Bridge structure flexibility prediction method and system based on quality disturbance strategy under natural traffic flow

Abstract

The invention discloses a bridge structure flexibility prediction method and system based on a quality disturbance strategy under natural traffic flow, and the method comprises the steps: arranging a vibration testing environment, carrying out the environment vibration testing of a to-be-predicted bridge structure under sparse and dense traffic flow, and measuring the acceleration response of the to-be-predicted bridge structure; for the acceleration response measured under the sparse traffic flow, identifying the basic modal parameters of the to-be-predicted bridge structure through the CMIF; for the acceleration response measured under the dense traffic flow, adopting a parameter optimization-based variational mode decomposition method to identify the instantaneous frequency of the bridge structure to be predicted; carrying out vehicle-mounted space-time distribution identification according to the acquired traffic flow distribution condition on the bridge floor, calculating the real-time position of a vehicle, and distributing the real-time position of the vehicle to each acceleration sensor by adopting a load equivalent distribution principle to obtain an additional mass matrix; according to the basic modal parameters, the instantaneous frequency and the additional mass matrix, the vibration mode scaling coefficient is calculated, the flexibility is obtained through further calculation, and the method can be widely applied to the flexibility recognition process of structural health detection.

Description

Technical Field

[0001] The invention relates to a method and system for predicting the flexibility of a bridge structure based on a mass disturbance strategy under natural traffic flow, and belongs to the technical field of structural flexibility prediction. Background Art

[0002] In the field of bridge structure health detection, flexibility is a very important parameter of bridge structure, and structural flexibility identification plays a key role in the process of bridge bearing capacity assessment. At present, a large number of studies have used environmental vibration test methods to output basic structural parameters such as frequency and vibration mode, and have also been applied to the health monitoring of many bridge structures. However, at this stage, the method of using only environmental vibration cannot identify the amplitude of the structural frequency response function, so the environmental vibration test method alone cannot directly support the flexibility identification of bridge structures. According to the research of Zhang et al., the amplitude of the frequency response function identified by the environmental vibration test is in a certain proportional relationship with the actual frequency response function amplitude, and this proportional relationship can be expressed by a scaling factor.

[0003] In environmental vibration testing, the calculation of scaling factor is the key to accurately identifying structural flexibility. In recent years, Parloo et al. proposed a vibration test method based on mass change strategy. In the field test, multiple concrete blocks or stationary trucks are used as additional mass to act on the bridge. The scaling factor is calculated by the change of modal parameters, that is, the structural natural frequency, so as to identify the structural flexibility. This method has the advantage of not requiring the measurement of input force, but the bridge needs to be closed during the loading process and it is time-consuming and labor-intensive, and the implementation process is relatively difficult.

[0004] In view of the shortcomings of the above methods, which require closing bridges and are time-consuming and labor-intensive, a large number of studies have used dynamic identification systems of vehicles and bridges to carry out bearing capacity testing and bridge structure maintenance. Dan et al. collaboratively processed and analyzed the data of the WIM system and multi-camera machine vision to extract and identify the characteristics and attributes of moving loads. Hou et al. described in detail the fusion method of the WIM system and bridge monitoring data. They combined the vehicle load data measured by the WIM system with the bridge monitoring data to analyze the response and behavior of the bridge under different load conditions. Wang et al. simulated the impact of actual traffic loads on suspension bridges by establishing traffic flow models and load conversion models, discussed the impact of different parameters and assumptions on the simulation results, and proposed reasonable models and method selection. Guo et al. proposed a method based on probability and statistical analysis to determine the traffic load pattern on the bridge by collecting and analyzing actual traffic flow data. This method takes into account factors such as different vehicle types, vehicle speeds, and lane distribution, and describes the changes in traffic loads with a reasonable probability distribution. However, there is no research on applying the identification of the spatiotemporal distribution of traffic flow to structural flexibility identification.

[0005] Under natural traffic flow, the vehicle will act as a mass block to disturb the dynamic parameters of the bridge structure, and the structural frequency will change in real time. In the process of calculating the scaling coefficient, it is crucial to accurately identify the real-time changes of the structural frequency. For the above-mentioned time-varying system parameter identification problem, the current methods are mainly divided into two categories: one is to segment the structure in the time domain based on the short-time invariance assumption, such as short-time Fourier transform (STFT), Wigner-Ville distribution (WVD), wavelet transform (WT) or time series model and subspace model, etc. This type of method has a relatively complete theoretical basis, and the identification accuracy is easy to guarantee, but it cannot take into account the identification efficiency at the same time. The second is based on the concept of pseudo-modal parameters, expressing the structural response as a superposition of multi-component time-varying signals and performing an overall decomposition, and then identifying the instantaneous modal parameters of the obtained single-component signal. This type of method has high computational efficiency and the component signals have certain physical meanings, but the theoretical basis is not yet perfect. A typical example is the Hilbert-Huang transform (HHT), which first uses empirical mode decomposition to decompose the response signal into each intrinsic mode function (IMF), and then performs Hilbert transform on the decomposed IMF to obtain the instantaneous frequency of each mode. This method is mainly based on empirical decomposition and lacks a theoretical basis, and there are problems such as mode aliasing and endpoint effect. Dragomiretskiy et al. proposed a variational mode decomposition (VMD) method, which iteratively searches for the optimal solution of the variational model in a completely non-recursive manner to determine the frequency center and bandwidth of each component, and separates the component signals of limited bandwidth from the frequency domain. The VMD method has a strict theoretical basis, does not have the problem of mode aliasing, has strong noise resistance, and the decomposed components have a certain degree of quasi-orthogonality, but the parameters required for its decomposition: mode number, quadratic penalty factor, noise tolerance, and convergence error limit need to be preset in advance.

