An integrated analysis method for geological engineering and surveying engineering data

Abstract

The present invention provides an integrated analysis method for geological engineering and surveying and mapping engineering data, belonging to the field of geological exploration technology. The present invention realizes multi-level decomposition and expression of geological bodies by constructing a multi-scale geological spatial data model, unifies the reference system of heterogeneous data by adopting a spatiotemporal registration algorithm, and reduces the computational complexity to O(ElogV) by optimizing dimensionality reduction processing in combination with a minimum spanning tree algorithm. A probabilistic association between geological and surveying data is established based on a probabilistic graph model, and an adaptive weight fusion algorithm is applied to integrate key information to form a fusion feature set, which is input into a geological surveying and mapping dual-domain model for in-depth analysis. A physical error correction module is provided to ensure the rationality of the results, and a computational precision balance function is introduced to dynamically adjust model parameters, thereby achieving an optimal balance between ensuring analysis quality and computational efficiency, and effectively solving the problems of computing resource consumption and real-time performance in the processing of massive geological surveying and mapping data.

Description

Technical Field

[0001] The present invention belongs to the field of geological exploration technology, and in particular relates to an integrated analysis method for geological engineering and surveying and mapping engineering data. Background Art

[0002] The fusion and analysis of geological engineering and surveying and mapping data has significant application value in fields such as mineral resource exploration, large-scale engineering construction, natural disaster monitoring, and urban planning. Traditional geological engineering relies on drilling and geophysical surveying to collect structural characteristics and physical properties of underground geological bodies, while surveying and mapping uses advanced technologies such as InSAR, GNSS, and 3D laser scanning to obtain surface deformation and displacement monitoring data. These technologies are relatively mature in their respective fields. For example, geological analysis uses 3D modeling and finite element analysis, while surveying and mapping utilizes differential interferometry and time series analysis to monitor surface deformation.

[0003] However, with advances in data acquisition technology, the fields of geology and surveying and mapping are facing the challenge of processing massive amounts of heterogeneous, multi-source data. Traditional fusion analysis methods, such as simple overlay and regression analysis, are inefficient when processing high-dimensional, multi-scale geological surveying and mapping big data. This is particularly true in large-scale engineering scenarios, where geological drilling sites are densely located and surveying and monitoring data is highly continuous. The exponential growth in data volume leads to significant computational resource consumption. Existing technologies typically use methods such as downsampling or regional partitioning to reduce computational complexity, but these methods often sacrifice data integrity, affecting the accuracy of analytical results.

[0004] Therefore, how to effectively process massive amounts of geological mapping data, reduce computing resource consumption, and achieve efficient or even real-time analysis and processing while ensuring the quality of fusion analysis has become a core issue that needs to be addressed in current technologies. This issue is particularly prominent in time-sensitive applications such as real-time monitoring and early warning during construction. Summary of the Invention

[0005] In view of this, the present invention provides an integrated analysis method for geological engineering and surveying and mapping engineering data, which can solve the problem in the existing technology that the amount of data in the fusion analysis of geological engineering and surveying and mapping engineering data is large, resulting in excessive consumption of computing resources and difficulty in real-time processing.

[0006] The present invention is implemented as follows: The present invention provides an integrated analysis method for geological engineering and surveying and mapping engineering data, including: constructing a multi-scale geological spatial data model and performing multi-level decomposition on geological bodies; using a spatiotemporal registration algorithm to standardize surveying and mapping data; using a minimum spanning tree algorithm to optimize and reduce the dimensionality of multi-source heterogeneous data, constructing a geological surveying and mapping data feature connection network, extracting key nodes and paths, and generating a dimensionality reduction parameter set; establishing association rules between geological body attributes and surface deformation parameters based on a probabilistic graph model; applying an adaptive weight fusion algorithm to integrate geological internal structure information and surface monitoring results to form a geological surveying and mapping fusion feature set; inputting the geological surveying and mapping fusion feature set into a geological surveying and mapping dual-domain model, dynamically adjusting model parameters in combination with a computational accuracy balance function, performing iterative optimization until convergence, and outputting the integrated analysis results of geological engineering and surveying and mapping engineering.

[0007] Among them, the multi-scale geological spatial data model extracts features of geological bodies at different scales through wavelet transform, constructs a multi-level mathematical expression including geological structure, physical parameters and spatial distribution characteristics, and realizes a complete description of geological bodies from regional to local.

[0008] Among them, the spatiotemporal registration algorithm is a data alignment method based on control point matching and non-rigid transformation. It establishes a spatial correspondence between data by identifying common feature points in geological data and surveying and mapping data, and realizes the unified expression of data collected at different times through time series analysis.

[0009] Among them, the minimum spanning tree algorithm is a graph theory optimization method that regards the characteristic points of geological surveying data as nodes of the graph and the characteristic similarity as the edge weight. By constructing a subgraph connecting all nodes with the minimum total weight, the key structure of the data is extracted. The complexity of the minimum spanning tree algorithm is O(ElogV), which reduces the computational complexity of multi-source heterogeneous data processing while retaining the core topological relationship between the data.

[0010] Among them, the dimensionality reduction parameter set includes the feature dimension reduction ratio, node connection threshold, graph structure sparsity, and data compression rate, which are used to quantify the data loss and information retention degree in the dimensionality reduction process. The dimensionality reduction parameter set serves as the input parameter of the computational accuracy balance function.

[0011] Among them, the probabilistic graphical model is a statistical model that uses graph theory to represent the dependency relationship between random variables. It uses nodes to represent geological and surveying parameters, and edges to represent the correlation between parameters. It establishes the conditional probability distribution between parameters and realizes the uncertainty reasoning of geological conditions and surface responses.

[0012] Among them, the adaptive weight fusion algorithm dynamically adjusts the weight coefficients of various types of data in the fusion process according to data quality, reliability and relevance, ensuring that high-quality data has a greater contribution to the final analysis results and improving analysis accuracy.

[0013] Among them, the geological surveying and mapping fusion feature set is a set of geological internal structure information and surface monitoring results integrated through an adaptive weight fusion algorithm. It includes the lithology distribution, structural characteristics, physical and mechanical parameters of the geological body, and surface deformation, displacement gradient, deformation rate and other data indicators in the surveying and mapping data that describe the core characteristics of geological engineering and surveying and mapping engineering.

[0014] Among them, the specific structure of the geological surveying and mapping dual-domain model is an encoder-decoder network based on the Transformer architecture, which includes two parallel branches: the geological domain encoder and the surveying and mapping domain encoder. Each branch is composed of a multi-layer self-attention module, and feature interaction is achieved through the cross-domain attention mechanism. The decoder part uses a multi-head attention mechanism to fuse dual-domain features, and the output layer contains a physical error correction module as a submodule.

[0015] Among them, the geological surveying and mapping dual-domain model uses adaptive gating units to screen features. The total parameter quantity of the geological surveying and mapping dual-domain model is dynamically adjusted according to the complexity of geological data and the density of surveying and mapping data. The number of self-attention heads is proportional to the complexity of geological body structure, and the attention dimension is proportional to the accuracy level of surveying and mapping data.

[0016] The physical error correction module is a submodule of the output layer of the dual-domain model of geological surveying and mapping. It is a constraint mechanism constructed based on the principles of geomechanics and the basic laws of surveying. It corrects data that do not conform to physical laws by detecting whether the analysis results violate physical constraints. The correction strength of the physical error correction module is determined by the output value of the calculation accuracy balance function.

[0017] The inputs of the computational accuracy balance function include the intensity of multi-source data variation features, the confidence of the neural network output, the dimensionality reduction parameter set, and the computing resource constraint index. By comprehensively evaluating the computing needs and accuracy requirements, the balance value is output. When the balance value is less than 0.3, the number of self-attention heads and the correction strength of the physical error correction module are increased to improve the accuracy. When the balance value is greater than 0.7, the number of self-attention layers and the correction strength of the physical error correction module are reduced to reduce the computational complexity.

