Optimization control system and method based on asymmetric excavation deformation of foundation pit

Abstract

The invention discloses an optimal control system and method based on asymmetric excavation deformation of a foundation pit, and belongs to the field of engineering construction management. Comprising a virtual simulation module, an acquisition preprocessing module, a historical data processing module, a deformation prediction module, a path simulation module, a decision adjustment module, a parameter optimization module, a dynamic adjustment module, a monitoring feedback module and a report generation module. According to the method, real-time adjustment can be carried out in a dynamic environment, the flexibility and adaptability of decision making can be improved, the overall construction strategy can be optimized, different excavation paths and deformation conditions thereof can be visually displayed, communication and negotiation among workers are facilitated, a personalized optimization scheme can be provided, optimization efficiency can be improved, calculation time can be shortened, and the construction efficiency can be improved. The deformation risk in the foundation pit excavation process is effectively reduced, the decision accuracy is improved, and the construction safety is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of engineering construction management. Specifically, it relates to an optimization control system and method based on the deformation of asymmetric excavation of foundation pits. Background Technique

[0002] Foundation pit excavation is a key civil engineering activity and is widely used in engineering projects such as buildings, bridges, and underground facilities. As an indispensable link in civil engineering, foundation pit excavation is involved in projects including buildings, bridges, and underground facilities. With the acceleration of the urbanization process, the scale and complexity of foundation pit excavation are also increasing continuously. Especially in the case of asymmetric excavation, the deformation problems that occur during the construction process are particularly prominent. These deformations may not only affect the structural safety of the foundation pit but also cause damage to surrounding buildings and infrastructure, bringing potential risks to the construction. Traditional methods for controlling foundation pit deformation rely on experience and simple monitoring means and are difficult to cope with complex soil conditions and changing environmental factors.

[0003] The existing optimization control systems and methods based on the deformation of asymmetric excavation of foundation pits cannot be adjusted in real time in a dynamic environment, reducing the flexibility and adaptability of decision-making and making it inconvenient for communication and negotiation among staff members. In addition, the existing optimization control systems based on the deformation of asymmetric excavation of foundation pits cannot provide personalized optimization schemes, reducing the optimization efficiency, increasing the calculation time, and having a relatively high deformation risk during the foundation pit excavation process. Therefore, we propose an optimization control system and method based on the deformation of asymmetric excavation of foundation pits. Summary of the Invention

[0004] The purpose of the present invention is to provide an optimization control system and method based on the deformation of asymmetric excavation of foundation pits, so as to solve the problems that the existing optimization control systems and methods based on the deformation of asymmetric excavation of foundation pits cannot be adjusted in real time in a dynamic environment, reducing the flexibility and adaptability of decision-making and making it inconvenient for communication and negotiation among staff members. In addition, the existing optimization control systems based on the deformation of asymmetric excavation of foundation pits cannot provide personalized optimization schemes, reducing the optimization efficiency, increasing the calculation time, and having a relatively high deformation risk during the foundation pit excavation process.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] An optimization control system based on the deformation of asymmetric excavation of foundation pits, comprising:

[0007] Historical data processing module, analyzing historical data and extracting characteristic data during foundation pit excavation; Acquisition and preprocessing module, collecting and processing parameter information of each device in real time during foundation pit excavation; Deformation prediction module, predicting the deformation of the upcoming excavation work and evaluating potential risks; Virtual simulation module, constructing a virtual foundation pit excavation environment and simulating the deformation conditions occurring during construction; Path simulation module, simulating different excavation paths and their corresponding deformation conditions; Decision-making and adjustment module, optimizing the excavation control strategy according to the deformation prediction and path simulation results; Parameter optimization module, optimizing the key parameters during foundation pit excavation; Dynamic adjustment module, dynamically adjusting the excavation path of the device and tool in real time according to the feedback results of the virtual model; Monitoring and feedback module, monitoring the deformation conditions during excavation in real time and feeding back the data to the system; Report generation module, generating a report on the excavation process, recording the deformation prediction, decision-making process and implementation effect.

