Foundation pit deformation prediction and control method and device based on graph neural network, computer equipment and storage medium

Abstract

The invention discloses a foundation pit deformation prediction and control method based on a graph neural network, and relates to the technical field of civil engineering, and the method comprises the steps: employing a space-time diagram attention network model, and employing a graph attention mechanism to capture spatial features between monitoring points around a steel support; inputting spatial features extracted by the time-space diagram attention network model into a gating circulation unit, extracting single representation of the whole diagram, and obtaining a monitoring point deformation prediction value; the method has the advantages that the space-time diagram attention network is used for analyzing and weighting the space relation among the multiple monitoring points, the importance of supporting axial force near all the monitoring points is distinguished, the model more accurately captures the spatial heterogeneity of soil deformation, and the deformation prediction accuracy is improved. Therefore, the accuracy of foundation pit deformation prediction is obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of civil engineering, and in particular to a method for predicting and controlling foundation pit deformation based on a graph neural network. Background Art

[0002] With the acceleration of urbanization, more and more buildings are being built near subway lines. The foundation pit projects adjacent to subway lines not only need to meet their own construction requirements, but also need to ensure the normal operation and structural safety of subway tunnels. Traditional steel support structures can usually only passively control soil deformation and are easily affected by axial force loss and temperature changes, making it difficult to meet strict deformation control requirements. The development of servo systems provides the possibility of active deformation control for steel support structures, which can achieve real-time adjustment by adjusting the support axial force.

[0003] However, the current system mainly relies on monitoring data for adjustment, so there is a problem of deformation control lag. In the deformation control of the support structure, this lag may lead to excessive deformation during the foundation pit excavation process and cannot be adjusted in time, thereby increasing the construction risk. In recent years, data-driven deep learning models have shown excellent performance in soil deformation prediction because they can capture complex nonlinear relationships and automatically extract features. Replacing real-time monitoring data with the prediction results generated by these methods can achieve early active deformation control of the foundation pit. However, existing research is mainly based on convolutional neural networks to extract spatial features, but these methods usually give the same weight to different monitoring points and cannot distinguish the importance of each monitoring point, resulting in the model being unable to focus on key areas that have a greater impact on deformation, limiting the accuracy of deformation prediction at multiple monitoring points and the effectiveness of active control. Summary of the invention

[0004] Purpose of the invention: The problem to be solved by the present invention is that existing research is mainly based on convolutional neural networks to extract spatial features, but these methods usually assign the same weight to different monitoring points and cannot distinguish the importance of each monitoring point, resulting in the model being unable to focus on key areas that have a greater impact on deformation, limiting the accuracy of deformation prediction at multiple monitoring points and the effectiveness of active control.

[0005] Technical solution: In a first aspect, the present invention provides a method for predicting and controlling foundation pit deformation based on a graph neural network, which includes: using a spatiotemporal graph attention network model and utilizing a graph attention mechanism to capture spatial features between monitoring points around steel supports;

[0006] The spatial features extracted by the spatiotemporal graph attention network model are input into the gated recurrent unit to extract a single representation of the entire graph and obtain the deformation prediction value of the monitoring point;

[0007] The wall deformation is controlled in advance based on the deformation prediction value.

[0008] Furthermore, the prediction and control method adopts a spatiotemporal graph attention network model and uses a graph attention mechanism to capture the spatial features between monitoring points around the steel support, including:

[0009] Determine the structure of the graph, treating each monitoring point as a node;

[0010] The characteristic vector h of each node i is composed of the wall offset data d and axial force data f of the monitoring point, which is expressed as follows:

[0011] h i =[d i , f i ];

[0012] Among them, h i is the feature vector of node i, d i is the wall offset data, f i is the axial force data of the wall;

[0013] The obtained eigenvector is transformed by the shared linear matrix p to produce the transformed spatial features, expressed as,

[0014] h i' =ph i ;

[0015] Among them, h i' is the spatial feature of node i, p is a linear matrix, h i is the feature vector of node i.

