Structural performance evaluation method guided by digital object fusion twin substructure and multi-physical parameter update

Abstract

The invention discloses a structure performance evaluation method guided by a number object fusion twin substructure and multi-physical parameter updating. The method comprises the following steps: automatically segmenting and extracting crack width information based on a deep learning model and an image processing algorithm; introducing the crack width into a mechanical model and solving a crack area rigidity reduction coefficient based on the bonding stress distribution and mechanical equilibrium relationship between the reinforcing steel bar and the concrete in the tensile area, and realizing the local crack area rigidity updating of the concrete structure; a structure multi-parameter identification method based on a substructure and monitoring data is provided, non-contact monitoring data is used as a system input quantity, a finite element model calculation result is used as an observed quantity, and identification of a plurality of constitutive parameters such as support rigidity is achieved through an unscented Kalman filtering method (VAFUKF) of a variance adaptive forgetting factor; and correcting the local stiffness reduction coefficient and the multiple structural parameters into a finite element model to realize deep fusion of non-contact sensing data and the finite element model, and performing structural performance evaluation. According to the method, deep fusion of'non-contact sensing data-finite element model 'is realized, and data support can be provided for structural safety assessment.

Description

Technical Field

[0001] The present invention relates to the fields of structural safety evaluation and data processing technology, and in particular to a digital-physical fusion twin structure and a multi-physical parameter updating-guided structural performance evaluation method. Background Art

[0002] Existing structures are prone to changes in structural stress due to defects such as cracks and changes in material properties, which have a great impact on the evaluation of structural service performance. Bridge structure health monitoring and detection technology can provide structural disease information in a timely and effective manner, making it possible to evaluate the condition of the structure. However, for structures with large size and complex stress states, it is difficult to truly and effectively analyze and evaluate the service performance of the structure using only structural inspection and monitoring data. The rapid development of finite element theory has facilitated the analysis of bridge structures, and finite element improvement methods using inspection and monitoring data have been widely studied. However, the existing methods still have the following problems:

[0003] (1) Currently, many new methods for rapid detection of external defects have been proposed, but they mainly focus on the qualitative identification, location, and quantification of structural defects based on detection data. The detection data cannot be well linked to the local structural degradation performance of the structure. In addition, the existing non-contact monitoring methods mainly focus on measuring structural deformation, and there are relatively few studies on the connection between monitoring data and the intrinsic performance of the structure. In addition, the academic community generally recognizes that it is difficult to effectively analyze and evaluate the safety performance of a structure using only structural detection or monitoring data.

[0004] (2) Finite element models have been widely studied and applied in the field of civil engineering, and are one of the important methods for integrating non-contact sensing data (including detection and monitoring data). However, existing model correction technologies mainly modify the internal parameters of the model to achieve consistency between theoretical values ​​and monitoring values ​​at the measuring points. The corrected finite element model parameters often reflect the overall structural properties, and the application and mining of the detection and monitoring data are not sufficient, and the structural evaluation results are prone to large errors. Therefore, if the non-contact sensing data can be deeply integrated with the finite element model during the simulation and analysis process, it will be expected to break through the bottleneck of bridge structure status analysis and evaluation.

[0005] In summary, the finite element analysis method used in the existing structural safety performance assessment process has a very limited degree of fusion of inspection and monitoring data, especially non-contact sensing data. When conducting structural analysis and assessment of existing structures, there are problems such as inaccurate assessment structure. Summary of the invention

[0006] Purpose of the invention: In view of the above technical problems, the present invention proposes a structural performance evaluation method guided by a digital-physical fusion twin structure and multi-physical parameter update, so as to overcome or partially overcome the bottleneck problem that non-contact sensing data and finite element models are difficult to integrate with each other, and provide data support for structural safety assessment.

[0007] In order to achieve the above technical objectives, the present invention adopts the following technical means:

[0008] The digital-physical fusion twin structure and multi-physical parameter update-oriented structural performance evaluation method includes the following steps:

[0009] S1. Automatically segment crack images based on deep learning models and image processing algorithms, and extract crack width information;

[0010] S2. Based on the bond stress distribution and mechanical equilibrium relationship between the steel bar and the concrete in the tension zone, the crack width is introduced into the mechanical model and the stiffness reduction coefficient of the crack area is solved to achieve the stiffness update of the local crack area of ​​the concrete structure;

[0011] S3, a structural multi-parameter identification method based on substructure and monitoring data, using non-contact monitoring data as system input and finite element model calculation results as observations, and realizing multiple constitutive parameter identification through the unscented Kalman filtering method with variance adaptive forgetting factor;

[0012] S4. Based on the local stiffness reduction coefficient and structural parameters, the finite element model is modified to achieve deep integration of non-contact sensing data and the finite element model to conduct structural performance evaluation.

[0013] The digital-physical fusion twin substructure integrates non-contact detection and monitoring data with the finite element model of the substructure. The non-contact monitoring data includes: intelligent crack detection data based on a deep learning model and structural deformation monitoring data based on laser scanning technology.

[0014] Step S1 specifically includes the following sub-steps:

[0015] S11, automatic segmentation of crack morphology using deep model;

[0016] S12, threshold segmentation is performed on the crack pixel-level recognition result, that is, the predicted image is binarized according to a specific threshold, and then the total number of crack pixels is calculated, thereby obtaining the pixel-level representation of the crack image, and then calculating the total crack area A c ;

[0017] S13, extract the crack skeleton by the medial axis method and calculate its total length. Skeletonization is to convert the multi-pixel representation of the crack into a skeleton with a single pixel width. After extracting the crack skeleton, further calculate the crack length L c ;

[0018] S14, automatically calculates the average width of cracks based on crack area and length: in, represents the average crack width in the image, Ac is the total crack area, L c is the crack length, H is the image height, W is the image width, i is the position in the image height direction, j is the position in the image width direction, represents the result of binary segmentation, sk(i,j) represents the extracted skeleton result, l represents the length of the pixel in the image, Indicates the ratio between the actual physical size and the image pixels.

[0019] Step S2 specifically includes the following sub-steps:

[0020] S21. Combine the constitutive assumptions of reinforced concrete and other materials and the strain distribution assumptions of the crack section to establish the equilibrium equation of reinforced concrete in the tension zone, and introduce the crack damage factor to establish the average stiffness calculation model of the cracked section;

[0021] S22. Based on the bond-slip theory, the bond force distribution near the crack is set, the crack width is introduced to combine the above equations to solve the damage factor by bisection method, and finally the stiffness reduction coefficient of the crack area is solved;

[0022] S23. A structural finite element model is established based on the stiffness method and units are divided. The units in the crack area are reduced based on the stiffness reduction coefficient calculated above, and finally the stiffness update of the local crack area of ​​the concrete structure is achieved.

