An arch dam body damage identification method

Abstract

The application provides an arch dam body damage identification method, comprising the following steps: constructing a normal operation condition model and a potential damage condition model; obtaining a three-dimensional static displacement field response of each condition model under different water level conditions; calculating a monitoring node displacement by using an interpolation function, and calculating an incremental displacement eigenvalue; constructing a displacement feature matrix and a state feature matrix under different conditions; training an arch dam body state prediction model by using the feature matrix, and realizing a two-class prediction of a normal / damaged state of the arch dam body based on engineering monitoring data. The arch dam body damage identification method provided by the application can effectively solve the limitations of the current arch dam body damage identification method by using a displacement feature to capture the damage state of the arch dam body.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of water conservancy and hydropower engineering safety, and particularly relates to an arch dam body damage identification method. BACKGROUND

[0002] At present, global climate change is becoming more and more significant, and the frequency of extreme natural disaster events continues to rise. As one of the most common dam types in water conservancy and hydropower engineering, arch dams have the risk of structural damage in such harsh environments. The traditional safety monitoring method mainly relies on absolute displacement threshold judgment or statistical model. After natural disasters occur, the weak displacement changes caused by early damage of the arch dam body structure are usually easily covered by measurement noise and environmental fluctuations, and the threshold judgment has obvious insufficient sensitivity; at the same time, the statistical model trained based on historical data is difficult to capture the nonlinear response mutation caused by sudden damage. Therefore, it is urgent to establish an arch dam body damage judgment and identification method that can effectively suppress environmental noise interference and accurately capture structural weak abnormal response and displacement characteristics. SUMMARY

[0003] In view of the defects in the prior art, the application provides an arch dam body damage identification method for accurately identifying whether the arch dam is damaged after encountering natural disasters.

[0004] The technical scheme adopted by the application is as follows:

[0005] The application provides an arch dam body damage identification method, comprising the following steps:

[0006] Step S1: according to the actual size of the arch dam body engineering, a three-dimensional finite element grid is used to establish a refined geometric model of the arch dam body; based on the refined geometric model of the arch dam body, M working condition models are constructed; wherein the M working condition models include one normal operation working condition model and M-1 potential damage working condition models;

[0007] Step S2: for each working condition model m, m = 1, 2, …, M, the static water load under different water levels is sequentially applied according to the application time sequence, and through finite element numerical simulation, the three-dimensional displacement field response of each finite element grid of the arch dam body under the action of the static water load corresponding to each application time is obtained;

[0008] Step S3: in the working condition model m, J monitoring nodes are selected, based on the three-dimensional displacement field response of the finite element grid, an interpolation function is used to calculate the three-dimensional displacement field response of each monitoring node j under the action of the static water load corresponding to each application time, j = 1, …, J;

[0009] Step S4: the displacement feature matrix x m and the state feature matrix X m of the working condition model m are constructed:

[0010] Step S4.1, for each monitoring node j of the working condition model m, the three-dimensional displacement field response obtained at each two adjacent application times is analyzed to obtain an incremental displacement eigenvalue; based on the incremental displacement eigenvalue, combined with the time attribute and the water level attribute, an incremental displacement eigenvector is obtained;

[0011] Step S4.2, the incremental displacement eigenvectors of all monitoring nodes of the working condition model m are spliced to obtain a displacement feature matrix X of the working condition model m m ;

[0012] Step S4.3, a state feature matrix Y of the working condition model m is constructed m ; wherein, if the working condition model m is a normal operation working condition model, each element in the state feature matrix Y m is 1; if the working condition model m is a potential damage working condition model, each element in the state feature matrix Y m is 0;

[0013] Step S5, the displacement feature matrix X m and the state feature matrix Y m of each working condition model m form a training sample m; each training sample is used to train a pre-constructed arch dam body state prediction model to obtain a trained arch dam body state prediction model; wherein, the displacement feature matrix X m is the model input, and the state feature matrix Y m is the model training label;

[0014] Step S6, an actual displacement feature matrix X * composed of J monitoring nodes of the arch dam body project is obtained and input into the trained arch dam body state prediction model, and the trained arch dam body state prediction model outputs the result of whether the arch dam body project is currently damaged.

[0015] Preferably, the normal operation working condition model and each of the potential damage working condition models are constructed by using entity elements; the normal operation working condition model has the same structure geometric parameters and material characteristics as the actual arch dam body project; the potential damage working condition model is obtained by setting different damage conditions on the basis of the normal operation working condition model.

