Underground engineering disaster monitoring and early warning method, system, medium and program

Abstract

The invention relates to the technical field of underground engineering, in particular to an underground engineering disaster monitoring and early warning method, system, medium and program, and the method comprises the steps: obtaining the multisource heterogeneous data of the geological structure and structural stress distribution in the underground engineering; performing data preprocessing on the multi-source heterogeneous data, and fusing the preprocessed multi-source heterogeneous data; the fused data are input into an intelligent monitoring and early warning model, the intelligent monitoring and early warning model outputs a disaster monitoring result of the underground engineering and generates a disaster early warning report according to the disaster monitoring result, and the intelligent monitoring and early warning model analyzes the fused data based on a machine learning algorithm to obtain the disaster monitoring result. Therefore, the problems that manual observation has great limitation, the efficiency is low, key data in the construction process is difficult to accurately and comprehensively obtain due to subjective factors and environment influences, and the underground engineering disaster monitoring result stability is poor in the related technology are solved.

Description

Technical Field

[0001] The present application relates to the field of underground engineering technology, and in particular to an underground engineering disaster monitoring and early warning method, system, medium and program. Background Art

[0002] With the acceleration of urbanization and the continuous gathering of population, the demand for the development and utilization of underground space in cities is growing. As an important part of urban underground projects, subways and underground comprehensive pipe corridors play a vital role in alleviating traffic pressure and ensuring the operation of urban infrastructure, but they also face many technical difficulties during construction and use. During the construction process, the complexity of geological conditions is the primary challenge. The stratigraphic structure in different regions may be uneven, and these geological differences will have different degrees of impact on the shield construction of subway tunnels and the excavation of pipe corridors. In addition, the groundwater conditions also have a significant impact on subway and pipe corridor projects. Factors such as the height of the groundwater level, water flow velocity, and water pressure may cause a series of disasters during the construction process. Traditional underground engineering construction monitoring methods have obvious shortcomings when facing these problems of subway and pipe corridor projects.

[0003] In related technologies, monitoring methods mainly rely on manual observation and some simple instrument measurements, and the instrument functions are relatively simple. Manual observation has great limitations. It is not only inefficient, but also greatly affected by subjective factors, making it difficult to accurately and comprehensively obtain key data in the construction process. The data obtained by some measurement methods are scattered and lack effective integration. And with the continuous expansion of the scale of subway and pipeline corridor projects, the amount of data generated during the construction process has increased massively. However, due to the lack of advanced data processing technology, most of these data are simply stored and cannot be effectively mined and utilized, resulting in the inability to take timely and effective countermeasures when facing sudden problems in the construction process. Summary of the invention

[0004] The present application provides an underground engineering disaster monitoring and early warning method, system, medium and program to solve the great limitations of manual observation in related technologies. Not only is the efficiency low, but also it is difficult to accurately and comprehensively obtain key data in the construction process due to subjective factors and environmental influences, resulting in poor stability of underground engineering disaster monitoring results.

[0005] The first aspect of the embodiments of the present application provides a method for monitoring and early warning of underground engineering disasters, including the following steps: obtaining multi-source heterogeneous data of the geological structure and structural stress distribution inside the underground engineering; performing data preprocessing on the multi-source heterogeneous data and fusing the preprocessed multi-source heterogeneous data; inputting the fused data into an intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground engineering and generates a disaster early warning report according to the disaster monitoring results, where the intelligent monitoring and early warning model analyzes the fused data based on a machine learning algorithm to obtain the disaster monitoring results.

[0006] Optionally, the multi-source heterogeneous data includes image data, real-time detection data, and abnormal data. Among them, the image data includes: the topographic and geomorphic features, geological structure, and engineering structure of the underground engineering; the real-time detection data includes: surrounding rock data and surface deformation data; the abnormal data includes abnormal groundwater data in the underground engineering pipelines, abnormal rock and soil body data in the underground engineering, and abnormal engineering structure data.

[0007] Optionally, the data preprocessing includes one or more of data cleaning, data feature extraction, and data standardization.

[0008] Optionally, fusing the preprocessed multi-source heterogeneous data includes: constructing a judgment matrix according to the preprocessed multi-source heterogeneous data and calculating the corresponding weights of each data source according to the judgment matrix; inputting the multi-source heterogeneous data and the corresponding weights of each data source into an expert model, and the expert model generates the adjusted corresponding weights of each data source; weighted-fusing the preprocessed multi-source heterogeneous data according to the adjusted corresponding weights of each data source.

[0009] Optionally, the machine learning algorithm includes the Transformer knowledge graph extraction algorithm and the ILocal-BART (Incremental Local Bayesian Additive Regression Trees) algorithm. The intelligent monitoring and early warning model includes an ensemble learning system, a real-time stream data analysis system, and a feedback early warning system. Among them, the ensemble learning system uses the Transformer knowledge graph extraction algorithm to extract knowledge features from the preprocessed multi-source heterogeneous data to construct a knowledge graph; the real-time stream data analysis system uses the ILocal-BART algorithm to perform incremental clustering analysis on real-time multi-source heterogeneous data to identify abnormal data and generate early warning information; the feedback early warning system is used to accurately push the early warning information to the target terminal.

