Damage prediction system for concrete multiple-arch dam under action of impact load
By combining three-dimensional finite element analysis and RBF neural network, the problem of the inability of traditional methods to accurately assess the damage of concrete arch dams has been solved, achieving efficient and accurate damage prediction under complex loads and improving the practicality of dam health monitoring.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- ANHUI SCI & TECH UNIV
- Filing Date
- 2024-12-19
- Publication Date
- 2026-06-25
AI Technical Summary
Traditional methods cannot assess the damage status of concrete arch dams under complex working conditions and dynamic loads in real time, especially under water flow impact loads. Existing finite element analysis cannot directly give the degree and location of damage to the dam body.
The response of a concrete arch dam is simulated using a three-dimensional finite element analysis module, the flexibility calculation module is used to analyze the change in flexibility curvature of the dam body, the damage function module is used to calculate the degree of damage, and the model is trained by RBF neural network to predict the location and degree of damage.
It enables more accurate assessment of dam damage under complex loads, reduces prediction errors, meets the needs of rapid response and accurate prediction in real-time monitoring, and improves the reliability and practicality of dam health monitoring.
Smart Images
Figure CN2024140618_25062026_PF_FP_ABST
Abstract
Description
Damage prediction system for concrete arch dams under impact load Technical Field
[0001] This invention belongs to the field of damage prediction for arch dams, and more specifically, it relates to a damage prediction system for concrete arch dams under impact loads. Background Technology
[0002] With the construction and operation of reservoirs and dams, the safety and stability of dam structures have become key factors in ensuring the long-term reliable operation of water conservancy projects. Due to the impact of natural disasters (such as floods and earthquakes) and human factors (such as engineering defects and material aging) on dams, especially under the influence of water flow impact loads, dams may suffer damage of varying degrees. Traditional dam damage assessment methods mainly rely on manual inspection, physical experiments, and empirical formulas. However, these methods are generally unable to assess the damage status of dams in real time, comprehensively, and accurately, especially when dealing with dam damage prediction under complex working conditions and dynamic loads, thus exhibiting certain limitations.
[0003] Finite element analysis (FEA) is a powerful numerical simulation method widely used in calculating the stress, strain, and displacement responses of dam structures. By performing three-dimensional finite element modeling of the dam, the response of the dam under complex loads such as water flow impact can be accurately simulated, providing fundamental mechanical data. However, traditional finite element analysis can only provide static data such as stress and deformation, and cannot directly indicate the degree and location of damage to the dam.
[0004] In view of this, the present invention is proposed. Summary of the Invention
[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a damage prediction system for concrete arch dams under impact load, which solves the problems mentioned in the background art.
[0006] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows:
[0007] A damage prediction system for concrete arch dams under impact load includes: a finite element analysis module for establishing a three-dimensional finite element model, simulating the response of the concrete arch dam under water flow impact load, and analyzing and extracting stress, strain and displacement data of the dam body.
[0008] The flexibility calculation module calculates the flexibility matrix of the dam body based on the data provided by the finite element analysis module, analyzes the changes in the flexibility curvature of the dam body, and extracts the sensitivity of structural deformation.
[0009] The damage function module calculates the degree of damage to the dam body and generates damage indicators based on the change in flexibility curvature using a preset damage function.
[0010] The RBF neural network module is used to train the neural network model by taking the damage index output by the damage function module as input. After training, the RBF neural network is used to predict the location and extent of damage to the dam body.
[0011] The damage prediction result output module is used to output the predicted damage location and damage degree.
[0012] The damage prediction system for a concrete arch dam under impact load according to claim 1 is characterized in that the steps for establishing a three-dimensional finite element model, simulating the response of the concrete arch dam under water flow impact load, and analyzing and extracting the stress, strain, and displacement data of the dam body are as follows:
[0013] Based on the design drawings or on-site measurement data of the concrete arch dam, the geometric dimensions and shape of the dam body are obtained. A three-dimensional geometric model of the concrete arch dam is generated using three-dimensional modeling software. Then, corresponding material properties are defined for different components of the dam body and corresponding material models are assigned to different parts. The material properties include elastic modulus, Poisson's ratio, and yield strength.
