Method and system for predicting long-term deformation of large-span rigid frame bridge

Abstract

The invention provides a long-term deformation prediction method and system for a large-span rigid frame bridge, and relates to the technical field of bridge deformation prediction. The method aims at solving the problems that in the prior art, a prediction method insufficiently considers the coupling effect of geometric nonlinearity and the concrete shrinkage and creep effect and lacks the dynamic response capacity to damage events in the operation period, and consequently the deviation between a prediction result and the reality is large. According to the method, synchronous coupling analysis of the two effects is realized by establishing a space finite element model and constructing a structure balance equation which simultaneously considers the geometric nonlinearity and the shrinkage and creep effect; through reference simulation and damage event identification based on a probability distribution model, dynamic damage correction and simulation restart are carried out on the model in combination with monitoring data, a closed-loop prediction mechanism for identification, correction and restart is constructed, the precision and accuracy of long-term deformation and internal force prediction are effectively improved, and the adaptability in a complex scene is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bridge deformation prediction, and particularly relates to a long-term deformation prediction method and system for a large-span rigid-frame bridge. BACKGROUND

[0002] Large-span rigid-frame bridges have been widely used in modern bridge engineering due to their good structural integrity, superior seismic performance, unobstructed view under the bridge, and low maintenance cost. With the progress of material science and the improvement of construction technology, the span record of rigid-frame bridges is constantly being broken. Rigid-frame bridges have unique advantages in crossing complex terrains such as deep valleys and straits, and have broad development prospects. In particular, in the construction of highway networks and urban expressways, large-span rigid-frame bridges have become an indispensable bridge type choice.

[0003] However, in terms of long-term service performance prediction of large-span rigid-frame bridges, the existing analysis methods still have obvious deficiencies. Traditional deformation prediction methods often separately calculate geometric nonlinearity and concrete shrinkage and creep effects and then linearly superimpose them. The influence of the coupling of the two on the long-term performance of the structure is not adequately considered. More importantly, existing methods generally lack dynamic response capability to damage events during operation, such as vehicle impact and earthquakes. After identifying structural damage, they cannot automatically modify the calculation model and re-predict, which results in a large deviation between the calculation results and the actual service state, making it difficult to provide reliable basis for the preventive maintenance and precise treatment of bridges. In particular, after encountering unexpected damage, traditional methods cannot dynamically adjust the prediction path and adaptively update the model, which seriously affects the accuracy of long-term deformation prediction and the universality in complex scenarios.

[0004] Therefore, there is an urgent need in the prior art for a long-term deformation prediction method and system for large-span rigid-frame bridges with higher accuracy and universality. SUMMARY

[0005] To this end, the present application provides a long-term deformation prediction method and system for a large-span rigid-frame bridge to overcome the problems in the prior art that the influence of the coupling of geometric nonlinearity and concrete shrinkage and creep effects on the long-term performance of the structure is not adequately considered, and the dynamic response capability to damage events during operation is lacking, resulting in a large deviation between the calculation results and the actual service state, making it difficult to provide reliable basis for the preventive maintenance and precise treatment of bridges.

[0006] To achieve the above-mentioned purpose, the present application provides a long-term deformation prediction method for a large-span rigid-frame bridge, comprising: S1, establishing a spatial finite element model based on the design drawings of the bridge, material properties, and boundary conditions; S2, constructing a structure equilibrium equation considering geometric nonlinearity and concrete shrinkage and creep effects based on the spatial finite element model; S3, performing a baseline simulation on the spatial finite element model based on the structural equilibrium equation to determine a baseline mechanical state evolution from the construction start to a target prediction year; S4, identifying a damage event based on a pre-established baseline probabilistic distribution model; S5, when a damage event is identified, pausing the baseline simulation and performing a damage correction on the spatial finite element model based on the monitoring data to obtain a corrected finite element model representing a state after the damage event, and constructing a corrected structural equilibrium equation based on the corrected finite element model; S6, re-performing the baseline simulation based on the corrected finite element model and the corrected structural equilibrium equation until no damage event is identified; S7, outputting a long-term deformation and internal force prediction result of the bridge at the target year; wherein the damage event at least includes vehicle impact and earthquake.

[0007] Further, the baseline simulation on the spatial finite element model includes: S31, starting from the construction start point, gradually advancing in time sequence to the target prediction year, dividing the total time into consecutive time steps; S32, in each current time step, performing a baseline simulation loop; S33, advancing the time to the next time step and repeating S32 until the simulation time reaches the target prediction year.

[0008] Further, the baseline simulation loop includes: S32a, applying the concrete shrinkage and creep effect up to the previous time step as equivalent node forces to the current structural configuration; S32b, solving the structural equilibrium equation to update the displacement and bar-end force of the structure through iterative calculation until the solution converges; S32c, based on the updated bar-end force, calculating the shrinkage and creep equivalent node forces of the next time step through a recursive formula fitting the creep coefficient based on an exponential function.

[0009] Further, the establishment of the baseline probabilistic distribution model includes: Collecting historical monitoring data of the bridge under normal operating conditions to form a training sample set; Based on the training sample set, establishing a baseline probabilistic distribution model of each mechanical state monitoring parameter and determining its statistical characteristic value; Based on the hypothesis testing method, setting the control upper limit and control lower limit of the abnormal index for identifying damage events.

[0010] Further, the damage event identification based on the pre-established baseline probability distribution model comprises: S41, acquiring monitoring data in an actual time interval corresponding to a current time step and not participating in the identification and judgment, as new monitoring data; S42, calculating a comprehensive abnormal index of the new monitoring data based on the baseline probability distribution model; S43, using a sliding time window technology to analyze the change trend of the comprehensive abnormal index in consecutive time steps; S44, when it is identified that the comprehensive abnormal index exceeds the upper control limit for consecutive k time steps or is below the lower control limit for consecutive k time steps, determining that a damage event occurs; Wherein, k is a preset sensitivity parameter.

[0011] Further, the calculation of the abnormal index of the new monitoring data based on the baseline probability distribution model comprises: S42a, calculating the standardized residual of each new monitoring data with respect to the baseline probability distribution model; S42b, obtaining a local abnormal index of each monitoring data through Mahalanobis distance calculation based on the standardized residual; S42c, weighting and fusing the local abnormal indexes according to the spatial distribution characteristics of the monitoring data to form a comprehensive abnormal index of the structure as a whole.

