Concrete high arch dam construction early-stage toppling deformation monitoring system and monitoring method

Abstract

The invention discloses a concrete high arch dam construction early-stage toppling deformation monitoring system and method, and the method comprises the steps: obtaining high arch dam design parameters, constructing a three-dimensional dam body model, carrying out the stress intensity analysis based on the three-dimensional dam body model, and selecting deformation characteristics; the method comprises the following steps: classifying inclination forms of a monitoring area according to similarity of inclination tracks between the bottom and the top of the monitoring area, respectively constructing a first deformation prediction model and a second deformation prediction model for uniform inclination and non-uniform inclination, and judging whether a potential risk exists or not based on the deformation prediction model corresponding to the monitoring area, and obtaining strong correlation characteristics of the first deformation prediction model and the second deformation prediction model and the optimal weight of the strong correlation characteristics based on security, and constructing a deformation monitoring model based on the strong correlation characteristics and the optimal weight. The method not only can improve the efficiency and real-time performance of toppling deformation monitoring, but also has good interpretability, and can be directly applied to a concrete high arch dam construction early toppling deformation monitoring system.

Description

Technical Field

[0001] The invention relates to the field of tilt deformation monitoring, and in particular to a system and method for monitoring early collapse deformation during construction of a concrete high arch dam. Background Art

[0002] With the rapid development of the Internet of Things, artificial intelligence and big data, the intelligence and automation of early collapse deformation monitoring methods during the construction of concrete high arch dams have been continuously improved, and the accuracy and timeliness of monitoring have become key factors in ensuring project safety.

[0003] Traditional tilt and deformation monitoring methods mainly rely on manual measurement and on-site observation. In the early stages of high arch dam construction, it is difficult to achieve continuous and real-time monitoring due to construction interference and environmental restrictions.

[0004] At present, although there are some methods for monitoring the collapse and deformation of high arch dams, most of them are manually intervened by using inclinometers or similar equipment to measure the monitoring points. A large number of monitoring equipment need to be placed to conduct comprehensive monitoring of the entire area, and the capture of the influencing elements of the construction site is not comprehensive enough. In addition, most of the monitoring time is when the construction is completed, and there is a lack of prediction of potential risks and comprehensive analysis using multi-source data and multiple factors. Summary of the invention

[0005] The purpose of the present invention is to provide a system and method for monitoring the early collapse deformation of a concrete high arch dam during construction.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0007] The first aspect of the present invention provides a method for monitoring the early collapse deformation of a concrete high arch dam during construction, comprising the following steps:

[0008] Step S1, obtaining design parameters of a high arch dam, constructing a three-dimensional dam body model according to the design parameters, performing stress intensity analysis based on the three-dimensional dam body model, and selecting deformation features based on the stress intensity analysis results;

[0009] Step S2, predicting the tilt trajectories of the bottom and top of the monitoring area according to the deformation characteristics of the monitoring area, and classifying the tilt forms of the monitoring area according to the similarity of the tilt trajectories between the bottom and the top; the tilt forms include uniform tilt and non-uniform tilt;

[0010] Step S3, constructing a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt respectively, and judging whether the monitoring area has potential risks based on the deformation prediction model corresponding to the monitoring area; the first deformation prediction model is a deformation prediction based on a surface, and the second deformation prediction model is a deformation prediction based on a point;

[0011] Step S4: perform feature association analysis on the first deformation prediction model and the second deformation prediction model to obtain strong correlation features, obtain the optimal weight of the strong correlation features based on safety, build a deformation monitoring model based on the strong correlation features and the optimal weight, and output abnormal monitoring results with potential risk areas.

[0012] Furthermore, the method for constructing a three-dimensional dam model according to the design parameters includes:

[0013] According to the design parameters, the overall outline and structural elements of the dam body are constructed through engineering software, and the concrete material properties are assigned to the three-dimensional dam body model. The constraints in the three-dimensional dam body model are defined. The expression is:

[0014]

[0015] Among them, f1 is the continuity constraint of the point on the three-dimensional dam model, C represents the side boundary of the dam, and T represents the stress tensor of each point. It is represented by the divergence of the stress tensor, S represents the contact surface between the dam and the foundation, dS represents the surface integral of the process, f2 is the compatibility constraint at the connection, V represents the spatial volume of the three-dimensional dam model, ∈ ij It is expressed as the degree of deformation of the material in the ij direction in the direction matrix, E ij It is represented as the material modulus tensor in the ij direction in the direction matrix, dV represents the volume integral, t represents the force vector acting on S, u represents the displacement vector of the material, f3 represents the curvature constraint, k(s) represents the curvature of the point on S, and They are respectively expressed as the bending change rate of the dam surface in the vertical and horizontal directions, and They represent the tangent directions of the dam surface in the vertical and horizontal directions, f4 is the disturbance constraint, w(x,z) represents the vertical disturbance of the dam, x represents the distance from the arch dam base, ξ represents the vertical section of the dam, ξ * is a point in ξ, M(ξ * ,z) is expressed as the dam body at ξ * The bending moment at the position, q(ξ * ,z) represents the load on the dam body in the dam axis direction, E represents the elastic modulus of the material, I represents the moment of inertia of the vertical section of the dam body, dξ * dξ represents the integration of all points on the vertical section of the dam body, f5 is the multi-physics field coupling constraint, ρ represents the concrete density, c p Expressed as the specific heat capacity of concrete, Expressed as the time rate of change of the temperature field, is represented as the divergence operator, k is represented as the thermal conductivity of the dam body, is the rate of change of temperature in space, Q is the heat generation, σ ij It is expressed as the stress state in the ij direction in the material direction matrix.

