Crustal stress measurement method based on fusion of radial deformation and differential strain of rock core

Abstract

The invention discloses a ground stress measurement method based on rock core radial deformation and differential strain fusion, which comprises the following steps: acquiring a drill hole wall image and a rock core surface image, identifying common features according to the drill hole wall image and the rock core surface image, and mapping azimuth information of the common features of the drill hole wall image into the rock core surface image; measuring radial deformation data of the rock core, calculating a radial deformation curve, determining a principal stress direction according to wave crest and wave trough positions of the radial deformation curve, cutting the rock core according to the radial deformation data, marking end surfaces along the wave crest and wave trough directions, and respectively arranging strain gauges on the three surfaces to measure three-dimensional strain. The main strain ratio is calculated according to the three-dimensional strain, three main stress absolute values are calculated according to the radial deformation of the rock core and the main strain ratio, and the complex behavior of the rock core in the stress release process is reflected more accurately through a machine learning algorithm, so that the measurement precision in the main stress direction is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rock mass mechanics, and particularly relates to a geostress measurement method based on fusion of core radial deformation and differential strain. BACKGROUND

[0002] With the continuous growth of global mineral resource demand, the shallow surface mineral resources are gradually exhausted, and the development of deep resources has become an inevitable trend. At present, the maximum mining depth of coal and metal mineral resources in China has exceeded 1500 meters, and increases by 8-25 meters per year; at the same time, the construction and design depth of underground space such as nuclear waste deep geological disposal, energy storage, core military protection engineering, railway transportation tunnel, water conservancy and hydropower engineering, etc. are all kilometer level, and the depth of geothermal exploitation and shale gas development is up to 3000-3500 meters. In these deep engineering, the rock mass deformation and failure and the stability of surrounding rock under the environment of high geostress, high temperature and high permeability pressure are particularly prominent. Accurate measurement of the initial geostress state of rock is the key to solving the above problems, and is also an important topic in the research of rock mass mechanics and engineering. Geostress measurement not only provides basic data for the design, construction and safety evaluation of underground engineering, but also directly affects the long-term stability and economic benefits of the engineering.

[0003] The core radial deformation analysis method (DCDA) measures the cross-sectional deformation of the core after stress relief to solve the horizontal differential stress, and the differential strain method (DSA) solves the three-dimensional principal stress ratio through strain recovery. However, a single method often has limitations, such as the DCDA method can only solve the two-dimensional plane stress difference, and the DSA method cannot directly determine the principal stress value. SUMMARY

[0004] The present application provides a geostress measurement method based on fusion of core radial deformation and differential strain, which more accurately reflects the complex behavior of the core in the stress release process through machine learning algorithm based on multi-dimensional information such as time sequence deformation data, environmental parameters, equipment state and geological characteristics, thereby significantly improving the measurement accuracy of the principal stress direction.

[0005] The present application provides a geostress measurement method based on fusion of core radial deformation and differential strain, comprising:

[0006] S101, collecting a borehole wall image and a core surface image, identifying common features according to the borehole wall image and the core surface image, and mapping the orientation information of the common features of the borehole wall image to the core surface image;

[0007] S102, measuring core radial deformation data, calculating a radial deformation curve according to the measurement data, and determining the principal stress direction according to the peak and valley positions of the radial deformation curve;

[0008] S103, cutting the core according to the radial deformation data, marking the end face in the wave peak and wave valley direction, the three faces of the cut core being respectively parallel to the maximum horizontal principal stress, the minimum horizontal principal stress and the vertical direction, and respectively setting strain gauges on the three faces to measure three-way strain, calculating the principal strain ratio according to the three-way strain, and calculating the absolute values of the three principal stresses according to the differential stress and the principal stress ratio.

[0009] Preferably, the step of mapping the orientation information of the common features of the borehole wall image to the core surface image is: marking the fracture wave peak position and orientation angle in the borehole wall image; marking the wave peak position of the same fracture in the core surface image, drawing a feature marker line; mapping the orientation angle information in the borehole image to the core surface image, drawing a directional reference line and marking the geographic orientation.

[0010] Preferably, the calculation formula of the radial deformation curve is: Δ rθ =A·sin(θ-φ)+B, wherein Δ rθ is the radial deformation of the core at the rotation angle θ, that is, the radial deformation curve, A is the deformation amplitude, θ is the rotation angle, φ is the phase angle, and B is the average deformation.