[0016] Preferably, the potential damage working condition model is constructed by the following method:

[0017] Performing structural analysis on the normal operation working condition model to determine the vulnerable area of the arch dam body project, including: determining the vulnerable area based on seismic dynamic response analysis, or determining the stress concentration area based on static force analysis as the vulnerable area;

[0018] One or more types of damage are set in the vulnerable area, and the potential damage condition model is constructed;

[0019] The damage is set in the following manner:

[0020] A contact joint element is introduced in the vulnerable area to simulate the cracks and dislocations of the arch dam body engineering; the cracks include surface cracks and internal cracks; the dislocations include shear slip along the structural plane or fissure plane;

[0021] Alternatively, the material parameters of the vulnerable area are modified to simulate material degradation, freeze-thaw damage or chemical erosion; the material parameters include elastic modulus and tensile strength;

[0022] Alternatively, the vulnerable area is geometrically adjusted to simulate local subsidence or deformation.

[0023] Preferably, each of the potential damage condition models is constructed by the following method:

[0024] When the contact joint element is introduced in the vulnerable area, the spatial position, geometric size, orientation and contact characteristics of the contact joint element are adjusted to construct different potential damage condition models; the spatial position includes elevation, dam section and bank slope position; the geometric size includes crack length, depth, width and penetration degree; the orientation includes river direction, transverse direction and inclined crack; the contact characteristics include normal and tangential stiffness, friction coefficient and opening and closing threshold, which respectively simulate crack opening, closing and shear slip behavior;

[0025] When material degradation is performed in the vulnerable area, the degradation range and degradation degree are adjusted to construct different potential damage condition models; the degradation range includes the number and distribution of affected elements; the degradation degree includes the reduction coefficient of material parameters.

[0026] Preferably, in step S2, the static water load under different water levels is sequentially applied in time sequence, which is characterized by time sequence T and water level sequence H, and is defined as:

[0027] T={t1,...,t I}

[0028] H={h1,...,h I}

[0029] Wherein: {t1,...,t I} is a discrete time sequence with equal time intervals, and {h1,...,h I} is the water level sequence at the corresponding time; time t i and water level value h iThis forms a one-to-one mapping relationship, where i = 1, ..., I.

[0030] Preferably, step S3 specifically includes:

[0031] Step S3.1, for monitoring node j, if it is located at e in the working condition model m j On the common boundary or vertex of any finite element mesh, the monitoring node j has corresponding three-dimensional coordinates (ξ,η,ζ) in the local coordinate system of any finite element mesh k, k=1,…,e j Using the shape function of the finite element mesh k and all of its N k Displacement of a known finite element mesh The displacement interpolation result of monitoring node j Represented as:

[0032]

[0033] In the formula, the shape function of the finite element mesh k This is determined by the properties of its finite element mesh;

[0034] Step S3.2, due to monitoring nodes j and e j Given two adjacent finite element meshes, to ensure the continuity and uniqueness of the displacement at the mesh boundaries, the final displacement value U at monitoring node j is determined. j The average of the interpolation results from all neighboring finite element meshes is expressed as:

[0035]

[0036] Monitoring node j displacement value U j This includes displacement components in three directions within the dam's global coordinate system (x, y, z), namely:

[0037]

[0038] Where: s j d j and v j , respectively, are the displacement values ​​U of monitoring node j. j Displacement components in the x, y, and z directions.

[0039] Preferably, step S3 specifically includes: step S4.1 specifically includes:

[0040] Step S4.1.1: For each monitoring node j of the working condition model m, there exists a corresponding radial displacement monitoring sequence S within the time series T. j , defined as: S j ={s j,1 ,...,s j,Iwherein the radial direction is the x direction;

[0041] Step S4.1.2, the incremental displacement feature value of each two adjacent application time is defined as:

[0042]

[0043] wherein: s j,i is the radial displacement of monitoring node j at time t i ; s j,i-1 is the radial displacement of monitoring node j at time t i-1 ; t i-1 and t i are two adjacent time points; λ j,i is the incremental displacement feature value of monitoring node j at time t i relative to time t i-1 ;

[0044] Step S4.1.3, the incremental displacement feature vector of monitoring node j of working condition model m at time t i is expressed as:

[0045]

[0046] wherein: Δt i = t i - t i-1 , is the time interval i between adjacent time points t i-1 and t i ; Δh i is the water level difference in the time interval i, i.e. Δh i = h i - h i-1 ; h i and h i-1 are the water levels at time t i and time t i-1 respectively; c jx , c jy , c jz are the coordinates of monitoring node j in x, y, z directions respectively.