[0016] Furthermore, the prediction and control method, wherein the spatial features extracted by the spatiotemporal graph attention network model are input into a gated recurrent unit to extract a single representation of the entire graph, includes:

[0017] The attention coefficient is calculated through a single linear layer, and the LeakyReLU activation function is used afterwards to focus on the node pairs and consider their transformed spatial features. The calculation structure is as follows:

[0018] e ij =LeakyReLU(a T [h i' ||h j' ]);

[0019] Among them, e ij is the attention coefficient, a T is the learnable parameter vector, h i' is the spatial feature of node i, h j' is the spatial feature of node j;

[0020] Attention coefficient e ijNormalization is performed using the softmax function to produce the final attention score, which is calculated as follows,

[0021]

[0022] Among them, α ij is the attention score, e ij is the attention coefficient, e ik is the attention coefficient between node i and its neighboring node k

[0023] The attention scores between different nodes are obtained through the multi-head attention mechanism, and the attention score α ij It allows weighted aggregation of the spatial features of neighboring nodes j to update the feature vector of each node and finally realize the representation of each node in the network:

[0024] h i' =∑ j∈N(i) α ij h j' ;

[0025] Among them, h i' is the spatial feature of node i, α ij is the attention score, h j' is the spatial feature of neighboring node j,

[0026] Aggregate the spatial features of all nodes to obtain a single representation h of the entire graph G ,

[0027]

[0028] Among them, h G is a single representation of the entire graph, V is the set of nodes in the graph, and h i' is the spatial feature of node i.

[0029] Furthermore, the prediction and control method, wherein the deformation prediction value of the monitoring point is obtained through a single representation of the entire graph, comprises:

[0030] A single representation h for the entire graph G Apply a fully connected layer and a nonlinear activation function ReLU,

[0031] z=ReLU(W·h G + b);

[0032] Among them, z is the middle layer feature, W and b are the weight and bias of the fully connected layer respectively, and h G is a single representation of the entire graph;

[0033] Finally, the output deformation prediction value y is expressed as,

[0034] y=W'·z+b';

[0035] Among them, y is the deformation prediction value, W' is the weight of the output layer, b' is the bias of the output layer, and z is the intermediate layer feature.

[0036] Furthermore, the prediction and control method, wherein the wall deformation is controlled in advance based on the deformation prediction value, comprises:

[0037] Calculate the error between the target deformation value and the deformation prediction value;

[0038] A PID controller is used to control the lateral displacement of the wall by controlling the axial force;

[0039] The proportional part adjusts the axial force according to the current error, which is the difference between the target deformation value and the deformation prediction value;

[0040] The integral part is responsible for calculating the cumulative sum of errors over a period of time;

[0041] The derivative part calculates the error rate of change by measuring the difference between the current error and the previous error and dividing it by the time interval;

[0042] The PID controller is expressed as,

[0043] P(t)=K p ·e(t);

[0044] I(t)=K i ∫e(t)dt;

[0045]

[0046] Among them, P(t) is the proportional part, K p is the proportional gain, e(t) is the error, I(t) is the integral part, K i is the integral gain, D(t) is the differential part, K d is the differential gain;

[0047] Combining the proportional part, the integral part and the differential part, the output control value is expressed as,

[0048]

[0049] Among them, u(t) is the PID control value, K p is the proportional gain, e(t) is the error, K i is the integral gain, K d is the differential gain,

[0050] The support axial force is controlled by PID control value to control the lateral displacement of the wall in advance.

[0051] Furthermore, the second aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for predicting and controlling foundation pit deformation based on graph neural network as described in the first aspect of the present invention is implemented.

[0052] Furthermore, the third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for foundation pit deformation prediction and control based on graph neural network as described in the first aspect of the present invention.

[0053] Beneficial effects: Compared with the prior art, the significant advantages of the present invention are: the present invention utilizes the space-time graph attention network to analyze and weight the spatial relationship between multiple monitoring points. By distinguishing the importance of the supporting axial force near each monitoring point, the model more accurately captures the spatial heterogeneity of soil deformation, thereby significantly improving the accuracy of foundation pit deformation prediction.