[0012] Further, the damage correction based on the monitoring data on the spatial finite element model comprises: S51, identifying a damage area in the structure based on the spatial distribution of the monitoring data triggering the damage event identification and the abnormal index thereof; S52, minimizing the residual between the dynamic characteristics or static response of the spatial finite element model and the measured data in the damage area by using a weighted least squares method, so as to identify and correct the structural parameters of the damage area; S53, correcting all node coordinates of the spatial finite element model based on the corrected structural parameters of the damage area to obtain a corrected finite element model.

[0013] Further, the monitoring data comprises data collected by at least one of the following sensors: strain data collected by a fiber grating sensor, rotation angle data collected by an inclinometer, and vibration data collected by an acceleration sensor.

[0014] Further, the long-term deformation and internal force prediction results of the bridge at the target age include: a full-bridge time-varying deformation envelope, a key section internal force time history, and a long-term stress distribution cloud map.

[0015] The application further provides a long-term deformation prediction system for a long-span rigid frame bridge, which adopts any of the long-term deformation prediction methods for the long-span rigid frame bridge and specifically comprises the following: a model establishing module, configured to establish a spatial finite element model based on design drawings, material properties and boundary conditions of the bridge; an equation constructing module, connected with the model establishing module, configured to construct a structure balance equation considering geometric nonlinearity and concrete shrinkage and creep effects based on the spatial finite element model; a cyclic simulation module, connected with the equation constructing module, configured to perform a benchmark simulation on the spatial finite element model based on the structure balance equation to determine a benchmark mechanical state evolution from the start of construction to a target prediction year; an event identifying module, connected with the cyclic simulation module, configured to identify a damage event based on a pre-established benchmark probability distribution model; an event correcting module, connected with the event identifying module, configured to pause the benchmark simulation when a damage event is identified and perform a damage correction on the spatial finite element model based on monitoring data to obtain a corrected finite element model representing a state after the damage event, and construct a corrected structure balance equation based on the corrected finite element model; a corrected simulation module, connected with the event correcting module, configured to perform the benchmark simulation again based on the corrected finite element model and the corrected structure balance equation; a result predicting module, connected with the corrected simulation module, configured to output a long-term deformation and internal force prediction result of the bridge at the target year.

[0016] Compared with the prior art, the application has the following beneficial effects: Firstly, the application realizes synchronous coupling analysis of geometric nonlinearity and concrete shrinkage and creep effects by using initial strain method and equivalent node force conversion technology, thereby avoiding error accumulation caused by linear superposition; meanwhile, a dynamic closed loop of damage event identification-model damage correction-simulation restart is constructed to accurately locate a damage area and correct structure parameters and geometric configuration based on improved DBSCAN spatial clustering and weighted least square method, so that the calculation model always fits the actual service state of the bridge, thereby solving the problem of large deviation between traditional one-time prediction and actual state and improving the accuracy and precision of long-term deformation and internal force prediction results; Secondly, the application improves long-term operation period simulation efficiency through non-uniform time step division and recursive calculation technology, avoids mass historical data storage pressure, constructs a comprehensive abnormality index through standardized residual error, Mahalanobis distance and weighted fusion technology, reduces false positive rate and false negative rate of damage event identification by combining sliding time window trend analysis, provides high-quality data input for model correction, realizes fine presentation and verification of full-bridge deformation, key cross-section internal force and stress distribution through spatial interpolation, grid optimization and visual modeling, further reduces deviation between predicted value and actual value through full-process multi-technology cooperation from data acquisition, model calculation to result output, and achieves higher precision and higher accuracy prediction effect. Thirdly, the application establishes a complete response mechanism from event identification, simulation adjustment to model correction to solve the pain point that the traditional method lacks dynamic response capability of damage events: firstly, the precise identification of damage events such as vehicle impact and earthquake is realized through the combination of baseline probability distribution model and sliding time window technology, standardized residual error, Mahalanobis distance weighted fusion calculation, effectively avoiding instantaneous fluctuation misjudgment and event false negative; secondly, after identifying the damage event, the baseline simulation can be immediately paused and the current mechanical state is saved to ensure the continuity of subsequent correction and simulation, solving the problem that the traditional prediction cannot be connected after interruption; finally, the improved DBSCAN spatial clustering is used to locate the damage area, the weighted least squares method is used to correct the structure parameters, and the radial basis function interpolation is used to update the node coordinates, realizing the adaptive adjustment of the model, so that the corrected model can truly reflect the structure shape and mechanical properties after damage, and restarting the simulation based on the corrected model can ensure that the long-term deformation and internal force prediction after the damage event can still fit the actual service state, completely changing the limitation that the traditional one-time prediction cannot respond to sudden damage, and greatly improving the universality and reliability of the prediction results in complex service scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The method flowchart of the long-term deformation prediction method of the long-span rigid frame bridge of the embodiment of the application is shown in the figure. Figure 2 The structural block diagram of the long-term deformation prediction system of the long-span rigid frame bridge of the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0018] In order to make the purpose and advantages of the application more clear and explicit, the application will be further described below in combination with embodiments; it should be understood that the specific embodiments described herein are only used to explain the application, and do not limit the application.

[0019] The preferred embodiments of the application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the application, and are not intended to limit the protection scope of the application.

[0020] It should be noted that in the description of the present application, the terms of direction or position relationship indicated by "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or position relationship shown in the drawings, which is only for the convenience of description, and does not indicate or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0021] In addition, it should be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, or the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0022] Embodiment 1 As Figure 1 shown, the present application proposes a long-span rigid frame bridge long-term deformation prediction method, comprising: S1, based on the design drawings, material properties and boundary conditions of the bridge, a spatial finite element model is established; In one possible implementation, according to the main beam and the bridge pier axis coordinates, a quadratic interpolation method is used to generate the key nodes of the structure, and the node spacing is controlled according to the ratio of the adjacent pier spacing to 20; the ironwood Xinke beam element considering shear deformation is used to simulate the concrete member; the actual centroid position of the member is accurately reproduced by section offset setting.