[0047] Preferably, the displacement feature matrix X m of working condition model m is expressed as:

[0048]

[0049] wherein: J is the total number of monitoring nodes of working condition model m, and I is the total number of time points of the time sequence;

[0050] The state feature matrix Y m of working condition model m is expressed as: ​

[0051]

[0052] wherein: the monitoring node j representing the working condition model m at time t i , l m is a working condition state constant, whose value depends on the nature of the working condition model m generating the state feature matrix Y m , if the working condition model m is a normal working condition model, then l m = 0; if the working condition model m is any kind of potential damage working condition model, then l m = 1.

[0053] Preferably, in step S5, when training the arch dam body state prediction model, the K-means spatial clustering algorithm is used to optimize sampling of the internal nodes of the dam body to determine a monitoring node set with optimal spatial distribution uniformity; in the model training iteration process, a step-by-step downsampling strategy is adopted, that is, the number of monitoring node samples is gradually reduced until it approaches the actual engineering monitoring node number, realizing smooth transition from a theoretical model to actual application.

[0054] Preferably, the arch dam body state prediction model is a mapping model established from a displacement feature matrix to a state feature matrix, which is used to realize binary classification prediction of the normal / damage state of the arch dam body.

[0055] The arch dam body damage identification method provided by the application has the following advantages:

[0056] The application provides an arch dam body damage identification method, which can accurately identify whether the arch dam is damaged after encountering natural disasters, and can effectively solve the limitations of current arch dam body damage identification methods. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flowchart of the arch dam body damage identification method provided by the application is shown in the figure;

[0058] Figure 2 The finite element model of the embodiment of the application is shown in the figure;

[0059] Figure 3 The damage area diagram of the embodiment of the application is shown in the figure;

[0060] Figure 4 The figure shows the incremental displacement feature value of a monitoring node under different working conditions along the water level;

[0061] Figure 5 The figure shows the spatial sampling of the monitoring node of the embodiment of the application;

[0062] Figure 6 A prediction model diagram for an embodiment of the present application;

[0063] Figure 7 A prediction result confusion matrix diagram for an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to make the technical problems solved by the present application, technical solutions and beneficial effects clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.

[0065] Referring to Figure 1 The present application provides an arch dam body damage identification method, comprising the following steps:

[0066] According to the actual size of the arch dam body engineering, a three-dimensional finite element grid is used to establish a refined geometric model of the arch dam body, and a working condition model including a normal operation condition model and a plurality of potential damage condition models is constructed.

[0067] For each working condition model, a static water load under different water level conditions is applied, and a complete three-dimensional displacement field response of the dam body under the static water load is obtained through finite element numerical calculation.

[0068] Based on the three-dimensional displacement field calculation results of each working condition model, an interpolation function is used to calculate the displacement data of the monitoring nodes, calculate the incremental displacement eigenvalues, and construct the displacement eigenvalue matrix and the state eigenvalue matrix under different working conditions.

[0069] The displacement eigenvalue matrix and the state eigenvalue matrix are used to train the arch dam body state prediction model, and the displacement eigenvalue matrix is reconstructed based on the engineering monitoring data as the model input, so as to realize the binary classification prediction of the normal / damage state of the arch dam body, and accurately identify and evaluate the engineering operation state.

[0070] The present application proposes a method for capturing the damage state of the arch dam body by using displacement characteristics, which can effectively solve the limitations of the current arch dam body damage identification method.

[0071] The present application will be described in detail as follows:

[0072] Step S1, according to the actual size of the arch dam body engineering, a three-dimensional finite element grid is used to establish a refined geometric model of the arch dam body; based on the refined geometric model of the arch dam body, M working condition models are constructed; wherein the M working condition models include one normal operation condition model and M-1 potential damage condition models;

[0073] The normal operation condition model and each of the potential damage condition models are constructed by solid elements; the normal operation condition model has the same structure geometric parameters and material properties as the actual arch dam body project, and strictly follows the structure geometric parameters and material properties of engineering design; the potential damage condition model is obtained by setting different damage conditions on the basis of the normal operation condition model.