[0006] In general, structural flexibility is an important indicator for bridge health monitoring. The current structural flexibility identification methods still have the following shortcomings: (1) Traditional impact vibration methods or truck static load mass disturbance methods require closing the bridge and are time-consuming and labor-intensive, while traffic flow spatiotemporal distribution identification technology has not been introduced into structural flexibility identification; (2) For the structural frequency changes caused by mass disturbance under natural traffic flow, the current algorithm cannot take into account accuracy, noise resistance and practicality. Therefore, it is necessary to design a new structural flexibility identification method to enable bridge flexibility to be directly identified through environmental vibration testing. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a bridge structure flexibility prediction method and system based on a mass disturbance strategy under natural traffic flow, apply the traffic spatiotemporal distribution identification technology to the structural flexibility identification, which can accurately locate the real-time position of the bridge traffic and calculate the real-time frequency of the structure, greatly reducing the difficulty of testing, and without the need to close the bridge, a more convenient and faster monitoring process can be achieved.

[0008] The present invention adopts the following technical solutions to solve the above technical problems:

[0009] The bridge structure flexibility prediction method based on mass disturbance strategy under natural traffic flow includes the following steps:

[0010] Step 1: Install a number of speed bumps on the bridge deck of the bridge structure to be predicted for environmental vibration testing, and install a number of acceleration sensors at the same intervals on the bridge deck to measure the structural response under environmental vibration;

[0011] Step 2: Perform an environmental vibration test on the bridge structure to be predicted under sparse traffic flow, measure the acceleration response of the bridge structure to be predicted, and identify the basic modal parameters of the bridge structure to be predicted by the modal identification method CMIF, including the natural circular frequency, vibration mode and damping information;

[0012] Step 3: Perform an environmental vibration test on the bridge structure to be predicted under dense traffic flow, measure the acceleration response of the bridge structure to be predicted, and collect the traffic distribution on the bridge deck;

[0013] Step 4: Based on the acceleration response measured by the environmental vibration test under dense traffic flow, the variational mode decomposition method based on parameter optimization is used to identify the instantaneous frequency of the bridge structure to be predicted;

[0014] Step 5: Identify the vehicle-borne spatiotemporal distribution according to the collected traffic distribution on the bridge deck, calculate the real-time position of the vehicle on the bridge deck, and distribute the real-time position of the vehicle to each acceleration sensor using the load equivalent distribution principle to obtain the additional mass matrix of the bridge structure to be predicted;

[0015] Step 6: Calculate the mode scaling coefficient according to the basic modal parameters, instantaneous frequency and additional mass matrix of the bridge structure to be predicted, substitute the mode scaling coefficient into the scaled modal flexibility matrix, and further calculate the flexibility of the bridge structure to be predicted.

[0016] A bridge structure flexibility prediction system based on mass disturbance strategy under natural traffic flow, the system comprising: an environment layout module, an environment vibration test module, a modal identification module, an instantaneous frequency identification module, a wheel identification and resolution module, a vehicle template matching algorithm module, a scaling coefficient calculation module and a scaling flexibility calculation module; wherein,

[0017] An environment arrangement module is used to install a number of speed bumps on the bridge deck of the bridge structure to be predicted, install a number of acceleration sensors at the same intervals on the bridge deck, and arrange cameras for capturing video images of vehicles on the bridge deck;

[0018] Environmental vibration test module, used to perform environmental vibration tests on the bridge structure to be predicted under sparse traffic flow and dense traffic flow, and measure the acceleration response of the bridge structure to be predicted through an acceleration sensor;

[0019] The modal identification module is used to obtain the basic modal parameters of the bridge structure to be predicted by using the modal identification method CMIF based on the acceleration response measured by the environmental vibration test under sparse traffic;

[0020] The instantaneous frequency identification module is used to identify the acceleration response measured by the environmental vibration test under dense traffic flow, using the variational mode decomposition method based on parameter optimization to obtain the instantaneous frequency of the bridge structure to be predicted;

[0021] The wheel recognition and identification module is used to identify all vehicles and corresponding license plate numbers in the image based on the vehicle video image captured by the camera;

[0022] The vehicle template matching algorithm module is used to determine the real coordinates of the wheels of each identified vehicle in the bridge deck coordinate system, and to determine whether the vehicles identified in different frame images are the same vehicle, so as to obtain the additional mass matrix of the bridge structure to be predicted;

[0023] A scaling factor calculation module is used to calculate the vibration mode scaling factor according to the basic modal parameters, instantaneous frequency and additional mass matrix of the bridge structure to be predicted;

[0024] The scaling flexibility calculation module is used to substitute the vibration mode scaling coefficient into the scaling modal flexibility matrix, and further calculate the flexibility of the bridge structure to be predicted.

[0025] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0026] 1. The present invention performs environmental vibration testing based on mass disturbance, and does not need to clearly determine the magnitude of the input force, thus greatly reducing the difficulty of the test.

[0027] 2. The present invention incorporates the technology of identifying the spatiotemporal distribution of traffic flow, and there is no need to close the bridge during bridge structure monitoring, which can achieve a more convenient and faster monitoring process.

