Method for predicting vehicle load effect extremum of large-span concrete beam bridge and application

Abstract

The invention relates to a large-span concrete beam bridge vehicle load effect extreme value prediction method and application, and the method comprises the steps: classifying vehicles according to the number of axles, and obtaining an edge distribution function F of each traffic flow load parameter by taking the axle loads and lane vehicle distances of different vehicle types as the traffic flow load parameters; in consideration of vehicle axle load correlation, edge distribution functions of all axle loads of the same type of vehicle models are combined to obtain a joint distribution function H of the axle loads; based on the joint distribution function of the axle load and the edge distribution function of the lane distance, sampling and establishing a random traffic flow load sample set considering the axle load correlation; and establishing a traffic flow load parameter-bridge response extreme value mapping relation model so as to obtain a bridge response extreme value of the current random traffic flow load sample and corresponding extreme value distribution. By establishing the mapping relation between the traffic flow load parameter and the bridge response extreme value, the response extreme value of the traffic flow load passing through the bridge at any moment to the bridge can be quickly and accurately calculated.

Description

Technical Field

[0001] The invention belongs to the field of health monitoring and safety assessment of civil engineering structures, and proposes a method and application for predicting extreme values ​​of vehicle load effects on long-span concrete beam bridges, which are used for assessing vehicle load effects on long-span concrete beam bridges. Background Art

[0002] Bridge structures are an important part of the road traffic system, especially long-span concrete beam bridges, whose operational safety is directly related to the normal operation of the traffic network. During the use of bridges, vehicle loads are the most important external loads, which can easily cause damage to bridge structures. With the development of the economy, transportation has become more convenient, especially with the increasing number of heavy vehicles, and overloading has become more serious. Bridge damage caused by vehicle loads has become an increasingly important issue. In recent years, incidents of partial damage or even collapse of bridges caused by overweight vehicles have occurred frequently.

[0003] The current method is to use finite element software (such as Abaqus or Madis Civil) to build a bridge model, substitute vehicle load data and use the software to calculate the response. This is a common method for calculating the response of bridge structures caused by external loads, but this method takes a long time to calculate, especially for random analysis. Therefore, it is crucial to quickly and accurately calculate the response of vehicle loads to bridge structures.

[0004] Zhao Hui et al. (Zhao Hui, Chen Shuisheng, Li Jinhua, et al. Prediction of the extreme value of bridge vehicle-induced displacement under random vehicle loads [J]. Journal of Harbin Engineering University, 2023, 44(02): 219-225.) calculated the bridge vehicle-induced displacement under random vehicle loads. When generating random traffic flow, the axle weight of each axle is proportionally distributed according to the total vehicle weight, without considering the correlation of axle weights. In general, it is assumed that different axle weights of the same vehicle are independent of each other, ignoring the influence of the correlation between the connecting axle and axle weight. Especially for multi-axle vehicles (such as six-axle vehicles), the mutual influence between axle weights is more obvious. Assuming that the axle weights are independently distributed cannot better reflect the actual axle weight of the vehicle.

[0005] Zong Zhouhong et al. (Zong Zhouhong, Xue Cheng, Yang Zegang, et al. Automobile load model of Jiangsu Province highway bridges based on WIM [J]. Journal of Southeast University (Natural Science Edition), 2020, 50(01): 143-152.) established a automobile load model of Jiangsu Province highway bridges and calculated the load effects of the bridges. The calculation amount was huge and it took 45 days to calculate using 4 computers.

[0006] In order to make the traffic flow model more suitable for actual vehicles and accurately and quickly calculate the response of vehicle load to bridges, the method of the present invention first improves the accuracy of the random traffic flow model by considering the correlation of axle weight, calculates the influence line of the bridge by load test, instead of extracting it by finite element software modeling, and then establishes a mapping relationship model between traffic flow load parameters and bridge response extreme values. The required calculated traffic flow load is substituted into the model to directly obtain the bridge response extreme value, without the need to use finite element software for complex calculations, thereby improving the calculation efficiency. Summary of the invention

[0007] The present invention aims to propose a method and application for predicting the extreme value of vehicle load effect on long-span concrete beam bridges. The method takes into account the correlation between vehicle axle weights, constructs a mapping relationship between traffic load parameters and bridge response extreme values, and realizes the rapid calculation of the extreme value response and probability distribution of the bridge structure under the action of random traffic flow. The vehicle load effect of long-span concrete beam bridges is then evaluated.