[0018] When the balance value is between 0.3 and 0.7, the self-attention parameters and the correction strength of the physical error correction module are kept unchanged, achieving a dynamic balance between system resources and analysis accuracy.

[0019] The steps for establishing the training data set for the dual-domain model of geological surveying and mapping include collecting historical data of geological exploration and mapping monitoring in multiple regions, classifying and organizing them according to geological structure types and mapping data types, using the sliding window method to split the time series data, constructing spatiotemporal registration annotations, using manual annotation and semi-supervised methods to establish the correspondence between geological anomalies and surface deformation, and using data enhancement technology to expand the training samples.

[0020] The steps of training the dual-domain model for mass surveying and mapping include first performing self-supervised pre-training on a large-scale geological dataset to learn geological feature representation, then performing self-supervised pre-training on surveying and mapping data to learn surveying and mapping feature representation, and then performing dual-domain joint fine-tuning to optimize the cross-domain attention mechanism and physical error correction module parameters. The entire training process adopts a dynamic learning rate strategy and automatically adjusts the optimizer parameters according to the performance of the validation set.

[0021] The data enhancement techniques include random rotation, scaling, and noise addition operations, which ultimately form a large-scale training set containing surveying and mapping response characteristics under standard geological scenarios.

[0022] The present invention achieves efficient processing of massive heterogeneous data by constructing a multi-scale geological spatial data model and minimum spanning tree data dimensionality reduction processing technology. This method uses a spatiotemporal registration algorithm to unify the data reference system, uses a probabilistic graphical model to establish a quantitative association between geological state and surface response, combines an adaptive weight fusion algorithm to integrate key information, and conducts in-depth analysis and prediction through a geological mapping dual-domain model. This method effectively solves the problem of low efficiency in massive data processing in traditional technologies, and reduces the computational complexity from O( ) is reduced to O(ElogV), which significantly reduces the consumption of computing resources. At the same time, through the optimized configuration of the dimensionality reduction parameter set, the key topological structure and characteristic information are retained while compressing the data volume, ensuring the accuracy of the analysis results. In particular, the introduction of the computational accuracy balance function can dynamically adjust the model parameters according to the application scenario requirements and computing resource constraints, and realize flexible switching from high-precision analysis to real-time processing. Therefore, the present invention solves the technical problem that the large amount of data in the fusion analysis of geological engineering and surveying and mapping engineering data leads to excessive consumption of computing resources and difficulty in real-time processing, making real-time monitoring, early warning and decision-making possible in complex geological engineering environments, and greatly improving the efficiency of geological engineering safety management and risk control. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the method of the present invention.

[0024] Figure 2 This is a diagram showing the effect of the spatiotemporal registration processing in Example 2.

[0025] Figure 3 A comparison chart of key information retained for the dimensionality reduction process in Example 2.

[0026] Figure 4 This is the effect diagram of the adaptive weight fusion in Example 2. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0028] like Figure 1 FIG. 1 is a flow chart of a method for integrated analysis of geological engineering and surveying and mapping engineering data provided by the present invention. The method comprises the following steps:

[0029] S01. Construct a multi-scale geological spatial data model, perform multi-level decomposition of geological bodies, realize the expression of geological information from macro to micro, and establish a mathematical description of the relationship between geological characteristic attributes and distribution;

[0030] S02. Standardize the surveying and mapping data using a spatiotemporal registration algorithm, determine a unified spatial coordinate system and time reference, and establish a positional correspondence between the surveying and mapping data and the geological data;

[0031] S03. Use the minimum spanning tree algorithm to optimize and reduce the dimensionality of multi-source heterogeneous data, build a feature connection network of geological surveying and mapping data, extract key nodes and paths, generate a dimensionality reduction parameter set, and achieve data compression and key information retention;

[0032] S04. Establish association rules between geological body attributes and surface deformation parameters based on the probabilistic graphical model to achieve quantitative expression of geological state and surface response;

[0033] S05. Apply the adaptive weight fusion algorithm to integrate geological internal structure information and surface monitoring results to form a geological mapping fusion feature set, laying the foundation for subsequent analysis;

[0034] S06. Input the geological surveying and mapping fusion feature set into the geological surveying and mapping dual-domain model, dynamically adjust the model parameters in combination with the calculation accuracy balance function, perform iterative optimization until convergence, and output the final geological engineering and surveying engineering integrated analysis results.

[0035] Among them, the multi-scale geological spatial data model extracts features of geological bodies at different scales through wavelet transform, constructs a multi-level mathematical expression including geological structure, physical parameters and their spatial distribution characteristics, and realizes a complete description of geological bodies from regional to local.

[0036] Among them, the spatiotemporal registration algorithm is a data alignment method based on control point matching and non-rigid transformation. It establishes a spatial correspondence between data by identifying common feature points in geological data and surveying and mapping data, and realizes the unified expression of data collected at different times through time series analysis.

[0037] Among them, the minimum spanning tree algorithm is a graph theory optimization method. It regards the characteristic points of geological surveying data as nodes of the graph, and the characteristic similarity as the edge weight. By constructing a subgraph connecting all nodes with the minimum total weight, the key structure of the data is extracted. The complexity of the minimum spanning tree algorithm is O(ElogV), which reduces the computational complexity of multi-source heterogeneous data processing while retaining the core topological relationship between the data.

[0038] Among them, the dimensionality reduction parameter set includes the feature dimension reduction ratio, node connection threshold, graph structure sparsity, and data compression rate, which are used to quantify the data loss and information retention degree in the dimensionality reduction process. The dimensionality reduction parameter set serves as the input parameter of the computational accuracy balance function.

[0039] Among them, the probabilistic graphical model is a statistical model that uses graph theory to represent the dependency relationship between random variables. It uses nodes to represent geological and surveying parameters, and edges to represent the correlation between parameters. It establishes the conditional probability distribution between parameters and realizes the uncertainty reasoning of geological conditions and surface responses.

[0040] Among them, the adaptive weight fusion algorithm dynamically adjusts the weight coefficients of various types of data in the fusion process according to data quality, reliability and relevance, ensuring that high-quality data has a greater contribution to the final analysis results and improving analysis accuracy.

[0041] Among them, the geological surveying and mapping fusion feature set is a set of geological internal structure information and surface monitoring results integrated through the adaptive weight fusion algorithm, including the lithology distribution, structural characteristics, physical and mechanical parameters of the geological body, and surface deformation, displacement gradient, deformation rate and other data indicators in the surveying and mapping data that describe the core characteristics of geological engineering and surveying and mapping engineering.

[0042] Among them, the specific structure of the geological surveying and mapping dual-domain model is an encoder-decoder network based on the Transformer architecture, which includes two parallel branches: a geological domain encoder and a surveying and mapping domain encoder. Each branch is composed of a multi-layer self-attention module, and feature interaction is achieved through a cross-domain attention mechanism. The decoder part uses a multi-head attention mechanism to fuse dual-domain features. The output layer includes a physical error correction module as a submodule, and uses an adaptive gating unit to filter features. The total parameter quantity of the geological surveying and mapping dual-domain model is dynamically adjusted according to the complexity of geological data and the density of surveying and mapping data. The number of self-attention heads is proportional to the complexity of the geological body structure, and the attention dimension is proportional to the accuracy level of the surveying and mapping data.

[0043] Among them, the physical error correction module is a sub-module of the output layer of the geological surveying and mapping dual-domain model. It is a constraint mechanism constructed based on the principles of geomechanics and the basic laws of surveying. It corrects data that does not conform to physical laws by detecting whether the analysis results violate physical constraints. The correction strength of the physical error correction module is determined by the output value of the calculation amount accuracy balance function. Strict correction is adopted in high-precision requirements scenarios, and loose correction is adopted in high-efficiency requirements scenarios.