[0028] 3. The variational modal decomposition method based on parameter optimization proposed in the present invention can solve the problem that the modal number and the quadratic penalty factor of the traditional variational modal decomposition method are difficult to determine, thereby making the identification accuracy of the instantaneous modal parameters higher. Experimental verification shows that the flexibility identification method of the present invention has high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of the bridge structure flexibility prediction method based on the mass disturbance strategy under natural traffic flow of the present invention;

[0030] Figure 2 It is a diagram of the vehicle-mounted distribution identification process of the present invention;

[0031] Figure 3 It is a diagram of the instantaneous frequency identification process of the present invention;

[0032] Figure 4 It is a layout diagram of the experiment site of the embodiment of the present invention;

[0033] Figure 5 It is a diagram of the identification process of modal parameters and vibration modes under sparse traffic flow of the present invention;

[0034] Figure 6 is a conversion diagram between the image coordinate system and the bridge deck world coordinate system of the embodiment of the present invention;

[0035] Figure 7 This is the vehicle distribution diagram on the bridge deck at the 12th second under the dense traffic flow test of the embodiment of the present invention;

[0036] Figure 8 is a nodal force equivalent diagram of the wheel load in an embodiment of the present invention;

[0037] Fig. 9 1 is a graph showing the variation of the third-order instantaneous frequency over time according to an embodiment of the present invention, wherein (a) is the first-order instantaneous frequency, (b) is the second-order instantaneous frequency, and (c) is the third-order instantaneous frequency;

[0038] Fig.10 1 is a diagram of the flexibility recognition accuracy prediction of an embodiment of the present invention, wherein (a) is the scaled flexibility matrix, and (b) is the comparison between the deflection prediction value and the measured value. DETAILED DESCRIPTION

[0039] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0040] like Figure 1 As shown, the present invention proposes a bridge structure flexibility prediction method based on mass disturbance strategy under natural traffic flow, and the specific steps are as follows:

[0041] (1) Conduct environmental vibration tests on the predicted bridge structure under sparse traffic flow, measure the structural acceleration response, and identify basic modal parameters, including frequency, damping ratio, vibration mode, etc., by using a modal identification method (the present invention adopts the CMIF method).

[0042] (2) A camera is set up above the bridge structure to be predicted to conduct environmental vibration tests under dense traffic flow and simultaneously collect images of the traffic position on the bridge.

[0043] (3) The acceleration response collected from the environmental vibration test under dense traffic flow is used to identify the instantaneous frequency of the structure using the optimized variational mode decomposition (VMD) method.

[0044] Since traffic movement is not constant in the time and space domains, the VMD method is proposed to identify the instantaneous dynamic characteristics of the structure when applying different dynamic characteristic parameters caused by different traffic flow states. In view of the problem that the modal number and quadratic penalty factor of the traditional variational mode decomposition method are difficult to determine, the present invention proposes a variational mode decomposition method based on parameter optimization to identify the instantaneous dynamic characteristics of the structure. VMD can decompose a real-valued acceleration signal into K intrinsic mode functions (IMFs) with frequency-amplitude modulation characteristics and sparse characteristics. The single-component signal obtained by variational mode decomposition has the property of quasi-orthogonality, and the following orthogonality indicators are used for split component identification:

[0045]

[0046] In the formula, u k is the k-th order component obtained by VMD decomposition, res is the residual component, and n is the number of discrete sampling points. In normal decomposition, the orthogonality index corresponding to each component approaches 0, and when a certain order mode is split, its corresponding orthogonality index will increase significantly. The false noise component is identified by calculating the ratio of the energy of each order IMF component to the total energy. The energy ratio of the false noise component is much smaller than that of the effective component. For the split components, the energy of the split components of the same order mode needs to be increased. Assume that the ratio of the energy of the k-th order component to the total energy is defined as follows:

[0047]

[0048] In the formula, u k,i is the i-th value of the k-th order component. Figure 3 As shown in the figure, the split components and noise components generated by over-decomposition can be identified through orthogonality index and energy ratio, so as to obtain the optimal mode number. Based on the optimal mode number, the variational energy entropy (VEE) of different quadratic penalty factors is calculated. Different quadratic penalty factors correspond to different center frequency bandwidths, which in turn cause changes in the total variational energy entropy. When the total variational energy entropy is minimized, the components of each order are best separated, and the corresponding quadratic penalty factor is the optimal value. Finally, the Hilbert-Huang transform (HHT) is used to identify the instantaneous frequency of the separated single-component signal.

[0049] (4) Identify the spatiotemporal distribution of vehicle loads under densely structured traffic conditions, calculate the real-time position of vehicles on the bridge, and distribute the real-time position of the traffic flow to each collection node using the load equivalent distribution principle.

[0050] In the actual bridge test, static load is applied to the structure, which is cumbersome and time-consuming. If the temporal and spatial distribution information of the traffic flow on the bridge deck can be fully utilized, the temporal and spatial distribution of the traffic flow at each moment can be regarded as a static load condition, and the known vehicle weight information, the position of the bridge deck vehicle load and the corresponding deflection information of the underbridge structure can be substituted into the formula to determine the strength of the environmental excitation force, and then the structural flexibility can be identified without closing the traffic. This method can reduce the consumption of human, material and other resources in static load tests. Figure 2 As shown, the present invention adopts the following method to perform vehicle-mounted spatiotemporal distribution recognition:

[0051] (1) Using Yolov5s network for vehicle recognition

[0052] The present invention uses Yolov5s network to identify the license plate frame from the segmented vehicle image, and LPRNet to identify the license plate number from the license plate area identified by Yolov5s. The LPRNet deep learning network is a real-time, high-quality, license plate recognition network structure that supports variable-length license plates. The structure does not require pre-segmentation of license plate characters, and has sufficiently good robustness and is not affected by various camera parameters, viewing angles, lighting, etc. In order to avoid unnecessary repeated calculations, the license plate recognition program for a vehicle is only executed when the mask of a vehicle is completely within the ROI.

