A tunnel water inflow prediction method based on horizontal drilling while drilling testing

Abstract

The application relates to the technical field of hydrogeological prediction, and discloses a tunnel water inflow prediction method based on horizontal drilling while drilling testing, which comprises the following steps: collecting multi-source while-drilling testing data of a horizontal advanced drill hole in real time to generate an original while-drilling testing data set; pre-processing and feature extraction are performed on the original while-drilling testing data set to generate a multi-dimensional drilling feature parameter set; based on the multi-dimensional drilling feature parameter set, the positions of geological interfaces and rock mass integrity changes are identified to generate geological interface identification data; through real-time collection of multi-source while-drilling testing data in the construction process of the horizontal advanced drill hole and synchronous collection of while-drilling geophysical exploration data, in-situ geological and hydrological information continuously distributed along the drill hole track is obtained, the rock mass characteristics and water-rich structures of unexcavated areas can be described, and therefore the continuity and reliability of a geological and hydrological structure model are improved, and a data foundation is laid for water inflow prediction.

Description

Technical Field

[0001] This invention relates to the field of hydrogeological prediction technology, specifically to a method for predicting tunnel water inflow based on horizontal drilling while drilling testing. Background Technology

[0002] As long and narrow underground structures, tunnels inevitably traverse different hydrogeological and engineering geological environments during construction. When a tunnel passes through a water-bearing section of rock and soil, the construction disrupts the original groundwater seepage conditions, turning the tunnel body into an underground corridor through which groundwater discharges outward in various forms such as seepage, dripping, streams, and large-scale water inrush, resulting in water inrush disasters.

[0003] Currently, in the early stages of tunnel construction, hydrogeological investigation and water inflow prediction mainly rely on discrete point-based survey data such as ground drilling and geophysical exploration. This makes it difficult to depict the continuous geological and hydrological structure of the unexcavated area in front of the tunnel face, resulting in discrepancies between the prediction model and the actual hydrogeological conditions. Consequently, the reliability of the water inflow prediction results is insufficient, and it cannot guide the design and construction decisions for drainage and flood control.

[0004] Therefore, a method for predicting tunnel water inflow based on horizontal drilling while drilling tests is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for predicting tunnel water inflow based on horizontal drilling while drilling tests, which solves the problem mentioned in the background technology that the water inflow prediction results are not reliable enough and cannot guide drainage design and construction decisions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting tunnel water inflow based on horizontal drilling while drilling testing, the method comprising the following steps: S1. Real-time acquisition of multi-source drilling test data from horizontal advanced boreholes to generate raw drilling test dataset; S2. Preprocess and extract features from the original drilling test dataset to generate a multi-dimensional drilling feature parameter set; S3. Based on the multi-dimensional drilling feature parameter set, identify the geological interface and the location of changes in rock mass integrity, and generate geological interface identification data. S4. Simultaneously collect and process geophysical exploration data while drilling in the borehole to generate spatial positioning data of water-rich structures. S5. Integrate the geological interface identification data with the spatial positioning data of the water-rich structure to construct a three-dimensional geological and hydrological generalized model of the area in front of the tunnel. S6. Based on the three-dimensional geological and hydrological generalization model and the multi-dimensional drilling characteristic parameter set, and using the hydrogeological numerical simulation method, dynamically invert the hydrogeological parameters of the water-bearing rock mass to generate a dynamic hydrogeological parameter set. S7. Based on the geological interface identification data and the spatial positioning data of the water-rich structure, perform water-rich classification and risk segment division, and generate segmented water inflow prediction interval data. S8. Based on the three-dimensional geological and hydrological generalization model, the dynamic hydrogeological parameter set, and the preset tunnel excavation disturbance conditions, calculate the predicted water inflow value in each segmented water inflow prediction interval, and generate segmented quantitative water inflow prediction data. S9. Integrate the segmented quantitative water inflow prediction data, the spatial positioning data of the water-rich structure, and related geological risk information to generate a dynamic prediction report of water inflow in front of the tunnel. S10. Based on the actual hydrogeological conditions and water inflow revealed during tunnel excavation, the three-dimensional geological and hydrological generalization model, the dynamic hydrogeological parameter set, and the water inflow prediction model are updated and adaptively learned.

[0007] Preferably, the real-time acquisition of multi-source drilling test data from the horizontal advanced borehole in step S1 includes the following steps: S11. During the horizontal advance drilling process of the tunnel boring machine, the drilling rig engineering parameters and borehole annulus parameters corresponding to the drilling depth are collected in real time and synchronously. The drilling rig engineering parameters include drill bit rotation speed, drilling pressure, drilling speed, torque and drilling rig vibration acceleration, and the borehole annulus parameters include borehole mud inlet pressure, outlet pressure, inlet flow rate, outlet flow rate and temperature. S12. Align and package the real-time collected drilling rig engineering parameters and borehole annulus parameters according to a unified timestamp and borehole depth coordinates to generate the original drilling test dataset with spatiotemporal tags.

[0008] Preferably, the preprocessing and feature extraction in S2 includes the following steps: S21. Perform data cleaning on the original drilling test dataset. The data cleaning includes removing abnormal jump values, smoothing filtering, and linear interpolation filling for missing signal segments. S22. For the drilling test data after cleaning, extract the characteristic parameters representing the mechanical and permeability properties of the rock mass from the time domain, frequency domain, and time-frequency domain respectively, according to a fixed drilling footage. Among them, the time-domain feature parameters include the mean, variance, peak-to-peak value and rate of change within each parameter window; the frequency-domain feature parameters include the main frequency amplitude and energy proportion of a specific frequency band obtained by fast Fourier transform; and the time-frequency domain feature parameters include the energy spectrum characteristics obtained by wavelet transform. S23. After normalizing the extracted time-domain, frequency-domain, and time-frequency-domain feature parameters, combine them according to the borehole depth sequence to generate the multi-dimensional drilling feature parameter set.