[0016] Furthermore, the method of performing stress intensity analysis based on the three-dimensional dam model and selecting deformation features based on the stress intensity analysis results includes:

[0017] Based on the finite element method, the stress analysis of the cross section of the three-dimensional dam model is carried out, and the expression is:

[0018]

[0019] Among them, σ self Expressed as self-weight stress, σ hydro Expressed as water pressure stress, σ temp It is expressed as temperature thermal stress, γ is expressed as the unit weight of the dam material, h is expressed as the height of the dam, g(z) is expressed as the change of gravitational acceleration along the height z direction, P is expressed as the pressure of the water body on the dam body, n is expressed as the direction of the water pressure on the dam body, E is expressed as the elastic modulus of the material, α is expressed as the thermal expansion coefficient of the material, ΔT is expressed as the temperature change experienced by the material, v is expressed as the Poisson's ratio of the material, and I is expressed as the moment of inertia of the vertical section of the dam body. is expressed as a tensor product operation, (3I-2vI) is the anisotropy of the material, It is expressed as the stress state in the plane orthogonal to the water pressure direction n of the dam body;

[0020] Damage mechanics is used to simulate the concrete damage process, and the concrete collapse is counted. The area where the concrete collapse rate is greater than the safety standard is regarded as the stress concentration area.

[0021] Features with localized feature changes greater than 20% in the stress concentration area are extracted as deformation features.

[0022] Furthermore, the method for predicting the tilt trajectories of the bottom and top of the monitoring area according to the deformation characteristics of the monitoring area includes:

[0023] Set a time window, use the time window as the monitoring period, and collect time series data of deformation characteristics at the bottom and top of the monitoring area through the tilt sensor in the monitoring area;

[0024] Decompose the time series data into several intrinsic mode functions and residual components using variational mode decomposition;

[0025] A deep learning model including input layer, CNN layer, LSTM layer, fully connected layer and output layer is designed. The intrinsic mode function and residual component obtained are used as input features, and the Adam optimization algorithm is used to train the deep learning model.

[0026] An iterative method is used to take the prediction result of the previous time step of the deep learning model as the input of the current time step to obtain the tilt trajectory of the bottom and top of the monitoring area.

[0027] Furthermore, the method for classifying the tilt form of the monitoring area according to the similarity of the tilt track between the bottom and the top includes:

[0028] Calculate the similarity of the bottom and top tilt trajectories, the expression is:

[0029]

[0030] Where T1 and T2 represent the bottom tilt trajectory and the top tilt trajectory respectively. It is expressed as the number of sliding times of the time window, i.e., the number of monitoring times, which includes the trajectory in three-dimensional space and the tilted trajectory in time dimension, d 3D (T1, T2) represents the Euclidean distance between the bottom tilt trajectory and the top tilt trajectory, DTW(T1, T2) represents the dynamic time warping distance between the bottom tilt trajectory and the top tilt trajectory, v1 and v2 represent the velocity vectors of the tilt angle changes of the bottom tilt trajectory and the top tilt trajectory, respectively;

[0031] The average value of the trajectory similarity between the top and bottom of each vertical section in the high arch dam design standard is used as the division threshold. The monitoring area less than the threshold is the non-uniform inclination area, and the monitoring area greater than the threshold is the uniform inclination area.

[0032] Furthermore, the method of constructing a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt, respectively, comprises:

[0033] Construct the forward model, the expression is:

[0034]

[0035] +H(u(t),T(t),ε(t))=F ext

[0036] Where M represents the inertial force of the high arch dam structure to resist the change of tilt acceleration, u(t) represents the tilt displacement of the high arch dam, and They are respectively expressed as tilt acceleration and tilt velocity, It is expressed as, It is expressed as the restoring force of concrete material under the toppling deformation, is the geometric rigidity matrix, H(u(t),T(t),ε(t)) is the environmental load vector, F ext is the external load vector;

[0037] Perform deformation analysis on the surface in the uniform tilt type area based on the forward model to extract spatial features, and perform deformation analysis on the points in the non-uniform tilt type area based on the forward model to extract temporal features;

[0038] The first deformation prediction model is constructed based on the convolutional neural network and trained by the extracted spatial features. The second deformation prediction model is constructed based on the long short-term memory algorithm and trained by the extracted temporal features.

[0039] Furthermore, the method for performing feature association analysis on the first deformation prediction model and the second deformation prediction model includes:

[0040] The data of the second deformation prediction model is converted into a continuous spatial surface Z by spatial interpolation. p (x, y), calculate the correlation coefficient between the features of the first deformation prediction model and the second deformation prediction model, the calculation formula is:

[0041]

[0042] Where g is the number of spatial grid points, is the deformation prediction value of the i-th point obtained by spatial interpolation of the data of the second deformation prediction model, is the deformation prediction value corresponding to the i-th point of the first deformation prediction model, and are respectively the average of the deformation prediction value of the data of the second deformation prediction model through spatial interpolation and the deformation prediction value of the first deformation prediction model;

[0043] Features with a feature correlation greater than 0.7 are considered strong correlation features.