[0011] Preferably, the formula for calculating the principal strain is: wherein ε1 to ε9 are 9 strain values measured in the experiment, ε xx , ε yy , and ε zz are the normal strains in the x, y, and z directions, respectively, the differential strain method stress measurement test changes from 9 strain components to be measured to 3 positions to be measured, and the formula that the three normal strain measurement results satisfy the proportional relationship is: ε xx :ε yy :ε zz =a:b:c, wherein a, b, and c are actual measurement ratios.

[0012] Preferably,

[0013] S201, collecting time sequence deformation, environmental parameter, equipment state and geological feature data, constructing a sub-network according to the collected data, and fusing each sub-network to obtain a fusion feature vector;

[0014] S202, constructing a prediction model according to the fusion feature vector, predicting the principal stress direction according to the prediction model, comparing the predicted principal stress direction with the actually measured principal stress direction, and correcting the predicted principal stress direction if the error exceeds a preset threshold.

[0015] Preferably, the sub-networks include a long short-term memory network, an environmental compensation network and a geological priori network.

[0016] Preferably, the extraction of the geological features is achieved by calculating the fracture density, and the formula for calculating the fracture density is: Wherein, F is the fracture density, N C is the number of fractures, L t is the core length.

[0017] Preferably, the time series features, the environmental compensation and the geological prior features are weighted and fused using attention weights, and the formula for calculating the attention weights is: Wherein, a i is the attention weight, f i is the feature vector, which is the vector of the time series features, the environmental compensation and the geological prior features respectively, f q is the query vector, and W is the weight matrix.

[0018] One or more technical solutions provided in the present application have at least the following technical effects or advantages: by fusing core radial deformation analysis (DCDA) and differential strain analysis (DSA), precise inversion of in-situ stress measurement from two dimensions to three dimensions is achieved, which has remarkable technical effects: first, based on feature matching of borehole images and core surface images (such as fracture sinusoidal curve peak positioning), directional orientation of the core with low cost and high precision (azimuth error < 3°) is achieved, which breaks through the cost limitation of traditional paleomagnetic orientation; second, high-precision laser ranging platform (accuracy up to 0.001 mm) is used to obtain radial deformation data, the principal stress direction is accurately determined through sinusoidal curve fitting (such as the peak corresponding to the maximum horizontal principal stress), and three-dimensional principal stress ratio is directly obtained by combining the oriented rock sample and simplifying the differential strain experiment (strain gauge from 9 to 3); finally, the absolute value of the principal stress is calculated through data fusion, which solves the limitation that the traditional DSA cannot independently determine the value, this method is simple to operate, low in cost, and suitable for deep "three high" environment, which provides reliable parameter support for stability analysis of underground engineering, and the deviation between the measured results and the hydraulic fracturing method is only 3°-17°, which verifies the engineering reliability.

[0019] Through multi-dimensional information such as time series deformation data, environmental parameters, equipment state and geological features, and through machine learning algorithm, the complex behavior of the core in the stress release process is more accurately reflected, thereby the measurement accuracy of the principal stress direction is significantly improved, in the process of deep resource development, the in-situ stress direction of the underground rock mass is accurately determined to optimize the roadway support design and mining scheme, and the safety and efficiency of the mine exploitation are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is the unfolded image of the core surface image of the in-situ stress measurement method based on core radial deformation and differential strain fusion of the present application;

[0021] Figure 2 is the unfolded image of the borehole wall image of the present application;

[0022] Figure 3 A schematic diagram showing the determination of the principal stress direction in this invention;

[0023] Figure 4 The graph shows the radial deformation measurement curve and fitting curve of the core sample in this invention.

[0024] Figure 5 This is a schematic diagram of the core directional processing of the present invention;

[0025] Figure 6 This is a schematic diagram of a strain gauge with nine strain gauges applied, illustrating the conventional differential strain analysis method of this invention.

