A tunnel geological information modeling and dynamic updating method and system

Abstract

The application discloses a tunnel geological information modeling and dynamic updating method and system. The method first collects drilling data before construction and stratum information exposed in a working face during construction, defines a lithology category set and converts the lithology category set into an indicator variable after standardization processing; calculates an empirical semi-variance based on the standardized data set, obtains a variation parameter through variogram fitting; determines a neighborhood radius according to the variation parameter, and constructs an initial geological model by using an indicator Kriging method; in the tunneling process, newly exposed stratum information is incrementally integrated, the variation parameter is adaptively adjusted, the geological model is updated, and uncertainty quantification is simultaneously performed; finally, the most possible lithology category of each spatial position is determined based on the updated geological model, and a lithology distribution map is generated. The application realizes lithology category probability prediction and geological model dynamic updating, can accurately depict tunnel geological spatial characteristics, quantitatively predict uncertainty, and improves prediction accuracy and reliability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel engineering geological modeling, and particularly relates to a tunnel geological information modeling and dynamic updating method and system. BACKGROUND

[0002] In the process of tunnel construction, the geological conditions have strong heterogeneity and significant uncertainty, and the structure and physical parameters of surrounding rock change rapidly along the spatial direction, which brings serious challenges to the safety control of construction. The traditional geological modeling method mainly relies on sparse drilling data before construction, and the model is usually static and updated with lag, which is difficult to reflect the real geological evolution characteristics in the process of tunneling.

[0003] In underground engineering, especially in mountain tunnel construction, the geological conditions often have strong heterogeneity, sudden change and spatial uncertainty. The changes of surrounding rock lithology, structural plane and weak interlayer have a decisive influence on construction safety, support parameter design and advanced prediction. However, due to the limitation of the small number of pre-drilling, sparse distribution and limited representativeness, the traditional tunnel geological model is mostly a static modeling result, which cannot reflect the real situation of the changing surrounding rock with the tunneling exposure, resulting in model lag, insufficient risk identification and limited prediction ability. This problem is particularly prominent in areas with obvious rock undulation, strong local structure or weak interlayer development. Therefore, how to build a geological model that can be continuously updated with the construction progress and serve the construction risk control and geological prediction has become a key scientific and engineering problem that needs to be broken through in the current tunnel industry.

[0004] At the same time, the importance of uncertainty expression in tunnel construction is increasingly prominent. Due to the large drilling interval and the characteristics of obtaining information step by step, the model often has high uncertainty near the layer boundary, structure disturbance area and unexposed area. If only relying on "single lithology map" for geological judgment, potential dangers may be covered up, affecting the construction organization design, support form selection and disaster prediction ability. Therefore, building a geological model that can express the probability distribution of lithology and quantify the prediction uncertainty is an important direction to improve the geological guarantee ability of tunnel construction.

[0005] At present, there are many geological modeling techniques applied in the field of underground engineering, such as Markov random field (MRF), multi-point geostatistics (MPS) and classification prediction model based on machine learning. These methods perform well in mining, oil reservoir and general stratum modeling, but still have deficiencies in the scene of "data exposed step by step, model needs real-time updating" such as tunnel construction, such as relying on experience, weak dynamic updating ability or lack of probability explanation.

[0006] In view of the above, it is necessary to provide a new tunnel engineering geological modeling method. SUMMARY

[0007] The purpose of the present application is to provide a tunnel geological information modeling and dynamic updating method and system, which can realize lithology category probability prediction.

[0008] The technical scheme provided by the present application is:

[0009] In a first aspect, the present application provides a tunnel geological information modeling and dynamic updating method, comprising:

[0010] Data acquisition and preprocessing: collecting drilling data before construction and stratum information exposed by the working face during construction, and performing standardization processing on the drilling data and the exposed information to obtain a standardized data set; based on the drilling data in the standardized data set, defining a lithology category set, and converting the lithology category variable into an indicator variable;

[0011] Variation function fitting: based on the standardized data set, calculating the empirical semivariance of the sample point indicator variable, fitting the empirical semivariance using a variation function, and obtaining a variation parameter;

[0012] Initial geological model construction: determining a neighborhood radius according to the variation parameter; estimating an initial geological model through the indicator kriging method, wherein the geological model (lithology probability field) includes a set of conditional probability distributions of each spatial position in the tunnel profile belonging to each type of lithology; wherein, for any prediction point in the tunnel profile, the lithology category indicator variable of the prediction point is obtained by weighting the lithology category indicator variables of each sample point in the neighborhood of the prediction point in the standardized data set; and the conditional probability of the prediction point belonging to each type of lithology is obtained according to the lithology category indicator variable of the prediction point.

[0013] Dynamic updating of the geological model: during tunnel excavation, based on the stratum information exposed by the working face continuously acquired in the data acquisition and preprocessing step, incrementally integrating the newly exposed stratum information into the standardized data set to form an updated data set; based on the updated data set, adaptively adjusting the variation parameter in the variation function, and then based on the adjusted variation parameter and the updated data set, updating the initial geological model through the indicator kriging method to obtain an updated geological model;

[0014] Generating a lithology distribution map: based on the updated geological model, determining the most likely lithology category of each spatial position, and generating a lithology distribution map.

[0015] In some possible implementation manners, in the data acquisition and preprocessing step, the standardization processing includes coordinate registration, lithology coding, and gridding processing, and the gridding processing is to discretize the tunnel profile along the line into a two-dimensional regular grid, so that the data is adapted to the subsequent indicator kriging modeling requirements.

[0016] In some possible implementation manners, in the initial geological model construction, the definition of the indicator variable is: if the position Lithology class of the point For the first class, the corresponding indicator variable = 1, otherwise = 0, where , is a set of lithology classes, , is the number of lithology classes.

[0017] In some possible implementations, for any prediction point (unknown location) , the corresponding lithology class indicator variable estimate value is calculated by the following formula:

[0018] ;

[0019] wherein, represents an indicator variable of whether the lithology class of the prediction point is the first class; is an indicator variable of whether the lithology class of the th sample point in the neighborhood of the prediction point is the first class, is the Kriging weight of the th sample point in the neighborhood of the prediction point , determined by a Kriging system satisfying the unbiasedness and minimum variance conditions; is the number of sample points in the neighborhood of the prediction point ;

[0020] The lithology class probability calculation formula is:

[0021] ;

[0022] In the formula, represents the probability estimate value of the lithology class of the prediction point being the first class;

[0023] The independently estimated probabilities of each lithology class are normalized, and the formula is: , wherein, represents the normalized probability of the lithology class of the prediction point being the first class; the normalized probabilities of each lithology class at each prediction point form a complete geological model.

