Geological attribute modeling method and system based on multi-source information joint constraint

Abstract

The invention discloses a geological attribute modeling method and system based on multi-source information joint constraint, and relates to the technical field of geological engineering, and the method comprises the steps: obtaining geological detection data, obtaining geological attribute data after preprocessing, and carrying out the continuous and discontinuous type division and quantification of the geological attribute data; based on the quantized geological attribute data, performing multi-source geological attribute correlation analysis to obtain correlation among different geological attributes; the method comprises the following steps: constructing a geologic structure model of a to-be-modeled region, performing unstructured network division on the model by combining correlation among different geologic attributes, and performing geologic unit grid attribute assignment based on 1D + 2D trend chart constraint for each grid region according to acquired different geologic attribute data, thereby constructing a geologic unit attribute model. According to the method, tunnel geological attribute modeling and geological attribute prediction with higher precision and reliability can be realized.

Description

Technical Field

[0001] The present invention relates to the field of geological engineering technology, and in particular to a geological attribute modeling method and system based on multi-source information joint constraints. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Tunnel construction usually faces complex geological conditions, such as different rock and soil types, crack conditions, groundwater distribution and seismic activity, which directly affect the stability and construction difficulty of the tunnel. Therefore, accurate evaluation and modeling of the geological characteristics around the tunnel is the key to the successful implementation of tunnel engineering. In order to improve the accuracy and reliability of tunnel geological modeling, a method of geological modeling using multi-point geostatistical technology has been extracted to construct a geological model that is more in line with the actual geological conditions. However, multi-point geostatistical technology still faces challenges in data acquisition and processing, especially in the fusion and correlation analysis of multi-source attribute data, and is unable to guarantee the accuracy of the final geological modeling. The main problems are:

[0004] (1) Traditional tunnel geological modeling methods mainly rely on drilling data and geological surveys. Although this method can provide basic geological information, it is costly and time-consuming, and still has problems such as data acquisition limitations and insufficient single data sources. In other words, during tunnel construction, drilling data is often limited by the number and distribution of drilling points, the uneven distribution of geological exploration points, or the hidden nature of cracks and faults, which may lead to the inability to accurately obtain data in certain areas, resulting in missing spatial data. This makes it difficult to fully reflect the complexity and variability of geological characteristics, especially in complex geological environments, where it is even more difficult to obtain sufficient geological data.

[0005] (2) Although geological surveys can obtain more geological information, they are limited by the professional knowledge and experience of investigators and the limitations of survey methods, and their accuracy and reliability are also insufficient. That is, there is a risk that the exploration points do not completely match the actual geological characteristics;

[0006] (3) In addition, the information from a single data source is often insufficient to fully reflect the true situation of geological characteristics, which affects the accuracy and reliability of geological modeling. The data from different sources are of uneven quality and may also differ in accuracy, coverage, timestamp, etc., resulting in inaccurate final modeling. Summary of the invention

[0007] To address the deficiencies of the above-mentioned prior art, the present invention provides a geological attribute modeling method and system based on multi-source information joint constraints. The correlation between multi-source geological attributes is analyzed, and the geological structure model is gridded according to the correlation. Then, 1D+2D trend chart constraints are used to assign values ​​to the geological unit grid attributes, thereby constructing a geological unit attribute model. For the attribute model, a dynamic fuzzy neural network and a multi-attribute probabilistic neural network are used to analyze the distribution and variability of geological attributes in space, so as to better describe and predict geological characteristics, optimize the tunnel geological attribute model, and effectively improve the accuracy and reliability of tunnel geological attribute modeling and geological attribute prediction.

[0008] In a first aspect, the present invention provides a geological attribute modeling method based on joint constraints of multi-source information.

[0009] A geological attribute modeling method based on joint constraints of multi-source information, comprising:

[0010] Obtain geological exploration data, obtain geological attribute data after preprocessing, and classify and quantify the geological attribute data into continuous and discontinuous types;

[0011] Based on the quantified geological attribute data, multi-source geological attribute correlation analysis is performed to obtain the correlation between different geological attributes;

[0012] Construct a geological structure model of the area to be modeled, divide the model into unstructured networks based on the correlation between different geological attributes, and for each grid area, assign the geological unit grid attributes based on 1D+2D trend chart constraints according to the different geological attribute data obtained, and construct a geological unit attribute model.

[0013] Further technical solutions include:

[0014] Based on the geological unit attribute model, the nonlinear relationship between different attributes in the model is extracted through dynamic fuzzy neural network, and then the multi-attribute probabilistic neural network modeling method is used to update and aggregate the geological attributes of each grid node, so as to output the predicted value of the geological attributes of each grid unit in the geological unit attribute model that changes with time and the excavation process.

[0015] In a second aspect, the present invention provides a geological attribute modeling system based on joint constraints of multi-source information.