[0023] The concrete material adopts C55 concrete constitutive, the elastic modulus is 3.55x10 4 MPa, the Poisson's ratio is 0.2, the density is 26kN / m³, the shrinkage and creep model adopts CEB-FIP2010 model, and the environmental humidity, member theoretical thickness and other parameters are considered; the prestressed steel bar adopts a three-fold elastic-plastic model, specifically, the elastic modulus is 1.95x10 5 MPa, and the yield strength is 1860MPa.

[0024] For the pier bottom consolidation, all six degrees of freedom are constrained; for the fixed support, the three translational degrees of freedom are constrained; for the sliding support, the longitudinal displacement of the beam body is released; the compression and shear stiffness of the support is simulated by spring element.

[0025] The mesh density is verified by h convergence method to ensure that the internal force error of the key section is less than 2%.

[0026] S2, based on the spatial finite element model, a structure equilibrium equation considering geometric nonlinearity and concrete shrinkage and creep effect is constructed; In a possible implementation, S21, a geometric nonlinear basic equation is established, specifically, a rotating coordinate method is adopted to establish a beam element balance equation in a local coordinate system of the element; Considering the large displacement effect, an UL column is used to establish an incremental form of virtual work equation; a tangent stiffness matrix of the element containing a geometric stiffness matrix is formed.

[0027] S22, shrinkage and creep initial strain is introduced, specifically, the shrinkage strain and the creep strain of the concrete are treated as initial strain; the shrinkage strain is calculated according to a specification formula, and the time and development law are considered; the creep strain is fitted by an exponential function.

[0028] S23, the equivalent node force is calculated, specifically, the shrinkage and creep initial strain is converted into the equivalent node force by integration; a recursive formula is established to avoid storing historical data; the recursive result is used to calculate the equivalent node force of the next time step.

[0029] S24, coordinate system conversion, specifically, the balance equation in the local coordinate system is converted to the overall coordinate system of the structure through a conversion matrix; the contribution of the geometric stiffness matrix is considered to form the overall tangent stiffness matrix of the structure.

[0030] S25, the coupled balance equation is established, specifically, the shrinkage and creep equivalent node force is taken as an additional load term into the balance equation; the final incremental balance equation is formed.

[0031] S26, a solving strategy is established, specifically, the Newton iteration method is used to solve the nonlinear equation; the convergence criterion is set as the ratio of the unbalanced force norm to the reference load norm being less than the tolerance; the iterative calculation is performed in each time step until convergence.

[0032] The present application solves the error accumulation problem of traditional separation calculation of geometric nonlinearity and shrinkage and creep effect by the initial strain method and equivalent node force conversion technology, realizes the synchronous coupling analysis of the two complex effects, and improves the accuracy and reliability of long-term deformation prediction; through the exponential function fitting and recursive calculation technology, the technical bottleneck of large historical data storage in creep analysis is solved, the efficient management of massive time step calculation data is realized, and the calculation resource consumption and storage demand are reduced; through the combination of the rotating coordinate method and the UL column, the precision problem of the balance equation established under the condition of large displacement is solved, the accurate description of the geometric nonlinearity effect is realized, and the calculation accuracy of the structure response under complex stress state is improved; by setting the Newton iteration method solving strategy, the convergence stability problem of the nonlinear equation set is solved, the efficient numerical solution of the complex coupling system is realized, and the stability and convergence efficiency of the calculation process are improved.

[0033] The alternative solutions of geometric nonlinear processing are TL array method, co-rotational method and finite displacement theory method. Specifically, the TL array method adopts full Lagrange array, establishes equilibrium equation based on initial configuration, and is suitable for large displacement and small strain condition; the co-rotational method decomposes rigid body motion and deformation at the element level, which can simplify nonlinear calculation; the finite displacement theory method adopts higher order strain displacement relationship, which improves the calculation accuracy in large rotation condition. The TL array method is theoretically rigorous but has large calculation amount, and is suitable for theoretical research; the co-rotational method has high calculation efficiency and is suitable for engineering application; the finite displacement theory method has high accuracy but is complex to realize.

[0034] The alternative solutions of shrinkage and creep effect calculation are age-adjusted effective modulus method, integral constitutive relationship method and internal variable theory method. Specifically, the age-adjusted effective modulus method converts the creep effect into the adjustment of elastic modulus, which can simplify the calculation process; the integral constitutive relationship directly adopts integral equation to describe the development of creep, and combines numerical integral method for solution; the internal variable theory method introduces internal state variable to establish a more accurate creep prediction model. The age-adjusted effective modulus method is simple to calculate but has limited accuracy; the integral constitutive relationship method has high accuracy but is complex to calculate; the internal variable theory method has clear physical meaning but has difficulty in parameter determination.

[0035] S3, based on the structure equilibrium equation, performing benchmark simulation on the spatial finite element model to determine the benchmark mechanical state evolution from the construction start to the target prediction period; Performing benchmark simulation on the spatial finite element model comprises: S31, starting from the construction starting point, gradually advancing in time sequence to the target prediction period, dividing the total time into continuous time steps; S32, in each current time step, performing a benchmark simulation loop; S33, advancing the time to the next time step and repeating S32 until the simulation time reaches the target prediction period.

[0036] Performing a benchmark simulation loop comprises: S32a, applying the concrete shrinkage and creep effect up to the last time step as equivalent node force to the current structure configuration; S32b, solving the structure equilibrium equation to update the displacement and bar end force of the structure through iterative calculation until the solution converges; S32c, based on the updated bar end force, calculating the shrinkage and creep equivalent node force of the next time step through the recursive formula fitting the creep coefficient based on the exponential function.

[0037] In one possible implementation, a non-uniform time step is adopted, for example, a step of 15 days for the first 2 years of construction, a step of 30 days for the initial 5 years of operation, and a step of 90 days for the long-term operation of 5-50 years, starting from the first concrete member pouring day of the bridge as the time origin, to establish the time series.

[0038] In each time step, the shrinkage and creep equivalent nodal force of the previous time step is read from the stored historical data and applied as an external load to the current structural configuration. The tangent stiffness matrix of the structure is assembled under the current configuration, and the Newton iteration method is used for solving, with the convergence criterion being that the ratio of the norm of the unbalanced force to the norm of the total load is less than 0.0001. The converged displacement and bar end force are output. Based on the converged bar end force, the creep effect of the next time step is calculated through a recursive formula, the shrinkage strain increment is calculated and converted into an equivalent nodal force, and the creep effect and shrinkage strain increment are stored for use in the next time step.