[0053] (2) Identify wheel position by transforming the coordinate system

[0054] The camera is calibrated using the pixel coordinate information of the feature points on the image and the known world space coordinate information. The pixel coordinate information of the feature points on the image is the pixel coordinates of the image point (x s ,y s ). The world coordinate system O w -x w y w z w The plane is placed on the bridge deck, and the origin of the coordinate system is O w Placed at the leftmost feature point on the image, y w The direction is the longitudinal extension direction of the bridge deck, i.e. the direction of vehicle travel, x w The direction is the lateral extension direction of the bridge deck, perpendicular to the y w Direction. w = 0, any characteristic point on the bridge deck satisfies the following equation:

[0055]

[0056] Where R = [r1r2r3] is the rotation matrix, t is the translation matrix, K is the camera intrinsic parameter matrix, (x s ,y s ) is the pixel coordinate of the image point, (x w ,y w ,z w ) is the world coordinate of the corresponding point. Then a homography mapping relationship can be established between the point on the bridge deck and its corresponding image point, which can be expressed as:

[0057]

[0058] Among them, H is the homography matrix, α * is an unknown proportional coefficient. It can be seen that at least 4 pairs of known spatial point and image point coordinates are required in a single photo to calculate the homography matrix H. Substituting the above formula into the pixel coordinates of the wheel in the picture, its real coordinate position on the bridge deck can be determined.

[0059] (3) Vehicle tracking based on correlation coefficient

[0060] The vehicle detection method based on Yolov5s can only detect all vehicles in each frame independently, but cannot determine whether the vehicles detected in different frames of the traffic video are the same vehicle. The present invention uses a template matching method based on correlation coefficient to track vehicles. For two vehicle sub-images, the correlation coefficient is calculated as follows:

[0061]

[0062] Where f is the grayscale function in the reference image (previous frame), and g is the grayscale function in the target image (next frame). m is the grayscale mean in the reference image, g m is the grayscale mean in the target image, and both can be calculated by the following formula:

[0063]

[0064] The correlation coefficient between the vehicle sub-image identified in each frame and the vehicle sub-image in the previous frame is calculated, and the correlation coefficient is used to determine whether the two vehicle sub-images are the same vehicle. The specific process includes the following three steps:

[0065] ① Convert all recognized vehicle sub-images in the next frame to a uniform size. The vehicle sub-images in the previous frame have been converted to a uniform size in the previous process.

[0066] ② Calculate the correlation coefficient of each vehicle sub-image in the next frame with all vehicle sub-images in the previous frame to form a correlation matrix. The number of rows in the correlation matrix is ​​equal to the number of vehicle sub-images in the previous frame, and the number of columns is equal to the number of vehicle sub-images in the next frame.

[0067] ③ Find the peak value of each row in the correlation matrix. The vehicles represented by the rows and columns where the peak value is located are the same vehicle.

[0068] It should be noted that: in order to deal with the unequal number of rows and columns of the correlation matrix, a threshold needs to be set, and the points greater than the threshold are extreme points. If the number of rows in the correlation matrix is ​​greater than the number of columns, if the number of rows in the matrix is ​​greater than the number of columns, and there is no extreme point (a point greater than the threshold) in one row, it means that the vehicle in the previous frame did not appear in the next frame. If the number of rows in the correlation matrix is ​​less than the number of columns, and there is no extreme point (a point greater than the threshold) in one column, it means that the vehicle in the next frame did not appear in the previous frame, that is, the vehicle is a new vehicle.

[0069] (5) Substitute the parameters identified or calculated above into the scaling factor calculation formula to calculate the vibration mode scaling factor α r .

[0070] For a bridge structure with N degrees of freedom, under proportional damping, the relationship between its eigenvalue and eigenvector is as follows:

[0071] [K]{δ r}=λ r [M]{φ r}

[0072] In the formula, {φ r} is the normalized vibration mode of the rth order mass, λ r is the square of the r-th order natural frequency.

[0073] The system matrix of the structure is the structural mass m k function, then by taking the derivative of the above formula we can get:

[0074]

[0075] Multiply both sides of the above equation by the transposed matrix of the eigenvector {φ r} T ,get:

[0076]

[0077] According to the orthogonality condition {φ r} T [M]{φ r}=1 and the characteristic equation {φ r} T [K]-λ r {φr} T [M] = 0, the above formula can be simplified to:

[0078]

[0079] According to the above formula, the first-order derivative of the eigenvalue to the quality can be obtained as:

[0080]

[0081] Assume that the change in mass does not affect the stiffness distribution of the structure, that is The above formula is simplified to:

[0082]

[0083] The displacement mode shape {ψ r} and mass normalized vibration mode {φ r} is represented as:

[0084] {φ r}=α r {ψ r}

[0085] Among them, α r is the scaling factor of the rth-order vibration mode. After derivation, the first-order sensitivity of the eigenvalue to the structural mass is:

[0086]

[0087] If additional masses of different weights are simultaneously introduced into the p degrees of freedom of the structure, the eigenvalue of the structure with additional mass can be expressed as the superposition of the eigenvalue of the structure without additional mass, that is:

[0088]

[0089] In the formula, λ 0r is the structural characteristic before adding mass, λ 1r is the characteristic value of the structure after adding mass, Δm k is the mass attached to the kth degree of freedom of the structure;

[0090] If only the first-order sensitivity is considered, the above formula can be obtained:

[0091]

[0092] Rearranging the above formula, the expression of the structural vibration mode scaling coefficient can be obtained as follows:

[0093]

[0094] The relationship between eigenvalue and natural frequency Substituting, we can get:

[0095]

[0096] In the formula, ω 0r is the natural circular frequency of the bridge structure without additional mass, ω 1r is the natural circular frequency of the bridge structure when the mass is added, ψ kr is the value of the r-th order unnormalized displacement mode in degree of freedom k identified from the environmental vibration test data; [ΔM] is the additional mass matrix, Δm k is the kth element on the main diagonal, and there are p elements on the main diagonal.

[0097] (6)α r After the value is obtained, it is substituted into the formula to calculate the scaled flexibility matrix, and the structural deflection prediction can be carried out.

