A method for identifying the damage state of a group anchor foundation by pulling

Abstract

This invention discloses a method for identifying pull-out damage state of anchor group foundations, relating to the field of transmission line foundation safety monitoring technology. The method includes the following steps: acquiring the real-time relative displacement sequence of the anchor group foundations and inputting it into a stiffness degradation surrogate model to obtain a damage matrix, then updating the initial coupled stiffness matrix; extracting the spectral characteristic change relationship of the coupled stiffness matrix before and after the update to identify the pull-out damage state of the anchor group foundations; wherein, the initial coupled stiffness matrix is ​​established based on the stress-deformation data of the anchor group foundations under undamaged conditions and the static equilibrium relationship; the surrogate model is generated based on a joint loss function including energy constraint terms and equilibrium constraint terms. The energy constraint terms are constructed based on the energy balance relationship between external force work, elastic strain energy, and concrete damage dissipation energy, while the equilibrium constraint terms are constructed based on the residual between the pull-out load and the tensile internal force of the anchor group. This invention addresses the problem of low reliability in damage assessment due to neglecting the differences in stiffness degradation of the anchor group and the lack of structural stress constraints.

Description

Technical Field

[0001] This invention relates to the field of power transmission line foundation monitoring technology, and more specifically, to a method for identifying pull-out damage status of anchored foundations. Background Technology

[0002] Existing methods for assessing damage to transmission tower foundations typically rely on on-site monitoring data such as displacement and strain, combined with empirical indicators or data analysis models, to determine the structural condition. However, due to the characteristic of multiple anchors sharing the load under pull-out loads in multi-anchor foundations, a complex correlation exists between structural response and internal damage state. Therefore, existing methods still have the following shortcomings in practical applications: 1. Existing damage assessment methods are usually based on overall deformation response or equivalent performance indicators, which are difficult to fully reflect the differences in structural stiffness caused by the interaction between anchor bolts within the anchor group foundation. This limits the ability to characterize the damage state and results in low accuracy in damage state judgment.

[0003] 2. Some data-driven damage assessment methods rely mainly on sample data to establish response mapping relationships, without fully incorporating the mechanical constraints during the structural stress process, resulting in discrepancies between model predictions and actual stress states. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for identifying pull-out damage state of anchor group foundations. By using a mechanically constrained surrogate model to update the initial coupling stiffness matrix and extracting the spectral feature change relationship to identify the damage state, this method addresses the problem that the prior art is unable to reflect the differences in stiffness degradation of anchor group foundations and lacks structural stress constraints, resulting in low reliability of damage assessment.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for identifying pull-out damage state of anchor group foundations includes the following steps: The real-time relative displacement sequence of the anchor group foundation is obtained and input into the pre-constructed stiffness degradation surrogate model. After obtaining the damage matrix characterizing the degree of stiffness attenuation, the pre-established initial coupling stiffness matrix is ​​updated. Extract the relationship between the spectral characteristics of the updated coupling stiffness matrix and the initial coupling stiffness matrix, and identify the pull-out damage state of the anchor group foundation. The initial coupling stiffness matrix is ​​established based on the stress and deformation data of the anchor group foundation under no-damage conditions and the static equilibrium relationship; the surrogate model is generated by training a joint loss function including energy constraint terms and equilibrium constraint terms. The energy constraint terms are constructed based on the energy balance relationship between external force work, elastic strain energy and concrete damage dissipation energy, and the equilibrium constraint terms are constructed based on the residual between pull-out load and tensile internal force of the anchor group.

[0006] In a preferred embodiment, the step of establishing the initial coupling stiffness matrix includes: acquiring the stress and deformation data of the pull-out load and corresponding relative displacement of the anchor group foundation under undamaged conditions; establishing a static equilibrium equation based on the stress and deformation data and the spatial distribution of the anchor bolts, and solving for the initial coupling stiffness matrix; wherein, the diagonal elements of the initial coupling stiffness matrix represent the pull-out stiffness of a single anchor bolt, and the off-diagonal elements represent the interaction stiffness between adjacent anchor bolts.

[0007] In a preferred embodiment, obtaining the real-time relative displacement sequence of the anchor group foundation includes: collecting the absolute displacement vector and pier motion parameters of the anchor group foundation during the pull-out process; calculating the displacement component caused by the pier motion based on the spatial coordinate vector of each anchor bolt relative to the centroid of the pier, combined with the pier motion parameters; and subtracting the displacement component from the absolute displacement vector to obtain the real-time relative displacement sequence.

