Prediction and early warning method of surface subsidence based on neural network

Abstract

The present invention provides a prediction and early warning method for surface settlement based on a neural network, which relates to the technical field of foundation pit settlement prediction, including: obtaining the structural data of the foundation pit and the soil within a preset range around the foundation pit, and at the same time, obtaining the monitoring data at multiple preset points in the foundation pit in real time, summarizing and outputting the initial data; selecting historical data matching the initial data, outputting the historical reference data, and combining the preset analysis model to conduct a comprehensive analysis of the initial data and the historical reference data, and outputting the real-time analysis results; combining the preset prediction model to conduct a prediction analysis on the initial data and the historical reference data, outputting the prediction analysis results, and combining the real-time analysis results to obtain the real-time state and the predicted state of the foundation pit structure in the result-state comparison table, outputting the state analysis table and conducting an early warning judgment according to the preset threshold conditions. The present invention can predict and warn the surface settlement data around the foundation pit during the construction period in real time, and take corresponding preventive measures in time.

Description

Technical Field

[0001] The invention relates to the technical field of foundation pit settlement prediction, and in particular to a prediction and early warning method for surface settlement based on a neural network. Background Art

[0002] With the development of cities, it is becoming more and more common to carry out underground construction projects such as subways, civil air defense, and high-rise building basements in complex environments with dense buildings and crisscrossing pipelines. Deep foundation pit construction will inevitably have adverse effects on existing adjacent foundation pit buildings and municipal pipelines. In order to protect the buildings, municipal pipelines and other facilities adjacent to the foundation pit, it is necessary to further explore the movement rules of the surrounding strata caused by deep foundation pit excavation, so as to more accurately predict the differential settlement caused by buildings, etc., and thus predict the impact of deep foundation pit excavation on buildings, etc. Therefore, it is of great practical significance to determine the surface settlement caused by foundation pit excavation.

[0003] At present, in the field of civil engineering, we often use formula methods or finite element software to predict settlement and other monitoring data. However, the soil conditions and boundary conditions are relatively complex and difficult to determine, and the prediction accuracy is low, making it difficult to control the potential settlement risks in the foundation pit in a timely manner.

[0004] Therefore, the present invention provides a prediction and early warning method for surface subsidence based on a neural network. Summary of the invention

[0005] The present invention provides a prediction and early warning method for surface subsidence based on a neural network, which is used to monitor and analyze the structure of the foundation pit in real time during the construction period. At the same time, the surface subsidence data is predicted and risk warning is carried out, and corresponding risk control measures are taken in time to reduce the impact of foundation pit construction on the surrounding environment.

[0006] The present invention provides a prediction and early warning method for surface subsidence based on a neural network, comprising:

[0007] Step 1: Obtain the structural data of the foundation pit and the soil within the preset range around the foundation pit. At the same time, obtain the monitoring data of multiple preset points in the foundation pit in real time through the preset monitoring equipment, and summarize and output the initial data;

[0008] Step 2: Select historical data matching the initial data from the historical database and output the historical reference data;

[0009] Step 3: Perform a comprehensive analysis on the initial data and historical reference data in combination with a preset analysis model, and output real-time analysis results;

[0010] Step 4: Perform predictive analysis on the initial data and historical reference data in combination with a preset predictive model, output the predictive analysis results, and obtain the real-time status and predicted status of the foundation pit structure in the result-status comparison table in combination with the real-time analysis results, output the status analysis table and make early warning judgments based on preset threshold conditions.

[0011] Preferably, the obtaining of structural data of the foundation pit and soil within a preset range around the foundation pit includes:

[0012] Obtain the construction duration data of the foundation pit and divide it into stages to obtain multiple construction sub-stages;

[0013] The surface settlement data of the foundation pit in each construction sub-stage and within each preset distance range around the foundation pit are obtained, and the surface settlement data of the construction stage are summarized and output.

[0014] Preferably, the real-time acquisition of monitoring data at a plurality of preset points in the foundation pit by using preset monitoring equipment includes:

[0015] Acquire first position data corresponding to the first support and the second support of the foundation pit during the construction process;

[0016] In combination with the first position data, multiple first preset points are respectively set at the first support and the second support, and a monitoring device array is constructed by installing preset monitoring devices at each of the first preset points. Based on the monitoring device array, the support axial force data of the first support and the second support during the construction of the foundation pit are obtained in real time, and the support axial force change data is output.

[0017] Preferably, the real-time acquisition of monitoring data at a plurality of preset points in the foundation pit by using a preset monitoring device also includes:

[0018] Acquire the structural data and second position data of the ground-connected wall enclosure structure of the foundation pit during the construction phase, and set a plurality of second preset points;

[0019] The displacement data of the ground-connected wall retaining structure at each of the second preset points during the construction of the foundation pit is obtained through the preset monitoring equipment, and the structure displacement change data is output.

[0020] Preferably, the aggregating and outputting the initial data includes:

[0021] Obtaining the timestamps corresponding to the structure data and the monitoring data, performing time alignment and information annotation, and summarizing and outputting the first data;

[0022] The first data is preprocessed using a preset preprocessing method to output initial data.

[0023] Preferably, step 2 includes:

[0024] Extracting features from the initial data, performing feature matching on the extracted data features in combination with a preset sedimentation feature database, and outputting sedimentation feature matching results;

[0025] Combined with the settlement feature matching result, historical foundation pit settlement data that meets preset screening conditions is screened in the historical database, and historical reference data is output.