[0098] If the mass normalized vibration mode is obtained, the frequency response function matrix of the structure can be expressed as:

[0099]

[0100] Among them, C r =1 / (2jω r ) is a coefficient related to the natural frequency of the structure, It is C r The complex conjugate of r is the r-th order pole of the system, Related to the structural natural frequency and damping ratio; is r The complex conjugate of Substituting the expression of the normalized vibration mode of mass into the above formula, we can obtain:

[0101]

[0102] Take the frequency response function at ω 0r The value of can be used to obtain the scaled modal flexibility matrix of the structure:

[0103]

[0104] It can be seen that the scaled displacement flexibility expression based on the mass perturbation strategy is related to the basic modal parameters of the structure identified from the initial structure under ambient vibration, such as frequency, damping and unscaled displacement mode shape, the basic modal parameters of the structure after the additional mass under ambient excitation, and the distribution of the additional mass.

[0105] Example

[0106] (1) Experimental arrangement

[0107] like Figure 4 As shown, an environmental vibration test was conducted on a simply supported beam in the laboratory to further verify the effectiveness of the scaling flexibility identification method proposed in the present invention. The steel beam is made of 25a hot-rolled channel steel with a length of 5.3 meters. A vehicle speed bump is set on it for environmental vibration testing. The steel beam is simply supported on a steel support. 14 Donghua test piezoelectric accelerometers, model 1A113E, are installed at equal intervals on the simply supported beam to measure the structural response under environmental vibration. The NIPXIe-1082 data acquisition system is used with a sampling frequency of 1000Hz.

[0108] At the same time Figure 4 As shown in the figure, two cameras are placed near the bridge to synchronously collect the traffic distribution on the bridge deck and the displacement response of the main beam. The resolution of the camera for monitoring vehicles is 3072x2048 pixels, and the synchronous acquisition frequency is 20Hz. The resolution of the camera for monitoring structural response is 2592x2048 pixels.

[0109] (2) Environmental vibration test under sparse traffic

[0110] First, an environmental vibration test is carried out under sparse traffic. A load-free vehicle is driven on the beam to excite the bridge. Several speed bumps are installed on the bridge deck to simulate the unevenness of the bridge deck and increase the impact vibration effect.

[0111] For the environmental vibration test results under sparse traffic, first extract the acceleration response of each node, select any node as the reference point to calculate the cross-correlation function, perform Fourier transform on the obtained cross-correlation function, and obtain the frequency response function; then call the CMIF algorithm to identify the basic modal parameters. The CMIF method is a modal identification algorithm based on frequency domain data. It performs singular value decomposition on the displacement frequency response function of all measuring points of the structure. By observing the peak value of the extracted characteristic singular value curve, the natural frequency, vibration shape and damping information of the structure can be picked up. Calculated by the CMIF algorithm, the frequencies of the first three orders of the steel beam are 6.57Hz, 25.58Hz and 55.45Hz respectively. And draw the first three order vibration shape diagrams of the steel beam, as shown Figure 5 and save the data.

[0112] (3) Environmental vibration testing under dense traffic flow and identification of spatiotemporal distribution of traffic flow

[0113] Environmental vibration test under dense traffic flow, the experiment is arranged with the same coefficient traffic conditions, four load-bearing vehicles are driven on the beam to excite the bridge, and several speed reduction belts are installed on the bridge deck to simulate the unevenness of the bridge deck to increase the impact vibration effect. The camera resolution used to synchronously collect the distribution of traffic flow on the bridge deck is 3072x2048 pixels, and the synchronous collection frequency is 20Hz.

[0114] ① Training of Yolov5s recognition network

[0115] Before using Yolov5s for vehicle detection, it needs to be trained to obtain a good vehicle recognition model. The training platform is a host with GPU (NVIDIA GeForce RTX 2070SUPER) and CPU (Intel Core i7-9700K@3.60GHz). The Yolov5s network was trained for 100 rounds with the data set, with batch=16 in each batch and an initial learning rate of 0.01. The trained network can be used to recognize the traffic video on the model, and can recognize all vehicles and give accurate masks. At the same time, the time used to recognize each picture is about 20ms, so about 50 pictures can be recognized within 1s.

[0116] ②Conversion between image coordinate system and bridge deck coordinate system

[0117] Light targets are arranged at the node positions on the bridge deck, which can be used to calibrate the image coordinate system and the bridge deck coordinate system. The camera is calibrated using the pixel coordinate information of the feature points on the image and the known actual space coordinate information. The pixel coordinate information of the feature points on the image is the pixel coordinate of the image point (x s ,y s ). The world coordinate system O w -x w y w z w The plane is placed on the bridge deck, then z w = 0. Establish the bridge deck coordinate system and image coordinate system as follows Figure 6 shown.

[0118] The bridge deck light targets are selected as reference points and named R1 to R23. The image coordinates and bridge deck coordinates corresponding to 20 reference points (R1 to R20) are selected for calibration, and the homography matrix H is determined according to the formula. The results are as follows:

[0119]

[0120] The remaining three reference points (R21-R23) are used to verify the accuracy of the above image coordinate to bridge deck coordinate transformation relationship. The results are shown in Table 1. It can be seen that the maximum error of the three reference points in the bridge width direction is 2%, and the maximum error in the bridge length direction is 0.85%, which fully meets the accuracy requirements of wheel alignment.

[0121] Table 1 Verification of coordinate transformation results of reference points

[0122]

[0123] ③ On-vehicle spatiotemporal distribution based on information fusion

[0124] The printed license plates were pasted on the front of the vehicles and weighed. The license plates and vehicle weights of the four vehicles were as follows: Sichuan FH7JA4 yellow truck weighed 21.785kg; Beijing PF9520 yellow truck weighed 15.785kg; Yun E6UA4Y yellow truck weighed 12.015kg; Ning AF0236 gray truck weighed 13.34kg. The vehicle weight information was associated with each vehicle according to the license plate number.