[0023] The specific steps of solving the crack damage factor include:

[0024] S211, set the upper limit of the crack damage factor D r =1 and lower limit D l =0, and the initial value of the crack damage factor D 0 =0.5;

[0025] S212, using the crack damage factor D of step k k , based on the cross-sectional force equilibrium equation, solve the steel stress and concrete tensile stress of the uncracked section;

[0026] S213, based on the force equilibrium equation along the longitudinal direction of the beam, solve the steel stress and concrete stress of the remaining sections.

[0027] Tensile stress;

[0028] S214, crack calculation formula based on bond-slip theory Solve for the calculated width w of the crack cal ;

[0029] S215, if the calculated width of the crack w cal Greater than the actual crack width w i , it means that the crack damage factor is too large, so Dr =D k , D k+1 =0.5(D l +D k ); otherwise, it means that the crack damage factor is too small, then D l =D k , D k+1 =0.5(D k +D r );

[0030] S216, repeat the above process until the calculated width of the crack w cal and the true crack width w i The difference is less than the tolerance.

[0031] The specific method for obtaining the structural deformation monitoring data based on the laser scanning technology comprises the following steps:

[0032] Based on 3D scanning technology, the 3D point cloud data of the structural deformation process is acquired in real time. Then, the reflectivity histogram is calculated based on the reflectivity of the point cloud data. Finally, the peripheral data with high reflectivity is eliminated through the outlier filtering algorithm, leaving only the circular target data.

[0033] First, the point cloud noise information is filtered using a filtering algorithm, then the point cloud is projected onto a plane based on the principal component analysis algorithm and the plane is fitted, and finally, the plane contour is extracted using the boundary function;

[0034] The Pratt least squares method is used to fit the contour line into a circle, from which the coordinates of the plane center can be calculated. By transforming this center point into the original coordinate system, the three-dimensional coordinates of the target center space in the actual coordinate system are further obtained, thereby obtaining the structural deformation monitoring data.

[0035] In step S3, the multi-physical parameter identification method based on substructure and monitoring data specifically includes the following sub-steps:

[0036] S31, selecting commonly used constitutive model parameters of steel bars and concrete of RC frame structure as parameters for subsequent updating, the commonly used constitutive model parameters of concrete include: support stiffness, steel bar elastic modulus, steel bar yield strength, concrete elastic modulus, concrete peak strain, and concrete strength;

[0037] S32, discretizing the loading process, the main structure module solves the overall response of the structure, extracts the displacement y, axial force N, and bending moment M interface condition parameters at the substructure interface, and transmits them to the substructure module;

[0038] S33, after the substructure receives the interface displacement y, axial force N and bending moment M information transmitted by the main structure, the substructure obtains the structural displacement response at this time through calculation, and starts the parameter update program when the response of the substructure exceeds the error range;

[0039] S34. Based on the VAFUKF algorithm, the optimal parameters are identified and updated to the main structure finite element model. The main structure finite element model carries out the next discrete process analysis according to the latest parameters. The above steps are repeated until the loading is completed.

[0040] The VAFUKF algorithm specifically includes the following steps:

[0041] Step 1, determine the information difference sequence between the actual observation value and the predicted value of the system: In the formula, y k+1 represents the measured value, represents the predicted value at step k;

[0042] Step 2: Determine the divergence criterion and adaptive forgetting factor θ k ,in In the formula, tr[] represents the rank of the matrix, represents the error vector of the k+1th step, E[] is the expected operation, is the transpose of the information difference vector;

[0043] Step 3, based on the adaptive forgetting factor θ k Further calculate the variance value of the k-th step observation: Where P yy represents the variance matrix of the k-th step observation, The weight coefficient of the i-th sampling point, is the predicted observation value of the i-th sampling point at the k+1th step, is the mean of the predicted observations at step k+1, R k+1 is the observation noise covariance matrix of the k+1th step;

[0044] Step 4, calculate the cross-covariance matrix of observation and state quantity, which can be expressed as:

[0045] Step 5, the Kalman gain matrix calculation process can be expressed as:

[0046] Step 6, finally based on the k+1th step observation value y k+1 and the Kalman gain matrix K k+1 Update the prior estimates of the mean and variance:

[0047] The method further includes the step of correcting the finite element model integrated with the non-contact sensing data, wherein the correction step includes the following steps:

[0048] For structures with existing cracks, the established finite element model first considers the cracks and other disease information, and the unit stiffness is weakened to ensure that the established finite element model takes the crack information into account;

[0049] Select local parts where the boundary or material properties are difficult to identify, such as columns connected to supports and beam-column connection nodes, and invert multiple parameters of the parts through monitoring data, including: steel bar elastic modulus, steel bar yield strength, concrete elastic modulus, concrete peak strain and concrete strength;

[0050] The initial finite element model is corrected and updated from two levels: detection and monitoring data. First, the stiffness reduction coefficient is calculated based on the crack detection data, and the stiffness reduction coefficient of the cracked area is updated in the finite element model. Then, the bottom column of the structure is selected as a substructure and a substructure finite element model is established. According to the deformation data of the actual structure, the substructure and the unscented Kalman filtering method of the variance adaptive forgetting factor are used to realize the multi-parameter identification of support stiffness, steel bar elastic modulus, steel bar yield strength, concrete elastic modulus, concrete peak strain and concrete strength, and the identified multi-parameters are updated in the finite element model.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] (1) Realize the fusion of non-contact sensing data and finite element models: In order to solve the problem of long-term separation between inspection and monitoring data and finite element models, this application proposes a framework for the fusion of non-contact sensing data (including inspection and monitoring data) and finite element models. The framework adopts a two-stage update strategy: first, the local unit stiffness information of the structure is updated based on the intelligent crack detection data and the mechanical model, and then multiple parameters such as the support stiffness of key parts are identified based on the non-contact monitoring data and the substructure finite element model. To a certain extent, a finite element model that conforms to the service status of the structure is established, providing strong support for structural safety assessment.

