A subway deep foundation pit construction safety early warning and auxiliary decision method and system

Abstract

This invention relates to a method and system for safety early warning and auxiliary decision-making in subway deep foundation pit construction, comprising: acquiring physical data of the subway deep foundation pit; the physical data including foundation pit entity information, foundation pit state information, and foundation pit environmental information; partitioning the subway deep foundation pit into zones based on the physical data using a clustering algorithm; performing geological parameter inversion for each zone by applying a CSA-BPNN neural network and combining it with simulation data output from a digital twin model of the subway deep foundation pit; inputting the predicted geological parameters, foundation pit entity simulation information, and foundation pit environmental simulation information obtained from the digital twin model of the subway deep foundation pit into a foundation pit state prediction model to obtain the predicted foundation pit state value for each zone and perform construction safety early warning and auxiliary decision-making. By refining the partitioning of the foundation pit, state prediction can be performed for each zone, improving the accuracy of state prediction and thus enabling more precise safety early warning and auxiliary decision-making.

Description

Technical Field

[0001] This invention relates to the field of monitoring technology for deep foundation pits in subways, and in particular to a method and system for early warning and auxiliary decision-making in the construction safety of deep foundation pits in subways. Background Technology

[0002] Foundation pit engineering is a type of urban construction project that is time-consuming, costly, and carries a high risk factor. Foundation pits with an excavation depth of 5 meters or more are classified as deep foundation pits. The open-cut method is commonly used in the construction of subway stations.

[0003] The monitoring of foundation pits is mostly handled by third-party monitoring units. The safety warning values ​​for foundation pits are based on GB50497-2009 "Technical Specification for Monitoring of Foundation Pit Engineering". The provisions for the safety warning values ​​for foundation pits are related to the type of foundation pit support and the depth of the foundation pit.

[0004] Currently, the analysis of detection data and the determination of early warning thresholds for foundation pit engineering do not consider spatiotemporal effects. Traditional foundation pit engineering safety early warning systems follow regulations, and the state control thresholds for the entire foundation pit area are uniform. This leads to two problems: firstly, thresholds that are too low in the project cause frequent alarms in localized areas; secondly, thresholds that are too high leave some areas in a dangerous state. There is limited data mining of existing foundation pit construction processes, making it difficult to fully utilize historical data to predict the future state of the foundation pit. This results in a lack of timely strategic guidance and improvement suggestions for foundation pit safety, insufficient analysis and visualization of the effectiveness of emergency rescue and control measures, and a tendency to miss optimal remedial opportunities or have poor remedial control effects, leading to dangerous conditions for the project. Therefore, there is an urgent need for a safety early warning and auxiliary decision-making method and system for deep foundation pit construction in subways that considers the spatiotemporal effects of foundation pits. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for early warning and auxiliary decision-making in the construction safety of deep foundation pits for subways. The method divides the foundation pit space into zones, performs geological parameter inversion and state prediction for each zone, which can greatly improve the accuracy of foundation pit state prediction. Furthermore, a state warning threshold is set for each zone to accurately provide early warning of construction safety and thus assist in decision-making.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A method for safety early warning and auxiliary decision-making in deep foundation pit construction of subways, comprising:

[0008] The physical data of the deep foundation pit of the subway is obtained; the physical data includes foundation pit entity information, foundation pit state information, and foundation pit environment information; the foundation pit entity information is foundation pit size data; the foundation pit state information is foundation pit stress and deformation data;

[0009] Based on the physical data, the K-Means algorithm is applied to partition the deep foundation pit of the subway.

[0010] For each partition, the first CSA-BPNN neural network is trained with the foundation pit entity information, the foundation pit environment information, and the corresponding foundation pit state information as inputs, and the geological parameters are used as labels to obtain a geological parameter prediction model; the geological parameters include elastic modulus and internal friction angle.

[0011] The physical data is input into the digital twin model of the deep foundation pit of the subway for simulation and prediction, and the simulation data is output. The simulation data is then input into the geological parameter prediction model to obtain the geological parameter prediction values.

[0012] The predicted geological parameters, the foundation pit entity information, and the foundation pit environment information are used as inputs, and the foundation pit state information is used as a label to train a second CSA-BPNN neural network to obtain a foundation pit state prediction model.

[0013] The predicted geological parameters, foundation pit entity simulation information, and foundation pit environment simulation information obtained based on the digital twin model of the deep foundation pit of the subway are input into the foundation pit state prediction model to obtain the foundation pit state prediction value for each of the partitions.