[0016] A geological attribute modeling system based on multi-source information joint constraints, comprising:

[0017] The data acquisition and preprocessing module is used to acquire geological exploration data, obtain geological attribute data after preprocessing, and classify and quantify the geological attribute data into continuous and discontinuous types;

[0018] The geological attribute analysis module is used to perform multi-source geological attribute correlation analysis based on the quantified geological attribute data to obtain the correlation between different geological attributes;

[0019] The geological attribute modeling module is used to construct a geological structure model of the area to be modeled. It divides the model into unstructured networks based on the correlation between different geological attributes. For each grid area, according to the different geological attribute data obtained, the geological unit grid attribute is assigned based on the 1D+2D trend chart constraints to construct a geological unit attribute model.

[0020] Further technical solutions include:

[0021] The geological attribute prediction module is used to extract the nonlinear relationship between different attributes in the model based on the geological unit body attribute model through dynamic fuzzy neural network, and then adopts the multi-attribute probabilistic neural network modeling method to update and aggregate the geological attributes of each grid node, so as to output the predicted value of the geological attributes of each grid unit in the geological unit body attribute model that changes with time and the excavation process.

[0022] In a third aspect, the present invention further provides an electronic device, comprising: a memory for storing executable instructions; and a processor for implementing the above-mentioned geological attribute modeling method based on joint constraints of multi-source information when executing the executable instructions stored in the memory.

[0023] In a fourth aspect, the present invention further provides a computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the above-mentioned geological attribute modeling method based on joint constraints of multi-source information.

[0024] In a fifth aspect, the present invention also provides a computer program product, which includes executable instructions stored in a computer-readable storage medium; wherein, when a processor of an electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the above-mentioned geological attribute modeling method based on joint constraints of multi-source information is implemented.

[0025] One or more of the above technical solutions have the following beneficial effects:

[0026] 1. The present invention provides a geological attribute modeling method and system based on multi-source information joint constraints. The correlation between multi-source geological attributes is analyzed, and the geological structure model is gridded according to the correlation. Then, 1D+2D trend chart constraints are used to assign values ​​to the geological unit grid attributes to construct a geological unit attribute model. For the attribute model, the distribution and variability of geological attributes in space are analyzed through dynamic fuzzy neural networks and multi-attribute probabilistic neural networks to better describe and predict geological characteristics, optimize the tunnel geological attribute model, and effectively improve the accuracy and reliability of tunnel geological attribute modeling and geological attribute prediction.

[0027] 2. The present invention can clearly distinguish different types of geological data by accurately quantitatively analyzing continuous attributes and discontinuous attributes, avoiding data confusion problems that may occur in traditional methods; by using statistical methods such as the Spearman rank coefficient correlation coefficient to perform correlation analysis on multi-source attributes, it can reveal the inherent relationship and interaction between attributes, help discover the mutual influence between different geological features, provide multi-dimensional data support for tunnel modeling, and provide a quantitative basis for risk assessment. Through comprehensive attribute analysis, potential geological risks can be accurately identified, such as fracture distribution zones, areas with high water content, and areas with high fragmentation, thereby providing more targeted early warnings for tunnel construction.

[0028] 3. The present invention can effectively overcome the discretization problem that may occur in traditional modeling methods through attribute coarsening, grid assignment and removal of sample deviation, and ensure the rationality and consistency of attribute value assignment through fine spatial constraint means; multi-attribute modeling can simultaneously consider multiple key attributes such as crack distribution, water content, degree of fragmentation, etc., to form a more refined attribute body model, which provides rich multi-dimensional support for subsequent tunnel construction design, risk assessment, resource exploration and other work; by introducing deterministic crack parameters and combining point constraint and surface constraint modeling methods, the spatial distribution and structural characteristics of cracks can be more accurately reproduced in the geological model. Compared with the traditional local modeling method, the present invention can make full use of spatial statistical data, can capture the spatial variability and trend of geological characteristics in a larger range, and compared with the traditional statistical modeling method, this method is more intuitive and clear, and can effectively improve the physical interpretability of the model. In addition, the present invention can significantly improve the model accuracy and reduce the error of the geological model through fine grid division and constraint assignment based on 1D+2D trend graph.

[0029] 4. The present invention enhances the application and control capabilities of crack parameters, which not only helps to accurately model the geometric characteristics of cracks, but also can effectively control the tunnel construction process. Through deterministic modeling of cracks, it can provide a specific basis for grouting, support design, etc., and provide effective protection for engineering safety.

[0030] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0032] Figure 1 The present invention is a flowchart of a geological attribute modeling method based on multi-source information joint constraints according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] It should be noted that the following detailed descriptions are exemplary only, are intended to describe specific embodiments, are intended to provide further explanation of the present invention, and are not intended to limit exemplary embodiments according to the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art to which the present invention belongs. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0034] Embodiment 1

[0035] This embodiment provides a geological attribute modeling method based on multi-source information joint constraints, such as Figure 1 As shown, the specific steps include:

[0036] Step S1, obtaining geological exploration data, obtaining geological attribute data after preprocessing, and classifying and quantifying the geological attribute data into continuous and discontinuous types;

[0037] Step S2: Based on the quantified geological attribute data, multi-source geological attribute correlation analysis is performed to obtain the correlation between different geological attributes;

[0038] Step S3, construct a geological structure model of the area to be modeled, combine the correlation between different geological attributes, divide the model into unstructured networks, and for each grid area, assign the geological unit grid attributes based on the 1D+2D trend map constraints according to the different geological attribute data obtained, and construct a geological unit attribute model.