[0008] The technical solution of the present invention is:

[0009] In a first aspect, the present invention provides a method for predicting the extreme value of vehicle load effect on a long-span concrete beam bridge, the process of the method is:

[0010] Step 1: Obtain vehicle data measured by the bridge dynamic weighing system, classify vehicles according to the number of axles, use the axle weight and lane distance of different vehicle models as traffic load parameters, the number of traffic load parameters is the sum of the axle weights of all vehicle models + the number of lane distances, use the non-parametric kernel density fitting estimation method to determine the probability density distribution of each traffic load parameter, and then obtain the marginal distribution function F of each traffic load parameter; considering the correlation of vehicle axle weights, the marginal distribution functions of all axle weights of the same vehicle model are combined according to the following formula to obtain the joint distribution function H of the axle weight;

[0011] H(x 1 ,…,x n )=Φ ρ (Φ -1 (x 1 ),…,Φ -1 (x n ))(F1(x 1 ),…F n (x n ))

[0012] Where: n is the number of axle weights in a certain type of vehicle, Φ v is the distribution function of the multivariate standard normal distribution with the correlation coefficient matrix v; Φ -1 Represents the inverse function of the distribution function of the standard normal distribution; x 1 represents the sample of the first axle weight; x nrepresents the sample of the nth axle weight;

[0013] Based on the joint distribution function of axle weight and the marginal distribution function of lane distance, a random traffic load sample set considering axle weight is established by sampling.

[0014] Step 2: Obtain a reasonable value range of axle weight and lane distance, and generate a traffic load sample set using stratified sampling method within the reasonable value range;

[0015] Step 3, obtaining the influence line of each mid-span section of the bridge through load test;

[0016] Step 4: Load the traffic load sample set in step 2 onto the influence line and calculate the traffic load response extreme value data set for the bridge; use these calculated response extreme value data sets and the corresponding traffic load sample set to establish the mapping relationship model between traffic load parameters and bridge response extreme values:

[0017]

[0018] In the formula, S is the extreme value of the response; M is the number of polynomials; r m is the coefficient; m (X) is a polynomial function; X is the traffic load parameter, corresponding to the axle weight and lane distance of all types of axles; ε is a function with a mean of 0 and a variance of σ 2 Normal distribution of

[0019] For each coefficient r m Introduce the prior distribution, construct the likelihood function, calculate the posterior distribution, and define the coefficient of variation ξ as:

[0020]

[0021] In the formula, x new is a new variable point, the axle weight of all axles of all types and the lane distance are a traffic load sample, and each traffic load sample is a point; S * , are the new variable points x new The mean and variance of the predicted response at ; is the prior variance; μ N is the posterior mean; σ N 2 is the posterior variance;

[0022] Each time, the first l1 data with the smallest coefficient of variation are selected to update the mapping relationship model until the model converges to obtain a convergent mapping relationship model;

[0023] Step 5: Input the random traffic load sample set obtained in step 1 considering the axle weight into the convergence mapping relationship model of step 4, that is, obtain the bridge response extreme value of the current random traffic load sample, and then further fit to obtain the corresponding extreme value distribution.

[0024] Furthermore, the method is used in the evaluation of vehicle load effects on long-span concrete beam bridges: the maximum permissible response value, i.e., the threshold value, is determined based on national or regional bridge design specifications and standards combined with statistical analysis of historical monitoring data;

[0025] Combined with the extreme value distribution of the response obtained in step 5, the probability of exceeding the threshold is calculated. This probability is the failure probability. The vehicle load effect on the bridge is evaluated based on the failure probability. If the failure probability is greater than 5%, it is determined that the vehicle load causes the bridge structure to fail.

[0026] Furthermore, the traffic load sample set generated in step 2 is defined as A q , select the initial traffic load sample l and the corresponding bridge response extreme value to train the mapping relationship model of traffic load parameter-bridge response extreme value;

[0027]

[0028] Where, l is the number of initial traffic load samples; j is the number of random traffic load parameters;

[0029] Then the remaining traffic load sample A q -l, substitute into In the calculation of S * and Take l1 traffic load samples with the smallest coefficient of variation and add them to the initial traffic load sample l, which is l+l1. Substitute l+l1 into Update the traffic load parameter-bridge response extreme value relationship model; then calculate the remaining traffic load sample A q -l+l1 corresponding to S * and Then take l1 traffic load samples with the smallest coefficient of variation and add them to the initial traffic load sample l, and repeat the above process until the model converges.

[0030] Preferably, the condition for model convergence is: ξ is less than or equal to 2.