[0014] Construction safety early warning and auxiliary decision-making are carried out based on the predicted values ​​of the foundation pit status for each of the aforementioned zones.

[0015] This invention also provides a safety early warning and auxiliary decision-making system for deep foundation pit construction in subways, comprising:

[0016] The foundation pit physical data acquisition module is used to acquire physical data of deep foundation pits in subways; the physical data includes foundation pit entity information, foundation pit state information, and foundation pit environmental information; the foundation pit entity information is foundation pit size data; the foundation pit state information is foundation pit stress and deformation data;

[0017] The partitioning module is used to partition the deep foundation pit of the subway using the K-Means algorithm based on the physical data.

[0018] The geological parameter prediction model construction module is used to train a first CSA-BPNN neural network for each zone, taking the foundation pit entity information, the foundation pit environment information, and the corresponding foundation pit state information as inputs, and using geological parameters as labels, to obtain a geological parameter prediction model; the geological parameters include elastic modulus and internal friction angle.

[0019] The geological parameter prediction module is used to input the physical data into the digital twin model of the deep foundation pit of the subway for simulation prediction, output simulation data, and input the simulation data into the geological parameter prediction model to obtain the geological parameter prediction value;

[0020] The foundation pit state prediction model construction module is used to train a second CSA-BPNN neural network with the predicted geological parameters, the foundation pit entity information and the foundation pit environment information as inputs, and the foundation pit state information as labels, to obtain the foundation pit state prediction model.

[0021] The foundation pit state prediction module is used to input the geological parameter prediction values, foundation pit entity simulation information and foundation pit environment simulation information obtained based on the digital twin model of the deep foundation pit of the subway into the foundation pit state prediction model to obtain the foundation pit state prediction value for each of the partitions.

[0022] The early warning and decision-making module is used to provide early warning and auxiliary decision-making for construction safety based on the predicted values ​​of the foundation pit status of each of the aforementioned zones.

[0023] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0024] This invention relates to a method and system for early warning and auxiliary decision-making in subway deep foundation pit construction, comprising: acquiring physical data of the subway deep foundation pit; the physical data including foundation pit entity information, foundation pit state information, and foundation pit environmental information; dividing the subway deep foundation pit into zones based on the physical data using the K-Means algorithm; performing geological parameter inversion for each zone by applying a first CSA-BPNN neural network and combining simulation data output from a digital twin model of the subway deep foundation pit; inputting the predicted geological parameters, foundation pit entity simulation information, and foundation pit environmental simulation information obtained based on the digital twin model of the subway deep foundation pit into a foundation pit state prediction model to obtain the predicted foundation pit state value for each zone; and performing construction safety early warning and auxiliary decision-making based on the predicted foundation pit state value for each zone. By refining the zoning of the foundation pit, state prediction can be performed for each zone, improving the accuracy of state prediction and thus enabling more precise safety early warning and auxiliary decision-making. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 Here is a flowchart of a safety early warning and auxiliary decision-making method for deep foundation pit construction in subways, provided in Embodiment 1 of the present invention;

[0027] Figure 2 A flowchart for training a second CSA-BPNN neural network is provided for Embodiment 1 of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The purpose of this invention is to provide a method and system for early warning and auxiliary decision-making in the construction safety of deep foundation pits for subways. The method divides the foundation pit space into zones, performs geological parameter inversion and state prediction for each zone, which can greatly improve the accuracy of foundation pit state prediction. Furthermore, a state warning threshold is set for each zone to accurately provide early warning of construction safety and thus assist in decision-making.

[0030] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Example 1

[0032] like Figure 1 As shown in the figure, this embodiment provides a method for safety early warning and auxiliary decision-making in subway deep foundation pit construction, including:

[0033] S1: Obtain physical data of the deep foundation pit of the subway; the physical data includes foundation pit entity information, foundation pit status information and foundation pit environment information; the foundation pit entity information is foundation pit size data; the foundation pit status information is foundation pit stress and deformation data.

[0034] The physical data is obtained by various sensors from the on-site physical model (the actual foundation pit). The amount of data increases over time and includes information about the foundation pit entity, foundation pit status, and the surrounding environment of the foundation pit.

[0035] The physical information of the foundation pit refers to the spatial geometric dimensions of the foundation pit and the dimensions and spatial location of the designed retaining structure. The spatial geometric dimensions of the foundation pit include the length, width, and depth of the foundation pit. The dimensions and spatial location of the retaining structure include the type of retaining structure (diaphragm wall, internal bracing, continuous piles, etc.), the length of the retaining structure, the dimensions (length, width, and height) of the internal bracing, and the relative position of the foundation pit.