[0039] Further technical solutions include:

[0040] Step S4: Based on the geological unit body attribute model, the nonlinear relationship between different attributes in the model is extracted through a dynamic fuzzy neural network, and then the geological attributes of each grid node are updated and aggregated using a multi-attribute probabilistic neural network modeling method, so as to output the predicted value of the geological attributes of each grid unit in the geological unit body attribute model that changes with time and the excavation process.

[0041] The following content introduces the geological attribute modeling method based on joint constraints of multi-source information proposed in this embodiment in more detail.

[0042] In step S1, geological exploration data is acquired, geological attribute data is obtained after preprocessing, and the geological attribute data is classified into continuous and discontinuous types and quantified.

[0043] In this embodiment, during the tunnel geological modeling process, the measured data of geological exploration in the tunnel area, such as cracks, resistivity, seismic waves and other data, are first obtained, and then the acquired data are inverted, extrapolated or estimated using inversion technology (i.e., preprocessing) to obtain corresponding key geological attribute data, such as water content based on resistivity inversion and degree of fragmentation based on seismic wave inversion. These attribute data are continuous or discrete in space, so these data are appropriately quantified and processed for subsequent analysis and modeling.

[0044] Specifically, the geological exploration data of the tunnel area (including crack width, crack distribution, resistivity, seismic waves, etc.) are obtained, and three attribute data of cracks, water content, and degree of fragmentation are inverted. The attribute data are divided into types of continuity and discreteness, and different data quantification preprocessing is performed according to the different attribute data types. Among them, the water content is obtained through resistivity inversion. Usually, in the tunnel, the resistivity of the area with high groundwater content is high, and the resistivity of the area with low water content is low; the degree of fragmentation is obtained by inversion through the propagation speed of seismic waves. The propagation speed of seismic waves is related to the degree of fragmentation of the rock layer. Generally speaking, the propagation speed of seismic waves in the fragmented area is slower.

[0045] Furthermore, the continuous attribute data is standardized or normalized to eliminate dimensional differences and numerical ranges; the non-continuous attribute data is encoded, such as using one-hot encoding or label encoding, so that it can be used for subsequent numerical calculations and association rule mining.

[0046] In this embodiment, for the three types of attribute data of fissures, water content, and degree of fragmentation, the width of fissures and the water content inverted based on resistivity belong to typical continuous attributes, and numerical expressions can be obtained through quantization processing. Specifically, the width of fissures usually has a continuous numerical range. To quantify the width of fissures, standardization or normalization can be performed on it so that it can be analyzed on a unified scale; the water content inverted based on resistivity usually shows continuous changes related to groundwater saturation and porosity, and there is usually an inverse relationship between the magnitude of the resistivity value and the water content. Normalization or standardization processing of the resistivity value is carried out.

[0047] The size of fissures and the degree of fragmentation inverted based on the seismic wave propagation velocity, etc. all belong to typical discontinuous attributes. Among them, the seismic wave propagation velocity is negatively correlated with the degree of fragmentation of the medium. Therefore, during the process of inverting the degree of fragmentation, the seismic wave propagation velocity data can be standardized and normalized, or the inverse of the seismic wave propagation velocity data can be transformed to obtain continuous numerical values of fragmentation. Then, for this continuous numerical value of fragmentation, it is converted into different grades, such as mild, moderate, severe, etc., and used as a discontinuous attribute for calculation. Specifically, for discrete attributes (i.e., discontinuous attributes) such as the existence of fissures, rock types, and fragmentation degree categories, for such discrete attributes, according to characteristics such as the size and shape of fissures, they are divided into different categories. For example, according to the size of fissures, they can be divided into "large fissures" (assigned a value of 3), "medium fissures" (assigned a value of 2), and "small fissures" (assigned a value of 1), and analyzed with unified numerical values; according to the degree of fragmentation in geology, it can be divided into multiple grades, such as "no fragmentation", "mild fragmentation", "moderate fragmentation", "severe fragmentation", etc. These fragmentation degree types can be quantitatively analyzed through water content information, and further quantitatively assigned through the water inflow. For example, "no fragmentation" (weak water-richness, q ≤ 0.1 L / m.s) is encoded as 0, "mild fragmentation" is encoded as 1 (medium water-richness, 0.1 < q ≤ 1.0 L / m.s), "moderate fragmentation" is encoded as 2 (strong water-richness, 1 < q ≤ 5.0 L / m.s), and "severe fragmentation" is encoded as 3 (extremely strong water-richness, q > 5.0 L / m.s), thus completing the assignment.

[0048] In step S2, based on the quantified geological attribute data, multi-source geological attribute correlation analysis is carried out to obtain the correlation between different geological attributes.