[0008] As a further solution of the present invention, the method for the historical data processing module to extract characteristic data during foundation pit excavation includes the following steps:

[0009] S1.1: Collect historical data, detect the data identifiers of each group, and at the same time screen out redundant data with the same ID identifier. Then detect the missing values existing in each group of historical data, and fill the existing missing values by mean filling or median filling;

[0010] S1.2: Draw the corresponding box plot based on the historical data, calculate the upper and lower limits of the outliers, mark the values in each historical data that exceed the upper limit and are lower than the lower limit as outliers, and at the same time screen out the marked outliers. Display the distribution and relationship of the data through histograms and scatter plots. Then use t-tests or analysis of variance to compare the mean differences between different groups;

[0011] S1.3: Then calculate the statistic and compare it with the significance level to determine whether the data conforms to the normal distribution, and screen out the data that does not conform to the normal distribution. Then draw a time series diagram to observe trends, seasonality and periodic changes. Then use Pearson or Spearman correlation coefficients to construct a correlation matrix to identify strongly correlated and weakly correlated features.

[0012] As a further solution of the present invention, the method for the deformation prediction module to evaluate potential risks includes the following steps:

[0013] S2.1: Perform Z-score standardization processing on each group of characteristic data extracted by the historical data processing module to unify it within the range of 0 to 1. Divide the data into a training set, a validation set and a test set, and design the DNN model structure based on the Keras deep learning framework, and set the input layer, hidden layer and output layer of the model;

[0014] S2.2: Input the training set as input data into the DNN model. The DNN model calculates the predicted values of each input data through the forward propagation algorithm. Then, calculate the difference between the predicted value and the actual value through the mean square error function. Next, through the backpropagation algorithm, calculate the gradients between the difference and each network layer, and adjust the network weights through the Adam optimizer;

[0015] S2.3: After each round of training, use the validation set data to evaluate the performance indicators of the DNN model, and adjust the hyperparameters of the model according to the evaluation results. Then, repeat the training and validation of the DNN model until the difference value converges to the preset range and stop. Evaluate the generalization ability of the final DNN model through the test set;

[0016] S2.4: Input the latest collected data into the DNN model. The DNN model processes the latest data through the input layer, hidden layer, and output layer in sequence through the forward propagation algorithm to obtain the predicted deformation amount. Then, combine the predicted deformation amount with the preset safety threshold to determine whether there is a risk. If it is higher than the safety threshold, it is determined that there is a risk.

[0017] As a further solution of the present invention, the parameter optimization module optimizes the key parameters in the foundation pit excavation process, including the following steps:

[0018] S3.1: Randomly generate multiple groups of parameter combinations, set the population size and the search range of the parameters according to the number of generated parameter combinations, and use each parameter combination as an individual in the population. Use the inverse function of deformation prediction to evaluate the performance of each group of individuals to obtain the fitness value of each group of individuals;

[0019] S3.2: Find the individual with the highest fitness value in the initial population and use it as the current global best solution. Update the position of each group of individuals according to the position of the current individual and the best individual, set the encirclement threshold, and at the same time, generate a random number between 0 and 1 during each round of position update;

[0020] S3.3: If the random number is less than the preset encirclement threshold, it means that the remaining individuals surround the optimal solution, and update the positions of the remaining individuals according to the optimal solution. If the random number is greater than the preset encirclement threshold, it means that the remaining individuals move away from the optimal solution, and perform exploration behavior through random search, select a random position in the population space, update the positions of the remaining individuals, and update the coefficient vector based on the linearly decreasing rule after each iteration;

[0021] S3.4: After updating the individual positions, check whether they exceed the search boundary. If the updated individual positions exceed the preset minimum or maximum values, limit them within the boundary, calculate the fitness of the new positions, and update the fitness values and position records;

[0022] S3.5: Repeat the iterative search until the preset maximum number of iterations is reached, then stop the search, record and display the currently found best parameter combination and its corresponding deformation prediction results, apply the found best parameter combination to the current construction plan, and simultaneously detect the construction situation of the foundation pit excavation in real time for dynamic parameter adjustment.