[0053] (2) The stiffness reduction coefficient of the crack area is easy to solve: In order to solve the problem that crack detection data and the local concrete structure degradation performance model are difficult to determine, the present invention proposes a method for updating the local stiffness of the structure based on intelligent crack detection data and a mechanical model. The method automatically extracts the crack width information based on a deep learning model and an image processing method, introduces the measured crack width into the above-mentioned mechanical model and solves the stiffness reduction coefficient of the crack area, thereby realizing a rapid update of the stiffness of the local crack area of ​​the concrete structure.

[0054] (3) Effectively identify the constitutive parameters of key parts of the structure: In order to solve the problem that the internal parameters of the key parts of the structure are difficult to obtain, the present invention proposes a substructure multi-parameter identification method based on substructure and VAFUKF. This method establishes a substructure finite element model of the key parts, uses the non-contact monitoring data obtained by laser scanning as the system input, and the calculation results of the substructure finite element model as the observation quantity. Through VAFUKF, the identification of multiple parameters such as support stiffness is realized. The identification accuracy is high and is closer to the actual response of the structure during static monotonic loading.

[0055] (4) Improving the performance prediction accuracy of complex structures with existing damage: In order to solve the problem of modeling and performance analysis of existing damage results, this paper considers existing damage such as cracks through detection data, identifies important structural parameters such as boundary conditions and material properties of the finite element model through monitoring data, and establishes a high-fidelity finite element model of the existing damaged structure. The calculation accuracy of the initial stiffness and ultimate bearing capacity of the structure is greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flow chart of the structural performance evaluation method for integrating non-contact sensing data and finite element model proposed in the present invention;

[0057] Figure 2 The intelligent crack segmentation model based on deep learning proposed in the present invention;

[0058] Figure 3 It is a flow chart of updating the local stiffness of an existing cracked structure proposed by the present invention;

[0059] Figure 4 Schematic diagram of the method for solving the crack damage factor proposed by the present invention;

[0060] Figure 5 This is a schematic diagram of structural deformation monitoring based on laser scanning technology and Pratt algorithm proposed in the present invention;

[0061] Figure 6 The structure multi-parameter identification method based on substructure and VAFUKF algorithm proposed in the present invention;

[0062] Figure 7 This is a flow chart of the VAFUKF structure multi-parameter identification algorithm proposed by the present invention;

[0063] Figure 8 This is a diagram of the integration of non-contact sensing data and finite element model proposed by the present invention;

[0064] Fig. 9 Schematic diagram of the experimental framework proposed for the example;

[0065] Fig.10 Figure 1. The test device diagram for the example;

[0066] Fig.11 (a) is the finite element model of the test frame structure proposed in the example.

[0067] Fig.11 (b) is the Joint2D unit of the test frame structure proposed in the example.

[0068] Fig.11 (c) Pinching4 constitutive model of the experimental frame structure proposed in the example;

[0069] Fig.12 The crack segmentation results based on the deep learning model proposed in the example;

[0070] Fig.13 (a) is the multi-parameter support stiffness identification result of the structure proposed in the example.

[0071] Fig.13 (b) is the identification result of the multi-parameter steel bar elastic modulus of the structure mentioned in the example.

[0072] Fig.13 (c) is the multi-parameter steel bar yield strength identification result of the structure mentioned in the example.

[0073] Fig.13 (d) is the identification result of the multi-parameter concrete elastic modulus of the structure proposed in the example.

[0074] Fig.13 (e) is the peak strain identification result of the multi-parameter concrete structure proposed in the example.

[0075] Fig.13 (f) is the multi-parameter concrete strength identification result of the structure proposed in the example;

[0076] Fig.14 (a) is the comparison of the first-level forward cycle results of the example.

[0077] Fig.14 (b) is a comparison of the results of the entire low-frequency reciprocating process mentioned in the example. DETAILED DESCRIPTION

[0078] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0079] Some of the definitions involved in this application are as follows:

[0080] Non-contact sensing data: refers to the detection and monitoring data obtained by non-contact measurement technology, specifically including intelligent crack detection data based on optical cameras and structural deformation monitoring data based on laser scanning technology.

[0081] Substructure and main structure: The main structure is the overall finite element model of the structure, and the substructure is the finite element model of the local area.

[0082] In view of the problems existing in the aforementioned prior art, this embodiment provides a structural performance evaluation method guided by a digital-physical fusion twin substructure and multi-physical parameter update to achieve structural service performance evaluation. First, in view of the difficulty in determining the crack detection data and the local concrete structural degradation performance model, based on the intelligent crack detection data and the crack mechanics model, the stiffness update of the local crack area of ​​the concrete structure is achieved by solving the stiffness reduction coefficient of the crack area. Then, in view of the difficulty in obtaining the internal parameters of the key parts of the structure, a structural multi-parameter identification method based on substructures and monitoring data is used to identify multiple constitutive parameters such as support stiffness through the variance adaptive forgetting factor unscented Kalman filtering method (VAFUKF). Finally, the structural parameters and the local stiffness reduction coefficient are updated to the finite element model to achieve deep fusion of non-contact sensing data and the finite element model.

[0083] See also Figure 1 The method mainly includes two parts: a method for processing non-contact sensing data (including non-contact detection and monitoring data) and a fusion of detection and monitoring data with a finite element model. The method for acquiring and processing non-contact sensing data corresponds to steps S1-S7, and the fusion of detection and monitoring data with a finite element model corresponds to steps S8-S9. The method for evaluating structural performance by fusing non-contact sensing data with a finite element model includes the following steps:

[0084] Step S1, using deep learning and image processing methods to achieve crack segmentation. Using deep learning models to achieve automatic crack morphology segmentation. Figure 2As shown in the figure, the model consists of multiple encoder modules, upsampling modules, feature fusion modules, side output modules and output modules. The encoder module has 5 stages, each of which includes 2 convolutional layers with a convolution kernel size of 3×3 and a stride of 1. There is only one activation function layer or normalization layer for each convolutional layer. There is a maximum pooling layer between each stage, and the size of the pooling layer is 2×2. The model includes three upsampling modules, each of which includes a 1×1 convolutional layer, a LN_VT layer, an H-Swish layer and a bilinear upsampling layer. The output module includes the main output OB module and the side output SOB module. The OB module mainly includes a conventional 1×1 convolutional layer and a transposed convolutional layer. The side output block SOB includes two 3×3 convolutional layers and a LN_VT layer. This module is not needed in the inference stage and does not bring any additional computational work. In addition, three feature fusion modules with similar structures are used in the deep learning model to enable information of different scales to flow freely in the network and effectively fuse the extracted features. This design enables the network to capture both high-level global semantic features and low-level fine-grained information, thereby improving the accuracy and efficiency of crack segmentation.