[0049] Specifically, by calculating the Spearman rank coefficient correlation coefficient between multi-source geological attributes, the correlation relationship between different geological attributes (such as fissures, water content, and degree of fragmentation) is evaluated, which specifically includes the following steps:

[0050] Step S2.1: Through the above step S1, the geological attribute data such as cracks, water content and degree of fragmentation of the tunnel area are obtained, and the obtained data are preprocessed, including: checking the missing values, outliers and duplicate values ​​in the data, and performing appropriate processing, such as filling missing values, deleting outliers or duplicate values, etc.

[0051] Step S2.2, using the Spearman rank correlation coefficient to evaluate the linear correlation between different geological attributes, measuring the linear correlation between two variables. The Spearman rank correlation coefficient between any two attributes in the attribute matrix is ​​calculated to obtain a correlation matrix, considering the correlation between the three geological attributes of attribute A (such as crack width), B (such as water content), and C (such as fragmentation degree), and constructing the following correlation matrix:

[0052]

[0053] Among them, r AB Represents the Spearman rank correlation coefficient between attributes A and B; r AC represents the Spearman rank correlation coefficient between attributes A and C; r BC Represents the Spearman rank correlation coefficient between attributes B and C.

[0054] By constructing the above correlation matrix, the correlation between multi-source geological attributes can be further clarified, laying the foundation for the subsequent uneven grid division of the address model.

[0055] In step S3, a multi-point geostatistical method is used to simulate the spatial distribution and variability of geological characteristics to build a geological unit attribute model. Specifically, the following steps are included:

[0056] Step S3.1: construct a geological structure model of the area to be modeled, and divide the model into unstructured networks based on the correlation between different geological attributes to describe the three-dimensional geological spatial structure of the tunnel area.

[0057] Specifically, when constructing the geological structure model, the inversion constraint method is used to ensure that the spatial boundary of the geological unit can accurately reflect the actual geological structure, and the surface constraint conditions of the geological unit are determined through exploration data, seismic inversion results and geological knowledge. On the basis of the surface constraint conditions, according to the distribution of fractures or tunnel faces, the geological unit grid is further divided into uneven parts through unstructured grid division to obtain more small units.

[0058] The determination of the above surface constraint conditions includes:

[0059] (1) Based on the exploration data (such as drilling and logging data), the spatial distribution information of the boundary of the geological unit body is determined. For example, the distribution and depth of the underground rock layer can be identified by the location and depth of the drill hole, and then the surface boundary function of the geological body can be determined, which can be expressed as:

[0060] f(x,y,z)=0;

[0061] In the above formula, f(x,y,z) is a function that describes the shape of the boundary surface of the geological unit, and (x,y,z) represents the spatial coordinates.

[0062] (2) The change in the velocity of seismic waves can reflect the boundaries of different geological units. For example, changes in fracture zones, faults, and rock layers will affect the propagation of seismic waves. The constraint surface is determined by this, which can be expressed as:

[0063] v 1 (x,y,z)≠v 2 (x,y,z);

[0064] In the above formula, v 1 and v 2 Represent the wave velocities of geological body 1 and geological body 2 respectively.

[0065] (3) Surface constraints of geological units also include:

[0066]

[0067] In the above formula, f i (x, y, z) represents the surface boundary function of the i-th geological unit obtained by inversion, g i The actual measurement data.

[0068] Furthermore, the geological unit model is grid-unevenly divided, that is, the face is divided according to the specific needs of the tunnel area, the distribution of fractures, the degree of fracture friction and other factors. In the division process, considering the complex geological environment around the tunnel, a single grid size may not be able to effectively express the spatial variability of geological properties. Therefore, a multi-level grid division method is adopted to divide the area with complex geological conditions of the face with finer grids, and use coarser grids for the peripheral area.

[0069] In this embodiment, according to the correlation between the multi-source geological attributes, a multi-level grid division method is used to divide the grid unevenly, so as to achieve the uneven grid division of the geological unit model, including:

[0070] (1) Based on the correlation between fracture data and water content, a finer mesh is used for areas with dense fractures during mesh generation;

[0071] (2) Areas with a higher degree of fragmentation usually have more cracks and larger crack widths. When dividing the grid, these areas can be divided into smaller grid units;

[0072] (3) Areas with severe fragmentation usually have higher water content. By introducing the correlation between water content and fragmentation into grid division, the density of the grid can be reasonably adjusted in the complex geological environment around the tunnel.

[0073] Step S3.2: for each grid area, according to the different geological attribute data obtained, the geological unit grid attribute assignment based on the 1D+2D trend map constraints is performed, the multi-attribute multi-dimensional numerical analysis is expanded, and the geological unit attribute model is preliminarily constructed.

[0074] Specifically, 1D trend graphs are used to describe the changes of a certain attribute over time or sequence. Through deep learning methods, trends and patterns can be learned from time series data, and time dependencies can be captured; 2D trend graphs are used to describe attribute changes in the spatial dimension; by combining 1D (time) and 2D (space) trend graphs, deep learning models can be used to simultaneously learn the relationship between time and space, so that spatial and temporal attribute predictions can be made, achieving more accurate modeling and prediction of geological attributes, and more accurately estimating the geological characteristics in the tunnel area.