[0044] Among them, the input of the computational accuracy balance function includes the intensity of multi-source data change characteristics, the confidence of the neural network output, the dimensionality reduction parameter set, and the computing resource constraint index. By comprehensively evaluating the computing requirements and accuracy requirements, the balance value is output. When the balance value is less than 0.3, the number of self-attention heads and the correction strength of the physical error correction module are increased to improve the accuracy. When the balance value is greater than 0.7, the number of self-attention layers and the correction strength of the physical error correction module are reduced to reduce the computational amount. When the balance value is between 0.3 and 0.7, the self-attention parameters and the correction strength of the physical error correction module are kept unchanged to achieve a dynamic balance between system resources and analysis accuracy.

[0045] Among them, the steps of establishing the training data set of the geological surveying and mapping dual-domain model specifically include collecting historical data of geological exploration and mapping monitoring in multiple regions, classifying and organizing them according to geological structure types and mapping data types, using the sliding window method to split time series data, constructing spatiotemporal registration annotations, using manual annotation and semi-supervised methods to establish the correspondence between geological anomalies and surface deformations, and using data enhancement technology to expand training samples, including random rotation, scaling, and noise addition operations, to finally form a large-scale training set containing mapping response characteristics under standard geological scenarios.

[0046] Among them, the steps of training the geological surveying and mapping dual-domain model specifically include first performing self-supervised pre-training on a large-scale geological dataset to learn geological feature representation, then performing self-supervised pre-training on surveying and mapping data to learn surveying and mapping feature representation, and then performing dual-domain joint fine-tuning to optimize the cross-domain attention mechanism and the physical error correction module parameters. The entire training process adopts a dynamic learning rate strategy and automatically adjusts the optimizer parameters according to the performance of the validation set.

[0047] The specific implementation of the above steps is described in detail below.

[0048] The specific implementation of step S01 is to perform multi-scale decomposition and feature extraction on the geological body through wavelet transform theory. First, a wavelet basis function suitable for the characteristics of the geological body is selected. Commonly used wavelet bases include Haar wavelet, Daubechies wavelet or Meyer wavelet. The selection is determined according to the complexity of the geological body structure. Generally, when the geological structure complexity is high, Daubechies wavelet is more suitable. Then, the three-dimensional data of the geological body is decomposed by wavelet according to different resolutions. Usually 4 to 5 decomposition levels are set. The first level represents macro-regional features, and the last level represents micro-local features. ; Then, the key features of the decomposition results at each level are extracted, including geological structure boundaries, physical parameter mutation points, and spatial distribution trends; then a multi-level feature matrix is constructed, in which the matrix elements contain spatial position coordinates, geological physical parameters and their change rates, and the matrix dimensions are linearly corresponding to the decomposition levels; finally, a mathematical expression of the spatial distribution of geological characteristics is established, and radial basis functions are used to continuously interpolate discrete feature points. When the Gaussian kernel is selected as the interpolation kernel function, the smoothing parameter threshold is usually set between 0.2 and 0.5. A smaller threshold is suitable for geological bodies with strong mutation, and a larger threshold is suitable for geological bodies with strong gradual change. This step realizes the complete feature expression of geological bodies from macroscopic regions to microscopic local areas, laying a data foundation for subsequent analysis.

[0049] The specific implementation of step S02 is to standardize the surveying and mapping data using a spatiotemporal registration algorithm based on control point matching and non-rigid transformation. First, common feature points in the geological data and the surveying and mapping data are identified, including obvious surface markers, geological structure outcrops, etc., as control points. The number of effective control points is not less than 3% of the total data points and not less than 12 points. Then, a unified spatial coordinate system is established, usually using the National Geodetic Coordinate System 2000 or WGS84 coordinate system, and all data are converted to this coordinate system through coordinate transformation. Then, a unified time base is determined, and one of them is selected. All time series data are normalized relative to a reference point in time. Nonrigid transformations are then applied to the surveying and mapping data for spatial registration. Thin-plate spline interpolation is used for smooth transformations when there are many control points, while affine transformations are used for basic registration when there are fewer control points. Registration accuracy is assessed using the root mean square error (RMS), with the threshold typically controlled within 1.5 times the measurement accuracy. Finally, data collected at different times are temporally registered using time series analysis methods. Linear or spline interpolation is used to uniformly represent data from different sampling periods, with the temporal resolution typically consistent with the minimum monitoring frequency. This step establishes a precise spatiotemporal correspondence between the surveying and mapping data and the geological data, ensuring consistency of data from different sources in subsequent analyses.

[0050] The specific implementation of step S03 is to use the minimum spanning tree algorithm to optimize the dimensionality reduction of multi-source heterogeneous data. First, each feature point of the geological mapping data is regarded as a node of the graph. The total number of nodes is usually 20% to 40% of the original data points. The sampling method is based on the feature significance score. Then, the feature similarity between nodes is calculated as the edge weight. The similarity calculation uses cosine distance or Mahalanobis distance, and the weight threshold is set between 0.3 and 0.7. Then, the Kruskal or Prim algorithm is applied to construct a minimum spanning tree, and the subgraph connecting all nodes with the smallest total weight is selected. The complexity of this algorithm is ,in represents the number of edges, The method then extracts key nodes and paths from the minimum spanning tree. Key nodes are defined as nodes with a connectivity greater than 1.5 times the average connectivity, and key paths are defined as paths connecting two key nodes with a weight less than 0.8 times the average weight. Finally, a set of dimensionality reduction parameters is generated, including a feature dimension reduction ratio set to 0.4 to 0.6, a node connection threshold set to 0.35 to 0.65, a graph structure sparsity control range of 0.1 to 0.3, and a data compression rate set between 60% and 85% depending on the application scenario. This step achieves effective dimensionality reduction and key information extraction from multi-source heterogeneous data, significantly reducing the computational complexity of subsequent processing while preserving core topological relationships.

[0051] The specific implementation of step S04 is to establish the association rules between geological body attributes and surface deformation parameters based on the probability graph model. First, the node variables of the probability graph are determined, including internal attributes such as the lithology category, structural characteristics, physical and mechanical parameters of the geological body, as well as surveying and monitoring parameters such as surface deformation, displacement gradient, and deformation rate; then, the conditional dependency relationship between nodes is constructed, and the probabilistic connection between variables is established using Bayesian network or Markov random field, and the dependency strength threshold is set to 0.25; then, the conditional probability distribution is learned, and the conditional probability table is used for discrete variables and the Gaussian mixture model is used for continuous variables. The expectation maximization algorithm is used for parameter estimation, and the iterative convergence threshold is set to ; Then the causal relationship between variables is inferred, and the causal direction between variables is determined by conditional independence test, with the significance level set at 0.05; finally, a quantitative expression of geological state and surface response is established in the form of conditional probability , indicating that in a given geological state Surface response under conditions This step achieves a quantitative description of the association rules between the internal attributes of the geological body and the surface deformation parameters, providing a probabilistic model support for geological state assessment and prediction.

[0052] The specific implementation of step S05 is to integrate geological internal structure information with surface monitoring results using an adaptive weighted fusion algorithm. First, the quality indicators of different data sources are evaluated, including data integrity, measurement accuracy, and spatiotemporal resolution, and quantified into quality scores ranging from 0 to 1. Then, the correlation between each data source and the target analysis task is calculated, quantified using mutual information or Pearson correlation coefficient, with a threshold typically set above 0.4. Next, the reliability of each data source is evaluated, including factors such as equipment accuracy rating and the degree of measurement environment impact, and quantified into reliability coefficients ranging from 0.6 to 1. Subsequently, initial weights are calculated based on the quality score, correlation, and reliability. The weight coefficients are dynamically adjusted during the iterative process based on the data fusion effect, with an adjustment step of 0.05 to 0.1. Finally, a geological mapping fusion feature set is generated, which includes core indicators such as lithologic distribution, structural characteristics, physical and mechanical parameters of geological bodies, and surface deformation, displacement gradient, and deformation rate in the mapping data. This step achieves adaptive fusion of multi-source data, ensuring that high-quality and highly relevant data contributes more to the final analysis, improving the representativeness and reliability of the fused feature set.