[0024] In some possible implementations, the value range of the neighborhood radius is ~ ,in The correlation length is the length of the fitted variation function.

[0025] In some possible implementations, the empirical semivariance of the sample point indicator variable is calculated using the following formula: ,in, For empirical semivariance, and Sample points and Is the lithology category the first? Class indicator variables; Represents all sample point pairs A set whose interval lies within the interval [ ]. Inside; Let be the logarithm of the sample points; where The lag distance, The lag width is used to determine the set of valid point pairs. ;

[0026] The variation function is an exponential variation function, and its expression is: ,in Value of a nugget. For sill values, For the relevant length; , and Together they constitute the set of variation parameters.

[0027] In some possible implementations, the dynamically updated geological model is divided into update stages according to the tunnel excavation ratio, with each 20% excavation stage constituting an update stage. The exposed data within the corresponding interval is then incorporated sequentially, and the updated dataset is a merged set of historical standardized data and newly exposed data.

[0028] In some possible implementations, the variation parameter adjustment in the dynamic geological model update step adopts a sliding weighted method, and the update formula is: ,in Indicates the previous stage, i.e., the phase. The set of variation parameters, This is the set of variation parameters obtained by refitting based on the new data. To update the weighting coefficients.

[0029] In some possible implementations, the step of dynamically updating the geological model further includes: quantifying the uncertainty of the updated geological model, whereby uncertainty quantification refers to characterizing the degree of ambiguity in lithology determination at each spatial location through a quantification index; wherein, uncertainty quantification is achieved by calculating information entropy, and the formula for calculating information entropy is: ,in, representative point information entropy of the representative point the lithology category of the representative point the probability of the representative point being of the first category is greater, the higher the uncertainty of the lithology determination of the representative point is greater, the higher the uncertainty of the lithology determination of the representative point

[0030] In some possible implementation manners, the step of dynamically updating the geological model further includes: model accuracy evaluation; and an overall accuracy evaluation index OA is calculated according to the following formula: wherein, is the total number of sample points used for verification (for example, the total number of lithology observation points used for verification in a tunnel profile), is the most probable lithology category predicted by the sample point is the actual lithology category (true value) of the sample point is the actual lithology category (true value) of the sample point is an indicator function, and the function value is 1 when is an indicator function, and the function value is 1 when is an indicator function, and the function value is 1 when

[0031] In a second aspect, the present application provides an electronic device, comprising: a memory and a processor;

[0032] the memory is configured to store a computer program;

[0033] the processor is configured to invoke the computer program to perform the method described above.

[0034] In a third aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is configured to cause an electronic device to perform the method described above when the computer program is run on the electronic device.

[0035] In a fourth aspect, the present application provides a computer program product, comprising a computer program, and the computer program is configured to cause an electronic device to perform the method described above when the computer program is run on the electronic device.

[0036] The specific implementation manners of the second to fourth aspects of the present application described above can refer to the implementation manners of the first aspect described above, and will not be described herein.

[0037] Advantages:

[0038] The application discloses a tunnel geological information modeling and dynamic updating method. The method is based on an indicator Kriging (IK) method, which can convert multiple types of lithology into multiple binary indicator fields, and then realize lithology probability prediction and uncertainty expression, and has better interpretability and adaptability in a complex stratum environment. The application constructs a dynamic indicator Kriging (DIK) tunnel geological modeling scheme, which can solve the problem that a traditional static geological modeling method is difficult to adapt to rapid evolution of strata in a tunnel construction process, and realizes automatic updating and iterative prediction of a geological model with exposed information in tunneling. The method first takes drilling data before construction as initial input to construct an initial geological model. In the construction process, new stratum information exposed by a working face is accumulated, and a dynamic updating mechanism is used to gradually correct the geological model and spatial correlation structure, so that the model can continuously adapt to the evolving underground geological conditions. On the basis of indicator variable conversion and Kriging estimation, the model can simultaneously perform probability expression on multiple types of lithology, and further realizes quantitative evaluation of prediction results and uncertainty through maximum a posteriori (MAP) prediction and information entropy, thereby identifying potential high-risk sections, and realizing enhanced interpretability of model results and significant highlighting of risk areas. Experiments based on a profile of the Zhaibei Mountain tunnel show that the proposed dynamic indicator Kriging method can accurately reconstruct main stratum structures by relying on limited drilling information, has good modeling capability in the initial stage, and after gradually introducing exposed data in the tunneling process, the model prediction accuracy is continuously improved, the lithology boundary is gradually clear, and the uncertainty gradually converges with the stage advancing. The research results show that the dynamic indicator Kriging framework can realize rolling updating and high-reliability prediction of a tunnel geological model, has theoretical robustness and engineering application value, and can provide effective technical support for geological information integration, risk prediction, intelligent decision-making and digital construction in the tunnel construction process. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 It is a method step schematic diagram of the application.

[0040] Figure 2 It is a borehole data marking and lithology class indicator variable schematic diagram, wherein (a) shows the lithology class label distribution of the borehole position; (b) corresponds to the indicator variable distribution of the lithology class 2.

[0041] Figure 3 It is a geological profile along the direction of the Zhaibei Mountain tunnel.

[0042] Figure 4 It is a known borehole distribution diagram.

[0043] Figure 5 It is an initial stage cross-validation result diagram.

[0044] Figure 6 The fitted curves are for the variation functions of each lithology category in the initial stage.

[0045] Figure 7 This is a probability distribution diagram of each lithology category in the initial stage.

[0046] Figure 8 This is the initial stage maximum a posteriori (MAP) lithology distribution map.

[0047] Figure 9 This is the information entropy diagram for the initial stage.

[0048] Figure 10 This is a distribution map of known information for each of the five phases (Phase 1-Phase 5).

[0049] Figure 11 This is a MAP lithology distribution map for each stage from the first to the fifth stage.

[0050] Figure 12 This is a diagram showing the information entropy distribution of each stage from the first to the fifth phase. Detailed Implementation

[0051] To enable those skilled in the art to better understand the present application, the technical solution of the present application will be further described in detail below with reference to the embodiments and accompanying drawings.

[0052] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0053] Specific embodiments according to this application will now be described with reference to the accompanying drawings.