[0036] The foundation pit status information includes the deformation, stress, and response of the foundation pit and its retaining structure after excavation, as well as the location information of measuring points. Deformation includes (horizontal displacement and top vertical displacement of the retaining structure, vertical deformation of the soil at the bottom of the pit, vertical deformation of the soil outside the pit, and horizontal deformation of the deep soil outside the pit); deformation and stress of the retaining structure (horizontal and vertical deformation of the internal supports, and axial force of the internal supports); and location information of measuring points (their position within the foundation pit, such as horizontal and vertical positions). The foundation pit status information is constructed by monitoring foundation pit inclination, internal support axial force, anchor cable stress, groundwater level, surface deformation, 3D point cloud data of the foundation pit, weather information (temperature, rainfall, etc.), surrounding loads, and tilt and settlement data of adjacent structures.

[0037] The environmental information of the foundation pit refers to the surrounding environment of the foundation pit. The environmental information is divided into hydrogeology, external loads, and nearby sensitive structures (risk projects). Among them, hydrogeology includes strata (stratum type, stratum parameters (resilience modulus, Poisson's ratio, unit weight, water content, internal friction angle, cohesion, dilatation angle, etc.), stratum depth, groundwater level, whether it contains confined water, etc.); external loads (surcharge around the foundation pit, surcharge location); and nearby sensitive structures (risk level, control standards, spatial location of the foundation pit (horizontal distance, vertical distance, etc.)).

[0038] Based on existing engineering information (physical data from the physical model), a physical model database for the foundation pit is established. However, due to insufficient data in the physical model database, the accuracy of the prediction model will be affected, thus requiring data expansion. Therefore, step 2 includes the following:

[0039] (1) Establish digital models of foundation pits with different support types;

[0040] Using FLAC3D finite difference software, numerical models of various common retaining structures (retaining wall support, pile support, diaphragm wall internal support, diaphragm wall anchor cable) were established for foundation pits consisting of end wells and intermediate conventional sections. Environmental factors were taken into account in the numerical models, and template libraries were created by encapsulating the digital model calculations of various types of supports.

[0041] (2) Multi-factor level tests were conducted on the digital models of the foundation pit with different support types to obtain test data.

[0042] Based on the encapsulated digital model template, a multi-factor level experimental design is adopted.

[0043] (3) The experimental data and the physical data are merged into a subsequent training database. The merged data is used to partition the deep foundation pit of the subway, and to train the first CSA-BPNN neural network.

[0044] S2: Based on the physical data, the K-Means algorithm is applied to partition the deep foundation pit of the subway. To improve the accuracy of the clustering results, the physical data here can be merged data.

[0045] Specifically, step S2 includes:

[0046] S21: Principal component analysis is used to preprocess the merged data. The preprocessing includes dimensionality reduction, and the removal and recombination of redundant data.

[0047] S22: Extract the data containing the preset percentage information from the preprocessed data and apply the K-Means algorithm to partition the deep foundation pit of the subway. For example, use the parameter representing 90% of the overall information as the input data for the K-Means algorithm, and calculate the number of spatial effect partitions and the distribution of partition parameters of the foundation pit according to K-means.

[0048] K-means algorithm:

[0049] a. Determine the optimal clustering parameter K for the K-means clustering algorithm: Use the Gap statistic method, Gap(K) = E(logD) k )-logD k D k Let E(logD) be the loss function. k ) refers to logD k The expected value is obtained through Monte Carlo simulation. A random sample library with the same number of samples as the original sample library is randomly generated in a uniform distribution within the region containing the historical foundation pit condition sample library. This random sample library is then adjusted by k (the number of spatial effect zones for the foundation pit) and K-Means is applied to obtain a partitioning error D. k Repeating this process many times, usually 20 times, yields 20 logD values. k The average of these 20 values ​​gives E(logD). k The approximate value of ) can be obtained. Finally, the Gap statistic can be calculated. The k corresponding to the maximum value of the Gap statistic is the optimal number of spatial effect zones for the foundation pit.

[0050] b. For each foundation pit historical status dataset, calculate the distance from the sample to the k cluster centers for foundation pit samples at different locations, and classify them according to their distance to the k cluster centers, assigning them to the category represented by the k with the smallest distance;

[0051] c. Adjust the positions of k cluster centers, change the positions of k cluster centers, and repeat step b until the termination condition is met (the total distance from all points to their classification center k is minimized). At this point, the number of pit spatial effect partitions k and the corresponding pit locations in space are obtained.

