Dam operation deformation prediction method and device and storage medium

Abstract

The invention relates to the technical field of dam health monitoring, in particular to a dam operation deformation prediction method, device and equipment and a computer storage medium. According to the invention, aiming at a structure monitoring mode of current point detection, one of core factors influencing arch dam deformation is solved in a targeted manner by combining a temperature field model. Therefore, a CNN-based deformation field reconstruction model is finally constructed based on data such as a temperature field and a water pressure field. For each predicted location, the parameters of the reconstructed model are universal and unified. The globally unified model can export high-precision and high-density deformation reconstruction data, and the overall deformation behavior of the arch dam can be analyzed more accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of dam health monitoring, and in particular to a dam operation deformation prediction method, device, equipment and computer storage medium. Background Art

[0002] The characterization of the deformation behavior of arch dams is the core of arch dam health monitoring. The current deformation behavior model establishes model parameters for different measuring points independently, completely ignoring the correlation between different measuring points. For an over-static structure such as an arch dam, its integrity is more important. Therefore, it is necessary to establish a unified spatial deformation field behavior model that considers multiple measuring points and the integrity of the arch dam.

[0003] The unified spatial deformation field behavior model not only poses challenges to the analysis method, but also has strict requirements on the monitoring data. The current structural monitoring mode is still mainly point observation, and the corresponding information is obtained through sensors placed at specific positions on the dam body. Therefore, for the existing statistical models, including the neural network machine learning model that has developed rapidly in recent years, it is still impossible to quantitatively construct a unified spatial deformation field model due to the current single-point structural monitoring mode, and it is impossible to conduct a more accurate analysis of the overall deformation behavior of the arch dam. Summary of the invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the problem of low accuracy in the analysis of the overall deformation behavior of the arch dam in the prior art.

[0005] In order to solve the above technical problems, the present invention provides a method for predicting dam operation deformation, comprising:

[0006] The water pressure field, temperature field and foundation constraint field are used as the input of the model, and the gradient of deformation along the height direction is used as the output of the model to establish a CNN model;

[0007] Training the CNN model using a pre-constructed training set until the mean square error between the measured gradient and the reconstructed gradient reaches a preset critical value;

[0008] Input the real-time monitoring data into the trained CNN model to obtain the target reconstruction gradient field;

[0009] The dam deformation field is calculated based on the target reconstructed gradient field.

[0010] Preferably, establishing a CNN model comprises:

[0011] An edge filling operation is introduced in the input process of the CNN model.

[0012] Preferably, the establishing of the CNN model further comprises:

[0013] After each convolutional layer in the CNN model, a pooling layer is set.

[0014] Preferably, the establishing of the CNN model further comprises:

[0015] The fully connected layers in the CNN model are removed.

[0016] Preferably, the calculation process of the water pressure field includes:

[0017] Reservoir water level is converted to a hydrostatic pressure field based on density and height.

[0018] Preferably, the calculation process of the temperature field includes:

[0019] The temperature sensor data is interpolated along the elevation to obtain the upstream temperature field;

[0020] According to the thermal stress generated by the hyperstatic structure of the arch dam and the temperature difference between the operation period and the annual average temperature, the difference between the reconstructed temperature field and the annual average temperature is calculated, and the multi-day average temperature field is calculated considering the delayed effect of the temperature field on deformation.

[0021] Preferably, the calculation process of the basic constraint field includes:

[0022] The boundary constraint strength is calculated based on the elevation corresponding to the position of the dam body close to the boundary, the horizontal distance from the position to the dam arch crown beam, the elevation of the bottom of the arch dam and the elevation of the top of the arch dam.