[0014] The three-dimensional geometric model was discretized into a finite element mesh using three-dimensional tetrahedral elements. Then, fixed boundary conditions were applied to the bottom and sides of the dam, and contact conditions between the dam and the foundation, and between the dam and the water flow, were applied according to actual conditions.
[0015] Water flow impact loads are applied using dynamic pressure and shock wave models. The time history, amplitude, and location of the load are defined and applied to the contact surface of the dam. The applied water flow impact load includes the kinetic energy of the water flow and the shock wave model. The amplitude, time history, and location of the load are defined according to the actual working conditions. Finally, after the dynamic analysis calculation is completed, the stress field, strain field, and displacement field of the dam are extracted from the finite element analysis results, with a focus on the stress, strain, and displacement changes in key parts of the dam.
[0016] The damage prediction system for a concrete arch dam under impact load according to claim 1 is characterized in that the steps of calculating the flexibility matrix of the dam body, analyzing the change in flexibility curvature of the dam body, and extracting the sensitivity of structural deformation based on the data provided by the finite element analysis module are as follows:
[0017] The finite element analysis module is used to obtain stress, displacement and deformation data of the dam body, and the nodal displacement data and stiffness matrix of the dam body are extracted based on the analysis results.
[0018] Using the stiffness matrix K from the finite element analysis results, the flexibility matrix F of the dam body is calculated according to the following formula: F = K -1Where K is the stiffness matrix and F is the flexibility matrix, the flexibility values of each part of the dam body are obtained through this calculation;
[0019] Based on the calculated flexibility matrix F, the variation of the flexibility curvature of the dam body under different working conditions is analyzed. Calculated using the following formula: Where ΔF represents the change in the compliance matrix under different impact loads, and F0 represents the initial compliance matrix under unloaded conditions. The change in flexibility curvature reflects the deformation sensitivity of the dam body under load.
[0020] By analyzing the changes in flexibility curvature, the deformation sensitivity of the dam body under impact load is extracted. The larger the sensitivity value, the stronger the response of a certain part of the dam body to load changes, and vice versa.
[0021] The damage prediction system for a concrete arch dam under impact load according to claim 1 is characterized in that the damage function module calculates the degree of damage to the dam body and generates damage indices based on the change in flexibility curvature using a preset damage function, as follows:
[0022] As the impact force of water flow increases, the local deformation and stress response of the dam body often increase nonlinearly. Especially under large impacts, the dam body undergoes significant deformation or failure. At this point, the relationship between the degree of damage and the impact load is no longer a simple linear one. Therefore, an exponential damage function is adopted. To determine the degree of damage, the first step is to calculate the change in flexibility curvature. The damage degree D of the dam body is obtained by inputting it into a preset damage function, where D is the damage degree. Let α be the change in flexibility curvature, and β be empirical constants.
[0023] According to claim 1, a damage prediction system for a concrete arch dam under impact load is characterized in that an RBF neural network module is established to take the damage index output by the damage function module as input, and the step of training the neural network model is as follows:
[0024] Based on the damage index, the structure of the RBF neural network is selected, the number of hidden layer nodes is determined, and the parameters of the RBF neural network are initialized, with the radial basis function type set to Gaussian function;
[0025] The damage indicators output from the damage function module are paired with the actual measured damage data to construct a dataset for training the neural network, ensuring that the dataset includes a complete pairing of damage indicators and actual damage conditions.
[0026] The training dataset is input into the RBF neural network. The neural network is trained using the backpropagation algorithm through supervised learning. The weights and biases of the neural network are adjusted to minimize the error between the predicted and actual values and optimize the neural network model.
[0027] The parameters of the network are optimized by cross-validation to ensure that the model can accurately predict the location and extent of damage to the dam and avoid overfitting.
[0028] According to claim 1, the damage prediction system for a concrete arch dam under impact load is characterized in that the RBF neural network uses radial basis functions as activation functions and consists of an input layer, a hidden layer and an output layer, wherein the number of nodes in the input layer is 3, the number of nodes in the hidden layer is 6, and the number of nodes in the output layer is 1.
[0029] According to claim 1, a damage prediction system for a concrete arch dam under impact load is characterized in that the step of using an RBF neural network to predict the location and extent of damage to the dam body is as follows:
[0030] The damage function module calculates and outputs the damage index of the dam body. The damage index represents the degree of damage to the dam body under different loads, denoted as D=(D1,D2,...,D...). n ), where D i This represents the numerical value of each damage index.