[0008] In a preferred embodiment, updating the pre-established initial coupling stiffness matrix includes: multiplying the damage matrix element-wise with the pre-established initial coupling stiffness matrix to obtain the updated coupling stiffness matrix.

[0009] In a preferred embodiment, the training steps of the stiffness degradation surrogate model include: inputting the historical relative displacement sequence of the entire pull-out process in the training samples into the neural network to be trained, and outputting the corresponding predicted damage matrix; based on the predicted damage matrix, combined with the historical initial coupling stiffness matrix in the training samples and the historical pull-out load of the entire pull-out process, calculating the energy constraint term and the equilibrium constraint term respectively; using the energy constraint term and the equilibrium constraint term to construct a joint loss function to train the neural network, thereby generating the stiffness degradation surrogate model.

[0010] In a preferred embodiment, the energy constraint term is constructed based on the energy balance relationship between external force work, elastic strain energy, and concrete damage dissipation energy; the external force work is obtained by integrating historical pull-out loads and historical relative displacement sequences; the elastic strain energy is obtained by multiplying the predicted damage matrix element-wise with the historical initial coupling stiffness matrix and performing a quadratic matrix operation with the historical relative displacement sequence; the concrete damage dissipation energy is obtained by volume integral of plastic dissipation work and damage energy release rate in the tensile region based on the concrete tensile damage constitutive relation.

[0011] In a preferred embodiment, the calculation step of the equilibrium constraint term includes: multiplying the predicted damage matrix element by element with the historical initial coupling stiffness matrix, and then performing matrix multiplication with the historical relative displacement sequence to obtain the tensile internal force of the anchor group; using the residual between the historical pull-out load and the tensile internal force of the anchor group as the equilibrium constraint term.

[0012] In a preferred embodiment, extracting the spectral feature change relationship between the updated coupling stiffness matrix and the initial coupling stiffness matrix includes: extracting the first minimum eigenvalue and corresponding eigenvector of the updated coupling stiffness matrix, and the second minimum eigenvalue of the initial coupling stiffness matrix; calculating the skewness coefficient of the eigenvector, and the change magnitude of the first minimum eigenvalue relative to the second minimum eigenvalue, to constitute the spectral feature change relationship.

[0013] In a preferred embodiment, identifying the pull-out damage state of the anchor group foundation includes: establishing a two-dimensional feature space based on the variation amplitude and skewness coefficient, and marking the region corresponding to each damage state; mapping the current variation amplitude and skewness coefficient to the two-dimensional feature space, and determining the pull-out damage state of the anchor group foundation according to the region to which its mapping position belongs.

[0014] An electronic device includes: a memory for storing a computer program; and a processor for executing the computer program stored in the memory to implement the steps of any of the methods described.

[0015] The technical effects and advantages of the method for identifying pull-out damage state of group anchor foundations according to the present invention are as follows: 1. This invention updates the initial coupling stiffness matrix using the damage matrix output by the surrogate model, and identifies the pull-out damage state based on the spectral feature changes before and after the update. This effectively reflects the stiffness degradation differences caused by the interaction between anchor bolts during the pull-out process of the group anchor foundation, thereby improving the accuracy of damage state identification of the transmission tower group anchor foundation under complex stress conditions.

[0016] 2. This invention trains a stiffness degradation surrogate model by constructing a joint loss function that includes energy constraints and equilibrium constraints. It introduces the energy conservation relationship and static equilibrium conditions into the model training process, so that the output of the surrogate model satisfies the corresponding mechanical constraints. This reduces the prediction bias of the pure data-driven model in the complex damage evolution process and improves the reliability of the damage state identification results. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of a method for identifying pull-out damage status of a group anchor foundation provided in an embodiment of the present invention.

[0018] Figure 2This is a thermal diagram of the initial coupling stiffness matrix provided in an embodiment of the present invention.

[0019] Figure 3 This is the updated coupling stiffness matrix heatmap provided in the embodiments of the present invention.

[0020] Figure 4 This is a structural block diagram of an exemplary electronic device provided for implementing embodiments of the present disclosure. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1, Figure 1 This invention provides a method for identifying pull-out damage status of anchor group foundations, comprising the following steps: Step 1: Train the stiffness degradation surrogate model 1.1 Construct the training sample set, including: 1.1.1 Establish a three-dimensional simulation model and obtain simulation experimental data, as detailed below: Three-dimensional physical simulation models of various specifications of transmission tower anchor foundations were established using finite element simulation software. When configuring the model's material parameters in the software, the distortion similarity law of equal stress and equal strain was adopted, meaning that the stress similarity constant and strain similarity constant between the simulation model and the real prototype were both set to 1. For the concrete material in the simulation model, a concrete damage-plasticity (CDP) constitutive model was selected and configured from the software's material library. This involved inputting the uniaxial tensile and compressive stress-strain curves of the concrete material, along with corresponding tensile and compressive damage variables that evolve with inelastic strain.