[0026] Preferably, step 3 includes:

[0027] Extracting features from the initial data, and constructing a current foundation pit feature set based on the extracted features;

[0028] Based on the current foundation pit feature set, data satisfying a first screening condition is selected from the historical reference data, and output as comparison data;

[0029] Obtaining the contrast between each of the comparison data and the initial data, and sorting and dividing the comparison data according to a preset sorting rule to obtain a plurality of groups of comparison sub-data with different comparison ranges;

[0030] At the same time, feature extraction is performed on each of the comparison data, and a comparison feature set is constructed based on the extracted features, and multiple data analysis models that meet the user's data analysis needs are selected from the model database in combination with the current foundation pit feature set;

[0031] The initial data and the comparison sub-data are combined according to a preset combination rule to obtain combined data under multiple combinations, and the combined data are respectively input into each of the data analysis models for real-time analysis, and the analysis results of the multiple data analysis models under each of the combined data are summarized to obtain a multi-model analysis result under each of the combined data;

[0032] Performing content analysis on the multi-model analysis results under each of the combined data to obtain analyzed data, and performing curve fitting and data annotation on the analyzed data using a preset curve construction algorithm to obtain a model curve-data graph corresponding to each data analysis model under each of the combined data;

[0033] The curve-data graphs under each combination of data are time-aligned, and the curve-data graphs that meet the preset data comparison conditions are compared and analyzed and data consensus is achieved to obtain a multi-model consensus result under each combination of data, and the real-time analysis results are summarized.

[0034] Preferably, step 4 includes:

[0035] In combination with the data types corresponding to the data in the historical reference data, the historical reference data are classified to obtain reference sub-data under each type, and the sub-data are summarized to obtain the data to be classified;

[0036] Using a preset data partitioning method to partition the data to be partitioned, to obtain a training data set, a verification data set, and a test data set;

[0037] Based on the training data set, the validation data set and the test data set, multiple first models selected from the model database are trained and optimized, and the first models that meet the preset performance conditions are output as preset prediction models;

[0038] By using at least two of the preset prediction models to simultaneously perform prediction analysis on the initial data, surface settlement data, support axial force data and support structure displacement data of the foundation pit in a future preset period of time, and output a model-prediction data comparison table;

[0039] A consensus is reached on the prediction results in the model-prediction data comparison table, and a prediction consensus result is output;

[0040] Combine the preset evaluation indicators and use the preset grade classification rules to establish a risk grade comparison table, and determine the corresponding fall-in judgment conditions for each risk grade;

[0041] Among them, the corresponding fall-in judgment conditions for each risk level are determined as follows:

[0042] ;

[0043] in, represents the critical Euclidean distance corresponding to the k-th risk level under the i-th preset evaluation indicator; Indicates the total number of preset evaluation indicators; represents the quantitative value corresponding to the j-th risk level under the i-th preset evaluation indicator, where i=1,2,…m; j=1,2,…n; represents the calculation weight corresponding to the i-th preset evaluation indicator;

[0044] Combined with the preset evaluation indicators, the real-time analysis results and the predicted consensus results are analyzed by a preset risk level evaluation algorithm to obtain the risk evaluation data of the current foundation pit structure, and compared with the fall-in judgment conditions corresponding to each level in the risk level comparison table to determine the real-time risk level of the current foundation pit structure and the predicted risk level within a preset period of time in the future, and output the risk status analysis results;

[0045] The preset risk level assessment algorithm is as follows:

[0046] ;

[0047] in, represents the risk assessment value corresponding to the i-th assessment indicator, which is used for comparative analysis with the critical Euclidean distance; represents the mean value corresponding to all statistical data under the i-th evaluation indicator; Represents the standard deviation corresponding to all statistical data under the i-th preset evaluation indicator;

[0048] Compare and analyze the risk status analysis results with the preset threshold conditions to determine the type of warning data, and select the prevention strategy and prevention measures corresponding to the warning data type in the strategy database;

[0049] Based on the above-mentioned prevention strategies and measures, the corresponding structures in the current foundation pit structure are transformed.

[0050] The neural network-based surface settlement prediction and early warning method provided by the present invention realizes real-time monitoring and early warning of the foundation pit structure status by combining real-time monitoring data, historical data and preset models for comprehensive analysis and prediction. The present invention improves the prediction accuracy of surface settlement during the construction of foundation pit engineering, effectively reduces the risks that may occur in foundation pit engineering, helps engineering personnel to find problems in time and take measures, thereby ensuring the safety and stability of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 It is a flow chart of a method for predicting and warning ground subsidence based on a neural network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] like Figure 1 As shown, the prediction and early warning method of surface subsidence based on neural network provided by the embodiment of the present invention includes:

[0055] Step 1: Obtain the structural data of the foundation pit and the soil within the preset range around the foundation pit. At the same time, obtain the monitoring data of multiple preset points in the foundation pit in real time through the preset monitoring equipment, and summarize and output the initial data;

[0056] Step 2: Select historical data that matches the initial data from the historical database and output the historical reference data;

[0057] Step 3: Combine the preset analysis model to conduct a comprehensive analysis of the initial data and historical reference data, and output real-time analysis results;

[0058] Step 4: Perform predictive analysis on the initial data and historical reference data in combination with the preset predictive model, output the predictive analysis results, and obtain the real-time status and predicted status of the foundation pit structure in the result-status comparison table in combination with the real-time analysis results, output the status analysis table and make early warning judgments based on the preset threshold conditions.

