Orthotropic steel bridge deck deflection deformation detection method based on influence surface

Abstract

This invention relates to the field of bridge structural health monitoring and safety assessment technology, and particularly to a method for detecting the deflection deformation of orthotropic steel bridge decks based on influence surfaces. The method includes: S1, establishing a layered spatiotemporal traffic flow information database specific to the orthotropic steel bridge deck to be tested, generating virtual loads based on time windows and vehicle tail reinforcement coefficients to simulate bridge traffic; S2, acquiring the static stress response of key bridge points when vehicles cross the bridge; S3, inverting the influence surface based on dimensional adaptive prior information, constructing a hybrid likelihood function, and reconstructing a two-dimensional influence surface matrix by integrating typical traffic and tail reinforcement information; S4, comparing the changes and responses before and after the influence surface to infer the deflection deformation. This application can generate virtual loads using traffic flow statistics and invert the entire field influence surface through stress measurement at a single key point; the innovative dimensional adaptive prior setting method significantly improves the inversion efficiency, accuracy, and stability.

Description

Technical Field

[0001] This invention relates to the field of bridge structural health monitoring and safety assessment technology, and in particular to a method for detecting the deflection deformation of orthotropic steel bridge decks based on influence surfaces. Background Technology

[0002] Orthotropic steel bridge decks are widely used in long-span bridges due to their advantages of being lightweight, high-strength, and quick to construct. However, this structure undergoes flexural deformation under wheel loads, exhibiting complex local stress distribution characteristics. In particular, stress concentration is prone to occur at the weld joints connecting the longitudinal ribs and the bridge deck, leading to the initiation of fatigue cracks.

[0003] Traditional impact surface identification methods suffer from the following limitations: methods based on one-dimensional beam lattice theory cannot accurately reflect the two-dimensional stress characteristics of orthotropic steel bridge decks; methods based on dense sensor arrays are costly to implement and have poor engineering applicability; existing inversion methods mostly rely on precise vehicle load information, which is difficult to reliably obtain in actual traffic monitoring. In recent years, some research has attempted to identify vehicle loads using computer vision technology, but this method is greatly affected by environmental conditions and requires separate model training for each bridge, resulting in poor universality. Furthermore, existing Bayesian inversion methods often use fixed, uninformative priors, failing to fully utilize the dimensional and order-of-magnitude information of the physical problem, leading to low inversion efficiency. Existing methods for simulating traffic loads often employ stationary stochastic assumptions or generalized load spectra, making it difficult to reflect the time-of-day differences (peak / night), weekday patterns (weekday / weekend), and extreme events at the tail of heavy loads in actual bridge service, resulting in conservative or inaccurate fatigue damage assessments. There is an urgent need for an orthotropic steel bridge deck influence surface identification method that does not require real-time vehicle identification, can be based on sparse measurements, and has an inversion process that is more in line with physical reality, so as to improve the authenticity of influence surface inversion in the assessment of fatigue hot spots caused by deflection deformation. Summary of the Invention

[0004] The technical problem to be solved by this invention is: in order to improve the accuracy of influence surface inversion in the assessment of fatigue hot spots caused by deflection deformation, this invention provides a method for detecting the deflection deformation of orthotropic steel bridge deck based on influence surface.

[0005] The technical solution adopted by this invention to solve its technical problem is: A method for detecting the deflection deformation of orthotropic steel bridge decks based on influence surfaces: S1. Establish a layered traffic flow spatiotemporal information database specific to the orthotropic steel bridge deck to be tested, and generate virtual loads based on time windows and vehicle rear enhancement coefficients to simulate bridge traffic. S2, Obtain the static stress response at key points of the bridge when a vehicle crosses the bridge. ; S3, based on dimensional adaptive prior inversion of the influence surface, constructs a hybrid likelihood function, and reconstructs a two-dimensional influence surface matrix by integrating typical traffic and tail reinforcement information; S4, compare the changes and responses before and after the influence to infer the amount of flexural deformation.