[0085] Step S2, such as Figure 3 As shown, the structural stiffness loss is quantified based on the crack distribution and mechanical model. First, the equilibrium equation of reinforced concrete in the tension zone is established by combining the constitutive assumptions of reinforced concrete and other materials and the strain distribution assumptions of the crack section, and the damage factor is introduced to establish the average stiffness calculation model; then, based on the bond-slip theory, the bond force distribution near the crack is set, and the crack width is introduced to combine the above equations to solve the damage factor by bisection, and finally the stiffness reduction coefficient of the crack area is solved; finally, based on the stiffness method, a structural finite element model is established and the units are divided, and the units in the crack area are reduced according to the stiffness reduction coefficient calculated above, so as to realize the local damage update of the structure. This step includes the following sub-steps:

[0086] Step S201, setting the constitutive model of the steel bar to a linear elastic model, that is: σ s =E s ε s The Hognestead E model is selected as the compressive constitutive model of concrete, and the linear elastic model is selected as the tensile constitutive model.

[0087] Step S202: the three sections at different positions are section 1, section 2 and section 3. Figure 3 As shown, j Represents the distance between the two cracks. For the uncracked section (x = 0), the steel bar and the concrete in the compression zone satisfy the plane section assumption, and the strain distribution of the concrete in the tension zone is in the form of a three-fold line, with the maximum stress occurring at the steel bar position. If the transmission length of the bond stress is large enough, it is considered that the steel bar strain ε of the section iss and the tensile strain of concrete at the reinforcement ε tc are equal, and the entire section conforms to the plane section assumption. In the actual structure, ε s and ε tc Due to partial bonding failure, they are no longer equal. Therefore, the damage factor D is introduced to consider the damage caused by slip, that is, ε tc =(1-D)ε s .

[0088] Step S203, the damage factor solution process is as follows: Figure 4 As shown, the upper and lower limits of the damage factor D are established. l =0 and D r = 1, and the initial value of the damage factor D 0 =0.5.

[0089] Step S204, using the damage factor D of step k k , based on the cross-sectional force equilibrium equation and Solve the steel stress and concrete tensile stress in the uncracked section.

[0090] Step S205, based on the force equilibrium equation along the longitudinal direction of the beam, solve the steel bar stress and concrete tensile stress of the remaining sections.

[0091] Step S206, using formula Solve for the crack width w cal .

[0092] Step S207, if the crack calculated width w cal Greater than the actual crack width w i , it means that the damage factor is too large, so D r =D k , D k+1 =0.5(D l +D k ). Otherwise, it means the damage factor is too small, then D l =D k , D k+1 =0.5(D k +D r ).

[0093] Step S208, repeat the above process until the crack width w is calculated. cal and the true crack width w i The difference is less than the tolerance.

[0094] Step S209, when D = 0, it proves that the bonding performance of the two is good, and the steel bar strain ε s and concrete tensile strain ε tcWhen D = 1, it indicates that the bonding performance is completely destroyed. <x<l j / 2), its strain distribution is similar to that of section 1, but the steel bars in the tension zone no longer satisfy the flat section assumption, which can be deduced by the longitudinal force equilibrium condition of the steel bars. j / 2), the tensile stress on concrete can be neglected on the basis of the above section 3.

[0095] Step S210: The steel bar strains of Section 2 and Section 3 need to be derived from the force balance of the steel bars along the longitudinal direction. From the balance condition of the concrete and steel bars in the tension zone, it can be known that:

[0096]

[0097] In the formula, Represents the diameter of the steel bar, A te Represents the effective tensile area of ​​concrete, A s represents the steel bar area, σ s (x) is the tensile reinforcement stress, σ tc (x) is the tensile stress of concrete at the point where the reinforcement is applied.

[0098] Step S211, the above concrete tensile three-fold strain distribution, the concrete can calculate the effective tensile area that is more consistent with the stress on the crack interface, specifically as follows:

[0099]

[0100] In the formula, b represents the width of the rectangular beam, a represents the height of the steel bar layer, and y ouc represents the height of the neutral axis of the uncracked section, h is the height of the beam section, and d represents the effective height of the beam section.

[0101] Step S212: From the equilibrium equation, it can be known that if σ at the x=0 section us and σ utc Known, the steel stress and concrete tensile stress in other sections can be expressed as:

[0102]

[0103] From the above formula, we can know that: when σ us and σ utc If the distribution of the bonding stress τ(x) is known, then we can get Any cross section σ s (x) and σ tc (x). The bonding stress is 0 at the uncracked section, but The bonding stress τ(x) is basically distributed linearly.

[0104] Step S213, the distribution of the bonding force τ(x) near the crack is expressed as:

[0105]

[0106] In the formula, τ b represents the bonding stress, τ f Represents the residual bonding stress (generally 0.15τ max ), τ max Represents the maximum bonding stress (usually taken as 2f ck ), f ck represents the standard value of concrete compressive strength, a represents the bonding coefficient, s 1 ,s 2 ,s 3 is the adhesion range, which takes values ​​as 0.6mm, 0.6mm and 1mm respectively.

[0107] Step S214, if the crack width information w i It is known that the average bonding stress near the crack can be expressed as On this basis, combined with the τ(x) distribution, we can further obtain The expression of τ(x) in the range:

[0108]

[0109] The strain state of each section needs to be determined by the reinforcement strain ε s , height of neutral axis y 0 , concrete tensile strain ε at the reinforcement position tc and section rotation There are four unknown quantities, but there are only two force equilibrium equations on the interface, namely:

[0110]

[0111] Step S215: Since the steel bars and the concrete in the compression zone meet the plane section assumption, the steel bar strain ε of the uncracked section is s The neutral axis height y 0 and section rotation In addition, the tensile strain of concrete can be expressed as ε utc =(1-D)ε us According to the stick-slip theory, cracks are generally caused by relative slip between concrete and steel bars, and the crack width can be expressed as:

[0112]

[0113] Step S216: When the damage factor D is known, The number of unknown quantities in any section is reduced from 4 to 2, which can be solved by the force equilibrium equation of the above interface. From this, we can know the section stiffness The average stiffness within a certain range can be expressed as

[0114]

[0115] The average stiffness of the cracked area unit is calculated and the regional stiffness of the cracked unit is updated by updating the reduction coefficient.