[0075] The construction of the above geological unit attribute model specifically includes the following steps:

[0076] Step S3.2.1, coarsening of geological attribute data. First, the acquired geological attribute data (cracks, water content, and fragmentation) are spatially downscaled, that is, the geological attribute data are grouped or divided according to a certain spatial resolution, and the original high-resolution geological attribute data (such as crack width, water content, etc.) are mapped to a larger and coarser grid through gridding, and converted into a low-resolution form. Through this spatial downscaling process, the complexity of the data can be reduced, and the data can be mapped to a larger grid, which can reduce the number of data points contained in each grid cell.

[0077] Step S3.2.2, Grid mapping of geological attribute data. Using refined grid subdivision technology, select the appropriate grid size according to the actual needs of the region, divide the region into multiple grid cells, each grid cell represents a certain spatial range, and assign a geological attribute value to each grid cell. In order to remove the deviation caused by uneven data samples or noise, the sample deviation removal process is adopted. By de-averaging the sample data, the values ​​of all data points are adjusted to be consistent with the global mean, so as to reduce the impact of local sample imbalance on the results.

[0078] Step S3.2.3, for the geological attribute values ​​assigned to each grid unit, generate a 1D time change trend graph of the geological attribute to show the law of geological attribute change over time, and use the generated 1D trend graph as a constraint to correct the grid unit assignment result; at the same time, by using a two-dimensional interpolation method, generate a 2D spatial distribution trend graph of the geological attribute on a cross section or other two-dimensional plane to show the law of geological attribute change with spatial position, and use the generated 2D trend graph as a constraint to correct the grid unit assignment result. Take the 2D trend graph as a constraint for correction as an example, if the attribute value of a grid unit does not match the spatial change in the trend graph, the attribute value of the grid unit can be adjusted according to the guidance of the 2D trend graph to ensure that it conforms to the actual spatial change law.

[0079] Step S3.2.4, based on the 2D trend graph, the depth from the surface or tunnel entrance is introduced to form a 3D trend graph, which is used to show the change trend of properties such as cracks, resistivity inversion water content, etc. in different strata under vertical changes at different depths. On this basis, based on the data of three dimensions in space, a complete 3D spatial distribution trend graph is generated through spatial interpolation, and the attribute values ​​of the grid cells are adjusted to make the attribute changes more consistent in three-dimensional space. For example, if the attribute values ​​of a certain grid cell deviate from the three-dimensional distribution trend in the actual geological environment, these attribute values ​​can be made consistent in different depths and lateral distributions after correction.

[0080] Based on the joint application of the above 1D+2D trend map constraints, by combining the 1D trend map, 2D trend map and 3D trend map constraints, each grid cell is finally assigned an attribute value that meets the actual geological conditions to ensure the spatial consistency of the assignment results.

[0081] Through the above interpolation method and trend map correction method, the spatial consistency of attribute values ​​between grid cells can be ensured. For example, in the spatial distribution of attributes such as cracks and water content, there are usually certain rules (such as the crack width gradually increases with the increase of depth, or the water content in some areas is high). The correction process uses interpolation, trend map constraints and other means to make the attribute value of each grid cell conform to this spatial variation law. In this way, the different variation trends of attributes such as cracks, resistivity inversion and water content in different strata can be effectively constrained and guided to assign grid units, and fracture attribute modeling based on deterministic fracture parameters can be realized.

[0082] In step S4, considering the use of deep learning technology, by constructing a deep neural network model, it is possible to automatically learn the complex relationship between data and perform accurate prediction and classification. For this reason, this embodiment introduces deep learning technology to make full use of the nonlinear relationship between multi-source attribute data, improve the accuracy and reliability of geological modeling, and achieve accurate prediction of address unit attributes.

[0083] Specifically, based on the geological unit body attribute model constructed above, a multi-source attribute probability fusion method based on a deep learning model is adopted to construct a dynamic fuzzy neural network and a multi-attribute probabilistic neural network. The nonlinear relationship between different attributes in the model is extracted by the dynamic fuzzy neural network. Then, a multi-attribute probabilistic neural network modeling method is adopted to update and aggregate the geological attributes of each grid node, and output the predicted geological attribute value of each grid unit in the geological unit body attribute model that changes with time and the excavation process. Specifically, the following steps are included:

[0084] Step S4.1, construct a dynamic fuzzy neural network (DFNN) to mine and integrate nonlinear relationships among multiple attributes in the modeling of the tunnel area geological unit body, so as to adaptively adjust the change value of the model attribute.

[0085] Since the three attributes of cracks, resistivity inversion water content, and seismic wave inversion fragmentation degree usually have strong spatial dependence and exist in multi-level nonlinear relationships, for example, there are complex interactive effects between the distribution of cracks and water content. Considering that in long-term projects such as tunnel construction, the changes in geological environment are very significant, this embodiment introduces a dynamic fuzzy neural network DFNN, which can effectively mine these nonlinear relationships through neural networks, so that DFNN can adaptively adjust the change values ​​of model attributes.