[0125] At a certain moment, the wheel recognition result is used to fit the wheel mask by ellipse, and the lowest point of the ellipse is taken as the contact point between the wheel and the bridge deck. Combined with the recognition result of the above homography matrix, the specific position of the wheel on the bridge deck can be obtained.

[0126] The wheel position results are shown in Table 2:

[0127] Table 2 Wheel alignment results at t = 12s

[0128]

[0129]

[0130] According to the identified traffic flow trajectory and license plate information, the temporal and spatial distribution of the vehicles on the bridge deck at each moment can be obtained by associating the two. Taking the result at t = 12s as an example, the vehicle distribution result at this moment is as follows: Figure 7 shown.

[0131] Since the wheel force does not act directly on the node, the static equivalence principle is used to equate the vehicle load by a group of forces acting on the node. Taking a three-axle vehicle as an example, its axle weight and equivalent load are as follows: Figure 8 As shown, the equivalent distribution is performed as follows:

[0132] P i +P j =W,P i x=P j (Lx)

[0133] Among them, x represents the distance from the equivalent node on the left end of the axle weight, W represents the axle weight of a certain axis, L represents the distance between the two nodes, and P i and P j The node force distribution results are shown in Table 3.

[0134] Table 3 Node force distribution results (N)

[0135] node 3 4 5 6 7 8 9 10 11 Mass(kg) 1.52 9.60 2.22 8.13 3.89 4.51 11.28 6.22 15.56

[0136] (4) Identification of instantaneous frequency based on parameter optimization

[0137] First, the optimal modal number K and quadratic penalty factor α are calculated using the acceleration response measured by environmental vibration. The parameter optimization algorithm proposed in the invention is used to perform VMD pre-decomposition on the obtained acceleration response to identify the optimal modal number K and quadratic penalty factor α. The initial value of K is set to 8, and α is set to the default value of 2000; the thresholds μ1 and μ2 are set to the mean of the orthogonality index and the energy ratio, respectively. The order of the split components is determined by judging the orthogonality index of each component, and then the noise component is excluded by combining the time domain and frequency domain information, so the optimal K value corresponding to the measured acceleration response is 7. The optimal modal number K = 7 is obtained, the optimization range of α is preset to [100, 10000], the step size is set to 5, and the variational energy entropy corresponding to each α is calculated. It can be obtained that when α = 6830, the variational energy entropy has a minimum value, that is, the optimal α value is 6830.

[0138] After determining the optimal modal number K and the quadratic penalty factor α, the acceleration data can be Fourier transformed to calculate the spectrum, and then VMD decomposition is performed to obtain the IMF component. The Hilbert-Huang transform (HHT) is used to identify the instantaneous frequency of the separated single component signal. Finally, the identified instantaneous frequency is fitted by eliminating the standard deviation. The identified third-order instantaneous frequency is as follows: Fig. 9 As shown in (a)-(c).

[0139] (5) Calculation of scaling factor and identification of structural flexibility

[0140] After the above steps, the identified parameters, including the frequency of the structure under sparse traffic flow, the instantaneous frequency and vehicle distribution at the same time under dense traffic flow, can be brought into the scaling factor calculation formula to calculate the scaling factor α r , the values ​​of the first three scaling coefficients are 0.3526, 0.1347, and 0.1258 respectively. Then the scaling coefficients are substituted into the flexibility calculation formula to calculate the true flexibility of the structure.

[0141] For comparison, the deflection curve measured in the static load test is plotted together with the deflection prediction curve. Fig.10 As shown in (a)-(b), it can be seen that the flexibility identified by the method proposed in the present invention can well predict the deformation of the structure, greatly reduce the difficulty of the test, and reduce the consumption of manpower and material resources.

[0142] Based on the above prediction method, the present invention also proposes a bridge structure flexibility prediction system based on mass disturbance strategy under natural traffic flow, the system comprising:

[0143] An environment arrangement module is used to install a number of speed bumps on the bridge deck of the bridge structure to be predicted, install a number of acceleration sensors at the same intervals on the bridge deck, and arrange cameras for capturing video images of vehicles on the bridge deck;

[0144] Environmental vibration test module, used to perform environmental vibration tests on the bridge structure to be predicted under sparse traffic flow and dense traffic flow, and measure the acceleration response of the bridge structure to be predicted through an acceleration sensor;

[0145] The modal identification module is used to obtain the basic modal parameters of the bridge structure to be predicted by using the modal identification method CMIF based on the acceleration response measured by the environmental vibration test under sparse traffic;

[0146] The instantaneous frequency identification module is used to identify the acceleration response measured by the environmental vibration test under dense traffic flow, using the variational mode decomposition method based on parameter optimization to obtain the instantaneous frequency of the bridge structure to be predicted;

[0147] The wheel recognition and identification module is used to identify all vehicles and corresponding license plate numbers in the image based on the vehicle video image captured by the camera;

[0148] The vehicle template matching algorithm module is used to determine the real coordinates of the wheels of each identified vehicle in the bridge deck coordinate system, and to determine whether the vehicles identified in different frame images are the same vehicle, so as to obtain the additional mass matrix of the bridge structure to be predicted;

[0149] A scaling factor calculation module is used to calculate the vibration mode scaling factor according to the basic modal parameters, instantaneous frequency and additional mass matrix of the bridge structure to be predicted;

[0150] The scaling flexibility calculation module is used to substitute the vibration mode scaling coefficient into the scaling modal flexibility matrix, and further calculate the flexibility of the bridge structure to be predicted.