[0006] Furthermore, based on the hierarchical traffic flow spatiotemporal information database, virtual vehicle queues that conform to the actual traffic load characteristics of the bridge are generated in batches, including the axle load sequence of each virtual vehicle. and wheelbase This leads to the construction of a virtual axis weight matrix for inversion. .

[0007] Furthermore, the virtual axis weight matrix Including complete time-varying matrices and pure tail enhancement submatrix The complete time-varying matrix is ​​spliced ​​together according to the weight ratio of time windows to represent typical traffic, and the pure tail-reinforcement submatrix is ​​used for extreme response and fatigue hotspot assessment.

[0008] Further, in step S3, the two-dimensional influence surface of the bridge deck is discretized into... The grid is then rearranged into column vectors. The relationship between stress response and influence surface can be expressed as: , The system matrix is ​​determined by the virtual axle load matrix and the vehicle movement path. Let the observation error vector be assumed to follow a multivariate normal distribution. , The variance is unknown.

[0009] Furthermore, according to Bayes' theorem, the influence of surface vectors sum of error variance The posterior distribution is proportional to the product of the prior distribution and the likelihood function: , in, It is the likelihood function; It is the prior distribution; For the posterior distribution; each component of the influence surface vector Follows an independent normal prior distribution: Among them, the prior standard deviation It is adaptively determined by the following formula: ,in, It is the standard deviation of the stress response vector, which characterizes the fluctuation range of the response; It is the average of the absolute values ​​of the axle loads in the virtual axle load matrix, representing the typical level of the load; It is a scaling factor that adapts the range of the prior distribution to the order-of-magnitude ratio of the stress response to the load for a specific problem.

[0010] Furthermore, the mixture likelihood function is expressed as: , Where λ≈0.7~0.85, and They are respectively and The generated system matrix row vectors.

[0011] Furthermore, based on posterior samples The sum of calculations yields the maximum a posteriori estimate. and posterior mean : ; ; Calculate the 95% confidence interval for each influence surface grid point; finally, calculate the influence surface vector. Reconstruct a two-dimensional influence surface matrix and output the stress influence surface identification results of orthotropic steel bridge deck.

[0012] Further, in step S4, the deformation is reconstructed by convolving the changed influence surface with the typical load to reconstruct the response, and then compared with the initial response; initial reference response: Where q(y) represents the load distribution, For the initial system matrix, The reference stress influence surface for a bridge in a healthy and intact state; ,in, Approximation Instead, if there is no real-time load, the same path is assumed. This represents the static stress response value of the bridge at time t during its service life. The system matrix of the bridge at time t during its service life; This represents the reference stress influence surface at time t during the bridge's service life.

[0013] Furthermore, a linear method is used to predict the deformation increment: deflection increment. Actual deformation The load amplification factor is the stress-deflection conversion constant, determined through calibration; relative deformation rate: .

[0014] Furthermore, a method based on combined measured-simulation calibration is used to estimate the deformation increment, based on the initial estimated value of the deformation increment. and based on and The simulated displacement values ​​of the displacement check points were obtained. Collect the actual measured displacement values ​​of the displacement check points. Calculate the calibration coefficient: The stress response is corrected using calibration coefficients: Based on the calibrated parameters, the deformation increment is recalculated: .

[0015] The beneficial effects of this invention are that it does not require real-time vehicle identification, but only traffic flow statistics to generate virtual loads, which greatly reduces the technical threshold. Furthermore, it generates a virtual vehicle load matrix that conforms to real traffic characteristics by introducing a Monte Carlo simulation method with a tail enhancement coefficient.

[0016] By deploying a single stress sensor on the outer side of the connection between the channel rib and the bridge deck of the orthotropic steel bridge deck and performing low-pass filtering, the technical effect of obtaining full-field response data reflecting the static behavior of the structure can be achieved solely based on sparse measurements.