[0010] Optionally, the processing process of the Transformer knowledge graph extraction algorithm includes: using the fastText (fast text model) embedding model and the BERT (Bidirectional Encoder Representations from Transformers) embedding model to convert multi-source heterogeneous data into corresponding word vector sequences, and classifying the word vector sequences into categories to generate category labels; inputting the word vector sequence into the Transformer model, and generating a context-associated word vector sequence after feature extraction of the Transformer model output; using the feature function of the CRF (Conditional Random Field) layer to capture the relationship between the word vector sequence and the category label, filter out the optimal label sequence for each word vector and generate the corresponding entity naming; based on the entity naming, the fastText embedding model, the BERT embedding model and the Transformer model are used again to extract knowledge and construct a knowledge graph.

[0011] Optionally, the processing process of the ILocal-BART algorithm includes: acquiring multi-source heterogeneous data received in real time; performing incremental clustering calculations on the multi-source heterogeneous data in each sliding window to generate cluster results, and screening out the winning cluster with the largest selection function value to match and confirm abnormal data; generating early warning information based on the abnormal data.

[0012] The second aspect of the present application provides an underground engineering disaster monitoring and early warning system, including: a data acquisition module, used to obtain multi-source heterogeneous data on the geological structure and structural stress distribution inside the underground engineering collected by various types of sensors; a multi-source data fusion processing module, used to perform data preprocessing on the multi-source heterogeneous data, and fuse the preprocessed multi-source heterogeneous data; an intelligent monitoring and early warning module, used to input the fused data into an intelligent monitoring and early warning model, the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground engineering, and generates a disaster early warning report based on the disaster monitoring results, wherein the intelligent monitoring and early warning model analyzes the fused data based on a machine learning algorithm to obtain the disaster monitoring results.

[0013] The third aspect of the present application provides a vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform the underground engineering disaster monitoring and early warning method as described in the above embodiment.

[0014] The fourth aspect of the present application provides a computer-readable storage medium on which a computer program is stored. The program is executed by a processor to perform the underground engineering disaster monitoring and early warning method as described in the above embodiment.

[0015] The fifth aspect of the present application provides a computer program product, including a computer program or instructions, characterized in that when the computer program or instructions are executed, the underground engineering disaster monitoring and early warning method as described in the above embodiment is implemented.

[0016] Therefore, this application has at least the following beneficial effects:

[0017] (1) The embodiment of the present application will comprehensively use a variety of advanced sensors and data acquisition equipment to accurately collect various key data in the construction process of subway and pipeline corridor projects, and provide sufficient and reliable data support for subsequent data analysis and disaster prediction; secondly, use a set of efficient data fusion and processing algorithms. Through this set of algorithms, data from different data sources and different types are organically integrated. Then, advanced machine learning technology is used to build an intelligent disaster prediction model, which can deeply learn various laws and patterns in the construction process of subway and pipeline corridor projects, and accurately predict possible disasters based on a large amount of historical data and real-time collected data; and establish a complete feedback warning and emergency response mechanism. When the disaster prediction model sends out a warning signal, it can immediately transmit the warning information to construction personnel, management personnel and relevant emergency rescue departments to ensure that losses can be minimized before the disaster occurs, and to ensure the safety and smooth progress of subway and pipeline corridor construction.

[0018] (2) The embodiment of the present application obtains multi-source heterogeneous data on the geological structure and structural stress distribution inside the underground project, ensuring the comprehensiveness and diversity of the monitoring system; preprocesses the multi-source heterogeneous data, fuses the preprocessed multi-source heterogeneous data, and cleans, denoises and standardizes the multi-source heterogeneous data to ensure the quality and consistency of the input data and improve the reliability of subsequent analysis; inputs the fused data into the intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground project, and generates a disaster early warning report based on the disaster monitoring results, which not only improves the comprehensiveness and accuracy of monitoring and early warning, but also significantly enhances construction safety and emergency response speed.

[0019] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0021] Figure 1 A flowchart of an underground engineering disaster monitoring and early warning method provided according to an embodiment of the present application;

[0022] Figure 2 A block diagram of an underground engineering disaster monitoring and early warning system provided according to an embodiment of the present application;

[0023] Figure 3 It is an overall schematic diagram of an underground engineering disaster monitoring and early warning system provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0025] The following describes the underground engineering disaster monitoring and early warning method, system, storage medium and program of the embodiments of the present application with reference to the accompanying drawings.

[0026] Specifically, Figure 1 A flow chart of an underground engineering disaster monitoring and early warning method provided in an embodiment of the present application.

[0027] like Figure 1 As shown, the underground engineering disaster monitoring and early warning method includes the following steps:

[0028] In step S101, multi-source heterogeneous data of geological structure and structural stress distribution inside the underground project are obtained.

[0029] Among them, multi-source heterogeneous data include image data, real-time detection data and abnormal data. Among them, image data include: underground engineering topography, geological structure and engineering structure, real-time detection data include: surrounding rock data and surface deformation data, abnormal data include groundwater abnormal data in underground engineering pipelines, rock and soil abnormal data in underground engineering and engineering structure abnormal data.