[0031] In a second aspect, the present invention provides a vehicle load effect evaluation device for a long-span concrete beam bridge, the device comprising:

[0032] Vehicle information collection unit: This unit is based on a dynamic weighing system and is equipped with high-precision sensors and signal receivers. It is responsible for real-time collection of axle weight and lane distance information of vehicles crossing the bridge. It pre-processes data through integrated chips and digital signal processors, removes abnormal values, and stores the processed data in a storage module connected to an external interface.

[0033] Random traffic flow generation unit: uses the vehicle parameter information transmitted by the vehicle information acquisition unit to generate a random traffic flow load sample set considering the axle weight;

[0034] Mapping relationship establishment unit: obtain the influence line of each mid-span section of the bridge through load test, consider the reasonable value range of axle weight and lane spacing, and use stratified sampling method to generate traffic load sample set within the reasonable value range; calculate the bridge response extreme value corresponding to the traffic load sample set; establish the mapping relationship between traffic load parameters and bridge response extreme value in the form of polynomial, substitute the random traffic load sample data generated by the random traffic generation unit considering axle weight into the mapping relationship to calculate the bridge response extreme value and obtain the bridge response extreme value distribution;

[0035] Evaluation unit: Set the maximum response value allowed by the bridge as the threshold, find the probability of exceeding the threshold in the extreme value distribution of the bridge response obtained by using the mapping relationship, and this probability is the failure probability. The vehicle load effect of the bridge is evaluated by the failure probability;

[0036] Control unit: The control unit is responsible for adjusting the input frequency and time nodes of vehicle information, including hour, day, month, quarter and year.

[0037] Compared with the prior art, the present invention has the following beneficial effects: first, the axle weight correlation of vehicles is taken into account. The present invention no longer performs independent sampling of axle weights, but considers the sampling of the joint distribution of vehicle axle weights; second, fast extreme value calculation. By establishing a mapping relationship between traffic load parameters and bridge response extreme values, the response extreme value of the traffic load on the bridge at any time passing through the bridge can be quickly and accurately calculated. The traditional finite element analysis method takes about 20 days to calculate a random traffic load sample one million times, while after using this method, it only takes 50 minutes to obtain the response value. Third, the result is accurate. When observing the extreme value data corresponding to the 95% quantile of the extreme value distribution of the bridge response, the error between the finite element calculation data and the traditional method is only 0.7%, showing the significant advantages of this method in accuracy and efficiency.

[0038] The device of the present invention combines advanced sensor technology, digital signal processing and high-performance computing to provide a comprehensive and accurate bridge response extreme value rapid calculation capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1The flowchart of the method for predicting the extreme value of vehicle load effect on long-span concrete beam bridge is shown in FIG.

[0040] Figure 2 It is the probability density fitting diagram of the axle weight of the first axle of the six-axle vehicle;

[0041] Figure 3 It is the probability density fitting diagram of the axle weight of the second axle of the six-axle vehicle;

[0042] Figure 4 It is the probability density fitting diagram of the axle weight of the third axle of the six-axle vehicle;

[0043] Figure 5 It is the probability density fitting diagram of the axle weight of the fourth axle of the six-axle vehicle;

[0044] Figure 6 It is the probability density fitting diagram of the axle weight of the fifth axle of the six-axle vehicle;

[0045] Figure 7 It is the probability density fitting diagram of the axle weight of the sixth axle of the six-axle vehicle;

[0046] Figure 2 to Figure 7 The probability density curve is fitted according to the axle weight data in order to integrate the probability density and obtain the marginal distribution of each axle weight.

[0047] Figure 8 The fifth and sixth axle weights of a six-axle vehicle are sampled for correlation. After considering the correlation of axle weights, the axle weights are sampled to obtain the axle weights of the same vehicle in the vehicle load data. Taking the fifth and sixth axle weights of a six-axle vehicle as an example, it can be seen from the figure that the axle weights are not dispersed, but have a certain linear relationship.

[0048] Figure 9-10 This is a comparison of two methods for calculating the probability density of bridge deflection and stress extremes. MCS is calculated using the finite element method. As can be seen from the figure, the distributions obtained by the two methods are very consistent, indicating that the results of the new method are reliable.

[0049] Fig.11 It is a schematic diagram of the block unit structure of the device of the present invention.

[0050] Fig.12 It is a schematic diagram of the equipment connection of the hardware required for the device of the present invention. DETAILED DESCRIPTION

[0051] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0052] The bridge dynamic weighing system in the present invention is a device installed on the bridge entity to collect information about vehicles passing the bridge, which is a well-known technology in the art.

[0053] The present invention uses a non-parametric kernel density estimation method to estimate the probability density of vehicle axle weight and lane distance, instead of using normal distribution, gamma distribution, generalized extreme value distribution, etc. to fit the data, so that the fitting curve fits the actual data very well. Figure 2 to Figure 7 .