[0009] Preferably, identifying the location of geological interfaces and changes in rock mass integrity in step S3 includes the following steps: S31. Select a subset of characteristic parameters that are sensitive to lithological changes and fracture development from the multi-dimensional drilling characteristic parameter set, and use it as a key discrimination index set for geological interface identification. S32. Using a sliding window-based mutation point detection algorithm, the change sequence of the key discrimination index set along the borehole depth is analyzed to identify the depth points where the characteristic parameters change, including step changes, trend reversals, and abnormal peaks. S33. The identified feature change depth points are initially determined as potential geological interface locations, and the geological interface identification data containing the location and change intensity is generated.

[0010] Preferably, the generation of water-rich structural spatial positioning data in step S4 includes the following steps: S41. During the horizontal advance drilling process, the borehole transient electromagnetic instrument is used simultaneously to continuously collect geophysical signals reflecting the electrical differences of the surrounding rock mass along the borehole depth direction to generate raw geophysical data while drilling. S42. Perform inversion interpretation processing on the original geophysical data while drilling, identify the low resistivity anomaly areas, and interpret these anomaly areas as water-rich fracture zones. S43. The location, geometry, size and relative spatial relationship of the water-rich anomaly are digitally characterized to generate the spatial positioning data of the water-rich structure, including the anomaly's starting depth, ending depth, center position and anomaly intensity.

[0011] Preferably, the construction of the three-dimensional geological and hydrological generalized model of the area in front of the tunnel in step S5 includes the following steps: S51. Based on the tunnel design axis and the advance drilling trajectory, construct an initial geological framework model of the area in front of the tunnel in a three-dimensional coordinate system. S52. The interface locations identified in the geological interface identification data are used as stratigraphic interfaces, and imported into and corrected in the initial geological framework model. S53. The water-rich anomalies interpreted from the spatial positioning data of the water-rich structures are treated as independent hydrogeological objects and embedded into the modified geological framework model according to their spatial coordinates and geometric shapes. S54. Assign initial hydrogeological parameter value ranges to different lithological units and water-rich structural units in the model, and finally generate the three-dimensional geological and hydrological generalized model containing stratigraphic structure, structural distribution and initial hydrogeological parameter zoning information.

[0012] Preferably, the dynamic inversion of hydrogeological parameters of the water-bearing rock mass in step S6 includes the following steps: S61. Establish a hydrogeological numerical simulation model based on the three-dimensional geological and hydrological generalization model as the grid, wherein the governing equations of the model include the groundwater flow equation. S62. Establish a quantitative relationship mapping function between the multi-dimensional drilling feature parameter set and key hydrogeological parameters, wherein the key hydrogeological parameters include the permeability coefficient and water storage coefficient of the aquifer; the mapping function is implemented by a machine learning model trained on historical data, and is used to initially convert the drilling feature parameters into hydrogeological parameter estimates. S63. Using the hydrological parameter estimates obtained in step S62 as the initial field for the inversion iteration, the genetic algorithm is used to perform parameter inversion by combining the hydrogeological numerical simulation model with the spatial distribution of water-rich areas reflected by the spatial location data of water-rich structures. S64. The inversion process aims to match the simulated borehole water level response, seepage flow rate, and the parameter change trend that indirectly reflects the seepage state during drilling tests with the measured data. The hydrogeological parameters of each model zone are continuously optimized and adjusted until the goodness of fit requirement is met, and finally the dynamic hydrogeological parameter set determined by the inversion is output.

[0013] Preferably, the water-richness classification and risk segmentation in step S7 includes the following steps: S71. Based on the geological interface identification data, continuous borehole segments that are determined to be of the same lithology and whose multi-dimensional drilling characteristic parameters do not exceed the preset threshold are divided into a basic prediction sub-region. S72. Within the basic prediction sub-region, based on the spatial positioning data of the water-rich structure, identify whether there is a water-rich structure and its intensity, mark the basic prediction sub-region containing strong water-rich structures as a high-risk segment, mark the basic prediction sub-region containing fractured rock mass as a medium-risk segment, and mark the basic prediction sub-region with intact rock mass and no water-rich structure as a low-risk segment. S73. Merge adjacent basic prediction sub-regions with the same water-bearing risk level to form the final segmented water inflow prediction intervals. Each interval is associated with its starting mileage, ending mileage, dominant lithology, water-bearing structural description, and risk level.

[0014] Preferably, the step S8 of generating segmented quantitative water inflow prediction data includes the following steps: S81. For each segmented water inflow prediction interval, extract the corresponding three-dimensional geological structure from the three-dimensional geological and hydrological generalization model and extract the corresponding hydrogeological parameters from the dynamic hydrogeological parameter set. S82. Set the simulation conditions when the tunnel is excavated to this section, including the location of the excavation face, the equivalent permeability of the support structure, and the range of excavation unloading influence. S83. Substitute the parameters extracted in step S81 and the conditions set in step S82 into the hydrogeological numerical simulation model to simulate and calculate the stable inflow of water into the tunnel through the tunnel excavation face and surrounding area under the set working conditions. S84. Repeat steps S81 to S83 for each segmented water inflow prediction interval to calculate the predicted water inflow for all intervals, and summarize to generate the segmented quantitative water inflow prediction data, which includes the mileage range of each interval, the predicted water inflow value, and the maximum water inflow.

[0015] Preferably, the step S9 of generating a dynamic prediction report of water inflow ahead of the tunnel includes the following steps: S91. Obtain the segmented quantitative water inflow prediction data, the spatial location data of the water-rich structure, and the relevant geological risk information; S92. The segmented quantitative water inflow prediction data, the spatial location data of water-rich structures, and the relevant geological risk information are correlated and fused. S93. Based on the fused data, generate a dynamic prediction report of water inflow ahead of the tunnel, which includes a distribution map of predicted water inflow along the tunnel mileage, a spatial distribution profile of water-rich structures, and a description table of geological risk sections.