[0044] Furthermore, the method for obtaining the optimal weight of the strongly correlated feature based on security includes:

[0045] Taking maximizing safety as the objective function, the expression is:

[0046]

[0047] Among them, ξ is the safety metric, θ1 is the inclination angle of the high arch dam, γ is the deadweight of the risk area, and κ is 1m 3 Volumetric planned bearing water pressure, Α is the bottom thickness of the high arch dam, Β is the top thickness of the high arch dam, n is the number of monitoring points, φ i is the thermal stress at the ith monitoring point, λ max is the maximum temperature reached by concrete during pouring, minis the lowest temperature reached by concrete during the cooling process, v is the rate at which concrete drops from the highest temperature to the lowest temperature, ω1 is the weight of temperature control accuracy, θ2 is the crown angle of the high arch dam, σ(i) is the deformation displacement of the ith monitoring point, j is the arch height ratio of the high arch dam, t is the curing time of concrete, ω2 is the weight of the compressive strength of concrete per unit volume, μ b is the flexural modulus of concrete, μ c is the compression modulus of concrete, D v 1m 3 The compressive strength of concrete;

[0048] The particle position with the highest safety metric value is selected as the optimal weight solution through the particle swarm algorithm.

[0049] Furthermore, the method for constructing a deformation monitoring model based on the strong correlation feature and the optimal weight includes:

[0050] According to the obtained optimal weight solution, the strongly correlated features are weighted, and a deformation monitoring model is constructed based on a deep neural network. The weighted features are used as the initial value of the connection weight from the input layer to the hidden layer. The mean square error is used as the objective function, and the stochastic gradient descent algorithm is selected to train the network. The expression of the update rule of stochastic gradient descent is:

[0051]

[0052] Among them, θ t is the model parameter of the tth iteration, η is the learning rate, Expressed as the parameter θ t The objective function J at the sample (x i ,y i )’s gradient.

[0053] The second aspect of the present invention also provides a system for monitoring the early collapse and deformation of a concrete high arch dam during construction, comprising:

[0054] Feature selection module, which obtains the design parameters of the high arch dam, builds a three-dimensional dam model based on the design parameters, performs stress strength analysis based on the three-dimensional dam model, and selects deformation features based on the stress strength analysis results;

[0055] The tilt deformation type classification module obtains the tilt deformation information of the monitoring area, obtains the tilt trajectories of the bottom and the top according to the tilt deformation information, and classifies the tilt form of the monitoring area according to the similarity of the tilt trajectories between the bottom and the top;

[0056] A potential risk area identification module constructs a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt, respectively, and determines whether there is a potential risk based on the deformation prediction model corresponding to the monitoring area;

[0057] The monitoring model construction module performs feature correlation analysis on the first deformation prediction model and the second deformation prediction model to obtain strong correlation features, obtains the optimal weights of the strong correlation features based on safety, constructs a deformation monitoring model based on the strong correlation features and the optimal weights, and outputs abnormal monitoring results of monitoring areas with potential risks.

[0058] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0059] (1) The present invention provides a system and method for monitoring the early collapse deformation of a concrete high arch dam during construction. By performing stress analysis based on a three-dimensional model, regions with weak stress intensity are selected, and feature selection is performed based on the weak stress regions, thereby improving monitoring efficiency and resource utilization efficiency.

[0060] (2) The present invention constructs a first deformation prediction model based on a surface and a second deformation prediction model based on a point by similarity classification based on tilt trajectories, so that the monitoring model fits the deformation features based on the surface and the points, thereby achieving more refined and comprehensive monitoring and improving the accuracy and pertinence of monitoring;

[0061] (3) The present invention can identify features that have a significant impact on deformation under different conditions through feature association analysis. By identifying and quantifying the impact of strongly correlated features on safety, risk monitoring and management can be performed more effectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The present invention is a flowchart of the steps of the method for monitoring the early collapse deformation of a concrete high arch dam during construction. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0064] Reference Figure 1 As shown, the present invention provides a system and method for monitoring the early collapse deformation of a concrete high arch dam during construction, comprising:

[0065] Step S1, obtaining design parameters of a high arch dam, constructing a three-dimensional dam body model according to the design parameters, performing stress intensity analysis based on the three-dimensional dam body model, and selecting deformation features based on the stress intensity analysis results;

[0066] In the actual evaluation, the design parameters of the high arch dam, including geometric dimensions, material properties, boundary conditions, etc., were obtained. The engineering software Autodesk 3ds Max was used to build a three-dimensional model of the dam body. The physical properties of elastic modulus, Poisson's ratio, density and thermal expansion coefficient were assigned to the concrete material in the three-dimensional model, and multiple constraints were applied to constrain the three-dimensional model. The three-dimensional model was imported into the ANSYS finite element analysis software for meshing. The quadratic tetrahedron was selected as the unit type, and the mesh size ratio was set to: the minimum mesh size h min =0.5m, maximum grid size h max =0.5m;

[0067] In the actual evaluation, solid boundary conditions are defined at the dam foundation to simulate the fixed connection with the foundation. The self-weight, water pressure and temperature load stress are applied to the cross-section of the three-dimensional dam model for stress analysis. According to the stress analysis results, the area where the concrete slump rate is greater than the safety standard value is selected as the stress concentration area. The average value of the deformation characteristics of the stress concentration area is analyzed, including the vertical displacement Δy of the section, the horizontal displacement Δx of the section, the average inclination rate avg_θ of the section, and the maximum offset max of the section. δ , Maximum rotation angle of the section The cross-sectional displacement change rate Δd_rate and the cross-sectional total displacement total_Δ are accumulated, and features with feature changes greater than 20% are selected as deformation features, including: cross-sectional vertical displacement Δy, cross-sectional horizontal displacement Δx, cross-sectional average inclination rate avg_θ, and cross-sectional displacement change rate Δd_rate.