[0017] An innovative dimensional adaptive prior setting method dynamically determines the prior distribution range based on the ratio of the standard deviation of stress response to the mean of axle load, transforming Bayesian inversion from "blind search" to "physically guided intelligent search," significantly improving inversion efficiency, accuracy, and stability. By constructing a hybrid likelihood function and integrating typical traffic and tail reinforcement information for posterior sampling, a sensitive identification of extreme load conditions that contribute most to fatigue damage is achieved. Furthermore, the structural deformation is inferred through changes in the influence surface, thus serving the technical effect of bridge health assessment and maintenance decision-making.

[0018] This method is specifically designed for the two-dimensional stress characteristics of orthotropic steel bridge decks. The identification results can better reflect the actual mechanical behavior of the structure, support the estimation of daily influence surfaces and deformation rates, and directly serve fatigue life assessment and maintenance decisions. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] Figure 1 This is a schematic diagram illustrating the implementation process of the orthotropic steel bridge deck influence surface identification method based on stress response and Bayesian inversion in this invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0022] Example 1: like Figure 1The image shows an embodiment of the present invention, a method for identifying the influence surface of orthotropic steel bridge deck based on stress response and Bayesian inversion, comprising the following steps: Step 1: Based on the target bridge's traffic spatiotemporal information database and tail reinforcement, generate a virtual axle load matrix and generate virtual loads. S1.1 Establish a layered spatiotemporal information database of traffic flow specific to the orthotropic steel bridge deck to be tested.

[0023] Through historical surveys, short-term monitoring, or traffic camera data, a multi-dimensional distribution of vehicles crossing the bridge was statistically obtained, including time dimension, vehicle type dimension, wheelbase condition, and the identification threshold and rear condition distribution of heavy / overloaded vehicles. Specifically: In terms of time dimension, time windows are set in layers, divided into g time windows (typically 4-6 windows can be selected, such as morning peak, daytime off-peak, evening peak, and night). Calculate the conditional probability of vehicle type within each time window. , where k represents the kth time window; For each vehicle model, the axle load distribution (quantile table or main distribution + GPD tail) and wheelbase distribution are statistically analyzed for each time window. Based on axle load distribution, screen heavy-load / overloaded vehicles, identify the threshold and tail condition distribution of heavy-load / overloaded vehicles (it is recommended to use the Peaks-Over-Threshold method to fit the GPD, Generalized Pareto Distribution).

[0024] S1.2 Based on the hierarchical traffic flow spatiotemporal information database, generate virtual vehicle queues in batches that conform to the actual traffic load characteristics of the bridge, including the axle load sequence of each virtual vehicle. and wheelbase This leads to the construction of a virtual axis weight matrix for inversion. .

[0025] S1.2.1 Axle load sequence

[0026] Generate total number of vehicles At that time, a tail enhancement factor was introduced. (Typical values ​​are selected as 5–8): For each virtual car : Extract time window k~Categorical( )( (Assign traffic volume weights for each time window); then draw uniformly random numbers u ~ U(0,1). like (Approximately 0.83–0.89), then proceed to the rear reinforcement branch, and firstly, extract a subset of heavy-duty vehicle models proportionally (e.g., heavy-duty vehicles accounting for the bottom 10–15% of the total vehicle models by axle load). Then, sample the axle load at the rear conditional distribution of this vehicle model (u>0.95 quantile, or directly sample from the Global Product Sample Database (GPD)). This is the trigger probability threshold for virtual vehicles entering the "heavy-load tail-end enhanced sampling branch" in Monte Carlo simulations. It's a parameter designed to prevent traffic load simulations from underrepresenting extreme heavy-load tail-end events, which could lead to conservative or inaccurate fatigue damage assessments of the steel bridge deck. If... If the sample size is too low, the tail-enhancing sample size is insufficient, failing to address the lack of extreme case data in existing methods; if If the value is too high, it will not only increase the computational load of the Monte Carlo simulation, but also cause the virtual load matrix to deviate excessively from the actual traffic flow, resulting in a decrease in the adaptability of the inversion effect to conventional loads. Otherwise, the conditional probability of the vehicle model is based on time window k. Extract the vehicle model type_c and analyze its complete conditional axle weight distribution over time window k. Axle load sequence of each virtual vehicle sampled in the middle .