[0023] The present invention also provides a dam operation deformation prediction device, comprising:

[0024] The model building module is used to take the water pressure field, temperature field and foundation constraint field as the input of the model, and the gradient of the deformation along the height direction as the output of the model to establish the CNN model;

[0025] A model training module, used to train the CNN model using a pre-constructed training set until the mean square error between the measured gradient and the reconstructed gradient reaches a preset critical value;

[0026] The target gradient prediction module is used to input the real-time monitoring data into the trained CNN model to obtain the target reconstruction gradient field;

[0027] The deformation prediction module is used to calculate the dam deformation field according to the target reconstruction gradient field.

[0028] The present invention also provides a dam operation deformation prediction device, comprising:

[0029] Memory for storing computer programs;

[0030] A processor is used to implement the steps of the above-mentioned dam operation deformation prediction method when executing the computer program.

[0031] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned dam operation deformation prediction method are implemented.

[0032] The above technical solution of the present invention has the following advantages compared with the prior art:

[0033] The dam operation deformation prediction method described in the present invention proposes a deformation field reconstruction model based on CNN, which uses a spatial unified model to describe the overall stress deformation behavior of the arch dam. In view of the current point detection structural monitoring mode, one of the core factors affecting the deformation of the arch dam is targetedly solved in combination with the temperature field model. Therefore, a deformation field reconstruction model based on CNN is finally constructed based on temperature field, water pressure field and other data. For each predicted position, the parameters of the reconstruction model are universal and unified. This globally unified model can derive high-precision and high-density deformation reconstruction data, and conduct a more accurate analysis of the overall deformation behavior of the arch dam. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0035] Figure 1 It is a flow chart of the implementation of a dam operation deformation prediction method provided by the present invention;

[0036] Figure 2 It is a schematic diagram of maximum pooling and structural continuity;

[0037] Figure 3 This is a schematic diagram of the CNN structure;

[0038] Figure 4 This is a schematic diagram of output field processing;

[0039] Figure 5 is a schematic diagram of the basic constraint field;

[0040] Figure 6 This is the vertical line layout of Xiaowan;

[0041] Figure 7 This is the layout diagram of the downstream surface temperature sensor;

[0042] Figure 8 This is the layout diagram of the upstream surface temperature sensor. DETAILED DESCRIPTION

[0043] The core of the present invention is to provide a dam operation deformation prediction method, device, equipment and computer storage medium, which effectively improves the accuracy of the overall deformation behavior analysis of the arch dam.

[0044] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of 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.

[0045] Please refer to Figure 1 , Figure 1 This is a flow chart of the implementation of a dam operation deformation prediction method provided by the present invention; the specific operation steps are as follows:

[0046] S101: Taking the water pressure field, temperature field and foundation constraint field as the input of the model, taking the gradient of the deformation along the height direction as the output of the model, and establishing a CNN model;

[0047] S102: training the CNN model using a pre-constructed training set until the mean square error between the measured gradient and the reconstructed gradient reaches a preset critical value;

[0048] S103: Input the real-time monitoring data into the trained CNN model to obtain the target reconstruction gradient field;

[0049] S104: Calculate the dam deformation field according to the target reconstructed gradient field.

[0050] Based on the above embodiments, this embodiment describes the CNN model in detail:

[0051] In the CNN model, operators are used After processing the input field, the shape of its output will be reduced. Therefore, in order to keep the input and output field dimensions the same, an edge padding operation can be introduced in CNN, where the input field will be preprocessed (edge ​​padding 0, constant, or padding with edge values) and expanded to a preset size. Therefore, once the convolution kernel size of the CNN model is determined, the parameter size is constant even if the size of the input feature changes. On the one hand, this makes the scale of training parameters much smaller than traditional methods. Therefore, for sparse monitoring data, this way of reducing the dimension of parameters will make the spatial generalization ability of the model better. On the other hand, the shared global parameters including the weight matrix and the bias matrix imply an important physical law, that is, the difference in structural responses at different positions is only related to the input features.

[0052] For the deformation behavior model of multiple measuring points of the structure, we know that there is a strong correlation between different locations of the structure, which is manifested in the strong continuity of the deformation of a specific measuring point and the surrounding measuring points. How to describe this strong continuity through data-based statistics or neural network models has always been a very difficult problem. Some scholars have analyzed this correlation from a qualitative level through correlation coefficients, clustering algorithms, etc.