[0054] Example 1

[0055] The method for predicting the extreme value of vehicle load effect on a long-span concrete beam bridge in this embodiment comprises the following specific steps:

[0056] Step 1: Establish a random traffic load sample set considering axle weight

[0057] (1) Collect vehicle load information on the bridge within one year, including vehicle type, axle weight, wheelbase, lane distance, etc. In this embodiment, the axle weight of the vehicle and the distance between vehicles in each lane (i.e., lane distance) are used to represent the random traffic load sample.

[0058] The vehicles are classified according to the number of axles, and divided into two-axle vehicles, three-axle vehicles, four-axle vehicles, five-axle vehicles and six-axle vehicles, a total of five categories, and outliers in the data are removed, such as abnormal axle weights and lane distances. In this embodiment, five-axle vehicles account for a small proportion and are not considered.

[0059] (2) Fit the probability density functions of the axle weight and lane spacing of each type of vehicle (classified by the number of axles). Each type of vehicle has a single axle weight with a probability density function. Integrate each probability density function to obtain its corresponding cumulative probability function, which is the marginal distribution function of each traffic load parameter.

[0060] Probabilistic statistical analysis was performed on 18 traffic load parameters, including axle weights of all vehicle models (the number of axle weights of different vehicle models is added together, i.e., 2+3+4+6, a total of 15 parameters) and lane distances (a total of 3 parameters for three lanes in this embodiment). The probability density distribution of axle weights and lane distances were obtained using the non-parametric kernel density fitting estimation method. The lane distance is divided by lane, i.e., the distance between vehicles traveling on each lane.

[0061]

[0062] In the formula, For x j The probability density function at , e is the number of corresponding vehicle types or the number of spacings on the corresponding lanes, h is the bandwidth, K(·) is the kernel function, and this embodiment uses the Gaussian kernel function, i.e. x j is the jth traffic load parameter variable, and its value is [1,18]; Represents the i-th sample of the j-th traffic load parameter (either the axle weight value or the lane distance value).

[0063] The marginal distribution function F(x j )for:

[0064]

[0065] The t in the integral function is an integral variable, which is used to represent the value of the integrand at different points during the integration process and is "accumulated" or "summed" after the integration is completed.

[0066] Considering the correlation of axle weights of various types of vehicles, the marginal distribution functions of the corresponding axle weights of the same type of vehicles are connected into a joint distribution function of axle weights, which is expressed as

[0067] H(x 1 ,…,x n )=Φ ρ (Φ -1 (x 1 ),…,Φ -1 (x n ))(F1(x 1 ),...F n (x n ))

[0068] Where: H is the n-dimensional joint distribution function, n is the number of axle weights in a certain type of vehicle, F n is the marginal distribution function of the nth axle weight of a certain type of vehicle, Φ ρ is the distribution function of the multivariate standard normal distribution with the correlation coefficient matrix ρ; Φ -1 Represents the inverse function of the standard normal distribution; ρ is a linear correlation coefficient matrix with all diagonal elements equal to 1.

[0069] Based on the joint distribution function of axle weight and the marginal distribution function of lane distance, a random traffic load sample set considering axle weight is established by sampling. Each random traffic load sample contains axle weight and the corresponding lane distance.

[0070] Step 2: Obtain a reasonable range of values ​​for axle weight and lane spacing, and use stratified sampling within this reasonable range to generate a traffic load sample set.

[0071] The stratified sampling method divides the reasonable value range of each variable (axle weight and lane spacing) into several equal intervals (layers), and stratified sampling is used to establish a traffic load sample set. If there are N samples in the traffic load sample set, the range of each variable is divided into N layers. For each variable, a sample point is randomly selected from the respective layer. This ensures that a sample point can be selected in each layer, avoiding the concentrated distribution of samples. The sample points of each variable in the traffic flow passing through the bridge are combined to form a sample matrix. Each row of this sample matrix represents a traffic load sample, and each column corresponds to a traffic load parameter.

[0072] Step 3: directly test the working state of the bridge structure through load test to obtain the deflection influence line and stress influence line of each mid-span section of the experimental test bridge;

[0073] Step 4: Load the traffic load sample set in step 2 along the two influence lines respectively. Every time the traffic load sample moves forward 0.1 meters, calculate the response of the bridge until the entire convoy (a traffic load sample) completely passes through the bridge. The maximum value of all calculated response values ​​is taken as the extreme value of the response of a traffic load sample to the bridge. Calculate the extreme value data of the response of all bridges corresponding to the traffic load sample set, and obtain the extreme value data set of the response of the bridge corresponding to the traffic load sample set.