[0027] Example: Total number of time windows Weight Morning rush hour ( =1, =0.3), peak ( =2, =0.4), evening rush hour ( =3, =0.2), nighttime ( =4, =0.1); tail enhancement coefficient , calculate Total number of virtual vehicles Heavy-duty subset: heavy-duty vehicles (semi-trailer tractors) with axle load of 15% or more, and GPD sampling axle load of 30-55t; regular vehicle types: light-duty vehicles (axle load 2-5t) and medium-duty vehicles (8-15t).

[0028] The process of generating a single virtual vehicle:

[0029] Output: Integrate all axle loads of 5 vehicles and stitch them together into a complete matrix according to time window weights (off-peak season accounts for 40%); Select 0.05~0.15×5≈1 extremely heavy-duty vehicle (such as axle load [40,40,48]t) from the 3 tail branch vehicles to form a pure tail sub-matrix.

[0030] S1.2.2 Wheelbase

[0031] wheelbase Sampling: The prokaryotic density can be used, or weaker conditions can be added (heavy-duty vehicles tend to have more axles and longer wheelbases), specifically: (1) Using the axial sampling of the prokaryotic density A field traffic survey was conducted on the target bridge. Measured wheelbase data for each vehicle type was collected according to vehicle type classification (c). Kernel density estimation was performed on the wheelbase data for each vehicle type to obtain the wheelbase kernel density distribution specific to each vehicle type. Once the model c of a virtual car is determined in the Monte Carlo simulation, the kernel density distribution corresponding to that model is directly used. One or a set of wheelbase values ​​are randomly selected from the data (multiple sets are selected for multi-axle vehicles, corresponding to the number of axles in the axle load sequence {W}) to form the wheelbase {D} of the virtual vehicle.

[0032] (2) Increase wheelbase sampling under weak conditions First, complete the "data collection + kernel density fitting" step (1) to obtain the basic wheelbase kernel density distribution of each vehicle model. For heavy-duty vehicles, an "bias adjustment" is made to their wheelbase distribution: In terms of the number of axles: heavy-duty vehicles are sampled based on the basic distribution by default, but the sampling probability for "multi-axle (3 axles and above)" is increased by 10%–20%; In terms of wheelbase value: the wheelbase kernel density distribution of heavy-duty vehicles… Maintain the original range, but shift the peak of the distribution towards the long axis, or increase the sampling probability of the long axis segment.

[0033] S1.2.3 Matrix [W] Output two types of virtual axis weight matrices. A complete time-varying matrix, assembled according to time window weights, representing typical traffic. It includes all virtual vehicles (regardless of whether u is less than ptail or greater than or equal to ptail), and is spliced ​​together according to time window weights to reflect the complete characteristics of typical traffic. Pure tail-enhancing submatrix (size approximately 0.05–0.15 × It is specifically designed for extreme response and fatigue hotspot assessment, and selects extreme heavy load samples (approximately 0.05–0.15 × Nveh) only from the tail reinforcement branches when u is less than ptail.

[0034] This method embeds time window conditional probability into virtual load generation, realistically reflecting actual service characteristics such as peak / nighttime and heavy-load nighttime conditions. By adjusting the tail enhancement coefficient, it significantly increases the sampling density for heavy-load, multi-axle, and overloaded vehicles, making the inversion results more sensitive to extreme working conditions that contribute the most to fatigue damage. It avoids complex and unstable real-time vehicle identification, making the load input more consistent with the actual service state of the bridge, and can significantly improve the engineering applicability and reliability of this solution.

[0035] Step 2: Synchronous acquisition and static extraction of stress response at key bridge points The virtual load is the actual traffic flow characteristics of the same bridge, time, and route when the bridge generates actual dynamic stress response (including regular traffic flow and heavy-load extreme traffic flow, such as the simulated traffic flow in step one). Otherwise, the subsequent system matrix [A] will be distorted, and the inverted influence surface will deviate from reality.