[0053] The pooling layer is a method of compressing dimensions. Its main purpose is to compress the parameter scale and reduce the possibility of overfitting. For the CNN model, its main purpose is to extract the main features of the image and perform feature dimensionality reduction. For a picture, it contains a lot of redundant information, but this information is not very meaningful for the problem to be solved by traditional CNN, so the redundant information can be removed through the pooling layer.

[0054] For structural continuity characteristics, especially for arch dam as a spatial shell structure, in its early small deformation stage, the local deformation of the dam body along the elevation direction is a convex function. The convex function is defined as follows:

[0055] f(q 1 x 1 +q 2 x 2 )≤q 1 f(x 1 )+q 2 f(x 2 ) Formula 1

[0056] In the above formula, q 1 and q 2 is any positive number that sums to 1.

[0057] The above formula implies a very important rule as follows:

[0058]

[0059] For any x 1 ≤x 0 ≤x 2 Established.

[0060] Therefore, Figure 2 As shown, the relationship between maximum pooling and structural continuity is:

[0061] The figure above shows the effect of a maximum pooling (pooling size is 1×3) of the CNN model. We can see that for the discontinuous intermediate variable g' 1 ,g' 2 ,...,g' 6, after one maximum pooling, the boundary is constrained, that is, the output result becomes a continuous variable. In our research, we believe that there are two improvements. First, after processing the maximum pooling convolutional layers of different scales, the gradient field of the dam body becomes a continuous gradient field. Secondly, for the water pressure-free area at the top of the dam body, due to the introduction of continuity, the CNN model of the deformation field at the top has a certain processing capability.

[0062]

[0063] Typically, CNN models are used to solve classification problems, such as the application of CNN in face recognition. Traditional CNN methods set up a fully connected layer to fuse the derived features and output the final classification result. In our model, the fully connected layer is removed because the output result of the present invention is a two-dimensional field rather than a classification result. Therefore, the final CNN model structure is as follows: Figure 3 shown.

[0064] Based on the above embodiments, this embodiment describes the data preprocessing methods for different input features:

[0065] In traditional statistical or machine learning methods, the effect of hydrostatic pressure on deformation is simulated by a polynomial function, as shown in Equation 3. The reservoir water level is the same at each monitoring point, but the reservoir water level has different effects on the locations of different monitoring points. Therefore, it is impossible for these methods to derive a universal model for each monitoring point.

[0066]

[0067] In the above formula, h(t) is the upstream reservoir water level, a i is the regression coefficient, K is the power exponent, which is 4 for arch dams and 3 for gravity dams.

[0068] In the present invention, the reservoir water level is converted into a hydrostatic pressure field based on density and height. Hydrostatic pressure varies with location, so a general model can be established. Hydrostatic pressure is calculated using Equation 4.

[0069]

[0070] In the above formula, P ij and Al ij is the water pressure value and elevation of the target location (i, j), and ρ is the density of water.

[0071] One of the most important loads during the operation of an arch dam is temperature stress. The present invention obtains an accurate reconstructed temperature field that takes into account spatial heterogeneity and the actual operating temperature, and uses this as input to the model for temperature field input factors:

[0072] For the temperature field on the upstream surface of the arch dam, the boundary conditions are simple because it is in contact with the reservoir water. Therefore, the upstream temperature field remains unchanged in the horizontal direction and is only related to the elevation. Therefore, the temperature sensor data can be used for interpolation along the elevation.

[0073] More importantly, the further processing of the temperature field must match the actual physical laws, including two steps: First, the hyperstatic structure of the arch dam and the temperature difference between the operation period and the annual average temperature produce thermal stress. From this, the difference between the reconstructed temperature field and the annual average temperature is calculated. Secondly, the delayed effect of the temperature field on the deformation is considered. The multi-day average temperature field is calculated and finally input into the CNN model.