[0116] Step S3, such as Figure 5 As shown, non-contact monitoring of structural deformation is performed based on laser scanning technology, and the method of measuring structural deformation by fitting the target center using the Pratt least squares method provides a data basis for subsequent structural parameter identification. This step includes the following sub-steps:

[0117] Step S31, based on the three-dimensional scanning technology, the three-dimensional point cloud data of the structure at different stages is acquired in real time, and then the reflectivity histogram is calculated according to the reflectivity of the point cloud data, and finally the peripheral data with high reflectivity is eliminated through the outlier filtering algorithm, so that only the circular target data is left.

[0118] Step S32: First, filter the point cloud noise information using a filtering algorithm. Then, project the point cloud onto a plane and fit the plane based on a principal component analysis (PCA) algorithm. Finally, use a boundary function to extract the plane contour.

[0119] Step S33, the contour line is fitted into a circle using the Pratt least squares method, thereby calculating the coordinates of the plane center. By transforming this center point into the original coordinate system, the three-dimensional coordinates of the target center space in the actual coordinate system can be further obtained, thereby obtaining the deformation information of the structure.

[0120] Step S4, inferring the current state parameters of the structure based on the monitoring information such as deformation, and selecting the commonly used constitutive model parameters of steel bars and concrete of the RC frame structure as the parameters for subsequent updating;

[0121] Step S5, discretize the loading process, the main structure module solves the overall response of the structure, extracts the displacement y, axial force N, bending moment M and other interface condition parameters at the substructure interface and transmits them to the substructure module, such as Figure 6 As shown;

[0122] Step S6, after the substructure receives the information of interface y, axial force N, bending moment M, etc. transmitted by the main structure, the substructure obtains the structural displacement response at this time through calculation, and starts the parameter update program when the response of the substructure exceeds the error range;

[0123] Step S7, Figure 7As shown, based on the VAFUKF algorithm, the optimal parameters are identified and updated to the main structure finite element model, and the main structure finite element model carries out the next discrete process analysis according to the latest parameters. The above steps are repeated until the loading is completed. The algorithm includes the following steps:

[0124] Step S71, determine the information difference sequence between the real observed value and the predicted value of the system:

[0125]

[0126] In the formula, y k+1 represents the measured value, Represents the predicted value at step k.

[0127] Step S72, determine the divergence criterion and the adaptive forgetting factor θ k

[0128]

[0129] Step S73, based on the adaptive forgetting factor θ k Further calculate the variance value of the k-th step observation:

[0130]

[0131] Step S74, calculating the cross-covariance matrix of the observation quantity and the state quantity can be expressed as:

[0132]

[0133] Step S75, the Kalman gain matrix calculation process can be expressed as:

[0134]

[0135] Step S76, finally based on the observed value y k+1 and K k+1 Update the prior estimates of the mean and variance:

[0136]

[0137] When identifying the internal parameters of the structure based on the UKF identification algorithm, the large noise of the monitoring data can easily affect the stability of the parameters, which eventually leads to problems such as the accumulation of identification errors and even calculation interruption. In response to this problem, this embodiment, based on the UKF, cross-combines the adaptive forgetting factor and the filter judgment criterion to propose an unscented Kalman filter method (VAFUKF) based on the variance adaptive forgetting factor, realizes the multi-parameter identification method of the substructure constitutive model, and enhances the robustness of the multi-parameter identification of the structure. The VAFUKF algorithm provided in this embodiment adds a variance adaptive module on the basis of the UKF algorithm. This module uses the square sum of the difference between the measured value and the predicted value as the basis for judging the filter divergence. On this basis, the forgetting factor is added to correct the observation covariance matrix and the mutual covariance matrix, which greatly avoids the divergence of the identification process. In addition, this module can effectively reduce the influence of the initial value error on the parameter identification results.

[0138] like Figure 8 As shown, in steps S8-S9, the non-contact sensing data (including detection and monitoring data) is fused with the finite element model to perform structural analysis and evaluation.

[0139] Step S8: The established finite element model needs to first consider the damage information such as cracks, and by weakening the unit stiffness and other methods, it is ensured to a certain extent that the established finite element model takes the crack information into account.

[0140] Step S9, select local parts whose boundaries or material properties are difficult to identify, such as columns connected to supports, invert multiple parameters such as support stiffness of the parts through monitoring data, and carry out correction and update of the initial finite element model from two levels of detection and monitoring data. This step includes the following sub-steps:

[0141] Step S91, calculating the stiffness reduction coefficient based on the crack detection data according to the above method, and updating the stiffness reduction coefficient of the cracked area in the finite element model.

[0142] Step S92, select the bottom column of the structure as a substructure and establish a substructure finite element model. According to the deformation data of the actual structure, the substructure and the Kalman filter method of the adaptive forgetting factor proposed in this embodiment are used to realize the identification of multiple parameters such as the support stiffness, and the identified multiple parameters are updated in the finite element model. Subsequently, the structural performance prediction can be carried out based on the finite element model after the inspection and monitoring data are integrated, providing new ideas for the safety assessment of existing structures.

[0143] The correction process of the finite element model integrating the non-contact sensing data specifically includes:

[0144] The finite element model of the frame structure was established using OpenSEES. The nonlinear fiber beam-column unit was used to simulate the beam-column components of the frame structure. Different cracking degrees corresponded to different local unit properties. The concrete constitutive model was Concrete02, and the steel constitutive model was Steel02. Considering that the node area of ​​the RC frame structure is prone to shear behavior during loading, the Joint2D unit was used to simulate the mechanical behavior of the node area. The constitutive model of the central shear spring of the Joint2D unit was Pinching4. The Pingching4 skeleton curve of the shear spring in the core area of ​​the node can be calculated by the modified pressure field theory. For the behavior of the reinforced concrete frame structure with slippage rather than rigid connection, the zero-length unit and the Hysteretic model were used to simulate this behavior. A total of two co-coordinate points were set at the column foot of the finite element model, and the spring constraint was realized by setting zero-length units between the co-coordinate points. The zero-length unit constitutive relationship uses the Hysteretic model to characterize the relationship between the force and displacement of the spring constraint at the column foot of the finite element model.