[0023] After establishing the simulation model, a strategy combining graded loading and destructive cyclic loading was adopted to apply continuous pull-out loads (i.e., applying nodal forces upward along the anchor bolt axis) simulating extreme wind load conditions to each simulation model. Specifically, firstly, when the foundation is in a undamaged state (i.e., the elastic stage where the internal stress has not exceeded the tensile yield limit of concrete), pull-out forces are applied in stages at 30%, 50%, and 100% of the design load. Subsequently, a destructive cyclic loading stage exceeding 20% ​​to 50% of the design load is entered. During this stage, the system continuously monitors the macroscopic bearing capacity and principal tensile strain of the model. Loading is stopped when the bearing capacity is found to have decreased to below 85% of the peak load, or when the concrete unit in the tensile zone fails physically due to the principal tensile strain exceeding the limit.

[0024] Simulation data of the entire pull-out process is exported from the finite element method (FEM) software solver. This simulation data includes historical pull-out load vectors, pile cap motion parameters (including translational displacement vectors and overturning angle vectors of the pile cap centroid nodes), historical absolute displacement vectors of the anchorage group, and corresponding historical concrete damage dissipation energy scalars for each analysis step. The historical absolute displacement vectors are obtained by combining the historical absolute displacement data of each anchor bolt according to a spatial distribution sequence. The historical concrete damage dissipation energy scalar is obtained by calling the FEM solver's post-processing module to calculate the volume fraction of plastic dissipation work and damage energy release rate within the tension zone of the anchorage group for each analysis step.

[0025] 1.1.2 Extract the historical relative displacement sequence, as follows: Based on the initial undeformed state of the 3D physical simulation model, the difference between the spatial coordinates of each anchor bolt node and the spatial coordinates of the foundation centroid node is calculated to obtain the spatial coordinate vectors of each anchor bolt relative to the foundation centroid. In conjunction with the pier motion parameters, the displacement components caused by the pier motion are calculated, and the calculation formula for the displacement components is as follows: (1) in, For the first The displacement component of the anchor bolts due to the movement of the pile cap. Let be the translational displacement vector of the centroid of the foundation in the foundation's motion parameters. Let be the overturning rotation angle vector of the centroid of the pier in the pier motion parameters. For the first The spatial coordinate vector of the anchor bolt relative to the centroid of the foundation.

[0026] The calculated displacement components of each anchor bolt caused by the pier cap movement are concatenated according to the anchor bolt node number and topological order that are the same as the historical absolute displacement vector of the anchor group foundation, to obtain the overall displacement component with the same dimension as the absolute displacement vector. The overall displacement component is then subtracted from the historical absolute displacement vector of the anchor group foundation to obtain the historical relative displacement sequence. This historical relative displacement sequence is essentially a time-series data matrix representing relative slippage. The row indices of this matrix correspond to each analysis step in the pull-out process, and the column indices correspond to the local relative displacement of each anchor bolt within the anchor group foundation.

[0027] 1.1.3 Solve for the historical initial coupling stiffness matrix, as follows: The historical pull-out loads in the elastic stage of the simulation experiment data, and the corresponding historical relative displacement sequences calculated in step 1.1.2, are used as the stress-deformation data under the no-damage state. Based on the stress-deformation data, combined with the spatial distribution of the anchor bolts (i.e., the spatial coordinate vectors of each anchor bolt relative to the centroid of the foundation), a set of static equilibrium equations for the group anchor foundation is established; the set of static equilibrium equations includes vertical force equilibrium equations and moment equilibrium equations, as follows: (2) (3) in, To stretch the load for history The overturning moment component caused by the external load. For the first The spatial coordinate vector of the anchor bolt. For the first The local pull-out force borne by the anchor bolts at the base. This represents the total number of anchor bolts.

[0028] It should be noted that when the line of action of the historical pull-out load passes perpendicularly through the centroid of the pier, the overturning moment component is zero, so as to constrain the internal force distribution of each anchor bolt to satisfy the resultant moment being zero; when the external load is applied eccentrically or includes bending moments transmitted from the upper structure of the transmission tower, the overturning moment component takes a non-zero value according to the actual stress conditions.