[0074] The potential damage condition model is constructed by the following method:

[0075] The normal operation condition model is subjected to structural analysis to determine the vulnerable area of the arch dam body project, including: determining the vulnerable area based on seismic dynamic response analysis, or determining the stress concentration area based on static force analysis as the vulnerable area; or determining the vulnerable area of the arch dam project according to arch dam engineering experience, such as the contact surface between the dam body and the bedrock, the dam abutment, the high-stress area of the dam body, or the known structural weak points, etc.

[0076] Subsequently, one or more types of damage are set in the identified vulnerable area to construct the potential damage condition model;

[0077] The setting method of the damage is:

[0078] Contact joint elements are introduced in the vulnerable area to simulate the cracks and dislocations of the arch dam body project; the cracks include surface cracks and internal cracks; the dislocations include shear slip along the structural surface or fissure surface;

[0079] Alternatively, the material parameters of the vulnerable area are modified to simulate the damage caused by material degradation, freeze-thaw damage, or chemical erosion, etc.; wherein the material parameters include elastic modulus and tensile strength;

[0080] Alternatively, the vulnerable area is subjected to geometric adjustment to simulate local subsidence or deformation.

[0081] Further, each of the potential damage condition models is constructed by the following method:

[0082] When the contact joint elements are introduced in the vulnerable area, the spatial position, geometric size, orientation and contact characteristics of the contact joint elements are adjusted to construct different potential damage condition models; wherein the spatial position includes elevation, dam section and bank slope position; the geometric size includes crack length, depth, width and penetration degree; the orientation includes river direction, transverse direction and inclined crack; the contact characteristics include normal and tangential stiffness, friction coefficient and opening and closing threshold, respectively simulating crack opening, closing and shear slip behavior;

[0083] When degrading materials in the vulnerable area, the degradation range and degree are adjusted to construct different potential damage condition models; wherein, the degradation range includes the number and distribution of influencing elements; and the degradation degree includes the reduction factor of material parameters.

[0084] Step S2: For each working condition model m, m = 1, 2, ..., M, apply the static water load under different water level conditions in sequence according to the application time sequence. Through finite element numerical simulation, obtain the three-dimensional displacement field response of each finite element mesh of the arch dam body under the static water load at each application time.

[0085] The static water loads under different water level conditions are applied sequentially according to the application time sequence, and are characterized by the time series T and the water level sequence H, which are defined as follows:

[0086] T = {t1,...,t} I}

[0087] H = {h1,...,h} I}

[0088] Where: {t1,...,t I Let {h1,...,h} be a discrete time series with equal time intervals. I} represents the water level sequence at the corresponding time; time t i With water level value h i This forms a one-to-one mapping relationship, where i = 1, ..., I.

[0089] Therefore, for each type of working condition model (i.e., normal operation working condition model and each potential damage working condition model), static water loads generated under different water level conditions (defined by the water level sequence H) are applied to it; the finite element numerical simulation method is used to solve the structural response of each working condition model under the action of each level of static water load, and the displacement results of each node in the finite element mesh in the x, y, and z directions are obtained.

[0090] Step S3: Select J monitoring nodes in the working condition model m. Based on the three-dimensional displacement field response of the finite element mesh, use the interpolation function to calculate the three-dimensional displacement field response of each monitoring node j under the action of hydrostatic load at each applied time, j=1,…,J;

[0091] Monitoring nodes are essentially displacement measuring points actually deployed in the project, including conventional monitoring facilities such as tension lines, upright lines, and inverted lines. The spatial coordinates of monitoring nodes in the dam finite element model are determined based on the dam's global coordinate system. The coordinates of any monitoring node j (j=1,…,J) can be represented as c j =(c j,x ,c j,y ,c j,z ).