[0086] First, each grid node in the three-dimensional space is used as a data point to perform numerical interpolation of geological attributes. For missing values, DFNN fills the missing data by interpolating the attribute values ​​of adjacent three-dimensional coordinate points. The interpolated data set contains not only the value of each attribute, but also its corresponding spatial coordinates (x, y, z).

[0087] Secondly, the spatial coordinates and attribute values ​​of each data point obtained after interpolation are used as input data of the DFNN input layer and input into the DFNN network for training. The network will automatically capture the complex nonlinear relationship between multi-source attributes. Among them, for the attribute data set after interpolation, DFNN defines the correlation between attributes based on the repeatability of attribute values, that is, the interaction effects between multiple geological attributes such as cracks, resistivity inversion water content, and seismic wave inversion fragmentation. DFNN models these interaction effects through activation functions, and finally outputs the geological attribute prediction value of each grid unit considering other attribute values ​​through the network.

[0088] Among them, repeatability refers to the frequency of occurrence of certain attribute values ​​between different data points in the input data, that is, the frequency or distribution density of a certain attribute value (such as crack width, resistivity inversion water content, seismic wave inversion fragmentation, etc.) in the data set; the repeatability can be calculated by counting the frequency of occurrence of a certain attribute or attribute combination in the data set. After calculating the repeatability of the obtained attributes, the nonlinear activation function of the hidden layer of the DFNN is used to consider the mutual influence and superposition effect of the three attributes. In the hidden layer of the neural network, the relationship between the attributes is learned through nonlinear activation functions and multi-layer networks, and the input of each attribute is mapped to a nonlinear space, thereby simulating the interaction effect between the attributes, and finally clarifying the correlation between the attributes. Based on this correlation, the geological attribute value of each grid unit is predicted and output.

[0089] The specific structure of the above dynamic fuzzy neural network (DFNN) is:

[0090] (1) Input layer: The input data is a data set after numerical interpolation, including geological attribute values ​​such as fractures, water content from resistivity inversion, and fragmentation degree from seismic wave inversion, as well as the corresponding spatial coordinates (x, y, z). Different geological attribute values ​​are input into different nodes of the network respectively;

[0091] (2) Hidden layer: The hidden layer considers the mutual influence and superposition effect of the three attributes (cracks, water content from resistivity inversion, and fragmentation degree from seismic wave inversion) through a nonlinear activation function;

[0092] (3) Fuzzy membership function parameter adjustment mechanism: DFNN dynamically adjusts the parameters of the fuzzy membership function according to the characteristics of different attribute data to adapt to the processing requirements of different attribute data.

[0093] During the DFNN model training process, DFNN dynamically adjusts the degree of fuzzification and flexibly adjusts the parameters of the fuzzy membership function according to the characteristics of different attribute data. In addition, the DFNN training process is adaptive and usually does not require a fixed training set. Usually, a data set processed by numerical interpolation is used as the training set, which contains the numerical values ​​of attributes such as cracks, resistivity inversion water content, seismic wave inversion fragmentation degree, and their corresponding spatial coordinates (x, y, z). The model initializes the network parameters by selecting different initial data and optimizes them through a dynamic training process. Different attribute data such as cracks, resistivity inversion water content, seismic wave inversion fragmentation degree, etc. are input into different nodes of the network for initial training of the model.

[0094] Preferably, during the model training process, DFNN dynamically adjusts the network parameters and gradually optimizes the model's predictive ability according to the characteristics of the data. DFNN can handle complex nonlinear relationships in tunnel geology, and its application in tunnel geological modeling can improve the prediction accuracy of the project, especially under complex geological conditions. Through DFNN, it is possible to establish accurate nonlinear relationships between multiple attributes such as crack distribution, resistivity inversion water content, and seismic wave inversion fragmentation degree, providing reliable technical support for the design and construction of tunnel projects. As the training progresses, the network will automatically learn the relationship between the attributes, and gradually eliminate the noise or deviation in the data to improve the prediction accuracy.

[0095] Step S4.2, adopting the multi-attribute probabilistic neural network (MPNN) modeling method, by transmitting and updating information between the geological attribute data of grid nodes, the fusion and integration of multi-attribute data is realized, and the geological attribute prediction value of each grid unit in the geological unit body attribute model is output.

[0096] The above DFNN can dig out the complex nonlinear relationship between different attributes, which can provide the basis for the internal connection between attributes in the process of node information transmission and fusion of MPNN. For example, the specific nonlinear relationship between cracks and water content analyzed by DFNN enables MPNN to more reasonably consider the spatial distribution and mutual influence of these two attributes when processing node features. On the basis of obtaining this nonlinear relationship, MPNN further integrates and synthesizes multi-attribute data. By defining the graph structure (nodes represent grid units, and edges represent spatial relationships), the initial feature vector of each node (including cracks, resistivity inversion water content values, and seismic wave inversion fragmentation values) is fused between nodes through message transmission and update, so that each node can integrate all information from the neighborhood. The final aggregated node feature vector can be used for subsequent prediction tasks (such as grouting probability prediction).