[0145] Based on the local stiffness loss calculation method of crack detection data, when the crack width and the bond force distribution near the crack are known, the equations for the distribution of steel bar strain and concrete tensile strain along the unit can be further derived through the longitudinal force balance between the steel bar and the concrete in the tensile zone, and finally the calculation of the average stiffness reduction factor within a certain range can be realized. According to the crack distribution, the stiffness involved in the crack is corrected, so that the established finite element model is more consistent with the actual state of the existing damage.

[0146] The material parameters of the local structure are identified based on the substructure and the corresponding monitoring data. The monitoring data used is the structural displacement obtained by non-contact measurement, and then the key parameters of the substructure finite element model are identified based on the VAFUKF method.

[0147] For reinforced concrete structures, the constitutive parameters that have a dominant influence on the structural output mainly include k, E s , f y , E c , ε c and f c 6 parameters. Among them, k represents the stiffness of the support reinforcement, E S represents the elastic modulus of the steel bar, f y Represents the yield strength of steel bars, E c represents the elastic modulus of concrete, ε c is the peak strain of concrete, f c is the concrete strength. The identified substructure multi-parameters are used to update the finite element results.

[0148] The solution of this embodiment is explained below by constructing an experimental scenario.

[0149] The test was conducted on a two-bay, two-span reinforced concrete frame structure. The geometric dimensions and cross-sectional dimensions of the reinforced concrete frame are as follows: Fig. 9 As shown. The material used is C40 concrete, the load-bearing longitudinal reinforcement model is HRB400, and the stirrup model is HPB400. The beam ends and column ends are encrypted according to the requirements of the specification. The beams and columns are prefabricated components, the node areas are connected by cast-in-place, and the steel bars are connected by welding. The loading area is pre-buried with steel plates and welded to the steel bars to avoid local pressure damage to the concrete during the loading process.

[0150] This test is divided into two stages. The first stage is the static monotonic loading stage, which is used to cause certain crack damage to the structure and obtain the inspection and monitoring data of the structure through non-contact measurement equipment, which is used to integrate the inspection and monitoring data to establish a finite element model that conforms to the actual state of the structure, laying a certain foundation for the subsequent structural performance prediction. The second stage is the low-cycle reciprocating stage, which is mainly used to verify whether the finite element model established previously can predict the unknown performance of the structure. The experimental device is arranged as follows: Fig.10 As shown, the MTS hydraulic servo actuator is responsible for lateral loading, and the loading mode is displacement loading. The vertical load is applied by a jack, and the oil pump has been calibrated in advance. In addition, two lateral supports are used to prevent the reinforced concrete structure frame from out-of-plane instability. In this test, lidar is used to obtain structural deformation monitoring data, while structural cracks are obtained using an optical camera.

[0151] In the first stage of static monotonic loading, the vertical load on the top of the column is first applied by loading the distribution beam, and then the vertical load is maintained constant. The MTS hydraulic servo actuator applies the frame structure to produce horizontal displacement of the top floor, loading to 18mm and then slowly unloading to 0mm. In the second stage, hysteresis loading is used, and the recording method is still displacement loading. The displacement level of loading is between 5mm-80mm, where 5mm is a loading level, and each cycle is repeated 3 times. When the bearing capacity of the reinforced concrete structure is reduced to 85% of the maximum bearing capacity, the loading process stops immediately.

[0152] like Fig.11 As shown in the figure, the initial finite element model of the frame structure is established using OpenSEES. During the modeling process, nonlinear fiber beam-column units are used to simulate the beam-column components of the frame structure. Considering the need for subsequent fusion of crack detection data, different cracking degrees correspond to different local unit properties. Therefore, multiple units are used to simulate the beam and column components taking into account the crack distribution. In terms of material constitutive model, the concrete constitutive model is Concrete02 and the steel bar constitutive model is Steel02.

[0153] like Fig.11As shown in (a), considering that the node area of ​​the RC frame structure is prone to shear behavior during loading, the Joint2D unit is used to simulate the mechanical behavior of the node area. The Joint2D model includes a total of 5 spring units. The 1 spring located inside the node is used to characterize the shear behavior of the core area, while the 4 springs located at the interface between the node and the beam and column units are mainly used to characterize the behavior of the unit end rotation caused by steel bar slippage.

[0154] like Fig.11 As shown in (b), the constitutive model of the shear spring at the center of the Joint2D unit is Pinching4. This model can effectively reflect the behaviors of structural stiffness degradation, strength degradation and pinching effect. The model includes skeletons such as monotonic loading and unloading-reloading during cyclic loading. The Pingching4 skeleton curve of the shear spring in the core area of ​​the node can be calculated by modified pressure field theory. For the behavior of the RC frame structure with slippage rather than rigid connection, the zero-length unit and Hysteretic model are used to simulate this behavior. A total of 2 co-ordinate points are set at the column foot of the finite element model, and the spring constraint is realized by setting zero-length units between the co-ordinate points. The zero-length unit constitutive relationship uses the Hysteretic model, which can characterize the relationship between the force and displacement of the spring constraint at the column foot of the finite element model.

[0155] According to the method disclosed in the present invention for fusing inspection and monitoring data with finite element models, the model is corrected. The following steps are carried out on the reinforced concrete frame structure in the embodiment: (1) Based on the pre-trained deep learning model, the crack images collected in the test are input to achieve automatic segmentation of cracks and automatic calculation of width information. Due to computer performance limitations, this study cuts the high-definition images obtained by the camera into images with a resolution of 600×800 pixels, and then intelligently segments the input crack images based on the pre-trained deep learning model. The segmentation results are shown in Figure 1. Fig.12 As shown. (2) Threshold segmentation is performed on the crack pixel-level recognition results, that is, the predicted image is binarized according to a specific threshold. Then the total number of crack pixels can be calculated, thereby obtaining the pixel-level representation of the crack image. (3) The total area of ​​the crack is calculated as Where s is the area of ​​a pixel in the image, H and W are the height and width of the image, Represents the scale factor. (4) The crack skeleton is extracted by the medial axis method and its total length is calculated. Skeletonization is to convert the multi-pixel representation of the crack into a skeleton with a single pixel width. After extracting the crack skeleton, the crack length L c The calculation formula is Where l represents the length or width of a pixel in the image (the length and width of a single pixel are equal), and Sk represents the skeleton of the crack. (5) Automatically calculate the average width of the crack in, Represents the average crack width in the image. (6) According to the bond-slip theory, it is generally believed that the occurrence of cracks is due to the relative slip between concrete and steel bars, and the crack width is equal to the relative slip between steel bars and concrete in the area near the crack. The relationship between the slip and the average bond stress can be calculated according to the CEB-FIP specification to obtain the bond force distribution near the crack. (7) The local stiffness loss calculation method based on crack detection data. When the crack width and the bond force distribution near the crack are known, the equations for the distribution of steel bar strain and concrete tensile strain along the unit can be further derived through the longitudinal force balance between the steel bar and the concrete in the tensile zone, and finally the calculation of the average stiffness reduction factor within a certain range can be realized. The stiffness reduction factor is shown in Table 1. The stiffness involved in the crack is corrected so that the established finite element model is more consistent with the actual state of the existing damage. (8) The monitoring data used selects the structural displacement obtained by non-contact measurement. The key parameters of the substructure finite element model are identified based on the VAFUKF method. The multi-parameter identification results are shown in Table 1. Fig.13 As shown, Fig.13 (a) is the support stiffness identification result, Fig.13 (b) is the identification result of the elastic modulus of the steel bar. Fig.13 (c) is the steel bar yield strength identification result, Fig.13 (d) is the identification result of concrete elastic modulus, Fig.13 (e) is the concrete peak strain identification result, Fig.13 (f) is the concrete strength identification result, and finally the identified substructure multi-parameters are updated to the finite element model.