[0054] Example 1:

[0055] Figure 1 This is a schematic diagram illustrating the method steps of this application. For example... Figure 1 As shown, this embodiment provides a method for tunnel geological information modeling and dynamic updating, including:

[0056] S1. Data Acquisition and Preprocessing: Collect drilling data before construction and stratum information exposed at the working face during construction. Standardize the drilling data and exposed information to obtain a standardized dataset. Based on the drilling data in the standardized dataset, define a set of lithology categories and transform the lithology category variables into indicator variables.

[0057] The following section explains the data representation and modeling approach used in this step.

[0058] In tunnel construction, geological information mainly comes from two types of data: (1) pre-construction drilling data, which provides initial lithological categories and stratum depths, serving as prior constraints for the geological model; and (2) stratum information revealed during construction. This new data can be gradually acquired as the tunneling progresses, serving as an important basis for model updates. This application takes the tunnel profile as the research object, discretizing it into a two-dimensional regular grid, with each grid unit corresponding to a unique spatial coordinate point. For the locations of boreholes and exposed points, the lithological categories are known; however, the lithological information of unexcavated areas is unknown and needs to be predicted through geological modeling. Let the set of lithological categories be... These represent different stratigraphic units (such as sand, silty clay, gravel, etc.). Indicates the number of lithology categories. For example... Figure 2 As shown in (a), the locations of drill holes are marked with labels, while non-drill locations are recorded as missing values ​​(unlabeled).

[0059] The core objective of geological modeling is to estimate the conditional probability distribution of a given lithology at all predicted points. That is, location The lithological category is the first The probability of class, where This probability can not only be used to obtain the most probable lithological distribution results, but also provides a basis for subsequent uncertainty analysis and risk identification.

[0060] To facilitate probabilistic modeling, the lithology category variable is transformed into an indicator variable. Let... For position The lithological category at the location, then for each category Define indicator variables:

[0061] ;

[0062] in, This indicates an indicator variable used to transform multi-lithological category lithology problems into spatial estimation problems involving multiple binary fields. If the location... Lithology at the location For the first Class, then ;otherwise .

[0063] like Figure 2 (b) for Figure 2 (a) Corresponding lithology category 2 (Indicator variable plot). Figure 2 In (a), the pixel , The label at this location is 2, and the pixel is... The label at that location is 3. Therefore, in the indicator variable plot, the cell... , and The indicator variables are respectively , , .

[0064] Indicator variables transform the multi-lithological classification problem into the estimation problem of several binary fields, allowing the Kriging method to be applied independently to spatial modeling and probabilistic prediction for each lithological category. Its conditional expectation is expressed as:

[0065] ;

[0066] in, This refers to data from the borehole, specifically the lithology of the borehole location.

[0067] because It takes values ​​in the range [0,1], and its expectation has a direct probabilistic interpretation. For example, when... When, the lithological category indicating location 𝑥 is number 1. The probability of this class is 80%, thus having direct geological probability interpretation significance.

[0068] S2. Variation function fitting: Based on the standardized dataset, calculate the empirical semivariogram of the indicator variable for the sample points, and fit the empirical semivariogram with the variation function to obtain the variation parameters.

[0069] The variogram is a core tool in geostatistics for characterizing spatial correlation and structural continuity. Its form reflects the statistical laws governing the variation of geological properties with spatial distance. For indicator kriging models, the variogram not only determines the allocation of weights but also directly affects the smoothness of stratigraphic interfaces and prediction accuracy, thus playing a crucial role in the modeling process.

[0070] (1) Definition of the variation function:

[0071] For any lithology category c∈L, the empirical semivariance estimation formula for its indicator variable is defined as follows:

[0072] ;

[0073] in, For empirical semivariance, and Sample points and Is the lithology category the first? Class indicator variables; Represents all sample point pairs A set whose interval lies within the interval [ ]. Inside; Let be the logarithm of the sample points; where is the lag distance, is the lag width, used to determine the set of valid point pairs : only when the distance between point pairs falls into the interval , it is used for the statistics of the lag distance. and are empirical parameters. The lag distance sequence can be obtained by equally dividing the maximum distance into several intervals, while is consistent with the segment width to ensure that each lag interval has a sufficient number of sample point pairs. The increase of empirical semivariogram with usually indicates that the spatial correlation of lithology category variables decays with distance.

[0074] (2) Theoretical model fitting

[0075] In practical engineering, empirical semivariogram often shows a discrete distribution due to the influence of sample size, spatial distribution, and observation error, so a theoretical model needs to be selected for smoothing fitting. This application adopts an exponential variogram, whose expression is:

[0076] ;

[0077] where denotes the theoretical semivariogram; is the correlation length, which characterizes the scale of spatial correlation decay with distance, and its value is automatically obtained by least squares fitting of the empirical semivariogram; is the sill value, which is approximately equal to the population variance of the indicator variable, reflecting the overall variation degree of lithology categories; is the nugget value, used to describe microscale changes or observation errors, and is also determined by the fitting process. The above parameters together constitute the variogram parameter set, which is the core input of the Kriging system, enabling IK to reasonably depict the spatial probability distribution of different lithology categories.

[0078] This function reflects the trend that the correlation between indicator variables decreases with the increase of spatial distance : when tends to 0, is approximately 0, indicating that the sample points are highly correlated; when is large enough, tends to the sill value ( ), indicating that there is no correlation between sample points.

[0079] Parameter fitting can use the least squares method or weighted least squares method to make the theoretical curve fit the empirical semivariogram points as much as possible. In tunnel geology applications, the correlation length is closely related to the geological structure characteristics: if the stratum has strong continuity along the line, then is large; if the vertical variation is fast, then are relatively small. Among them, , are the relevant lengths in different directions (such as the relevant lengths in the direction along the tunnel bottom layer and the vertical direction, are the equivalent isotropic correlation lengths obtained by anisotropic conversion based on , and other directional parameters, which are used to uniformly describe the spatial correlation characteristics in different directions. The correlation parameters in different directions provide a basis for subsequent anisotropic modeling.

[0080] (3) Anisotropic characteristics and their geological significance:

[0081] Tunnel surrounding rock usually presents obvious spatial anisotropic characteristics: on the one hand, the stratum continuity is strong along the tunnel axis direction, and the lithology changes relatively gently; on the other hand, the lithology is obviously layered in the vertical direction, and the layer boundary changes steeply. If the isotropic assumption is adopted, it is easy to lead to over-smoothing in the vertical direction or discontinuity along the line direction, which cannot truly reflect the stratum structure.