[0031] The acquired damage index D is used as input data to the already trained RBF neural network model. The input data is x = (x1, x2, ..., x...). n As the input layer node, the output h of the hidden layer node is... i (x) is obtained by calculating the radial basis functions, and its expression is: Where x = (x1, x2, ..., x n ) is the damage index input vector, c i Let h be the center vector of hidden layer node i, σ be the width parameter of the Gaussian function, and h be the center vector of the hidden layer node i. i (x) is the output of hidden layer node i.
[0032] The trained RBF neural network is based on the output h of the hidden layer nodes. i (x) Perform weighted summation and output the damage prediction result Y, predicting the damage extent or location: Among them, w i h represents the weights from hidden layer node i to the output layer. i (x) represents the output of the hidden layer node, b is the bias term, and Y is the final damage prediction result.
[0033] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art. Of course, any product implementing the present invention does not necessarily need to achieve all of the following advantages at the same time:
[0034] This invention combines flexibility curvature variation and damage function modules to calculate the damage degree of dam bodies, exhibiting low error in damage degree prediction, especially under high impact loads, effectively reflecting the damage evolution of the dam body. Compared with existing technologies, this system has significantly smaller prediction errors and can more accurately assess dam damage under complex loads. By incorporating the RBF neural network model, this system can rapidly perform calculations and predictions after finite element analysis and damage data acquisition, with short computation time and real-time output of damage prediction results. This makes the system more computationally efficient in real-time monitoring, meeting the needs of rapid response and accurate prediction for damage assessment in practical engineering, and significantly improving the reliability and practicality of dam health monitoring.
[0035] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description
[0036] The accompanying drawings described below are merely some embodiments. Those skilled in the art can obtain other drawings based on these drawings without any creative effort. In the drawings:
[0037] Figure 1 is a block diagram of the damage prediction system for a concrete arch dam under impact load.
[0038] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation
[0039] The invention will now be described in further detail with reference to the accompanying drawings.
[0040] Please refer to Figure 1. In this embodiment, a damage prediction system for concrete arch dams under impact load is provided, including a finite element analysis module for establishing a three-dimensional finite element model, simulating the response of concrete arch dams under water flow impact load, and analyzing and extracting stress, strain and displacement data of the dam body.
[0041] The flexibility calculation module calculates the flexibility matrix of the dam body based on the data provided by the finite element analysis module, analyzes the changes in the flexibility curvature of the dam body, and extracts the sensitivity of structural deformation.
[0042] The damage function module calculates the degree of damage to the dam body based on the change in flexibility curvature using a preset damage function, and generates damage indicators. The damage function is a linear relationship based on the change in flexibility curvature, and the degree of damage is proportional to the change in flexibility curvature.
[0043] The RBF neural network module is used to build the RBF neural network module. It takes the damage index output by the damage function module as input to train the neural network model. After training, the RBF neural network is used to predict the location and extent of damage to the dam body.
[0044] The damage prediction result output module is used to output the predicted damage location and damage degree.
[0045] In this embodiment, the steps for establishing a three-dimensional finite element model to simulate the response of a concrete arch dam under water flow impact load, and for analyzing and extracting the stress, strain, and displacement data of the dam body are as follows:
[0046] Based on the design drawings or on-site measurement data of the concrete arch dam, the geometric dimensions and shape of the dam body are obtained. A three-dimensional geometric model of the concrete arch dam is generated using three-dimensional modeling software. Then, corresponding material properties are defined for different components of the dam body and corresponding material models are assigned to different parts. The material properties include elastic modulus, Poisson's ratio, and yield strength.
[0047] The three-dimensional geometric model was discretized into a finite element mesh using three-dimensional tetrahedral elements. Then, fixed boundary conditions were applied to the bottom and sides of the dam, and contact conditions between the dam and the foundation, and between the dam and the water flow, were applied according to actual conditions.