[0074] The mapping relationship is established using the traffic load parameters and the response extreme values. Each influence line corresponds to a mapping relationship, and the output of the system is expressed as a polynomial combination of the input variables:

[0075]

[0076] It can also be written as: S = rΓ + ε

[0077] In the formula, S is the extreme value of the response; M is the number of polynomials; r m is the coefficient; m (X) is a polynomial function; X is the traffic load parameter; ε is a function with a mean of 0 and a variance of σ 2 Normal distribution; r is the weight vector to be determined, Γ is a Q×M matrix, Q rows represent Q traffic load samples, and each row has M basis functions.

[0078] By inputting the probability distribution of random variables, the probability distribution of the system output can be predicted through polynomial expansion, which can effectively evaluate the uncertainty and risk of the output. The direct calculation of the coefficient output is a point estimate of the regression coefficient, which does not directly reflect the uncertainty and may be affected by outliers, resulting in estimation bias. Therefore, in order to quantify the uncertainty of the model coefficient, the coefficient r m Probability estimation. For each coefficient r m Introducing the prior distribution,

[0079]

[0080] Where μ0 is the prior mean. is the prior variance, r is the coefficient, N represents the normal distribution, and p(r) represents the prior probability.

[0081] Based on the observed data and model output, the likelihood function of S is:

[0082]

[0083] Combining the prior distribution and the likelihood function, calculate the posterior distribution p(r|S,X):

[0084]

[0085] The quantification of uncertainty is achieved through the posterior distribution.

[0086] Posterior mean:

[0087]

[0088] Posterior variance:

[0089]

[0090] Get the variable point x new The predicted response mean S at * and its variance

[0091] S * =γ m (x new )μ N ,

[0092]

[0093] In the mapping relationship model of traffic load parameters and bridge response extreme values, the closer the sample data is to the test point, the greater the impact on the predicted value of the test point, while the farther the sample data is, the smaller the impact. If enough sample points are selected, the data with a great impact on the test point can be more comprehensively covered, thereby establishing a more accurate nonlinear relationship, but this will lead to lower efficiency. Therefore, in order to select the optimal data more efficiently, the ratio of the predicted response mean and variance is defined as the coefficient of variation ξ, which is used to characterize the degree of discreteness of the data. Each time a point is selected, the first l1 data with the smallest coefficient of variation are taken to update the mapping relationship model.

[0094]

[0095] The traffic load sample set generated in step 2 is defined as A q, select a certain initial traffic load sample l and the corresponding bridge response extreme value to train the mapping relationship model of traffic load parameter-bridge response extreme value, that is, in

[0096]

[0097] Where, l is the number of initial traffic load samples; j is the number of traffic load parameters;

[0098] Then the remaining traffic load sample A q -l, substitute into In the calculation of S * and Take l1 traffic load samples with the smallest coefficient of variation and add them to the initial traffic load sample l, which is l+l1. Substitute l+l1 into Update the traffic load parameter-bridge response extreme value relationship model; then calculate the remaining traffic load sample A q -l+l1 corresponding to S * and Then take l1 traffic load samples with the smallest coefficient of variation and add them to the initial traffic load sample l, and repeat the above process until the model converges. * +σ * With S * -σ * The closer the two are, the more they have reached convergence, or the convergence condition is met when ξ is defined to be less than or equal to 2.

[0099] Step 5: Input the random traffic load sample set considering axle weight into the convergence mapping relationship model of step 4, that is, obtain the bridge response extreme value of the current random traffic load sample, and then further fit to obtain the corresponding extreme value distribution.

[0100] Sample U1, U2, ... U from uniform distribution U(0, 1) d . (d is the number of random variables), U i Mapped to the original spatial data, for each marginal distribution F i , using its inverse function F i -1

[0101]

[0102] X iis the result of joint distribution sampling, and the generated data takes into account the axle load equivalence of the actual vehicle load. Repeated sampling can obtain a random traffic load sample set C2 that takes into account the axle load correlation. Substituting it into the convergence mapping relationship model of traffic load parameters-bridge response extreme values, the corresponding bridge response extreme values ​​can be obtained. The response extreme value data is fitted using non-parametric kernel density estimation to establish the bridge response extreme value distribution.

[0103] Step 6: Evaluate the effects of vehicle loads on the bridge.