[0023] The present invention also discloses an optimization control method based on the deformation of the asymmetric excavation of the foundation pit, which is implemented through the above optimization control system. The control method includes the following steps:

[0024] The first step is to collect and preprocess the historical excavation data and real-time monitoring data;

[0025] The second step is to construct a foundation pit excavation model and analyze the deformation under complex geological conditions;

[0026] The third step is to predict the deformation during the excavation work based on the historical data and real-time monitoring data;

[0027] The fourth step is to simulate different excavation paths and evaluate the safety and effect of each excavation path;

[0028] The fifth step is to dynamically adjust the excavation strategy and parameters and continuously monitor the actual deformation of the foundation pit.

[0029] Among them, the fourth step specifically includes the following steps:

[0030] S4.1: Collect the construction data of each group of foundation pit excavations, construct the corresponding state space according to the construction data of each group, and each node in the state space represents a state. Then construct a group of root nodes, input the current initial excavation conditions and the corresponding deformation data into the root nodes, and generate new child nodes according to the subsequent feasible excavation paths of the root nodes to simulate the corresponding excavation decisions;

[0031] S4.2: Starting from the root node, select the child node with the highest UCB value layer by layer through the UCB strategy until the selected child node is a leaf node that has not been visited and not fully expanded, then stop the selection, select feasible decisions according to the current excavation conditions and environment, and create new child nodes for each feasible decision to save their state information;

[0032] S4.3: Update the environmental state of each child node according to the selected path, and use the previously trained DNN model to input the feature information of each child node for simulation to obtain the corresponding deformation prediction value, and calculate the evaluation value according to the predicted deformation amount and the risk threshold. After the simulation is completed, trace back the simulation results from the child nodes to the root node and update the evaluation values of all nodes on the path;

[0033] S4.4: Repeat the steps of selection, expansion, simulation, and backtracking until the change in the evaluation value converges to a preset termination range. Starting from the root node, select the child node with the highest evaluation value as the optimal path, record the corresponding deformation conditions, and at the same time analyze the selected path. Combining the predicted deformation amount and safety standards, evaluate its feasibility and risks.

[0034] As a further solution of the present invention, the state space described in S4.1 includes the excavation depth, excavation path, soil type, surrounding environment, equipment status, deformation history, construction stage, safety threshold, construction strategy, and construction personnel allocation.

[0035] As a further solution of the present invention, the fifth step specifically includes the following steps:

[0036] S4.1: Collect all the excavation states of the foundation pit to construct the corresponding state set, and construct the corresponding action set according to different excavation paths and excavation equipment parameters. Then, through historical data or simulation calculations, obtain the probabilities of transferring to new states after performing different actions in each state. According to the deformation amounts and safety thresholds obtained from the simulation and historical data, calculate the immediate rewards obtained by performing different actions in each state.

[0037] S4.2: Analyze the deformation conditions in different states, and construct a transition probability matrix based on the calculated sets of transition probabilities. Under the current strategy, according to the corresponding immediate rewards or the results of policy evaluation, determine the optimal actions to be taken in each state, and in each state, select the actions that can maximize the value and update the policy.

[0038] S4.3: Repeat policy evaluation and policy improvement until the change range of the immediate reward reaches within the convergence threshold, stop policy improvement, extract the final optimal policy from the iterative results, transform the obtained optimal policy into an actual excavation plan to guide every decision-making step in the construction process, continuously monitor the deformation conditions during the implementation process, and adjust the policy according to the actual situation.