[0074] The boundary constraints of arch dams are crucial to the stability of arch dams, but they have not been covered by statistical methods and machine learning methods before. According to the structural characteristics of arch dams, the closer the dam body is to the boundary, the stronger the constraint. Therefore, Equation 5 using position coordinates is used to describe the boundary constraints as follows:

[0075]

[0076] In the above formula, C ij is the boundary constraint strength, Al ij is the elevation corresponding to position (i, j), Al bot and Al top is the bottom and top elevation of the arch dam. ij is the horizontal distance from position (i, j) to the dam crown beam. In addition, the boundary area constraints are filled with constants. The final basic constraint field and the symbols in the formula are as follows Figure 4 As shown:

[0077] Based on the above embodiments, this embodiment describes the output deformation field:

[0078] Unlike traditional statistical or machine learning methods, the output variable of this model is not deformation, but the gradient of deformation along the height direction. Due to the weakness of machine learning in physical laws and prior knowledge, the CNN model cannot directly handle this type of problem, that is, due to the continuity of the dam body, the deformation of the dam crest is not zero, and CNN cannot directly obtain this relationship, and will output zero under a given input value (water pressure is 0).

[0079] Therefore, the gradient of the deformation with height is calculated as the output of the neural network. However, not all deformation gradients at different positions are available. Figure 5 The single vertical line measuring area shown can only calculate the total gradient within the range of the vertical line. Therefore, the CNN model will output all gradients, but the sum of the range will be used to train the neural network parameters.

[0080]

[0081] Based on the above embodiments, the present invention describes the training process of the CNN model:

[0082] During the training process, all input features are fully sampled matrices, and the output domain of the CNN model is also a fully sampled matrix. Only the measured gradient of the vertical line is a downsampled matrix, and only the corresponding position has relevant data. We train the parameters of the CNN model until the mean square error between the measured gradient and the reconstructed gradient reaches the critical value ε, and the training is completed.

[0083] Based on the above embodiments, this embodiment uses the Xiaowan Arch Dam on the Lancang River as an analysis example and verification case. The dam is 294.5m high, 16m wide at the top, and the total reservoir capacity is expected to be 15 billion cubic meters. The reservoir has a multi-year regulation capacity. There are 43 pre-installed vertical line measurement points on the dam body, divided into 7 groups, located in different dam sections. These vertical lines are used to measure the relative deformation of different elevation ranges, such as Figure 6 The total area of ​​the downstream surface of Xiaowan Arch Dam is about several hundred thousand square meters. Obviously, the vertical measurement points are very sparse. Figure 7 As shown in Figure 2, only eight temperature sensors were pre-installed on the downstream surface of the dam, and the new reconstruction method was used to expand the entire temperature field, as shown in Figure 2. Figure 8 ,There are several temperature sensors at different heights on the upstream surface of the dam.

[0084] The total number of CNN model training and testing iterations is 3,000. In each iteration, all input samples are randomly divided into 8 small sample batches to train the model faster. After 2,500 iterations, the training error and test error reach convergence. Using a GTX1080Ti graphics card, the total training time is about three hours.

[0085] For the single-point inversion prediction accuracy verification part, since the current multi-point models, such as linear interpolation models and Kriging interpolation methods, can only interpolate spatial fields based on monitoring data and do not have prediction capabilities, we use the traditional HST model and CNN inversion model results for comparative analysis in this part. All sensor data are used as training sets, the training period is from January 2015 to August 2017, and the verification period is from August to December 2017. The CNN deformation field reconstruction model CNN-T and the HST regression model are trained respectively.

[0086] We use mean square error to evaluate the accuracy of the model. We can conclude that the accuracy of the multi-measurement point CNN inversion model and the FEM model is worse than that of the single-measurement point HST model. The average mean square error of the CNN model is 2.3mm, the average mean square error of the HST model is 0.8mm, and the average mean square error of the FEM model is 2.4mm. This is because the HST model uses different model parameters for different measurement points for training, so it is understandable that its model accuracy is better.