[0082] Therefore, in some embodiments, a geometric anisotropic modeling method is introduced. By performing linear transformation on the coordinate system, the spatial distances in different directions are scaled in proportion, so that the model remains isotropic in mathematics, while embodying the directional differences in physical space. Let the original coordinates be , and the transformed coordinates be:

[0083] ;

[0084] wherein and are the scaling coefficients of the horizontal (along the line direction) and vertical (depth direction) respectively. If , it means that the horizontal correlation length is larger, and the model is smoother along the line direction; on the contrary, if , it reflects the geological structure that the vertical layering is significant.

[0085] This method can effectively depict the spatial anisotropic characteristics of the tunnel surrounding rock, and balance the model between the layer boundary transition and local changes.

[0086] The optimal values of and are determined by parameter scanning and model accuracy verification within a predetermined range, and the average negative log-likelihood (NLL) and overall accuracy (OA) can be used as evaluation indexes for the model accuracy verification. By introducing an anisotropic variogram fitting strategy, the spatial structure characteristics of the tunnel, which are strong in continuity along the line direction and significant in vertical differentiation, are accurately depicted, providing core parameter support for the construction and dynamic updating of the tunnel geological model.

[0087] S3. Initial geological model construction: Determine the neighborhood radius based on the variation parameters; estimate the initial geological model using the indicator kriging method. The geological model includes a set of conditional probability distributions for each spatial location within the tunnel profile belonging to various lithologies; wherein, for any predicted point within the tunnel profile, the lithology category indicator variable for the predicted point is obtained by weighting the lithology category indicator variables of each sample point in the neighborhood of the predicted point in the standardized dataset; and obtain the conditional probability of the predicted point belonging to various lithologies based on the lithology category indicator variable of the predicted point.

[0088] The following is an explanation of the instructional Kriging framework.

[0089] After obtaining the indicator variable field, it is necessary to establish its spatial correlation model and perform probability estimation. The Indicator Kriging (IK) method uses a semivariogram function to characterize the spatial variability of geological properties, and on this basis, performs linear optimal estimation of the indicator variables, thereby achieving probabilistic prediction of each lithology category at unobserved locations. This method takes into account both geological continuity and uncertainty representation, and is suitable for tunnel geological environments with sparse drilling and complex stratigraphic variations.

[0090] (1) Spatial estimation principle:

[0091] For any predicted point (unknown location) The corresponding lithological category indicator variable estimate is expressed as follows:

[0092] ;

[0093] in, This indicates a prediction point. Is the lithology at this location the first? Class indicator variables; For prediction points Within the neighborhood sample points Is the lithology category the first? Class indicator variables, For prediction points Within the neighborhood sample points Kriging weights, For prediction points The number of sample points in the neighborhood. The value of can be determined by the neighborhood search strategy: (1) Fixed neighborhood radius: Within a given spatial range (e.g., with the prediction point as the center and radius R), all sample points falling within this range are taken as neighborhood points. =Equal to the number of sample points within the spatial range. (2) Fixed number of neighborhood points: Select no more than the preset upper limit from the known samples closest to the predicted location (e.g., participate in the estimation.

[0094] The above formula is actually a weighted average of the indicator variables of the sample points (known points), and the weights are determined by the Kriging system that satisfies the unbiasedness and minimum variance conditions:

[0095] ;

[0096] In the formula, μ is the Lagrange multiplier, μ is an unknown quantity, and the Kriging weight is solved by a linear equation system, which ensures that the weight satisfies the unbiased constraint; is the semi-variance between the sample points and , and the calculation formula is:

[0097] .

[0098] By solving the linear system, the optimal value of the weight vector can be obtained, and then the .

[0099] For a binary indicator variable, the expected value of the prediction can be explained as the lithology class probability:

[0100] ;

[0101] In the formula, represents the estimated probability that the lithology class of the prediction point is the th class.

[0102] This probability directly reflects the spatial distribution trend of the geological body and the prediction confidence.

[0103] (2) Probability normalization and classification output:

[0104] Since each lithology class is independently estimated, the sum of its probabilities may not be strictly equal to 1, so normalization processing is required:

[0105] ;

[0106] In the formula, represents the normalized probability that the lithology class of the prediction point is the th class.

[0107] The normalized probability vector has complete probability meaning and can be used for subsequent classification and uncertainty analysis.

[0108] In tunnel construction, the maximum a posteriori criterion (MAP) is used to determine the most likely lithology class for easy geological interpretation and risk assessment:

[0109] ;

[0110] wherein, i.e. the prediction point the predicted most probable lithology class.

[0111] The hard classification result, i.e. the most probable stratigraphic distribution map, is obtained by the above steps.

[0112] (3) Calculation of the implementation and neighborhood constraints:

[0113] In order to take into account the calculation efficiency and geological rationality, IK modeling generally adopts a local neighborhood search strategy. For each prediction point , only sample points within a certain search radius are selected to participate in the estimation, in order to meet the "local correlation" assumption of the geological attribute. In some embodiments, the neighborhood radius is in the range of ~ , where is the correlation length (horizontal variation range) of the fitted variogram; this range can ensure that the number of neighborhood sample points is sufficient, the spatial correlation structure is reasonably expressed, and the calculation efficiency of IK in the dynamic updating process is guaranteed.

[0114] In some embodiments, a KD-Tree-based neighborhood search algorithm is used to achieve fast sample point retrieval, and the prediction grid is divided into several blocks to achieve batch calculation, thereby ensuring that the model can be quickly reconstructed in the dynamic updating process.

[0115] This localized calculation method not only reduces the dimension of the Kriging matrix, but also enables real-time correction and local updating of the geological model along the line as the tunnel advances.

[0116] (4) Characteristics and applicability of the method:

[0117] The indicator Kriging framework has characteristics such as probabilistic expression, strong geological continuity, high data fusion capability, and strong result interpretability. The output result is a continuous lithology class probability distribution, which can not only be used for stratigraphic identification, but also quantitatively characterize the prediction uncertainty; by describing the spatial correlation through the semi-variogram, the model can maintain the reasonable transition and continuity of the strata; at the data level, the IK method can simultaneously use drilling data and excavation exposure information as sample constraints, and is naturally suitable for dynamic updating modeling of tunnel geology; at the same time, the model result can be intuitively interpreted through probability maps, maximum probability values, and information entropy, etc., providing a reliable basis for geological risk identification and construction decision-making. In summary, the indicator Kriging not only enables reliable spatial interpolation and prediction of the stratigraphic structure during the tunnel construction phase, but also provides a mathematical foundation and probabilistic expression framework for subsequent dynamic updating mechanisms.