[0039] Advantages of the present invention:

[0040] 1. The present invention constructs a corresponding state space based on each group of construction data, then constructs a set of root nodes, generates new child nodes according to the subsequent feasible excavation paths of the root nodes, simulates the corresponding excavation decisions. Starting from the root nodes, the child nodes with the highest UCB values are selected layer by layer through the UCB strategy until the selected child node is an unvisited and incompletely expanded leaf node, at which point the selection stops. According to the current excavation conditions and environment, feasible decisions are selected, new child nodes are created for each feasible decision, and their state information is saved. The environmental state of each child node is updated according to the selected path, and simulations are carried out to obtain the corresponding deformation prediction values. The evaluation values are calculated based on the predicted deformation amounts and risk thresholds. After the simulation is completed, the simulation results are traced back from the child nodes to the root nodes, and the evaluation values of all nodes on the path are updated. The steps of selection, expansion, simulation, and backtracking are repeated until the change in the evaluation value converges to a preset termination range. Starting from the root nodes, the child node with the highest evaluation value is selected as the optimal path, and the corresponding deformation conditions are recorded. At the same time, the selected path is analyzed, and its feasibility and risks are evaluated in combination with the predicted deformation amounts and safety standards. Meanwhile, a construction report is generated to display the evaluation results of different excavation paths and the corresponding deformation conditions, which can be adjusted in real time in a dynamic environment, improve the flexibility and adaptability of decision-making, optimize the overall construction strategy, and can intuitively display different excavation paths and their deformation conditions, facilitating communication and negotiation among staff members.

[0041] 2. The present invention randomly generates multiple groups of parameter combinations, sets the population size and the search range of the parameters according to the number of generated parameter combinations, calculates the fitness values of each group of individuals using the inverse function of deformation prediction, updates the positions of each group of individuals according to the positions of the current individuals and the best individuals, sets a hunting threshold. At the same time, in each round of position update process, a random number is generated. If the random number is less than the preset hunting threshold, it means that the remaining individuals surround the optimal solution and their positions are updated according to the optimal solution. If the random number is greater than the preset hunting threshold, it means that the remaining individuals move away from the optimal solution, and exploratory behaviors are performed through random search, randomly selecting positions in the population space to update the positions of the remaining individuals. After each iteration ends, after updating the positions of the individuals, it is checked whether the boundaries are exceeded. If the updated individual position exceeds the preset minimum or maximum values, it is restricted within the boundaries. The iterative search is repeated until the preset maximum number of iterations is reached, at which point the search stops, and the best parameter combination found is applied to the current construction plan. At the same time, the construction situation of the foundation pit excavation is detected in real time for dynamic parameter adjustment, which can provide personalized optimization solutions, improve the optimization efficiency, reduce the calculation time, effectively reduce the deformation risk during the foundation pit excavation process, improve the accuracy of decision-making, and ensure construction safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention will be further described below with reference to the accompanying drawings.

[0043] Figure 1 This is the system block diagram of an optimization control system based on the deformation of asymmetric excavation of foundation pits proposed by the present invention;

[0044] Figure 2 This is the flow block diagram of an optimization control method based on the deformation of asymmetric excavation of foundation pits proposed by the present invention. Specific implementation manners

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0046] Embodiment 1

[0047] An optimization control system based on the deformation of asymmetric excavation of foundation pits, as Figure 1 shown, includes a virtual simulation module, a collection and preprocessing module, a historical data processing module, a deformation prediction module, a path simulation module, a decision-making and adjustment module, a parameter optimization module, a dynamic adjustment module, a monitoring and feedback module, and a report generation module;

[0048] The virtual simulation module is used to construct a virtual foundation pit excavation environment and simulate the deformation conditions that occur during the construction process;

[0049] The collection and preprocessing module is used to collect and process the parameter information of each device in real time during the foundation pit excavation process;

[0050] The historical data processing module is used to analyze historical data and extract characteristic data during the foundation pit excavation process;

[0051] Specifically, the historical data processing module collects historical data, detects the data identifiers of each group, and screens out redundant data with the same ID identifier. Then, it detects the missing values in each group of historical data, fills in the existing missing values by mean filling or median filling, draws the corresponding box plot based on the historical data, calculates the upper and lower limits of the outliers, marks the values exceeding the upper limit and below the lower limit in each historical data as outliers, and screens out the marked outliers at the same time. It displays the distribution and relationship of the data through histograms and scatter plots. Then, it uses t-tests or analysis of variance to compare the mean differences between different groups, calculates the statistic and compares it with the significance level to determine whether the data conforms to the normal distribution, and screens out the data that does not conform to the normal distribution. After that, it draws a time series plot to observe trends, seasonality, and cyclic changes, and then uses Pearson or Spearman correlation coefficients to construct a correlation matrix to identify strongly and weakly correlated features.