[0016] Compared with the prior art, the present invention provides a method for predicting tunnel water inflow based on horizontal drilling while drilling test, which has the following beneficial effects: 1. In this invention, by collecting multi-source drilling test data in real time during the horizontal advanced drilling construction process and simultaneously collecting drilling geophysical exploration data, in-situ geological and hydrological information continuously distributed along the borehole trajectory is obtained, which can characterize the rock mass characteristics and water-rich structures of the unexcavated area, thereby improving the continuity and reliability of the geological and hydrological structure model and laying a data foundation for water inflow prediction.

[0017] 2. In this invention, a three-dimensional geological and hydrological generalized model is constructed by integrating geological interface identification data and spatial positioning data of water-rich structures. Based on the drilling characteristic parameter set, hydrogeological parameters are dynamically inverted using hydrogeological numerical simulation methods. This allows the model parameters to be dynamically updated and calibrated during the drilling process, thereby achieving a quantitative characterization of the permeability characteristics of water-bearing rock masses, reducing prediction model bias, and improving the accuracy and reliability of water inflow prediction.

[0018] 3. In this invention, the water-bearing capacity of the predicted area is classified and the risk sections are divided according to geological and water-bearing information. The quantitative water inflow of each section is calculated based on a three-dimensional geological and hydrological generalization model and dynamic parameters. Finally, a dynamic prediction report containing risk sections and quantitative prediction values ​​is generated. This allows the prediction results to be dynamically coupled with the tunnel excavation progress and to guide the drainage design in an intuitive and segmented manner. This enhances the practical application value of the prediction results in the dynamic management and control of construction risks and the formulation of differentiated prevention and control measures. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for predicting tunnel water inflow based on horizontal drilling while drilling tests according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] For specific implementation examples, please refer to: Figure 1 A method for predicting tunnel water inflow based on horizontal drilling while drilling tests, the method includes the following steps: S1. Real-time acquisition of multi-source drilling test data from horizontal advanced boreholes to generate raw drilling test dataset; S2. Preprocess and extract features from the original drilling test dataset to generate a multi-dimensional set of drilling feature parameters; S3. Based on a multi-dimensional drilling feature parameter set, identify the location of geological interfaces and changes in rock mass integrity, and generate geological interface identification data. S4. Simultaneously collect and process geophysical exploration data while drilling in the borehole to generate spatial positioning data of water-rich structures. S5. Integrate geological interface identification data with water-rich structural spatial positioning data to construct a three-dimensional geological and hydrological generalized model of the area in front of the tunnel. S6. Based on the three-dimensional geological and hydrological generalization model and the multi-dimensional drilling characteristic parameter set, and using the hydrogeological numerical simulation method, the hydrogeological parameters of the water-bearing rock mass are dynamically inverted to generate a dynamic hydrogeological parameter set. S7. Based on geological interface identification data and spatial positioning data of water-rich structures, water-richness is classified and risk sections are divided to generate segmented water inflow prediction interval data. S8. Based on the three-dimensional geological and hydrological generalization model, dynamic hydrogeological parameter set and preset tunnel excavation disturbance conditions, calculate the predicted water inflow value in each segment water inflow prediction interval and generate segmented quantitative water inflow prediction data. S9. Integrate segmented quantitative water inflow prediction data, spatial location data of water-rich structures, and related geological risk information to generate a dynamic prediction report of water inflow in front of the tunnel. S10. Based on the actual hydrogeological conditions and water inflow revealed during tunnel excavation, the three-dimensional geological and hydrological generalization model, dynamic hydrogeological parameter set, and water inflow prediction model are updated and adaptively learned through feedback.

[0022] The real-time acquisition of multi-source drilling test data from horizontally advanced boreholes in S1 includes the following steps: S11. During the horizontal advance drilling process of the tunnel boring machine, the drilling rig engineering parameters and borehole annulus parameters corresponding to the drilling depth are collected in real time and synchronously. Among them, the drilling rig engineering parameters include drill bit speed, drilling pressure, drilling speed, torque and drilling rig vibration acceleration, and the borehole annulus parameters include the inlet pressure, outlet pressure, inlet flow rate, outlet flow rate and temperature of the drilling mud; S12. Align and package the real-time collected drilling rig engineering parameters and borehole annulus parameters according to a unified timestamp and borehole depth coordinates to generate an original drilling test dataset with spatiotemporal labels.