[0054] This method uses the predicted value of the ST-GAT model instead of the actual monitoring value, and realizes the proactive control of wall deformation in advance by reasonably adjusting the servo axis force. This pre-adjustment can effectively prevent potential structural risks, reduce the lag of deformation control, and improve construction safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the foundation pit deformation prediction and control method based on graph neural network in Example 1;

[0056] Figure 2 is an attention weight map between the features of the foundation pit deformation prediction and control method based on graph neural network in Example 2;

[0057] Figure 3 It is a wall lateral displacement control diagram under different proportional item settings of the foundation pit deformation prediction and control method based on graph neural network in Example 2;

[0058] Figure 4 It is a wall displacement control diagram under different integral item settings of the foundation pit deformation prediction and control method based on graph neural network in Example 2;

[0059] Figure 5 This is a comparison chart of the axial forces before and after adjustment of the foundation pit deformation prediction and control method based on graph neural network in Example 2. DETAILED DESCRIPTION

[0060] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0061] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0062] Example 1

[0063] Reference Figure 1 , which is the first embodiment of the present invention, provides a method for predicting and controlling foundation pit deformation based on graph neural network, comprising the following steps:

[0064] Determine the structure of the graph, treating each monitoring point as a node;

[0065] The characteristic vector h of each node i is composed of the wall offset data d and axial force data f of the monitoring point, which is expressed as follows:

[0066] h i =[d i , f i ];

[0067] Among them, h i is the feature vector of node i, d i is the wall offset data, f i is the axial force data of the wall;

[0068] The obtained eigenvector is transformed by the shared linear matrix p to produce the transformed spatial features, expressed as,

[0069] h i' =ph i ;

[0070] Among them, h i' is the spatial feature of node i, p is a linear matrix, h i is the feature vector of node i.

[0071] The attention coefficient is calculated through a single linear layer, and the LeakyReLU activation function is used afterwards to focus on the node pairs and consider their transformed spatial features. The calculation structure is as follows:

[0072] e ij =LeakyReLU(a T [h i' ||h j' ]);

[0073] Among them, e ij is the attention coefficient, a T is the learnable parameter vector, h i'is the spatial feature of node i, h j' is the spatial feature of node j;

[0074] Attention coefficient e ij Normalization is performed using the softmax function to produce the final attention score, which is calculated as follows,

[0075]

[0076] Among them, α ij is the attention score, e ij is the attention coefficient, e ik is the attention coefficient between node i and its neighboring node k

[0077] The attention scores between different nodes are obtained through the multi-head attention mechanism, and the attention score α ij It allows weighted aggregation of the spatial features of neighboring nodes j to update the feature vector of each node and finally realize the representation of each node in the network:

[0078] h i' =∑ j∈N(i) α ij h j' ;

[0079] Among them, h i' is the spatial feature of node i, α ij is the attention score, h j' is the spatial feature of neighboring node j,

[0080] Aggregate the spatial features of all nodes to obtain a single representation h of the entire graph G ,

[0081]

[0082] Among them, h G is a single representation of the entire graph, V is the set of nodes in the graph, and h i' is the spatial feature of node i.

[0083] A single representation h for the entire graph G Apply a fully connected layer and a nonlinear activation function ReLU,

[0084] z=ReLU(W·h G + b);

[0085] Among them, z is the middle layer feature, W and b are the weight and bias of the fully connected layer respectively, and h G is a single representation of the entire graph;

[0086] Finally, the output deformation prediction value y is expressed as,

[0087] y=W'·z+b';

[0088] Among them, y is the deformation prediction value, W' is the weight of the output layer, b' is the bias of the output layer, and z is the intermediate layer feature.