[0039] When the time represented by the current time step is greater than the target service life, the loop is terminated, otherwise, the current state is saved and the next time step calculation is performed in S32.

[0040] In the time step division process, alternative solutions are adaptive time step method and logarithmic time coordinate method. The adaptive time step method dynamically adjusts the time step according to the convergence speed and response change rate. The step size is automatically reduced when the response is intense, and the step size is increased when the response is stable. The advantages are balancing the calculation efficiency and accuracy, and the disadvantages are complex implementation and the need to set adjustment criteria. The logarithmic time coordinate method divides the time step on the time axis using a logarithmic coordinate. The step size is small in the early stage and gradually increases in the later stage. The advantages are consistent with the shrinkage and creep development law, and the disadvantages are that important changes in the later stage may be missed.

[0041] In the time-varying effect calculation process, alternative solutions are age-adjusted effective modulus method and genetic integral method. The age-adjusted effective modulus method converts the creep effect into the time-dependent change of the elastic modulus, and uses the effective elastic modulus to replace the instantaneous elastic modulus. The advantages are avoiding historical data storage and simple calculation, and the disadvantages are relatively low accuracy, which is suitable for approximate calculation. The genetic integral method directly calculates the creep integral using numerical integration method, and uses trapezoidal rule or Simpson's rule for integral approximation. The advantages are theoretical rigor and high accuracy, and the disadvantages are large calculation amount and the need to store complete stress history.

[0042] In the equation solving process, alternative solutions are quasi-Newton method and explicit time integration method. The quasi-Newton method approximates the Hessian matrix by rank-two update, reducing the number of stiffness matrix reformations. The advantages are improving the nonlinear solving efficiency, and the disadvantages are slightly worse convergence than Newton method. The explicit time integration method uses central difference method for time integration, without the need for iterative solution of nonlinear equations. The advantages are avoiding iteration convergence problems, and the disadvantages are conditional stability and limited time step size.

[0043] The application solves the problems of insufficient precision in the construction period and low efficiency in the operation period of the traditional fixed step method by the non-uniform time step division technology, realizes the optimized allocation of the computing resources on the time axis, improves the computing efficiency and precision of the whole life cycle simulation, solves the technical problem of huge storage of historical data in long-term simulation by the recursive calculation technology of shrinkage effect, realizes the efficient management of the data of the mass time step calculation, reduces the computing and storage requirements, improves the upper limit of the prediction time, solves the convergence problem of the coupled analysis of geometric nonlinearity and material nonlinearity by the Newton iteration method and convergence control technology, realizes the efficient and stable solution of the complex nonlinear system, and improves the reliability and robustness of the computing process, solves the technical connection problem of the continuous simulation of the construction process and the operation stage by the phased cyclic simulation architecture, realizes the tracking of the complete mechanical state evolution from the beginning of construction to the target year, and establishes the complete database of the development of the whole life performance of the structure.

[0044] S4, identifying damage events based on the pre-established baseline probability distribution model, including: The establishment of the baseline probability distribution model includes: Collecting historical monitoring data of the bridge under normal operation state to form a training sample set; Based on the training sample set, a baseline probability distribution model of each mechanical state monitoring parameter is established and the statistical characteristic value is determined; Based on the hypothesis testing method, the upper and lower control limits of the abnormal index for identifying damage events are set.

[0045] In one possible implementation, the damage event is an event or state change that occurs during the operation of the bridge and has a significant impact on the safety or performance of the structure, which will cause the mechanical state of the bridge structure to deviate abnormally and exceed the prediction range under the normal operation state, for example, vehicle impact and earthquake. Specifically, at least 2 years of continuous historical monitoring data of the bridge in the normal operation period are collected, and the monitoring data includes the deflection of the main beam, the displacement of the tower, the structure frequency, the strain of the key section, etc. The historical monitoring data is sequentially subjected to data cleaning, data alignment, environmental factor correction and data normalization, and the data set is divided by quarter to consider the influence of seasonal changes, and a sample database is established, containing fields such as timestamp, monitoring value and environmental parameter.

[0046] Distribution fitting test is performed on each monitoring parameter, and the commonly used distribution types are normal distribution, lognormal distribution and Weibull distribution. The K-S test is used to determine the optimal distribution type, the marginal probability distribution model of each parameter is established, the mean and standard deviation of the distribution of each parameter are calculated, the quantile values of the 95% and 99% confidence intervals are determined, the shape parameters such as skewness and kurtosis are calculated, and the correlation coefficient matrix between parameters is established.

[0047] The preliminary control limit is set by the 3 standard deviation principle, the upper control limit is the sum of the mean and 3 times the standard deviation, the lower control limit is the difference between the mean and 3 times the standard deviation, the control limit is adjusted based on the actual false positive rate, the control limit is verified and fine-tuned through historical data to ensure that the target false positive rate is controlled within 0.5%.

[0048] The model performance is tested using the reserved validation data set, the false positive rate and the false negative rate of the model are calculated, and the control limit setting is optimized based on the false positive rate and the false negative rate to further balance the sensitivity and specificity.

[0049] S41, obtaining the monitoring data in the actual time interval corresponding to the current time step and not participating in the identification judgment, as new monitoring data; S42, calculating the comprehensive anomaly index of the new monitoring data based on the reference probability distribution model; S43, using the sliding time window technology to analyze the change trend of the comprehensive anomaly index in the continuous time steps; S44, when the comprehensive anomaly index is identified to exceed the upper control limit for k consecutive time steps, or is below the lower control limit for k consecutive time steps, it is determined that a damage event occurs; Wherein, k is a preset sensitivity parameter.

[0050] Based on the reference probability distribution model, the anomaly index of the new monitoring data is calculated, including: S42a, for each new monitoring data, calculating its standardized residual relative to the reference probability distribution model; S42b, based on the standardized residual, obtaining the local anomaly index of each monitoring data by Mahalanobis distance calculation; S42c, according to the spatial distribution characteristics of the monitoring data, the local anomaly indexes are weighted and fused to form the comprehensive anomaly index of the structure as a whole.