[0052] S3: For each partition, the first CSA-BPNN neural network is trained with the foundation pit entity information, the foundation pit environment information, and the corresponding foundation pit state information as inputs and the geological parameters as labels to obtain a geological parameter prediction model; the geological parameters include elastic modulus and internal friction angle.

[0053] Low-sensitivity parameters were obtained from geological reports, while parameters with greater impact were obtained through parameter inversion. To reduce computational load in geological parameter inversion, the foundation pit entity information and foundation pit environmental information, which have little impact on the final foundation pit state information, were not used as target parameters for parameter inversion. The foundation pit state information in the training model was used as the target, and the foundation pit entity information and foundation pit environmental information were used as influencing factors. The correlation between the parameters of the two was analyzed. The Morris global sensitivity parameter calculation method was used to calculate the sensitivity parameters of each parameter (foundation pit entity and foundation pit environmental information parameters) corresponding to the foundation pit state information. Based on the sensitivity parameters, the main and secondary influencing factors were determined. Parameter sensitivity analysis was conducted at different state stages and in different foundation pit areas, and the analysis results were used as reference parameters for zonal geological parameter inversion.

[0054] Sensitivity analysis of the parameters is performed to identify the main influencing parameters and reduce the computational load of subsequent parameter inversion. Step S3 specifically includes:

[0055] For each partition, the key foundation pit entity information and key foundation pit environmental information in the physical data are determined using the parameter sensitivity analysis method.

[0056] The formula used in the sensitivity analysis is as follows:

[0057]

[0058] In the formula: q is the total number of levels in the level test; Y i Y is the calculated value for the i-th level; i+1 The calculated value for level i+1; P i P represents the parameter value at the i-th level. i+1 Y0 is the initial value of the (i+1)th level parameter; P0 is the initial value of the calculation result; P0 is the initial value of the calculation parameter.

[0059] Using the key foundation pit entity information, the key foundation pit environmental information, and the corresponding foundation pit state information as inputs, and the geological parameters as labels, a first CSA-BPNN neural network is trained to obtain the geological parameter prediction model.

[0060] The Clonal Selection Algorithm (CSA) optimizes the weights, thresholds, and number of neurons per layer to adjust the negative feedback neural network (BPNN) for predicting geological parameters in each area. The response values ​​(stress and deformation) of the retaining structure in the foundation pit area are used as input, and the geological parameters (elastic modulus, internal friction angle, etc.) are used as output. The established training database is used for model training.

[0061] S4: Input the physical data into the digital twin model of the deep foundation pit of the subway for simulation prediction, output the simulation data, and input the simulation data into the geological parameter prediction model to obtain the geological parameter prediction value.

[0062] A digital twin model is constructed based on a physical model of a deep foundation pit. After physical data is input into the digital twin model, it can contain the attribute information of the physical entity of the foundation pit and visualize the model's operation process. The digital twin model of the subway deep foundation pit performs simulation based on the input physical data and outputs simulation data in real time.

[0063] S5: Using the predicted geological parameters, the foundation pit entity information, and the foundation pit environment information as inputs, and the foundation pit state information as labels, train a second CSA-BPNN neural network to obtain a foundation pit state prediction model.

[0064] CSA-BPNN algorithm: (taking deformation prediction as an example)

[0065] a. Create a BPNN network, using an N-layer neural network model. The input and hidden layers are connected using the ELU (Exponential LinearUnits) function, as follows:

[0066]

[0067] The training function is chosen as traingd, and the hidden layer transfer function and the hidden layer-output layer function are chosen as Tanh functions.

[0068] The input parameters are the geometric, mechanical, and geological parameters of the foundation pit, and the output parameter is the displacement response of the foundation pit.

[0069] b. Use the learning rate η, the number of hidden layers N, and the number of neurons per layer mi of the BPNN neural network as the optimization objective function of the CSA algorithm.

[0070] c. CSA-BPNN model training: A training dataset is established by normalizing the training database. Based on the established dataset, the dataset is divided into three sets: training set, validation set, and test set, with a ratio of 7:1.5:1.5.