[0023] Preprocessing and feature extraction in S2 include the following steps: S21. Perform data cleaning on the original drilling test dataset. Data cleaning includes removing outlier values, smoothing filtering, and linear interpolation filling for missing signal segments. S22. For the drilling test data after cleaning, extract the characteristic parameters representing the mechanical and permeability properties of the rock mass from the time domain, frequency domain, and time-frequency domain respectively, according to a fixed drilling footage. Among them, the time-domain feature parameters include the mean, variance, peak value and rate of change within each parameter window; the frequency-domain feature parameters include the main frequency amplitude and energy proportion of a specific frequency band obtained by fast Fourier transform; and the time-frequency domain feature parameters include the energy spectrum characteristics obtained by wavelet transform. The extraction of characteristic parameters representing the mechanical and permeability properties of rock masses from the time domain includes the following steps: S221. For the drilling footage test data within a fixed drilling footage window, calculate its statistical characteristics, including the mean, variance, and rate of change of drilling pressure, torque, drilling speed, and mud inlet pressure within the window; wherein, for any parameter In length Data sequence within the window its window mean ,variance rate of change Specifically, it is calculated using the following formula: ; ; ; in, For any parameter tested while drilling, For parameters In length The first sliding window Data points, This represents the number of data points within the sliding window. For data points The corresponding timestamp; S222. The calculated window mean, variance and rate of change are combined as time-domain characteristic parameters characterizing the comprehensive strength and degree of fragmentation of the rock mass in this advance section. Extracting characteristic parameters from the frequency domain to characterize the mechanical and permeability properties of rock masses includes the following steps: S223. Perform a fast Fourier transform on the drilling rig vibration acceleration signal within the fixed drilling footage window to obtain its frequency domain energy distribution; S224. Extract the preset low-frequency band from the frequency domain energy distribution. With high frequency band Energy percentage and The energy percentage is calculated using the following formula: ; ; in, , These represent the preset energy proportions of the low-frequency band and the high-frequency band, respectively. For frequency, To perform a fast Fourier transform on the vibration acceleration signal, at the frequency... The amplitude in the frequency domain at that point, Let Nyquist be the frequency of the signal; S225. The combination of the low-frequency band energy ratio and the high-frequency band energy ratio is used as a frequency domain characteristic parameter to characterize the integrity of the rock mass and the degree of fracture development in this advance section. Extracting characteristic parameters from the time-frequency domain to characterize the mechanical and permeability properties of rock masses includes the following steps: S226. Perform wavelet packet transform on the drilling rig vibration acceleration signal within the fixed drilling footage window and decompose it to a set level. S227. Calculate the nodes of each wavelet packet. signal energy Construct the wavelet packet energy spectrum of the advance segment signal; where the nodal energy... Calculated using the following formula: ; in, After wavelet packet transform, located at scale Location index Signal energy at the node, After the signal undergoes wavelet packet transform, at the node Wavelet packet coefficients on The scale for wavelet packet decomposition. To be at a specific scale The position index on, For time; S228. Use the energy values ​​of specific nodes in the wavelet packet energy spectrum that characterize signal abrupt changes and energy distribution in different frequency bands as time-frequency domain characteristic parameters that characterize the heterogeneity of the rock mass and potential seepage channels in this advance section. S23. After normalizing the extracted time-domain, frequency-domain, and time-frequency-domain feature parameters, combine them according to the borehole depth sequence to generate a multi-dimensional drilling feature parameter set.

[0024] Identifying geological interfaces and changes in rock mass integrity in S3 includes the following steps: S31. Select a subset of characteristic parameters that are sensitive to lithological changes and fracture development from the multi-dimensional drilling characteristic parameter set as the key discrimination index set for geological interface identification. S32. A sliding window-based abrupt change detection algorithm is used to analyze the change sequence of key discrimination index set along the borehole depth, and to identify the depth points where the characteristic parameters change. The changes include step changes, trend reversals, and abnormal peaks. The steps include: S321. Along the drilling depth direction, slide the data sequence of the key discrimination index set with a window of preset length; S322. Within each sliding window, calculate the difference in characteristic parameter statistics between the subsequence at the end of the window and the subsequence at the beginning of the window; where, the difference... Calculated using the following formula: ; in, and These are the mean and standard deviation of the first subsequence of the window, respectively. and These are the mean and standard deviation of the subsequence at the back of the window, respectively. These are the weighting coefficients; S323. When the difference exceeds the historical data, the depth point corresponding to the center of the window is determined to be the abrupt change point where the feature parameters change. S33. The identified feature change depth points are initially determined as potential geological interface locations, and geological interface identification data containing location and change intensity are generated.

[0025] The steps involved in generating spatial location data for water-rich structures in S4 are as follows: S41. During the horizontal advance drilling process, the borehole transient electromagnetic instrument is used simultaneously to continuously collect geophysical signals reflecting the electrical differences of the surrounding rock mass along the borehole depth direction to generate raw geophysical data while drilling. S42. Perform inversion interpretation processing on the original geophysical data while drilling, identify low resistivity anomaly zones, and interpret these anomaly zones as water-rich fracture zones, including the following steps: S421. Based on the borehole trajectory and instrument parameters, establish an initial electrical structure model of the rock mass surrounding the borehole, whose resistivity distribution is as follows: ,in It is a spatial position vector; S422, Using forward calculation algorithm Calculate the theoretical electromagnetic response of the initial electrical structure model ; S423, Theoretical Electromagnetic Response Measured electromagnetic response compared with original geophysical data while drilling Compare and calculate the residuals. : ; in, The residual between theoretical calculations and measured data. For measured electromagnetic response data, These are theoretical electromagnetic response data; S424. Adjust the distribution of electrical parameters in the initial electrical structure model through an iterative optimization algorithm until the residuals meet the preset convergence criteria, and obtain the inverted resistivity profile. ; S43. Digitally characterize the location, geometry, size, and relative spatial relationship of the interpreted water-rich anomaly to the borehole, generating spatial location data of the water-rich structure, including the anomaly's starting depth, ending depth, center location, and anomaly intensity, including the following steps: S431. In the three-dimensional coordinate system, the low-resistivity anomaly region is delineated from the inverted resistivity profile based on the resistivity anomaly threshold. S432. Record the boundary vertex coordinates of the low-resistivity anomaly region and fit them with a geometric body, including an ellipsoid and a cuboid. When fitting to an ellipsoid, the central coordinates are used. , semi-axis length The spatial orientation is used for characterization; when the fit is a cuboid, the coordinates of the minimum point are used. and the coordinates of the maximum point To characterize; S433. Calculate the center position coordinates, spatial distribution length, width, height and volume of the water-rich anomaly based on the fitted geometry, as the scale parameters for its digital representation. The steps involved in constructing a three-dimensional geological and hydrological generalization model of the area in front of the tunnel in S5 are as follows: S51. Based on the tunnel design axis and the advance drilling trajectory, construct an initial geological framework model of the area in front of the tunnel in a three-dimensional coordinate system. S52. Use the interface locations identified in the geological interface identification data as stratigraphic interfaces, import them into and correct the initial geological framework model. S53. The water-rich anomalies interpreted from the spatial location data of water-rich structures are treated as independent hydrogeological objects and embedded into the modified geological framework model according to their spatial coordinates and geometric shapes. S54. Assign initial hydrogeological parameter ranges to different lithological units and water-rich structural units in the model, and finally generate a three-dimensional geological and hydrological generalized model containing stratigraphic structure, structural distribution and initial hydrogeological parameter zoning information.