[0097] Specifically, to construct the MPNN model, the graph structure is first defined. The graph structure consists of nodes and edges. Each node represents a grid unit in the tunnel geological model. The grid units are connected by edges to represent the spatial relationship between nodes. Each node has an initial feature vector, which includes the values ​​of cracks, resistivity inversion water content and seismic wave inversion fragmentation. For each grid node, the corresponding values ​​of cracks, resistivity inversion water content and fragmentation are extracted, and these values ​​are input into the neural network as the initial features of the node. A message passing mechanism is set up in the MPNN. Through several rounds of message passing, nodes exchange information and weights and update their own feature vectors. Each node exchanges information with adjacent nodes, passes the feature vectors of adjacent nodes, and combines its own features to update its own feature vector through aggregation operations. After several rounds of message passing, each node in the network integrates all information from its neighborhood. After the last round of message passing, the feature vector of each node is aggregated to obtain the aggregated node feature vector, which can be used for subsequent prediction tasks. For example, the final node feature representation can be used as a prediction basis for grouting probability.

[0098] Preferably, after preliminary data processing, the attribute values ​​of cracks, water content from resistivity inversion, and fragmentation degree from seismic wave inversion related to each grid unit are extracted from the geological exploration data to form a training data set for MPNN, and the training data set is used to train the MPNN network. By fusing and integrating multi-source attribute data through MPNN, the node feature vectors are continuously updated and aggregated, which can optimize the actual characteristics of the geological body and output the predicted geological attribute values ​​of each grid unit in the geological unit body attribute model that change with time and the excavation process.

[0099] Embodiment 2

[0100] This embodiment provides a geological attribute modeling system based on multi-source information joint constraints, specifically including:

[0101] The data acquisition and preprocessing module is used to acquire geological exploration data, obtain geological attribute data after preprocessing, and classify and quantify the geological attribute data into continuous and discontinuous types;

[0102] The geological attribute analysis module is used to perform multi-source geological attribute correlation analysis based on the quantified geological attribute data to obtain the correlation between different geological attributes;

[0103] The geological attribute modeling module is used to construct a geological structure model of the area to be modeled. It divides the model into unstructured networks based on the correlation between different geological attributes. For each grid area, according to the different geological attribute data obtained, the geological unit grid attribute is assigned based on the 1D+2D trend chart constraints to construct a geological unit attribute model.

[0104] Further technical solutions include:

[0105] The geological attribute prediction module is used to extract the nonlinear relationship between different attributes in the model based on the geological unit body attribute model through dynamic fuzzy neural network, and then adopts the multi-attribute probabilistic neural network modeling method to update and aggregate the geological attributes of each grid node, so as to output the predicted value of the geological attributes of each grid unit in the geological unit body attribute model that changes with time and the excavation process.

[0106] Embodiment 3

[0107] This embodiment provides an electronic device, including: a memory, used to store executable instructions; and a processor, used to implement the above method provided in this embodiment when executing the executable instructions stored in the memory.

[0108] Embodiment 4

[0109] This embodiment also provides a computer-readable storage medium storing executable instructions. When the executable instructions are executed by a processor, the processor will be caused to execute the above method provided in this embodiment.

[0110] Embodiment 5

[0111] This embodiment provides a computer program product, which includes an executable instruction, which is a computer instruction; the executable instruction is stored in a computer-readable storage medium. When a processor of an electronic device reads the executable instruction from the computer-readable storage medium and the processor executes the executable instruction, the electronic device executes the above method provided in this embodiment.

[0112] The steps involved in the above embodiments 2 to 5 correspond to the method embodiment 1. For the specific implementation, please refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0113] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0114] The above description is only a preferred embodiment of the present invention. Although the specific implementation mode of the present invention is described in conjunction with the accompanying drawings, it is not a limitation of the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the protection scope of the present invention.

Claims

1. A geological attribute modeling method based on joint constraints of multi-source information, characterized in that: include: Obtain geological exploration data, obtain geological attribute data after preprocessing, and classify and quantify the geological attribute data into continuous and discontinuous types; Based on the quantified geological attribute data, multi-source geological attribute correlation analysis is performed to obtain the correlation between different geological attributes; Construct a geological structure model of the area to be modeled, divide the model into unstructured networks based on the correlation between different geological attributes, and for each grid area, assign the geological unit grid attributes based on 1D+2D trend chart constraints according to the different geological attribute data obtained to construct a geological unit attribute model.

2. A geological attribute modeling method based on multi-source information joint constraints as claimed in claim 1, characterized in that: Also includes: Based on the geological unit attribute model, the nonlinear relationship between different attributes in the model is extracted through dynamic fuzzy neural network, and then the multi-attribute probabilistic neural network modeling method is used to update and aggregate the geological attributes of each grid node, so as to output the predicted value of the geological attributes of each grid unit in the geological unit attribute model that changes with time and the excavation process.