[0071] d. Determine the optimal parameters of the BPNN model using the training and validation sets, and set the learning rate η, the number of hidden layers N, and the number of neurons per layer m. i Train the model M-times under these parameters, and calculate the error of the model trained under these parameters using the validation set. If the error does not meet the preset error, adjust the new learning rate η, the number of hidden layers N, and the number of neurons per layer m. i Then retrain for M times until the preset error is met, or the prediction error of the trained model reaches its minimum value, and output the optimal learning rate η, the number of hidden layers N, and the number of neurons per layer m. i .

[0072] e. Based on optimal network parameters (learning rate η, number of hidden layers N, and number of neurons per layer m) i We perform optimal training of BPNN with the best hyperparameters and network structure parameters under this architecture (equivalent to the parameters of the model) to improve model training efficiency and prediction accuracy. After the model training is completed, we test the model.

[0073] Explanation: The training accuracy and fit of BPNN are mainly related to its network structure. Therefore, it is necessary to select the grid structure parameters suitable for foundation pit deformation prediction under this training database. Thus, through continuous trial calculations at different levels and CSA assistance in adjusting the parameters (optimization, CSA is the clone selection algorithm), the goal is to find the network architecture most suitable for this training database.

[0074] The established training dataset is combined with geological parameters, foundation pit entity, and foundation pit environment parameters obtained by inverting the regional geological parameters as inputs and foundation pit state parameters as outputs to train the established CSA-BPNN model until the target accuracy is achieved, thus completing the training of the prediction model.

[0075] S6: Input the predicted geological parameters, foundation pit entity simulation information, and foundation pit environment simulation information output from the digital twin model of the deep foundation pit of the subway into the foundation pit state prediction model to obtain the foundation pit state prediction value for each of the partitions.

[0076] Based on the CSA-BPNN foundation pit zoning parameters, the geological parameters are calibrated by inversion. The calibrated geological parameters, foundation pit entity and environmental information are then substituted into the trained model to predict the foundation pit status information for the next construction stage. At the same time, the foundation pit status information of the zoning can also be predicted through the encapsulated digital modeling template library (digital model calculations for various types of support are encapsulated).

[0077] S7: Conduct construction safety early warning and auxiliary decision-making based on the predicted values ​​of the foundation pit status for each of the aforementioned zones.

[0078] Specifically, step S7 includes:

[0079] (1) Based on the physical model of the deep foundation pit of the subway, set a foundation pit status early warning value for each of the partitions.

[0080] Based on the safety level of the foundation pit and the safety protection level of the surrounding environment, the safety status threshold of the foundation pit in each area is selected.

[0081] (2) Compare the predicted value of the foundation pit status of each partition with the corresponding early warning value of the foundation pit status to conduct construction safety early warning and auxiliary decision-making.

[0082] When the predicted state value of the foundation pit exceeds the set safety threshold, the frequency of on-site foundation pit monitoring increases, and early warning information is promptly sent to the engineering construction, design, construction and management parties through the server.

[0083] Construction support decision-making after a safety risk warning for the foundation pit (e.g., on-site construction causes the foundation pit to be in a dangerous state; through digital twin model investigation, it is found that the state exceeds the limit due to a sudden rise in groundwater during on-site construction; construction control measures are provided for this type of case; combining the current foundation pit state information, entity and environmental information, it is analyzed which control measures have the highest control efficiency, providing construction technicians with construction decision support. The possible causes of risk are summarized below, and will be supplemented in a timely manner if situations exceeding this coverage occur later. Because the auxiliary measures are added after the risk points appear, it affects the application of the second CSA-BPNN foundation pit state prediction method mentioned above; for such special cases, an encapsulated digital modeling template library is used).

[0084] Emergency Assistance Decision Making After Foundation Pit Warning:

[0085] a. Hydrogeological Risks: When the exposed geological conditions suddenly deviate from the survey during excavation, leading to an increase in the pit's condition, the control effect can be enhanced by strengthening local supports and supplementing with localized grouting reinforcement. When extreme rainfall or other factors cause a rise in the groundwater level within the pit's impact area, surface water diversion and pit drainage should be increased, while strengthening pit condition monitoring. If necessary, local structural support should be reinforced to prevent pit collapse. A digital modeling template library is used to predict pit conditions caused by geological abrupt changes and groundwater level fluctuations, and the condition control effect after implementing auxiliary measures is analyzed.

[0086] b. Surrounding environment risks: Consider the surrounding underground pipelines, buildings, tunnels, etc. Before construction, formulate a construction plan based on risk rating and expert opinions, strictly control the state of existing structures caused by foundation pit excavation, strengthen foundation pit monitoring, and if necessary, take measures such as grouting to reinforce the strata, adding foundation pit steel supports, and reinforcing existing structures. Analyze the state control effect through a digital modeling template library.