[0051] In one possible implementation, a data interface with the bridge health monitoring system is established, and after the calculation of each simulation time step is completed, the monitoring data in the actual time interval corresponding to the time step is automatically collected. The following processing is performed on the collected raw data: Data validity check: eliminate invalid data caused by sensor failure; Timestamp alignment: unify the time base of each monitoring point data; Data identification: adding "non-participation in identification judgment" status identification to new monitoring data; Data storage: store the valid data in the monitoring database of the current time step.

[0052] For each new monitoring data, the following calculation process is performed: Call the corresponding benchmark probability distribution model to obtain the mean μ and standard deviation σ of the monitoring parameter; Calculate the standardized residual Z=(x-μ) / σ, where x is the monitoring value; Perform outlier screening on the standardized residual, and eliminate obviously abnormal data with |Z|>10.

[0053] Based on the standardized residual, calculate the local anomaly index by the following method: Group multiple related parameters of the same monitoring point into a feature vector X:[Z1,Z2,...,Z p ]; Obtain the covariance matrix Σ from the benchmark probability distribution model; Calculate Mahalanobis distance , where the superscript T represents the transpose operation of the vector, and the superscript-1 represents the inverse operation of the matrix; Take the Mahalanobis distance D 2 as the local anomaly index of the monitoring point.

[0054] According to the spatial distribution characteristics of the monitoring data, weighted fusion is performed, specifically including: Determine the weight coefficient W i of each monitoring point, where i takes the value range [1,2,3,...,i], and the factors include the structural importance of the measurement point position, the key degree of the monitoring parameter, and the reliability of the historical data; Calculate the weighted comprehensive anomaly index I, and the calculation formula is ; Normalize the comprehensive anomaly index to make it conform to the standard normal distribution.

[0055] Trend analysis is performed using the sliding time window technique, specifically including: Set the window length to m time steps, and determine the specific value according to the monitoring frequency; Calculate the moving average, change slope and fluctuation amplitude of the comprehensive anomaly index within the window; Analyze the persistence and development trend of the anomaly index.

[0056] Optionally, the optional implementation range of the sensitivity parameter k is [3,8], and the preferred embodiment of the sensitivity parameter k is 4.

[0057] The application solves the problems of repeated use of monitoring data and lack of timeliness in the traditional method by real-time data acquisition and state identification technology, realizes accurate synchronization of monitoring data and simulation time steps, and improves the freshness and accuracy of event identification; solves the problem of low reliability of single parameter abnormality judgment by a multi-level abnormality index calculation system, realizes systematic abnormality evaluation from the parameter level, the measuring point level to the structure level, and improves the comprehensiveness and accuracy of abnormality identification; realizes the standardization and quantization of the abnormality degree of multiple parameters by the fusion calculation of standardized residual error and Mahalanobis distance, reduces the false positive rate and the false negative rate; solves the problem of misjudgment of instantaneous fluctuation as structural damage by sliding time window trend analysis, realizes continuous tracking of abnormal development trend, and improves the reliability and stability of event identification; solves the problem of poor adaptability of single threshold judgment by a multi-condition joint determination mechanism, realizes intelligent event identification based on statistical significance, and improves the scientificity and accuracy of damage event judgment.

[0058] S5, when a damage event is identified, pausing the reference simulation, and performing damage correction on the spatial finite element model based on the monitoring data to obtain a corrected finite element model representing the state after the damage event, and constructing a corrected structure balance equation based on the corrected finite element model; and performing damage correction on the spatial finite element model based on the monitoring data, comprising: S51, identifying a damage area in the structure based on the spatial distribution of the monitoring data triggering damage event identification and the abnormality index thereof; S52, minimizing the residual error between the dynamic characteristics or static response of the spatial finite element model and the measured data in the damage area by the weighted least squares method, so as to identify and correct the structural parameters of the damage area; S53, correcting all node coordinates of the spatial finite element model based on the corrected structural parameters of the damage area to obtain a corrected finite element model.

[0059] The monitoring data includes data collected by at least one of the following sensors: strain data collected by a fiber grating sensor, rotation angle data collected by an inclinometer, and vibration data collected by an acceleration sensor.

[0060] In a possible implementation, when a damage event is identified, the reference simulation process is immediately paused, all calculation states of the current simulation time step are saved, the time stamp of the occurrence of the damage event, the peak value and spatial distribution of the abnormality index are recorded, the complete state data of the current spatial finite element model is backed up, and then a model correction dedicated thread is started, and the main calculation thread is kept in a paused state.

[0061] The feature parameters of the damage event are extracted, specifically including: the maximum abnormal index value and its occurrence position, the duration of the abnormal index exceeding the threshold value, and the spatial distribution range parameter, a preliminary damage event evaluation report is generated, including the possible damage type and severity.

[0062] An improved DBSCAN spatial clustering algorithm is used for spatial clustering analysis of abnormal data to generate an abnormal index spatial distribution heat map with a resolution of 0.5 m x 0.5 m. The optimal cluster number is determined by the contour coefficient method to ensure that the clustering quality is greater than 0.7. Among them, the search radius of the improved DBSCAN spatial clustering algorithm is adaptively determined according to the average distance between the measuring points, and the minimum point number is set according to the monitoring network density.

[0063] A damage area identification matrix is established, a convex hull algorithm is used to determine the damage area boundary, and the geometric features of each damage area are calculated, such as the area center coordinates, the equivalent damage radius, and the area / volume.

[0064] S52a, determining the parameters to be corrected and the target response, specifically including: Select the parameter set to be corrected θ = [E, γ, α, β], where E is the elastic modulus, γ is the Poisson's ratio, α is the stiffness reduction coefficient, and β is the mass density correction coefficient; After the damage event, the measured response data of the damage area and the surrounding key measuring points are collected, mainly including static responses such as displacement, strain, and dynamic characteristics such as vibration frequency and mode shape.