[0030] It can be understood that the embodiments of the present application can use the image acquisition system, real-time detection system, and abnormal data monitoring system in the data acquisition module to obtain multi-source heterogeneous data on the geological structure and structural stress distribution inside the underground project to improve the accuracy of data acquisition.

[0031] It should be noted that the data acquisition module uses a multi-type sensor collaborative operation mode, integrating geological detection, structural deformation monitoring and other sensors, and optimizing the layout according to the layout characteristics of underground projects. Wireless transmission technology is used to build a stable transmission link, realize real-time data upload, ensure data diversity and timeliness, and provide a basic data source for a comprehensive understanding of the project status.

[0032] Specifically, if Figure 3As shown, the image acquisition system is equipped with a 3D laser scanner, which uses the 3D laser scanner to photograph the underground topography, geological structure, etc. of the project, and comprehensively derives and integrates it into the 3D spatial information of the underground project for modeling, quality monitoring and deformation monitoring.

[0033] The real-time detection system includes a seismic wave detection unit, a SAR (Synthetic Aperture Radar) interferometer and an adaptive filter. The seismic wave detection unit conducts advance detection of the surrounding rock of the engineering body to obtain preliminary data of the surrounding rock. The SAR interferometer conducts real-time detection of the deformation of the engineering body and the surrounding surface to obtain preliminary data of surface deformation. The adaptive filter denoises the obtained surrounding rock data and surface deformation data to obtain accurate surrounding rock information and surface deformation information. The data is denoised through the adaptive filter to establish an underground engineering working condition model.

[0034] The abnormal data monitoring system includes intelligent hydrophones and Brillouin optical time-domain reflectometers. During the construction and use of the project, the intelligent hydrophones are used to monitor the flow rate, water level changes and other related information of groundwater in the underground project pipelines, which serves as the main data source for judging whether there are hidden dangers of disasters such as seepage and pipe bursts in the underground project; the Brillouin optical time-domain reflectometer is used to sense the changes in physical quantities such as stress and strain of rock and soil bodies in underground projects, and the force and deformation of engineering structures, detect abnormal changes in physical quantities that may cause disasters, and comprehensively derive them into hydrological condition vectors and rock and soil physical state vectors to establish an engineering environment monitoring model.

[0035] In step S102, data preprocessing is performed on the multi-source heterogeneous data, and the preprocessed multi-source heterogeneous data is fused.

[0036] Among them, data preprocessing includes one or more of data cleaning, data feature extraction and data standardization.

[0037] It can be understood that the embodiments of the present application can perform data preprocessing on multi-source heterogeneous data and fuse the preprocessed multi-source heterogeneous data, which not only improves the quality and consistency of the data, but also provides a solid foundation for subsequent intelligent monitoring and early warning.

[0038] It should be noted that the collected data is transmitted to the data preprocessing system via wired or wireless transmission, which checks, identifies and fills the collected data, extracts feature vectors, intelligently analyzes the distribution characteristics of the data, and chooses to use the Z-score normalization method or the Min-Max normalization method to standardize the data. For abnormal data, the data preprocessing system will mark it; if it exceeds a given threshold, it will send a warning message to the feedback warning system for preliminary warning.

[0039] Specifically, the formulas for the Z-score standardization method and the Min-Max normalization method are as follows:

[0040]

[0041] Where X is the collected data, μ is the mean, σ is the standard deviation, X max With X min They are the maximum and minimum values ​​in the collected data set respectively; for abnormal data, the data preprocessing system will mark it; if it exceeds the given threshold, an early warning message will be sent to the feedback early warning system for preliminary warning.

[0042] In an embodiment of the present application, the preprocessed multi-source heterogeneous data is fused, including: constructing a judgment matrix based on the preprocessed multi-source heterogeneous data, and calculating the corresponding weight of each data source based on the judgment matrix; inputting the multi-source heterogeneous data and the corresponding weight of each data source into an expert model, and the expert model generates an adjusted corresponding weight of each data source; and weightedly fusing the preprocessed multi-source heterogeneous data according to the adjusted corresponding weight of each data source.

[0043] It can be understood that the embodiments of the present application can construct a judgment matrix based on the preprocessed multi-source heterogeneous data, and calculate the corresponding weight of each data source based on the judgment matrix; the multi-source heterogeneous data and the corresponding weight of each data source are input into the expert model, and the expert model generates the corresponding weight of each data source after adjustment; the preprocessed multi-source heterogeneous data is weighted and fused according to the corresponding weight of each data source after adjustment, thereby improving the accuracy of anomaly detection and prediction, so as to identify potential risks in advance.

[0044] It should be noted that this application is based on a fusion algorithm based on data feature association, which first extracts and classifies features of different sensor data. Through an intelligent matching algorithm, it finds potential associations between data and fuses related feature data, so that the fused data can comprehensively reflect the various aspects of underground engineering, enhance the integrity and effectiveness of the data, and facilitate accurate analysis.