[0087] c. Key protected projects: When the foundation pit project is being carried out near ancient buildings, precision laboratories or major projects, strengthen the monitoring of the construction process, and take necessary auxiliary measures in a timely manner based on the foundation pit status monitoring data and early warning information: strengthen the support (control the source of disturbance), use isolation piles (block the propagation path) and reinforce the existing structure (reinforce the existing structure) to ensure the safety of key protected projects, and analyze the status control effect through the digital modeling template library.

[0088] d. Foundation pit support failure and foundation pit edge overload: When the foundation pit is overloaded or exceeds the limit, the monitoring and measurement information should be analyzed, and experts should be organized in a timely manner to analyze and judge the early warning information and the historical information of the foundation pit to determine whether the foundation pit retaining structure is stable and whether it is necessary to add auxiliary support measures. The status control effect should be analyzed through the digital modeling template library to avoid possible foundation pit construction risks in a timely manner.

[0089] e. Freezing: Foundation pit projects inevitably encounter winter construction. Problems such as incomplete compaction, spalling, low strength, and anchor cable failure caused by winter construction can lead to ground settlement and cracking. Detailed special plans for winter construction should be adopted to avoid major safety hazards.

[0090] In this embodiment, 1. Zoning based on the spatiotemporal effect of the foundation pit (zoning prediction and early warning based on the state patterns and characteristics of different areas of the foundation pit); 2. Prediction of foundation pit state zoning: using dynamic monitoring data and artificial intelligence algorithms to correct soil parameters, and performing dynamic prediction of foundation pit state through artificial intelligence algorithms and encapsulated digital models; 3. Foundation pit safety early warning and auxiliary decision-making (by combining the inversion of zoning geological parameters with the encapsulated digital modeling template library and the foundation pit state prediction of the second CSA-BPNN, foundation pit safety prediction, auxiliary decision-making, and prediction of the effectiveness of risk classification auxiliary measures are realized).

[0091] Existing foundation pit prediction methods cannot take into account the spatiotemporal effects of foundation pits. Traditional prediction algorithms do not consider the spatial effects of foundation pit geology, resulting in foundation pits not being predicted and controlled in a segmented and detailed manner. This also fails to evaluate the effectiveness of construction auxiliary measures, which is not conducive to rapid emergency decision-making after foundation pit early warning.

[0092] Therefore, compared with the prior art, the method of this embodiment combines the characteristics of the foundation pit state to establish a zoning method based on the spatial effect of the foundation pit, and performs refined monitoring, accurate prediction and early warning of the foundation pit state in different areas; it uses dynamic monitoring data to calibrate the geological parameters of the foundation pit, so as to achieve the accuracy of the foundation pit state prediction, and combines encapsulated digital modeling to predict the state and auxiliary effects, so as to provide a reference for auxiliary decision-making; it establishes a responsive early warning and emergency auxiliary decision-making system, which improves the safety of deep foundation pit construction in subways.

[0093] Example 2

[0094] This embodiment provides a safety early warning and auxiliary decision-making system for deep foundation pit construction in subways, including:

[0095] The foundation pit physical data acquisition module T1 is used to acquire physical data of the deep foundation pit of the subway; the physical data includes foundation pit entity information, foundation pit state information and foundation pit environment information; the foundation pit entity information is foundation pit size data; the foundation pit state information is foundation pit stress and deformation data;

[0096] The partitioning module T2 is used to partition the deep foundation pit of the subway based on the physical data using the K-Means algorithm.

[0097] Specifically, the partitioning module T2 includes:

[0098] The data preprocessing unit is used to preprocess the merged data using principal component analysis. The preprocessing includes dimensionality reduction, and the removal and recombination of redundant data.

[0099] The partitioning unit is used to extract data containing a preset percentage from the preprocessed data and apply the K-Means algorithm to partition the deep foundation pit of the subway.

[0100] The geological parameter prediction model construction module T3 is used to train a first CSA-BPNN neural network for each partition, taking the foundation pit entity information, the foundation pit environment information, and the corresponding foundation pit state information as inputs, and using geological parameters as labels, to obtain a geological parameter prediction model; the geological parameters include elastic modulus and internal friction angle.

[0101] Specifically, the geological parameter prediction model construction module T3 includes:

[0102] The key parameter acquisition unit is used to determine the key foundation pit entity information and key foundation pit environmental information in the physical data for each partition using parameter sensitivity analysis methods.