[0052] The deformation prediction module is used to predict the deformation of the upcoming excavation work and evaluate potential risks;

[0053] Specifically, the characteristic data of each group extracted by the historical data processing module are standardized by Z-score to the range of 0 to 1, the data is divided into a training set, a validation set, and a test set, and the DNN model structure is designed based on the Keras deep learning framework. The input layer, hidden layer, and output layer of the model are set. The training set is used as the input data and passed into the DNN model. The DNN model calculates the predicted values of each input data through the forward propagation algorithm. Then, it calculates the difference between the predicted value and the actual value through the mean squared error function. Then, through the backpropagation algorithm, it calculates the gradient between the difference and each network layer, and adjusts the network weights through the Adam optimizer. After each round of training, the validation set data is used to evaluate the performance indicators of the DNN model, and the hyperparameters of the model are adjusted according to the evaluation results. Then, the training and validation of the DNN model are repeated until the difference value converges to the preset range and then stops. The generalization ability of the final DNN model is evaluated through the test set. The latest collected data is input into the DNN model. The DNN model processes the latest data through the input layer, hidden layer, and output layer in turn through the forward propagation algorithm to obtain the predicted deformation amount. Then, combined with the predicted deformation amount and the preset safety threshold, it judges whether there is a risk. If it is higher than the safety threshold, it judges that there is a risk and formulates corresponding control measures.

[0054] The path simulation module is used to simulate different excavation paths and their corresponding deformation conditions to provide a reference for decision-making;

[0055] The decision-making adjustment module is used to optimize the excavation control strategy according to the deformation prediction and path simulation results;

[0056] The parameter optimization module is used to optimize the key parameters in the foundation pit excavation process;

[0057] Specifically, multiple groups of parameter combinations are randomly generated, and the population size and the search range of the parameters are set according to the number of generated parameter combinations. Each parameter combination is used as an individual in the population. The inverse function of deformation prediction is used to evaluate the performance of each group of individuals to obtain the fitness value of each group of individuals. The individual with the highest fitness value is found in the initial population and used as the current global best solution. The positions of each group of individuals are updated according to the positions of the current individuals and the best individuals. A hunting threshold is set. At the same time, during each round of position update, a random number between 0 and 1 is generated. If the random number is less than the preset hunting threshold, it means that the remaining individuals surround the optimal solution and the positions of the remaining individuals are updated according to the optimal solution. If the random number is greater than the preset hunting threshold, it means that the remaining individuals move away from the optimal solution and perform exploration behavior through random search, select random positions in the population space, and update the positions of the remaining individuals. After each iteration ends, the coefficient vector is updated based on the rule of linear decrease. After updating the individual positions, it is checked whether the positions exceed the search boundaries. If the updated individual positions exceed the preset minimum or maximum values, they are restricted within the boundaries. The fitness of the new positions is calculated, and the fitness values and position records are updated. The iterative search is repeated until the preset maximum number of iterations is reached, then the search is stopped. The best parameter combination found currently and its corresponding deformation prediction results are recorded and displayed, and the best parameter combination found is applied to the current construction plan. At the same time, the construction situation of the foundation pit excavation is monitored in real time for dynamic parameter adjustment.

[0058] The dynamic adjustment module is used to dynamically adjust the excavation paths of the equipment and tools in real time according to the feedback results of the virtual model;

[0059] The monitoring and feedback module is used to monitor the deformation situation during the excavation process in real time and feed the data back into the system;

[0060] The report generation module is used to generate a report on the excavation process, recording the deformation prediction, decision-making process, and implementation effect.