[0029] Based on the principles of continuum mechanics, the local pull-out force of each anchor bolt and the historical relative displacement sequence of the anchor group satisfy the following constitutive physical relationship: (4) in, For the first The relative displacement scalar of the anchor bolts at the base; Historical initial coupling stiffness matrix The elements in, when The time element is a diagonal element, representing the pull-out stiffness of a single anchor bolt (i.e., the proportional relationship between the tensile force on a single anchor bolt and its relative displacement). The time element is a non-diagonal element, representing the interaction stiffness between adjacent anchor bolts (i.e., the group anchor coupling effect caused by stress field diffusion).

[0030] Substituting the constitutive physical relations into the static equilibrium equations for elimination yields a simultaneous equation system relating the pull-out load, relative displacement, and coupling stiffness matrix. To ensure the uniqueness of the solution to this simultaneous equation system, a priori physical constraints based on the spatial distribution of anchor bolts are introduced for parameterization and dimensionality reduction, thereby reducing the number of independent parameters to be solved in the simultaneous equation system. Specifically, the constraint condition is that the historical initial coupling stiffness matrix is ​​set to be a symmetric matrix (i.e.,...). Furthermore, the interaction stiffness values ​​represented by the off-diagonal elements decrease negatively correlated with the linear spatial distance between the corresponding two anchor bolts.

[0031] Based on the aforementioned physical prior constraints, an overdetermined set of equations is constructed using force and deformation data from multiple sampling points within the elastic stage. The objective function is to minimize the L2 regularized sum of squares of the force residual and the moment residual. The least squares method (such as the ridge regression algorithm) is used to solve the equations. During the solution process, conventional numerical optimization function libraries (such as lsqnonlin based on the Levenberg-Marquardt algorithm) can be directly called to obtain the uniquely determined historical initial coupling stiffness matrix.

[0032] 1.1.4 Generating the training sample set The historical initial coupling stiffness matrix, historical pull-out load vector, historical concrete damage dissipation energy, and historical relative displacement sequence corresponding to each anchor foundation group are aligned according to time steps, and the aligned sequence data is segmented using a fixed-time sliding window to obtain multiple samples. In each sample, the historical relative displacement data is organized into a three-dimensional tensor in the format of [batch, time step, anchor bolt feature dimension], which is used as the input of the subsequent neural network. At the same time, the initial coupling stiffness matrix, historical concrete damage dissipation energy, and historical pull-out load vector corresponding to the sample are associated and stored with the three-dimensional tensor to form a training sample set.

[0033] 1.2 Train a neural network using the training sample set to generate a stiffness degradation surrogate model. 1.2.1 Input and output data and topology of neural networks Input data: historical relative displacement sequence of the anchorage foundation, in three-dimensional tensor format.

[0034] Output data: Predicted damage matrix. The dimensions of this predicted loss matrix are the same as those of the historical initial coupling stiffness matrix. Square formation.

[0035] The topology of the neural network consists of an input layer, a hidden layer, and an output layer.

[0036] The number of nodes in the input layer is consistent with the feature dimension of the historical relative displacement sequence; The hidden layer adopts a 3-layer LSTM network architecture with 128 hidden units per layer, and the hidden state is transmitted between layers in a fully connected manner. The hidden layer relies on the temporal gating mechanism inside the LSTM architecture to extract the historical cumulative effect of damage evolution at the concrete bonding interface of the anchor bolt. The output layer maps the output of the hidden layer to a fully connected layer. An initial square matrix of dimension is obtained. After symmetrization, the elements in the square matrix are mapped to the (0,1) interval by the Sigmoid activation function, and the predicted damage matrix is ​​output. Here, 0 indicates that the foundation is in a completely elastic and undamaged state, and 1 indicates that the foundation is completely destroyed.

[0037] 1.2.2 Construct a joint loss function that includes energy constraint terms and balance constraint terms, as follows: 1.2.2.1 Based on the energy balance relationship between external force work, elastic strain energy, and concrete damage dissipation energy, an energy constraint term is constructed. The formula for the energy constraint term is as follows: (5) in, For energy constraint terms, for External force at all times for Elastic strain energy at time t, for Energy dissipation due to concrete damage at any given time. It represents the square of the Euclidean norm.

[0038] The work done by the external force is obtained by integrating the historical pull-out load with the historical relative displacement sequence, as shown in the following formula: (6) in, Between the initial time and the current time The time integral variable between them for Historical pull-out load vector at any given moment for The historical relative displacement vector at any given time.

[0039] The elastic strain energy is obtained by element-wise multiplication of the predicted damage matrix and the historical initial coupling stiffness matrix, followed by a quadratic matrix operation with the historical relative displacement sequence, as shown in the following formula: (7) in, for The historical relative displacement vector at any given time. It is a matrix of all ones with the same dimension as the predicted damage matrix. To predict the damage matrix, This is for element-wise multiplication.