[0103] The training unit is used to train a first CSA-BPNN neural network with the key foundation pit entity information, the key foundation pit environmental information, and the corresponding state information as inputs, and the geological parameters as labels, to obtain the geological parameter prediction model.

[0104] The geological parameter prediction module T4 is used to input the physical data into the digital twin model of the deep foundation pit of the subway for simulation prediction, output simulation data, and input the simulation data into the geological parameter prediction model to obtain the geological parameter prediction value.

[0105] The foundation pit state prediction model construction module T5 is used to train a second CSA-BPNN neural network with the predicted geological parameters, the foundation pit entity information, and the foundation pit environment information as inputs, and the foundation pit state information as labels, to obtain the foundation pit state prediction model.

[0106] The foundation pit state prediction module T6 is used to input the geological parameter prediction values, foundation pit entity simulation information and foundation pit environment simulation information obtained based on the digital twin model of the deep foundation pit of the subway into the foundation pit state prediction model to obtain the foundation pit state prediction value for each of the partitions.

[0107] The early warning and decision-making module T7 is used to provide early warning and auxiliary decision-making for construction safety based on the predicted value of the foundation pit status of each of the aforementioned zones.

[0108] Specifically, the early warning and decision-making module T7 includes:

[0109] The status warning threshold setting unit is used to set a foundation pit status warning value for each of the deep foundation pits of the subway based on the physical model of the foundation pit.

[0110] The early warning and decision-making unit is used to compare the predicted value of the foundation pit status of each of the aforementioned partitions with the corresponding early warning value of the foundation pit status to conduct construction safety early warning and auxiliary decision-making.

[0111] The system further includes: a data expansion module T8; the data expansion module T8 includes:

[0112] The digital model building unit T81 for foundation pits is used to establish digital models of foundation pits with different support types.

[0113] The test data acquisition unit T82 is used to conduct multi-factor level tests on the digital model of the foundation pit with different support types to obtain test data.

[0114] The data merging unit T83 is used to merge the experimental data and the physical data; the merged data is used to partition the deep foundation pit of the subway and to train the first CSA-BPNN neural network and the second CSA-BPNN neural network.

[0115] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0116] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for safety early warning and auxiliary decision-making in deep foundation pit construction of subways, characterized in that, include: The physical data of the deep foundation pit of the subway is obtained; the physical data includes foundation pit entity information, foundation pit state information, and foundation pit environment information; the foundation pit entity information is foundation pit size data; the foundation pit state information is foundation pit stress and deformation data; Based on the physical data, the K-Means algorithm is applied to partition the deep foundation pit of the subway. For each partition, the first CSA-BPNN neural network is trained with the foundation pit entity information, the foundation pit environment information, and the corresponding foundation pit state information as inputs, and the geological parameters are used as labels to obtain a geological parameter prediction model; the geological parameters include elastic modulus and internal friction angle. The physical data is input into the digital twin model of the deep foundation pit of the subway for simulation and prediction, and the simulation data is output. The simulation data is then input into the geological parameter prediction model to obtain the geological parameter prediction values. Using the predicted geological parameters, the foundation pit entity information, and the foundation pit environment information as inputs, and the foundation pit state information as labels, a second CSA-BPNN neural network is trained to obtain a foundation pit state prediction model. The predicted geological parameters, foundation pit entity simulation information, and foundation pit environment simulation information obtained based on the digital twin model of the deep foundation pit of the subway are input into the foundation pit state prediction model to obtain the foundation pit state prediction value for each of the partitions. Construction safety early warning and auxiliary decision-making are carried out based on the predicted values ​​of the foundation pit status of each of the aforementioned zones; The process of partitioning the deep foundation pit of the subway based on the physical data using the K-Means algorithm also includes: Establish digital models of foundation pits with different support types; Multi-factor level tests were conducted on the digital models of the foundation pit with different support types to obtain test data; The experimental data and the physical data are merged; the merged data is used to partition the deep foundation pit of the subway and to train the first CSA-BPNN neural network and the second CSA-BPNN neural network. Specifically, the step of applying the K-Means algorithm to partition the deep foundation pit of the subway based on the physical data includes: Principal component analysis is used to preprocess the merged data. The preprocessing includes dimensionality reduction, and the removal and recombination of redundant data. The data containing the preset percentage information in the preprocessed data is extracted and the K-Means algorithm is applied to partition the deep foundation pit of the subway.