[0118] S4, dynamically updating the geological model: in the tunneling process, based on the stratum information continuously acquired by the data acquisition and preprocessing step of the tunnel face, the newly exposed stratum information is incrementally integrated into the standardized data set to form an updated data set; based on the updated data set, the variation parameters in the variation function are adaptively adjusted, and then based on the adjusted variation parameters and the updated data set, the initial geological model is updated by the indicator Kriging method to obtain the updated geological model.

[0119] Tunnel construction has obvious space-time dynamic characteristics. With the continuous advancement of tunneling, new stratum information is continuously exposed, and the prediction accuracy and reliability of the original geological model will change. If the model cannot timely absorb this new information, the prediction result will lag behind the actual geological conditions. Therefore, based on the indicator Kriging (IK) framework, a dynamic updating mechanism for the geological model of tunnel construction is constructed to realize the rolling correction and adaptive evolution of the geological model.

[0120] (1) Overall idea:

[0121] The basic idea of the dynamic updating mechanism is to regard the geological modeling process as an iterative cycle of "prior modeling-new data fusion-posterior correction". In the early stage of construction, the model takes drilling data as the main input to establish an initial geological model; with the tunneling, the newly exposed surrounding rock information is gradually collected and input into the model to correct the spatial structure parameters and stratum probability distribution, thereby forming a new prediction result.

[0122] This process can be represented as:

[0123] ;

[0124] Wherein, represents the geological model of stage .

[0125] The newly added data set exposed by tunnel excavation is the "new data" introduced by dynamic updating, which actually contains multiple types of observations. To avoid symbol confusion, it must be unified into a set. For some positions, "spatial coordinates + lithology category".

[0126] is an update operator, and the update process can be represented as: (a) merge the new and old sample sets and ; (b) recalculate the local empirical semivariance; (c) fit to obtain a new set of variation parameters; (d) update the variation parameters using a sliding weighting method; (e) perform indicator Kriging estimation under the new variation parameters to obtain the updated probability field; finally form stage Complete geological model This allows the model to continuously evolve as the tunneling process progresses.

[0127] This iteration is performed at each excavation step length, enabling the continuous evolution of the geological model.

[0128] (2) Components of the update mechanism

[0129] The dynamic update mechanism proposed in this application consists of four core steps, which together achieve the phased correction and rolling evolution of the geological model. The logic of the entire update process is as follows: incremental data update → adaptive adjustment of variogram → geological model update → uncertainty quantification and evaluation.

[0130] (2.1) Incremental data update steps:

[0131] During the tunneling process, the tunnel face continuously reveals new lithological information, forming an incremental dataset. These data, after coordinate registration, lithology coding, and gridding, are compared with existing drilling and previous exposure data. Together they serve as input for the next stage of model updates.

[0132] Data flow:

[0133] ;

[0134] in: For the stage The samples that have already been used for modeling include borehole data and historical exposure data; For the stage → Newly revealed lithology and interface location data during the tunneling process.

[0135] The two, when combined, constitute a complete sample set for the new phase.

[0136] (2.2) Adaptive adjustment steps of the variation function:

[0137] During tunnel excavation, new geological data are continuously revealed, and local geological structures may change. If fixed variation parameters are continued to be used, the model will be unable to reflect these dynamic characteristics.

[0138] Therefore, this application introduces an adaptive update strategy for the variation function during tunnel excavation.

[0139] That is, at each tunneling stage, the local empirical semivariogram is recalculated based on the newly added sample points. This semivariogram reflects the local spatial correlation structure of lithological indicator variables under the constraint of the new samples. Subsequently, a new set of variation parameters is obtained through model fitting. and update the model parameters in a sliding weighted form (avoiding drastic fluctuations in parameters):

[0140]

[0141] wherein, denotes the set of variation parameters of the previous stage, i.e., stage , , , , is the set of variation parameters obtained by re-fitting based on the new data (new sample points); is the update weight coefficient, which is an empirical parameter and can be determined by comprehensively considering the stability of the model and the response ability to new information. In some embodiments, the value range is 0.2-0.3 based on the experience of geostatistics and numerical test verification, for example, = 0.2. This method can gradually reflect the dynamic evolution of geological information while maintaining the stability of the model.

[0142] (2.3) Geological model updating step:

[0143] Under the new variation parameters , the probability estimation is performed for each lithology class using the indicator Kriging to obtain the multi-lithology conditional probability field of stage :

[0144]

[0145] wherein: the latest sample set, including old data and newly exposed data; is the conditional probability of the lithology class being the th class at the position of stage .

[0146] (2.4) Uncertainty quantification and evaluation step:

[0147] In some embodiments, this step further includes: quantifying the uncertainty of the updated geological model.

[0148] The goal of tunnel geological modeling is not only to obtain the most likely stratum distribution, but also to quantitatively describe the credibility and uncertainty range of the prediction results. In the dynamic environment of gradually exposed construction information, the uncertainty of the geological model presents the characteristics of evolution over time, so it is necessary to establish a systematic uncertainty quantification and precision evaluation framework to evaluate the model performance and guide subsequent updating and optimization.

[0149] (2.4.1) Uncertainty quantification method:

[0150] ​​Based on the modeling results of the indicator Kriging, the conditional probability distribution of each lithology class at any location can be obtained:

[0151]

[0152] where, is the probability that the lithology class at location is the th class. This probability is derived from the updated geological model. This probability distribution not only provides the basis for determining the most likely lithology type, but also can be used to measure the uncertainty of the prediction results.

[0153] To quantify this uncertainty, the information entropy index is introduced:

[0154]

[0155] where, denotes the information entropy of point (location) .

[0156] When is smaller, the prediction result is more certain (a certain lithology class probability is significantly dominant); when is larger, it indicates that the probabilities of different lithology classes in this area are close, the model reliability is low, and the uncertainty of lithology determination at this location is higher. By calculating the entropy value distribution of the entire profile grid, an "uncertainty map" can be obtained, which intuitively displays the high confidence area and potential risk area of the model prediction.

[0157] After the geological model is updated, the information entropy of the new geological model is calculated:

[0158] and the uncertainty change compared with the previous stage is compared:

[0159]

[0160] Meaning: : Uncertainty is reduced, and the update is effective; : New data does not significantly improve model cognition; : The model may have overfitting or disturbance, and attention should be paid to the stability of the model.

