Geotechnical engineering investigation safety monitoring system and method based on artificial intelligence

Abstract

The application discloses a kind of based on artificial intelligence's geotechnical engineering survey safety monitoring system and method, it is related to rock-soil mass deformation safety monitoring technical field, specifically including the following steps: obtaining multi-source data, including the deformation of rock-soil mass, stress and environmental data, model input data source is established;Data fusion processing is carried out;With the result of data fusion processing as input, through data simulation, the stress and deformation correlation of rock-soil mass is obtained;Based on deep learning, with the stress and deformation correlation of rock-soil mass as basement, through the method of time series prediction, short-term recursive calculation model and long-term trend analysis model are respectively constructed;Real-time multi-source data value is obtained, the displacement prediction vector of rock-soil mass is obtained according to the short-term recursive calculation model, compared with the preset threshold value, when exceeding the threshold value, real-time early warning is carried out, and the long-term trend analysis result is visualized.The application enhances engineering safety, improves the long-term stability and reliability of infrastructure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geotechnical deformation safety monitoring, in particular to a geotechnical engineering investigation safety monitoring system and method based on artificial intelligence. BACKGROUND

[0002] Geotechnical engineering investigation is a crucial link in infrastructure construction, involving the evaluation of rock and soil stability and safety. Traditional monitoring methods mainly rely on field measurements and experience-based judgments, such as total station monitoring, inclinometer measurement, and pore water pressure testing. However, these methods have problems such as data update lag, limited measurement points, and precision affected by human factors, making it difficult to meet the requirements of modern engineering for high precision, real-time performance, and intelligence. At the same time, due to the complex stress of rock and soil and the nonlinear and time-varying characteristics of deformation, traditional monitoring methods are difficult to accurately predict future stability, affecting the accuracy of engineering safety evaluation. In recent years, the development of Internet of Things and artificial intelligence technology has promoted the intelligentization of geotechnical engineering monitoring. Internet of Things can realize large-scale data collection, and artificial intelligence can improve prediction accuracy through big data analysis and deep learning. However, existing monitoring methods still face difficulties in multi-source data fusion, traditional time series models do not fully consider the nonlinear characteristics of rock and soil, and geotechnical engineering investigation relies too much on human resources, making it difficult to detect small rock and soil deformations and unable to balance short-term mutations and long-term trend analysis. Therefore, there is an urgent need for more efficient, accurate, and intelligent rock and soil deformation safety monitoring methods. SUMMARY

[0003] The present application aims to provide a geotechnical engineering investigation safety monitoring system and method based on artificial intelligence to solve the problems in the prior art.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical solution: a geotechnical engineering investigation safety monitoring method based on artificial intelligence, which specifically includes the following steps:

[0005] Step S100, acquire multi-source data, including rock and soil deformation, stress, and environmental data, and establish model input data sources;

[0006] Step S200, perform data fusion processing on the acquired multi-source data;

[0007] Step S300, input the results of data fusion processing, simulate data, and use the finite element method to solve the stress and deformation of continuous rock and soil to obtain the stress and deformation correlation of rock and soil;

[0008] Step S400, based on deep learning, take the stress and deformation correlation of rock and soil as the basis, and use the time series prediction method to construct a short-term recursive calculation model and a long-term trend analysis model respectively;

[0009] Step S500, acquiring real-time multi-source data values, obtaining the displacement prediction vector of the rock mass according to the short-term recursive calculation model, comparing with the preset threshold value, and performing real-time early warning when the threshold value is exceeded, and visualizing the long-term trend analysis result.

[0010] In step S100, multi-source data is acquired, including deformation, stress and environmental data of the rock mass, and a model input data source is established, specifically:

[0011] An Internet of Things device is provided to monitor in real time and acquire deformation, stress and environmental data of the rock mass, and establish a model input data source;

[0012] It includes surface monitoring, underground monitoring, vibration monitoring and environmental monitoring;

[0013] The surface monitoring includes laser ranging and inclination measurement;

[0014] The underground monitoring is used to monitor the pore water pressure of the rock mass;

[0015] The vibration monitoring device includes a MEMS accelerometer and a seismic array;

[0016] The environmental monitoring includes temperature and humidity, rainfall and remote sensing images.

[0017] In step S200, the acquired multi-source data is subjected to data fusion processing, specifically:

[0018] The data fusion processing specifically includes the following steps:

[0019] Step S201, replacing abnormal values of the acquired multi-source data;

[0020] Step S202, setting a time stamp, aligning multi-source data with different space-time based on the time scale between the time stamps, and realizing data fusion of the multi-source data;

[0021] Step S203, keeping different data at the same scale by standardization transformation, specifically: X'=(X-μ) / a; wherein X' represents the data after standardization transformation; X represents the data after data fusion; μ represents the average value of the data X; and a represents the standard deviation of the data X.