[0089] Calculate the error between the target deformation value and the deformation prediction value;

[0090] A PID controller is used to control the lateral displacement of the wall by controlling the axial force;

[0091] The proportional part adjusts the axial force according to the current error, which is the difference between the target deformation value and the deformation prediction value;

[0092] The integral part is responsible for calculating the cumulative sum of errors over a period of time;

[0093] The derivative part calculates the error rate of change by measuring the difference between the current error and the previous error and dividing it by the time interval;

[0094] The PID controller is expressed as,

[0095] P(t)=K p ·e(t);

[0096] I(t)=K i ∫e(t)dt;

[0097]

[0098] Among them, P(t) is the proportional part, K p is the proportional gain, e(t) is the error, I(t) is the integral part, K i is the integral gain, D(t) is the differential part, K d is the differential gain;

[0099] Combining the proportional part, the integral part and the differential part, the output control value is expressed as,

[0100]

[0101] Among them, u(t) is the PID control value, K p is the proportional gain, e(t) is the error, K i is the integral gain, K d is the differential gain,

[0102] The support axial force is controlled by PID control value to control the lateral displacement of the wall in advance.

[0103] This embodiment also provides a computer device, which is suitable for the case of a foundation pit deformation prediction and control method based on a graph neural network, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the foundation pit deformation prediction and control method based on a graph neural network as proposed in the above embodiment.

[0104] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0105] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for predicting and controlling foundation pit deformation based on graph neural network proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, disk or optical disk.

[0106] In summary, the beneficial effects of the present invention are as follows: the present invention utilizes the space-time graph attention network to analyze and weight the spatial relationship between multiple monitoring points. By distinguishing the importance of the supporting axial force near each monitoring point, the model more accurately captures the spatial heterogeneity of soil deformation, thereby significantly improving the accuracy of foundation pit deformation prediction.

[0107] This method uses the predicted value of the ST-GAT model instead of the actual monitoring value, and realizes the proactive control of wall deformation in advance by reasonably adjusting the servo axis force. This pre-adjustment can effectively prevent potential structural risks, reduce the lag of deformation control, and improve construction safety.

[0108] Example 2

[0109] Referring to Table 1, which is the second embodiment of the present invention, in order to further verify the technical solution of the present invention, experimental simulation data of the foundation pit deformation prediction and control method based on graph neural network are given.

[0110] This specific embodiment is a foundation pit project with an excavation depth between 23.85 meters and 25.41 meters. The surrounding environment is complex, including underground pipelines and complex geological conditions. The area is densely populated, surrounded by residential buildings, schools and complex road networks, and is adjacent to sensitive facilities such as schools and subway tunnels, so deformation control is crucial.

[0111] The main structure of the station is 175 meters long, the standard section is 22.79 meters wide, the end shaft section is 27 meters wide, the top plate is covered with soil about 3.05 meters, the bottom of the center foundation of the station is buried 30.85 meters deep, and the end shaft foundation is buried 32.41 meters deep. The main retaining structure is a 1,200 mm thick underground continuous wall with four concrete supports and five steel supports.

[0112] The project uses a servo support system, combining servo control technology with traditional steel supports to control the deformation of the retaining wall. Specifically, load sensors are used to monitor the axial force, and the support force is adjusted in real time by controlling the displacement of the jack. By detecting changes in the axial force, the displacement of the jack is precisely adjusted to control the deformation of the retaining wall and adapt to environmental changes.

[0113] Table 1 shows the comparison of the performance of the ST-GAT model of the present invention with other alternative models.

[0114] In the present invention, the evaluation indicators MSE (mean square error) and MAE (mean absolute error) are used to measure the accuracy of wall displacement prediction. The smaller the value, the better the prediction effect. In terms of alternative models, several common deep learning models are considered, including graph convolutional long short-term memory neural network (GCN-LSTM), graph convolutional gated recurrent unit (GCN-GRU), convolutional neural network (CNN), long short-term memory neural network (LSTM), long short-term memory gated recurrent unit network (LSTM-GRU), temporal convolutional network (TCN) and bidirectional long short-term memory neural network (BiLSTM).