[0065] S52b, constructing an error minimization objective function, specifically including: For each item of measured data of each monitoring point, calculate the difference (i.e., residual) between it and the corresponding value calculated by the finite element model under the current parameters; According to the reliability and importance of different types of monitoring data, different weights are assigned to them. For example, it is believed that displacement data is more reliable than strain data, so displacement data is given a higher weight; Sum the weighted residuals of all monitoring points, and this sum is the objective function value that needs to be minimized. The smaller this value is, the more consistent the model is with the actual situation.

[0066] S52c, solving the optimal parameters by using an iterative optimization algorithm, specifically including: The sequence quadratic programming optimization algorithm is adopted, an approximate quadratic programming sub-problem is solved in each iteration to quickly approach the optimal solution, starting from a set of initial parameter guess values, which are the design values or identification values in the healthy state, the target function value and its gradient are calculated according to the current parameters, a search direction and step length are determined to make the target function decrease, thereby generating a new set of more optimal parameter values, the finite element model is updated using the new parameters, and the model response and target function value are recalculated, and the above iteration process is repeated until one of the following convergence criteria is met: The target function change rate is small enough, for example, the change of two consecutive iterations is less than 1%, indicating that the optimization has stabilized; The change of the parameter itself is small enough, for example, the change is less than 0.5%, indicating that the parameter value has been basically determined; The maximum number of iterations is reached, for example, 50 times, to prevent infinite loops.

[0067] S52d, output and verification, specifically including: When the iteration converges, output the final optimal structure parameter set; Substitute the final optimal structure parameter set into the finite element model and run a complete analysis to verify whether the response calculated by the model, such as displacement and frequency, is highly consistent with the measured data to ensure the effectiveness of the correction.

[0068] Due to the limited number of monitoring points, the displacement of each model node cannot be directly obtained. Therefore, the continuous and smooth displacement field of the entire structure needs to be calculated from the sparse monitoring point displacement data through mathematical methods, specifically including: Collect the displacement data of all monitoring points after the damage event, such as linear displacement and rotation angle; Radial basis function interpolation method is used to process irregularly distributed scattered point data, each monitoring point is regarded as an influence source, its displacement value spreads to the surrounding space, the closer the distance, the greater the influence, by superimposing the influence of all monitoring points, the displacement value of any node in the model can be calculated, thereby constructing a continuous three-dimensional displacement field covering the entire structure.

[0069] After obtaining the complete displacement field, the geometry of the original model can be updated, specifically including: The three direction displacement components of each node calculated in the first step are superimposed on the original coordinate values of the node, and the calculated new coordinates are used to replace the node coordinates in the original finite element model in batches. At this time, the grid topology relationship of the model remains unchanged, but the geometric shape of the entire structure has changed, such as the inclination of the pier and the new sag of the main beam.

[0070] In the area of severe deformation, direct updating of coordinates may cause mesh distortion, affecting the calculation accuracy. Therefore, mesh optimization is needed, including: Checking whether there are shape abnormal distorted elements in the corrected model, such as excessively narrow triangles or quadrilateral elements; If distorted elements are found, use Laplace smoothing algorithm for optimization. This algorithm moves the nodes slightly towards the geometric center of their adjacent nodes, making the mesh distribution more uniform and reasonable, while maintaining the main deformation trend determined by the monitoring data.

[0071] After all the above operations are completed, a corrected finite element model is obtained, which is consistent with the actual shape of the structure after the damage event. This model provides an accurate physical basis for re-establishing the balance equation and continuing the simulation. Through this systematic geometric correction process, it is ensured that the calculation model is no longer an idealized initial design shape, but a digital model that can truly reflect the geometric characteristics of the structure in the current damage state.

[0072] Based on the corrected finite element model, the corrected structure balance equation considering geometric nonlinearity and concrete shrinkage and creep effect is re-established.

[0073] The present application improves the DBSCAN spatial clustering and heat map visualization technology, solves the problems of fuzzy damage area positioning and inaccurate boundary definition, improves the accuracy and spatial resolution of damage area identification; by using weighted least squares method and sequential quadratic programming optimization algorithm, the problems of low fusion efficiency of multi-source monitoring data and slow optimization convergence in the process of damage parameter identification are solved, the accuracy and calculation efficiency of structure parameter correction are improved; by using radial basis function interpolation and node coordinate updating technology, the problem of sparse monitoring data difficult to support full structure geometric correction is solved, the consistency of finite element model geometric configuration and actual damage state is improved; by using Laplace smoothing mesh optimization technology, the problem of mesh distortion caused by geometric correction of damage area is solved, the numerical calculation error is reduced, the reliability of subsequent simulation is improved; by using precise positioning of damage area and local parameter correction technology, the problems of resource waste and insufficient precision caused by traditional global model correction are solved, the calculation complexity is reduced, the pertinence and efficiency of model correction are improved; by using multi-type monitoring data fusion correction technology, the problems of low reliability and weak anti-interference ability of single data source correction are solved, the robustness and comprehensiveness of damage correction are improved.

[0074] S6, based on the corrected finite element model and the corrected structure balance equation, re-perform the baseline simulation until no damage event is identified; In a possible implementation, the modified finite element model complete data in S5 includes the modified node coordinates, the structural parameters of the damaged area such as the elastic modulus E, the Poisson's ratio γ, the stiffness reduction coefficient α and the like, the mesh topological relationship and the optimized element shape; the modified structure balance equation considering the geometric nonlinearity and the concrete shrinkage and creep effect is constructed based on the modified model, and the tangent stiffness matrix and the equivalent node force term in the equation are matched with the modified model parameters.

[0075] The simulation state at the time of suspension in S5 is restored, including the current time step number, the completed simulation time length, the concrete shrinkage and creep equivalent node force data of the last time step, the structure displacement and the rod end force history record, to ensure the continuity of the simulation process and avoid the fracture of the time axis and the mechanical state caused by model modification.

[0076] From the time step at which S5 is suspended, the simulation is sequentially advanced by a preset time step, and the actual time interval corresponding to each current time step is recorded, for example, if the simulation is suspended at the 10th time step in the 3rd year of the operation period, the actual time is the 270th-300th day of the 3rd year, and the next time step is the 300th-330th day of the 3rd year.

[0077] If a new damage event is identified, the current modification simulation is immediately suspended, and the damage modification process in S5 is repeated: a new damage area is identified based on the new monitoring data and the abnormal index spatial distribution, the structural parameters of the area are modified by the weighted least squares method, the model node coordinates are updated and the mesh is optimized, and the modified structure balance equation is reconstructed; after the modification is completed, the simulation is restarted from the current suspended time step.