[0151] Based on the same inventive concept, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the aforementioned method for predicting the flexibility of a bridge structure based on a mass disturbance strategy under natural traffic flow are implemented.

[0152] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the aforementioned bridge structure flexibility prediction method based on the mass disturbance strategy under natural traffic flow.

[0153] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0154] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0155] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0156] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0157] The above embodiments are only for illustrating the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. A bridge structure flexibility prediction method based on mass disturbance strategy under natural traffic flow, characterized by: The steps include: Step 1: Install a number of speed bumps on the bridge deck of the bridge structure to be predicted for environmental vibration testing, and install a number of acceleration sensors at the same intervals on the bridge deck to measure the structural response under environmental vibration; Step 2: Perform an environmental vibration test on the bridge structure to be predicted under sparse traffic flow, measure the acceleration response of the bridge structure to be predicted, and identify the basic modal parameters of the bridge structure to be predicted by the modal identification method CMIF, including the natural circular frequency, vibration mode and damping information; Step 3: Perform an environmental vibration test on the bridge structure to be predicted under dense traffic flow, measure the acceleration response of the bridge structure to be predicted, and collect the traffic distribution on the bridge deck; Step 4: Based on the acceleration response measured by the environmental vibration test under dense traffic flow, the variational mode decomposition method based on parameter optimization is used to identify the instantaneous frequency of the bridge structure to be predicted; Step 5: Identify the vehicle-borne spatiotemporal distribution according to the collected traffic distribution on the bridge deck, calculate the real-time position of the vehicle on the bridge deck, and distribute the real-time position of the vehicle to each acceleration sensor using the load equivalent distribution principle to obtain the additional mass matrix of the bridge structure to be predicted; Step 6: Calculate the mode scaling coefficient according to the basic modal parameters, instantaneous frequency and additional mass matrix of the bridge structure to be predicted, substitute the mode scaling coefficient into the scaled modal flexibility matrix, and further calculate the flexibility of the bridge structure to be predicted.

2. The method according to claim 1, characterized in that: The specific process of step 2 is as follows: A load-free vehicle is driven on the bridge structure to be predicted to perform bridge vibration excitation and measure the acceleration response of the bridge structure to be predicted. Each acceleration sensor is taken as a node, and the acceleration response of each node is extracted. An arbitrary node is selected as a reference point to calculate the cross-correlation function, and the cross-correlation function is subjected to Fourier transform to obtain the frequency response function. The frequency response function is identified using the modal identification method CMIF to obtain the basic modal parameters of the bridge structure to be predicted.

3. The method according to claim 1, characterized in that: The specific process of step 3 is as follows: Four load-bearing vehicles are used to drive on the bridge structure to be predicted to perform bridge vibration, measure the acceleration response of the bridge structure to be predicted, and use cameras to collect video images of vehicles on the bridge deck.

4. The method according to claim 1, characterized in that: The specific process of step 4 is as follows: Step 3.1, initialize the modal number K and the quadratic penalty factor α, and use the variational mode decomposition method to decompose the acceleration response measured by the environmental vibration test under dense traffic flow to obtain the intrinsic mode component IMF; Step 3.2: Calculate the orthogonality index I based on the intrinsic mode component IMF o , judge I o Is it less than the first threshold μ1? If so, proceed to step 3.3; otherwise, keep the α value unchanged, update the K value and return to step 3.1 to decompose again; Step 3.3, calculate the energy ratio E of the split component and the noise component calculated based on the orthogonality index r , judge E r Is it less than the second threshold μ2? If so, proceed to step 3.4; otherwise, keep the α value unchanged, update the K value and return to step 3.1 to decompose again; Step 3.4 will simultaneously make I o is less than the first threshold μ1 and E r The K value that is less than the second threshold μ2 is taken as the optimal mode number, and the optimization range and step size of the α value are set; Step 3.5, keeping the optimal modal number unchanged, within the optimization range, decompose the acceleration response measured by the environmental vibration test under dense traffic flow using the variational mode decomposition method every other step to obtain the intrinsic mode component IMF at the α value every other step; Step 3.6, based on the intrinsic mode component IMF at every other step value of α, calculate the corresponding variational energy entropy value, and select the α value that minimizes the variational energy entropy value as the optimal quadratic penalty factor; Step 3.7, perform Fourier transform on the acceleration response measured in the environmental vibration test under dense traffic to obtain a spectrum, and decompose the spectrum using the variational mode decomposition method under the optimal mode number and the optimal quadratic penalty factor to obtain a single component; Step 3.8, using Hilbert-Huang transform to identify the instantaneous frequency of the single component obtained in step 3.7, eliminating the outliers in the identified instantaneous frequency, and obtaining the instantaneous frequency of the bridge structure to be predicted.