[0059] In this embodiment, the preset range is an area for data collection and monitoring within a preset range around the foundation pit. For example, a range with a radius of 50 meters can be set to monitor changes in the soil around the foundation pit;

[0060] In this embodiment, structural data: data describing the structural characteristics of the foundation pit and surrounding soil, such as soil layer type, density, water content, etc.;

[0061] In this embodiment, the preset monitoring equipment: equipment pre-installed around the foundation pit for real-time monitoring of data, such as displacement sensors, strain gauges, etc.;

[0062] In this embodiment, preset points: monitoring point locations pre-determined around the foundation pit, used for real-time monitoring data, for example, multiple monitoring points can be set around the foundation pit;

[0063] In this embodiment, monitoring data: data collected in real time from a preset monitoring device, including displacement, pressure, deformation and other information;

[0064] In this embodiment, initial data: i.e., the summary output of structure data and real-time monitoring data;

[0065] In this embodiment, historical database: a database storing past monitoring data for comparison and analysis with initial data;

[0066] In this embodiment, historical reference data: data selected from a historical database that matches the initial data and is used as a reference for analysis;

[0067] In this embodiment, the preset analysis model: a model used to analyze the initial data and historical reference data, which may be a neural network model or other data analysis model;

[0068] In this embodiment, real-time analysis results: real-time results obtained by combining initial data and historical reference data for analysis are used to understand the current state of the foundation pit structure;

[0069] In this embodiment, the preset prediction model is a model used to perform predictive analysis on the initial data and historical reference data, which is usually a model built based on a neural network or other machine learning methods, and can predict future surface subsidence based on historical data and real-time data.

[0070] In this embodiment, the prediction analysis results are: the surface settlement prediction results obtained by analyzing the initial data and historical reference data through the preset prediction model. These results can help predict the settlement of the soil around the foundation pit.

[0071] In this embodiment, the result-state comparison table: a table containing the mapping relationship between the real-time analysis results, the predictive analysis results and the states, used to determine the corresponding state according to the input real-time analysis results or predictive analysis results;

[0072] In this embodiment, the real-time state and the predicted state are: the current state of the foundation pit structure obtained based on the real-time monitoring data, and the predicted state is the future state obtained based on the prediction model. The comparison of the two states can help engineering personnel understand the development trend of the foundation pit project;

[0073] In this embodiment, the state analysis table: a table for summarizing the state information of the foundation pit structure;

[0074] In this embodiment, the preset threshold conditions are: a set of standards or conditions used to determine whether the state of the foundation pit structure is within a safe range. Once the monitoring data exceeds these thresholds, early warning processing is required;

[0075] In this embodiment, early warning judgment: the real-time analysis results and the predictive analysis results are judged according to preset threshold conditions. If the set threshold is exceeded, an early warning is triggered to notify relevant personnel to take corresponding measures.

[0076] Implementation principle and beneficial effects of this embodiment: The present invention combines real-time monitoring data, historical data and preset models for comprehensive analysis and prediction, thereby achieving real-time monitoring and early warning of the state of the foundation pit structure. The present invention improves the prediction accuracy of surface settlement during the construction of foundation pit engineering, effectively reduces the risks that may occur in foundation pit engineering, and helps engineering personnel to discover problems in a timely manner and take measures, thereby ensuring the safety and stability of the project.

[0077] The neural network-based surface settlement prediction and early warning method provided in the embodiment of the present invention obtains the structural data of the foundation pit and the soil within a preset range around the foundation pit, including:

[0078] Obtain the construction duration data of the foundation pit and divide it into stages to obtain multiple construction sub-stages;

[0079] Obtain the surface settlement data of the foundation pit in each construction sub-stage and within the preset distance range around the foundation pit, and summarize and output the surface settlement data of the construction stage.

[0080] In this embodiment, the construction time data: the time data spent on the construction of the foundation pit project at different stages, which can reflect the progress of the foundation pit project;

[0081] In this embodiment, stage division: the entire foundation pit construction process is divided into multiple stages, each stage having specific construction content and time range;

[0082] In this embodiment, the construction sub-stage is a sub-stage further divided into each stage, which is used to monitor and analyze the surface settlement of the foundation pit project in more detail;

[0083] In this embodiment, the preset distance range: a distance range preset in each construction sub-stage, used to monitor the surface settlement of the foundation pit and the soil around the foundation pit;

[0084] In this embodiment, the surface settlement data includes monitoring data of the foundation pit and the surface settlement within a preset distance range acquired in each construction sub-stage;

[0085] In this embodiment, the surface settlement data during the construction phase refers to a data set that summarizes and outputs the surface settlement data obtained in each construction sub-phase.

[0086] Implementation principle and beneficial effects of this embodiment: The present invention obtains surface settlement data of different construction sub-stages, combines construction duration data and stage division, and uses neural network and other technologies for comprehensive analysis and prediction to achieve monitoring and early warning of surface settlement of foundation pit engineering. By analyzing surface settlement data of different construction stages, the present invention can better understand the deformation of soil during the construction of foundation pit engineering, and provide a scientific basis for engineering management and decision-making.