[0022] In step S201, the acquired multi-source data is subjected to abnormal value replacement, specifically:

[0023] Step S211, sorting different types of multi-source data in descending order to obtain a descending sequence;

[0024] Step S212, according to the descending sequence, calculating the upper quartile, median and lower quartile of the data; the upper quartile, median and lower quartile are respectively the upper quartile point Q1, median point Q2 and lower quartile point Q3 corresponding to the data in the descending sequence; Wherein, m represents the number of data in the descending sequence, m is a positive integer; Q1 and Q2 are rounded up; Q3 is rounded down;

[0025] Step S213, calculating the difference between the upper quartile and the lower quartile, taking the absolute value of the difference as IQR;

[0026] Step S214, respectively taking the upper quartile and the lower quartile as the upper limit and the lower limit, determining the data greater than the upper quartile plus 1.5 times IQR and the data less than the lower quartile minus 1.5 times IQR as abnormal values;

[0027] Step S215, replacing the abnormal values with the value of the upper quartile plus 1.5 times IQR, and replacing the abnormal values less than the lower quartile minus 1.5 times IQR with the value of the lower quartile minus 1.5 times IQR.

[0028] In step S300, taking the result of data fusion processing as input, solving the stress and deformation of the continuous rock-soil mass by data simulation and using finite element method, obtaining the stress and deformation correlation of the rock-soil mass, specifically:

[0029] Step S301, according to the balance relationship between the stress tensor of the rock-soil mass and the external volume force of the rock-soil mass under static conditions, constructing the balance equation of the rock-soil mass, the balance equation is characterized as: Wherein, δ represents the stress tensor; f represents the external volume force; represents the stress divergence;

[0030] Wherein, the stress tensor δ is composed of normal stress and shear stress;

[0031]

[0032] δ xx represents the normal stress along the x direction (acting on the yz plane);

[0033] δ yy represents the normal stress along the y direction (acting on the xz plane);

[0034] δ zz represents the normal stress along the z direction (acting on the xy plane);

[0035] τ xy represents the shear stress on the xy plane (acting on the y axis along the x direction);

[0036] τ yz represents the shear stress in the yz plane (acting on the z axis in the y direction);

[0037] τ zx represents the shear stress in the zx plane (acting on the x axis in the z direction);

[0038] Due to static equilibrium, the shear stress satisfies symmetry;

[0039] As preferred, the external volume force f includes gravity and the pore water pressure of the rock-soil body;

[0040] The gravity f g = ρg; wherein, ρ represents the density of the rock-soil body; g represents the acceleration of gravity;

[0041] The pore water pressure of the rock-soil body Wherein, ρ w represents the density of the pore water of the rock-soil body; represents the gradient of the pore water head height of the rock-soil body, the negative sign indicates that the direction of the pore water pressure of the rock-soil body is opposite to the direction of the water head gradient, and depends on the direction of water flow;

[0042] In step S302, the stress and strain relationship of the rock-soil body is solved to obtain the displacement vector of the rock-soil body.

[0043] In step S303, the stress and strain relationship of the rock-soil body is solved to obtain the displacement vector of the rock-soil body, specifically including the following steps:

[0044] Step S311, discretizing the rock-soil body into a finite triangular element;

[0045] Step S312, for each triangular element, using a shape function to approximate the deformation displacement to obtain the displacement interpolation relationship of each point in the triangular element;

[0046] Wherein, as preferred, the shape function is specifically:

[0047]

[0048] Wherein, u(x, y, z) represents the deformation displacement of the rock-soil body at the spatial coordinates (x, y, z), which is obtained by interpolating the node displacement through the shape function N i (x, y, z);

[0049] u i is the displacement value of node i, and n is the total number of nodes participating in interpolation;

[0050] The displacement of any point in the element is obtained by summing the products of the shape functions of all nodes and the corresponding node displacements;

[0051] Step S313, calculating a stiffness matrix according to the deformation displacement of each triangular element; the stiffness matrix is used to describe the relationship between the deformation and force of the rock-soil body when subjected to force;

[0052] The specific formula for representing the stiffness matrix is:

[0053] K = ∫ V B T DBdV;

[0054] Wherein, K represents the stiffness matrix; B represents the shape function derivative matrix; B T represents the transpose matrix of the shape function derivative matrix;

[0055] The shape function derivative matrix B is composed of partial derivatives of shape functions with respect to coordinates, and is used to represent the relationship between stress and strain in the triangular element and the node displacement;

[0056] D represents an elastic matrix;

[0057] As preferred, the elastic matrix is determined by the elastic modulus and Poisson's ratio, and is used to convert strain into stress;

[0058] V represents the volume of the triangular element;

[0059] Step S314, calculating the rock-soil body displacement vector according to the stiffness matrix.