[0053] The specific implementation of step S06 is to input the geological surveying and mapping fusion feature set into the geological surveying and mapping dual-domain model for integrated analysis. First, the fusion feature set is divided into two parts: geological domain data and surveying and mapping domain data; then the geological domain data is input into the geological domain encoder, and the surveying and mapping domain data is input into the surveying and mapping domain encoder. Both encoders are composed of multi-layer self-attention modules. The number of attention heads is automatically set according to the complexity of the geological body structure, and the range is usually 4 to 12; then the feature interaction between the two domains is realized through the cross-domain attention mechanism, and the correlation pattern between the domains is learned; then the multi-head attention mechanism is applied to perform feature fusion, and the results are analyzed. The decoder generates preliminary analysis results. The results are then corrected using a physical error correction module. This module constructs constraints based on the principles of geomechanics and geodesy to correct data that does not conform to physical laws. The correction strength is determined by the output of the computational precision balance function. Finally, the model parameters are dynamically adjusted based on the output of the computational precision balance function. When the balance value is less than 0.3, the number of attention heads and the strength of physical error correction are increased to improve accuracy. When the balance value is greater than 0.7, the number of attention layers and the strength of physical error correction are reduced to reduce the computational effort. When the balance value is between 0.3 and 0.7, the model parameters remain unchanged. This step achieves a deep fusion analysis of geological engineering and surveying and mapping engineering data through a dual-domain model, outputting physically meaningful integrated analysis results while balancing computational efficiency and analysis accuracy.

[0054] The detailed structure of the geological surveying and mapping dual-domain model is based on the Transformer architecture and consists of two parallel branches: a geological domain encoder and a surveying and mapping domain encoder. The geological domain encoder processes geological internal structural information and consists of six layers of self-attention modules. Each layer consists of a multi-head self-attention sublayer and a feedforward network sublayer. Residual connections are used between layers, and LayerNorm is used for normalization. The number of self-attention heads is dynamically adjusted based on the complexity of the geological volume: 12 heads are used for complexity scores above 0.8, 8 heads are used for complexity scores between 0.4 and 0.8, and 4 heads are used for complexity scores below 0.4. The dimension of each attention head is set to 64, and the dimension of the feedforward network hidden layer is set to 2048. The surveying and mapping domain encoder processes surface monitoring data. Its structure is similar to that of the geological domain encoder, but the number of layers is typically set to 4. The number of attention heads is related to the density of the surveying and mapping data: 10 heads are used for data density scores above 0.7, 6 heads are used for scores between 0.3 and 0.7, and 3 heads are used for scores below 0.3. The cross-domain attention mechanism is located between the two encoders and consists of a bidirectional cross-attention layer to achieve inter-domain information interaction. The weight matrix is initialized using the Xavier method, and the bias term is initialized to a zero vector. The decoder part consists of a 4-layer multi-head attention module. Each layer contains a self-attention sublayer, an encoder attention sublayer, and a feedforward network sublayer. After fusing the dual-domain features, the analysis results are output through a fully connected layer. The output layer contains a physical error correction module as a submodule. This module constructs a regularization term based on geomechanical constraints and filters features through adaptive gating units. The gating threshold is dynamically adjusted according to the output of the computational accuracy balance function to correct prediction results that do not conform to physical laws. The total number of parameters of the entire model is dynamically adjusted. When the geometric mean of the geological data complexity score and the mapping data density score is greater than 0.6, the parameter amount is Magnitude; when the average value is between 0.3 and 0.6, the parameter value is Magnitude; when the average value is less than 0.3, the parameter value is The number of self-attention heads is proportional to the complexity of the geological structure. For every 0.1 increase in complexity, the number of heads increases by 1. The attention dimension is proportional to the accuracy level of the surveying and mapping data. For every level increase in the accuracy level, the dimension increases by 16.

[0055] The detailed steps for establishing the training data set of the geological surveying and mapping dual-domain model are to first collect historical data of geological exploration and mapping monitoring in multiple regions, covering different geological structure types, including fold structures, fault structures, igneous rock bodies, etc., as well as different mapping data types, including leveling, GNSS measurement, InSAR monitoring, etc.; then classify and organize them according to the geological structure type and mapping data type, and build a hierarchical database structure, with each type of data containing at least 50 typical cases; then use the sliding window method to segment the time series data, and the window length is set according to the characteristic time scale of the geological process. The window for rapid change processes is set to 7 to 15 days, and the window for slow change processes is set to 30 to 90 days. The window sliding step is 25% to 50% of the window length; then construct the spatiotemporal registration annotation, select the surface feature points and the underground reference points to establish the corresponding relationship, and the density of the registration control points is not less than 1 / 3 of each A point; then, the correspondence between geological anomalies and surface deformation is established by manual annotation by professionals and semi-supervised methods. The annotation accuracy is ensured to be no less than 90% through cross-validation; then, data enhancement technology is used to expand the training samples, including random rotation operations, with a rotation angle range of , scaling operation, the scaling factor range is 0.8 to 1.2, noise operation is added, and the signal-to-noise ratio is controlled above 15dB. Through these operations, the original samples are expanded to 3 to 5 times; finally, a large-scale training set containing surveying and mapping response characteristics under standard geological scenarios is formed, and the training set size is not less than 10,000 sample pairs. The detailed steps of geological surveying and mapping dual-domain model training are first self-supervised pre-training on a large-scale geological dataset, and the masked self-encoding method is used to learn the geological feature representation. The mask ratio is set to 15% to 25%, and the pre-training lasts for 100 to 200 rounds; then self-supervised pre-training is performed on the surveying and mapping data, and the masked self-encoding method is also used to learn the surveying and mapping feature representation. The mask ratio is set to 10% to 20%, and the pre-training lasts for 80 to 150 rounds; then the dual-domain joint fine-tuning is performed, using the geological surveying and mapping paired data as input, optimizing the cross-domain attention mechanism and the physical error correction module parameters, and the fine-tuning lasts for 50 to 100 rounds; the entire training process adopts a dynamic learning rate strategy, and the initial learning rate is set to , decay according to the cosine annealing strategy, evaluate the performance of the validation set every 10 rounds, and if there is no performance improvement for 3 consecutive rounds, the learning rate is reduced to 0.5 times the original value, and the minimum learning rate is not less than ; The optimizer uses the Adam algorithm, The parameter is set to 0.9, The parameter is set to 0.999, and the weight decay coefficient is set to During the training process, an early stopping strategy is adopted. If the performance of the validation set does not improve for 10 consecutive rounds, the training is stopped. Finally, the model parameters with the best performance on the validation set are selected.

[0056] It should be noted that the present invention uses the minimum spanning tree algorithm to optimize the dimensionality reduction of multi-source heterogeneous data of geological engineering and surveying and mapping engineering, regards the characteristic points of geological surveying and mapping data as nodes of the graph, and uses the characteristic similarity as the edge weight. By constructing a subgraph that connects all nodes and has the smallest total weight, the key structure of the data is extracted. Compared with traditional data dimensionality reduction methods such as principal component analysis or linear dimensionality reduction technology, the minimum spanning tree algorithm has the advantage of retaining the topological relationship of the data. It can maintain the core structural association between the data while reducing the data dimension, effectively avoiding information loss. This algorithm reduces the computational complexity from O( ) is reduced to O(ElogV), which is suitable for the efficient processing of massive geological mapping data. At the same time, by adjusting parameters such as the node connection threshold and the sparsity of the graph structure, the degree of dimensionality reduction and the proportion of information retention can be flexibly controlled, achieving a good balance between data compression and feature retention.

[0057] The geological surveying and mapping dual-domain model designed in the present invention is based on the Transformer architecture, and includes two parallel branches: a geological domain encoder and a surveying and mapping domain encoder. Feature interaction is achieved through a cross-domain attention mechanism. Compared with traditional data fusion models, this dual-domain model fully considers the heterogeneous characteristics of geological data and surveying and mapping data, retains the unique expression of knowledge in each domain through domain processing, and the cross-domain attention mechanism can capture the deep connection between the two types of data. The model screens important features through the self-attention mechanism, avoiding indiscriminate processing of all data, greatly improving computational efficiency; at the same time, the physical error correction module of the output layer constructs a constraint mechanism based on the principles of geomechanics and the basic laws of surveying to ensure that the analysis results conform to physical laws, avoiding additional correction calculations caused by unreasonable model output results.