[0040] The energy dissipated due to concrete damage is theoretically based on the tensile damage constitutive relation of concrete (i.e., the CDP constitutive model configured in the finite element simulation software). It is obtained by volume integral of the plastic dissipation work and the damage energy release rate in the tensile region, as shown in the following formula: (8) in, and These represent the plastic dissipation work density and the damage energy release rate density, respectively, within the tensile region. Let be the volume of the tensioned region.

[0041] It should be noted that, during the training of the surrogate model, in order to avoid spatial integration calculations during each forward propagation, the concrete damage dissipation energy is read as a known physical scalar true value from the pre-constructed training sample set and directly substituted into the joint loss function.

[0042] 1.2.2.2 Based on the residuals of the pull-out load and the tensile internal forces of the anchor group, construct equilibrium constraint terms. The formula for the balance constraint term is as follows: (9) in, To balance the constraint terms, for The anchors at any given moment are subjected to tensile internal forces.

[0043] The tensile internal force of the anchor group is obtained by multiplying the predicted damage matrix element-by-element with the historical initial coupling stiffness matrix and then performing matrix multiplication with the historical relative displacement sequence, as shown in the following formula: (10) 1.2.2.3 The specific formula for the joint loss function is as follows: (11) in, For the joint loss function; The pre-defined penalty weights used to balance the magnitudes, with dimensions m. -2 (That is, the reciprocal of square meters).

[0044] 1.2.3 Train the neural network and solidify it to obtain the stiffness degradation surrogate model. In training the neural network using the training sample set, the entire network weights are first initialized and hyperparameters are configured: the initial learning rate is set to 0.001 and a cosine annealing strategy is used to dynamically decay it with each training round; the batch size is set to 64; and the maximum number of training rounds is set to 500. In each training round, the system extracts a batch of samples and inputs them into the neural network, performs forward propagation to calculate the current predicted damage matrix, and substitutes it into the calculation to obtain the joint loss function value; then, backpropagation is performed according to the chain rule to calculate the gradient of the joint loss function with respect to the network weight matrix and bias vector; finally, the entire network parameters are updated using the Adam optimizer according to the gradient descent optimization rule. The system continuously executes the above training rounds until the joint loss function value converges to below the preset loss threshold (e.g., 0.01) or reaches the maximum number of training rounds. At this point, training ends and the neural network parameters are saved, thus solidifying the stiffness degradation surrogate model.

[0045] This step trains a stiffness degradation surrogate model by constructing a joint loss function that includes energy constraints and equilibrium constraints. This ensures that the predicted damage matrix satisfies the energy relationship between external force work, elastic strain energy, and concrete damage dissipation energy, as well as the equilibrium relationship between pull-out load and tensile internal forces of the anchor group. This improves the consistency between the surrogate model output and the actual stress state of the anchor group foundation, providing reliable damage parameters for subsequent damage state identification.

[0046] Step 2: Obtain the real-time relative displacement sequence of the anchor group foundation and input it into the stiffness degradation surrogate model. After obtaining the damage matrix characterizing the degree of stiffness attenuation, update the pre-established initial coupling stiffness matrix, including: 2.1 Establish the initial coupling stiffness matrix of the anchorage group foundation, as follows: At the designated centroid of the transmission tower foundation, a real-time dynamic positioning (RTK) sensor with static measurement accuracy down to the millimeter level and a high-precision biaxial inclinometer with a resolution better than 0.001° are fixedly installed. Simultaneously, a rigid reference frame, independent of the foundation and unaffected by surrounding soil deformation, is constructed around the tower foundation. The base of the guy wire displacement gauge is fixed to this reference frame, and the end of the measuring line is vertically connected to the top of each anchor bolt. Furthermore, a local three-dimensional coordinate system is established on-site using a high-precision total station, with the centroid of the foundation as the origin. Coordinate mapping is performed on the exposed ends of each anchor bolt to obtain the spatial distribution of the anchor bolts, i.e., the initial spatial coordinate vectors of each anchor bolt relative to the centroid of the foundation.