[0087] However, compared with the traditional statistical model that is trained separately for each measuring point, the advantages of CNN inversion and FEM model are that it can take into account the correlation and unified rules between measuring points, which helps us analyze the overall safety of high arch dams. In addition, more importantly, the CNN inversion model can obtain the deformation field of high arch dams through a unified model, that is, we can obtain the deformation prediction value of this accuracy level (i.e. 2.3mm) through the CNN inversion model at the location where no monitoring sensors are deployed. Traditional single-measurement point models such as HST cannot achieve this goal. Next, we evaluate the spatial expansion capability of the CNN inversion model.

[0088] Currently, there are two main types of models that can obtain the deformation field of arch dams. The first type is the interpolation model that only relies on monitoring data and does not consider the physical mechanism, such as cubic spline interpolation, Kriging method interpolation, etc. The second type is the finite element model that considers both monitoring data and physical mechanisms. Therefore, this section evaluates and discusses the spatial expansion capabilities of interpolation models, finite element models, and CNN deformation field reconstruction models.

[0089] We evaluate the spatial expansion capability from two perspectives. The first perspective selects different sensor data for verification. In this section, A15-02 and A22-01 are selected as verification data sets to compare the interpolation and extrapolation capabilities of the interpolation model, CNN inversion model, and finite element model.

[0090] According to the comparison of the predicted value time series of the corresponding measuring points of different models' spatial prediction and the measured results, we can see that the prediction sequence accuracy of different models for the A15-PL-02 measuring point, that is, the interpolation prediction, is relatively high. Further, for the extrapolation prediction (A22-PL-01), we can see that the interpolation method does not consider the physical mechanism of the dam, so its interpolation result has a very large deviation, while for the CNN deformation field reconstruction model, we can see that its accuracy has been significantly improved.

[0091] For the second evaluation angle of spatial expansion capability, we evaluate the reconstruction accuracy of the overall spatial field. Before this task, since we cannot obtain the monitoring results of the spatial deformation field, this section selects the deformation field of finite element inversion as the benchmark model, so we first need to evaluate the error of the benchmark finite element model.

[0092] We use the finite element model parameters inverted in the previous article to extract the deformation field of the downstream surface of the arch dam. We further obtain the relative displacement corresponding to different vertical measuring points through the deformation field. Finally, we evaluate the accuracy of the finite element model through the finite element model and the measured data as the accuracy threshold. According to the mean square error results of the deformation at different measuring points, we believe that the average mean square error of the finite element benchmark model for the reconstruction of the global deformation field is about 56.76mm, and the average absolute error is about 5.66mm. Therefore, we will compare the different deformation field models in the previous article with the finite element model to determine the reconstruction accuracy of the overall deformation field.

[0093] Next, we derive the deformation fields of the aforementioned three models (CNN-T, CNN-A15-02, and CNN-A22-01) and define the spatial mean absolute error as the accuracy measure of the finite element deformation field and the CNN inversion model deformation field, as shown in the following formula.

[0094]

[0095] In the above formula, N is the total number of nodes. is the node displacement of the CNN inversion model, is the nodal displacement of the finite element model.

[0096] According to the MAEs between the deformation fields of different models, we can see that the difference between the deformation fields of the CNN reconstruction model and the finite element model obtained by using different measurement point training sets at different times is not large (the average value is less than 3 times the error between the finite element model and the measured value). In addition, the difference between the deformation fields of different CNN reconstruction models is very small, which indirectly demonstrates the robustness and robustness of the CNN reconstruction model.