[0058] The computational precision balance function introduced in this invention is an innovative mechanism for dynamically adjusting system performance. By comprehensively evaluating the intensity of multi-source data variation characteristics, the confidence level of the neural network output, the dimensionality reduction parameter set, and the computational resource constraint index, the output balance value guides the adjustment of model parameters. Unlike traditional fixed-parameter analysis methods, this function can be adaptively adjusted according to the application scenario requirements and resource conditions. When high precision is required, the number of self-attention heads and the correction strength of the physical error correction module are increased, and when high efficiency is required, the computational parameters are reduced. This dynamic mechanism enables the system to flexibly switch between different operating modes, achieving a smooth transition from high-precision analysis to real-time processing, and effectively solving the technical problem of balancing precision and efficiency in the processing of massive geological surveying and mapping data.

[0059] The three core technical ideas mentioned above form a complete technical chain from data processing to model building to system optimization, and through synergy, they produce comprehensive advantages that surpass individual technologies. The minimum spanning tree algorithm achieves preliminary dimensionality reduction and information screening at the data input level, providing more streamlined high-quality data for subsequent processing; the geological surveying and mapping dual-domain model fully utilizes the characteristics of the data after dimensionality reduction, extracting key information through domain parallel processing and cross-domain interaction; and the adaptive computational accuracy balance function runs through the entire analysis process, dynamically adjusting system parameters to adapt to the needs of different scenarios. This multi-level optimization strategy enables the system to significantly reduce computing resource consumption while maintaining analysis quality when facing massive amounts of heterogeneous data. At the same time, it flexibly adjusts the working mode according to application requirements, realizing efficient processing and accurate analysis of integrated analysis of geological engineering and surveying engineering data, and providing strong technical support for real-time monitoring, early warning and decision-making in engineering practice.

[0060] Specifically, the principle of the present invention is: the core principle of the integrated analysis method of geological engineering and surveying engineering data proposed in the present invention is to achieve efficient processing and accurate analysis of massive heterogeneous data through hierarchical data expression, optimized dimensionality reduction processing and adaptive computing framework.

[0061] First, at the data representation level, this method employs a multi-scale geological spatial data model using wavelet transforms to perform a multi-level decomposition of geological volumes, achieving a hierarchical representation from the macroscopic to the microscopic level. This multi-scale representation aligns with the inherent structural characteristics of geological volumes, flexibly selecting information at the appropriate scale based on analytical needs, and avoiding the computational burden of processing data simultaneously at all scales. Furthermore, a unified coordinate system and time base are established through a spatiotemporal registration algorithm, addressing the prerequisites for heterogeneous data fusion and laying the foundation for subsequent analysis.

[0062] Secondly, at the data processing level, this method uses the minimum spanning tree algorithm to optimize the dimensionality reduction of multi-source heterogeneous data. By treating the characteristic points of geological surveying data as nodes of the graph and the feature similarity as the edge weight, a subgraph connecting all nodes with the minimum total weight is constructed. This processing process effectively extracts the key topological relationships between data, significantly reduces the data dimension, and the computational complexity is reduced from O( ) is reduced to O(ElogV), significantly improving processing efficiency. At the same time, a probabilistic graphical model is used to establish association rules between geological body attributes and surface deformation parameters, converting the massive amount of data in these two fields into probabilistic distribution expressions, further compressing the data expression space.

[0063] At the model construction level, this method designs a dual-domain model for geological surveying and mapping based on the Transformer architecture. This model extracts domain features through parallel encoders for the geological domain and the mapping domain, respectively, and then implements cross-domain attention mechanisms to facilitate information exchange. This structural design enables the model to focus on the most relevant features, avoiding indiscriminate processing of all data and significantly improving computational efficiency. In particular, the physical error correction module in the output layer ensures that the analysis results conform to the fundamental laws of geomechanics and geodesy, effectively improving the reliability of the results while avoiding unnecessary iterative calculations.

[0064] The most innovative aspect of this method is the introduction of a computational effort-accuracy balance function. This function dynamically adjusts model complexity and physical error correction strength by comprehensively considering data characteristics, model confidence, dimensionality reduction parameters, and computational resource constraints. When the balance value is less than 0.3, the system prioritizes analysis accuracy; when the balance value is greater than 0.7, the system prioritizes computational efficiency; and when the balance value is between 0.3 and 0.7, the system maintains the parameters unchanged. This adaptive mechanism enables the system to flexibly adjust based on actual application scenarios and resource conditions, achieving an optimal balance between high-precision analysis and real-time processing.

[0065] Furthermore, this method utilizes a large-scale training dataset and a phased training strategy, combining self-supervised pre-training with dual-domain joint fine-tuning to fully exploit the inherent patterns in geological mapping data. The model training process employs a dynamic learning rate strategy, automatically adjusting optimizer parameters based on validation set performance. This ensures the model's robustness and generalization capabilities when processing massive amounts of data, further improving the system's performance in practical applications.

[0066] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows.

[0067] The specific implementation of step S01 is to construct a multi-scale geological spatial data model, perform multi-level decomposition of geological bodies through wavelet transform, and realize multi-scale expression of geological information. First, select the wavelet basis function that is suitable for the characteristics of the geological body. For example, when the geological structure boundary is obvious, Haar wavelet is selected, and when the geological body structure has a smooth transition, Daubechies wavelet is selected. Then, the geological data in the three-dimensional space is transformed into the wavelet basis function. Perform multi-scale wavelet decomposition, which is specifically expressed as:

[0068] ;

[0069] Where, is the spatial distribution function of the geological body; To decompose the scale, is the maximum decomposition scale, usually 4 to 5; is the spatial position index; is the wavelet detail coefficient; For scale The approximate coefficient on ; is the wavelet basis function; is the scaling function.

[0070] Based on the multi-scale decomposition results, the geological body feature vector is constructed , representing geological features at different scales:

[0071] ;

[0072] Where, For scale The eigenvector under ; For the characteristic components, including geological structure, physical parameters, etc.; is the feature dimension, which decreases as the scale increases. ,in is the original feature dimension.

[0073] In order to express the spatial distribution relationship of geological characteristics, radial basis functions are used to construct a continuous mathematical model:

[0074] ;

[0075] Where, is the geological characteristic interpolation function; is the weight coefficient; is the spatial position vector; is the known feature point position; is the radial basis function, usually the Gaussian kernel function ; is a smoothing parameter ranging from 0.2 to 0.5.

[0076] Weight coefficient By solving the linear equations:

[0077] ;

[0078] Where, is the coefficient matrix, ; is the weight vector; is the target value vector. This step achieves a complete description of the geological body from macro to micro through multi-scale decomposition and spatial interpolation, providing a data basis for subsequent analysis.

[0079] The specific implementation of step S02 is to standardize the surveying and mapping data using a spatiotemporal registration algorithm. First, identify the common feature points in the geological data and the surveying and mapping data, and establish a control point set. and ,in and are the number of control points in geological data and surveying data respectively. Then a spatial coordinate transformation model is established, using affine transformation or non-rigid transformation, which is specifically expressed as:

[0080] ;

[0081] Where, is the original coordinate; is the transformed coordinate; is the rotation and scaling parameter; is the translation parameter.

[0082] When the number of control points is large and irregularly distributed, thin plate spline interpolation is used to achieve non-rigid transformation:

[0083] ;

[0084] Where, is the transformation function; is the global affine transformation parameter; is the weight coefficient; is the basis function; is the position of the point to be transformed; is the control point location.

[0085] For time series data, establish an expression under a unified time base:

[0086] ;

[0087] Where, For time Surveying and mapping data at the moment; Base time Surveying and mapping data at the moment; is a function of the rate of change.