[0047] For the anchorage foundation monitored on-site, the system performs a static load test during the final acceptance stage after construction and full curing of the concrete to obtain stress and deformation data under undamaged conditions. Specifically: In the static load test, a test load is applied by a servo jack on-site, and relying on the aforementioned fixed monitoring device, a uniform 50Hz sampling frequency is used, and the data is synchronized with a hard-synchronized clock by a centralized data acquisition instrument on-site to synchronously collect the pull-out load, the pile cap motion parameters (translational displacement vector and overturning angle vector), and the absolute displacement vector of the anchorage foundation under undamaged conditions. Based on the initial spatial coordinate vector and the pile cap motion parameters, the displacement component caused by the pile cap motion is calculated according to formula (1) in step 1.1.2, and the relative displacement sequence is obtained by subtracting the displacement component from the absolute displacement vector. Finally, the pull-out load and the relative displacement sequence together constitute the stress and deformation data under undamaged conditions.

[0048] Extract the discrete stress and deformation data of high-frequency continuous sampling during the elastic stage of the static load test, and based on the spatial distribution of the anchor bolts obtained by surveying, reuse the macroscopic static equilibrium equations constructed in step 1.1.3 and the least squares method with physical constraints to identify parameters and solve for the uniquely determined initial coupling stiffness matrix of the group anchor foundation.

[0049] 2.2 Obtain the real-time relative displacement sequence of the anchor group foundation, as follows: After entering the online monitoring stage, relying on the monitoring device pre-fixed in step 2.1, the sampling rate of 50Hz is maintained to synchronously collect the pier motion parameters (i.e., translational displacement vector and overturning angle vector) of the pier centroid under extreme wind load conditions, as well as the absolute displacement vector of each anchor bolt under pull-out force.

[0050] Based on the initial spatial coordinate vectors of the anchor bolts obtained in step 2.1 and the real-time collected pier motion parameters, the displacement components caused by the pier motion are calculated according to formula (1) in step 1.1.2; the displacement components are subtracted from the real-time collected absolute displacement vector to obtain the real-time relative displacement sequence.

[0051] 2.3 Update the initial coupling stiffness matrix as follows: A sliding window of fixed time length (e.g., 50 to 200 steps, corresponding to a time window of 1 to 4 seconds at a 50Hz sampling rate) is used to truncate the real-time relative displacement sequence. The truncated relative displacement sequence is then reconstructed into a three-dimensional tensor according to the dimensional structure of [1, number of time steps in the sliding window, and anchor group feature dimension]. The inference batch size of the first dimension is constantly set to 1 to accommodate the requirements of single online real-time forward inference. The third dimension, the anchor group feature dimension, represents the number of relative displacement scalars of each anchor bolt along the axial direction. The reconstructed three-dimensional tensor is input into the stiffness degradation surrogate model pre-built and trained in step S1, and the corresponding damage matrix is ​​output through network forward propagation.

[0052] The damage matrix is ​​multiplied element-wise with the measured initial coupling stiffness matrix of the anchor group foundation to obtain the updated coupling stiffness matrix; the calculation formula for the updated coupling stiffness matrix is ​​as follows: (12) in, Current monitoring time The updated coupling stiffness matrix, It is a square matrix of all 1s with the same dimension as the initial coupling stiffness matrix. The initial coupling stiffness matrix of the anchorage foundation is obtained from actual measurements. Current monitoring time A damage matrix characterizing the degree of stiffness attenuation, wherein the value range of each element in the damage matrix is ​​mapped by a neural network and converges to the interval (0,1).

[0053] It should be noted that the calculation formula of the updated coupling stiffness matrix achieves nonlinear mechanical reduction by calculating the difference between the all-1 square matrix and the damage matrix and performing a Hadamard product operation with the initial coupling stiffness matrix: For the updated coupling stiffness matrix, its diagonal elements represent the remaining pull-out stiffness of a single anchor bolt in the current damage state, and its off-diagonal elements represent the interaction stiffness between adjacent anchor bolts weakened by the evolution of micro-cracks in the concrete.

[0054] like Figure 2 and Figure 3 As shown, Figure 2 , Figure 3These are heatmaps of the initial coupling stiffness matrix of the anchor group foundation in the elastic, undamaged state, and the updated coupling stiffness matrix at a typical damage moment, respectively. Comparative analysis shows that the initial coupling stiffness matrix is ​​symmetrical, with the main diagonal elements dominating, consistent with the structural mechanics characteristics of the elastic stage. However, in the updated coupling stiffness matrix, the values ​​of the main diagonal elements at specific locations significantly decrease, indicating a degradation in the pull-out stiffness of the anchor bolts at those locations. Simultaneously, the off-diagonal elements show a non-uniform decrease in value, reflecting the reduced interaction stiffness between adjacent anchor bolts due to localized cracking and debonding of the concrete. This comparison demonstrates that this method can reflect the location and extent of damage to the anchor group foundation through changes in the values ​​of elements in the stiffness matrix, providing a basis for subsequent extraction of spectral feature indicators.