[0048] Water flow impact loads are applied using dynamic pressure and shock wave models. The time history, amplitude, and location of the load are defined and applied to the contact surfaces of the dam body. The applied water flow impact load includes the kinetic energy of the water flow and the shock wave model. The amplitude, time history, and location of the load are defined according to the actual working conditions. Finally, after the dynamic analysis calculation is completed, the stress field, strain field, and displacement field of the dam body are extracted from the finite element analysis results, with a focus on the stress, strain, and displacement changes in key parts of the dam body. The applied water flow impact load includes the kinetic energy of the water flow and the shock wave model, as well as the amplitude, time history, and location of the load. Key parts of the dam body include the dam crest, arch ribs, dam foundation, and joints.
[0049] In this embodiment, the steps of calculating the flexibility matrix of the dam body, analyzing the change in flexibility curvature of the dam body, and extracting the sensitivity of structural deformation based on the data provided by the finite element analysis module are as follows:
[0050] The finite element analysis module is used to obtain stress, displacement and deformation data of the dam body, and the nodal displacement data and stiffness matrix of the dam body are extracted based on the analysis results.
[0051] Using the stiffness matrix K from the finite element analysis results, the flexibility matrix F of the dam body is calculated according to the following formula: F = K -1 Where K is the stiffness matrix and F is the flexibility matrix, the flexibility values of each part of the dam body are obtained through this calculation;
[0052] Based on the calculated flexibility matrix F, the variation of the flexibility curvature of the dam body under different working conditions is analyzed. Calculated using the following formula: Where ΔF represents the change in the compliance matrix under different impact loads, and F0 represents the initial compliance matrix under unloaded conditions. The change in flexibility curvature reflects the deformation sensitivity of the dam body under load.
[0053] By analyzing the changes in flexibility curvature, the deformation sensitivity of the dam body under impact load is extracted. The larger the sensitivity value, the stronger the response of a certain part of the dam body to load changes, and vice versa.
[0054] In this embodiment, the damage function module calculates the degree of damage to the dam body and generates damage indices based on the change in flexibility curvature using a preset damage function. The steps are as follows:
[0055] As the impact force of water flow increases, the local deformation and stress response of the dam body often increase nonlinearly. Especially under large impacts, the dam body undergoes significant deformation or failure. At this point, the relationship between the degree of damage and the impact load is no longer a simple linear one. Therefore, an exponential damage function is adopted. To determine the degree of damage, the first step is to calculate the change in flexibility curvature. The damage degree D of the dam body is obtained by inputting it into a preset damage function, where D is the damage degree. Let α and β be the change in flexibility curvature, and let α and β be empirical constants, where α = 0.1-1 and β = 1-2, with the optimal values being α = 0.5 and β = 1.2.
[0056] The damage function is a linear or polynomial function based on the change in flexibility curvature, and the degree of damage D varies with the change in flexibility curvature. It increases as the value increases.
[0057] In this embodiment, the steps for establishing an RBF neural network module, which takes the damage index output by the damage function module as input, and training the neural network model are as follows:
[0058] Based on the damage index, the structure of the RBF neural network is selected, the number of hidden layer nodes is determined, and the parameters of the RBF neural network are initialized, with the radial basis function type set to Gaussian function;
[0059] The damage indicators output from the damage function module are paired with the actual measured damage data (including damage location and damage extent) to construct a dataset for training the neural network. This ensures the dataset includes a complete pairing of damage indicators and actual damage conditions. The training dataset includes both damage indicators and actual measured damage data, which includes the damage location and extent of damage to the dam body under different working conditions. The actual measured damage data is collected in real-time by installing stress sensors, displacement sensors, accelerometers, and other monitoring devices at the dam crest, arch ribs, dam foundation, and joints. By monitoring the stress and strain at different parts of the dam body, the response of the dam body under different loads can be identified. For example, localized stress concentration usually indicates that the material in that area may have undergone plastic deformation or damage.
[0060] The training dataset is input into the RBF neural network. The neural network is trained using the backpropagation algorithm through supervised learning. The weights and biases of the neural network are adjusted to minimize the error between the predicted and actual values and optimize the neural network model.
[0061] The parameters of the network are optimized by cross-validation to ensure that the model can accurately predict the location and extent of damage to the dam and avoid overfitting.
[0062] In this embodiment, the RBF neural network uses radial basis functions as activation functions and consists of an input layer, a hidden layer, and an output layer. The input layer has 3 nodes, the hidden layer has 6 nodes, and the output layer has 1 node.