[0078] If no damage event is identified, all data of the entire modification simulation stage are archived, including the structure mechanical state, the shrinkage and creep equivalent node force history, the damage event occurrence time and the damage modification record, the monitoring data and the abnormal index change curve of each time step.

[0079] In S7, the long-term deformation and internal force prediction results of the bridge in the target service life are output, including: the time-varying deformation envelope of the whole bridge, the internal force time history of the key section and the long-term stress distribution cloud diagram.

[0080] In a possible implementation, the complete time series data from the construction starting point to the target prediction year, for example, 50 years, is extracted, including the simulation time stamp of each time step and the corresponding structure mechanical state calculation result; the displacement data of all nodes of the whole bridge, the internal force data of the key section and the stress data of all elements of the whole bridge are extracted, and at the same time, the abnormal data generated in the simulation process due to iteration non-convergence is removed, the linear interpolation method is used to complete the data missing time step, to ensure the data continuity, and all data units are unified and standardized.

[0081] Envelope modeling and visualization, specifically including: Based on the axis projection of the bridge three-dimensional space finite element model, import the full bridge node coordinates in CAD or BIM software, establish a simplified model consistent with the actual bridge geometry, for each time interval, superimpose the maximum / minimum displacement value of the nodes in the simplified model, connect the displacement extreme points of adjacent nodes using cubic spline interpolation method, form the deformation envelope of the interval, mark the displacement extreme value of the key nodes and the corresponding time in the graph, mark the displacement unit, time interval legend.

[0082] Key section internal force time history generation, specifically including: Using Origin software, taking the operating life as the horizontal axis and the internal force value as the vertical axis, setting the range according to the internal force type, such as the bending moment range of-5000~5000kN m, draw 3 time history curves for each key section, respectively for axial force, bending moment, shear force, different internal force types are distinguished by different colors, such as bending moment with red, axial force with blue, shear force with green; Mark the characteristic points on the curve, including: Internal force peak point, such as the maximum bending moment in the construction period closure stage, and the corresponding time; Internal force mutation point, such as the node where the internal force suddenly increases / decreases after the damage event, and the event description; Long-term stable value, that is, the numerical range of internal force that tends to be stable in the later operation period; Draw the internal force time history curves of the same type of key section in the same graph to compare and analyze the internal force difference of different positions, add the specification limit line in the graph to intuitively judge whether the internal force at each time node meets the target requirement.

[0083] Long-term stress distribution cloud map generation, specifically including: Group the stress data of all elements in the full bridge according to the time node; Use radial basis function interpolation method to interpolate the discrete element stress data to the full bridge continuous space grid to generate the full bridge stress field data at each key time node; Import the corrected full bridge finite element model into ANSYS or ABAQUS and other finite element analysis software, and map the interpolated stress field data to the corresponding elements and nodes of the model; Use blue-green-yellow-red, gradient color mapping scheme to set the corresponding relationship between stress value and color; Mark the stress concentration area in the cloud map, such as the bottom of the pier, the connection between the main beam and the support, and the parts with stress value exceeding the surrounding area by more than 30%, marked with a red circle and numbered; Add text explanation below the cloud map, including: The stress distribution characteristics of each key time node, such as operating for 10 years, the compressive stress at the bottom of the main beam in the middle of the span is stabilized at-6 to-8 MPa; Cause analysis of stress concentration area, such as stress concentration at the bottom of the pier due to uneven settlement of the foundation; Operation and maintenance suggestions, such as suggesting that the stress concentration area No. 1 be subjected to non-destructive testing every 2 years.

[0084] The present application visually presents the maximum or minimum displacement extreme value and deformation trend of each node of the whole bridge at different stages from the construction period to the target life through the time-varying deformation envelope diagram of the whole bridge, solves the problem of scattered deformation data in traditional text reports and the difficulty for engineering personnel to quickly grasp the overall deformation law, greatly improves the visual cognitive efficiency of the spatial distribution characteristics and time evolution law of long-term deformation of the bridge, provides a direct basis for judging whether the bridge has excessive sagging, lateral deviation and other risks, and helps maintenance personnel to develop targeted deformation control measures in advance; the dynamic changes of axial force, bending moment and shear force of the core stress position in the whole life cycle are clearly displayed through the time history of internal forces of the key section, the internal force peak point, the mutation point and the long-term stable value are accurately marked, the problem that the time fluctuation law of the internal force of the key section cannot be quantified in the traditional method is solved, the engineering personnel can quickly identify the abnormal change of internal force, and through comparison with the specification limit line, it is directly judged whether the internal force at each stage meets the design requirements, which provides quantitative data support for the reinforcement time selection and material fatigue life evaluation of the key section, and improves the scientificity and accuracy of the maintenance decision.

[0085] Embodiment 2 As Figure 2 shown, the present application also provides a long-term deformation prediction system for a long-span rigid frame bridge, which uses any one of the long-term deformation prediction methods for a long-span rigid frame bridge in embodiment 1, and specifically comprises: A model establishing module is used to establish a spatial finite element model based on the design drawings, material properties and boundary conditions of the bridge; An equation constructing module is connected with the model establishing module and is used to construct a structure balance equation considering geometric nonlinearity and concrete shrinkage and creep effect based on the spatial finite element model; A cycle simulation module is connected with the equation constructing module and is used to perform a benchmark simulation on the spatial finite element model based on the structure balance equation to determine the evolution of the benchmark mechanical state from the start of construction to the target prediction life; An event identification module is connected with the cycle simulation module and is used to identify damage events based on a pre-established benchmark probability distribution model; An event correction module, connected with the event identification module, is configured to pause the reference simulation when an injury event is identified, and perform injury correction on the spatial finite element model based on the monitoring data to obtain a corrected finite element model representing the state after the injury event, and construct a corrected structure balance equation based on the corrected finite element model; A corrected simulation module, connected with the event correction module, is configured to perform the reference simulation again based on the corrected finite element model and the corrected structure balance equation; A result prediction module, connected with the corrected simulation module, is configured to output the long-term deformation and internal force prediction results of the bridge in the target service life.