[0055] This step extracts the relative displacement sequence reflecting the true damage inside the concrete matrix by using the absolute displacement vector of the anchor group foundation and the motion parameters of the pile cap. This filters out the rigid displacement error introduced by the overall translation and rotation of the pile cap under complex wind loads. Furthermore, the damage matrix output by the stiffness degradation surrogate model is used to perform an element-wise multiplication reduction operation on the initial coupled stiffness matrix, realizing the mapping of the damage state of the anchor group foundation from network prediction parameters to the stiffness matrix. This provides an objective physical benchmark for subsequent pull-out damage state identification based on spectral features.

[0056] Step 3: Extract the spectral characteristic change relationship between the updated coupling stiffness matrix and the initial coupling stiffness matrix, and identify the pull-out damage state of the anchor group foundation, including: 3.1 Extract the spectral characteristic change relationship between the updated coupling stiffness matrix and the initial coupling stiffness matrix, as follows: Eigenvalue decomposition is performed on the updated coupling stiffness matrix and the initial coupling stiffness matrix respectively; the first minimum eigenvalue and its corresponding eigenvector are extracted from the eigenvalue set of the updated coupling stiffness matrix, and the second minimum eigenvalue is extracted from the eigenvalue set of the initial coupling stiffness matrix.

[0057] The magnitude of change of the first minimum eigenvalue relative to the second minimum eigenvalue, and the skewness coefficient of the distribution of each element in the eigenvector are calculated as the spectral feature change relationship; the formulas for calculating the magnitude of change and the skewness coefficient are as follows: (13) (14) in, Current monitoring time The range of change, It is the second smallest eigenvalue of the initial coupling stiffness matrix. This is the first minimum eigenvalue of the updated coupling stiffness matrix; Current monitoring time skewness coefficient, The first in the feature vector Each component value The arithmetic mean of all components in the feature vector is given, and the total number of components in the feature vector is the same as the total number of anchor bolts.

[0058] It should be noted that the variation range is used to characterize the degree of softening of the overall pull-out stiffness of the anchor group foundation, and the skewness coefficient is used to characterize the local force asymmetry within the anchor group foundation.

[0059] 3.2 Identify the pull-out damage state of the anchorage foundation as follows: A two-dimensional feature space is pre-established based on the change amplitude and skewness coefficient, wherein the horizontal axis is set as the change amplitude representing the overall stiffness attenuation, and its effective coordinate range is [0,1); the vertical axis is set as the absolute value of the skewness coefficient representing the local force asymmetry, and its effective coordinate range is defined as [0,+∞).

[0060] The boundaries of the judgment regions corresponding to each preset damage state are marked in the two-dimensional feature space. As an exemplary region division rule, each damage state and its corresponding interval range are defined as follows: The region where the variation range is in the range of [0, 0.05) and the absolute value of the skewness coefficient is in the range of [0, 0.5) is defined as the elastic lossless stage; The regions with variation ranges in the range [0, 0.05) and absolute values ​​of skewness coefficients in the range [0.5, +∞), as well as the regions with variation ranges in the range [0.05, 0.20), are defined as the local asymmetric damage evolution stage. The region with a change range in the range of [0.20, 1) is designated as the overall cone pull-out failure warning stage.

[0061] During online monitoring, the calculated change amplitude and absolute value of the skewness coefficient are used as two-dimensional coordinate points and directly mapped to the two-dimensional feature space. The corresponding judgment result is output according to the calibration area to which the mapped position belongs, thus completing the identification of the pull-out damage state of the anchor group foundation.

[0062] This step extracts the change range of the minimum eigenvalue of the coupling stiffness matrix before and after the update, and calculates the skewness coefficient of the corresponding eigenvector. It then transforms and projects the high-dimensional stiffness matrix into a two-dimensional feature space, thereby achieving the decoupling identification of the overall deterioration mode and the local specific failure mode of the anchor group foundation. This provides an objective basis for judging the damage state and distinguishing the physical failure mode of the transmission tower foundation.

[0063] Example 2, Figure 4An electronic device is provided, comprising: a memory for storing a computer program; and a processor for executing the computer program stored in the memory to implement the steps of any of the methods described.