[0063] In this embodiment,
[0064] The damage function module calculates and outputs the damage index of the dam body. The damage index represents the degree of damage to the dam body under different loads, denoted as D=(D1,D2,...,D...). n ), where D i This represents the numerical value of each damage indicator;
[0065] The acquired damage index D is used as input data to the already trained RBF neural network model. The input data is x = (x1, x2, ..., x...). n As the input layer node, the output h of the hidden layer node is... i (x) is obtained by calculating the radial basis functions, and its expression is: Where x = (x1, x2, ..., x n ) is the damage index input vector, ci Let h be the center vector of hidden layer node i, σ be the width parameter of the Gaussian function, and h be the center vector of the hidden layer node i. i (x) represents the output of hidden layer node i;
[0066] The trained RBF neural network is based on the output h of the hidden layer nodes. i (x) Perform weighted summation and output the damage prediction result Y, predicting the damage extent or location: Among them, w i h represents the weights from hidden layer node i to the output layer. i (z) represents the output of the hidden layer node, b is the bias term, and Y is the final damage prediction result. The damage indicators include the dam's flexibility curvature change, stress-strain data, etc., used to characterize the dam's stress state and damage degree. The damage prediction result Y includes the damage location, damage type (such as cracks, deformation, etc.), and damage degree (such as mild, moderate, severe) of the dam.
[0067] To illustrate the beneficial effects of the present invention, a comparative experiment was conducted between the present invention and existing technical solutions.
[0068] Comparative Example 1: Dam damage assessment method based on traditional finite element analysis
[0069] This approach typically employs finite element analysis (FEA) to simulate the dam's response under different loads, assessing damage by acquiring stress, strain, and displacement data. A simplified linear damage function is usually used for damage calculation, with simple corrections made based on laboratory data. This approach primarily estimates the degree of damage to the dam through the calculation of conventional stiffness and flexibility matrices.
[0070] Comparative Example 2: Dam Damage Assessment Method Based on Structural Health Monitoring (SHM) System
[0071] By installing sensors (such as stress sensors, displacement sensors, accelerometers, etc.) at key parts of the dam body, the stress, displacement, vibration and other data of the dam body are monitored in real time, and the monitoring data is combined with a simplified damage model for damage assessment.
[0072] In damage prediction, relevant damage indicators were extracted from actual monitoring data and finite element analysis. These indicators included changes in flexibility curvature, displacement, stress, and strain. Specifically, stress sensors, displacement sensors, and accelerometers were deployed at key parts of the dam body (such as the dam crest, arch ribs, foundation, and joints) to monitor the dam's response in real time. Stress sensors monitored stress changes at various points on the dam body; displacement sensors primarily monitored displacement changes, especially at critical locations where cracks or deformations might occur; and accelerometers measured the dam's vibration response, helping to further assess its deformation behavior under different loads. Data was collected hourly for three months, continuously collecting data through sensors to record the dam's deformation under various loads, obtaining a large amount of real-world data for training and testing the neural network model.
[0073] Finite Element Analysis Data: During the finite element model construction process, a three-dimensional finite element model was constructed based on the dam design drawings and on-site measurement data. The stress field, strain field, and displacement field of the dam were extracted through finite element analysis. Under different load conditions, water flow impact loads of varying intensities were applied, and dynamic analysis was performed to obtain the flexibility matrix and flexibility curvature changes of the dam under different conditions. To ensure the reliability of the neural network model training, a training dataset containing 1000 samples was designed. Each sample corresponds to one damage prediction, including damage indicators calculated from changes in flexibility curvature and sensor data such as stress, displacement, and strain. In addition, each sample also includes actual damage data, i.e., the damage location and degree of damage at key parts of the dam. This data was calibrated using on-site sensor data and laboratory test results. Damage locations include the dam crest, arch ribs, dam foundation, and joints, and damage degrees are categorized as mild, moderate, and severe.
[0074] Experimental Design: To compare the effects of this invention and existing technical solutions, the following experimental conditions were set: the Foziling Reservoir arch dam was used as the finite element analysis model, and water flow impact load was used as the load type, with an amplitude range of 500 N / m. 2 Up to 2000 N / m 2 The time history of the load was described using a shock wave model. Loads were applied at the dam crest, arch ribs, dam foundation, and joints to simulate the dam's response under different conditions. The simulated conditions included normal loads, extreme water flow impact loads, and fatigue under sustained loads, to comprehensively evaluate the dam's performance and damage under various load conditions.