[0060] The formula for calculating the rock-soil body displacement vector is:

[0061] KU = f;

[0062] Wherein, U represents the displacement vector, which is a column vector composed of displacement values of all nodes.

[0063] In step S400, based on deep learning, taking the stress and deformation correlation of the rock-soil body as the basis, a short-term recursive calculation model and a long-term trend analysis model are respectively constructed by a time series prediction method, specifically:

[0064] Step S401, constructing a short-term recursive calculation model by a time series prediction method, the short-term recursive calculation model including input gate, forget gate, output gate, cell state and hidden state part;

[0065] Step S402, using a random forest method, constructing multiple subsets by randomly sampling with replacement from the original training data, using the method of step S401 for secondary training on each subset, taking the average value of all secondary training subsets as the final prediction result, and constructing a long-term trend analysis model through secondary training results for a certain time.

[0066] In step S500, real-time multi-source data values are obtained, displacement prediction vectors of the rock-soil mass are obtained according to the short-term recursive calculation model, a preset threshold value is compared, real-time early warning is performed when the threshold value is exceeded, long-term trend analysis results are visualized, and specifically:

[0067] In step S501, real-time multi-source data values are obtained, and displacement prediction vectors of the rock-soil mass are obtained according to the short-term recursive calculation model.

[0068] In step S502, a displacement threshold value of the rock-soil mass is set, a preset threshold value is compared, and real-time early warning is performed when the threshold value is exceeded.

[0069] In step S503, long-term trend analysis results are visualized.

[0070] A rock-soil engineering investigation safety monitoring system based on artificial intelligence, comprising a data acquisition module, a data fusion module, a correlation analysis module, a deep learning module, and an intelligent early warning and visualization module.

[0071] The data acquisition module is used to equip an Internet of Things device, obtain deformation, stress, and environmental data of the rock-soil mass, and establish a model input data source.

[0072] The data fusion module is used to fuse and process the collected multi-source data, including the processing steps of abnormal value replacement, time alignment, and data standardization, match data from different sources through timestamps, and ensure data scale consistency by using a standardization method.

[0073] The correlation analysis module is used to establish a correlation between stress and deformation of the rock-soil mass based on the finite element method.

[0074] The deep learning module is used to construct a short-term recursive calculation model and a long-term trend analysis model based on a time series prediction method, and introduce an attention mechanism to improve prediction accuracy, so as to fully consider the nonlinear and time-varying characteristics of the rock-soil mass.

[0075] The intelligent early warning and visualization module is used to perform real-time early warning and visualization analysis according to the prediction results.

[0076] Compared with the prior art, the method of the present application introduces the Internet of Things and deep learning technology, significantly improves the intelligent level of geotechnical engineering monitoring, can realize large-scale and real-time data acquisition, and can improve the monitoring accuracy through multi-source data fusion to avoid the errors caused by the limited measurement points and data lag of the traditional method; at the same time, the intelligent prediction model based on deep learning can fully consider the nonlinear and time-varying characteristics of the rock and soil mass, greatly improve the accuracy of deformation trend and stability prediction; in addition, combined with the intelligent early warning mechanism, real-time response to geological environment changes can be realized, potential risks can be found in advance and effective measures can be taken to enhance engineering safety, improve the long-term stability and reliability of infrastructure, and reduce the waste of human resources. BRIEF DESCRIPTION OF DRAWINGS

[0077] Fig. 1 A step schematic diagram of the geotechnical engineering investigation safety monitoring method based on artificial intelligence of the present application;

[0078] Fig. 2 A structure schematic diagram of the geotechnical engineering investigation safety monitoring system based on artificial intelligence of the present application. DETAILED DESCRIPTION

[0079] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0080] Embodiment: as shown in the figure, the present application provides a technical solution, a geotechnical engineering investigation safety monitoring method based on artificial intelligence, as shown in the figure, the geotechnical engineering investigation safety monitoring method specifically includes the following steps: Figs. 1-2 Fig. 1 Step S100, acquiring multi-source data, including deformation, stress and environmental data of rock and soil mass, establishing model input data source;

[0081] Step S200, data fusion processing is performed on the acquired multi-source data;

[0082] Step S300, taking the result of data fusion processing as input, solving the stress and deformation of continuous rock and soil mass by data simulation and using finite element method, obtaining the stress and deformation correlation of rock and soil mass;

[0083] Step S300, taking the result of data fusion processing as input, solving the stress and deformation of continuous rock and soil mass by data simulation and using finite element method, obtaining the stress and deformation correlation of rock and soil mass;

[0084] Step S400, based on deep learning, taking the stress and deformation correlation of rock and soil mass as the basis, constructing short-term recursive calculation model and long-term trend analysis model respectively by time series prediction method;​

[0085] Step S500, acquiring real-time multi-source data values, obtaining displacement prediction vectors of the rock-soil mass according to the short-term recursive calculation model, comparing with a preset threshold value, and performing real-time early warning when the threshold value is exceeded, and visualizing long-term trend analysis results.