[0088] For discrete sampling data, the interpolation method is used to achieve a unified time expression:

[0089] ;

[0090] Where, is the time interpolation coefficient, satisfying ; For time point The data collected. The root mean square error is used to evaluate the registration quality:

[0091] ;

[0092] Where, is the root mean square error of registration; is the transformation function; is the source data point; is the target data point; is the number of assessment points. The registration accuracy threshold is typically set to within 1.5 times the measurement accuracy. This step establishes the positional correspondence between the surveying and mapping data and the geological data, laying the foundation for subsequent fusion analysis.

[0093] The specific implementation of step S03 is to use the minimum spanning tree algorithm to optimize the dimensionality reduction of multi-source heterogeneous data. First, the characteristic points of geological survey data are regarded as the Node, node set , edge set . Then calculate the feature similarity between nodes as the edge weight:

[0094] ;

[0095] Where, For the edge The weight of For nodes and similarity; and Node and The eigenvector of Represents the vector norm.

[0096] Then apply Kruskal algorithm to construct the minimum spanning tree :

[0097] 1. Sort all edges by weight from small to large;

[0098] 2. Initialize an empty tree ;

[0099] 3. Examine each edge in sorted order If you add If it does not form a ring, then ;

[0100] 4. When Include The algorithm terminates when there are edges.

[0101] The complexity of this algorithm is ,in is the number of edges, is the number of nodes. Based on the minimum spanning tree, key nodes and paths are extracted. Key nodes are defined as:

[0102] ;

[0103] Where, is the key node set; For nodes degree; is the average node degree; is a coefficient, and its value is usually 1.5.

[0104] The critical path is defined as:

[0105] ;

[0106] Where, is the critical path set; is the average edge weight; is a coefficient, and its value is usually 0.8.

[0107] The resulting dimensionality reduction parameter set includes:

[0108] ;

[0109] Where, is the feature dimension reduction ratio, ranging from 0.4 to 0.6; is the node connection threshold, ranging from 0.35 to 0.65; is the graph structure sparsity, ranging from 0.1 to 0.3; The data compression ratio ranges from 60% to 85%. This step achieves dimensionality reduction of multi-source heterogeneous data through graph theory optimization, preserving key information while reducing computational complexity.

[0110] The specific implementation of step S04 is to establish association rules between geological body attributes and surface deformation parameters based on the probability graph model. First, construct a probability graph. , node set ,in represents geological parameters, represents the mapping parameters. Then, conditional dependencies between nodes are established to construct a Bayesian network or Markov random field. For a Bayesian network, the conditional probability is expressed as:

[0111] ;

[0112] Where, is the joint probability distribution; For nodes About its parent node The conditional probability of .

[0113] For Markov random fields, the joint probability distribution is expressed as:

[0114] ;

[0115] Where, is the normalization factor; is the largest clique set in the graph; For the group The potential function on ; For the group The node collection in .

[0116] For continuous variables, the Gaussian mixture model is used to represent the conditional probability:

[0117] ;

[0118] Where, For a given geological state Subsurface response The conditional probability of For the The weights of the mixture components satisfy ; The mean is , the covariance matrix is Gaussian distribution; is the amount of mixed ingredients.

[0119] Mean function Expressed as a function of geological parameters:

[0120] ;

[0121] Where, is the weight matrix; is the bias vector.

[0122] Parameter estimation uses the expectation maximization algorithm, and the iterative process is:

[0123] 1. Step E: Calculate the posterior probability ;

[0124] 2.M step: update parameters 、 、 and .

[0125] Iterate until convergence, the convergence threshold is set to This step establishes a probabilistic correlation model between geological body attributes and surface deformation parameters, achieving a quantitative expression of geological state and surface response.

[0126] The specific implementation of step S05 is to use the adaptive weight fusion algorithm to integrate the geological internal structure information and the surface monitoring results. First, the quality indicators of different data sources are evaluated and the quality score vector is constructed. , . Then calculate the correlation between each data source and the target task:

[0127] ;

[0128] Where, For the Relevance indicators of data sources; For data sources With the target variable mutual information between them; is the joint probability distribution; and are marginal probability distributions respectively.

[0129] For reliability assessment, construct the reliability coefficient vector , Based on the above three indicators, calculate the initial fusion weight:

[0130] ;

[0131] Where, For the The initial weight of each data source.

[0132] During the iteration process, the weights are dynamically adjusted according to the fusion effect:

[0133] ;

[0134] Where, For the The first iteration The weight of each data source; is the learning rate, usually ranging from 0.05 to 0.1; is the fusion effect evaluation function; Weight The gradient of the evaluation function.

[0135] The final generated fusion feature set is expressed as:

[0136] ;

[0137] Where, is the fusion feature set; For the The feature set of each data source; This step realizes the adaptive fusion of heterogeneous data and ensures that high-quality data plays a dominant role in the analysis results.

[0138] The specific implementation of step S06 is to input the geological survey fusion feature set into the geological survey dual domain model and perform integrated analysis. First, the fusion feature set is divided into geological domain data and surveying domain data . Then input the corresponding encoder for feature extraction:

[0139] ;

[0140] ;

[0141] Where, and are the coding features of the geological domain and the mapping domain respectively; and They are geological domain encoder and surveying and mapping domain encoder respectively.

[0142] The encoder is based on the Transformer architecture, and the self-attention mechanism is expressed as:

[0143] ;

[0144] Where, 、 、 They are query matrix, key matrix and value matrix respectively; is the dimension of the key vector.

[0145] The multi-head attention mechanism is expressed as:

[0146] ;

[0147] ;

[0148] Where, is the number of attention heads; 、 、 and is the parameter matrix.

[0149] The cross-domain attention mechanism is used to achieve feature interaction between the two domains:

[0150] ;

[0151] ;

[0152] Where, It represents the attention characteristics of the geological domain to the mapping domain; Represents the attention characteristics of the surveying and mapping domain to the geological domain.

[0153] Then the dual domain features are fused through the decoder:

[0154] ;

[0155] Where, This is the preliminary analysis result output by the decoder.

[0156] Finally, the results are corrected through the physical error correction module:

[0157] ;

[0158] Where, is the analysis result after correction; is the correction intensity coefficient; is a correction function based on physical constraints.

[0159] Correction intensity factor Determined by the calculation accuracy balance function:

[0160] ;

[0161] Where, is the intensity of the changing features of multi-source data; Output confidence for the neural network; is the dimension reduction parameter set; It is the computing resource constraint indicator.

[0162] Balance value output by the balance function Determines the model parameter adjustment strategy:

[0163] ;

[0164] Where, is the equilibrium value, and its value range is [0, 1]; is the sigmoid function; is the weight coefficient; is the bias term.

[0165] when When , increase the number of attention heads: ,in ;

[0166] when When , reduce the number of attention layers: ,in ;

[0167] when , keeping the model parameters unchanged.

[0168] This step realizes the deep integration analysis of geological engineering and surveying and mapping engineering data through the dual-domain model, ensuring the analysis accuracy while taking into account the computational efficiency.

[0169] The detailed structure of the geological surveying and mapping dual-domain model is based on the Transformer architecture and consists of two parallel branches: a geological domain encoder and a mapping domain encoder. The geological domain encoder processes geological internal structure information and consists of multiple layers of self-attention modules, each containing a multi-head self-attention sublayer and a feedforward network sublayer. The calculation process of the self-attention mechanism is as follows:

[0170] ;

[0171] ;

[0172] Where, Enter characteristics for the geological domain; 、 、 is the parameter matrix; is the key vector dimension.

[0173] The multi-head attention mechanism is expressed as:

[0174] ;

[0175] ;

[0176] Where, is the number of attention heads, and is related to the complexity of geological structure Related: , ; 、 、 and is the parameter matrix.

[0177] The feedforward network sublayer is represented as:

[0178] ;

[0179] Where, 、 、 、 are the parameter matrix and bias vector.