[0104] According to the national or regional bridge design specifications and standards and combined with the statistical analysis of historical monitoring data, the maximum allowable response value - the threshold value is determined. Combined with the obtained extreme value distribution of the response, the probability of exceeding the threshold value is calculated. This probability is the failure probability, and then the vehicle load effect of the bridge is evaluated. If the failure probability is greater than 5%, it is determined that the vehicle load causes the bridge structure to fail.

[0105] When the calculated response extreme value is greater than the threshold, it indicates failure. The final calculated response extreme value is an extreme value distribution function. We can see how many values ​​of this distribution function are greater than the threshold. The ratio of these values ​​greater than the threshold to the total number is the failure probability.

[0106] Example 2

[0107] The vehicle data of Xingtai North Ring South-to-North Water Diversion Bridge in the past year is used to verify the effectiveness of this method. During the data preprocessing, the outliers of axle weight and lane distance are removed to ensure the accuracy of the verification results.

[0108] The details are as follows:

[0109] (1) The parameter variables of random traffic flow include axle weight and lane spacing. First, the vehicle types are classified according to the number of axles, and the axle weight probability model of vehicles with different axle numbers is established. The data is fitted using the nonparametric kernel density estimation method. Taking the Xingtai North Ring South-to-North Water Diversion Bridge as an example, six-axle vehicles account for 64.9%, and the axle weight correlation of six-axle vehicles is the strongest. The axle weight kernel density fitting results of six-axle vehicles are shown in Figure 2. Figure 2 to Figure 7 shown.

[0110] (2) Use Matlab to call the copulastat function to calculate the Spear of the axle weight m An rank correlation coefficient is used to establish a joint distribution function related to axle weight. The joint distribution and the marginal distribution function of lane distance are sampled to generate a random traffic load sample set considering axle weight correlation. Figure 8 The relevant axle weight sampling points for a six-axle vehicle are shown.

[0111] Table 1 shows the Spearman rank correlation coefficient of axle weights. The larger the value, the stronger the correlation. Taking a six-axle vehicle as an example, it can be seen from the table that the 5th and 6th axles have the strongest correlation, followed by the 4th and 5th axles. The closer the distance between axle weights, the stronger the correlation. This shows that the correlation of axle weights of actual vehicles exists and needs to be considered when establishing a random traffic load model.

[0112] Table 1 Spearman rank correlation coefficient of six-axle axle weight

[0113]

[0114] (3) After determining the reasonable value ranges of the parameters of axle weight and lane distance, stratified sampling is performed on the parameters. 1000 samples are drawn for each parameter to form sample library A. q . Calculate sample library A by influence line loading q The bridge response extreme value of each sample in the database B is obtained. q . Select A q and B q The initial sample points in are substituted into the model to establish the mapping relationship between traffic load parameters and bridge response extreme values. Each time, 50 sample points (i.e., l1=50) are selected from the remaining candidate samples and substituted into the training of the mapping relationship model. According to the calculated mean and variance, the coefficient of variation is introduced, and the point with the smallest value is added to the initial sample point for update iteration. After reaching the allowable error range, the mapping relationship is established and a converged mapping relationship model is obtained.

[0115] (4) Input the generated random traffic load sample set considering axle weight into the convergence mapping relationship model to obtain the extreme value distribution of the bridge response. Calculate the probability that the extreme value of the response exceeds the threshold, and then evaluate the vehicle load effect on the bridge. If the failure probability is greater than 5%, it is determined that the vehicle load causes the failure of the bridge structure. Compared with the traditional finite element calculation method, this method greatly shortens the time for calculating the bridge response, and the error of the extreme value of the 95% quantile of the extreme value distribution of the bridge response is controlled below 1%. In addition, this method realizes the selection of the optimal point for the training relationship, thereby improving the efficiency of model establishment. Fitting curve comparison Figure 9-10 shown.

[0116] Fig. 9 and Fig.10The results of calculating deflection and stress by the method of the present invention are compared with those of the traditional MCS calculation method (finite element software is used to establish a bridge model, a large number of vehicle samples are input, and then the corresponding response extreme value calculation is performed). It can be seen that the response extreme value distribution curves of the two results are very close, especially in the tail data. Taking deflection as an example, the mean of the bridge response extreme value distribution and the extreme value data corresponding to the 95% quantile are observed. The mean calculated by the finite element method is 0.9710, and the mean calculated by this method is 0.9677, with a difference of only 0.05; the extreme value data corresponding to the 95% quantile calculated by the finite element method is 1.4134, and the extreme value data corresponding to the 95% quantile calculated by this method is 1.4117, with a difference of 0.0017. For the extreme value response distribution of the bridge caused by vehicle loads, the tail data, that is, the data with large values, has a very important impact on the bridge, so this method improves the calculation efficiency while also ensuring the accuracy of the data.