[0045] Specifically, a fusion algorithm based on AHP (Analytic Hierarchy Process) and expert analysis is used to classify and store the original data and processed data, and then export them as fused feature vector groups for transmission to the intelligent monitoring and early warning module.

[0046] In step S103, the fused data is input into the intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground project and generates a disaster early warning report based on the disaster monitoring results. The intelligent monitoring and early warning model obtains the disaster monitoring results by analyzing the fused data based on a machine learning algorithm.

[0047] It can be understood that the embodiment of the present application inputs the fused data into the intelligent monitoring and early warning model, the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground project, and generates a disaster early warning report based on the disaster monitoring results, which not only improves the comprehensiveness and accuracy of monitoring and early warning, but also significantly enhances construction safety and emergency response speed.

[0048] Specifically, during the construction and commissioning of underground projects, if the monitoring data exceeds a given threshold and cannot match the knowledge graph and incremental clustering model, the feedback warning system receives warning signals from the abnormal monitoring system, evaluates and analyzes the abnormal data through the trained ILocal-BART model, and generates a comprehensive disaster warning report to explain the location, cause and potential impact of the disaster, and provide reference response suggestions.

[0049] It should be noted that this application obtains fused data and feature vectors, uses a fusion algorithm based on AHP and expert analysis to determine input features with high correlation, and uses a multi-head attention mechanism to extract knowledge in an integrated learning system and construct it into a Transformer knowledge graph as a feature input to a real-time streaming data analysis system; the real-time streaming data system receives various types of data and knowledge graphs, performs dimensionless processing on them, integrates incremental data, and constructs an ILocal-BART data incremental model;

[0050] An intelligent system combining neural network and fuzzy logic is used. The neural network learns complex data patterns, and the fuzzy logic processes data uncertainty. The two work together, and the output of the neural network is used as the input of the fuzzy logic. The warning results are obtained through fuzzy reasoning, which improves the adaptability and judgment accuracy of the warning system to complex situations and effectively prevents engineering disasters. The details are as follows:

[0051] In an embodiment of the present application, the machine learning algorithm includes a Transformer knowledge graph extraction algorithm and an ILocal-BART algorithm, and the intelligent monitoring and early warning model includes an integrated learning system, a real-time streaming data analysis system, and a feedback early warning system. Among them, the integrated learning system uses the Transformer knowledge graph extraction algorithm to extract knowledge features from multi-source heterogeneous data after fusion preprocessing to construct a knowledge graph, and the real-time streaming data analysis system uses the ILocal-BART algorithm to perform incremental clustering analysis on real-time multi-source heterogeneous data to identify abnormal data and generate early warning information. The feedback early warning system is used to accurately push the early warning information to the target terminal.

[0052] It can be understood that the embodiments of the present application can provide rich background information and association relationships by constructing a knowledge graph, thereby enhancing the ability to interpret complex problems; combining historical data and real-time data to provide more accurate predictions and risk assessments, and identify potential risks in advance; through efficient feature extraction and pattern recognition, the probability of false alarms and missed alarms is reduced, and the reliability of the system is improved; real-time processing and automated report generation are achieved to ensure timely warnings and quick decision-making.

[0053] In an embodiment of the present application, the processing process of the Transformer knowledge graph extraction algorithm includes: using the fastText embedding model and the BERT embedding model to convert multi-source heterogeneous data into corresponding word vector sequences, and classifying the word vector sequences into categories to generate category labels; inputting the word vector sequence into the Transformer model, and generating a context-associated word vector sequence after feature extraction from the Transformer model output; using the feature function of the CRF layer to capture the relationship between the word vector sequence and the category label, screening out the optimal label sequence for each word vector to generate the corresponding entity naming; based on the entity naming, the fastText embedding model, the BERT embedding model and the Transformer model are used again to extract knowledge and construct a knowledge graph.

[0054] It can be understood that the embodiments of the present application can combine the advantages of fastText and BERT to generate high-quality word vectors. The self-attention mechanism of the Transformer model enhances the understanding of context. The CRF layer improves the accuracy of entity naming through global optimization, constructs a comprehensive knowledge graph, provides rich background information and association relationships, and improves the reliability of monitoring.

[0055] Specifically, the process of building a Transformer knowledge graph includes:

[0056] (1) Entity naming. The information in the text is used as the input data of the entity naming process to name the entity. The fastText and Bert embedding technologies are used to give the word vector sequence indexed in the future generated knowledge graph based on the vector of the input text information word. The Transformer model with reference attention mechanism is used to extract features and export vectors for the data. Finally, the relationship between labels is constructed using the feature function F(y|x) through CRF (Conditional Random Field). The output layer formula is as follows:

[0057]

[0058] (2) The structure of knowledge extraction is basically the same as that of entity naming. The output of the entity naming process is used as input data, embedded using fastText and Bert embedding technologies, and the Transformer model is used to extract features from the data. Finally, self-attention is used for knowledge extraction.