[0180] The structure of the surveying and mapping domain encoder is similar to that of the geological domain encoder, but the parameter settings are different. The cross-domain attention mechanism enables feature interaction between the two domains:

[0181] ;

[0182] ;

[0183] ;

[0184] Where, and They are the coding features of the geological domain and the surveying and mapping domain respectively.

[0185] The decoder consists of a multi-layer attention module that fuses the dual-domain features and outputs the analysis results through a fully connected layer. The physical error correction module constructs constraints based on geomechanics principles to correct prediction results that violate physical laws:

[0186] ;

[0187] Where, Loss of physical restraint; For the A physical constraint function; is the constraint weight; is the number of constraints.

[0188] The correction process is expressed as:

[0189] ;

[0190] Where, is the result after correction; To correct the intensity, it is determined by the output value of the calculation accuracy balance function; is the gradient of the physical constraint loss with respect to the output.

[0191] The detailed steps for establishing the training dataset for the geological surveying and mapping dual-domain model include collecting historical geological survey and mapping monitoring data from multiple regions, classifying and organizing them according to geological structure type and mapping data type. Then, the sliding window method is used to segment the time series data:

[0192] ;

[0193] ;

[0194] Where, is the input time window; is the output time window; is the window length; is the predicted length; The starting time index.

[0195] By constructing spatiotemporal registration annotations, the correspondence between geological anomalies and surface deformation is established. Finally, data augmentation techniques are used to expand the training samples, including random rotation, scaling, and noise addition:

[0196] ;

[0197] Where, is the enhanced sample; is the rotation matrix, is the rotation angle; is the scaling matrix, is the scaling factor; is Gaussian noise.

[0198] The entire training process consists of two stages: self-supervised pre-training and supervised fine-tuning. Self-supervised pre-training uses masked self-encoding to learn feature representations:

[0199] ;

[0200] ;

[0201] Where, The input after adding the mask; To reconstruct the output; To rebuild the losses.

[0202] During the supervised fine-tuning phase, the cross-domain attention mechanism and physical error correction module parameters are optimized:

[0203] ;

[0204] Where, is the fine-tuning loss; To predict losses; Loss of physical restraint; is the balance coefficient.

[0205] The entire model training process adopts a dynamic learning rate strategy, and the learning rate update rule is:

[0206] ;

[0207] Where, For the The learning rate of the round iteration; is the initial learning rate; The optimizer parameters are automatically adjusted based on the validation set performance, and the model parameters with the best performance are finally selected.

[0208] The following explains some common formulas in the prior art involved in this embodiment.

[0209] Optionally, in step S02, the rate of change function : ;

[0210] Where, For time The rate of change of time; is the exponential decay coefficient, ranging from 0.001 to 0.01; is the attenuation constant, ranging from 0.1 to 0.5; is the amplitude of periodic variation, ranging from 0.005 to 0.02; is the angular frequency; is the initial phase; is the base time.

[0211] The parameter acquisition method is: The method adopts historical data fitting, including step 1: collecting the change data of the same type of geological bodies in the past 10 years; step 2: fitting the exponential trend term using the least squares method. The spectrum analysis is used to obtain the data, including step 1: Fourier transform of time series data; step 2: extract the amplitude of the main frequency component as value.

[0212] Optionally, in step S03, the calculation amount precision balance function : ;

[0213] Where, is the output value of the balance function, and its value range is [0,1]; , , , is the weight coefficient; is the bias term; is the intensity of the changing features of multi-source data; Output confidence for the neural network; is the comprehensive score of the dimension reduction parameter set; It is the computing resource constraint indicator.

[0214] The parameter acquisition method is: The statistical analysis is used to obtain the data, including step 1: calculating the coefficient of variation of each data source within the time window; step 2: taking the weighted average of the coefficient of variation to obtain value. The system monitoring is used to obtain the information, including step 1: real-time monitoring of CPU and memory usage; step 2: calculating the constraint strength based on resource usage, .

[0215] Optionally, in step S04, the potential function : ;

[0216] Where, For the group The potential function on ; For nodes and The strength of the interaction between is the incompatibility measure between nodes, defined as ,in A small constant to prevent division by zero.

[0217] Normalization factor : ;

[0218] Where, is the normalization factor, ensuring the normalization condition of the probability distribution; the sum traverses all possible variable configurations ; is the largest clique in the graph.

[0219] Optionally, in step S05, the fusion effect evaluation function : ;

[0220] Where, is the comprehensive evaluation function; is the mean square error; is the KL divergence; is the consistency indicator; , , is the weight coefficient; For the the relevance of the data sources; is the average correlation.

[0221] Gradient function : ;

[0222] Where, ; ; .

[0223] Optionally, in step S06, the physical constraint correction function : ;

[0224] Where, is a correction function based on physical constraints; is the number of constraints; For the The correction strength of each constraint ranges from 0.1 to 0.5; For the The sensitivity parameter of each constraint ranges from 1 to 5; The output Quantity It is a reference value determined based on physical laws.

[0225] Physical constraint function : ;

[0226] Where, For the A physical constraint function; For the Mathematical expression of physical laws, including stress equilibrium conditions , deformation continuity conditions wait; For the The tolerance threshold of each constraint; is the stress tensor; is the density; is the acceleration due to gravity; is the displacement vector.

[0227] The parameter acquisition method is: Cross-validation is used to obtain the results, including step 1: testing different Correction effect of the value; Step 2: Select the optimal comprehensive index value. Obtained by statistical analysis, including step 1: collecting the distribution of physical constraint violation in historical data; step 2: taking the 95% quantile as the tolerance threshold . The value range of is 1 to 5. In order to better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: In a tunnel project in a certain mountainous area, researchers applied an integrated analysis method of geological engineering and surveying engineering data to evaluate and predict geological risks during tunnel construction. The tunnel is 3.5 kilometers long and passes through a variety of complex geological environments, including fault zones, karst areas, and high-stress areas, posing a high construction safety risk. In order to predict possible geological disasters in advance and take effective measures, researchers collected multi-source heterogeneous data, including geological drilling data, geophysical survey results, and surface deformation monitoring data, for integrated analysis.

[0228] First, the researchers constructed a multiscale geological spatial data model according to step S01. Based on the geological data acquired through drilling, they performed a wavelet transform multiscale decomposition of the geological bodies along the tunnel, using the Daubechies-4 wavelet basis function and a decomposition level of four. Table 1 shows the accuracy of geological feature representation at different decomposition levels.

[0229] Table 1 Wavelet transform multi-scale decomposition accuracy

[0230]

[0231] During radial basis function interpolation, the smoothing parameter β was set to 0.35. This interpolation resulted in a continuous geological distribution function that accurately represents geological structural features such as fault zones and karst caves. The multiscale geological spatial data model can represent geological information at multiple levels, from macroscopic to microscopic, providing a data foundation for subsequent risk analysis.

[0232] The researchers then performed the spatiotemporal registration process in step S02. Forty-two control points were placed along and around the tunnel, including 18 surface landmarks and 24 underground reference points, as shown in Table 2:

[0233] Table 2 Distribution of spatiotemporal registration control points

[0234]

[0235] The thin plate spline interpolation method was used for spatial registration, and the RMSE value of the registration accuracy was 3.2mm, which was lower than the 1.5 times threshold of the measurement accuracy of 4.5mm. In terms of temporal registration, a monitoring frequency of 7 days was used as the standard time base to unify the surveying and mapping data collected at different time periods. Figure 2 As shown in the figure, the spatiotemporal registration process significantly improves the consistency of multi-source data and lays the foundation for subsequent analysis.

[0236] In step S03, the researchers performed dimensionality reduction using the minimum spanning tree algorithm on the registered multi-source heterogeneous data. 30% of the feature points in the original data were extracted as graph nodes, for a total of 1286 nodes. Inter-node feature similarity was calculated using cosine distance, with a similarity threshold of 0.45. The Kruskal algorithm was used to construct a minimum spanning tree, as shown in Table 3:

[0237] Table 3 Minimum spanning tree dimensionality reduction effect table

[0238]

[0239] 37 key nodes and 52 key paths were extracted from the minimum spanning tree as key indicators for geological risk assessment. The dimensionality reduction parameter set is {0.52, 0.45, 0.18, 78%}, which means the feature dimension reduction ratio is 0.52, the node connection threshold is 0.45, the graph structure sparsity is 0.18, and the data compression rate is 78%. Figure 3 As shown in Figure 3, the dimensionality reduction process retains key geological structure information while significantly reducing the amount of calculation.