[0092] Since the mesh nodes of the model often cannot perfectly coincide with the positions of these monitoring nodes during finite element numerical simulation, it is necessary to use an interpolation function to convert the displacement results of each node in the x, y, and z directions in the finite element mesh into the displacement results of the monitoring nodes. The specific method is as follows:

[0093] Step S3.1, for monitoring node j, if it is located at e in the working condition model m j On the common boundary or vertex of any finite element mesh, the monitoring node j has corresponding three-dimensional coordinates (ξ,η,ζ) in the local coordinate system of any finite element mesh k, k=1,…,e j Using the shape function of the finite element mesh k and all of its N k Displacement of a known finite element mesh The displacement interpolation result of monitoring node j Represented as:

[0094]

[0095] In the formula, the shape function of the finite element mesh k This is determined by the properties of its finite element mesh;

[0096] Step S3.2, due to monitoring nodes j and e j Given two adjacent finite element meshes, to ensure the continuity and uniqueness of the displacement at the mesh boundaries, the final displacement value U at monitoring node j is determined. j The average of the interpolation results from all neighboring finite element meshes is expressed as:

[0097]

[0098] Monitoring node j displacement value U j This includes displacement components in three directions within the dam's global coordinate system (x, y, z), namely:

[0099]

[0100] Where: s j d j and v j , respectively, are the displacement values ​​U of monitoring node j. j Displacement components in the x, y, and z directions.

[0101] Step S4: Construct the displacement feature matrix X of the working condition model m. m and state feature matrix Y m :

[0102] Step S4.1, analyzing the three-dimensional displacement field response obtained by each monitoring node j of the working condition model m at each two adjacent application times, to obtain an incremental displacement eigenvalue; based on the incremental displacement eigenvalue, combining the time attribute and the water level attribute, to obtain an incremental displacement eigenvector;

[0103] Step S4.1 is specifically:

[0104] Step S4.1.1, for each monitoring node j of the working condition model m, in the time sequence T, there is a corresponding radial displacement monitoring sequence S j , defined as: S j ={s j,1 ,...,s j,I}; wherein, the radial direction is the x direction;

[0105] In the formula, {s j,1 ,...,s j,I} is the radial displacement of the monitoring node at equal time intervals in the time sequence T; the time t i , the water level value h i and the displacement s j,i (i=1,…,I) are one-to-one corresponding;

[0106] Step S4.1.2, the incremental displacement eigenvalue of each two adjacent application times is defined as:

[0107]

[0108] Wherein: s j,i is the radial displacement of the monitoring node j at time t i ; s j,i-1 is the radial displacement of the monitoring node j at time t i-1 ; t i-1 and t i are two adjacent time points; λ j,i is the incremental displacement eigenvalue of the monitoring node j at time t i relative to time t i-1 ;

[0109] Step S4.1.3, the incremental displacement eigenvector of the monitoring node j of the working condition model m at time t i is represented as:

[0110]

[0111] Wherein: Δt i =t i -t i-1 , is the time interval i between adjacent time points t i-1 and t i ; Δh iThe water level difference in the time interval i, i.e. Δh i = h i - h i-1 ; h i and h i-1 are the water levels at time t i and time t i-1 , respectively; c jx , c jy , c jz are the coordinates of the monitoring node j in the x, y, z directions, respectively.

[0112] Step S4.2, concatenating the incremental displacement feature vectors of all monitoring nodes of the working condition model m to obtain the displacement feature matrix X m of the working condition model m.

[0113] The displacement feature matrix X m of the working condition model m is expressed as:

[0114]

[0115] wherein J is the total number of monitoring nodes of the working condition model m, and I is the total number of time series;

[0116] Step S4.3, constructing the state feature matrix Y m of the working condition model m; wherein if the working condition model m is a normal operation working condition model, each element in the state feature matrix Y m is 1; if the working condition model m is a potential damage working condition model, each element in the state feature matrix Y m is 0.

[0117] The state feature matrix Y m of the working condition model m is expressed as:

[0118]

[0119] wherein: represents the state of the monitoring node j of the working condition model m at time t i , l m is a working condition state constant, and its value depends on the nature of the working condition model m generating the state feature matrix Y m ; if the working condition model m is a normal operation working condition model, l m = 0; if the working condition model m is any potential damage working condition model, l m = 1.

[0120] Step S5, the displacement feature matrix X m and the state feature matrix Y mform a training sample m; each training sample is used to train the pre-constructed arch dam body state prediction model to obtain a trained arch dam body state prediction model; wherein the displacement feature matrix X m is the model input, the state feature matrix Y m is the model training label.