2. The method according to claim 1, characterized in that, For each partition, the first CSA-BPNN neural network is trained using the foundation pit entity information, the foundation pit environment information, and the corresponding state information as inputs, and geological parameters as labels, to obtain a geological parameter prediction model, specifically including: For each partition, the key foundation pit entity information and key foundation pit environmental information in the physical data are determined using the parameter sensitivity analysis method. Using the key foundation pit entity information, the key foundation pit environmental information, and the corresponding foundation pit state information as inputs, and the geological parameters as labels, a first CSA-BPNN neural network is trained to obtain the geological parameter prediction model.

3. The method according to claim 1, characterized in that, The process of providing construction safety early warning and auxiliary decision-making based on the predicted foundation pit status values ​​for each of the aforementioned zones specifically includes: Based on the physical model of the deep foundation pit of the subway, a foundation pit status early warning value is set for each of the aforementioned zones; The predicted value of the foundation pit status for each partition is compared with the corresponding early warning value of the foundation pit status to provide early warning and auxiliary decision-making for construction safety.

4. A subway deep foundation pit construction safety early warning and auxiliary decision-making system based on the subway deep foundation pit construction safety early warning and auxiliary decision-making method according to any one of claims 1 to 3, characterized in that, include: The foundation pit physical data acquisition module is used to acquire physical data of deep foundation pits in subways; the physical data includes foundation pit entity information, foundation pit state information, and foundation pit environmental information; the foundation pit entity information is foundation pit size data; the foundation pit state information is foundation pit stress and deformation data; The partitioning module is used to partition the deep foundation pit of the subway using the K-Means algorithm based on the physical data. The geological parameter prediction model construction module is used to train a first CSA-BPNN neural network for each zone, taking the foundation pit entity information, the foundation pit environment information, and the corresponding state information as inputs, and using geological parameters as labels, to obtain a geological parameter prediction model; the geological parameters include elastic modulus and internal friction angle. The geological parameter prediction module is used to input the physical data into the digital twin model of the deep foundation pit of the subway for simulation prediction, output simulation data, and input the simulation data into the geological parameter prediction model to obtain the geological parameter prediction value; The foundation pit state prediction model construction module is used to train a second CSA-BPNN neural network with the predicted geological parameters, the foundation pit entity information and the foundation pit environment information as inputs, and the foundation pit state information as labels, to obtain the foundation pit state prediction model. The foundation pit state prediction module is used to input the geological parameter prediction values, foundation pit entity simulation information and foundation pit environment simulation information obtained based on the digital twin model of the deep foundation pit of the subway into the foundation pit state prediction model to obtain the foundation pit state prediction value for each of the partitions. The early warning and decision-making module is used to provide early warning and auxiliary decision-making for construction safety based on the predicted values ​​of the foundation pit status of each of the aforementioned zones. The system further includes: a data expansion module; the data expansion module includes: The foundation pit digital model building unit is used to establish foundation pit digital models with different support types; The test data acquisition unit is used to conduct multi-factor level tests on the digital model of the foundation pit with different support types to obtain test data. The data merging unit is used to merge the experimental data and the physical data; the merged data is used to partition the deep foundation pit of the subway and to train the first CSA-BPNN neural network and the second CSA-BPNN neural network. Specifically, the partitioning module includes: The data preprocessing unit is used to preprocess the merged data using principal component analysis. The preprocessing includes dimensionality reduction, and the removal and recombination of redundant data. The partitioning unit is used to extract data containing a preset percentage from the preprocessed data and apply the K-Means algorithm to partition the deep foundation pit of the subway.

5. The system according to claim 4, characterized in that, The geological parameter prediction model construction module specifically includes: The key parameter acquisition unit is used to determine the key foundation pit entity information and key foundation pit environmental information in the physical data for each partition using parameter sensitivity analysis methods. The training unit is used to train a first CSA-BPNN neural network with the key foundation pit entity information, the key foundation pit environmental information, and the corresponding state information as inputs, and the geological parameters as labels, to obtain the geological parameter prediction model.

6. The system according to claim 4, characterized in that, The early warning and decision-making module specifically includes: The status warning threshold setting unit is used to set a foundation pit status warning value for each of the deep foundation pits of the subway based on the physical model of the foundation pit. The early warning and decision-making unit is used to compare the predicted value of the foundation pit status of each of the aforementioned partitions with the corresponding early warning value of the foundation pit status to conduct construction safety early warning and auxiliary decision-making.

Images