[0026] Figure 7 This is a schematic diagram of strain gauge bonding in the fusion analysis method of the present invention. Detailed Implementation

[0027] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0028] It should be noted that the terms "vertical," "horizontal," "up," "down," "left," "right," and similar expressions used in this article are for illustrative purposes only and do not represent the only possible implementation.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0030] Example 1: Figure 1 This is a flowchart illustrating a geostress measurement method based on the fusion of core radial deformation and differential strain according to an embodiment of the present invention, including:

[0031] S101, acquire borehole wall images and core surface images, identify common features based on borehole wall images and core surface images, and map the orientation information of the common features in the borehole wall images to the core surface images;

[0032] Specifically, a digital panoramic borehole camera system is used to acquire borehole wall images. This system consists of a panoramic imaging probe, a main unit, a depth encoder, a winch, and dedicated cables. It integrates electronic, video, digital, and computer application technologies, employing a truncated conical mirror panoramic imaging structure design. It can observe a 360° panoramic view of the borehole wall in real time. The panoramic imaging probe has a diameter of 72 mm. The probe is lowered at a constant speed of 90 m / h to a 1000-meter section of borehole ZK-2024 for scanning. Before lowering the probe, the water quality inside the borehole is ensured to be clear, and a Wel... The lCAD software unfolds the acquired annular borehole wall image into a planar image, and marks the depth range and azimuth range on the image. The core surface image is acquired by a core scanner, which consists of a camera, light source, slide rail, and rolling shaft. The core sample is taken out of the borehole and placed stably on the rolling shaft of the core scanner. The camera is set 45 cm directly above the core. When the image scanning begins, the two rolling shafts rotate slowly in the same direction, causing the cylindrical core above to roll slowly in the same direction. At this time, the camera begins to continuously scan and image the outer surface of the core.

[0033] The acquired borehole wall images and core surface images were compared to identify common features. The borehole images contain azimuth information; the left side of the image shows depth information, recording the depth location of structural features; the top shows azimuth information, recording the geographical orientation of the developed borehole wall image. When bedding or fractures are perpendicular to the borehole axis, they appear as straight lines on the developed borehole planar image. If they are oblique to the borehole axis, they form a sine curve on the developed image. The larger the angle between the oblique plane and the borehole axis, the larger the amplitude of the sine curve, with peaks corresponding to high points on the oblique plane and troughs corresponding to low points. When common features (oblique fractures) are identified in the borehole images... Afterwards, the developed image of the borehole shows a sine curve. The azimuth corresponding to the peak of this sine curve is found. The precise measurement yields an azimuth of 140°, which corresponds to 40° east of due south. Markings (such as arrows and circles) are used on the borehole wall image to label the fracture and its peak location, along with the azimuth information. When obtaining cores using non-directional coring methods, due to rotation and other factors during the coring process, the core itself does not carry natural azimuth information. When a bedding plane or fracture exists on the core surface perpendicular to the core's axial direction, it will appear as a straight line on the developed image of the core surface. Conversely, if a bedding plane or fracture is oblique to the core's axial direction, it will appear as a sine curve on the developed image of the core surface. The larger the angle between the slope and the core axis, the larger the amplitude of the sine curve. The peak of the curve corresponds to the high point of the slope, and the trough corresponds to the low point of the slope. For the core surface image, based on the fracture characteristics identified in the borehole image, carefully locate the same fracture, mark the peak position of the fracture in the core surface image, and draw a feature marker line. This marker line should accurately pass through the peak position. Map the fracture peak azimuth angle of 140° measured in the borehole image to the corresponding position in the core surface image. In the core surface image, draw an orientation reference line parallel to the previously drawn feature marker line along the core axis. Based on the transmitted azimuth information, synchronously determine the geographical azimuth of the orientation reference line as 140°, and indicate the geographical azimuth information of the orientation reference line in the image.

[0034] S102, Measure the radial deformation data of the core, calculate the radial deformation curve based on the measurement data, and determine the direction of the principal stress based on the position of the peak and trough of the radial deformation curve;