[0115] By comparing the data in the table, it can be found that the ST-GAT model is superior to other models in prediction accuracy. Specifically, the MSE of ST-GAT is 0.485 mm.2 , MAE is 0.567mm, compared with GCN-LSTM (MSE = 1.33mm 2 , MAE=1.37mm), GCN-GRU (MSE=1.41mm 2 , MAE=1.10mm), CNN (MSE=1.91mm 2 , MAE=2.88mm), LSTM (MSE=1.91mm 2 , MAE=1.14mm), LSTM-GRU (MSE=1.53mm 2 , MAE=1.36mm), TCN (MSE=1.62mm 2 , MAE = 2.13 mm) and BiLSTM (MSE = 3.28 mm) 2 , MAE=3.24mm) model, ST-GAT shows significantly lower error, proving its superiority in wall displacement prediction.

[0116] Table 1 Comparison of the performance of the proposed ST-GAT model with other alternative models

[0117]

[0118]

[0119] Figure 2 The distribution of attention weights in two layers of the Graph Attention Network (GAT) is shown. The nodes labeled 2, 3, 4, and 6 represent the corresponding monitoring points, and the weights on the edges indicate the strength of attention between these nodes. For example, in the first layer, node 2 shows a strong connection with node 3 (weights 0.53 and 0.5), while node 6 has a weaker connection with other nodes such as 0.31 and 0.35. These weights reflect how GAT assigns importance to different monitoring points when predicting wall deformation based on spatial relationships.

[0120] Figure 3 The wall deformation control is shown when kp is set to 0.01, 0.8 and 0.9. In these cases, ki is set to 0. The blue dashed lines represent the set target deformation values, which are 7.5 mm, 9.6 mm, 17 mm and 30 mm for monitoring points 2, 3, 4 and 6, respectively. The results show that the control model reacts more quickly with the increase of kp value. Specifically, the adjusted wall deformation tends to the target deformation more quickly with the increase of kp value. This trend can be clearly seen from the graph. As kp increases from 0.01 to 0.9, the alignment with the target deformation line (blue dashed line) is closer. However, there are still small errors in the adjusted wall deformation for monitoring points 2 and 3, while the errors for monitoring points 4 and 6 are larger. This difference suggests that the integral part should be further incorporated into the control model.

[0121] Figure 4 The effect of adjusting the ki setting on the wall deformation is demonstrated. The error of ki = 0.01 is significantly greater than that of ki = 0.1 and 0.2. ki = 0.1 and 0.2 reach the target deformation value after about 18 days of adjustment. Among them, ki = 0.2 has better deformation results at monitoring points 4 and 6. Although there is a slight overshoot at wall deformation point 3, the deformation exceeds the target by less than 0.3 mm, but it quickly stabilizes within one to two days. The results show that by properly setting ki and kp, the proposed intelligent control system can effectively control the wall deformation and make it close to the target deformation.

[0122] Figure 5 The comparison of the axial forces before and after adjustment is shown. The reasonable range is determined by the historical maximum and minimum values ​​of these steel supports. The axial force range of the steel supports is as follows: support 2 ranges from 2000kN to 3000kN, support 3 ranges from 3000kN to 4000kN, and support 4 and support 6 range from 4000kN to 5200kN. The results show that the axial forces of the four steel supports after adjustment are all within the reasonable range, proving the effectiveness of the proposed model.

Claims

1. A method for predicting and controlling foundation pit deformation based on graph neural network, characterized in that: include, A spatiotemporal graph attention network model is adopted to capture the spatial features between monitoring points around the steel support using the graph attention mechanism. The spatial features extracted by the spatiotemporal graph attention network model are fed into a gated recurrent unit to extract a single representation of the entire graph. The deformation prediction value of the monitoring point is obtained through a single representation of the entire graph; The wall deformation is controlled in advance based on the deformation prediction value.