[0087] The neural network-based surface settlement prediction and early warning method provided in the embodiment of the present invention acquires monitoring data at multiple preset points in a foundation pit in real time through preset monitoring equipment, including:

[0088] Acquire first position data corresponding to the first support and the second support of the foundation pit during the construction process;

[0089] Based on the first position data, multiple first preset points are respectively set at the first support and the second support, and a monitoring device array is constructed by installing preset monitoring devices at each first preset point. The support axial force data of the foundation pit at the first support and the second support during the construction process are obtained in real time based on the monitoring device array, and the support axial force change data is output.

[0090] In this embodiment, the first support is usually located at the edge of the foundation pit, relatively shallow from the ground surface, and is used to prevent soil collapse and protect the stability of the foundation pit slope in the early stage of foundation pit excavation;

[0091] In this embodiment, the second support is usually located in the deep part of the foundation pit, and is used to continue to support the soil and protect the stability of the foundation pit slope after the depth of the foundation pit increases;

[0092] In this embodiment, the first position data refers to the specific position data corresponding to the first support and the second support during the construction of the foundation pit, for example, specific setting position information;

[0093] In this embodiment, the first preset point: a plurality of monitoring points set at the first support and the second support according to the first position data, for monitoring the deformation and axial force of the support structure;

[0094] In this embodiment, the monitoring device array is a group of monitoring device systems constructed by installing monitoring devices at each first preset point, which is used to monitor the support axial force data of the foundation pit at the first support and the second support in real time;

[0095] In this embodiment, support axial force data: the axial force data of the support structure at the first support and the second support during the construction of the foundation pit acquired in real time by the monitoring device array, reflecting the stress condition of the support structure;

[0096] In this embodiment, the support axial force change data: data obtained based on the support axial force data obtained by the monitoring device array, which is the data analyzed and outputted on the change of the axial force of the support structure.

[0097] The implementation principle and beneficial effects of this embodiment: The present invention monitors the support axial force data at multiple preset points in the foundation pit in real time, and combines neural network and other technologies for analysis and prediction to achieve real-time monitoring and early warning of the status of the support structure of the foundation pit project. The present invention improves the monitoring accuracy of the deformation and axial force changes of the support structure of the foundation pit project, which helps to timely discover possible problems with the support structure and take corresponding measures to ensure the safe construction of the foundation pit project. At the same time, by monitoring the support axial force change data, the stability of the support structure can be effectively evaluated, providing an important reference basis for project management and decision-making.

[0098] The neural network-based surface settlement prediction and early warning method provided in the embodiment of the present invention acquires monitoring data at multiple preset points in the foundation pit in real time through preset monitoring equipment, and also includes:

[0099] Acquire the structural data and second position data of the ground-connected wall enclosure structure of the foundation pit during the construction phase, and set a plurality of second preset points;

[0100] The displacement data of the ground-anchored wall retaining structure at each second preset point during the construction process of the foundation pit is obtained through the preset monitoring equipment, and the structural displacement change data is output.

[0101] In this embodiment, the ground-connected wall enclosure structure is a structure used to support and protect the soil around the foundation pit, usually composed of steel sheet piles, concrete piles, etc.

[0102] In this embodiment, the construction data includes the specific construction information of the ground-connected wall enclosure structure, including the materials used, design parameters, construction methods, etc.;

[0103] In this embodiment, the second position data refers to the relative position data between the ground-connected wall enclosure structure and the foundation pit during the construction phase;

[0104] In this embodiment, the second preset points are: a plurality of monitoring points set at the ground-connected wall enclosure structure according to the second position data, and are used to monitor the deformation and displacement of the ground-connected wall enclosure structure;

[0105] In this embodiment, the displacement data of the ground-connected wall enclosure structure is obtained in real time by a preset monitoring device at each second preset point of the foundation pit during the construction process, reflecting the deformation of the ground-connected wall enclosure structure;

[0106] In this embodiment, the structural offset change data: data obtained by summarizing the offset change of the ground-connected wall enclosure structure can help to subsequently predict the deformation and offset data of the ground-connected wall enclosure structure.

[0107] Implementation principle and beneficial effects of this embodiment: The present invention realizes real-time monitoring and early warning of the condition of the ground-connected wall retaining structure by real-time monitoring of the displacement data of the ground-connected wall retaining structure at multiple preset points in the foundation pit, and combines neural network and other technologies for analysis and prediction. The present invention improves the monitoring accuracy of the deformation and displacement of the ground-connected wall retaining structure of the foundation pit project, helps to timely discover possible problems and take corresponding measures to ensure the safe construction of the foundation pit project.

[0108] The prediction and early warning method for surface subsidence based on a neural network provided in an embodiment of the present invention summarizes and outputs initial data, including:

[0109] Obtaining timestamps corresponding to the structure data and monitoring data, performing time alignment and information annotation, and summarizing and outputting the first data;

[0110] The first data is preprocessed using a preset preprocessing method to output initial data.