[0064] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0065] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0066] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0067] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0068] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0069] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying pull-out damage state of a group anchor foundation, characterized in that, Includes the following steps: The real-time relative displacement sequence of the anchor group foundation is obtained and input into the pre-constructed stiffness degradation surrogate model. After obtaining the damage matrix characterizing the degree of stiffness attenuation, the pre-established initial coupling stiffness matrix is ​​updated. Extract the relationship between the spectral characteristics of the updated coupling stiffness matrix and the initial coupling stiffness matrix, and identify the pull-out damage state of the anchor group foundation. The initial coupling stiffness matrix is ​​established based on the stress and deformation data of the anchor group foundation under no-damage conditions and the static equilibrium relationship; the surrogate model is generated by training a joint loss function including energy constraint terms and equilibrium constraint terms. The energy constraint terms are constructed based on the energy balance relationship between external force work, elastic strain energy and concrete damage dissipation energy, and the equilibrium constraint terms are constructed based on the residual between pull-out load and tensile internal force of the anchor group.

2. The method according to claim 1, characterized in that, The steps for establishing the initial coupling stiffness matrix include: Obtain the stress and deformation data of pull-out load and corresponding relative displacement under the undamaged state of the anchor group foundation; Based on the stress-deformation data and the spatial distribution of the anchor bolts, a static equilibrium equation is established, and the initial coupling stiffness matrix is ​​obtained by solving it. The diagonal elements of the initial coupling stiffness matrix represent the pull-out stiffness of a single anchor bolt, while the off-diagonal elements represent the interaction stiffness between adjacent anchor bolts.

3. The method according to claim 1, characterized in that, The acquisition of the real-time relative displacement sequence of the anchor group foundation includes: The absolute displacement vector of the anchor foundation of the transmission tower group and the motion parameters of the pile cap were collected during the pull-out process; Based on the spatial coordinate vectors of the anchor bolts relative to the centroid of the pier cap, and combined with the pier cap motion parameters, the displacement components caused by the pier cap motion are calculated. The real-time relative displacement sequence is obtained by subtracting the displacement component from the absolute displacement vector.

4. The method according to claim 1, characterized in that, The update of the pre-established initial coupling stiffness matrix includes: The damage matrix is ​​multiplied element-wise with the pre-established initial coupling stiffness matrix to obtain the updated coupling stiffness matrix.

5. The method according to claim 1, characterized in that, The training steps of the stiffness degradation surrogate model include: Input the historical relative displacement sequence of the entire pull-out process in the training samples into the neural network to be trained, and output the corresponding predicted damage matrix; Based on the predicted damage matrix, and combined with the historical initial coupling stiffness matrix and the historical pull-out load of the entire pull-out process in the training samples, the energy constraint term and the equilibrium constraint term are calculated respectively. The neural network is trained by constructing a joint loss function using the energy constraint term and the equilibrium constraint term to generate the stiffness degradation surrogate model.

6. The method according to claim 5, characterized in that, The energy constraint term is constructed based on the energy balance relationship between external force work, elastic strain energy and concrete damage dissipation energy. The external force work is obtained by integrating the historical pull-out load and the historical relative displacement sequence; The elastic strain energy is obtained by multiplying the predicted damage matrix element-wise with the historical initial coupling stiffness matrix and performing a quadratic matrix operation with the historical relative displacement sequence. The energy dissipated by concrete damage is obtained by volume fractionating the plastic dissipation work and damage energy release rate in the tensile region based on the tensile damage constitutive relation of concrete.

7. The method according to claim 5, characterized in that, The calculation steps for the balance constraint term include: After multiplying the predicted damage matrix element by element with the historical initial coupling stiffness matrix, matrix multiplication is performed with the historical relative displacement sequence to obtain the tensile internal forces of the anchor group. The residual between the historical pull-out load and the tensile internal force of the anchor group is used as the equilibrium constraint term.

8. The method according to claim 1, characterized in that, The extraction of the spectral feature change relationship between the updated coupling stiffness matrix and the initial coupling stiffness matrix includes: Extract the first minimum eigenvalue and corresponding eigenvector of the updated coupling stiffness matrix, and the second minimum eigenvalue of the initial coupling stiffness matrix; The skewness coefficient of the eigenvector and the magnitude of change of the first minimum eigenvalue relative to the second minimum eigenvalue are calculated to form the spectral feature change relationship.

9. The method according to claim 8, characterized in that, The identification of the pull-out damage state of the anchor group foundation includes: A two-dimensional feature space is established based on the aforementioned variation amplitude and skewness coefficient, and the corresponding regions for each damage state are calibrated. The current variation amplitude and skewness coefficient are mapped to the two-dimensional feature space, and the pull-out damage state of the anchor group foundation is determined according to the region to which the mapped position belongs.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing a computer program stored in the memory to implement the steps of the method as described in any one of claims 1-9.

Images