[0068] Step S2, predicting the tilt trajectories of the bottom and top of the monitoring area according to the deformation characteristics of the monitoring area, and classifying the tilt forms of the monitoring area according to the similarity of the tilt trajectories between the bottom and the top; the tilt forms include uniform tilt and non-uniform tilt;

[0069] In the actual evaluation, there are 4 monitoring areas, namely area A, area B, area C and area D. Tilt sensors are installed at the top and bottom of the monitoring area, and the time window step is set to 1 minute. The time series of deformation characteristics of the top and bottom of the monitoring area are periodically obtained, including horizontal displacement, vertical displacement and tilt angle. The obtained deformation characteristics are fused with the time series data to form a comprehensive data set. The comprehensive data set is decomposed into several intrinsic mode functions and residual components using variational mode decomposition. The first 4 IMFs and residual components are selected as input features. The feature dimension of each IMF is 5. The deep learning model is set , including: input layer, receiving 20 features (4 IMFs*5 features / IMF+residual component), CNN layer, Conv1D(64,kernel_size=3,activation='relu'), LSTM layer, LSTM(50,activation='relu'), fully connected layer, Dense(10,activation='relu') and output layer, Dense(1), predicting the deformation features of the next time step, using the deep learning model for iterative prediction, forming the tilt trajectories at the bottom and top of the monitoring area. Further calculate the similarity of the top and bottom tilt trajectories, and classify the tilt type of the monitoring area based on the similarity of the tilt trajectories, and obtain that areas A, B and D are uniform tilt areas, and area C is non-uniform tilt area.

[0070] Step S3, constructing a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt respectively, and judging whether there is a potential risk based on the deformation prediction model corresponding to the monitoring area; the first deformation prediction model is a deformation prediction based on a surface, and the second deformation prediction model is a deformation prediction based on a point;

[0071] In the actual assessment, a forward model of toppling deformation is constructed, and deformation analysis is performed on the surface of the uniform tilt area to extract spatial characteristics, including the uniformity of deformation distribution and the amplitude of deformation. Deformation analysis is performed on the points of the non-uniform tilt area to extract temporal characteristics, including the trend and periodicity of deformation over time.

[0072] In the actual evaluation, a convolutional neural network is used to construct the first deformation prediction model, and the spatial features extracted from the uniform tilt area are used to train the first deformation prediction model. A long short-term memory network is used to construct the second deformation prediction model, and the time features extracted from the non-uniform tilt area are used to train the second deformation prediction model. The number of convolutions is selected as 10, the convolution kernel size is 2 rows and 1 column, the root mean square error RMSE is selected as the error function, the Adam algorithm is the optimization algorithm of the model, the number of neurons in the LSTM layer is 32, the epoch is 1000 times, and the batchsize is 100. The tilt deformation prediction curves of the first deformation prediction model and the second deformation prediction model are obtained for area B. The tilt safety standard range is: the safety threshold of the vertical displacement is ±0.05mm. In the first prediction model of area B, the vertical displacement of the surface of area B is +0.6mm, which exceeds the safety threshold, and area B is judged to be a potential risk area.

[0073] Step S4: perform feature association analysis on the first deformation prediction model and the second deformation prediction model to obtain strong correlation features, obtain the optimal weight of the strong correlation features based on safety, build a deformation monitoring model based on the strong correlation features and the optimal weight, and output abnormal monitoring results with potential risk areas.

[0074] In the actual evaluation, the data of the first deformation prediction model and the second deformation prediction model are collected, including: obtaining the data of the first deformation prediction model and the data of the second deformation prediction model in area B, using the inverse distance weighted method to convert the point data of the second model into the same spatial grid surface as the first deformation prediction model, for each spatial grid point, calculating the correlation coefficient of the prediction values ​​of the two models, selecting feature points with a correlation coefficient greater than 0.7 as strong correlation features, including: tilt angle, deformation rate, temperature change rate, concrete temperature difference, water pressure and concrete curing time, using the particle swarm algorithm to maximize the safety metric value, and obtaining the optimal weight solutions, which are 0.23, 0.12, 0.14, 0.18, 0.17 and 0.16 respectively, and constructing a monitoring model based on the deep learning network based on the optimal weight solution and the strong correlation features, and performing 50 epochs of training in the network training process, with the learning rate set to 0.01, to obtain the trained deformation monitoring model and input it into The abnormal monitoring results of area B are obtained, and the abnormal degree and abnormal time of the combination of the tilt angle, deformation rate, temperature change rate, concrete temperature difference, water pressure and concrete curing time of area B are obtained, including: the normal tilt angle of area B is between 0.01 degrees and 0.05 degrees, and the tilt angle is predicted to increase to 0.2 degrees within 12 hours; the normal deformation rate of area B is 0.01mm / day, and the deformation rate is predicted to increase to 0.5mm / day; the temperature change rate of area B is predicted to increase from the normal 0.5℃ / day to 2℃ / day within 17 hours; the temperature difference between the inside and the surface of the concrete should not exceed 10℃, and it is predicted that the temperature difference will reach 20℃ within 4 hours; the water pressure of area B is predicted to increase from the normal 100kPa to 200kPa within 14 hours; the concrete curing time of area B is predicted to be 22 days in the second month, which is less than the standard curing time of 28 days; the construction personnel optimize the construction process based on the abnormal degree and abnormal time.