[0111] In this embodiment, timestamp refers to a time mark of a data record, which is usually a number or a string used to identify the time point of the data record;

[0112] In this embodiment, time alignment: aligning data from different data sources or different monitoring points in time order to ensure that they are compared and analyzed at the same time point;

[0113] In this embodiment, information annotation: adding additional labels or annotations to the data to help better understand and analyze the data, usually including information such as the source, type, and unit of the data;

[0114] In this embodiment, the first data: the summary data after time alignment and information annotation;

[0115] In this embodiment, the pre-processing method is a pre-set method for processing the first data, including steps such as data cleaning, feature extraction, and normalization, so as to convert the data into a form suitable for input to the neural network model;

[0116] In this embodiment, preprocessing is used to clean, convert and process the first data to improve data quality and model training effect.

[0117] The implementation principle and beneficial effects of this embodiment: The present invention aligns the structure data and the monitoring data by timestamp and annotates the information to form the first data, and then processes the first data by a preset preprocessing method to prepare the input data for the neural network model. The present invention can improve the quality and applicability of the data, and help enhance the accuracy and stability of the neural network model.

[0118] The neural network-based surface subsidence prediction and early warning method provided in the embodiment of the present invention comprises, in step 2:

[0119] Extract features from the initial data, perform feature matching on the extracted data features in combination with a preset settlement feature database, and output the settlement feature matching results;

[0120] Combined with the settlement feature matching results, historical foundation pit settlement data that meets the preset screening conditions are screened in the historical database, and historical reference data are output.

[0121] In this embodiment, a preset settlement feature database is a pre-established database containing various possible settlement feature information, which is used to match the extracted data features to identify and describe the characteristics of surface settlement;

[0122] In this embodiment, the settlement feature matching result: a matching result describing the surface settlement feature obtained by matching the extracted data feature with a preset settlement feature database;

[0123] In this embodiment, the preset screening conditions are conditions set according to specific standards or requirements, and are used to screen out data that meets the conditions in the historical database for use as historical reference data;

[0124] In this embodiment, historical foundation pit settlement data: surface settlement data recorded in past foundation pit projects, is used as a reference and comparison to help predict possible settlement conditions in current foundation pit projects.

[0125] Implementation principle and beneficial effects of this embodiment: The present invention extracts features from the initial data and matches them with a preset settlement feature database to obtain settlement feature matching results. Then, according to the matching results and preset screening conditions, the historical foundation pit settlement data that meets the conditions is screened out from the historical database as historical reference data for further analysis and prediction of the surface settlement of the current foundation pit project. The present invention can use historical data as a reference to help predict the possible surface settlement of the current foundation pit project, improve the accuracy and reliability of the prediction, and thus effectively warn of potential safety risks.

[0126] The method for predicting and warning ground subsidence based on a neural network provided in an embodiment of the present invention comprises, in step 3:

[0127] Extract features from the initial data, and construct the current foundation pit feature set based on the extracted features;

[0128] Based on the current foundation pit feature set, data satisfying the first screening condition is selected from the historical reference data, and output as comparison data;

[0129] Obtaining the contrast between each comparison data and the initial data, and sorting and dividing the comparison data according to a preset sorting rule to obtain multiple groups of comparison sub-data with different comparison ranges;

[0130] At the same time, feature extraction is performed on each comparison data, and a comparison feature set is constructed based on the extracted features. In addition, a variety of data analysis models that meet the user's data analysis needs are selected from the model database in combination with the current foundation pit feature set.

[0131] The initial data is combined with each comparison sub-data according to the preset combination rules to obtain the combined data under multiple combinations, and the combined data are respectively input into each data analysis model for real-time analysis, and the analysis results of multiple data analysis models under each combined data are summarized to obtain the multi-model analysis results under each combined data;

[0132] Perform content analysis on the multi-model analysis results under each combination of data to obtain analysis data, and use a preset curve construction algorithm to perform curve fitting and data annotation on the analysis data to obtain a model curve-data graph corresponding to each data analysis model under each combination of data;

[0133] The curves and data graphs under each combination of data are time-aligned, and the curves and data graphs that meet the preset data comparison conditions are compared and analyzed and data consensus is achieved to obtain the multi-model consensus results under each combination of data, and the real-time analysis results are summarized.

[0134] In this embodiment, the current foundation pit feature set is a feature set obtained by extracting features from the initial data, and is used to describe the key features of the current foundation pit project;

[0135] In this embodiment, the first screening condition: the screening condition set when selecting data from the historical reference data, used to screen out data that meets the condition as comparison data;

[0136] In this embodiment, the comparison data: data that meets the first screening condition selected from the historical reference data, used for comparison and analysis with the initial data;

[0137] In this embodiment, contrast refers to the similarity or difference between the comparison data and the initial data, which is used to evaluate the relationship between the two sets of data;

[0138] In this embodiment, the preset sorting rule: a sorting method set according to certain standards or rules, used to sort and divide the comparison data;

[0139] In this embodiment, the contrast range refers to the range of contrast corresponding to each group of contrast sub-data, which is used to distinguish different groups of contrast sub-data;

[0140] In this embodiment, the comparison sub-data: a sub-data set obtained by dividing the comparison data according to a preset rule;

[0141] In this embodiment, the comparison feature set is a feature set obtained by extracting features from the comparison sub-data, and is used to describe the feature conditions of the comparison data;

[0142] In this embodiment, the model database: a database storing various data analysis models, used to select a variety of data analysis models that meet the user's data analysis needs for real-time analysis;

[0143] In this embodiment, user data analysis requirements: user data analysis requirements refer to the specific needs and requirements of users for data analysis, including the type of data to be analyzed, the purpose of analysis, etc.;