[0086] In step S100, multi-source data is acquired, including deformation, stress and environmental data of the rock-soil mass, and model input data sources are established, specifically:

[0087] IoT devices are equipped to monitor in real time, acquire deformation, stress and environmental data of the rock-soil mass, and establish model input data sources;

[0088] It includes surface monitoring, underground monitoring, vibration monitoring and environmental monitoring;

[0089] The surface monitoring includes laser ranging and inclination measurement;

[0090] The underground monitoring is used to monitor the pore water pressure of the rock-soil mass;

[0091] The vibration monitoring device includes a MEMS accelerometer and a seismic array;

[0092] The environmental monitoring includes temperature and humidity, rainfall and remote sensing images.

[0093] In step S200, the acquired multi-source data is subjected to data fusion processing, specifically:

[0094] The data fusion processing specifically includes the following steps:

[0095] Step S201, replacing abnormal values of the acquired multi-source data;

[0096] Step S202, setting time stamps, aligning multi-source data with different space-time based on the time scale between the time stamps, and realizing data fusion of the multi-source data;

[0097] Step S203, keeping different data at the same scale by standardization transformation, specifically: X'=(X-μ) / a; wherein X' represents the data after standardization transformation; X represents the data after data fusion; μ represents the average value of the data X; and a represents the standard deviation of the data X.

[0098] In step S201, the acquired multi-source data is subjected to abnormal value replacement, specifically:

[0099] Step S211, sorting different types of multi-source data in descending order to obtain a descending sequence;

[0100] Step S212, according to the descending sequence, calculating the upper quartile, median and lower quartile of the data; the upper quartile, median and lower quartile are respectively the upper quartile point Q1, the median point Q2 and the lower quartile point Q3 corresponding to the data in the descending sequence; Wherein, m represents the number of data in the descending sequence, m is a positive integer; Q1 and Q2 are rounded up; Q3 is rounded down;

[0101] Step S213, calculating the difference between the upper quartile and the lower quartile, taking the absolute value of the difference as IQR;

[0102] Step S214, respectively taking the upper quartile and the lower quartile as the upper limit and the lower limit, determining the data greater than the upper quartile plus 1.5 times IQR and the data less than the lower quartile minus 1.5 times IQR as abnormal values;

[0103] Step S215, replacing the abnormal values with the value of the upper quartile plus 1.5 times IQR, and replacing the abnormal values less than the lower quartile minus 1.5 times IQR with the value of the lower quartile minus 1.5 times IQR.

[0104] In step S300, the results of data fusion processing are input, and the stress and deformation of the continuous rock-soil mass are solved by data simulation and finite element method to obtain the stress and deformation correlation of the rock-soil mass, specifically:

[0105] Step S301, according to the balance relationship between the stress tensor of the rock-soil mass and the external volume force of the rock-soil mass under static conditions, constructing the balance equation of the rock-soil mass, the balance equation is characterized as: Wherein, δ represents the stress tensor; f represents the external volume force; represents the stress divergence;

[0106] Wherein, the stress tensor δ is composed of normal stress and shear stress;

[0107]

[0108] δ xx represents the normal stress along the x direction (acting on the yz plane);

[0109] δ yy represents the normal stress along the y direction (acting on the xz plane);

[0110] δ zz represents the normal stress along the z direction (acting on the xy plane);

[0111] τ xy represents the shear stress on the xy plane (acting on the y axis along the x direction);

[0112] τ yz represents the shear stress in the yz plane (acting on the z axis in the y direction);

[0113] τ zx represents the shear stress in the zx plane (acting on the x axis in the z direction);

[0114] Due to static equilibrium, the shear stress satisfies symmetry;

[0115] As preferred, the external volume force f includes gravity and the pore water pressure of the rock-soil body;

[0116] The gravity f g = ρg; wherein, ρ represents the density of the rock-soil body; g represents the acceleration of gravity;

[0117] The pore water pressure of the rock-soil body Wherein, ρ w represents the density of the pore water of the rock-soil body; represents the gradient of the pore water head height of the rock-soil body, the negative sign indicates that the direction of the pore water pressure of the rock-soil body is opposite to the direction of the water head gradient, and depends on the direction of water flow;

[0118] Step S302, solving the stress and strain relationship of the rock-soil body to obtain the displacement vector of the rock-soil body.