3. A geological attribute modeling method based on multi-source information joint constraints as claimed in claim 1, characterized in that: The geological attributes include fractures, water content based on resistivity inversion, and degree of fragmentation based on seismic wave inversion; Among them, the width of the crack and the water content based on resistivity inversion are continuous geological attributes, and the size of the crack and the degree of fragmentation based on seismic wave propagation velocity inversion are non-continuous geological attributes; the continuous geological attribute data are standardized or normalized, and the non-continuous geological attribute data are coded.

4. A geological attribute modeling method based on multi-source information joint constraints as claimed in claim 1, characterized in that: When constructing the geological structure model, the inversion constraint method is used to make the spatial boundary of the geological unit body reflect the actual geological structure, and the surface constraint conditions of the geological unit body are determined through exploration data and seismic inversion results. On the basis of the surface constraint conditions, according to the distribution of fractures or tunnel faces, several small units are obtained through unstructured grid division; The determination of the surface constraint condition includes: According to the survey data, the spatial distribution information of the boundary of the geological unit body is determined, including: the distribution and depth of the underground rock layer can be identified by the location and depth of the drilling hole, and the surface boundary function of the geological body can be determined, which is expressed as: f(x,y,z)=0; In the above formula, f(x,y,z) is a function that describes the shape of the boundary surface of the geological unit, and (x,y,z) represents the spatial coordinates; The velocity change of seismic waves is used to reflect the boundaries of different geological units, thereby determining the constraint surface, which is expressed as: v1(x,y,z)≠v2(x,y,z); In the above formula, v1 and v2 represent the wave velocities of geological body 1 and geological body 2 respectively; Determine the surface constraints of the geological unit volume, including: In the above formula, f i (x, y, z) represents the surface boundary function of the i-th geological unit obtained by inversion, g i The actual measurement data.

5. A geological attribute modeling method based on multi-source information joint constraints as claimed in claim 4, characterized in that: The construction of the geological unit attribute model includes: Coarsening of geological attribute data; that is, converting the acquired geological attribute data into a low-resolution form by spatially downscaling the data; Grid mapping of geological attribute data; that is, using refined grid subdivision technology, the grid size is selected according to the actual needs of the region, the region is divided into multiple grid cells, each grid cell represents a certain spatial range, and each grid cell is assigned a geological attribute value; For the geological attribute values ​​assigned to each grid unit, a 1D time variation trend graph of the geological attribute is generated to show the law of geological attribute variation over time. The generated 1D trend graph is used as a constraint to correct the grid unit assignment result. At the same time, a 2D spatial distribution trend graph of the geological attribute is generated on a cross section or other two-dimensional plane by using a two-dimensional interpolation method to show the law of geological attribute variation with spatial position. The generated 2D trend graph is used as a constraint to correct the grid unit assignment result. On the basis of the 2D trend chart, the depth from the surface or tunnel entrance is introduced to form a 3D trend chart, which is used to show the changing trends of different geological attributes at different depths under vertical changes; based on the data of three dimensions in space, a complete 3D spatial distribution trend chart is generated through spatial interpolation to adjust the attribute values ​​of the grid units.

6. A geological attribute modeling system based on multi-source information joint constraints, characterized in that: include: The data acquisition and preprocessing module is used to acquire the geological exploration data of the tunnel area to be modeled, and obtain continuous attribute data and discontinuous attribute data after quantization processing; The geological attribute analysis module is used to obtain the correlation between different geological attributes through multivariate attribute correlation analysis based on the quantified geological attribute data; The geological attribute modeling module is used to construct the geological structure model of the area to be modeled and divide the cell network. According to the different geological attribute data obtained, the cell grid attribute is assigned based on the 1D+2D trend chart constraints to preliminarily build the geological attribute model. The geological attribute model optimization module is used to extract the nonlinear relationship between different attributes in the model based on the geological attribute model through dynamic fuzzy neural network, and then adopt the multi-attribute probabilistic neural network modeling method to update and aggregate the geological characteristics of each grid node to obtain the final geological attribute model.

7. A geological attribute modeling system based on multi-source information joint constraints as claimed in claim 6, characterized in that: Also includes: The geological attribute prediction module is used to extract the nonlinear relationship between different attributes in the model based on the geological unit body attribute model through dynamic fuzzy neural network, and then adopts the multi-attribute probabilistic neural network modeling method to update and aggregate the geological attributes of each grid node, so as to output the predicted value of the geological attributes of each grid unit in the geological unit body attribute model that changes with time and the excavation process.

8. An electronic device, characterized in that: include: A memory for storing executable instructions; The processor is used to implement the geological attribute modeling method based on multi-source information joint constraints as described in any one of claims 1 to 5 when executing the executable instructions stored in the memory.

9. A computer-readable storage medium, characterized in that: Executable instructions are stored, which are used to cause the processor to execute the executable instructions to implement the geological attribute modeling method based on multi-source information joint constraints as described in any one of claims 1-5.

10. A computer program product, characterized in that The computer program product includes executable instructions stored in a computer-readable storage medium; When the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the geological attribute modeling method based on multi-source information joint constraints described in any one of claims 1 to 5 is implemented.

Images