[0026] The dynamic inversion of hydrogeological parameters of water-bearing rock masses in S6 includes the following steps: S61. Establish a hydrogeological numerical simulation model based on a three-dimensional geological and hydrological generalized model. The governing equations of the model include the groundwater flow equation, specifically: A three-dimensional numerical model of groundwater flow based on the finite difference method is established, and its governing equations are solved discretly using unsteady flow partial differential equations of the following form: ; in, , , For water-containing medium along , , The main permeability coefficient in the direction, For the water head, For source and sink items, For water storage rate, For time; S62. Establish a quantitative mapping function between a multi-dimensional set of drilling feature parameters and key hydrogeological parameters, where the key hydrogeological parameters include the permeability coefficient and water storage coefficient of the aquifer. The mapping function is implemented using a machine learning model trained on historical data, and is used to initially convert drilling feature parameters into hydrogeological parameter estimates, including the following steps: S621. Collect a set of multi-dimensional drilling feature parameters corresponding to borehole sections in historical engineering projects where hydrogeological parameters have been accurately obtained, to form a training sample set. ,in For feature vectors, ,in, For the first The target output vector corresponding to each training sample. For the first The aquifer permeability coefficient corresponding to each training sample For the first The aquifer storage coefficient corresponding to each training sample It is the transpose of the vector; S622. A feedforward neural network is selected as the machine learning model, with multi-dimensional drilling feature parameters as input and the logarithm of the permeability coefficient and water storage coefficient as output. S623. Supervised training of the feedforward neural network is performed using the training sample set, and the network weights and biases are adjusted through the backpropagation algorithm to minimize the loss function. The calculation formula is: ; in, The total number of training samples. For the first A log-valued vector of the true hydrogeological parameters of each sample. For the neural network to the first The predicted output vectors for each sample are processed until the error between the model's predicted output and the actual hydrogeological parameter values ​​is minimized, thus obtaining the trained quantitative relational mapping function. ; S63. Using the hydrological parameter estimates obtained in S62 as the initial field for the inversion iteration, and combining the hydrogeological numerical simulation model with the spatial distribution of water-rich areas reflected by the spatial location data of water-rich structures, a genetic algorithm is used to perform parameter inversion, including the following steps: S631. Encode the aquifer permeability coefficient and water storage coefficient to be inverted into chromosomes, and randomly generate chromosomes containing... The initial population of chromosomes ; S632, Based on each chromosome of the current population The corresponding parameter values ​​were used to run a hydrogeological numerical simulation model. The goodness of fit between the simulation results and the measured data in S64 was calculated, and the reciprocal of the goodness of fit was used as the fitness of the individual. Fitness is defined as: ; in, For when using chromosomes When running hydrogeological numerical simulations with corresponding parameters, the residuals between the simulation results and the measured data are... The first in the genetic algorithm population chromosome, Chromosomes The residuals between the simulation results and the measured data under the corresponding parameters; S633. Based on fitness, perform roulette wheel selection and retain individuals with high fitness; S634. Perform single-point crossover and uniform mutation operations on the selected chromosomes to generate a new generation of population; S635. Repeat steps S632 to S634 until the preset maximum number of iterations is reached. Decode the chromosome with the highest fitness in the final population to obtain the inverted hydrogeological parameters. S64. The inversion process aims to match the simulated borehole water level response, seepage flow rate, and the changing trends of parameters indirectly reflecting the seepage state during drilling tests with the measured data. It continuously optimizes and adjusts the hydrogeological parameters of each model zone until the goodness-of-fit requirement is met, ultimately outputting a dynamic hydrogeological parameter set determined by the inversion, specifically including: S641. Calculate the root mean square error between the simulated borehole water level response and seepage flow rate and the measured data. Among them, for those with Water level data at one comparison point: ; in, The number of water level data points used for comparison. and The first Simulated water level and measured water level at each point; S642. The inversion iteration process continues until either of the following conditions is met: the root mean square error is less than a preset error threshold. and continuous The decrease in root mean square error in each iteration Less than the preset convergence threshold .

[0027] The following steps are involved in classifying water abundance and delineating risk zones in S7: S71. Based on geological interface identification data, continuous borehole sections that are determined to be of the same lithology and whose multi-dimensional drilling characteristic parameters do not exceed a preset threshold are divided into a basic prediction sub-region. The specific determination method is as follows: S711. Select multiple historical borehole sections that represent uniform and complete rock masses, and calculate the standard deviation of their multi-dimensional drilling characteristic parameters. S712. Multiply the statistical mean of the standard deviations of each characteristic parameter by an empirical coefficient ranging from 1.5 to 2.5 to obtain a preset threshold; for the first... Each feature parameter has a preset threshold. Determine by the following formula: ; in, It is the first On the first reference uniform borehole section The standard deviation of each characteristic parameter For reference, the total number of borehole sections, This is an empirical coefficient, with a value between 1.5 and 2.5. S72. Within the basic prediction sub-region, based on the spatial location data of water-rich structures, identify whether there are water-rich structures and their intensity. Mark the basic prediction sub-region containing strong water-rich structures as high-risk sections, the basic prediction sub-region containing fractured rock masses as medium-risk sections, and the basic prediction sub-region with intact rock masses and no water-rich structures as low-risk sections. S73. Merge adjacent basic prediction sub-regions with the same water-bearing risk level to form the final segmented water inflow prediction intervals. Each interval is associated with its starting mileage, ending mileage, dominant lithology, water-bearing structural description, and risk level.