[0036] Stress sensors are installed on the outer side of the connection between the channel rib and the bridge deck of the orthotropic steel bridge deck. This location is most sensitive to the lateral movement of vehicles and experiences the largest stress range. Dynamic stress responses at this location are simultaneously acquired as vehicles cross the bridge. The dynamic stress signals are then low-pass filtered to remove high-frequency components caused by vehicle-bridge coupled vibrations, extracting the static stress response that reflects the static behavior of the structure. .

[0037] Step 3: Bayesian influence surface inversion based on dimensional adaptive priors S3.1 Establish an influence surface inversion model A two-dimensional influence surface is established based on the bridge deck, and the two-dimensional influence surface to be identified is discretized into... The grid is then rearranged into column vectors. The relationship between stress response and influence surface can be expressed as:

[0038] in, The observed static response vector; The system matrix is ​​composed of the virtual axis weight matrix ( or ) and vehicle movement path determination; This represents the influence surface vector of this type of response to be identified; Let the observation error vector be assumed to follow a multivariate normal distribution. , The variance is unknown. The determination process is as follows: (1) Discretize the two-dimensional stress influence surface of the steel bridge deck into a P×Q grid, rearrange it into column vectors, and determine the correspondence between the grid point coordinates and the vector index; (2) Extract complete parameters of all virtual vehicles from [W]: axle load, wheelbase, lane / lateral position, vehicle speed / longitudinal starting coordinate; (3) Calibrate the physical coordinates of the stress sensor and calculate the static stress value (unit load stress contribution coefficient) generated by 1 unit load at the acquisition point of each grid point through finite element method or field test. (4) Match the core action grid points (static most unfavorable positions) of each axle load on the influence surface grid according to the passage parameters of a single virtual vehicle. (5) The stress contribution coefficient of the axle load at the corresponding grid point is obtained by multiplying the axle load by the unit load stress contribution coefficient. The corresponding value is then assigned to the specified column of the matrix row. (6) The stress contribution of all axles of a single virtual vehicle is superimposed onto the columns of the corresponding rows of the matrix, and the columns with no effect are assigned 0, forming a line of; (7) Concatenate all virtual cars row by row to obtain the initial matrix. After dimension normalization and outlier removal, the final matrix is ​​obtained. .

[0039] S3.2, Dimensionally Adaptive Prior According to Bayes' theorem, the parameters to be determined (influenced by the surface vector) sum of error variance The posterior distribution of a probability is proportional to the product of its prior distribution and the likelihood function.

[0040] in, It is the likelihood function; It is the prior distribution; It is a posterior distribution.

[0041] This scheme proposes an innovative method for dimensionally adaptive prior setting. Let each component affecting the surface vector... Follows an independent normal prior distribution:

[0042] Among them, the prior standard deviation It is adaptively determined by the following formula:

[0043] in, It is the standard deviation of the stress response vector, which characterizes the fluctuation range of the response; It is the average of the absolute values ​​of the axle loads in the virtual axle load matrix, representing the typical level of the load; It is a scaling factor (usually 1 to 10), with a preferred value of 5.

[0044] This formula automatically adapts the range of the prior distribution to the order-of-magnitude ratio of the stress response to the load in a specific problem. When the stress response is large or the load is small, it indicates a potentially large influence surface area, so a wider prior is set; conversely, a narrower prior is set. This avoids the blindness of manual setting, resulting in faster convergence and more stable results for Bayesian inference. Essentially, it provides a physical estimate of the order of magnitude of the impact area. Scaling factor. This ensures that the prior distribution is broad enough to cover all physically plausible solutions, while avoiding excessive broadness that could lead to computational inefficiency.

[0045] For error variance We employ an information-free conjugate prior, i.e., the inverse Gamma distribution:

[0046] Where hyperparameters are taken This indicates that there is almost no prior information about the error variance.