[0121] The training stage is realized based on the finite element model, and the input data is composed of the displacement feature matrix X m and the corresponding state feature matrix Y m ; the standard data set division strategy is used to randomly allocate all data into a training set and a test set, the model parameters are optimized through the training set, so that the mapping relationship from X m to Y m (normal / damage) can be established, the test set is used to verify the generalization ability of the model, and it is ensured that the model can accurately identify the actual engineering working condition; the prediction and actual deployment stage is realized based on the actual monitoring data of the project.

[0122] In the present application, when training the arch dam body state prediction model, for the normal operation condition model and each potential damage condition model, the K-means spatial clustering algorithm is used to optimize the sampling of the internal nodes of the dam body, and the monitoring node set with the optimal spatial distribution uniformity is determined; in the model training iteration process, a step-by-step downsampling strategy is adopted, that is, the number of monitoring node samples is gradually reduced until it approaches the actual engineering monitoring node number, and a smooth transition from a theoretical model to an actual application is realized. Based on the finally determined preferred node set, the normalized displacement feature matrix and the corresponding state feature matrix are extracted and constructed; the random forest classification algorithm is used to train and establish the mapping model from the displacement feature matrix to the state feature matrix, and the binary classification prediction of the normal / damage state of the arch dam body is realized.

[0123] In the present application, the arch dam body state prediction model is used to establish the mapping model from the displacement feature matrix to the state feature matrix, and is used to realize the binary classification prediction of the normal / damage state of the arch dam body.

[0124] Step S6, an actual displacement feature matrix X * composed of J monitoring nodes of the arch dam body project is obtained and input into the trained arch dam body state prediction model, and the trained arch dam body state prediction model outputs the result of whether the arch dam body project is damaged at present.

[0125] As a preferred mode, the prediction and actual deployment stage needs to be realized based on the following process: the monitoring records meeting the quality requirements are screened from the engineering historical monitoring database; the appropriate sampling time sequence T and the adjacent time interval Δt iExtract the water level sequence H and radial displacement monitoring sequence S that strictly correspond to this time series. j Based on the above sequence data, displacement feature vectors and state feature vectors of each monitoring node are constructed according to the aforementioned process, and used as input features of the model to determine whether damage exists.

[0126] The following is an example:

[0127] Reference Figure 2 , Figure 3 In this embodiment of the invention, a hyperbolic arch dam is used as the research prototype, and a three-dimensional finite element model is established; all working conditions are based on hexahedral solid elements to construct the dam geometry and achieve accurate simulation. Figure 2 The finite element model shown strictly adheres to the original geometric parameters and material property parameters of the engineering design, fully reflecting the dam's design state, and serves as the model for normal operating conditions; for example... Figure 3 As shown, crack damage is introduced at the left bank abutment of the finite element model. This damage is accurately simulated using contact joint elements to form a potential damage condition model. The location, geometry, and related parameters of the contact joint elements are determined based on the seismic response analysis results of the normal operating condition model.

[0128] Reference Figure 4 This is a graph showing the incremental displacement characteristic value of a monitoring node along the water level under the normal operation condition model and the potential damage condition model of this invention. For the displacement response of the two models under the same time series conditions but different water level conditions, the radial displacement value of the monitoring node is calculated using the method of this invention, and the incremental displacement characteristic value of the monitoring node under the two conditions is calculated using an interpolation function. This amplifies the difference in displacement characteristics under different conditions, and then a complete displacement characteristic vector and state characteristic vector of the monitoring node are constructed according to the method of this invention.

[0129] Reference Figure 5 This is a schematic diagram of spatial sampling of monitoring nodes in an embodiment of the present invention. For the normal operation model and various potential damage models, the K-means spatial clustering algorithm is used to sample the nodes inside the dam finite element model (…). Figure 5 (Left) Optimize sampling to determine the monitoring node set with the best spatial distribution uniformity. Figure 5 (Right); The displacement feature vector and state feature vector of the above-mentioned optimal monitoring node set are both calculated based on the interpolation function. The displacement feature matrix and state feature matrix of each model are constructed according to the method of the present invention.

[0130] Reference Figure 6FIG. 1 is a schematic diagram of a prediction model according to an embodiment of the present application. The prediction model according to the present embodiment adopts a random forest model structure, and uses a displacement feature matrix as an input feature and a state feature matrix as a target label for training. In the training stage, the model constructs multiple decision trees, each of which is trained using a different subset of training data and a feature subspace. Each decision tree learns the mapping relationship from displacement features to state features (normal / damaged) by recursively splitting nodes until a stop condition is reached.