[0028] The steps involved in generating segmented quantitative inflow prediction data in S8 are as follows: S81. For each segmented water inflow prediction interval, extract the corresponding three-dimensional geological structure from the three-dimensional geological and hydrological generalization model and extract the corresponding hydrogeological parameters from the dynamic hydrogeological parameter set. S82. Set the simulation conditions when the tunnel is excavated to this section, including the location of the excavation face, the equivalent permeability of the support structure, and the range of excavation unloading influence. S83. Substitute the parameters extracted in S81 and the conditions set in S82 into the hydrogeological numerical simulation model to simulate and calculate the steady inflow of water into the tunnel through the tunnel excavation face and surrounding areas under the set working conditions. This specifically includes the following steps: S831. The three-dimensional geological structure information of the interval extracted from the three-dimensional geological and hydrological generalization model in S81, the permeability coefficient and water storage coefficient extracted from the dynamic hydrogeological parameter set, and the tunnel excavation face location, the equivalent permeability of the support structure and the excavation unloading influence range set in S82 are jointly set as the input conditions of the hydrogeological numerical simulation model. S832. Run the hydrogeological numerical simulation model and perform groundwater seepage numerical simulation under the set input conditions to calculate the water head distribution and flow direction in the surrounding rock mass after tunnel excavation. S833. Extract the total flow rate into the tunnel interior space from the model simulation results, passing through the tunnel excavation face and the boundary within the excavation unloading influence range defined in S82. S834. Continue simulation calculation until the total flow rate into the tunnel tends to stabilize over time. Output the stable flow rate value as the stable flow rate prediction value of the segmented water inflow prediction interval under the set working conditions. S84. Repeat S81 to S83 for each segmented water inflow prediction interval to calculate the predicted water inflow for all intervals, and summarize to generate segmented quantitative water inflow prediction data that includes the mileage range of each interval, the predicted water inflow value, and the maximum water inflow.

[0029] The steps involved in generating a dynamic prediction report of water inflow ahead of the tunnel in S9 are as follows: S91. Obtain segmented quantitative water inflow prediction data, spatial location data of water-rich structures, and related geological risk information; S92. Link and integrate segmented quantitative water inflow prediction data, spatial location data of water-rich structures, and relevant geological risk information; S93. Based on the fused data, generate a dynamic prediction report of water inflow ahead of the tunnel, which includes a distribution map of predicted water inflow along the tunnel mileage, a spatial distribution profile of water-rich structures, and a description table of geological risk sections.

[0030] The feedback update and adaptive learning of the three-dimensional geological and hydrological generalization model, dynamic hydrogeological parameter set, and water inflow prediction model in S10 includes the following steps: S101. After the tunnel is excavated and exposed, obtain the lithology, structural attitude, and measured water inflow and location of the actual exposed surface. S102. Compare the actual exposure information obtained with the model prediction information of the corresponding mileage. When the lithology and structural occurrence of the local strata do not match the model, the geological attributes and structural geometry of the corresponding area in the three-dimensional geological and hydrological generalization model are corrected. S103. When the deviation between the measured inflow and the predicted value exceeds the preset tolerance, the measured inflow data will be used as the new constraint to re-execute the hydrogeological parameter inversion, updating the parameters of the corresponding area in the dynamic hydrogeological parameter set. Specifically, this includes: The measured water inflow data obtained in S101 is compared with the predicted value of the corresponding interval in the segmented quantitative water inflow prediction data. When the deviation exceeds the preset tolerance, the parameter update process is triggered. Using measured water inflow data as new target constraints, and combining the multi-dimensional drilling feature parameter set corresponding to the deviation interval, we prepare the input data for re-inversion. The hydrogeological numerical simulation model established by S61 was called, and the genetic algorithm of S63 was used as the optimization tool, with the measured water inflow data as the fitting target, and the parameter inversion calculation was re-executed. The hydrogeological parameters obtained from this inversion calculation are updated to the local parameters in the corresponding deviation interval of the dynamic hydrogeological parameter set to complete the parameter feedback correction. S104. The multi-dimensional drilling feature parameter set corresponding to this excavation section, the corrected model parameters, and the measured water inflow are combined to form a new sample and added to the historical engineering training sample library for subsequent retraining and optimization of the quantitative relationship mapping function.

[0031] The operation steps of this method for predicting tunnel water inflow based on horizontal drilling while drilling testing are as follows: Step 1: Synchronous acquisition and processing of multi-source drilling test data: First, during the horizontal advance drilling process of the tunnel boring machine, two key types of in-situ data were collected in real time and synchronously. The first is the drilling rig engineering parameters and borehole annulus parameters, generating a raw, spatiotemporally labeled test-while-drilling dataset. The second is the borehole transient electromagnetic instrument used to collect geophysical exploration data reflecting the differences in the electrical properties of the surrounding rock mass. By cleaning the raw test-while-drilling data and extracting features from the time and frequency domains according to a fixed drilling depth, a multi-dimensional set of drilling characteristic parameters characterizing the continuous changes in rock mechanics and permeability properties is generated.

[0032] Step 2: Integrated modeling of geological structure and water-rich structures: Based on a multi-dimensional set of drilling feature parameters, a sliding window-based abrupt change detection algorithm was used to identify the location of geological interfaces, generating geological interface identification data. Simultaneously, geophysical exploration data from drilling was inverted and interpreted to identify low resistivity anomalies and interpret them as water-rich fracture zones, generating spatial location data of water-rich structures including their spatial location, morphology, and scale. Subsequently, these two types of data were fused, and a three-dimensional geological and hydrological generalized model of the area ahead of the tunnel, including stratigraphic interfaces and water-rich anomalies, was constructed based on the tunnel axis and borehole trajectory.