[0075] In this embodiment, the method for constructing a three-dimensional dam model according to the design parameters includes:

[0076] According to the design parameters, the overall outline and structural elements of the dam body are constructed through engineering software, and the concrete material properties are assigned to the three-dimensional dam body model. The constraints in the three-dimensional dam body model are defined. The expression is:

[0077]

[0078] Among them, f1 is the continuity constraint of the point on the three-dimensional dam model, C represents the side boundary of the dam, and T represents the stress tensor of each point. It is represented by the divergence of the stress tensor, S represents the contact surface between the dam and the foundation, dS represents the surface integral of the process, f2 is the compatibility constraint at the connection, V represents the spatial volume of the three-dimensional dam model, ∈ij It is expressed as the degree of deformation of the material in the ij direction in the direction matrix, E ij It is represented as the material modulus tensor in the ij direction in the direction matrix, dV represents the volume integral, t represents the force vector acting on S, u represents the displacement vector of the material, f3 represents the curvature constraint, k(s) represents the curvature of the point on S, and They are respectively expressed as the bending change rate of the dam surface in the vertical and horizontal directions, and They represent the tangent directions of the dam surface in the vertical and horizontal directions, f4 is the disturbance constraint, w(x,z) represents the vertical disturbance of the dam, x represents the distance from the arch dam base, ξ represents the vertical section of the dam, ξ * is a point in ξ, M(ξ * ,z) is expressed as the dam body at ξ * The bending moment at the position, q(ξ * ,z) represents the load on the dam body in the dam axis direction, E represents the elastic modulus of the material, I represents the moment of inertia of the vertical section of the dam body, dξ * dξ represents the integration of all points on the vertical section of the dam body, f5 is the multi-physics field coupling constraint, ρ represents the concrete density, c p Expressed as the specific heat capacity of concrete, Expressed as the time rate of change of the temperature field, is represented as the divergence operator, k is represented as the thermal conductivity of the dam body, is the rate of change of temperature in space, Q is the heat generation, σ ij It is expressed as the stress state in the ij direction in the material direction matrix.

[0079] In this embodiment, the method of performing stress intensity analysis based on the three-dimensional dam model and selecting deformation features based on the stress intensity analysis results includes:

[0080] Based on the finite element method, the stress analysis of the cross section of the three-dimensional dam model is carried out, and the expression is:

[0081]

[0082] Among them, σ self Expressed as self-weight stress, σ hydro Expressed as water pressure stress, σ tempIt is expressed as temperature thermal stress, γ is expressed as the unit weight of the dam material, h is expressed as the height of the dam, g(z) is expressed as the change of gravitational acceleration along the height z direction, P is expressed as the pressure of the water body on the dam body, n is expressed as the direction of the water pressure on the dam body, E is expressed as the elastic modulus of the material, α is expressed as the thermal expansion coefficient of the material, ΔT is expressed as the temperature change experienced by the material, v is expressed as the Poisson's ratio of the material, and I is expressed as the moment of inertia of the vertical section of the dam body. is expressed as a tensor product operation, (3I-2vI) is the anisotropy of the material, It is expressed as the stress state in the plane orthogonal to the water pressure direction n of the dam body;

[0083] Damage mechanics is used to simulate the concrete damage process, and the concrete collapse is counted. The area where the concrete collapse rate is greater than the safety standard is regarded as the stress concentration area.

[0084] Features with localized feature changes greater than 20% in the stress concentration area are extracted as deformation features.

[0085] In this embodiment, the method for predicting the tilt trajectories of the bottom and top of the monitoring area according to the deformation characteristics of the monitoring area includes:

[0086] Set a time window, use the time window as the monitoring period, and collect time series data of deformation characteristics at the bottom and top of the monitoring area through the tilt sensor in the monitoring area;

[0087] Decompose the time series data into several intrinsic mode functions and residual components using variational mode decomposition;

[0088] A deep learning model including input layer, CNN layer, LSTM layer, fully connected layer and output layer is designed. The intrinsic mode function and residual component obtained are used as input features, and the Adam optimization algorithm is used to train the deep learning model.

[0089] An iterative method is used to take the prediction result of the previous time step of the deep learning model as the input of the current time step to obtain the tilt trajectory of the bottom and top of the monitoring area.

[0090] In this embodiment, the method for classifying the tilt form of the monitoring area according to the similarity of the tilt track between the bottom and the top includes:

[0091] Calculate the similarity of the bottom and top tilt trajectories, the expression is:

[0092]

[0093] Where T1 and T2 represent the bottom tilt trajectory and the top tilt trajectory respectively. It is expressed as the number of sliding times of the time window, i.e., the number of monitoring times, which includes the trajectory in three-dimensional space and the tilted trajectory in time dimension, d 3D (T1, T2) represents the Euclidean distance between the bottom tilt trajectory and the top tilt trajectory, DTW(T1, T2) represents the dynamic time warping distance between the bottom tilt trajectory and the top tilt trajectory, v1 and v2 represent the velocity vectors of the tilt angle changes of the bottom tilt trajectory and the top tilt trajectory, respectively;

[0094] The average value of the trajectory similarity between the top and bottom of each vertical section in the high arch dam design standard is used as the division threshold. The monitoring area less than the threshold is the non-uniform inclination area, and the monitoring area greater than the threshold is the uniform inclination area.