5. The method according to claim 3, characterized in that: The specific process of step 5 is as follows: Step 5.1, based on the collected traffic distribution on the bridge deck, i.e., the bridge deck vehicle video image, all vehicles and corresponding license plate numbers are identified from each frame of the bridge deck vehicle image using the pre-trained Yolov5s vehicle recognition model; Step 5.2, by converting the image coordinate system and the bridge deck coordinate system, determine the real coordinates of each vehicle's wheel in the bridge deck coordinate system; the details are as follows: The world coordinate system O w -x w y w z w The plane is placed on the bridge deck, with the origin O w is the feature point in the upper left corner of the bridge deck vehicle image, i.e., the first point along the vehicle travel direction, y w The direction is the longitudinal extension direction of the bridge deck, i.e. the direction of vehicle travel, x w The direction is the lateral extension direction of the bridge deck, perpendicular to the y w Direction, when z w = 0, any characteristic point on the bridge deck satisfies: In the formula, α * is the proportionality coefficient, (x s ,y s ) is the pixel coordinate of the feature point on the image, K is the camera internal parameter matrix, R = [r1r2r3] is the rotation matrix, t is the translation matrix, (x w ,y w ,z w ) is the coordinate of the space point in the world coordinate system corresponding to the feature point on the image, then the homography mapping relationship between the feature point on the image and the space point corresponding to the feature point is established, which is expressed as follows: Where H is the homography matrix, h 11 、h 12 、h 13 、h 21 、h 22 、h 23 、h 31 and h 32 All are elements of the homography matrix; the homography matrix is ​​obtained through at least 4 pairs of known spatial points and pixel coordinate points, and the real coordinates of the wheel in the bridge deck coordinate system are determined according to the homography matrix and the pixel coordinates of the wheel in the image; Step 5.3, based on the correlation coefficient, determine whether the vehicles identified in different frame images are the same vehicle. The calculation formula of the correlation coefficient is as follows: In the formula, represents the correlation coefficient, f(x,y) is the grayscale function of the reference image, i.e. the previous frame image, g(x′,y′) is the grayscale function of the target image, i.e. the next frame image, and f m is the grayscale mean of the reference image, g m is the grayscale mean of the target image, and M is half the length of the image side; Calculate the correlation coefficient between the vehicle sub-image identified in the current frame image and the vehicle sub-image identified in the previous frame image, and determine whether the two vehicle sub-images are the same vehicle based on the size of the correlation coefficient. Specifically: 1) Convert the vehicle sub-image identified in the current frame image into a uniform size, where the uniform size is a pre-set size; 2) Calculate the correlation coefficient between each vehicle sub-image identified in the current frame image and all vehicle sub-images in the previous frame to form a correlation matrix. The number of rows in the correlation matrix is ​​equal to the number of vehicle sub-images in the previous frame, and the number of columns is equal to the number of vehicle sub-images in the current frame. 3) Find the peak value of each row in the correlation matrix. The row and column where the peak value is located represent the same vehicle. 4) Preset threshold: if the number of rows of the correlation matrix is ​​greater than the number of columns, and there are rows whose correlation coefficients are all less than the preset threshold, it means that the vehicle corresponding to the row whose correlation coefficients are all less than the preset threshold has left the current frame; If the number of rows of the correlation matrix is ​​less than the number of columns, and there are columns whose correlation coefficients are all less than the preset threshold, it means that the vehicles corresponding to the columns whose correlation coefficients are all less than the preset threshold are the newly added vehicles in the current frame.

6. The method according to claim 1, characterized in that: In step 6, the calculation formula of the vibration mode scaling coefficient is as follows: In the formula, α r is the vibration mode scaling factor, ω 0r is the natural circular frequency of the bridge structure without additional mass, ω 1r is the natural circular frequency of the bridge structure when the mass is added, ψ kr is the value of the rth order unnormalized displacement mode in degree of freedom k identified from the environmental vibration test data, [ΔM] is the additional mass matrix, ψ r is the rth order unnormalized displacement mode vector identified from the environmental vibration test data; Δm k is the kth element on the main diagonal, and there are p elements on the main diagonal; The scaled modal flexibility matrix is ​​calculated as follows: Where F is the scaled flexibility, m is the identified modal order, and C r =1 / (2jω r ), C r is the coefficient related to the natural frequency of the bridge structure to be predicted, C r The complex conjugate of r is the r-th order pole of the bridge structure system, is r The complex conjugate of ξ r is the damping ratio of the bridge structure, ω r is the natural frequency of the bridge structure, and j is an imaginary unit.

7. Bridge structure flexibility prediction system based on mass disturbance strategy under natural traffic flow, characterized by: The system includes: an environment arrangement module, an environment vibration test module, a modal identification module, an instantaneous frequency identification module, a wheel identification and resolution module, a vehicle template matching algorithm module, a scaling coefficient calculation module and a scaling flexibility calculation module; wherein, An environment arrangement module is used to install a number of speed bumps on the bridge deck of the bridge structure to be predicted, install a number of acceleration sensors at the same intervals on the bridge deck, and arrange cameras for capturing video images of vehicles on the bridge deck; Environmental vibration test module, used to perform environmental vibration tests on the bridge structure to be predicted under sparse traffic flow and dense traffic flow, and measure the acceleration response of the bridge structure to be predicted through an acceleration sensor; The modal identification module is used to obtain the basic modal parameters of the bridge structure to be predicted by using the modal identification method CMIF based on the acceleration response measured by the environmental vibration test under sparse traffic; The instantaneous frequency identification module is used to identify the acceleration response measured by the environmental vibration test under dense traffic flow, using the variational mode decomposition method based on parameter optimization to obtain the instantaneous frequency of the bridge structure to be predicted; The wheel recognition and identification module is used to identify all vehicles and corresponding license plate numbers in the image based on the vehicle video image captured by the camera; The vehicle template matching algorithm module is used to determine the real coordinates of the wheels of each identified vehicle in the bridge deck coordinate system, and to determine whether the vehicles identified in different frame images are the same vehicle, so as to obtain the additional mass matrix of the bridge structure to be predicted; A scaling factor calculation module is used to calculate the vibration mode scaling factor according to the basic modal parameters, instantaneous frequency and additional mass matrix of the bridge structure to be predicted; The scaling flexibility calculation module is used to substitute the vibration mode scaling coefficient into the scaling modal flexibility matrix, and further calculate the flexibility of the bridge structure to be predicted.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor executes the computer program, the steps of the bridge structure flexibility prediction method based on the mass disturbance strategy under natural traffic flow as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the bridge structure flexibility prediction method based on mass disturbance strategy under natural traffic flow as described in any one of claims 1 to 6 are implemented.

Images