[0156] (9) Based on the above steps, the finite element model is updated to effectively improve the accuracy of the prediction of the initial stiffness and maximum bearing capacity of the structure, better simulate the stress state of the existing structure, and improve the accuracy of the performance analysis of large structures under complex stress states.

[0157] In order to further verify the effectiveness of the method proposed in this paper, the corresponding low-cycle reciprocating loading is performed on the finite element after fusion of the existing inspection and monitoring data. Fig.14 (a) is the comparison of the results of the first cycle (i.e. the loading level is 5 mm) before and after the non-contact sensing data fusion. Fig.14 As shown in (a), after the non-contact sensing data is fused, the finite element simulation is closer to the test results, while the stiffness of the finite element model before updating is relatively high, which has a large deviation from the test results.

[0158] Fig.14 (b) is the comparison of the results of the entire low-frequency reciprocating process before and after the non-contact sensing data fusion. Fig.14As shown in (b), the initial finite element simulation results are significantly higher than the test results, while the finite element analysis results after non-contact sensing data fusion are closer to the test data. It can be seen that the non-contact sensing data method proposed in this paper can more accurately establish the finite element model of the existing structure and simulate its low-cycle reciprocating loading and other behaviors, and the results are closer to the actual situation.

[0159] Table 1 Element stiffness reduction factor

[0160]

[0161]

[0162] Table 1 shows the calculation results of the stiffness reduction coefficient of the unit where the crack area is located in the example; it is subsequently updated to the finite element model.

[0163] Table 2 Comparison of the predicted results of initial stiffness and ultimate bearing capacity of the structure

[0164]

[0165] As shown in Table 2, the initial stiffness of the first positive cycle of the finite element model before the fusion of non-contact sensing data is about 6.20kN / mm, while the test result is 4.68kN / mm, with an error of up to 32.47%. The initial stiffness of the first positive cycle of the finite element model after the fusion of non-contact sensing data is 4.79kN / mm, and the error is reduced to 2.35%. In terms of maximum bearing capacity, the maximum bearing capacity of the structure is about 132.50kN, and the maximum bearing capacity of the finite element model before the fusion of non-contact sensing data is about 143.66kN, with an error of 8.42%. The maximum bearing capacity of the finite element model after the fusion of non-contact sensing data is 133.63kN, and the error is reduced to 0.85%. It can be seen that the calculation accuracy of the initial stiffness and ultimate bearing capacity of the structure of the finite element model that integrates non-contact sensing data has been greatly improved.

[0166] In summary, the specific embodiments verify the feasibility of the scheme proposed in the present invention. The experimental results show that the finite element model established by integrating the inspection and monitoring data can better characterize the service behavior of the structure and effectively improve the accuracy of the prediction of the initial stiffness and maximum bearing capacity of the structure. The scheme proposed in the present invention can well simulate the stress state of the existing structure and improve the accuracy of the performance analysis of large structures under complex stress states.

[0167] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A structural performance evaluation method guided by digital-physical fusion twin structure and multi-physical parameter update, characterized in that: The steps include: S1. Automatically segment crack images based on deep learning models and image processing algorithms, and extract crack width information; S2. Based on the bond stress distribution and mechanical equilibrium relationship between the steel bar and the concrete in the tension zone, the crack width is introduced into the mechanical model and the stiffness reduction coefficient of the crack area is solved to achieve the stiffness update of the local crack area of ​​the concrete structure; S3, a structural multi-parameter identification method based on substructure and monitoring data, using non-contact monitoring data as system input and finite element model calculation results as observations, and realizing multiple constitutive parameter identification through the unscented Kalman filtering method with variance adaptive forgetting factor; S4. Based on the local stiffness reduction coefficient and structural parameters, the finite element model is modified to achieve deep integration of non-contact sensing data and the finite element model to conduct structural performance evaluation.

2. The method for structural performance evaluation guided by digital-physical fusion twin structure and multi-physical parameter updating according to claim 1 is characterized in that: The digital-physical fusion twin substructure integrates non-contact detection and monitoring data with the finite element model of the substructure. The non-contact monitoring data includes: intelligent crack detection data based on a deep learning model and structural deformation monitoring data based on laser scanning technology.

3. The method for structural performance evaluation guided by digital-physical fusion twin structure and multi-physical parameter updating according to claim 1 is characterized in that: Step S1 specifically includes the following sub-steps: S11, automatic segmentation of crack morphology using deep model; S12, threshold segmentation is performed on the crack pixel-level recognition result, that is, the predicted image is binarized according to a specific threshold, and then the total number of crack pixels is calculated, thereby obtaining the pixel-level representation of the crack image, and then calculating the total crack area A c ; S13, extract the crack skeleton by the medial axis method and calculate its total length. Skeletonization is to convert the multi-pixel representation of the crack into a skeleton with a single pixel width. After extracting the crack skeleton, further calculate the crack length L c ; S14, automatically calculates the average width of cracks based on crack area and length: in, represents the average crack width in the image, A c is the total crack area, L c is the crack length, H is the image height, W is the image width, i is the position in the image height direction, j is the position in the image width direction, represents the result of binary segmentation, sk(i,j) represents the extracted skeleton result, l represents the length of the pixel in the image, Indicates the ratio between the actual physical size and the image pixels.