[0131] Referring to Figure 7 FIG. 2 is a confusion matrix of a prediction result according to an embodiment of the present application. For the present embodiment, the prediction result of the dam body damage classification is excellent, with an overall accuracy of 96.19% (202 / 210) among all samples input for model training; the model is relatively accurate in identifying normal and damaged states, with a low misjudgment rate, only 6 normal samples being misjudged as damaged and 2 damaged samples being misjudged as normal. When the accuracy of the model meets the requirements, the model can be deployed: selecting monitoring records meeting the quality requirements from the engineering historical monitoring database; manually determining appropriate sampling time sequence T and adjacent time interval Δt i ; extracting the water level sequence H and the radial displacement monitoring sequence S j strictly corresponding to the time sequence, based on the above sequence data, constructing the displacement feature vector and the state feature vector of each monitoring node according to the method of the present application as the input feature of the model, and performing real-time damage judgment.

[0132] The above only describes the preferred embodiments of the present application, and it should be noted that those of ordinary skill in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered within the scope of the present application.

Claims

1. An arch dam body damage identification method, characterized in that, The method comprises the following steps: Step S1, according to the actual size of the arch dam body engineering, a three-dimensional finite element grid is used to establish a refined geometric model of the arch dam body; based on the refined geometric model of the arch dam body, M working condition models are constructed; wherein the M working condition models include one normal operation working condition model and M-1 potential damage working condition models; Step S2, for each working condition model m, m=1, 2, …, M, the static water load under different water level conditions is sequentially applied according to the time sequence, and the three-dimensional displacement field response of each finite element grid of the arch dam body under the action of the static water load corresponding to each application time is obtained through finite element numerical simulation; Step S3, selecting a monitoring node in the working condition model m, based on the three-dimensional displacement field response of the finite element grid, using an interpolation function, calculating the three-dimensional displacement field response of each monitoring node under the static water load corresponding to each application time corresponding to each application time, ;​ Step S4, constructing the displacement feature matrix of the working condition model m and the state feature matrix ; Step S4.1, for each monitoring node of the working condition model m The three-dimensional displacement field responses obtained at every two adjacent application times are analyzed to obtain an incremental displacement eigenvalue; based on the incremental displacement eigenvalue, in combination with time attributes and water level attributes, an incremental displacement eigenvector is obtained; Step S4.2, concatenating the incremental displacement eigenvectors of all the monitoring nodes of the working condition model m to obtain the displacement eigenvector matrix of the working condition model m ; Step S4.3, constructing a state feature matrix of the working condition model m ; wherein, if the working condition model m is a normal operation working condition model, each element in the state feature matrix is 1; if the working condition model m is a potential damage working condition model, each element in the state feature matrix is 0. Step S5, displacement feature matrix of each working condition model m and state feature matrix form a training sample m; each training sample is used to train the pre-constructed arch dam body state prediction model to obtain a trained arch dam body state prediction model; wherein the displacement feature matrix is the model input, and the state feature matrix is the model training label; Step S6, the actual displacement feature matrix composed of the monitoring nodes of the arch dam body engineering is input into the trained arch dam body state prediction model, and the trained arch dam body state prediction model outputs the result of whether the arch dam body engineering is damaged currently. , input to the trained arch dam body state prediction model, and the trained arch dam body state prediction model outputs the result of whether the arch dam body engineering is damaged currently.​ Step S3 is specifically: Step S3.1, for monitoring nodes If it is located in the working condition model m Monitoring nodes are located on the common boundaries or vertices of the finite element meshes. In any finite element mesh In the local coordinate system, there are corresponding three-dimensional coordinates ( ), Using finite element mesh shape function and all of them Displacement of a known finite element mesh , will monitor nodes displacement interpolation results Represented as: ； where the shape functions of the finite element mesh are determined by its finite element mesh properties; Step S3.2, the displacement value of the monitoring node adjacent to the known finite element mesh, in order to ensure the continuity and uniqueness of the displacement on the boundary of the finite element mesh, the final monitoring node displacement value is derived from the average of the interpolation results of all adjacent finite element meshes, expressed as: ​​ ； Monitoring nodes Displacement value Including the dam's global coordinate system ( The displacement components in the three directions below, namely: ； wherein: , and are displacement values in three directions. components of displacement.​ 2. The method according to claim 1, wherein, The normal operation working condition model and each potential damage working condition model are constructed by using solid elements; the structure geometric parameters and material characteristics of the normal operation working condition model are the same as those of the actual arch dam body engineering; the potential damage working condition model is obtained by setting different damage conditions on the basis of the normal operation working condition model.