[0117] For bridges, smaller deflections or stresses will not affect the structure of the bridge. It is the larger data that can damage the bridge structure. Therefore, this method calculates the extreme value distribution of the bridge response. In the extreme value distribution, the tail data is the larger data among all extreme values, and this part of the data has a great impact on the bridge. The calculation results of this method overlap well with the traditional finite element distribution calculation results in the tail data, indicating that the calculation results are very reliable based on the improvement of calculation efficiency.

[0118] Example 3

[0119] This embodiment is a vehicle load effect evaluation device for a long-span concrete beam bridge, comprising:

[0120] Vehicle information collection unit: This unit is based on a dynamic weighing system and is equipped with high-precision sensors and signal receivers. It is responsible for real-time collection of axle weight and lane distance information of vehicles crossing the bridge. Data preprocessing is performed through integrated chips and digital signal processors (DSPs), outliers are removed, and the processed data is stored in a storage module connected to an external interface to ensure efficient data management and subsequent analysis.

[0121] Random traffic flow generation unit: uses the vehicle parameter information transmitted by the vehicle information acquisition unit to generate a random traffic flow load sample set that takes into account the axle weight. This unit uses the processor to perform complex calculation tasks and provides the necessary input data for calculating the extreme values ​​of bridge response.

[0122] Mapping relationship establishment unit: obtain the influence line of each mid-span section of the bridge through load test, consider the reasonable value range of axle weight and lane spacing, and use stratified sampling method to generate traffic load sample set within the reasonable value range; calculate the bridge response extreme value corresponding to the traffic load sample set; establish the mapping relationship between traffic load parameters and bridge response extreme value in the form of polynomial, substitute the random traffic load sample data generated by the random traffic generation unit considering axle weight into the mapping relationship to calculate the bridge response extreme value and obtain the bridge response extreme value distribution;

[0123] Evaluation unit: Set the maximum response value allowed by the bridge as the threshold, find the probability of exceeding the threshold in the extreme value distribution of the bridge response obtained by using the mapping relationship, and this probability is the failure probability. The vehicle load effect of the bridge is evaluated by the failure probability;

[0124] Control unit: The control unit is responsible for adjusting the input frequency and time nodes of vehicle information, including multiple settings such as hour, day, month, quarter, year, etc. Through the external interface of the server terminal, users can adjust the input data in real time to meet different analysis needs and research purposes.

[0125] The pressure of vehicles on the bridge is transmitted to the bridge deck through the weight of each axle. To establish a random traffic load sample set, it is necessary to perform probability statistics on the marginal distribution of the axle weight data and sample from the marginal distribution to obtain samples. However, each axle weight in an actual vehicle does not exist independently, especially the coupling design of multi-axle vehicles, which makes the axle weights affect each other. If only the marginal distribution of each axle weight is considered separately for sampling, it cannot reflect the axle weight of the actual vehicle well, so the correlation of the axle weights needs to be considered. The method of the present invention uses a polynomial function to establish a mapping relationship between traffic load parameters and bridge response extremes, which can quickly calculate the bridge response extremes caused by vehicle loads, consider the correlation of vehicle axle weights, and improve the accuracy of the random traffic load model.

[0126] Any matters not described in the present invention are applicable to the prior art.