[0097] The embodiment of the present invention also provides a dam operation deformation prediction device; the specific device may include:

[0098] The model building module is used to take the water pressure field, temperature field and foundation constraint field as the input of the model, and the gradient of the deformation along the height direction as the output of the model to establish the CNN model;

[0099] A model training module, used to train the CNN model using a pre-constructed training set until the mean square error between the measured gradient and the reconstructed gradient reaches a preset critical value;

[0100] The target gradient prediction module is used to input the real-time monitoring data into the trained CNN model to obtain the target reconstruction gradient field;

[0101] The deformation prediction module is used to calculate the dam deformation field according to the target reconstruction gradient field.

[0102] The dam operation deformation prediction device of this embodiment is used to implement the aforementioned dam operation deformation prediction method. Therefore, the specific implementation method of the dam operation deformation prediction device can be seen in the embodiment part of the dam operation deformation prediction method mentioned above. For example, the model building module, the model training module, the target gradient prediction module, and the deformation prediction module are respectively used to implement steps S101, S102, S103, and S104 in the above-mentioned dam operation deformation prediction method. Therefore, its specific implementation method can refer to the description of the corresponding embodiments of each part, which will not be repeated here.

[0103] A specific embodiment of the present invention further provides a dam operation deformation prediction device, comprising: a memory for storing a computer program; and a processor for implementing the steps of the above-mentioned dam operation deformation prediction method when executing the computer program.

[0104] A specific embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned dam operation deformation prediction method are implemented.

[0105] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0106] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0109] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for predicting dam operation deformation, characterized in that: include: The water pressure field, temperature field and foundation constraint field are used as the input of the model, and the gradient of deformation along the height direction is used as the output of the model to establish a CNN model; Training the CNN model using a pre-constructed training set until the mean square error between the measured gradient and the reconstructed gradient reaches a preset critical value; Input the real-time monitoring data into the trained CNN model to obtain the target reconstruction gradient field; The dam deformation field is calculated based on the target reconstructed gradient field.

2. The dam operation deformation prediction method according to claim 1, characterized in that: The CNN model establishment includes: An edge filling operation is introduced in the input process of the CNN model.

3. The dam operation deformation prediction method according to claim 2, characterized in that: The CNN model establishment further includes: After each convolutional layer in the CNN model, a pooling layer is set.

4. The dam operation deformation prediction method according to claim 3 is characterized in that: The CNN model establishment further includes: The fully connected layers in the CNN model are removed.

5. The dam operation deformation prediction method according to claim 1, characterized in that: The calculation process of the water pressure field includes: Reservoir water level is converted to a hydrostatic pressure field based on density and height.

6. The dam operation deformation prediction method according to claim 1, characterized in that: The calculation process of the temperature field includes: The temperature sensor data is interpolated along the elevation to obtain the upstream temperature field; According to the thermal stress generated by the hyperstatic structure of the arch dam and the temperature difference between the operation period and the annual average temperature, the difference between the reconstructed temperature field and the annual average temperature is calculated, and the multi-day average temperature field is calculated considering the delayed effect of the temperature field on deformation.

7. The dam operation deformation prediction method according to claim 1, characterized in that: The calculation process of the basic constraint field includes: The boundary constraint strength is calculated based on the elevation corresponding to the position of the dam body close to the boundary, the horizontal distance from the position to the dam arch crown beam, the elevation of the bottom of the arch dam and the elevation of the top of the arch dam.

8. A dam operation deformation prediction device, characterized in that: include: The model building module is used to take the water pressure field, temperature field and foundation constraint field as the input of the model, and the gradient of the deformation along the height direction as the output of the model to establish the CNN model; A model training module, used to train the CNN model using a pre-constructed training set until the mean square error between the measured gradient and the reconstructed gradient reaches a preset critical value; The target gradient prediction module is used to input the real-time monitoring data into the trained CNN model to obtain the target reconstruction gradient field; The deformation prediction module is used to calculate the dam deformation field according to the target reconstruction gradient field.

9. A dam operation deformation prediction device, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of a dam operation deformation prediction method as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a dam operation deformation prediction method as described in any one of claims 1 to 7 are implemented.

Images