[0061] Embodiment 2

[0062] Refer to Figure 2 , an optimization control method based on the deformation of asymmetric excavation of a foundation pit. The specific steps of this control method are as follows:

[0063] Step 1, collect and preprocess historical excavation data and real-time monitoring data;

[0064] Step 2, construct a foundation pit excavation model and analyze the deformation situation under complex geological conditions;

[0065] Step 3: Predict the deformation during the excavation work based on historical data and real-time monitoring data;

[0066] Step 4: Simulate different excavation paths and evaluate the safety and effectiveness of each excavation path;

[0067] Specifically, collect the construction data of each group of foundation pit excavations, construct the corresponding state space according to the construction data of each group, and each node in the state space represents a state. Then construct a set of root nodes, input the current initial excavation conditions and the corresponding deformation data into the root nodes. According to the subsequent feasible excavation paths of the root nodes, generate new child nodes, simulate the corresponding excavation decisions. Starting from the root node, layer by layer select the child node with the highest UCB value through the UCB strategy until the selected child node is an unvisited and not fully expanded leaf node, then stop the selection. According to the current excavation conditions and environment, select feasible decisions, create new child nodes for each feasible decision, save their state information, update the environmental state of each child node according to the selected path, and use the previously trained DNN model to input the feature information of each child node for simulation to obtain the corresponding deformation prediction value, and calculate the evaluation value according to the predicted deformation amount and the risk threshold. After the simulation is completed, trace back the simulation results from the child nodes to the root node, and update the evaluation values of all nodes on the path. Repeat the steps of selection, expansion, simulation, and backtracking until the change of the evaluation value converges to the preset termination range. Starting from the root node, select the child node with the highest evaluation value as the optimal path, and record the corresponding deformation situation. At the same time, analyze the selected path, and combine the predicted deformation amount with the safety standard to evaluate its feasibility and risk, and generate a construction report to show the evaluation results of different excavation paths and the corresponding deformation situations.

[0068] It should be further noted that the state space specifically includes excavation depth, excavation path, soil type, surrounding environment, equipment status, deformation history, construction stage, safety threshold, construction strategy, and construction personnel configuration.

[0069] Step 5: Dynamically adjust the excavation strategy and parameters and continuously monitor the actual deformation of the foundation pit.

[0070] Specifically, all the excavation states of the foundation pit are collected to construct the corresponding state set, and the corresponding action set is constructed according to different excavation paths and excavation equipment parameters. Then, the probabilities of transferring to new states after performing different actions in each state are obtained through historical data or simulation calculations. According to the deformation amounts and safety thresholds obtained from the simulation and historical data, the immediate rewards obtained by performing different actions in each state are calculated. The deformation conditions in different states are analyzed, and a transition probability matrix is constructed based on the calculated sets of transition probabilities. Under the current strategy, according to the corresponding immediate rewards or the results of policy evaluation, the optimal actions to be taken in each state are determined. In each state, the actions that can maximize the value are selected, and the strategy is updated. The policy evaluation and improvement are repeated until the change range of the immediate reward reaches within the convergence threshold, and the policy improvement is stopped. The final optimal strategy is extracted from the iterative results, and the obtained optimal strategy is transformed into an actual excavation plan to guide every decision-making step in the construction process. During the implementation process, the deformation conditions are continuously monitored, and the strategy is adjusted according to the actual situation.

[0071] The above content is only an example and illustration of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods for substitution, as long as they do not deviate from the invention or exceed the scope defined by this claim book, they should fall within the protection scope of the present invention.

Claims

1. An optimization control system based on asymmetric excavation deformation of foundation pit, characterized in that: include: Historical data processing module, which analyzes historical data and extracts characteristic data during foundation pit excavation; The acquisition preprocessing module collects and processes the parameter information of each device during the foundation pit excavation process in real time; The deformation prediction module predicts deformation of the upcoming excavation work and assesses potential risks; Virtual simulation module, which builds a virtual foundation pit excavation environment and simulates the deformation that occurs during the construction process; Path simulation module, simulating different excavation paths and their corresponding deformations; The decision adjustment module optimizes the excavation control strategy based on deformation prediction and path simulation results; Parameter optimization module, which optimizes key parameters in the foundation pit excavation process; Dynamic adjustment module, which dynamically adjusts the excavation path of equipment and tools in real time based on the feedback from the virtual model; Monitoring and feedback module, which monitors the deformation during excavation in real time and feeds the data back to the system; The report generation module generates excavation process reports, recording deformation prediction, decision-making process and implementation results.