[0035] Further, place the core to be measured horizontally at the center of a high-precision laser ranging platform and fix it with a special fixture to ensure that the core axis coincides with the platform rotation axis, avoiding offset or vibration during the scanning process. The laser ranging platform has a micrometer-level positioning accuracy and a stable rotation mechanism. Using the already marked directional reference line on the core surface (geographical azimuth 140°, that is, 40° south by west) as the starting measurement reference line, define this direction as the rotation angle 0°. The rotation angle starts from the directional reference line, and the clockwise direction is positive, with a range of 0° - 360°. The radial deviation of a certain point on the core surface from the ideal circumference is the radial deformation. The radial radius expanding outwards is positive, and contracting inwards is negative. Use a laser transceiver to continuously rotate and scan around the core for 360°, and collect diameter data every 0.17°. A total of data points are collected. The formula for the measured diameter changing with the direction rotation angle is: D θ = D0 + 2Δ rθ , where D θ is the measured diameter of the core at the rotation angle θ, D0 is the ideal diameter of the core (when there is no deformation), Δ rθ is the radial deformation amount of the core at the rotation angle θ, 2Δ rθ is the total deformation amount in the diameter direction. After the core is affected by in-situ stress, the cross-section changes from a circle to an ellipse, and its diameter will change with the direction rotation angle. The measured diameter is equal to the ideal diameter plus the deformation amount in the diameter direction. If the core expands in a certain direction (Δ rθ > 0), then D θ > D0; if it contracts (Δ rθ < 0), then D θ < D0, which is related to the principal stress direction. The peak (maximum diameter) of the radial deformation curve corresponds to the long axis direction of the ellipse, that is, the maximum horizontal principal stress direction, and the trough (minimum diameter) corresponds to the short axis direction of the ellipse, that is, the minimum horizontal principal stress direction. Fit the obtained diameter data product with a sine function to obtain the radial deformation curve. The formula is: Δ rθ = A·sin(θ - φ)+B, where Δ rθ is the radial deformation amount of the core at the rotation angle θ, A is the deformation amplitude (half-wave amplitude), θ is the rotation angle, φ is the phase angle (phase shift of the sine curve), B is the average deformation amount (vertical shift of the sine curve). After the core is affected by the horizontal principal stress (σ1 > σ2), the cross-section deforms into an ellipse, and its radial deformation amount Δ rθThe direction θ varies sinusoidally; the principle for determining the direction of principal stress is as follows: under the action of the maximum horizontal principal stress (σ1), the core will form an elliptical cross-section, and its major axis is consistent with the direction of σ1; the position of the peak of the radial deformation curve (110° rotation angle) corresponds to the direction of the major axis of the ellipse, that is, the direction of the maximum horizontal principal stress. The geographical azimuth of the starting reference point is 140°, and the peak rotation angle is 110°. Then the geographical azimuth of the maximum principal stress is α=140°+110°=250°. 250° corresponds to 70° south of due west (270°-250°=20°, that is, 20° south of due west, but usually described as NW70° (70° northwest, that is, rotating 70° clockwise from due north to the direction of principal stress) with north as the reference.

[0036] S103, the core was cut according to the radial deformation data, and the end face was marked along the crest and trough direction. The three faces of the cut core were parallel to the maximum horizontal principal stress, the minimum horizontal principal stress and the vertical direction, respectively. Strain gauges were set on the three faces to measure the triaxial strain. The principal strain ratio was calculated based on the triaxial strain, and the absolute values ​​of the three principal stresses were calculated based on the differential stress and the principal stress ratio.

[0037] Specifically, the differential strain method is based on the microfracture closure characteristics of rock cores under hydrostatic pressure. When confining pressure is applied to oriented rock samples, the strain differences generated in different directions reflect the anisotropic characteristics of the original geostress field. After stress release, hydrostatic pressure is applied to the core, causing the microfractures inside the core to close, such as... Figure 6 As shown, by collecting strain data in nine directions on three adjacent, orthogonal surfaces of the rock sample, the normal strain and shear strain of the sample during loading can be calculated using these nine directions of strain data. The formula is as follows:

[0038]

[0039] Where, ε xx ε yy ε zz ε represents the normal strain in the x, y, and z directions, respectively, where the x, y, and z directions are arbitrary; xy ε yz ε xz ε1 represents the shear strain in the xy, yz, and xz planes, respectively; ε2 to ε3 represent the strain values ​​measured by nine strain gauges attached to the cubic specimen. The strain matrix is ​​determined based on the calculated normal and shear strains, and the principal strains in the three directions are further obtained using the following cubic equation:

[0040] In this equation, ε represents the principal strain, and the equation has three roots, corresponding to the three principal strains ε. 11 ε 22 ε 33 , ε 11>ε 22 >ε 33 .