[0161] (2.4.2) Model accuracy evaluation index:

[0162] To objectively evaluate the accuracy of the model prediction, this application uses a variety of statistical indicators for comprehensive analysis. Assuming that the predicted result and the actual revealed stratum type are and , the evaluation index of the overall accuracy can be defined as:

[0163] ;​​​

[0164] wherein, is the total number of sample points for validation (such as the total number of lithology observation points in the tunnel profile for validation), is the sample point the most likely lithology class of prediction, is the sample point the actual lithology class (true value) of, is an indicator function, which is 1 when ; otherwise, 0.

[0165] S5, generating a lithology distribution map: based on the updated geological model, determining the most likely lithology class of each spatial position, and generating a lithology distribution map.

[0166] The final output of the present application can include: an updated probability field, a MAP classification map, and an information entropy map.

[0167] In summary, the dynamic updating mechanism established by the present application realizes the transformation of the geological model from static cognition to dynamic evolution, and provides a theoretical basis and engineering application support for geological information fusion, model adaptive updating and risk control in tunnel construction.

[0168] The present application is aimed at the engineering requirements of complex geological conditions, gradual data disclosure, and dynamic response of the model in tunnel construction. The present application proposes a dynamic updating modeling framework for tunnel geology based on indicator Kriging (DIK) that integrates drilling data and construction disclosure data, realizes the rolling correction and evolution of the geological model, introduces anisotropic spatial structure modeling and uncertainty quantification mechanism in the DIK method, and makes the model more consistent with the spatial characteristics of tunnel geology and the risk identification requirements. The present application provides a new idea and technical approach for geological information integration, risk prediction and intelligent decision-making in tunnel construction, and has important theoretical and engineering significance for promoting the digitization, intelligentization and dynamic management of geological information in tunnel engineering.

[0169] To verify the effectiveness of the present application method (DIK method) in real engineering scenarios, a multi-lithology two-dimensional model is constructed based on the profile of Zhaibei Mountain Tunnel, and the disclosure process under different tunneling stages is carried out. The effectiveness of the method in dynamic information fusion, prediction accuracy improvement and risk identification is verified through the engineering case.

[0170] (1) Engineering background and geological conditions:

[0171] The present application selects a section of Zhaibei Mountain Tunnel located in Meizhou City, Guangdong Province as an engineering case. Based on the intensive drilling and geological reconnaissance data before construction, experienced geologists give a two-dimensional stratigraphic interpretation profile of this section as shown in Figure 3The profile shows that the section mainly develops four types of strata, from top to bottom, they are tuffaceous sandstone (Ts), argillaceous sandstone (As), quartz sandstone (Qs), and metamorphic sandstone (Ms). The strata are distributed in layers from bottom to top, and there is a local fluctuation and thickness mutation at about 700-800m, indicating the existence of tectonic disturbance. This application regards the interpreted profile as the "true value field" and carries out dynamic modeling driven by tunnel exposure.

[0172] To perform initial constraints on the model, 5 boreholes are uniformly distributed within the profile range, as shown in Figure 4 The positions correspond to 0m, 250m, 500m, 750m and 1000m of the total length of the profile, and penetrate the full depth range. The borehole data includes the interface depth and lithology code information of each lithology layer, which is used to establish the initial geological model and as the benchmark input for subsequent dynamic updating. This drilling method can not only reflect the typical survey density, but also provide clear prior constraint conditions in model verification.

[0173] The tunnel excavation line is arranged about 45m below the profile, and the excavation profile of the tunnel body can be approximately regarded as an elliptical section with a height of 12m and a width of 14m. For ease of display, the tunnel excavation line is marked in red solid line in Figure 3 , clearly showing its spatial position in the strata and its relationship with different lithology interfaces. This line will be the main source of dynamic exposure data in the subsequent stage, used to simulate the gradual accumulation of face exposure information and the model correction process during construction advancement.

[0174] (2) Research methods and experimental design:

[0175] During the construction process, new geological information is continuously obtained through face geological logging and related test detection. These new data cover key elements such as lithology category, which can reflect the dynamic changes of surrounding rock properties with the excavation process. Exposure data is gradually accumulated according to the construction step length, which is used as the stage input for dynamic correction and updating of the model.

[0176] In the data processing stage, all geological information is processed through spatial coordinate registration, attribute standardization and regular gridding to ensure the spatial consistency and numerical comparability between different source data. Finally, all samples are mapped to the tunnel profile coordinate system (s-z plane) to form a standardized two-dimensional data set suitable for indicator kriging modeling and dynamic prediction analysis, providing consistent coordinates and attribute basis for subsequent spatial statistical modeling and result evaluation.

[0177] (3) Experimental results:

[0178] To verify the effectiveness of the proposed dynamic update mechanism in tunnel geological prediction, this application designed a dynamic modeling experimental scheme based on the tunnel stage excavation ratio (each stage being 20%). During the simulated construction process, the adaptability and prediction accuracy evolution of the model were systematically evaluated. Through a five-stage (initial stage to fifth stage) dynamic modeling process, this application systematically simulated the entire process of model updating during tunnel construction, providing a repeatable experimental framework for evaluating the dynamic adaptability of the indicator kriging method under actual complex geological conditions.

[0179] (3.1) Initial stage modeling results:

[0180] To analyze the modeling performance of the Dynamic Indicator Kriging (DIK) method under initial conditions, this section takes the initial stage as an example, using borehole data ( Figure 4 An initial geological model M0 is constructed, and the model's parameter optimization process, probability distribution characteristics of each lithology, and spatial uncertainty are systematically displayed.

[0181] (3.1.1) Selection of optimal values ​​for anisotropic scaling parameters:

[0182] Indicator S of anisotropic scaling parameter in Kriging model s (Horizontal direction, i.e., the front) ) and S z (Vertical, i.e., the front) This is used to control the scaling ratio of spatially related structures in different directions, thereby affecting the model's depiction of bedding extension and vertical differentiation. To determine the optimal spatial scale relationship, this application fixes S. z Under the condition that = 1.0, for S s The parameters are scanned within the range of [1,12], and the mean negative log-likelihood (NLL) and overall accuracy (OA) are used as comprehensive evaluation indicators.