2. The optimization control system based on asymmetric excavation deformation of foundation pit according to claim 1 is characterized in that: The method for extracting characteristic data during foundation pit excavation by the historical data processing module comprises the following steps: S1.1: Collect historical data and detect the identifiers of each group of data, and filter out redundant data with the same ID identifier. Then detect the missing values ​​in each group of historical data and fill the missing values ​​by mean filling or median filling. S1.2: Draw the corresponding box plot based on the historical data, and calculate the upper and lower limits of the outliers. Mark the values ​​in each historical data that exceed the upper limit and are lower than the lower limit as outliers. At the same time, filter out the marked outliers, and display the distribution and relationship of the data through histograms and scatter plots. Then use t-tests or variance analysis to compare the mean differences between different groups. S1.3: Recalculate the statistic and compare it with the significance level to determine whether the data conforms to the normal distribution and filter out the data that does not conform to the normal distribution. Then draw a time series graph to observe trends, seasonality, and cyclical changes. Use the Pearson or Spearman correlation coefficient to construct a correlation matrix and identify strong and weak correlation features.

3. The optimization control system based on asymmetric excavation deformation of foundation pit according to claim 2 is characterized in that: The method for evaluating potential risks by the deformation prediction module comprises the following steps: S2.1: Perform Z-score standardization on each set of feature data extracted by the historical data processing module to a range of 0 to 1, divide the data into training set, validation set and test set, and design the DNN model structure based on the Keras deep learning framework, setting the model input layer, hidden layer and output layer; S2.2: The training set is passed into the DNN model as input data. The DNN model calculates the predicted value of each input data through the forward propagation algorithm, and then calculates the difference between the predicted value and the actual value through the mean square error function. Then, the back propagation algorithm is used to calculate the difference and the gradient between each network layer, and the network weight is adjusted through the Adam optimizer. S2.3: After each round of training, the validation set data is used to evaluate the performance indicators of the DNN model, and the hyperparameters of the model are adjusted according to the evaluation results. The DNN model is then repeatedly trained and validated until the difference value converges to the preset range. The generalization ability of the final DNN model is evaluated through the test set. S2.4: The latest collected data is input into the DNN model. The DNN model processes the latest data through the input layer, hidden layer and output layer in sequence through the forward propagation algorithm to obtain the predicted deformation. The predicted deformation is then combined with the preset safety threshold to determine whether there is a risk. If it is higher than the safety threshold, it is determined that there is a risk.

4. The optimization control system based on asymmetric excavation deformation of foundation pit according to claim 3 is characterized in that: The parameter optimization module optimizes the key parameters in the foundation pit excavation process and includes the following steps: S3.1: Randomly generate multiple groups of parameter combinations, and set the population size and parameter search range according to the number of generated parameter combinations. Each parameter combination is regarded as an individual in the population, and the inverse function of deformation prediction is used to evaluate the performance of each group of individuals to obtain the fitness value of each group of individuals. S3.2: Find the individual with the highest fitness value in the initial population and use it as the current global optimal solution. Update the position of each group of individuals according to the position of the current individual and the position of the best individual, set the round-up threshold, and generate a random number between 0 and 1 in each round of position update; S3.3: If the random number is less than the preset capture threshold, it means that the remaining individuals are surrounding the optimal solution, and the positions of the remaining individuals are updated according to the optimal solution. If the random number is greater than the preset capture threshold, it means that the remaining individuals are far away from the optimal solution. The exploration behavior is performed through random search, random positions are selected in the population space, and the positions of the remaining individuals are updated. After each iteration, the coefficient vector is updated based on the linear decreasing rule. S3.4: After updating the individual position, check whether it exceeds the search boundary. If the updated individual position exceeds the preset minimum or maximum value, limit it within the boundary, calculate the fitness of the new position, and update the fitness value and position record; S3.5: Repeat the iterative search until the preset maximum number of iterations is reached, then stop the search, record and display the currently found best parameter combination and its corresponding deformation prediction results, apply the found best parameter combination to the current construction plan, and detect the foundation pit excavation construction status in real time to make dynamic parameter adjustments.