[0059] When extracting knowledge, the activation function softmax of the fully connected layer obtains the relationship probability corresponding to the entity in the sentence. In establishing the index graph, in addition to the initialization setting type, the other type is added to indicate the relationship type that does not belong to the initially defined type;

[0060] (3) Train the Transformer model and build a knowledge graph. The encoder and decoder of the Transformer model have different tasks during training and testing. During training, the original entity is input into the encoder, and the correct entity is input into the decoder. During testing, the original entity is input into the encoder, and the decoder is responsible for proofreading the output information. In the Transformer, the encoder layer includes a self-attention mechanism layer and a fully connected layer. The ReLU activation function is used at the output of an encoder layer, and the formula is as follows:

[0061] FFN(x)=max(0,xW 1 +b 1 )W 2 +b 2 (4)

[0062] Among them, W 1 ,W 2 is the weight; b 1 ,b 2 is the bias coefficient; x is the calculated entity value;

[0063] Based on the translation assumption, we know that:

[0064] h+r=t (5)

[0065] r=th (6)

[0066] Among them, h, r, t are the triple sets in the knowledge graph The three represent entity, relationship and tail entity respectively;

[0067] From this, the original embedding R of the relationship can be calculated through entity embedding 0 :

[0068] R 0 =S t -S h (7)

[0069] Among them, S t ,Sh is the embedded entity of the head and tail;

[0070] The original embedding, the embedding position information, and the relational embedding of the previous module are combined into a ternary embedding to complete the combination of the reference set and the query set, which is then input into the L-layer Transformer:

[0071]

[0072] in, It is the entity or relationship feature after the l-layer encoding layer. This feature encodes the semantic role of the entity and can identify the fine-grained meaning of different entity relationships.

[0073] In an embodiment of the present application, the processing process of the ILocal-BART algorithm includes: obtaining multi-source heterogeneous data received in real time; performing incremental clustering calculations on the multi-source heterogeneous data in each sliding window to generate cluster results, and screening out the winning cluster with the largest selection function value to match and confirm abnormal data; generating early warning information based on the abnormal data.

[0074] It is understandable that the embodiments of the present application

[0075] Specifically, the process of training the ILocal-BART data incremental model includes:

[0076] (1) Obtaining collected data;

[0077] (2) Initialization. Let N = 1, q = 1, and take the first data x in the qth snapshot 1 [q] as the first cluster C in this snapshot 1 Initialization mean vector μ of [q] 1 [q], initialize the covariance matrix ∑ 1 [q] is a diagonal matrix with the same dimension as the input, and the initialization formula is

[0078] μ 1 [q] = x 1 [q] (10)

[0079]

[0080] The input data set is D = [x 1 ,x 2 ,...,x N ]

[0081] (3) Incremental clustering of the data in each snapshot.

[0082] Ⅰ. Calculation of the selection function. i[q] matches all existing clusters, and the kth cluster C in the qth snapshot k [q] When matching, the calculation formula of the selection function is as follows:

[0083]

[0084] Among them, |C k [q]| is the number of clusters in the qth snapshot; P(C k [q]) is the cluster C in the qth snapshot k The prior probability of k = 1, 2, ..., |C k [q]|;P(C k [q]x i [q]) is the i-th data x in the q-th snapshot i [q] Select cluster C in the qth snapshot k [q] is the probability of the most suitable cluster, and the calculation formula is:

[0085]

[0086] Among them, N k [q] is the cluster C in the qth snapshot k [q] is the number of data, l is the dimension of the input, μ k [q] is the cluster C in the qth snapshot k The mean vector of [q], ∑ k [q] is the cluster C in the qth snapshot k The covariance matrix of [q], N is the number of processed data;

[0087] Ⅱ. Select the largest selection function value. The winning cluster J has the largest selection function value:

[0088]

[0089] Where C[q] is the number of existing clusters in the qth snapshot;

[0090] III. Matching test. Calculate data x i [q] The change in the covariance determinant after adding the winning cluster J. The calculation formula is as follows:

[0091] d(i)=N k [q]×log(∑ J [q])+log(Σ(x i [q]))-(N k [q]+1)×log(Σ com (i)) (16)

[0092] in,

[0093]

[0094] Among them, Σ com (i) is the data x i [q] The combined covariance matrix after merging into the winning cluster J, Σ(x i [q]) is the assumed data x i [q] is the covariance matrix of a single cluster, μ J [q] is the mean vector of the winning cluster J in the qth snapshot, Σ J [q] is the covariance matrix of the winning cluster J in the qth snapshot, N J [q] is the number of data in the winning cluster J in the qth snapshot, diag() is the vector diagonalization function, mean() is the mean function, and I is the identity matrix.

[0095] According to the underground engineering disaster monitoring and early warning method proposed in the embodiment of the present application, multi-source heterogeneous data of the geological structure and structural stress distribution inside the underground engineering are obtained, thereby ensuring the comprehensiveness and diversity of the monitoring system; data preprocessing is performed on the multi-source heterogeneous data, and the preprocessed multi-source heterogeneous data is fused. By cleaning, denoising and standardizing the multi-source heterogeneous data, the quality and consistency of the input data are ensured, and the reliability of subsequent analysis is improved; the fused data is input into the intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground engineering, and generates a disaster early warning report based on the disaster monitoring results, which not only improves the comprehensiveness and accuracy of monitoring and early warning, but also significantly enhances construction safety and emergency response speed.