[0119] In step S303, the stress and strain relationship of the rock-soil body is solved to obtain the displacement vector of the rock-soil body, specifically including the following steps:

[0120] Step S311, discretizing the rock-soil body into a finite triangular element;

[0121] Step S312, for each triangular element, using a shape function to approximate the deformation displacement to obtain the displacement interpolation relationship of each point in the triangular element;

[0122] Wherein, as preferred, the shape function is specifically:

[0123]

[0124] Wherein, u(x, y, z) represents the deformation displacement of the rock-soil body at the spatial coordinates (x, y, z), which is obtained by interpolating the node displacement through the shape function N i (x, y, z);

[0125] u i is the displacement value of node i, and n is the total number of nodes participating in interpolation;

[0126] The displacement of any point in the element is obtained by summing the products of the shape functions of all nodes and the corresponding node displacements;

[0127] Step S313, calculating a stiffness matrix according to the deformation displacement of each triangular element; the stiffness matrix is used to describe the relationship between the deformation and force of the rock-soil body when subjected to force;

[0128] The representation formula of the stiffness matrix is specifically:

[0129] K = ∫ V B T DBdV;

[0130] Wherein, K represents the stiffness matrix; B represents the shape function derivative matrix; B T represents the transpose matrix of the shape function derivative matrix;

[0131] The shape function derivative matrix B is composed of partial derivatives of shape functions with respect to coordinates, and is used to represent the relationship between stress and strain in the triangular element and the node displacement;

[0132] D represents an elastic matrix;

[0133] As preferred, the elastic matrix is determined by the elastic modulus and Poisson's ratio, and is used to convert strain into stress;

[0134] V represents the volume of the triangular element;

[0135] Step S314, calculating the rock-soil body displacement vector according to the stiffness matrix.

[0136] The calculation formula of the rock-soil body displacement vector is:

[0137] KU = f;

[0138] Wherein, U represents the displacement vector, which is a column vector composed of displacement values of all nodes.

[0139] In step S400, based on deep learning, taking the rock-soil body stress and deformation correlation as the basis, a short-term recursive calculation model and a long-term trend analysis model are respectively constructed by a time series prediction method, specifically:

[0140] Step S401, constructing a short-term recursive calculation model by a time series prediction method, the short-term recursive calculation model including an input gate, a forget gate, an output gate, a cell state and a hidden state part;

[0141] Wherein,

[0142] The forget gate is represented as: g t = σ(W g ·[h t-1 ,H t ]+b g ); g tan output vector of the forget gate, whose elements take values in the range [0, 1], for determining which information in the cell state C t-1 at the previous time step needs to be forgotten; σ denotes a Sigmoid activation function that maps input values to the interval [0, 1]; W g denotes a weight matrix of the forget gate; [h t-1 ,H t ] denotes concatenation of the hidden state h t-1 at the previous time step and the input H t at the current time step; b g denotes a bias vector of the forget gate;

[0143] The input gate is characterized by: I t = σ(W i · [h t-1 ,H t ] + b I ); I t denotes an output vector of the input gate, for determining how much new information in the input H t at the current time step is to be added to the cell state; W I denotes a weight matrix of the input gate; b I denotes a bias vector of the input gate;

[0144] The output gate is characterized by: o t = σ(W o · [h t-1 ,H t ] + b o );

[0145] o t denotes an output vector of the output gate, for determining how much information in the cell state C t at the current time step is to be output to the hidden state h t ; W o denotes a weight matrix of the output gate; b o denotes a bias vector of the output gate;

[0146] The cell state is characterized by: C t = g t ⊙ C t-1 + I t ⊙ tanh(W c · [h t-1 ,H t ] + b c ); C t denotes the cell state at the current time step, for storing and transmitting information; ⊙ denotes an element-wise multiplication operation; W c denotes a weight matrix for updating the cell state; b cdenotes the bias vector for updating the cell state; tanh denotes the hyperbolic tangent activation function, which maps the input value to the interval [0, 1];

[0147] The hidden state is represented as: h t = o t tanh(C t ); h t denotes the hidden state at the current time, which is the output of the short-term recurrent calculation model and is transmitted to the next time step, and can also be used as the final output result for prediction;

[0148] As preferred, an attention mechanism is added, specifically:

[0149] P1 = W p1 H; P2 = W p2 H; P3 = W p3 H; P1 denotes a query matrix used to find information related to itself in the attention mechanism;

[0150] P2 denotes a key matrix used to match the query matrix to calculate the relevance;

[0151] P3 denotes a value matrix containing information in the input sequence;

[0152] W p1 , W p2 and W p3 are weight matrices used to generate the query matrix, the key matrix and the value matrix, respectively;

[0153] H denotes the input time series data;

[0154]

[0155] Attention(P1, P2, P3) denotes the output of the attention mechanism, which is the result of weighted summation of the input sequence;

[0156] Softmax denotes a normalization function to obtain the attention weight of each position, so that the sum of the weights is 1;

[0157] d K denotes the dimension of the key vector, and the square root is used to prevent the dot product result from being too large, avoiding the problem of gradient disappearance or gradient explosion.