[0144] In this embodiment, data analysis model: a mathematical model used to analyze and predict data, which may be a neural network model, a regression model, a decision tree model, etc.;

[0145] In this embodiment, the preset combination rule: a combination method set according to a certain rule is used to combine the initial data with the comparison sub-data;

[0146] In this embodiment, combined data: a data set obtained by combining initial data with comparison sub-data according to a preset combination rule;

[0147] In this embodiment, the multi-model analysis result refers to the result obtained after each data analysis model analyzes the same combination of data. By summarizing these results, a more comprehensive analysis conclusion can be obtained;

[0148] In this embodiment, parsed data: parsed data is data obtained after content parsing of the multi-model analysis results;

[0149] In this embodiment, the preset curve construction algorithm is used to construct a model curve according to the features in the parsed data, so as to better understand the changing trend of the data;

[0150] In this embodiment, curve fitting: according to the distribution of data points, a curve that best fits the data trend is found through a mathematical model to better describe the variation pattern of the data;

[0151] In this embodiment, data annotation refers to marking key data points or information on the curve graph so that users can understand the meaning of the data more intuitively;

[0152] In this embodiment, the model curve-data graph is a graph that combines data points with corresponding model curves to display the changing trend of data and the fitting of the model;

[0153] In this embodiment, the preset data comparison conditions refer to the conditions set for comparing and analyzing data, such as the relative difference between data, trend consistency, etc.;

[0154] In this embodiment, data consensus refers to the process of consensus analysis of the analysis data of each model in the model curve-data graph to reach a consistent conclusion;

[0155] In this embodiment, the multi-model consensus result: the final conclusion obtained through comparative analysis and data consensus based on the multi-model analysis results under each combination of data.

[0156] The implementation principle and beneficial effects of this embodiment: Based on the current foundation pit feature set, the present invention selects comparative data that meets the first screening condition from the historical reference data, extracts features from the comparative data, and constructs a comparative feature set. Then, according to the preset sorting rules and combination rules, the initial data is combined with the comparative sub-data, and multiple data analysis models are input for real-time analysis, and finally a multi-model consensus result is obtained. The present invention can improve the accuracy and reliability of predictions through the aggregation of multi-model analysis and consensus results, help engineering personnel to promptly discover potential problems and take corresponding measures, thereby improving the safety and stability of foundation pit projects.

[0157] The neural network-based surface subsidence prediction and early warning method provided by the embodiment of the present invention, in step 4, includes:

[0158] Combine the data types corresponding to each data in the historical reference data, classify the historical reference data, obtain reference sub-data under each type, and summarize to obtain the data to be classified;

[0159] The data to be divided is divided by using a preset data division method to obtain a training data set, a verification data set and a test data set;

[0160] Based on the training data set, the validation data set and the test data set, multiple first models selected from the model database are trained and optimized, and the first models that meet the preset performance conditions are output as preset prediction models;

[0161] The initial data is predicted and analyzed synchronously by at least two preset prediction models to obtain the surface settlement data, support axial force data and support structure displacement data of the foundation pit in the future preset period, and output the model-prediction data comparison table;

[0162] Reach consensus on the prediction results in the model-prediction data comparison table and output the prediction consensus results;

[0163] Combine the preset evaluation indicators and use the preset grade classification rules to establish a risk grade comparison table, and determine the corresponding fall-in judgment conditions for each risk grade;

[0164] Among them, the corresponding fall-in judgment conditions for each risk level are determined as follows:

[0165] ;

[0166] in, represents the critical Euclidean distance corresponding to the k-th risk level under the i-th preset evaluation indicator; Indicates the total number of preset evaluation indicators; represents the quantitative value corresponding to the j-th risk level under the i-th preset evaluation indicator, where i=1,2,…m; j=1,2,…n; represents the calculation weight corresponding to the i-th preset evaluation indicator;

[0167] Combined with the preset evaluation indicators, the real-time analysis results and the predicted consensus results are analyzed through the preset risk level evaluation algorithm to obtain the risk assessment data of the current foundation pit structure, and compared with the corresponding fall-in judgment conditions of each level in the risk level comparison table to determine the real-time risk level of the current foundation pit structure and the predicted risk level in the future preset period, and output the risk status analysis results;

[0168] Among them, the preset risk level assessment algorithm is as follows:

[0169] ;

[0170] in, represents the risk assessment value corresponding to the i-th assessment indicator, which is used for comparative analysis with the critical Euclidean distance; represents the mean value corresponding to all statistical data under the i-th evaluation indicator; Represents the standard deviation corresponding to all statistical data under the i-th preset evaluation indicator;

[0171] Compare and analyze the risk status analysis results with the preset threshold conditions to determine the type of warning data, and select the prevention strategy and prevention measures corresponding to the warning data type in the strategy database;

[0172] Based on various prevention strategies and measures, the corresponding structures in the current foundation pit structure are transformed.