[0075] This invention uses 1000 samples for damage prediction, with 95% of the samples accurately predicting the damage location of the dam. This high accuracy is mainly attributed to the damage prediction method based on flexibility curvature changes and RBF neural networks used in this scheme. Compared with traditional schemes, existing scheme 1 has lower prediction accuracy under complex working conditions, approximately 75%, while existing scheme 2 has even lower accuracy, with only 60% of the samples accurately predicting the damage location. The high accuracy stems from the precise capture of the deformation sensitivity of key parts of the dam through flexibility curvature changes.
[0076] This invention achieves a damage degree prediction error of less than 5%, particularly effective in predicting dam damage under high impact loads. This low error is attributed to the combination of flexibility curvature variation and RBF neural network in this scheme. The flexibility curvature variation Δφ (Delta / varphiΔφ) accurately reflects the dam's deformation sensitivity under different loads, enabling the model to accurately capture subtle changes in damage evolution. In contrast, the existing scheme 1 has a damage degree error of 10-15%, failing to accurately reflect damage evolution under complex loads, especially under dynamic loads. The existing scheme 2 has a larger error, typically exceeding 20%, especially under dynamic loads, leading to inaccurate damage degree prediction. The key to low error lies in the sensitivity analysis of flexibility curvature variation to local dam damage, providing the RBF neural network with more accurate damage indicators, thereby optimizing the accuracy of damage prediction.
[0077] This invention is not limited to the embodiments described above. Anyone should understand that structural changes made under the guidance of this invention, and any technical solutions that are the same as or similar to this invention, fall within the protection scope of this invention. Technical aspects, shapes, and structures not described in detail in this invention are all publicly known technologies.
Claims
1. A damage prediction system for concrete arch dams under impact load, characterized in that, include: The finite element analysis module is used to build a three-dimensional finite element model, simulate the response of a concrete arch dam under water flow impact load, and analyze and extract stress, strain and displacement data of the dam body. The flexibility calculation module calculates the flexibility matrix of the dam body based on the data provided by the finite element analysis module, analyzes the changes in the flexibility curvature of the dam body, and extracts the sensitivity of structural deformation. The damage function module calculates the degree of damage to the dam body and generates damage indicators based on the change in flexibility curvature using a preset damage function. The RBF neural network module is used to train the neural network model by taking the damage index output by the damage function module as input. After training, the RBF neural network is used to predict the location and extent of damage to the dam body. The damage prediction result output module is used to output the predicted damage location and damage degree.
2. The damage prediction system for a concrete arch dam under impact load according to claim 1, characterized in that, The steps for establishing a three-dimensional finite element model to simulate the response of a concrete arch dam under water flow impact load, and to analyze and extract the stress, strain, and displacement data of the dam body are as follows: Based on the design drawings or on-site measurement data of the concrete arch dam, the geometric dimensions and shape of the dam body are obtained. A three-dimensional geometric model of the concrete arch dam is generated using three-dimensional modeling software. Then, corresponding material properties are defined for different components of the dam body and corresponding material models are assigned to different parts. The material properties include elastic modulus, Poisson's ratio, and yield strength. The three-dimensional geometric model was discretized into a finite element mesh using three-dimensional tetrahedral elements. Then, fixed boundary conditions were applied to the bottom and sides of the dam, and contact conditions between the dam and the foundation, and between the dam and the water flow, were applied according to actual conditions. Water flow impact loads are applied using dynamic pressure and shock wave models. The time history, amplitude, and location of the load are defined and applied to the contact surface of the dam. The applied water flow impact load includes the kinetic energy of the water flow and the shock wave model. The amplitude, time history, and location of the load are defined according to the actual working conditions. Finally, after the dynamic analysis calculation is completed, the stress field, strain field, and displacement field of the dam are extracted from the finite element analysis results, with a focus on the stress, strain, and displacement changes in key parts of the dam.