5. An optimization control method based on asymmetric excavation deformation of foundation pit, implemented by the optimization control system according to any one of claims 1 to 4, characterized in that: The control method comprises the following steps: The first step is to collect and preprocess historical excavation data and real-time monitoring data; The second step is to construct the foundation pit excavation model and analyze the deformation under complex geological conditions; The third step is to predict deformation during excavation work based on historical data and real-time monitoring data; The fourth step is to simulate different excavation paths and evaluate the safety and effectiveness of each excavation path; The fifth step is to dynamically adjust the excavation strategy and parameters and continuously monitor the actual deformation of the foundation pit.

6. The optimization control method based on asymmetric excavation deformation of foundation pit according to claim 5 is characterized in that: The fourth step specifically includes the following steps: S4.1: Collect each set of construction data of foundation pit excavation, and construct the corresponding state space according to each set of construction data, and each node in the state space represents a state, then construct a set of root nodes, and input the current initial excavation conditions and corresponding deformation data into the root node, and generate new child nodes according to the subsequent feasible excavation path of the root node to simulate the corresponding excavation decision; S4.2: Starting from the root node, the child node with the highest UCB value is selected layer by layer through the UCB strategy until the selected child node is a leaf node that has not been visited and has not been fully expanded. Then the selection stops, and a feasible decision is selected according to the current excavation conditions and environment. A new child node is created for each feasible decision, and its status information is saved; S4.3: Update the environmental state of each child node according to the selected path, and use the previously trained DNN model to input the feature information of each child node for simulation to obtain the corresponding deformation prediction value, and calculate the evaluation value based on the predicted deformation amount and risk threshold. After the simulation is completed, trace the simulation results from the child node back to the root node, and update the evaluation values ​​of all nodes on the path; S4.4: Repeat the selection, expansion, simulation and backtracking steps until the evaluation value changes converge to the preset termination range. Starting from the root node, select the child node with the highest evaluation value as the optimal path, and record the corresponding deformation. At the same time, analyze the selected path and combine the predicted deformation with the safety standard to evaluate its feasibility and risk.

7. The optimization control method based on asymmetric excavation deformation of foundation pit according to claim 6 is characterized in that: The state space described in S4.1 includes excavation depth, excavation path, soil type, surrounding environment, equipment status, deformation history, construction stage, safety threshold, construction strategy, and construction personnel configuration.

8. The optimization control method based on asymmetric excavation deformation of foundation pit according to claim 6 is characterized in that: The fifth step specifically includes the following steps: S4.1: Collect all excavation states of the foundation pit to construct a corresponding state set, and construct a corresponding action set based on different excavation paths and excavation equipment parameters. Then, calculate the probability of transferring to a new state after performing different actions in each state through historical data or simulation. Calculate the instant reward obtained by performing different actions in each state based on the deformation and safety threshold obtained from simulation and historical data. S4.2: Analyze the deformation under different states, and construct a transition probability matrix based on the calculated groups of transition probabilities. Under the current strategy, determine the optimal action to be taken in each state according to the corresponding immediate reward or the result of strategy evaluation, and in each state, select the action that maximizes the value and update the strategy; S4.3: Repeat strategy evaluation and strategy improvement until the immediate reward change range reaches the convergence threshold, stop strategy improvement, extract the final optimal strategy from the iteration results, convert the obtained optimal strategy into an actual excavation plan to guide each decision in the construction process, continuously monitor the deformation during implementation, and adjust the strategy according to the actual situation.

Images