[0047] Based on the assumption of observation error, the likelihood function is:

[0048] S3.3, Posterior sampling Due to the posterior distribution Due to the complex and high-dimensional characteristics of the posterior distribution, it cannot be solved directly. This scheme employs an efficient transition-free sampler (E-NUTS) for posterior distribution sampling, which is an advanced Hamiltonian Monte Carlo (HMC) method for drawing samples from the posterior distribution. The specific sampling process of E-NUTS is as follows: First, initialize the parameters. .for arrive (Total sampling times): Based on the current state, a continuous trajectory is constructed in the parameter space by simulating Hamiltonian dynamics; the optimal trajectory length is automatically determined by the no-turn criterion to avoid random walk behavior; new samples are collected from the trajectory. Before finally discarding One sample (pre-burning period) was used to eliminate the influence of initial values, and the remaining samples were used. Each sample will be analyzed further.

[0049] S3.4 Enhanced Inversion Strategy By combining typical traffic and tail-end reinforcement information, a hybrid likelihood function is constructed:

[0050] Wherein, λ is the weighting coefficient in the mixed likelihood function, used to balance the contribution ratio of typical traffic load and extreme heavy tail load in the Bayesian inversion likelihood function, λ≈0.7~0.85 (which can be determined through cross-validation or fatigue sensitivity analysis). For the j-th vehicle crossing the bridge event, the measured static stress response sample value at the key points of the bridge is given. and They are respectively and The generated system matrix row vectors.

[0051] Based on posterior samples And calculate the maximum a posteriori estimate and posterior mean : ; ; Where T is the total number of iterations (total sampling count) of the E-NUTS sampler sampling the posterior distribution; B is the number of samples during the pre-burning period of the sampling process. For practical engineering applications, this addresses a key aspect of the Bayesian inversion identification results because the entire identification process contains unavoidable inherent errors (sensor observation errors, statistical errors in virtual load simulation, random errors in Bayesian sampling, structural material / manufacturing deviations, etc.). Using only point estimates such as MAP or posterior mean would deprive subsequent bridge safety assessments and fatigue damage analyses of risk references. Therefore, the 95% confidence interval for each influence surface grid point is calculated to quantify uncertainty. The purpose of this quantification is to provide a scientific risk basis for bridge health monitoring and maintenance decisions, transforming the identification results from "theoretical values" into "engineering-usable conclusions." It is the influence surface vector solution with the highest probability in the Bayesian inversion posterior distribution, and it is the basis for the final value of the influence surface vector. Its role is to directly anchor the optimal solution direction of the influence surface identification result, and ultimately determine the influence surface vector. Reconstruct a two-dimensional influence surface matrix and output the stress influence surface identification results of orthotropic steel bridge deck.

[0052] S4, Deformation amount inferred based on changes in the influence surface. The deformation (such as deflection δ or local displacement u) can be used to reconstruct the response by convolving the changed influence surface with the typical load, and then comparing it with the initial response.

[0053] First, select a typical load (using...) Or standard vehicle load, such as highway-class I vehicles): unit load P=1 (calculate the influence surface itself) or actual typical vehicle load sequence: axle load vector a, wheelbase d, vehicle speed v.

[0054] The steps to refactor the response include: Initial baseline response (health status):

[0055] Where x and y are the standard engineering orthogonal coordinates in the two-dimensional plane of the orthotropic steel bridge deck, x is the longitudinal coordinate of the bridge (along the bridge span direction); y is the transverse coordinate of the bridge (along the bridge deck width direction); q(y) is the load distribution. For the initial system matrix, This serves as the reference stress influence surface for a bridge in a healthy and intact state.

[0056] Specifically: When the bridge is in a healthy state at t=0, and a unit concentrated load P=1 is applied to the bridge deck at coordinates (x,y), the static stress value generated at the stress observation point is denoted as: When an arbitrary lateral distributed load q(y) (unit: N / m) is applied to the bridge deck, taking a lateral infinitesimal element dy, the infinitesimal element load is: The stress response produced by the infinitesimal load at the longitudinal position x is: Integrating the response of all transverse infinitesimal loads across the full width of the bridge deck, we obtain the total initial reference response at the longitudinal position x: .