[0086] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after these changes or replacements will all fall within the protection scope of the present application.

Claims

1. A long-span rigid frame bridge long-term deformation prediction method, characterized in that, The method comprises the following steps: S1, establishing a spatial finite element model based on the design drawings of the bridge, material properties and boundary conditions; S2, constructing a structure balance equation considering geometric nonlinearity and concrete shrinkage and creep effects based on the spatial finite element model; S3, performing a benchmark simulation on the spatial finite element model based on the structure balance equation to determine the benchmark mechanical state evolution from the start of construction to the target prediction year; S4, identifying damage events based on a pre-established benchmark probability distribution model; S5, when a damage event is identified, pausing the benchmark simulation and performing damage correction on the spatial finite element model based on the monitoring data to obtain a corrected finite element model representing the state after the damage event, and constructing a corrected structure balance equation based on the corrected finite element model; S6, resuming the benchmark simulation based on the corrected finite element model and the corrected structure balance equation until no damage event is identified; S7, outputting the long-term deformation and internal force prediction results of the bridge at the target year; Wherein, the damage events at least include vehicle impact and earthquake.

2. The long-term deformation prediction method for long-span rigid frame bridges according to claim 1, characterized in that, The benchmark simulation on the spatial finite element model comprises: S31, starting from the construction starting point, gradually advancing to the target prediction year in time sequence, dividing the total time into consecutive time steps; S32, performing a benchmark simulation loop in each current time step; S33, advancing the time to the next time step and repeating S32 until the simulation time reaches the target prediction year.

3. The long-term deformation prediction method for long-span rigid frame bridges according to claim 2, characterized in that, The execution of the benchmark simulation loop comprises: S32a, applying the concrete shrinkage and creep effects up to the previous time step as equivalent node forces to the current structure configuration; S32b, solving the structure balance equation to update the structure's displacement and bar-end force through iterative calculation until the solution converges; S32c, based on the updated bar-end force, calculating the shrinkage and creep equivalent node forces of the next time step through a recursive formula that fits the creep coefficient based on an exponential function.

4. The long-term deformation prediction method for long-span rigid frame bridges according to claim 1, characterized in that, The establishment of the benchmark probability distribution model comprises: Collecting historical monitoring data of the bridge under normal operating conditions to form a training sample set; Based on the training sample set, establishing a benchmark probability distribution model for each mechanical state monitoring parameter and determining its statistical characteristic value; Based on the hypothesis testing method, setting the control upper limit and control lower limit of the abnormal index for identifying damage events.

5. The long-term deformation prediction method for long-span rigid frame bridges according to claim 1, characterized in that, The identification of damage events based on the pre-established benchmark probability distribution model comprises: S41, obtaining monitoring data in the actual time interval corresponding to the current time step that has not participated in the identification judgment, and taking it as new monitoring data; S42, calculating the comprehensive abnormal index of the new monitoring data based on the benchmark probability distribution model; S43, using the sliding time window technology to analyze the change trend of the comprehensive abnormal index in consecutive time steps; S44, when the comprehensive abnormal index exceeds the control upper limit for k consecutive time steps or is below the control lower limit for k consecutive time steps, it is determined that a damage event has occurred; Wherein, k is a preset sensitivity parameter.

6. The long-term deformation prediction method for long-span rigid frame bridges according to claim 5, characterized in that, The calculation of the abnormal index of the new monitoring data based on the benchmark probability distribution model comprises: S42a, for each new monitoring data, calculate its standardized residual relative to the baseline probability distribution model; S42b, based on the standardized residual, obtain the local anomaly index of each monitoring data by Mahalanobis distance calculation; S42c, according to the spatial distribution characteristics of the monitoring data, the local anomaly indexes are weighted and fused to form the comprehensive anomaly index of the whole structure.

7. The long-term deformation prediction method for long-span rigid frame bridges according to claim 1, characterized in that, The damage correction based on the monitoring data on the spatial finite element model comprises: S51, based on the spatial distribution of the monitoring data and its anomaly index identified by triggering the damage event, identify the damage area in the structure; S52, in the damage area, minimize the residual between the dynamic characteristics or static response of the spatial finite element model and the measured data by weighted least squares method, so as to identify and correct the structural parameters of the damage area; S53, based on the structural parameters of the corrected damage area, correct all node coordinates of the spatial finite element model to obtain the corrected finite element model.

8. The long-term deformation prediction method for long-span rigid frame bridges according to claim 1, characterized in that, The monitoring data includes data collected by at least one of the following sensors: strain data collected by fiber grating sensors, rotation angle data collected by inclinometers, and vibration data collected by acceleration sensors.

9. The long-term deformation prediction method for long-span rigid frame bridges according to claim 1, characterized in that, The long-term deformation and internal force prediction results of the bridge in the target service life include: time-varying deformation envelope of the whole bridge, internal force time history of key sections, and long-term stress distribution cloud diagram.

10. A long-span rigid frame bridge long-term deformation prediction system characterized by, The system adopts the long-term deformation prediction method of the long-span rigid frame bridge according to any one of claims 1 to 9, and specifically comprises: A model establishing module is configured to establish a spatial finite element model based on design drawings, material properties and boundary conditions of the bridge; An equation constructing module is connected with the model establishing module and configured to construct a structure balance equation considering geometric nonlinearity and concrete shrinkage and creep effect based on the spatial finite element model; A cyclic simulation module is connected with the equation constructing module and configured to perform baseline simulation on the spatial finite element model based on the structure balance equation to determine the baseline mechanical state evolution from the start of construction to the target prediction year; An event identifying module is connected with the cyclic simulation module and configured to identify damage events based on a pre-established baseline probability distribution model; An event correcting module is connected with the event identifying module and configured to pause the baseline simulation when a damage event is identified, and perform damage correction on the spatial finite element model based on the monitoring data to obtain a corrected finite element model representing the state after the damage event, and construct a corrected structure balance equation based on the corrected finite element model; A correction simulation module is connected with the event correcting module and configured to perform baseline simulation again based on the corrected finite element model and the corrected structure balance equation; A result predicting module is connected with the correction simulation module and configured to output long-term deformation and internal force prediction results of the bridge in the target service life.

Images