3. The method according to claim 2, wherein, The potential damage working condition model is constructed by the following method: Performing structural analysis on the normal operation working condition model to determine the vulnerable area of the arch dam body engineering, including: determining the vulnerable area based on seismic dynamic response analysis, or determining the stress concentration area based on static analysis as the vulnerable area; One or more types of damage are set in the vulnerable area to construct the potential damage working condition model; Wherein, the setting mode of the damage is: Introducing a contact joint element in the vulnerable area to simulate cracks and dislocations of the arch dam body engineering; the cracks include surface cracks and internal cracks; the dislocations include shear slip along the structural surface or crack surface; Alternatively, modify the material parameters of the vulnerable area to simulate material degradation, freeze-thaw damage or chemical erosion; wherein the material parameters include elastic modulus and tensile strength; Alternatively, adjust the geometry of the vulnerable area to simulate local subsidence or deformation.

4. The method according to claim 3, wherein, Each potential damage working condition model is constructed by the following method: When the contact joint element is introduced in the vulnerable area, the spatial position, geometric size, trend and contact characteristics of the contact joint element are adjusted to construct different potential damage working condition models; wherein the spatial position includes elevation, dam section and bank slope position; the geometric size includes crack length, depth, width and penetration degree; the trend includes river direction, transverse direction and inclined crack; the contact characteristics include normal and tangential stiffness, friction coefficient and opening degree threshold, which respectively simulate crack opening, closing and shear slip behavior; When material degradation is performed in the vulnerable area, the degradation range and degradation degree are adjusted to construct different potential damage working condition models; wherein the degradation range includes the number and distribution of affected elements; the degradation degree includes the reduction coefficient of material parameters.

5. The method of claim 1, wherein, In step S2, the hydrostatic load under different water level conditions is applied in sequence according to the time sequence of application, and the time sequence and the water level sequence are used to characterize, which is defined as: ； ； wherein: is a discrete-time sequence of equal time intervals, is a water level sequence at corresponding time instants; time instant and water level value constitute a one-to-one mapping relationship, .

6. The method of claim 1, wherein, Step S3 is specifically: Step S4.1 is specifically: Step S4.1.1, for each monitoring node of the working condition model m , there is a corresponding radial displacement monitoring sequence , defined as: ; where the radial direction is the direction ;​ Step S4.1.2, the incremental displacement eigenvalue of each two adjacent application times is defined as: ； wherein: is the monitoring node at time a radial displacement; is the monitoring node at time a radial displacement; and are two adjacent times; is the monitoring node at time an incremental displacement eigenvalue relative to time ; Step S4.1.3, monitoring node of working condition model m At time Incremental displacement feature vector of is expressed as: ; wherein: is the time interval between adjacent time instants and ; ; is the water level difference within the time interval , i.e. ; and are the water levels at time instants and , respectively; , are the coordinates of the monitoring node in the direction , respectively.

7. The method according to claim 6, wherein, Displacement characteristic matrix of operating condition model m is expressed as: ； wherein: is the total number of monitoring nodes for the working condition model is the total number of time series; and is the total number of time instants of the time series. State feature matrix of operating condition model m is expressed as: ； wherein: monitoring node representing a working condition model m at time , , is a working condition state constant, whose value depends on the nature of the working condition model that generates the state feature matrix , if the working condition model is a normal operation working condition model, then ; if the working condition model is any kind of potential damage working condition model, then .

8. The method of claim 1, wherein, In step S5, in the training of the arch dam body state prediction model, the K-means spatial clustering algorithm is used to optimize the sampling of the internal nodes of the dam body, and the monitoring node set with the optimal spatial distribution uniformity is determined; in the model training iteration process, a step-by-step down-sampling strategy is adopted, that is, the number of monitoring node samples is gradually reduced until it is close to the actual engineering monitoring node number, and a smooth transition from a theoretical model to an actual application is realized.

9. The method of claim 1, wherein, The arch dam body state prediction model is a mapping model for establishing a mapping from a displacement feature matrix to a state feature matrix, and is used for realizing the binary classification prediction of the normal / damaged state of the arch dam body.

Images