[0240] In step S04, a probability graph model is constructed based on the data after dimensionality reduction, and the association rules between geological body attributes and surface deformation parameters are established. The key geological parameters selected include 15 variables such as rock mass strength, fault dip, and groundwater level; the surveying and monitoring parameters include 12 variables such as surface settlement, horizontal displacement, and tilt angle. The Bayesian network is used to construct the conditional probability model, and the dependency strength threshold is set to 0.28. The number of Gaussian mixture model components is set to 3, and the parameters are trained using the expectation maximization algorithm. The iterative convergence threshold is , and finally get the conditional probability distribution , as shown in Table 4:

[0241] Table 4. Association probability between geological conditions and surface responses

[0242]

[0243] In step S05, the researchers applied an adaptive weight fusion algorithm to integrate geological structure information and monitoring results. The quality scores, relevance, and reliability of different data sources were evaluated, and the initial weights were assigned as follows: geological drilling data 0.35, geophysical exploration data 0.28, InSAR monitoring data 0.22, and GNSS monitoring data 0.15. During the iterative fusion process, the weights were dynamically adjusted according to the data performance, and the final weights were adjusted to: geological drilling data 0.32, geophysical exploration data 0.23, InSAR monitoring data 0.28, and GNSS monitoring data 0.17. The adjustment step size was set to 0.08, and the weights stabilized after 5 iterations. Figure 4 As shown, adaptive weight fusion significantly improves the accuracy of risk prediction.

[0244] Finally, in step S06, the fused feature set is input into the geological surveying and mapping dual-domain model for integrated analysis. The model adopts the Transformer architecture. The geological domain encoder adopts an 8-layer self-attention structure with 10 attention heads; the surveying domain encoder adopts a 6-layer structure with 8 attention heads. The total number of model parameters is The physical error correction module is based on geomechanical constraints, and the correction strength is determined by the calculation accuracy balance function. The balance value is 0.45, indicating a state of equilibrium. Through this model analysis, three potential risk areas during tunnel construction were accurately predicted: the fault fracture zone at K1+250, the karst caves at K2+420, and the high-stress area at K3+180. The predicted settlements were 28.5mm, 35.7mm, and 21.3mm, respectively.

[0245] Traditional tunnel geological risk assessment relies primarily on geological surveys and empirical judgment, resulting in information isolation, low data integration, and insufficient prediction accuracy. Single-domain models are often used to process geological and surveying data separately, followed by manual integration and analysis, leading to information loss and subjective bias. In this example, the traditional method predicted settlements of 33.2 mm, 41.5 mm, and 16.8 mm in three risk areas, differing from the actual monitored values by 18.7%, 13.2%, and 25.7%, respectively, with an average error of 19.2%. However, using the integrated analysis method for geological and surveying data presented in this paper, the predicted settlements differed from the actual monitored values (30.1 mm, 36.8 mm, and 22.5 mm, respectively) by only 5.3%, 3.0%, and 5.3%, with an average error of 4.5%. Overall, this paper significantly improves the integrated analysis capabilities of geological and surveying data through the integration of multi-scale spatial representation, spatiotemporal registration, graph optimization, probabilistic modeling, and dual-domain deep learning, providing more reliable technical support for engineering safety and decision-making.

[0246] It should be noted that the variables involved in the present invention are explained in detail as shown in Tables 5, 6, 7 and 8 below.

[0247] Table 5 Variable Explanation Table (Part I)

[0248]

[0249] Table 6 Variable Explanation Table (Part II)

[0250] Table 7 Variable Explanation Table (Part 3)

[0251]

[0252] Table 8 Variable Explanation Table (Part 4)

[0253]

[0254] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for integrated analysis of geological engineering and surveying and mapping engineering data, characterized in that: include: Construct a multi-scale geological spatial data model and perform multi-level decomposition of geological bodies; use spatiotemporal registration algorithms to standardize surveying and mapping data; The minimum spanning tree algorithm is used to optimize and reduce the dimensionality of multi-source heterogeneous data, construct a feature connection network of geological surveying and mapping data, extract key nodes and paths, and generate a dimensionality reduction parameter set; based on the probabilistic graphical model, the association rules between geological body attributes and surface deformation parameters are established; the adaptive weight fusion algorithm is used to integrate geological internal structure information and surface monitoring results to form a geological surveying and mapping fusion feature set; the geological surveying and mapping fusion feature set is input into the geological surveying and mapping dual-domain model, and the model parameters are dynamically adjusted in combination with the calculation amount and accuracy balance function. Iterative optimization is performed until convergence, and the integrated analysis results of geological engineering and surveying and mapping engineering are output; the spatiotemporal registration algorithm is a data alignment method based on control point matching and non-rigid transformation. By identifying common feature points in geological data and surveying and mapping data, the spatial correspondence between data is established, and the unified expression of data collected at different times is achieved through time series analysis.

2. The method according to claim 1, characterized in that The multi-scale geological spatial data model extracts features of geological bodies at different scales through wavelet transform, constructs a multi-level mathematical expression including geological structure, physical parameters and spatial distribution characteristics, and realizes a complete description of geological bodies from regional to local.

3. The method according to claim 2, characterized in that The minimum spanning tree algorithm is a graph theory optimization method that regards the characteristic points of geological surveying data as nodes of the graph and the feature similarity as the edge weight. It realizes the extraction of key data structures by constructing a subgraph that connects all nodes and has the minimum total weight.

4. The method according to claim 3, characterized in that The dimensionality reduction parameter set includes the feature dimension reduction ratio, node connection threshold, graph structure sparsity, and data compression rate, which are used to quantify the data loss and information retention degree in the dimensionality reduction process. The dimensionality reduction parameter set serves as the input parameter of the computational accuracy balance function.

5. The method according to claim 4, characterized in that The probabilistic graphical model is a statistical model that uses graph theory to represent the dependency relationship between random variables. It uses nodes to represent geological and surveying parameters, and edges to represent the correlation between parameters. It establishes the conditional probability distribution between parameters and realizes the uncertainty reasoning of geological state and surface response.

6. The method according to claim 5, characterized in that The adaptive weight fusion algorithm dynamically adjusts the weight coefficients of various types of data in the fusion process according to data quality, reliability and relevance, ensuring that high-quality data has a greater contribution to the final analysis results and improving analysis accuracy.

7. The method according to claim 6, characterized in that The geological surveying and mapping fusion feature set is a collection of geological internal structure information and surface monitoring results integrated through an adaptive weight fusion algorithm. It includes the lithologic distribution, structural characteristics, physical and mechanical parameters of the geological body, as well as the surface deformation, displacement gradient, and deformation rate in the surveying and mapping data, which are data indicators that describe the core characteristics of geological engineering and surveying and mapping engineering.

8. The method according to claim 7, characterized in that The specific structure of the geological surveying and mapping dual-domain model is an encoder-decoder network based on the Transformer architecture, which includes two parallel branches: the geological domain encoder and the surveying and mapping domain encoder. Each branch is composed of a multi-layer self-attention module, and feature interaction is achieved through a cross-domain attention mechanism. The decoder part uses a multi-head attention mechanism to fuse dual-domain features, and the output layer contains a physical error correction module as a submodule.

9. The method according to claim 8, characterized in that The geological surveying and mapping dual-domain model uses adaptive gating units to screen features. The total parameters of the geological surveying and mapping dual-domain model are dynamically adjusted according to the complexity of geological data and the density of surveying and mapping data. The number of self-attention heads is proportional to the complexity of geological body structure, and the attention dimension is proportional to the accuracy level of surveying and mapping data.

Images