[0158] Step S402, using the random forest method, a plurality of subsets are constructed by randomly sampling with replacement from the original training data, and the method of step S401 is used for secondary training on each subset, and the average value of all secondary training subsets is used as the final prediction result, and a long-term trend analysis model is constructed through a certain time of secondary training results;

[0159] As preferred, the formula of the long-term trend analysis model is represented as:

[0160]

[0161] wherein, represents the predicted deformation average value at time t+t', wherein t is the current time and t' is the time span;

[0162] l represents the number of subsets constructed in the random forest;

[0163] w j represents the weight of the jth subset;

[0164] f j (H) is the prediction result of the jth subset to the input H;

[0165] In step S500, real-time multi-source data values are obtained, a displacement prediction vector of the rock-soil mass is obtained according to the short-term recursive calculation model, a preset threshold value is compared, and when the threshold value is exceeded, real-time early warning is performed, and the long-term trend analysis result is visualized, specifically:

[0166] Step S501, real-time multi-source data values are obtained, and a displacement prediction vector of the rock-soil mass is obtained according to the short-term recursive calculation model;

[0167] Step S502, a displacement threshold value of the rock-soil mass is set, a preset threshold value is compared, and when the threshold value is exceeded, real-time early warning is performed;

[0168] Step S503, the long-term trend analysis result is visualized.

[0169] Embodiment

[0170] A section of K12+300~K12+700 of a certain highway is a deep excavation slope, with a maximum excavation depth of 35m, the stratum is strongly weathered sandstone with mudstone, joint development, and shallow landslide is prone to occur in the rainy season; the design adopts "prestressed anchor cable + lattice beam" support, which needs to monitor the slope deformation, stress and environmental factors in real time to ensure the safety of construction and operation;

[0171] Monitoring system deployment and data acquisition:

[0172] Sensor selection and arrangement:

[0173] Surface deformation monitoring:

[0174] Laser range finder (Leica TM50, accuracy ±0.3mm): 3 measuring lines are arranged along the slope opening, slope foot and platform, and monitoring points are set every 20m on each measuring line;

[0175] North-compass GNSS receiver (Trimble R10, positioning accuracy ±2 mm + 1 ppm): Reference stations and mobile stations are set up at the top of the slope and the surrounding stable area to monitor three-dimensional displacement;

[0176] Underground stress monitoring:

[0177] Vibrating wire strain gauge (VWP-1000, accuracy ±0.01% FS): Arranged at the anchorage segment of the anchor cable to monitor the loss of prestress;

[0178] Pore water pressure gauge (Dynagage PT-500, range 0-5 MPa, accuracy 0.1% FS): Drilled and buried to the depth of the sliding zone (15 m) to monitor the underground water pressure;

[0179] Environmental monitoring:

[0180] Rain gauge (dumper type, accuracy ±0.5 mm): Arranged at the top of the slope; Temperature and humidity sensor (SHT30, accuracy ±0.5°C, ±2% RH);

[0181] Video monitoring (200 million pixels): Real-time recording of crack development on the slope surface;

[0182] During construction: high-frequency acquisition (1 time / 10 min); During operation: low-frequency acquisition (1 time / h), automatically switched to 1 time / 5 min during rainfall;

[0183] Finite element modeling and stress deformation analysis:

[0184] Geometric model and parameters:

[0185] The slope is simplified as a two-dimensional plane strain model, with triangular elements (minimum edge length 1 m) and 5000+ nodes;

[0186] Geotechnical parameters: Elastic modulus E = 50 MPa, Poisson's ratio v = 0.3, density p = 2.2 g / cm 3 , cohesion c = 20 kPa, internal friction angle

[0187] Balancing equation solution:

[0188] Stress tensor (principal stress): s1 = 0.8 MPa (compressive stress), s3 = 0.3 MPa (compressive stress);

[0189] Body force: gravity f g = 22 kN / m 3 , pore water pressure ( fp = -100 kPa);

[0190] Displacement calculation:

[0191] The shape function adopts linear interpolation: u = N1u1 + N2u2 + N3u3;

[0192] The stiffness matrix K = ∫B T DBdV, the maximum displacement of the slope top is U = 12 mm (the finite element calculation time is about 15 min, and the parallel calculation is optimized to 5 min);

[0193] Short-term recurrent model (LSTM):

[0194] Input features: displacement rate, anchor stress, pore water pressure, rainfall (time window = 72h, step = 1h);

[0195] Model structure: 2-layer LSTM (64 units per layer) + attention mechanism, training data for 1 year of construction period;

[0196] Long-term trend model (random forest):

[0197] Feature engineering: integrate 10 features such as total annual rainfall, seasonal temperature difference, and support structure stress relaxation rate;

[0198] An artificial intelligence-based geotechnical engineering investigation safety monitoring system, as shown in Fig. 2 The geotechnical engineering investigation safety monitoring system includes a data acquisition module, a data fusion module, a correlation analysis module, a deep learning module, and an intelligent early warning and visualization module;

[0199] The data acquisition module is used to equip Internet of Things devices to obtain deformation, stress, and environmental data of the rock and soil mass, and to establish a model input data source;

[0200] The data fusion module is used to fuse and process the collected multi-source data, including the steps of abnormal value replacement, time alignment, and data standardization. Different sources of data are matched through timestamps, and standardized methods are used to ensure data scale consistency;

[0201] The correlation analysis module is used to establish the correlation between the stress and deformation of the rock and soil mass based on the finite element method;

[0202] The deep learning module is used to construct short-term recurrent calculation models and long-term trend analysis models based on time series prediction methods, and introduce attention mechanisms to improve prediction accuracy, taking into full account the nonlinear and time-varying characteristics of the rock and soil mass;

[0203] The intelligent early warning and visualization module is used for real-time early warning and visualization analysis based on the prediction results;

[0204] This module acquires real-time data, calculates the predicted displacement vector of soil and rock mass using a short-term prediction model, and compares it with preset threshold values. An early warning is triggered when the threshold value is exceeded. Simultaneously, the module provides a visual interface for long-term trend analysis, allowing engineering managers to intuitively view monitoring results, optimize decision-making, and improve the safety and stability of geotechnical engineering projects.

[0205] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A geotechnical engineering investigation safety monitoring method based on artificial intelligence, characterized in that: The geotechnical engineering investigation safety monitoring method specifically comprises the following steps: Step S100, acquiring multi-source data, including deformation, stress and environmental data of the rock-soil mass, and establishing a model input data source; Step S200, performing data fusion processing on the acquired multi-source data; Step S300, taking the data fusion processing result as input, solving the stress and deformation of the continuous rock-soil mass by using the finite element method through data simulation, and obtaining the stress and deformation correlation of the rock-soil mass; Step S400, based on deep learning, taking the stress and deformation correlation of the rock-soil mass as a basis, constructing a short-term recursive calculation model and a long-term trend analysis model through a time series prediction method, specifically: Step S401, constructing a short-term recursive calculation model through a time series prediction method, the short-term recursive calculation model comprising an input gate, a forget gate, an output gate, a cell state and a hidden state part; Step S402, using a random forest method, constructing multiple subsets through random sampling with replacement from the original training data, performing secondary training on each subset using the method of step S401, taking the average value of the secondary training of all subsets as the final prediction result, and constructing a long-term trend analysis model through the secondary training results of a certain time; Step S500, acquiring real-time multi-source data values, obtaining a displacement prediction vector of the rock-soil mass according to the short-term recursive calculation model, comparing with a preset threshold value, and performing real-time early warning when the threshold value is exceeded, and visualizing the long-term trend analysis result.

2. The method according to claim 1, wherein the method is characterized by: In step S100, multi-source data is acquired, including deformation, stress and environmental data of the rock-soil mass, and a model input data source is established, specifically: Internet of Things devices are equipped to monitor in real time, acquire deformation, stress and environmental data of the rock-soil mass, and establish a model input data source; It includes surface monitoring, underground monitoring, vibration monitoring and environmental monitoring; The surface monitoring includes laser ranging and inclination measurement; The underground monitoring is used for monitoring the pore water pressure of the rock-soil mass; The vibration monitoring device comprises a MEMS accelerometer and a seismic array; The environmental monitoring includes temperature and humidity, rainfall and remote sensing images.

3. The method of claim 2, wherein the method is based on artificial intelligence. In step S200, the acquired multi-source data is subjected to data fusion processing, specifically: The data fusion processing specifically comprises the following steps: Step S201, replacing abnormal values in the acquired multi-source data; Step S202, setting a time stamp, aligning multi-source data with different space-time based on the time scale between time stamps, and realizing data fusion of the multi-source data; Step S203, maintaining different data at the same scale by using a standardization transformation method, and the specific formula is: X'=(X-μ) / a; wherein X' represents the data after standardization transformation; X represents the data after data fusion; μ represents the average value of the data X; and a represents the standard deviation of the data X.