[0183] like Figure 5 As shown, with S s Increase, NLL in S s S remains at a low level within the range of ≤6. s The minimum value of 0.163 is obtained when S = 3. s >6 rises rapidly; OA, on the other hand, rises rapidly after S. s The peak value of 0.94 is reached at point 3. When the horizontal correlation scale is too large, the model bedding is over-smoothed, and the vertical resolution is reduced. Considering both indicators, the optimal parameter is determined to be S. s =3、S z =1.0, at which point NLL=0.163 and OA=0.940. This parameter combination can effectively reflect the vertical differences between rock strata while maintaining the continuity of bedding.

[0184] (3.1.2) Variation function fitting curves for the four lithological categories:

[0185] In determining S s =3、S z After setting the value to 1.0, the experimental semivariogram function of the indicator variable was calculated for each of the four lithological categories, and an exponential model was used for fitting. The results are as follows: Figure 6 As shown, the blue scatter plots represent experimental values, and the orange curve represents the fitting results. The semivariogram of each lithology gradually increases and then stabilizes with increasing lag distance, indicating that different lithologies exhibit significant spatial autocorrelation within a finite spatial scale. This result verifies the rationality of the DIK method in identifying lithological spatial structure and provides a reliable basis for adjusting the variability parameters in the subsequent dynamic update stage.

[0186] (3.1.3) Probability field distribution of each lithological category:

[0187] With the support of borehole constraints and variation models, probabilistic predictions were made for the two-dimensional mesh in the study area, such as... Figure 7 As shown, different colors represent the probability fields of Ts, As, Qs, and Ms. Overall, the probability distribution of each lithology exhibits a distinct layered structure, consistent with the spatial extension direction of the stratigraphy. High-probability areas match the lithology revealed by boreholes, while low-probability areas present a smooth transitional "fuzzy band," reflecting the model's spatial quantitative expression of geological uncertainties. This result indicates that, under the constraint of sparse boreholes, the DIK model in the initial stage can still reasonably reconstruct the main stratigraphic pattern, providing a basic reference for subsequent dynamic updates.

[0188] (3.1.4) Maximum a posteriori lithology distribution (MAP):

[0189] Based on various geological models, a maximum a posteriori (MAP) lithology distribution map is generated by selecting the lithology category with the highest probability for each grid cell, such as... Figure 8 As shown, the predicted profile is highly consistent with the true field morphology, with a clear and continuous bedding structure. The overall model accuracy OA=94.07%, indicating that the DIK method can achieve high initial modeling accuracy under the condition of relying solely on borehole constraints.

[0190] (3.1.5) Uncertainty distribution (information entropy diagram):

[0191] The information entropy of each grid cell was calculated based on the geological models of each region, and the information entropy distribution map was plotted, such as... Figure 9The information entropy reflects the degree of uncertainty of the model in determining the lithology at this location. The higher the value, the greater the uncertainty. The results show that the entropy value is lowest in the stable drilling and bedding area, and higher in the complex structure or data scarce area. This spatial distribution is highly consistent with the geological structure characteristics, revealing the sensitive response of the DIK model to data density and stratigraphic complexity, providing a quantitative reference for the subsequent dynamic exposure stage model evolution.

[0192] (3.2) Dynamic updating stage modeling results:

[0193] Based on the initial stage model M0, the updating strategy of every 20% excavation is adopted, and the exposure data in the interval of 0-20%, …, 0-100% is introduced in turn to form M1-M5 models. Each stage is optimized by cross-validation to update the anisotropy parameters, and the prediction field is updated accordingly.

[0194] (3.2.1) Time sequence extension of drilling and exposure information:

[0195] Figure 10 The known information distribution of each stage from the first stage to the fifth stage (stage1~stage5) is shown, including the fixed drilling position and the gradually accumulated tunnel exposure zone. With the advancement of excavation, the exposure data extends along the tunnel direction in 10% steps, forming a continuous known zone covering the entire tunnel section. By the fifth stage, the entire section has been exposed, and the model spatial constraint has been significantly enhanced. This process simulates the dynamic mechanism of the advancing of the working face and the exposure data-driven model correction in actual construction, providing a realistic reference for analyzing the model evolution law.

[0196] Figure 11 The MAP lithology distribution maps of each stage from the first stage to the fifth stage (stage1~stage5) are given. The overall trend shows that as the exposure ratio increases, the model structure gradually stabilizes. The areas near the drilling and exposure zones stabilize first, while the boundary areas far from the data sources lag behind in convergence; the structure disturbance zone still shows obvious adjustment in the middle stage, but eventually tends to be smooth and highly consistent with the geological interpretation. This evolution law reflects the adaptive characteristics of the DIK model: as the data accumulates, the model gradually transitions from prior estimation to posterior correction, achieving dynamic approximation of complex geological structures.

[0197] Figure 12The information entropy distribution of the first stage to the fifth stage is shown. The overall trend is that the information entropy value continues to decrease with the increase of the exposure ratio, from 0.185 at the beginning to 0.097, indicating that the uncertainty of the model cognition converges significantly. Spatially, the high-entropy area is mainly concentrated in the unexposed or complex geological structure area, while the uncertainty of the exposed section and the surrounding of the borehole decreases rapidly. This spatial law of uncertainty convergence reflects the self-learning process of the model: as the exposure data continues to be supplemented, the model's understanding of the stratigraphic structure expands from local to global, and the uncertainty gradually decreases from significant to controllable, finally forming a stable and reliable geological model.

[0198] Table 1 lists the modeling accuracy indicators of the initial stage to the fifth stage. OA increases from 94.07% to 97.40%, and the average information entropy decreases from 0.196 to 0.093, showing a significant monotonic convergence trend.

[0199] Table 1 Modeling accuracy indicators of the initial stage to the fifth stage

[0200] ;

[0201] Overall, the model performance improves steadily with the advancement of exposure, the prediction accuracy and the consistency of geological cognition increase synchronously, and the uncertainty decreases significantly. This indicates that the dynamic indicator kriging method can effectively integrate the temporal exposure information, realize the continuous optimization and reliable convergence of the geological model, and provide a feasible technical path for intelligent geological perception and dynamic risk assessment in tunnel construction.

[0202] The results show that by introducing anisotropic variation structure, indicator variable probability field, maximum a posteriori (MAP) prediction and information entropy uncertainty analysis, the model can establish a high-reliability initial geological model (initial stage) under the condition of only sparse boreholes, with an overall accuracy of 94.07%, indicating that the model can accurately depict the main stratigraphic structure and reasonably quantify the ambiguity of lithological boundaries in the initial stage. The DIK method can accurately reconstruct the main surrounding rock pattern under the constraint of sparse drilling; as the exposure data is gradually added in stages, the model prediction accuracy improves significantly, and the layer boundary identification becomes clearer.