[0033] Step 3: Dynamic inversion and quantification of hydrogeological parameters: A hydrogeological numerical simulation model with a three-dimensional geological and hydrological generalized model as its spatial framework was established. A machine learning model trained on historical data was used to establish a quantitative mapping function between a multi-dimensional set of drilling feature parameters and key hydrogeological parameters, obtaining initial parameter estimates. These estimates served as the initial field for inversion iterations. Combined with spatial location data of water-rich structures, a genetic algorithm was used to drive the hydrogeological numerical simulation model for parameter inversion, ensuring optimal matching between the simulation results and measured data. Finally, a calibrated dynamic hydrogeological parameter set was output.

[0034] Step 4: Risk Classification and Quantitative Prediction of Water Inrush: Based on geological interface identification data and spatial location data of water-rich structures, the area ahead of the tunnel is classified into water-rich zones and risk sections. Areas containing strongly water-rich structures are marked as high-risk sections, areas with fractured rock mass are marked as medium-risk sections, and areas with intact rock mass and no water-rich structures are marked as low-risk sections, forming segmented water inflow prediction intervals. For each interval, based on a three-dimensional geological and hydrological generalization model and a dynamic hydrogeological parameter set, excavation conditions are set, and the stable water inflow is calculated through a hydrogeological numerical simulation model, thereby generating segmented quantitative water inflow prediction data. This data is then integrated into a dynamic prediction report of water inflow ahead of the tunnel, including prediction maps and tables.

[0035] Step 5: Model Feedback Updates and Self-Learning The predicted results are compared with the actual lithology, structural occurrence, and measured water inflow revealed after tunnel excavation. If the actual revealed information does not match the model, the three-dimensional geological and hydrological generalization model is revised. If the deviation between the measured water inflow and the predicted value exceeds a preset tolerance, the hydrogeological parameter inversion is re-executed using the measured data as new constraints, and the dynamic hydrogeological parameter set is updated. Simultaneously, data from newly revealed sections are used as samples to supplement the historical database to optimize the quantitative relationship mapping function, achieving adaptive learning of the method and continuous improvement of prediction accuracy.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for predicting tunnel water inflow based on horizontal drilling while drilling testing, characterized in that: The method includes the following steps: S1. Real-time acquisition of multi-source drilling test data from horizontal advanced boreholes to generate raw drilling test dataset; S2. Preprocess and extract features from the original drilling test dataset to generate a multi-dimensional drilling feature parameter set; S3. Based on the multi-dimensional drilling feature parameter set, identify the geological interface and the location of changes in rock mass integrity, and generate geological interface identification data. S4. Simultaneously collect and process geophysical exploration data while drilling in the borehole to generate spatial positioning data of water-rich structures. S5. Integrate the geological interface identification data with the spatial positioning data of the water-rich structure to construct a three-dimensional geological and hydrological generalized model of the area in front of the tunnel. S6. Based on the three-dimensional geological and hydrological generalization model and the multi-dimensional drilling characteristic parameter set, and using the hydrogeological numerical simulation method, dynamically invert the hydrogeological parameters of the water-bearing rock mass to generate a dynamic hydrogeological parameter set. S7. Based on the geological interface identification data and the spatial positioning data of the water-rich structure, perform water-rich classification and risk segment division, and generate segmented water inflow prediction interval data. S8. Based on the three-dimensional geological and hydrological generalization model, the dynamic hydrogeological parameter set, and the preset tunnel excavation disturbance conditions, calculate the predicted water inflow value in each segmented water inflow prediction interval, and generate segmented quantitative water inflow prediction data. S9. Integrate the segmented quantitative water inflow prediction data, the spatial positioning data of the water-rich structure, and related geological risk information to generate a dynamic prediction report of water inflow in front of the tunnel. S10. Based on the actual hydrogeological conditions and water inflow revealed during tunnel excavation, the three-dimensional geological and hydrological generalization model, the dynamic hydrogeological parameter set, and the water inflow prediction model are updated and adaptively learned.

2. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 1, characterized in that: The real-time acquisition of multi-source drilling test data from horizontal advanced boreholes in S1 includes the following steps: S11. During the horizontal advance drilling process of the tunnel boring machine, the drilling rig engineering parameters and borehole annulus parameters corresponding to the drilling depth are collected in real time and synchronously. The drilling rig engineering parameters include drill bit rotation speed, drilling pressure, drilling speed, torque and drilling rig vibration acceleration, and the borehole annulus parameters include borehole mud inlet pressure, outlet pressure, inlet flow rate, outlet flow rate and temperature. S12. Align and package the real-time collected drilling rig engineering parameters and borehole annulus parameters according to a unified timestamp and borehole depth coordinates to generate the original drilling test dataset with spatiotemporal tags.

3. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 2, characterized in that: The preprocessing and feature extraction in S2 includes the following steps: S21. Perform data cleaning on the original drilling test dataset. The data cleaning includes removing abnormal jump values, smoothing filtering, and linear interpolation filling for missing signal segments. S22. For the drilling test data after cleaning, extract the characteristic parameters representing the mechanical and permeability properties of the rock mass from the time domain, frequency domain, and time-frequency domain respectively, according to a fixed drilling footage. Among them, the time-domain feature parameters include the mean, variance, peak value and rate of change within each parameter window; the frequency-domain feature parameters include the main frequency amplitude and the energy proportion of a specific frequency band obtained by fast Fourier transform; and the time-frequency domain feature parameters include the energy spectrum features obtained by wavelet transform. S23. After normalizing the extracted time-domain, frequency-domain, and time-frequency-domain feature parameters, combine them according to the borehole depth sequence to generate the multi-dimensional drilling feature parameter set.

4. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 3, characterized in that: The identification of geological interfaces and locations of changes in rock mass integrity in step S3 includes the following steps: S31. Select a subset of characteristic parameters that are sensitive to lithological changes and fracture development from the multi-dimensional drilling characteristic parameter set, and use it as a key discrimination index set for geological interface identification. S32. Using a sliding window-based mutation point detection algorithm, the change sequence of the key discrimination index set along the borehole depth is analyzed to identify the depth points where the characteristic parameters change, including step changes, trend reversals, and abnormal peaks. S33. The identified feature change depth points are initially determined as potential geological interface locations, and the geological interface identification data containing the location and change intensity is generated.

5. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 1, characterized in that: The process of generating spatial location data for water-rich structures in S4 includes the following steps: S41. During the horizontal advance drilling process, the borehole transient electromagnetic instrument is used simultaneously to continuously collect geophysical signals reflecting the electrical differences of the surrounding rock mass along the borehole depth direction to generate raw geophysical data while drilling. S42. Perform inversion interpretation processing on the original geophysical data while drilling, identify the low resistivity anomaly areas, and interpret these anomaly areas as water-rich fracture zones. S43. The location, geometry, size and relative spatial relationship of the water-rich anomaly are digitally characterized to generate the spatial positioning data of the water-rich structure, including the anomaly's starting depth, ending depth, center position and anomaly intensity.

6. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 5, characterized in that: The steps involved in constructing the three-dimensional geological and hydrological generalization model of the area in front of the tunnel in S5 are as follows: S51. Based on the tunnel design axis and the advance drilling trajectory, construct an initial geological framework model of the area in front of the tunnel in a three-dimensional coordinate system. S52. The interface locations identified in the geological interface identification data are used as stratigraphic interfaces, and imported into and corrected in the initial geological framework model. S53. The water-rich anomalies interpreted from the spatial positioning data of the water-rich structures are treated as independent hydrogeological objects and embedded into the modified geological framework model according to their spatial coordinates and geometric shapes. S54. Assign initial hydrogeological parameter value ranges to different lithological units and water-rich structural units in the model, and finally generate the three-dimensional geological and hydrological generalized model containing stratigraphic structure, structural distribution and initial hydrogeological parameter zoning information.

7. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 6, characterized in that: The dynamic inversion of hydrogeological parameters of the water-bearing rock mass in S6 includes the following steps: S61. Establish a hydrogeological numerical simulation model based on the three-dimensional geological and hydrological generalization model as the grid, wherein the governing equations of the model include the groundwater flow equation. S62. Establish a quantitative relationship mapping function between the multi-dimensional drilling feature parameter set and key hydrogeological parameters, wherein the key hydrogeological parameters include the permeability coefficient and water storage coefficient of the aquifer; the mapping function is implemented by a machine learning model trained on historical data, and is used to initially convert the drilling feature parameters into hydrogeological parameter estimates. S63. Using the hydrological parameter estimates obtained in step S62 as the initial field for the inversion iteration, the genetic algorithm is used to perform parameter inversion by combining the hydrogeological numerical simulation model with the spatial distribution of water-rich areas reflected by the spatial location data of water-rich structures. S64. The inversion process aims to match the simulated borehole water level response, seepage flow rate, and the parameter change trend that indirectly reflects the seepage state during drilling tests with the measured data. The hydrogeological parameters of each model zone are continuously optimized and adjusted until the goodness of fit requirement is met, and finally the dynamic hydrogeological parameter set determined by the inversion is output.

8. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 7, characterized in that: The water-richness classification and risk segmentation in S7 include the following steps: S71. Based on the geological interface identification data, continuous borehole segments that are determined to be of the same lithology and whose multi-dimensional drilling characteristic parameters do not exceed the preset threshold are divided into a basic prediction sub-region. S72. Within the basic prediction sub-region, based on the spatial positioning data of the water-rich structure, identify whether there is a water-rich structure and its intensity, mark the basic prediction sub-region containing strong water-rich structures as a high-risk segment, mark the basic prediction sub-region containing fractured rock mass as a medium-risk segment, and mark the basic prediction sub-region with intact rock mass and no water-rich structure as a low-risk segment. S73. Merge adjacent basic prediction sub-regions with the same water-bearing risk level to form the final segmented water inflow prediction intervals. Each interval is associated with its starting mileage, ending mileage, dominant lithology, water-bearing structural description, and risk level.

9. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 8, characterized in that: The step of generating segmented quantitative water inflow prediction data in S8 includes the following steps: S81. For each segmented water inflow prediction interval, extract the corresponding three-dimensional geological structure from the three-dimensional geological and hydrological generalization model and extract the corresponding hydrogeological parameters from the dynamic hydrogeological parameter set. S82. Set the simulation conditions when the tunnel is excavated to this section, including the location of the excavation face, the equivalent permeability of the support structure, and the range of excavation unloading influence. S83. Substitute the parameters extracted in step S81 and the conditions set in step S82 into the hydrogeological numerical simulation model to simulate and calculate the stable inflow of water into the tunnel through the tunnel excavation face and surrounding area under the set working conditions. S84. Repeat steps S81 to S83 for each segmented water inflow prediction interval to calculate the predicted water inflow for all intervals, and summarize to generate the segmented quantitative water inflow prediction data, which includes the mileage range of each interval, the predicted water inflow value, and the maximum water inflow.

10. The method for predicting tunnel water inflow based on horizontal drilling while drilling testing according to claim 9, characterized in that: The process of generating a dynamic prediction report of water inflow ahead of the tunnel in S9 includes the following steps: S91. Obtain the segmented quantitative water inflow prediction data, the spatial location data of the water-rich structure, and the relevant geological risk information; S92. The segmented quantitative water inflow prediction data, the spatial location data of water-rich structures, and the relevant geological risk information are correlated and fused. S93. Based on the fused data, generate a dynamic prediction report of water inflow ahead of the tunnel, which includes a distribution map of predicted water inflow along the tunnel mileage, a spatial distribution profile of water-rich structures, and a description table of geological risk sections.

Images