[0173] In this embodiment, data type: types of different data in the historical reference data, for example, surface settlement data, support axial force data, support structure displacement data, etc.;

[0174] In this embodiment, the reference sub-data: a sub-data set corresponding to each data type is obtained after classification according to the historical reference data;

[0175] In this embodiment, the data to be divided: namely, the summary data of each reference sub-data;

[0176] In this embodiment, a data partitioning method is preset: the data to be partitioned is partitioned according to certain rules and methods, and the data is usually partitioned into a training data set, a validation data set, and a test data set;

[0177] In this embodiment, the training data set, the validation data set and the test data set: the training data set is used to train the model, the validation data set is used to adjust the model parameters and prevent overfitting, and the test data set is used to evaluate the model performance;

[0178] In this embodiment, the first model: an untrained and unoptimized prediction model selected from a model database, used to predict surface settlement data, etc.;

[0179] In this embodiment, the preset performance conditions are: performance requirements that the model needs to meet during training and optimization, such as prediction accuracy, generalization ability, etc.;

[0180] In this embodiment, the future preset time period: a future time period set in the prediction analysis, used to predict the surface settlement data, support axial force data and support structure displacement data of the foundation pit during this period;

[0181] In this embodiment, the model-prediction data comparison table: includes a comparison table of the corresponding relationship between the preset prediction model and the prediction data obtained by the preset prediction model;

[0182] In this embodiment, the prediction consensus result: a consensus result obtained by consensus of the prediction analysis results of multiple prediction models;

[0183] In this embodiment, the preset evaluation index is a set of indicators for evaluating the risk of foundation pit structure, including surface settlement rate, support axial force change rate, etc.;

[0184] In this embodiment, the preset level classification rules are rules formulated according to preset evaluation indicators, which are used to classify risk levels into different levels, such as no risk, low risk, medium risk, high risk, etc.;

[0185] In this embodiment, the risk level comparison table includes the specific definitions and descriptions corresponding to each risk level, and a table of related fall-in determination conditions;

[0186] In this embodiment, the judgment condition is fallen into: the critical Euclidean distance calculated according to the preset evaluation index and weight is used to subsequently determine the real-time risk level of the current foundation pit structure and the predicted risk level within the future preset period;

[0187] In this embodiment, a preset risk level assessment algorithm is an algorithm for calculating the risk assessment value of the foundation pit structure according to preset assessment indicators and statistical data. The risk level of the foundation pit structure can be determined by the algorithm;

[0188] In this embodiment, risk assessment data: risk assessment value of the foundation pit structure calculated by a preset risk level assessment algorithm, used to determine the real-time risk level and the predicted risk level;

[0189] In this embodiment, the real-time risk level is: the risk level of the current foundation pit structure calculated based on the risk assessment data;

[0190] In this embodiment, the predicted risk level is the predicted risk level of the foundation pit structure in a future preset period of time calculated based on the predicted consensus result and the preset risk level assessment algorithm;

[0191] In this embodiment, the risk status analysis result: the risk status analysis result of the foundation pit structure obtained by comparing the real-time risk level and the predicted risk level;

[0192] In this embodiment, the warning data type: the risk type that needs warning determined according to the risk status analysis result, for example, abnormal surface settlement, excessive support axial force, etc.;

[0193] In this embodiment, the strategy database: contains prevention strategies and prevention measures for different warning data types, which are used to take corresponding measures to deal with potential risks;

[0194] In this embodiment, prevention strategies and prevention measures are early warning measures for specific risks selected based on information in the strategy database, such as strengthening the supporting structure, reducing the speed of foundation pit excavation, and excavating in sections.

[0195] The implementation principle and beneficial effects of this embodiment: The present invention trains and optimizes a preset prediction model through a training data set, a verification data set, and a test data set, and then predicts and analyzes the initial data through multiple prediction models to obtain surface settlement data, support axial force data, and support structure displacement data, and finally outputs the prediction consensus results and risk status analysis results. Finally, risk assessment and early warning are carried out to ensure the safety and stability of the foundation pit structure. Through multi-model prediction and risk assessment, the present invention helps engineering personnel better understand the risk situation of the foundation pit structure, take preventive measures in time, and ensure the safety and stability of the foundation pit project.

[0196] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting and warning ground subsidence based on a neural network, characterized in that: include: Step 1: Obtain the structural data of the foundation pit and the soil within the preset range around the foundation pit. At the same time, obtain the monitoring data of multiple preset points in the foundation pit in real time through the preset monitoring equipment, and summarize and output the initial data; Step 2: Select historical data matching the initial data from the historical database and output the historical reference data; Step 3: Perform a comprehensive analysis on the initial data and historical reference data in combination with a preset analysis model, and output real-time analysis results; Step 4: Perform predictive analysis on the initial data and historical reference data in combination with a preset predictive model, output the predictive analysis results, and obtain the real-time status and predicted status of the foundation pit structure in the result-status comparison table in combination with the real-time analysis results, output the status analysis table and perform early warning judgment according to the preset threshold conditions; Wherein, step 3 includes: Extracting features from the initial data, and constructing a current foundation pit feature set based on the extracted features; Based on the current foundation pit feature set, data satisfying a first screening condition is selected from the historical reference data, and output as comparison data; Obtaining the contrast between each of the comparison data and the initial data, and sorting and dividing the comparison data according to a preset sorting rule to obtain a plurality of groups of comparison sub-data with different comparison ranges; At the same time, feature extraction is performed on each of the comparison data, and a comparison feature set is constructed based on the extracted features, and multiple data analysis models that meet the user's data analysis needs are selected from the model database in combination with the current foundation pit feature set; The initial data and the comparison sub-data are combined according to a preset combination rule to obtain combined data under multiple combinations, and the combined data are respectively input into each of the data analysis models for real-time analysis, and the analysis results of the multiple data analysis models under each of the combined data are summarized to obtain a multi-model analysis result under each of the combined data; Performing content analysis on the multi-model analysis results under each of the combined data to obtain analyzed data, and performing curve fitting and data annotation on the analyzed data using a preset curve construction algorithm to obtain a model curve-data graph corresponding to each data analysis model under each of the combined data; The curve-data graphs under each combination of data are time-aligned, and the curve-data graphs that meet the preset data comparison conditions are compared and analyzed and data consensus is achieved to obtain a multi-model consensus result under each combination of data, and the real-time analysis results are summarized.