3. The damage prediction system for a concrete arch dam under impact load according to claim 1, characterized in that, Based on the data provided by the finite element analysis module, the following steps are taken to calculate the flexibility matrix of the dam body, analyze the changes in the flexibility curvature of the dam body, and extract the sensitivity to structural deformation: The finite element analysis module is used to obtain stress, displacement and deformation data of the dam body, and the nodal displacement data and stiffness matrix of the dam body are extracted based on the analysis results. Using the stiffness matrix K from the finite element analysis results, the flexibility matrix F of the dam body is calculated according to the following formula: F = K -1 Where K is the stiffness matrix and F is the flexibility matrix, the flexibility values of each part of the dam body are obtained through this calculation; Based on the calculated flexibility matrix F, the variation of the flexibility curvature of the dam body under different working conditions is analyzed. Calculated using the following formula: Where ΔF represents the change in the compliance matrix under different impact loads, and F0 represents the initial compliance matrix under unloaded conditions. The change in flexibility curvature reflects the deformation sensitivity of the dam body under load. By analyzing the changes in flexibility curvature, the deformation sensitivity of the dam body under impact load is extracted. The larger the sensitivity value, the stronger the response of a certain part of the dam body to load changes, and vice versa.
4. The damage prediction system for a concrete arch dam under impact load according to claim 1, characterized in that, The damage function module calculates the degree of damage to the dam body and generates damage indices based on changes in flexibility curvature using a preset damage function. The steps are as follows: As the impact force of water flow increases, the local deformation and stress response of the dam body often increase nonlinearly. Especially under large impacts, the dam body undergoes significant deformation or failure. At this point, the relationship between the degree of damage and the impact load is no longer a simple linear one. Therefore, an exponential damage function is adopted. To determine the degree of damage, the first step is to calculate the change in flexibility curvature. The damage degree D of the dam body is obtained by inputting it into a preset damage function, where D is the damage degree. Let α be the change in flexibility curvature, and β be empirical constants.
5. The damage prediction system for a concrete arch dam under impact load according to claim 1, characterized in that, The steps for establishing an RBF neural network module, which takes the damage index output by the damage function module as input, and training the neural network model are as follows: Based on the damage index, the structure of the RBF neural network is selected, the number of hidden layer nodes is determined, and the parameters of the RBF neural network are initialized, with the radial basis function type set to Gaussian function; The damage indicators output from the damage function module are paired with the actual measured damage data to construct a dataset for training the neural network, ensuring that the dataset includes a complete pairing of damage indicators and actual damage conditions. The training dataset is input into the RBF neural network. The neural network is trained using the backpropagation algorithm through supervised learning. The weights and biases of the neural network are adjusted to minimize the error between the predicted and actual values and optimize the neural network model. The parameters of the network are optimized by cross-validation to ensure that the model can accurately predict the location and extent of damage to the dam and avoid overfitting.
6. The damage prediction system for a concrete arch dam under impact load according to claim 1, characterized in that, The RBF neural network uses radial basis functions as activation functions and consists of an input layer, a hidden layer, and an output layer. The input layer has 3 nodes, the hidden layer has 6 nodes, and the output layer has 1 node.
7. The damage prediction system for a concrete arch dam under impact load according to claim 1, characterized in that, The steps for predicting the location and extent of damage to a dam using an RBF neural network are as follows: The damage function module calculates and outputs the damage index of the dam body. The damage index represents the degree of damage to the dam body under different loads, denoted as D=(D1,D2,...,D...). n ), where D i This represents the numerical value of each damage index. The acquired damage index D is used as input data to the already trained RBF neural network model. The input data is x = (x1, x2, ..., x...). n As the input layer node, the output h of the hidden layer node is... i (x) is obtained by calculating the radial basis functions, and its expression is: Where x = (x1, x2, ..., x n ) is the damage index input vector, c i Let h be the center vector of hidden layer node i, σ be the width parameter of the Gaussian function, and h be the center vector of the hidden layer node i. i (x) is the output of hidden layer node i. The trained RBF neural network is based on the output h of the hidden layer nodes. i (x) Perform weighted summation and output the damage prediction result Y, predicting the damage extent or location: Among them, w i h represents the weights from hidden layer node i to the output layer. i (x) represents the output of the hidden layer node, b is the bias term, and Y is the final damage prediction result.