[0096] Next, the underground engineering disaster monitoring and early warning system proposed according to the embodiment of the present application is described with reference to the accompanying drawings.

[0097] Figure 2 It is a block diagram of an underground engineering disaster monitoring and early warning system according to an embodiment of the present application.

[0098] like Figure 2 As shown, the underground engineering disaster monitoring and early warning system 10 includes: a data acquisition module 100, a multi-source data fusion processing module 200 and an intelligent monitoring and early warning module 300.

[0099] Among them, the data acquisition module 100 is used to obtain multi-source heterogeneous data of geological structure and structural stress distribution inside the underground project collected by various types of sensors; the multi-source data fusion processing module 200 is used to perform data preprocessing on the multi-source heterogeneous data and fuse the preprocessed multi-source heterogeneous data; the intelligent monitoring and early warning module 300 is used to input the fused data into the intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground project and generates a disaster warning report based on the disaster monitoring results. Among them, the intelligent monitoring and early warning model analyzes the fused data based on the machine learning algorithm to obtain the disaster monitoring results.

[0100] Specifically, if Figure 3 As shown, the underground engineering disaster monitoring and early warning system of the present application includes a data acquisition module, a multi-source data fusion processing module and an intelligent monitoring and early warning module, wherein the data acquisition module is used to obtain image acquisition data, real-time detection data, and abnormal monitoring data in engineering construction; the multi-source data fusion processing module is configured to obtain engineering environment feature vectors and surface deformation monitoring vectors according to image acquisition data and real-time detection data, and obtain hydrological condition feature vectors and temperature strain comprehensive monitoring vectors according to abnormal monitoring data; the intelligent monitoring and early warning module receives the image acquisition data, real-time detection data, and abnormal monitoring data of the data acquisition module, and performs real-time identification, evaluation, early warning and feedback on engineering environment and geological conditions, and exports the analysis results to the user end in real time, specifically:

[0101] (1) Data acquisition module The data acquisition module includes image acquisition system, real-time detection system and abnormal data monitoring system.

[0102] Among them, the image acquisition system mainly includes a three-dimensional laser scanner; the three-dimensional laser scanner can obtain three-dimensional spatial information such as underground engineering topography, geological structure and engineering structure with high precision, which is used for modeling, quality monitoring and deformation monitoring; the real-time detection system includes a seismic wave detection unit, a SAR interferometer and an adaptive filter; the seismic wave detection unit is used to conduct advance detection of the surrounding rock of the engineering geological body to obtain preliminary data of the surrounding rock; the SAR interferometer is used to measure the surface deformation around the underground project and obtain preliminary data of the surface deformation around the engineering body; the adaptive filter performs denoising on the obtained surrounding rock data and surface deformation data to obtain accurate surrounding rock information and surface deformation information; the abnormal data monitoring system includes an intelligent hydrophone and a Brillouin optical time-domain reflectometer; the intelligent hydrophone is used to monitor the flow rate, water level changes and other related information of groundwater in the underground engineering pipeline, as the main data source for judging whether there are hidden dangers of disasters such as seepage and pipe bursts in the underground project; the Brillouin optical time-domain reflectometer is used to perceive the changes in physical quantities such as stress and strain of rock and soil bodies and force and deformation of engineering structures in underground projects, and detect abnormal physical quantity changes that may cause disasters

[0103] (2) The multivariate data fusion processing module includes data preprocessing system, data fusion system, data transmission system and data storage system.

[0104] Among them, the data preprocessing system includes a data cleaning unit, a feature extraction unit and a data standardization unit; a cleaning and intelligent feature extraction method based on data rules is adopted. The cleaning link identifies and corrects abnormal data based on the normal fluctuation range and logical relationship of the data. Feature extraction uses deep learning algorithms to mine key data features, remove redundant information, make the data more refined and representative, and create favorable conditions for subsequent data in-depth analysis. Among them, the data cleaning unit is used to check the collected data and identify and process abnormal values ​​based on a given threshold. For missing values, if the missing data ratio is small (<5%) and presents statistical rules, mean filling or linear interpolation filling is used; for data with a large missing ratio, a prediction model is constructed based on the relevance and importance of the data; for abnormal values, they are marked and retained for subsequent analysis; if the abnormal value exceeds the given threshold, an early warning signal is sent to the feedback early warning system for early warning. The feature extraction unit is used for feature data in the construction process of underground engineering; the data standardization unit analyzes the distribution characteristics of the data and determines the appropriate data standardization method; if the data approximately obeys the normal distribution, the Z-score standardization method is used; if the data has a clear range of values, the Min-Max normalization method is used.

[0105] The data fusion system adopts a fusion algorithm based on AHP and expert analysis to integrate, evaluate and assign weights to data from different data sources and different modes. Finally, a fusion algorithm based on weighted average is used for data fusion. For each data feature or data point, it is multiplied by the corresponding weight and then summed up to obtain the fused feature vector. Finally, a three-dimensional underground engineering comprehensive working condition model and a disaster monitoring and early warning model are constructed.