2. The method for predicting and controlling foundation pit deformation based on graph neural network according to claim 1, characterized in that: The spatiotemporal graph attention network model is adopted to capture the spatial features between the monitoring points around the steel support using the graph attention mechanism, including: Determine the structure of the graph, treating each monitoring point as a node; The characteristic vector h of each node i is composed of the wall offset data d and axial force data f of the monitoring point, which is expressed as follows: h i =[d i ,fi]; Among them, h i is the feature vector of node i, d i is the wall offset data, f i is the axial force data of the wall; The obtained eigenvector is transformed by the shared linear matrix p to produce the transformed spatial features, expressed as, h i' =ph i ; Among them, h i' is the spatial feature of node i, p is a linear matrix, h i is the feature vector of node i.

3. The method for predicting and controlling foundation pit deformation based on graph neural network according to claim 2, characterized in that: The spatial features extracted by the spatiotemporal graph attention network model are input into the gated recurrent unit to extract a single representation of the entire graph, including: The attention coefficient is calculated through a single linear layer, and the LeakyReLU activation function is used afterwards to focus on the node pairs and consider their transformed spatial features. The calculation structure is as follows: have been ij =LeakyReLU(a T [h i' ||h j' ]); Among them, e ij is the attention coefficient, a T is the learnable parameter vector, h i' is the spatial feature of node i, h j' is the spatial feature of node j; Attention coefficient e ij Normalization is performed using the softmax function to produce the final attention score, which is calculated as follows, Among them, α ij is the attention score, e ij is the attention coefficient, e ik is the attention coefficient between node i and its neighboring node k; The attention scores between different nodes are obtained through the multi-head attention mechanism, and the attention score α ij It allows weighted aggregation of the spatial features of neighboring nodes j to update the feature vector of each node and finally realize the representation of each node in the network: h i' =∑ j∈N(i) a ij h j' ; Among them, h i' is the spatial feature of node i, α ij is the attention score, h j' is the spatial feature of neighboring node j, Aggregate the spatial features of all nodes to obtain a single representation h of the entire graph G , Among them, h G is a single representation of the entire graph, V is the set of nodes in the graph, and h i' is the spatial feature of node i.

4. The method for predicting and controlling foundation pit deformation based on graph neural network according to claim 3, characterized in that: The deformation prediction value of the monitoring point is obtained by a single representation of the entire graph, including: A single representation h for the entire graph G Apply a fully connected layer and a nonlinear activation function ReLU, z=ReLU(W h G +b); Among them, z is the middle layer feature, W and b are the weight and bias of the fully connected layer respectively, and h G is a single representation of the entire graph; Finally, the output deformation prediction value y is expressed as, y = W'·z+b'; Among them, y is the deformation prediction value, W' is the weight of the output layer, b' is the bias of the output layer, and z is the intermediate layer feature.

5. The method for predicting and controlling foundation pit deformation based on graph neural network according to claim 4, characterized in that: The method of controlling the wall deformation in advance based on the deformation prediction value includes: Calculate the error between the target deformation value and the deformation prediction value; A PID controller is used to control the lateral displacement of the wall by controlling the axial force; The proportional part adjusts the axial force according to the current error, which is the difference between the target deformation value and the deformation prediction value; The integral part is responsible for calculating the cumulative sum of errors over a period of time; The derivative part calculates the error rate of change by measuring the difference between the current error and the previous error and dividing it by the time interval; The PID controller is expressed as, P(t)=K p ·e(t); I(t)=K i ∫e(t)dt; Among them, P(t) is the proportional part, K p is the proportional gain, e(t) is the error, I(t) is the integral part, K i is the integral gain, D(t) is the differential part, K d is the differential gain; Combining the proportional part, the integral part and the differential part, the output control value is expressed as, Among them, u(t) is the PID control value, K p is the proportional gain, e(t) is the error, K i is the integral gain, K d It is the differential gain, and the support axial force is controlled by the PID control value to control the wall lateral displacement in advance.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for predicting and controlling foundation pit deformation based on graph neural network described in any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting and controlling foundation pit deformation based on graph neural network described in any one of claims 1 to 5 are implemented.

Images