4. The method for structural performance evaluation guided by digital-physical fusion twin structure and multi-physical parameter update according to claim 1 is characterized in that: Step S2 specifically includes the following sub-steps: S21. Combine the constitutive assumptions of reinforced concrete and other materials and the strain distribution assumptions of the crack section to establish the equilibrium equation of reinforced concrete in the tension zone, and introduce the crack damage factor to establish the average stiffness calculation model of the cracked section; S22. Based on the bond-slip theory, the bond force distribution near the crack is set, the crack width is introduced to combine the above equations to solve the damage factor by bisection method, and finally the stiffness reduction coefficient of the crack area is solved; S23. A structural finite element model is established based on the stiffness method and units are divided. The units in the crack area are reduced based on the stiffness reduction coefficient calculated above, and finally the stiffness update of the local crack area of ​​the concrete structure is achieved.

5. The method for structural performance evaluation guided by digital-physical fusion twin structure and multi-physical parameter updating according to claim 4 is characterized in that: The specific steps of solving the crack damage factor include: S211, set the upper limit of the crack damage factor D r =1 and lower limit D l =0, and the initial value of the crack damage factor D0 = 0.5; S212, using the crack damage factor D of step k k , based on the cross-sectional force equilibrium equation, solve the steel stress and concrete tensile stress of the uncracked section; S213, based on the force equilibrium equation along the longitudinal direction of the beam, solve the steel stress and concrete stress of the remaining sections. Tensile stress; S214, crack calculation formula based on bond-slip theory Solve for the calculated width w of the crack cal ; S215, if the calculated width of the crack w cal Greater than the actual crack width w i , it means that the crack damage factor is too large, so D r =D k , D k+1 =0.5(D l +D k ); otherwise, it means that the crack damage factor is too small, then D l =D k , D k+1 =0.5(D k +D r ); S216, repeat the above process until the calculated width of the crack w cal and the true crack width w i The difference is less than the tolerance.

6. The method for structural performance evaluation guided by digital-physical fusion twin structure and multi-physical parameter update according to claim 2 is characterized in that: The specific method for obtaining the structural deformation monitoring data based on the laser scanning technology comprises the following steps: Based on 3D scanning technology, the 3D point cloud data of the structural deformation process is acquired in real time. Then, the reflectivity histogram is calculated based on the reflectivity of the point cloud data. Finally, the peripheral data with high reflectivity is eliminated through the outlier filtering algorithm, leaving only the circular target data. First, the point cloud noise information is filtered using a filtering algorithm, then the point cloud is projected onto a plane based on the principal component analysis algorithm and the plane is fitted, and finally, the plane contour is extracted using the boundary function; The Pratt least squares method is used to fit the contour line into a circle, from which the coordinates of the plane center can be calculated. By transforming this center point into the original coordinate system, the three-dimensional coordinates of the target center space in the actual coordinate system are further obtained, thereby obtaining the structural deformation monitoring data.

7. The structural performance evaluation method of integrating non-contact sensing data and finite element model according to claim 3 is characterized in that: In step S3, the multi-physical parameter identification method based on substructure and monitoring data specifically includes the following sub-steps: S31, selecting the commonly used constitutive model parameters of steel bars and concrete of RC frame structure as the parameters for subsequent update; the commonly used constitutive model parameters of concrete include: support stiffness, steel bar elastic modulus, steel bar yield strength, concrete elastic modulus, concrete peak strain, and concrete strength; S32, discretizing the loading process, the main structure module solves the overall response of the structure, extracts the displacement y, axial force N, and bending moment M interface condition parameters at the substructure interface, and transmits them to the substructure module; S33, after the substructure receives the interface displacement y, axial force N and bending moment M information transmitted by the main structure, the substructure obtains the structural displacement response at this time through calculation, and starts the parameter update program when the response of the substructure exceeds the error range; S34. Based on the VAFUKF algorithm, the optimal parameters are identified and updated to the main structure finite element model. The main structure finite element model carries out the next discrete process analysis according to the latest parameters. The above steps are repeated until the loading is completed.

8. The structural performance evaluation method of integrating non-contact sensing data and finite element model according to claim 7 is characterized in that: The VAFUKF algorithm specifically includes the following steps: Step 1, determine the information difference sequence between the actual observation value and the predicted value of the system: In the formula, y k+1 represents the measured value, represents the predicted value at step k; Step 2: Determine the divergence criterion and adaptive forgetting factor θ k ,in In the formula, tr[] represents the rank of the matrix, represents the error vector of the k+1th step, E[] is the expected operation, is the transpose of the information difference vector; Step 3, based on the adaptive forgetting factor θ k Further calculate the variance value of the k-th step observation: Where P yy represents the variance matrix of the k-th step observation, The weight coefficient of the i-th sampling point, is the predicted observation value of the i-th sampling point at the k+1th step, is the mean of the predicted observations at step k+1, R k+1 is the observation noise covariance matrix of the k+1th step; Step 4, calculate the cross-covariance matrix of observation and state quantity, which can be expressed as: Step 5, the Kalman gain matrix calculation process can be expressed as: Step 6, finally based on the k+1th step observation value y k+1 and the Kalman gain matrix K k+1 Update the prior estimates of the mean and variance:

9. The method for structural performance evaluation guided by digital-physical fusion twin structure and multi-physical parameter update according to claim 1 is characterized in that: The method further includes the step of correcting the finite element model integrated with the non-contact sensing data, wherein the correction step includes the following steps: For structures with existing cracks, the established finite element model first considers the cracks and other disease information, and the unit stiffness is weakened to ensure that the established finite element model takes the crack information into account; Select local parts where the boundary or material properties are difficult to identify, such as columns connected to supports, and invert multiple parameters of the support stiffness of the parts through monitoring data; The initial finite element model is corrected and updated from two levels: detection and monitoring data. First, the stiffness reduction coefficient is calculated based on the crack detection data, and the stiffness reduction coefficient of the cracked area is updated in the finite element model. Then, the bottom column of the structure is selected as a substructure and a substructure finite element model is established. According to the deformation data of the actual structure, the Kalman filtering method of the substructure and the adaptive forgetting factor is used to realize the multi-parameter identification of the support stiffness, and the identified multi-parameters are updated in the finite element model.

Images