[0203] As the exposure data is gradually added in 20% steps (first stage to fifth stage), the dynamic updating mechanism significantly improves the model cognition ability. The MAP prediction results show that the lithological interface gradually becomes clear from fuzzy, the structural disturbance area is automatically corrected, and the uncertainty area gradually converges with data accumulation; the average information entropy decreases from 0.196 to 0.093, and the model accuracy also improves steadily from 94.07% to 97.40%. This process reveals the self-learning characteristics of the DIK model, indicating that the method is suitable for dynamic geological prediction needs in real construction scenarios.

[0204] Embodiment Two

[0205] The embodiment provides an electronic device, comprising a memory and a processor.

[0206] The memory is configured to store a computer program.

[0207] The processor is configured to invoke the computer program to execute the method in Embodiment One.

[0208] Embodiment Three

[0209] The embodiment provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to enable an electronic device to implement the method in Embodiment One when the computer program is run on the electronic device.

[0210] Embodiment Four

[0211] The embodiment provides a computer program product, comprising a computer program, and the computer program is configured to enable an electronic device to implement the method in Embodiment One when the computer program is run on the electronic device.

[0212] The specific implementation manners of the system, the electronic device, the computer-readable storage medium and the computer program product provided in the embodiment of the present application can refer to the specific embodiments of the above method, which will not be repeated here.

[0213] Obviously, those skilled in the art should understand that each unit or each step of the above application can be implemented by a general computing device, and they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device and executed by a computing device, or they can be made into each integrated circuit step, or multiple steps or steps among them can be made into a single integrated circuit step. Thus, the present application is not limited to any specific combination of hardware and software.

[0214] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for modeling and dynamically updating tunnel geological information, characterized in that, include: Data Acquisition and Preprocessing: Collect drilling data before construction and stratigraphic information exposed at the working face during construction. Standardize the drilling data and exposed information to obtain a standardized dataset. Based on the drilling data in the standardized dataset, define a set of lithological categories and transform the lithological category variables into indicator variables. Variation function fitting: Based on the standardized dataset, the empirical semivariogram of the indicator variable of the sample points is calculated, and the variation function is used to fit the empirical semivariogram to obtain the variation parameters; Initial geological model construction: The neighborhood radius is determined based on the variation parameters; the initial geological model is estimated using the indicator kriging method. The geological model includes a set of conditional probability distributions for each spatial location within the tunnel profile belonging to various lithologies; for any predicted point within the tunnel profile, the lithology category indicator variable for the predicted point is obtained by weighting the lithology category indicator variables of each sample point in the neighborhood of the predicted point in the standardized dataset; the conditional probability of the predicted point belonging to each lithology is obtained based on the lithology category indicator variable of the predicted point. Dynamically update the geological model: During tunnel excavation, based on the stratigraphic information continuously acquired from the data acquisition and preprocessing steps, the newly revealed stratigraphic information is incrementally integrated into the standardized dataset to form an updated dataset. Based on the updated dataset, the variogram parameters in the variogram function are adaptively adjusted. Then, based on the adjusted variogram parameters and the updated dataset, the initial geological model is updated using the indicator kriging method to obtain the updated geological model. Generating lithology distribution maps: Based on the updated geological model, the most likely lithology category for each spatial location is determined, and a lithology distribution map is generated.

2. The method according to claim 1, characterized in that, For any prediction point The corresponding lithology category indicator variable estimate is calculated using the following formula: ； in, This indicates a prediction point. Is the lithology category the first? Class indicator variables; For prediction points Within the neighborhood sample points Is the lithology category the first? Class indicator variables, For prediction points Within the neighborhood sample points The Kriging weights are determined by a Kriging system that satisfies the conditions of unbiasedness and minimum variance. For prediction points The number of sample points in the neighborhood; The formula for calculating the probability of lithology category is: ； In the formula, Indicates the prediction point The lithological category is the first The probability estimate of the class; The probabilities of independently estimated values ​​for each lithological category are normalized using the following formula: In the formula, Indicates the prediction point The lithological category is the first The normalized probability of each class; a complete geological model is formed by the normalized probability of each prediction point belonging to each lithological class.

3. The method according to claim 2, characterized in that, The neighborhood radius The range of values ​​is ~ ,in The correlation length is the length of the fitted variation function.

4. The method according to claim 1, characterized in that, The formula for calculating the empirical semivariance of the indicator variable for the sample points is: ,in, For empirical semivariance, and Sample points and Is the lithology category the first? Class indicator variables; Represents all sample point pairs A set whose interval lies within the interval [ ]. Inside; The number of sample points is denoted by ; where The lag distance, The lag width is used to determine the set of valid point pairs. ; The variation function is an exponential variation function, and its expression is: ,in Value of a nugget. For sill values, For the relevant length; , For stilt value and Together they constitute the set of variation parameters.

5. The method according to claim 1, characterized in that, In the step of dynamically updating the geological model, the variation parameters are adjusted using a sliding weighted method, and the update formula is as follows: ,in This indicates the previous stage, i.e., the phase. The set of variation parameters, This is the set of variation parameters obtained by refitting based on the new data. To update the weighting coefficients.

6. The method according to claim 1, characterized in that, The step of dynamically updating the geological model further includes: quantifying the uncertainty of the updated geological model, wherein uncertainty quantification refers to characterizing the degree of ambiguity in lithology determination at each spatial location through quantitative indicators; wherein uncertainty quantification is achieved by calculating information entropy, and the formula for calculating information entropy is: ,in, Point Information entropy For point The lithological category is the first The probability of a class The larger the value, the more likely it is to be a point. The higher the uncertainty in lithological determination.

7. The method according to claim 1, characterized in that, The dynamic updating geological model step also includes: model accuracy evaluation; the overall accuracy evaluation index OA is calculated using the following formula: ,in, The total number of sample points used for validation. For sample points The most likely lithology predicted For sample points The actual lithological category, For indicator functions, when When the condition is met, the function value is 1; otherwise, it is 0.

8. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer programs; The processor is configured to invoke the computer program to perform the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed on an electronic device, causes the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is run on an electronic device, it causes the electronic device to perform the method as described in any one of claims 1 to 7.

Images