bridge response deep learning training theory sample library is established according to the random traffic flow samples and the bridge coupling vibration analysis model; The bridge coupling vibration analysis model includes a vibration model of the bridge to be monitored, a vehicle vibration equation of the bridge to be monitored, and an interaction force between the bridge to be monitored and the vehicle; The vibration model of the bridge to be monitored is expressed as: ； The vehicle vibration equation of the bridge to be monitored is expressed as: ； The interaction force between the bridge to be monitored and the vehicle is expressed as: ； Among them, V i is the vehicle speed, m i For vehicle quality, is the vehicle interval time, N is the total number of vehicles in the random traffic flow, w(x,t) is the displacement of the bridge at position x and time t, E is the Young's modulus of the beam material, I is the section moment of inertia of the beam, ρ is the density of the beam, A is the cross-sectional area of ​​the beam, q(x,t) is the distributed load along the beam length x and time t, y v (t) is the vertical displacement of the vehicle, C V is the damping coefficient of the vehicle, K V is the spring stiffness of the vehicle, g is the acceleration due to gravity, δ(xI t ) is the Dirac delta function; Inputting the random traffic flow samples into the bridge coupled vibration analysis model to obtain a random traffic flow bridge response deep learning training theory sample library; According to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, the dynamic response data and noise data of the bridge to be monitored are collected in real time, wherein the sensing data collected by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there is no vehicle passing is the noise data; The noise data and the random traffic flow bridge response deep learning training theory sample library are fused by a data fusion algorithm to obtain training samples; Performing neural network model training through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network; The dynamic response data is input into the vehicle dynamic weighing deep learning model, and the weight of the vehicle passing through the bridge to be monitored is output.

2. The vehicle dynamic weighing method based on neural network according to claim 1 is characterized in that: The step of collecting the dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored comprises: The dynamic displacement sensor is used to measure the displacement change of the bridge to be monitored when a vehicle passes by; The accelerometer is used to measure the acceleration response of the bridge to be monitored; The dynamic strain gauge is used to measure the strain change of the bridge to be monitored; The bridge type of the bridge to be monitored is obtained, and the minimum sampling frequency of the dynamic response data of the bridge to be monitored is set according to the bridge type.

3. The vehicle dynamic weighing method based on neural network according to claim 1 is characterized in that: The step of fusing the noise data with the random traffic flow bridge response deep learning training theory sample library through a data fusion algorithm to obtain training samples includes: Performing spectrum analysis, noise source identification processing and preprocessing on the noise data to obtain final noise data; The Kalman filtering algorithm is used to fuse the final noise data and the deep learning training theoretical sample library of random traffic bridge response, and the estimated value of the response of the bridge to be monitored is dynamically updated. The Bayesian estimation method is used to fuse the final noise data and the deep learning training theoretical sample library of random traffic bridge response to update and correct the probability model.

4. The vehicle dynamic weighing method based on neural network according to claim 1 is characterized in that: The step of training the neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network includes: The loss function of the vehicle dynamic weighing deep learning model includes data error and physical error. The data error is expressed as: , used to measure the error between the deflection of the vehicle dynamic weighing deep learning model and the measured deflection, where N is the number of data points, x i and t i are the position and time of the ith data point, respectively. is the output deflection, W measured is the measured deflection; The physical error is expressed as: , which is used to measure whether the output deflection of the vehicle dynamic weighing deep learning model meets the physical constraints, where N is the number of data points, E is the Young's modulus of the beam material, and I is the section moment of inertia of the beam.

5. The vehicle dynamic weighing method based on neural network according to claim 1 is characterized in that: After the step of inputting the dynamic response data into the vehicle dynamic weighing deep learning model and outputting the weight of the vehicle passing through the bridge to be monitored, the method further includes: Collecting the periodic dynamic response data of the bridge to be monitored according to a preset period, fusing the periodic dynamic response data with the training samples to obtain periodic training samples; Generate adversarial samples based on the periodic training samples; The data error and adversarial error of the vehicle dynamic weighing deep learning model are optimized according to the adversarial sample.

6. A vehicle dynamic weighing device based on neural network, characterized in that: The device comprises: An acquisition module, used to acquire the passing vehicle information of the bridge to be monitored, and generate a random traffic flow sample according to the passing vehicle information; The passing vehicle information includes vehicle type ratio distribution, vehicle weight distribution, vehicle speed distribution and vehicle distance distribution, determining the parameter distribution of the passing vehicle information, wherein the vehicle type ratio distribution is represented by a polynomial distribution, the vehicle weight distribution is represented by a normal distribution or a uniform distribution, the vehicle speed distribution is represented by a normal distribution, and the vehicle distance distribution is represented by an exponential distribution, and generating a random traffic flow sample of the passing vehicle information based on a random number generation method; Establishing a module for establishing a bridge coupling vibration analysis model of the bridge to be monitored based on a finite element method, and establishing a random traffic flow bridge response deep learning training theory sample library according to the random traffic flow samples and the bridge coupling vibration analysis model; The bridge coupling vibration analysis model includes a vibration model of the bridge to be monitored, a vehicle vibration equation of the bridge to be monitored, and an interaction force between the bridge to be monitored and the vehicle; The vibration model of the bridge to be monitored is expressed as: ； The vehicle vibration equation of the bridge to be monitored is expressed as: ； The interaction force between the bridge to be monitored and the vehicle is expressed as: ； Among them, V i is the vehicle speed, m i For vehicle quality, is the vehicle interval time, N is the total number of vehicles in the random traffic flow, w(x,t) is the displacement of the bridge at position x and time t, E is the Young's modulus of the beam material, I is the section moment of inertia of the beam, ρ is the density of the beam, A is the cross-sectional area of ​​the beam, q(x,t) is the distributed load along the beam length x and time t, y v (t) is the vertical displacement of the vehicle, C V is the damping coefficient of the vehicle, K V is the spring stiffness of the vehicle, g is the acceleration due to gravity, δ(xI t ) is the Dirac delta function; Inputting the random traffic flow samples into the bridge coupled vibration analysis model to obtain a random traffic flow bridge response deep learning training theory sample library; A collection module, used for collecting dynamic response data and noise data of the bridge to be monitored in real time according to the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored, wherein the sensing data collected by the dynamic displacement sensor, accelerometer and dynamic strain gauge installed on the bridge to be monitored when there is no vehicle passing is the noise data; A fusion module, used to fuse the noise data with the random traffic bridge response deep learning training theory sample library through a data fusion algorithm to obtain a training sample; A training module, used to train a neural network model through the training samples to obtain a vehicle dynamic weighing deep learning model based on a physical information neural network; An output module is used to input the dynamic response data into the vehicle dynamic weighing deep learning model and output the weight of the vehicle passing through the bridge to be monitored.

7. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the vehicle dynamic weighing method based on a neural network as described in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the vehicle dynamic weighing method based on a neural network according to any one of claims 1 to 5 are implemented.

Images