[0095] In this embodiment, the method of constructing the first deformation prediction model and the second deformation prediction model for uniform tilt and non-uniform tilt respectively includes:

[0096] Construct the forward model, the expression is:

[0097]

[0098] Where M represents the inertial force of the high arch dam structure to resist the change of tilt acceleration. and They are respectively represented as the tilt acceleration and tilt velocity, u(t) represents the tilt displacement of the high arch dam, It is expressed as the damage variable of concrete material under the toppling deformation, It is expressed as the restoring force of concrete material under the toppling deformation, is the geometric rigidity matrix, H(u(t),T(t),ε(t)) is the environmental load vector, F ext is the external load vector;

[0099] Perform deformation analysis on the surface in the uniform tilt type area based on the forward model to extract spatial features, and perform deformation analysis on the points in the non-uniform tilt type area based on the forward model to extract temporal features;

[0100] The first deformation prediction model is constructed based on the convolutional neural network and trained by the extracted spatial features. The second deformation prediction model is constructed based on the long short-term memory algorithm and trained by the extracted temporal features.

[0101] In this embodiment, the method for performing feature association analysis on the first deformation prediction model and the second deformation prediction model includes:

[0102] The data of the second deformation prediction model is converted into a continuous spatial surface Z by spatial interpolation. p(x, y), calculate the correlation coefficient between the features of the first deformation prediction model and the second deformation prediction model, the calculation formula is:

[0103]

[0104] Where g is the number of spatial grid points, is the deformation prediction value of the i-th point obtained by spatial interpolation of the data of the second deformation prediction model, is the deformation prediction value corresponding to the i-th point of the first deformation prediction model, and are respectively the average of the deformation prediction value of the data of the second deformation prediction model through spatial interpolation and the deformation prediction value of the first deformation prediction model;

[0105] Features with a feature correlation greater than 0.7 are considered strong correlation features.

[0106] In this embodiment, the method for obtaining the optimal weight of the strong correlation feature based on security includes:

[0107] Taking maximizing safety as the objective function, the expression is:

[0108]

[0109] Among them, ξ is the safety metric, θ1 is the inclination angle of the high arch dam, γ is the deadweight of the risk area, and κ is 1m 3 Volumetric planned bearing water pressure, Α is the bottom thickness of the high arch dam, Β is the top thickness of the high arch dam, n is the number of monitoring points, φ i is the thermal stress at the ith monitoring point, λ max is the maximum temperature reached by concrete during pouring, min is the lowest temperature reached by concrete during the cooling process, v is the rate at which concrete drops from the highest temperature to the lowest temperature, ω1 is the weight of temperature control accuracy, θ2 is the crown angle of the high arch dam, σ(i) is the deformation displacement of the ith monitoring point, j is the arch height ratio of the high arch dam, t is the curing time of concrete, ω2 is the weight of the compressive strength of concrete per unit volume, μ b is the flexural modulus of concrete, μ c is the compression modulus of concrete, D v 1m 3 The compressive strength of concrete;

[0110] The particle position with the highest safety metric value is selected as the optimal weight solution through the particle swarm algorithm.

[0111] In this embodiment, the method for constructing a deformation monitoring model based on the strong correlation feature and the optimal weight includes:

[0112] According to the obtained optimal weight solution, the strongly correlated features are weighted, and a deformation monitoring model is constructed based on a deep neural network. The weighted features are used as the initial value of the connection weight from the input layer to the hidden layer. The mean square error is used as the objective function, and the stochastic gradient descent algorithm is selected to train the network. The expression of the update rule of stochastic gradient descent is:

[0113]

[0114] Among them, θ t is the model parameter of the tth iteration, η is the learning rate, Expressed as the parameter θ t The objective function J at the sample (x i ,y i )’s gradient.

[0115] The second aspect of the present invention also provides a system for monitoring the early collapse and deformation of a concrete high arch dam during construction, comprising:

[0116] Feature selection module, which obtains the design parameters of the high arch dam, builds a three-dimensional dam model based on the design parameters, performs stress strength analysis based on the three-dimensional dam model, and selects deformation features based on the stress strength analysis results;

[0117] The tilt deformation type classification module obtains the tilt deformation information of the monitoring area, obtains the tilt trajectories of the bottom and the top according to the tilt deformation information, and classifies the tilt form of the monitoring area according to the similarity of the tilt trajectories between the bottom and the top;

[0118] A potential risk area identification module constructs a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt, respectively, and determines whether there is a potential risk based on the deformation prediction model corresponding to the monitoring area;

[0119] The monitoring model construction module performs feature correlation analysis on the first deformation prediction model and the second deformation prediction model to obtain strong correlation features, obtains the optimal weights of the strong correlation features based on safety, constructs a deformation monitoring model based on the strong correlation features and the optimal weights, and outputs abnormal monitoring results of monitoring areas with potential risks.