2. The method for predicting and warning ground subsidence based on a neural network according to claim 1, characterized in that: The obtaining of structural data of the foundation pit and the soil within a preset range around the foundation pit includes: Obtain the construction duration data of the foundation pit and divide it into stages to obtain multiple construction sub-stages; The surface settlement data of the foundation pit in each construction sub-stage and within each preset distance range around the foundation pit are obtained, and the surface settlement data of the construction stage are summarized and output.

3. The method for predicting and warning ground subsidence based on neural network according to claim 1, characterized in that: The real-time acquisition of monitoring data at a plurality of preset points in the foundation pit by using preset monitoring equipment includes: Acquire first position data corresponding to the first support and the second support of the foundation pit during the construction process; In combination with the first position data, multiple first preset points are respectively set at the first support and the second support, and a monitoring device array is constructed by installing preset monitoring devices at each of the first preset points. Based on the monitoring device array, the support axial force data of the first support and the second support during the construction of the foundation pit are obtained in real time, and the support axial force change data is output.

4. The method for predicting and warning ground subsidence based on neural network according to claim 3 is characterized in that: The real-time acquisition of monitoring data at a plurality of preset points in the foundation pit by using a preset monitoring device also includes: Acquire the structural data and second position data of the ground-connected wall enclosure structure of the foundation pit during the construction phase, and set a plurality of second preset points; The displacement data of the ground-connected wall retaining structure at each of the second preset points during the construction of the foundation pit is obtained through the preset monitoring equipment, and the structure displacement change data is output.

5. The method for predicting and warning ground subsidence based on neural network according to claim 1, characterized in that: The summary output of initial data includes: Obtaining the timestamps corresponding to the structure data and the monitoring data, performing time alignment and information annotation, and summarizing and outputting the first data; The first data is preprocessed using a preset preprocessing method to output initial data.

6. The method for predicting and warning ground subsidence based on neural network according to claim 1, characterized in that: Step 2 includes: Extracting features from the initial data, performing feature matching on the extracted data features in combination with a preset sedimentation feature database, and outputting sedimentation feature matching results; Combined with the settlement feature matching result, historical foundation pit settlement data that meets preset screening conditions is screened in the historical database, and historical reference data is output.

7. The method for predicting and warning ground subsidence based on a neural network according to claim 1, characterized in that: Step 4 includes: In combination with the data types corresponding to each data in the historical reference data, the historical reference data are classified to obtain reference sub-data under each type, and the data to be classified are summarized; Using a preset data partitioning method to partition the data to be partitioned, to obtain a training data set, a verification data set, and a test data set; Based on the training data set, the validation data set and the test data set, multiple first models selected from the model database are trained and optimized, and the first models that meet the preset performance conditions are output as preset prediction models; By using at least two of the preset prediction models to simultaneously perform prediction analysis on the initial data, surface settlement data, support axial force data and support structure displacement data of the foundation pit in a future preset period of time, and output a model-prediction data comparison table; A consensus is reached on the prediction results in the model-prediction data comparison table, and a prediction consensus result is output; Combine the preset evaluation indicators and use the preset classification rules to establish a risk level comparison table, and determine the corresponding fall-in judgment conditions for each risk level; Among them, the corresponding fall-in judgment conditions for each risk level are determined as follows: ； in, represents the critical Euclidean distance corresponding to the k-th risk level under the i-th preset evaluation indicator; Indicates the total number of preset evaluation indicators; represents the quantitative value corresponding to the j-th risk level under the i-th preset evaluation indicator, where i=1,2,…m; j=1,2,…n; represents the calculation weight corresponding to the i-th preset evaluation indicator; Combined with the preset evaluation indicators, the real-time analysis results and the predicted consensus results are analyzed by a preset risk level evaluation algorithm to obtain the risk evaluation data of the current foundation pit structure, and compared with the fall-in judgment conditions corresponding to each level in the risk level comparison table to determine the real-time risk level of the current foundation pit structure and the predicted risk level within a preset period of time in the future, and output the risk status analysis results; The preset risk level assessment algorithm is as follows: ； in, represents the risk assessment value corresponding to the i-th assessment indicator, which is used for comparative analysis with the critical Euclidean distance; represents the mean value corresponding to all statistical data under the i-th evaluation indicator; Represents the standard deviation corresponding to all statistical data under the i-th preset evaluation indicator; Compare and analyze the risk status analysis results with the preset threshold conditions to determine the type of warning data, and select the prevention strategy and prevention measures corresponding to the warning data type in the strategy database; Based on the above-mentioned prevention strategies and measures, the corresponding structures in the current foundation pit structure are transformed.

Images