Claims

1. A method for predicting the extreme value of vehicle load effect on a long-span concrete beam bridge, characterized in that: The process of the method is: Step 1: Obtain vehicle data measured by the bridge dynamic weighing system, classify vehicles according to the number of axles, use the axle weight and lane distance of different vehicle models as traffic load parameters, the number of traffic load parameters is the sum of the axle weights of all vehicle models + the number of lane distances, use the non-parametric kernel density fitting estimation method to determine the probability density distribution of each traffic load parameter, and then obtain the marginal distribution function F of each traffic load parameter; considering the correlation of vehicle axle weights, the marginal distribution functions of all axle weights of the same vehicle model are combined according to the following formula to obtain the joint distribution function H of the axle weight; H(x 1 ,…,x n )=Φ ρ (Φ -1 (x 1 ),…,Φ -1 (x n ))(F1(x 1 ),…F n (x n )) Where: n is the number of axle weights in a certain type of vehicle, Φ ρ is the distribution function of the multivariate standard normal distribution with the correlation coefficient matrix ρ; Φ -1 Represents the inverse function of the distribution function of the standard normal distribution; x 1 represents the sample of the first axle weight; x n represents the sample of the nth axle weight; Based on the joint distribution function of axle weight and the marginal distribution function of lane distance, a random traffic load sample set considering axle weight is established by sampling. Step 2: Obtain a reasonable value range of axle weight and lane distance, and generate a traffic load sample set using stratified sampling method within the reasonable value range; Step 3, obtaining the influence line of each mid-span section of the bridge through load test; Step 4: Load the traffic load sample set in step 2 onto the influence line and calculate the traffic load response extreme value data set for the bridge; use these calculated response extreme value data sets and the corresponding traffic load sample set to establish the mapping relationship model between traffic load parameters and bridge response extreme values: In the formula, S is the extreme value of the response; M is the number of polynomials; r m is the coefficient; m (X) is a polynomial function; X is the traffic load parameter, corresponding to the axle weight and lane distance of all types of axles; ε is a function with a mean of 0 and a variance of σ 2 Normal distribution of For each coefficient r m Introduce the prior distribution, construct the likelihood function, calculate the posterior distribution, and define the coefficient of variation ξ as: In the formula, x new is a new variable point, the axle weight of all axles of all types and the lane distance are a traffic load sample, and each traffic load sample is a point; S * , are the new variable points x new The mean and variance of the predicted response at ; is the prior variance; μ N is the posterior mean; σ N 2 is the posterior variance; Each time, the first l1 data with the smallest coefficient of variation are selected to update the mapping relationship model until the model converges to obtain a convergent mapping relationship model; Step 5: Input the random traffic load sample set obtained in step 1 considering the axle weight into the convergence mapping relationship model of step 4, that is, obtain the bridge response extreme value of the current random traffic load sample, and then further fit to obtain the corresponding extreme value distribution.

2. The method according to claim 1, characterized in that The method is used in the evaluation of vehicle load effects on long-span concrete beam bridges: according to national or regional bridge design specifications and standards and combined with statistical analysis of historical monitoring data, the maximum permissible response value - the threshold value - is determined; Combined with the extreme value distribution of the response obtained in step 5, the probability of exceeding the threshold is calculated. This probability is the failure probability. The vehicle load effect on the bridge is evaluated based on the failure probability. If the failure probability is greater than 5%, it is determined that the vehicle load causes the bridge structure to fail.

3. The method according to claim 1, characterized in that The traffic load sample set generated in step 2 is defined as A q , select the initial traffic load sample l and the corresponding bridge response extreme value to train the mapping relationship model of traffic load parameter-bridge response extreme value; Where, l is the number of initial traffic load samples; j is the number of random traffic load parameters; Then the remaining traffic load sample A q -l, substitute into In the calculation of S * and Take l1 traffic load samples with the smallest coefficient of variation and add them to the initial traffic load sample l, which is l+l1. Substitute l+l1 into Update the traffic load parameter-bridge response extreme value relationship model; then calculate the remaining traffic load sample A q -l+l1 corresponding to S * and Then take l1 traffic load samples with the smallest coefficient of variation and add them to the initial traffic load sample l, and repeat the above process until the model converges.

4. The method according to claim 1, characterized in that: The condition for model convergence is: ξ is less than or equal to 2.

5. A vehicle load effect evaluation device for a long-span concrete beam bridge, characterized in that: The device comprises: Vehicle information collection unit: This unit is based on a dynamic weighing system and is equipped with high-precision sensors and signal receivers. It is responsible for real-time collection of axle weight and lane distance information of vehicles crossing the bridge. It pre-processes data through integrated chips and digital signal processors, removes abnormal values, and stores the processed data in a storage module connected to an external interface. Random traffic flow generation unit: uses the vehicle parameter information transmitted by the vehicle information acquisition unit to generate a random traffic flow load sample set considering the axle weight; Mapping relationship establishment unit: obtain the influence line of each mid-span section of the bridge through load test, consider the reasonable value range of axle weight and lane spacing, and use stratified sampling method to generate traffic load sample set within the reasonable value range; calculate the bridge response extreme value corresponding to the traffic load sample set; establish the mapping relationship between traffic load parameters and bridge response extreme value in the form of polynomial, substitute the random traffic load sample data generated by the random traffic generation unit considering axle weight into the mapping relationship to calculate the bridge response extreme value and obtain the bridge response extreme value distribution; Evaluation unit: Set the maximum response value allowed by the bridge as the threshold, find the probability of exceeding the threshold in the extreme value distribution of the bridge response obtained by using the mapping relationship, and this probability is the failure probability. The vehicle load effect of the bridge is evaluated by the failure probability; Control unit: The control unit is responsible for adjusting the input frequency and time nodes of vehicle information, including hour, day, month, quarter and year.

Images