[0106] The data transmission system adopts a combination of wired transmission and wireless transmission, and performs real-time verification of the transmitted data through data verification mechanisms such as CRC. The collected multimodal data is transmitted in real time to the multivariate data fusion processing module and the remote monitoring room, and the three modules are connected to realize information sharing among the modules within the system and inside and outside the system. Security technology based on identity authentication and encrypted transmission. Identity authentication strictly identifies the devices and users within the system, and encrypted transmission uses advanced encryption algorithms to encrypt data before transmission. Prevent illegal access and data theft from the source, ensure the confidentiality and security of data during transmission and interaction, and maintain stable operation of the system.

[0107] The data storage system uses HDFS as the main storage medium, and classifies and stores the collected raw data, pre-processed data, and fused data, while establishing a time series index. When data needs to be read, the data storage system retrieves and extracts the corresponding data from HDFS according to the type and requirements of the data request. A distributed cloud storage architecture is used in combination with data index optimization. Distributed cloud storage uses multiple cloud nodes to store data to ensure reliability and scalability. Data index optimization establishes multi-layer indexes based on data attributes and application scenarios, speeds up data query and call speed, meets the needs of underground engineering big data storage and frequent retrieval, and improves data management efficiency.

[0108] (3) The intelligent monitoring and early warning module includes an integrated learning system, a real-time streaming data analysis system, and a feedback early warning system.

[0109] Among them, the integrated learning system adopts the trained integrated Transformer knowledge graph extraction algorithm and the multi-head attention mechanism to extract valuable knowledge from multimodal data, construct it into a knowledge graph, and enrich the content of the knowledge graph by mining the mutual influence relationship between different monitoring parameters. Finally, the constructed knowledge graph is input as a feature into the real-time streaming data analysis system; the real-time streaming data analysis system adopts the trained ILocal-BART algorithm to realize incremental clustering of monitoring data. When abnormal data exceeding the threshold is monitored, an early warning signal is issued and early warning information is generated based on multimodal data information, and transmitted to the feedback early warning system; the feedback early warning system receives multimodal data and model information, monitors incremental clustering key data in real time, and establishes a working condition disaster early warning model; when an abnormality occurs, it gives priority to receiving early warning information and generates a disaster early warning report in real time, explaining the location, cause and potential impact of the disaster, and giving comprehensive response suggestions.

[0110] In summary, (1) the data acquisition module of this application uses a variety of high-precision sensors, which can sensitively perceive various subtle changes in underground projects, and can accurately capture stratum displacement and structural deformation, providing accurate data basis for project status assessment. The construction party can use this to gain insight into potential risks in advance, effectively optimize the construction plan, and greatly reduce project delays and rework caused by geological disasters or structural hazards, thereby significantly improving the efficiency of underground project construction and shortening the construction period.

[0111] (2) The data fusion processing technology of this application cleverly integrates multi-source heterogeneous data, and uses advanced algorithms to deeply mine hidden connections between data, so that the complex geological structure and structural stress distribution inside the underground project can be clearly presented. This helps engineering and technical personnel to fully grasp the overall picture of the project, and then formulate a more scientific and reasonable construction plan, effectively avoid safety accidents caused by misjudgment of geological conditions, effectively protect the lives of construction personnel, reduce the risk of economic losses in the project, and ensure the smooth implementation of the project.

[0112] (3) The intelligent monitoring and early warning module of this application has excellent disaster identification capabilities by virtue of the organic integration of deep learning and expert experience. In the early stages of underground engineering disasters, it can accurately detect signs of disasters such as water gushing and collapse, and issue early warning signals in a timely manner. The construction team can thus obtain sufficient time to implement response measures, such as strengthening support structures and transferring important equipment, thereby effectively reducing the degree of harm caused by the disaster, ensuring that the project progress is steadily advanced according to the scheduled plan, reducing the additional costs caused by delays in the construction period, and improving the economic and social benefits of the project.

[0113] (4) The data processing unit of this application adopts an efficient data cleaning and feature extraction method, which can accurately remove impurities and redundant information in the data and extract valuable key features. Based on this, the subsequent data analysis and model building work can be carried out more accurately and efficiently, and the development trend of the project at different construction stages, such as changes in structural stability, can be more accurately predicted. In addition, the construction parameters can be reasonably adjusted in advance, the project quality can be improved, material waste and resource consumption can be reduced, the benefits of underground projects can be maximized, and the sustainable development of projects can be promoted.

[0114] According to the underground engineering disaster monitoring and early warning system proposed in the embodiment of the present application, multi-source heterogeneous data of the geological structure and structural stress distribution inside the underground engineering are obtained, thereby ensuring the comprehensiveness and diversity of the monitoring system; data preprocessing is performed on the multi-source heterogeneous data, and the preprocessed multi-source heterogeneous data is fused. By cleaning, denoising and standardizing the multi-source heterogeneous data, the quality and consistency of the input data are ensured, and the reliability of subsequent analysis is improved; the fused data is input into the intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground engineering, and generates a disaster early warning report based on the disaster monitoring results, which not only improves the comprehensiveness and accuracy of monitoring and early warning, but also significantly enhances construction safety and emergency response speed.