[0057] Continuous influence surfaces cannot be directly used in this project, so the bridge deck needs to be discretized into a mesh. The bridge deck is divided into P×Q uniform meshes, with a horizontal dimension Δy and a vertical dimension Δx. Each mesh is globally numbered j=1,2,...,M, and the center coordinates of the j-th mesh are... Discretize the continuous influence surface into grid point values. The reference influence surface value for the j-th grid is: The influence surface values ​​of all grid points are arranged in order as an M-dimensional column vector: = .

[0058] The continuously distributed load q(y) is discretized into concentrated loads at grid points, generating N virtual vehicles corresponding to N load cases, numbered i=1,2,...,N. The axle load, wheelbase, and travel path of the i-th virtual vehicle uniquely determine the load distribution of each axle at each grid point. Let the load at the j-th grid under the i-th load case be denoted as... .

[0059] After discretizing the continuous bridge deck, influence surface, and load into M grid points, the exact integration is approximated as a discrete summation. The total response at the longitudinal position xi under the i-th load case is: To rewrite the summation into a concise matrix multiplication form, a definitional merging is performed: Let the initial system matrix... The element in the i-th row and j-th column is: = Substituting it into the discrete summation formula, we get: Combine the responses of all N operating conditions into a vector, and all Combining these into a matrix, we finally obtain: = For a single vertical position x, it simplifies to scalar form: .

[0060] Current response (after degradation):

[0061] in, Approximation Instead, if there is no real-time load, the same path is assumed. This represents the static stress response value of the bridge at time t during its service life. The system matrix of the bridge at time t during its service life; This represents the reference stress influence surface at time t during the bridge's service life.

[0062] Deformation increment estimation (using the method assuming linear elasticity): Deflection increment (Directly corresponding to the deflection increment under unit load), actual deformation amount The load amplification factor is the stress-deflection conversion constant, which can be determined through calibration; a more precise relative deformation rate:

[0063] Local deformation sensitive points (near the weld): Local compliance changes at the weld can be identified by hotspot mesh. By directly representing this, multiplying by a typical axle load can convert the "unit load compliance change" into the "local displacement increment under actual load".

[0064] Example 2: The difference from Example 1 is that in step S4 of this example, the deformation increment is predicted by a method based on joint measurement-simulation calibration. This method is suitable for scenarios where the bridge has multiple stress and a small number of displacement monitoring points (in addition to the core stress acquisition point, there are 1 to 2 displacement verification points), the monitoring data is sufficient, and high-precision calibration of the deformation is required (such as maintenance decisions for key segments of a bridge).

[0065] (1) Obtain the initial estimated value of the deformation increment using the linear method described above. and based on and The simulated displacement values ​​of the displacement check points were obtained. The simulation process is as follows: Under the premise of linear elasticity, the displacement influence surface With the inverted stress influence surface It has a fixed linear ratio ;Simultaneously convert proportionally to obtain Completely isomorphic displacement system matrix Reuse Based on the virtual vehicle parameters and driving path, and according to the principle of influence surface superposition, the displacement time history of each vehicle during the entire bridge crossing process is calculated for each vehicle and each position. Low-pass filtering is applied to the simulated displacement time history to extract the static displacement peak value, and the predicted displacement value of the check point is obtained after statistical matching. .

[0066] (2) Collect the actual measured displacement values ​​of the displacement check points Calculate the calibration coefficient: ( (The ratio of measured to simulated values). (3) Correct the stress response using calibration coefficients: ; (4) Based on the calibrated parameters, recalculate the deformation increment: .

[0067] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for detecting the deflection deformation of orthotropic steel bridge decks based on influence surfaces, characterized in that: S1. Establish a layered traffic flow spatiotemporal information database specific to the orthotropic steel bridge deck to be tested, and generate virtual loads based on time windows and vehicle rear enhancement coefficients to simulate bridge traffic. S2, Obtain the static stress response at key points of the bridge when a vehicle crosses the bridge. ; S3, based on dimensional adaptive prior inversion of the influence surface, constructs a hybrid likelihood function, and reconstructs a two-dimensional influence surface matrix by integrating typical traffic and tail reinforcement information; S4, compare the changes before and after the influence to infer the amount of flexural deformation.

2. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 1, characterized in that: Based on the hierarchical traffic flow spatiotemporal information database, virtual vehicle queues that conform to the actual traffic load characteristics of the bridge are generated in batches, including the axle load sequence of each virtual vehicle. and wheelbase This leads to the construction of a virtual axis weight matrix for inversion. .

3. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 2, characterized in that: The virtual axis weight matrix Including complete time-varying matrices and pure tail enhancement submatrix The complete time-varying matrix is ​​spliced ​​together according to the weight ratio of time windows to represent typical traffic, and the pure tail-reinforcement submatrix is ​​used for extreme response and fatigue hotspot assessment.

4. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 2, characterized in that: In step S3, the two-dimensional influence surface of the bridge deck is discretized into... The grid is then rearranged into column vectors. The relationship between stress response and influence surface can be expressed as: , The system matrix is ​​determined by the virtual axle load matrix and the vehicle movement path. Let the observation error vector be assumed to follow a multivariate normal distribution. , The variance is unknown.

5. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 4, characterized in that: According to Bayes' theorem, the influence of surface vectors sum of error variance The posterior distribution is proportional to the product of the prior distribution and the likelihood function: in, Let be the likelihood function. For the prior distribution, For the posterior distribution, each component of the influence surface vector Follows an independent normal prior distribution: Among them, the prior standard deviation It is adaptively determined by the following formula: ,in, It is the standard deviation of the stress response vector, which characterizes the fluctuation range of the response; It is the average of the absolute values ​​of the axle loads in the virtual axle load matrix, representing the typical level of the load; It is a scaling factor that adapts the range of the prior distribution to the order-of-magnitude ratio of the stress response to the load for a specific problem.

6. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 5, characterized in that: The mixture likelihood function is expressed as: ， Where λ is the weighting coefficient in the mixture likelihood function. For the j-th vehicle crossing the bridge event, the measured static stress response sample value of the bridge key points is given. and They are respectively and The generated system matrix row vectors.

7. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 6, characterized in that: Based on posterior samples The sum of calculations yields the maximum a posteriori estimate. and posterior mean : ; T is the total number of iterations (total sampling count) by the E-NUTS sampler to sample the posterior distribution; B is the number of samples during the pre-burning period in the sampling process. The confidence interval for each influence surface grid point is calculated, and finally, the influence surface vector is... Reconstruct a two-dimensional influence surface matrix and output the stress influence surface identification results of orthotropic steel bridge deck.

8. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 1, characterized in that: In step S4, the deformation is used to reconstruct the response by convolving the changed influence surface with the typical load, and then compared with the initial response; initial reference response: Where q(y) represents the load distribution, For the initial system matrix, The reference stress influence surface for a bridge in a healthy and intact state; ,in, Approximation Instead, if there is no real-time load, the same path is assumed. This represents the static stress response value of the bridge at time t during its service life. The system matrix of the bridge at time t during its service life; This represents the reference stress influence surface at time t during the bridge's service life.

9. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 8, characterized in that: Deformation increment is estimated using a linear method: deflection increment Actual deformation Wherein, the load amplification factor is the stress-deflection conversion constant, determined through calibration; relative deformation rate: .

10. The method for detecting the deflection deformation of orthotropic steel bridge deck based on the influence surface according to claim 8, characterized in that: The deformation increment is estimated using a method based on combined measured and simulated calibration, based on the initial estimated value of the deformation increment. and based on and The simulated displacement values ​​of the displacement check points were obtained. Collect the actual measured displacement values ​​of the displacement check points. Calculate the calibration coefficient: The stress response is corrected using calibration coefficients: Based on the calibrated parameters, the deformation increment is recalculated: .

Images