4. The method according to claim 3, wherein the method is characterized by: In step S201, the acquired multi-source data is subjected to abnormal value replacement, specifically: Step S211, sorting different types of multi-source data in descending order to obtain a descending sequence; Step S212, calculating the upper quartile, the median and the lower quartile of the data according to the descending sequence; the upper quartile, the median and the lower quartile are respectively corresponding data of the upper quartile point Q1, the median point Q2 and the lower quartile point Q3 in the descending sequence; ; ; ; wherein m represents the number of data in the descending sequence, m is a positive integer; Q1 and Q2 are rounded up; Q3 is rounded down. Step S213, calculating the difference between the upper quartile and the lower quartile, and taking the absolute value of the difference as IQR; Step S214, the upper quartile and the lower quartile are respectively taken as the upper limit and the lower limit, data greater than the upper quartile plus 1.5 times IQR and data less than the lower quartile minus 1.5 times IQR are determined as abnormal values; Step S215, the abnormal values are replaced with the value of the upper quartile plus 1.5 times IQR, and the abnormal values less than the lower quartile minus 1.5 times IQR are replaced with the value of the lower quartile minus 1.5 times IQR.

5. The method of claim 4, wherein the method is based on artificial intelligence. In step S300, the stress and deformation of the continuous rock-soil mass are solved by data simulation using the finite element method with the result of the data fusion processing as input, and the stress and deformation correlation of the rock-soil mass is obtained, specifically: Step S301, according to the balance relationship between the stress tensor of the rock-soil mass and the external volume force of the rock-soil mass under static conditions, a balance equation of the rock-soil mass is constructed, and the balance equation is characterized as: ∇·δ+f=0; wherein, δ represents the stress tensor; f represents the external volume force; ∇·δ represents the stress divergence; Step S302, the stress and strain relationship of the rock-soil mass is solved to obtain the displacement vector of the rock-soil mass.

6. The method of claim 5, wherein the method is based on artificial intelligence. In step S303, the stress and strain relationship of the rock-soil mass is solved to obtain the displacement vector of the rock-soil mass, specifically including the following steps: Step S311, the rock-soil mass is discretized into a finite triangular element; Step S312, for each triangular element, the shape function is used to approximate the deformation displacement to obtain the displacement interpolation relationship of each point in the triangular element; Step S313, according to the deformation displacement of each triangular element, the stiffness matrix is calculated; the stiffness matrix is used to describe the relationship between the deformation and the force of the rock-soil mass under stress; Step S314, according to the stiffness matrix, the displacement vector of the rock-soil mass is calculated.

7. The method according to claim 6, wherein the method is characterized by: In step S500, real-time multi-source data values are obtained, the displacement prediction vector of the rock-soil mass is obtained according to the short-term recursive calculation model, the preset threshold value is compared, and real-time early warning is performed when the threshold value is exceeded, and the long-term trend analysis result is visualized, specifically: Step S501, real-time multi-source data values are obtained, and the displacement prediction vector of the rock-soil mass is obtained according to the short-term recursive calculation model; Step S502, set the displacement threshold value of the rock-soil mass, compare the preset threshold value, and perform real-time early warning when the threshold value is exceeded; Step S503, the long-term trend analysis result is visualized.

8. A geotechnical engineering investigation safety monitoring system based on artificial intelligence, applying a geotechnical engineering investigation safety monitoring method based on artificial intelligence according to any one of claims 1-7, characterized in that: The rock-soil engineering investigation safety monitoring system comprises a data acquisition module, a data fusion module, an association analysis module, a deep learning module and an intelligent early warning and visualization module; The data acquisition module is used to equip the Internet of Things device, acquire the deformation, stress and environmental data of the rock-soil mass, and establish a model input data source; The data fusion module is used to fuse the collected multi-source data, including the processing steps of abnormal value replacement, time alignment and data standardization, match data from different sources through time stamp, and ensure the consistency of data scale by using the standardization method; The association analysis module is used to establish the correlation between the stress and deformation of the rock-soil mass based on the finite element method; The deep learning module is used to construct a short-term recursive calculation model and a long-term trend analysis model based on a time series prediction method, and introduce an attention mechanism to improve prediction accuracy, so as to fully consider the nonlinear and time-varying characteristics of the rock-soil mass; The intelligent early warning and visualization module is used for real-time early warning and visualization analysis according to the prediction result.

Images