[0115] An embodiment of the present application also provides a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the above-mentioned underground engineering disaster monitoring and early warning method is implemented.

[0116] The embodiment of the present application also provides a computer program product, including a computer program or instructions, characterized in that when the computer program or instructions are executed, the above-mentioned underground engineering disaster monitoring and early warning method is implemented.

[0117] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0118] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0119] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0120] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one or a combination of multiple of the following technologies known in the art: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0121] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

Claims

1. A method for monitoring and early warning of underground engineering disasters, characterized in that: The following steps are involved: Obtain multi-source heterogeneous data on geological structure and structural stress distribution inside underground projects; Performing data preprocessing on the multi-source heterogeneous data, and fusing the preprocessed multi-source heterogeneous data; The fused data is input into an intelligent monitoring and early warning model, which outputs disaster monitoring results of the underground project and generates a disaster early warning report based on the disaster monitoring results, wherein the intelligent monitoring and early warning model analyzes the fused data based on a machine learning algorithm to obtain the disaster monitoring results.

2. The underground engineering disaster monitoring and early warning method according to claim 1 is characterized in that: The multi-source heterogeneous data includes image data, real-time detection data and abnormal data, wherein the image data includes: underground engineering topography, geological structure and engineering structure; the real-time detection data includes: surrounding rock data and surface deformation data; the abnormal data includes groundwater abnormal data in underground engineering pipelines, rock and soil abnormal data in underground engineering and engineering structure abnormal data.

3. The underground engineering disaster monitoring and early warning method according to claim 1 is characterized in that: The data preprocessing includes one or more of data cleaning, data feature extraction and data standardization.

4. The underground engineering disaster monitoring and early warning method according to claim 1 is characterized in that: The fusion preprocessed multi-source heterogeneous data includes: Constructing a judgment matrix based on the preprocessed multi-source heterogeneous data, and calculating the corresponding weight of each data source based on the judgment matrix; Inputting the multi-source heterogeneous data and the corresponding weight of each data source into an expert model, wherein the expert model generates an adjusted corresponding weight of each data source; The preprocessed multi-source heterogeneous data are weightedly fused according to the corresponding weight of each data source after the adjustment.

5. The underground engineering disaster monitoring and early warning method according to claim 1 is characterized in that: The machine learning algorithm includes a Transformer knowledge graph extraction algorithm and an ILocal-BART algorithm, and the intelligent monitoring and early warning model includes an integrated learning system, a real-time streaming data analysis system and a feedback early warning system. The integrated learning system uses a Transformer knowledge graph extraction algorithm to extract knowledge features from multi-source heterogeneous data after fusion preprocessing to construct a knowledge graph. The real-time streaming data analysis system uses an ILocal-BART algorithm to perform incremental clustering analysis on real-time multi-source heterogeneous data to identify abnormal data and generate early warning information. The feedback early warning system is used to accurately push the early warning information to the target terminal.

6. The underground engineering disaster monitoring and early warning method according to claim 5 is characterized in that: The processing of the Transformer knowledge graph extraction algorithm includes: Use the fastText embedding model and the BERT embedding model to convert multi-source heterogeneous data into corresponding word vector sequences, and classify the word vector sequences into categories to generate category labels; The word vector sequence is input into a Transformer model, and the Transformer model outputs feature extraction to generate a context-related word vector sequence; The characteristic function of the CRF layer is used to capture the relationship between the word vector sequence and the category label, and the optimal label sequence of each word vector is selected to generate the corresponding entity name; Based on the entity naming, the fastText embedding model, the BERT embedding model and the Transformer model are used again to extract knowledge and build a knowledge graph.

7. The underground engineering disaster monitoring and early warning method according to claim 5 is characterized in that: The processing of the ILocal-BART algorithm includes: Acquire multi-source heterogeneous data received in real time; Perform incremental clustering calculations on multi-source heterogeneous data in each sliding window to generate cluster results, and select the winning cluster with the largest selection function value to match and confirm abnormal data; Generate warning information based on the abnormal data.

8. An underground engineering disaster monitoring and early warning system, characterized in that: include: The data acquisition module is used to obtain multi-source heterogeneous data of geological structure and structural stress distribution inside underground projects collected by various types of sensors; A multi-source data fusion processing module is used to perform data preprocessing on the multi-source heterogeneous data and fuse the preprocessed multi-source heterogeneous data; The intelligent monitoring and early warning module is used to input the fused data into the intelligent monitoring and early warning model, and the intelligent monitoring and early warning model outputs the disaster monitoring results of the underground project and generates a disaster early warning report based on the disaster monitoring results, wherein the intelligent monitoring and early warning model analyzes the fused data based on a machine learning algorithm to obtain the disaster monitoring results.

9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by the processor, it is used to implement the underground engineering disaster monitoring and early warning method as described in any one of claims 1-7.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed, the underground engineering disaster monitoring and early warning method as described in any one of claims 1-7 is implemented.

Images