[0120] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for monitoring the early collapse deformation of a concrete high arch dam during construction, characterized in that: The following steps are involved: Step S1, obtaining design parameters of a high arch dam, constructing a three-dimensional dam body model according to the design parameters, performing stress intensity analysis based on the three-dimensional dam body model, and selecting deformation features based on the stress intensity analysis results; Step S2, predicting the tilt trajectories of the bottom and top of the monitoring area according to the deformation characteristics of the monitoring area, and classifying the tilt forms of the monitoring area according to the similarity of the tilt trajectories between the bottom and the top; the tilt forms include uniform tilt and non-uniform tilt; Step S3, constructing a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt respectively, and judging whether the monitoring area has potential risks based on the deformation prediction model corresponding to the monitoring area; the first deformation prediction model is a deformation prediction based on a surface, and the second deformation prediction model is a deformation prediction based on a point; Step S4: perform feature association analysis on the first deformation prediction model and the second deformation prediction model to obtain strong correlation features, obtain the optimal weight of the strong correlation features based on safety, build a deformation monitoring model based on the strong correlation features and the optimal weight, and output abnormal monitoring results with potential risk areas.

2. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method for constructing a three-dimensional dam body model according to the design parameters comprises: According to the design parameters, the overall outline and structural elements of the dam body are constructed through engineering software, and the concrete material properties are assigned to the three-dimensional dam body model. The constraints in the three-dimensional dam body model are defined. The expression is: Among them, f1 is the continuity constraint of the point on the three-dimensional dam model, C represents the side boundary of the dam, and T represents the stress tensor of each point. It is represented by the divergence of the stress tensor, S represents the contact surface between the dam and the foundation, dS represents the surface integral of the process, f2 is the compatibility constraint at the connection, V represents the spatial volume of the three-dimensional dam model, ∈ ij It is expressed as the degree of deformation of the material in the ij direction in the direction matrix, E ij It is represented as the material modulus tensor in the ij direction in the direction matrix, dV represents the volume integral, t represents the force vector acting on S, u represents the displacement vector of the material, f3 represents the curvature constraint, k(s) represents the curvature of the point on S, and They are respectively expressed as the bending change rate of the dam surface in the vertical and horizontal directions, and They represent the tangent directions of the dam surface in the vertical and horizontal directions, f4 is the disturbance constraint, w(x,z) represents the vertical disturbance of the dam, x represents the distance from the arch dam base, ξ represents the vertical section of the dam, ξ * is a point in ξ, M(ξ * ,z) is expressed as the dam body at ξ * The bending moment at the position, q(ξ * ,z) represents the load on the dam body in the dam axis direction, E represents the elastic modulus of the material, I represents the moment of inertia of the vertical section of the dam body, dξ * dξ represents the integration of all points on the vertical section of the dam body, f5 is the multi-physics field coupling constraint, ρ represents the concrete density, c p Expressed as the specific heat capacity of concrete, Expressed as the time rate of change of the temperature field, is represented as the divergence operator, k is represented as the thermal conductivity of the dam body, is the rate of change of temperature in space, Q is the heat generation, σ ij It is expressed as the stress state in the ij direction in the material direction matrix.

3. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method of performing stress intensity analysis based on the three-dimensional dam body model and selecting deformation features based on the stress intensity analysis result includes: Based on the finite element method, the stress analysis of the cross section of the three-dimensional dam model is carried out, and the expression is: Among them, σ self Expressed as self-weight stress, σ hydro Expressed as water pressure stress, σ temp It is expressed as temperature thermal stress, γ is expressed as the unit weight of the dam material, h is expressed as the height of the dam, g(z) is expressed as the change of gravitational acceleration along the height z direction, P is expressed as the pressure of the water body on the dam body, n is expressed as the direction of the water pressure on the dam body, E is expressed as the elastic modulus of the material, α is expressed as the thermal expansion coefficient of the material, ΔT is expressed as the temperature change experienced by the material, v is expressed as the Poisson's ratio of the material, and I is expressed as the moment of inertia of the vertical section of the dam body. is expressed as a tensor product operation, (3I-2vI) is expressed as the anisotropy of the material, It is expressed as the stress state in the plane orthogonal to the water pressure direction n of the dam body; Damage mechanics is used to simulate the concrete damage process, and the concrete collapse is counted. The area where the concrete collapse rate is greater than the safety standard is regarded as the stress concentration area. Features with localized feature changes greater than 20% in the stress concentration area are extracted as deformation features.

4. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method for predicting the tilt trajectories of the bottom and top of the monitoring area according to the deformation characteristics of the monitoring area includes: Set a time window, use the time window as the monitoring period, and collect time series data of deformation characteristics at the bottom and top of the monitoring area through the tilt sensor in the monitoring area; Decompose the time series data into several intrinsic mode functions and residual components using variational mode decomposition; A deep learning model including input layer, CNN layer, LSTM layer, fully connected layer and output layer is designed. The intrinsic mode function and residual component obtained are used as input features, and the Adam optimization algorithm is used to train the deep learning model. An iterative method is used to take the prediction result of the previous time step of the deep learning model as the input of the current time step to obtain the tilt trajectory of the bottom and top of the monitoring area.

5. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method for classifying the tilt form of the monitoring area according to the similarity of the tilt track between the bottom and the top includes: Calculate the similarity of the bottom and top tilt trajectories, the expression is: Where T1 and T2 represent the bottom tilt trajectory and the top tilt trajectory respectively. It is expressed as the number of sliding times of the time window, i.e., the number of monitoring times, which includes the trajectory in three-dimensional space and the tilted trajectory in time dimension, d 3D (T1, T2) represents the Euclidean distance between the bottom tilt trajectory and the top tilt trajectory, DTW(T1, T2) represents the dynamic time warping distance between the bottom tilt trajectory and the top tilt trajectory, v1 and v2 represent the velocity vectors of the tilt angle changes of the bottom tilt trajectory and the top tilt trajectory, respectively; The average value of the trajectory similarity between the top and bottom of each vertical section in the high arch dam design standard is used as the division threshold. The monitoring area less than the threshold is the non-uniform inclination area, and the monitoring area greater than the threshold is the uniform inclination area.

6. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method of constructing a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt respectively comprises: Construct the forward model, the expression is: Where M represents the inertial force of the high arch dam structure to resist the change of tilt acceleration, u(t) represents the tilt displacement of the high arch dam, and They are respectively expressed as tilt acceleration and tilt velocity, It is expressed as, It is expressed as the restoring force of concrete material under the toppling deformation, is the geometric rigidity matrix, H(u(t),T(t),ε(t)) is the environmental load vector, which may include the temperature distribution of the dam and the surrounding environment, and F ext is the external load vector; Perform deformation analysis on the surface in the uniform tilt type area based on the forward model to extract spatial features, and perform deformation analysis on the points in the non-uniform tilt type area based on the forward model to extract temporal features; The first deformation prediction model is constructed based on the convolutional neural network and trained by the extracted spatial features. The second deformation prediction model is constructed based on the long short-term memory algorithm and trained by the extracted temporal features.

7. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method for performing feature association analysis on the first deformation prediction model and the second deformation prediction model includes: The data of the second deformation prediction model is converted into a continuous spatial surface Z by spatial interpolation. p (x, y), calculate the correlation coefficient between the features of the first deformation prediction model and the second deformation prediction model, the calculation formula is: Where g is the number of spatial grid points, is the deformation prediction value of the i-th point obtained by spatial interpolation of the data of the second deformation prediction model, is the deformation prediction value corresponding to the i-th point of the first deformation prediction model, and are respectively the average of the deformation prediction value of the data of the second deformation prediction model through spatial interpolation and the deformation prediction value of the first deformation prediction model; Features with a feature correlation greater than 0.7 are considered strong correlation features.

8. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method for obtaining the optimal weight of the strongly correlated feature based on security includes: Taking maximizing safety as the objective function, the expression is: Among them, ξ is the safety metric, θ1 is the inclination angle of the high arch dam, γ is the deadweight of the risk area, and κ is 1m 3 Volumetric planned bearing water pressure, Α is the bottom thickness of the high arch dam, Β is the top thickness of the high arch dam, n is the number of monitoring points, φ i is the thermal stress at the ith monitoring point, λ max is the maximum temperature reached by concrete during pouring, min is the lowest temperature reached by concrete during the cooling process, v is the rate at which concrete drops from the highest temperature to the lowest temperature, ω1 is the weight of temperature control accuracy, θ2 is the crown angle of the high arch dam, σ(i) is the deformation displacement of the ith monitoring point, j is the arch height ratio of the high arch dam, t is the curing time of concrete, ω2 is the weight of the compressive strength of concrete per unit volume, μ b is the flexural modulus of concrete, μ c is the compression modulus of concrete, D v 1m 3 The compressive strength of concrete; The particle position with the highest safety metric value is selected as the optimal weight solution through the particle swarm algorithm.

9. The method for monitoring the early collapse deformation of a concrete high arch dam during construction according to claim 1, characterized in that: The method for constructing a deformation monitoring model based on the strong correlation feature and the optimal weight includes: According to the obtained optimal weight solution, the strongly correlated features are weighted, and a deformation monitoring model is constructed based on a deep neural network. The weighted features are used as the initial value of the connection weight from the input layer to the hidden layer. The mean square error is used as the objective function, and the stochastic gradient descent algorithm is selected to train the network. The expression of the update rule of stochastic gradient descent is: Among them, θ t+1 and θ t are the model parameters for the t+1th and tth iterations respectively, η is the learning rate, Expressed as the parameter θ t The objective function J at the sample (x i ,y i )’s gradient.

10. The early collapse deformation monitoring system for concrete high arch dam construction is characterized by: The system comprises: Feature selection module, which obtains the design parameters of the high arch dam, builds a three-dimensional dam model based on the design parameters, performs stress strength analysis based on the three-dimensional dam model, and selects deformation features based on the stress strength analysis results; The tilt deformation type classification module obtains the tilt deformation information of the monitoring area, obtains the tilt trajectories of the bottom and the top according to the tilt deformation information, and classifies the tilt form of the monitoring area according to the similarity of the tilt trajectories between the bottom and the top; A potential risk area identification module constructs a first deformation prediction model and a second deformation prediction model for uniform tilt and non-uniform tilt, respectively, and determines whether there is a potential risk based on the deformation prediction model corresponding to the monitoring area; The monitoring model construction module performs feature correlation analysis on the first deformation prediction model and the second deformation prediction model to obtain strong correlation features, obtains the optimal weights of the strong correlation features based on safety, constructs a deformation monitoring model based on the strong correlation features and the optimal weights, and outputs abnormal monitoring results of monitoring areas with potential risks.

Images