[0041] Based on the radial deformation data, the major axis direction and the starting baseline are obtained. These are then marked on the core end face. Figure 5 As shown, a straight line parallel to the major axis of the core surface is drawn through the center of the end face. This line represents the projection direction of the maximum horizontal principal stress on the end face, i.e., the major axis marking line. A straight line perpendicular to the major axis marking line is also drawn through the center of the end face. This line represents the projection direction of the minimum horizontal principal stress on the end face. The direction perpendicular to the core axis of the end face is the vertical principal stress. Using a CNC diamond wire cutter, the cylindrical core is cut and ground into cubic specimens with sides of 50 mm along the marked major and minor axes on the end face. The three cubic specimens are then divided into three groups. A vertical plane was parallel to the directions of the maximum horizontal principal stress, the minimum horizontal principal stress, and the vertical principal stress, respectively, resulting in an oriented cubic rock sample with known principal stress directions. By constructing a collaborative analytical framework for the radial deformation induced by core stress relief and the differential strain response characteristics under hydrostatic pressure, a method for measuring geostress by fusing core radial deformation and differential strain analysis was proposed. After stress release, the core radial deformation becomes elliptical, with the major axis aligned with the direction of the maximum horizontal principal stress and the minor axis aligned with the direction of the minimum horizontal principal stress; that is, the core radial deformation information includes information about the horizontal principal stress direction. Figure 7 As shown, the major axis of the ellipse is designated as the x-axis, the minor axis as the y-axis, and the vertical direction as the z-axis. A resistance strain gauge is placed on each of the three mutually perpendicular faces of the cubic specimen. The specimen with the strain gauges is placed in a high-pressure confining cylinder. The strain gauge leads are connected to a static resistance strain acquisition instrument (DataTaker DT80) outside the cylinder via a plug. Hydrostatic pressure is applied at a constant rate until it exceeds the estimated vertical stress of the stratum where the rock sample is located. The formulas for normal strain and shear strain can be simplified as follows: Among them, ε1 to ε9 are the nine strain values ​​measured in the experiment, ε xx ε yy ε zz The normal strains in the x, y, and z directions are respectively. The differential strain method for geostress measurement changes the number of strain components to be measured from nine to three locations, and the three normal strain measurement results satisfy a certain proportional relationship, expressed by the formula: ε xx :ε yy :ε zz = a:b:c, where a, b, and c are the actual measured ratios, such as Figure 6 As shown, conventional differential strain analysis requires attaching nine strain gauges according to the pattern. However, due to the need for directional cutting of cubic specimens for core radial deformation analysis, [further details are needed]. Figure 7As shown, given that the principal stress direction is perpendicular to the specimen end face, three strain gauges are sufficient. Based on linear elasticity theory and considering small deformations, the following results are obtained when stretched along the directions of maximum and minimum horizontal principal stress:

[0042]

[0043]

[0044] Where, ε max ε is the tensile strain in the direction of the maximum horizontal principal stress in the core. min The tensile strain in the direction of the minimum horizontal principal stress of the core is given by ε, which is the tensile strain along the direction of the minimum horizontal principal stress. Based on the positional relationship between the major and minor axes (x and y axes) of the radial deformation of the core and the direction of the horizontal principal stress, ε is... max =ε xx ε min =ε yy E is the elastic modulus of the rock, ν is Poisson's ratio, and σ max For the maximum horizontal principal stress in the core, σ min For the minimum horizontal principal stress, σ Z For the vertical principal stress of the rock core, according to Hooke's law, we get: Where, ε zz This represents the strain value in the vertical direction.

[0045] At the same time, ε max and ε min It can be calculated from the core diameter deformation:

[0046]

[0047] Where d0 is the initial diameter of the core, d max d is the diameter along the major axis of the ellipse after radial deformation of the core. min It is the diameter along the minor axis of the ellipse after radial deformation of the rock core.

[0048] Based on the above radial deformation and principal strain ratio of the core sample, and derived from elasticity theory, we obtain:

[0049]

[0050] Based on the above formula, after obtaining the major and minor axis diameters and principal strain ratios of the elliptical deformation of the rock core, the maximum horizontal principal stress σ can be calculated. max Minimum horizontal principal stress σ min and vertical principal stress σ z The absolute value of.

[0051] The technical solutions described in the embodiments of this application above have at least the following technical effects or advantages: By integrating core radial deformation analysis (DCDA) and differential strain analysis (DSA), accurate inversion of geostress measurement from two-dimensional to three-dimensional is achieved, with significant technical effects: First, based on feature matching between borehole images and core surface images (such as fracture sinusoidal curve peak positioning), low-cost and high-precision orientation of the core (azimuth error < 3°) is achieved, breaking through the cost limitations of traditional paleomagnetic orientation; second, a high-precision laser ranging platform (accuracy up to 0.001 mm) is used to obtain... Radial deformation data is collected, and the direction of principal stress is accurately determined by fitting a sine curve (e.g., the peak corresponds to the maximum horizontal principal stress). Combined with a simplified differential strain test of directional rock samples (the number of strain gauges is reduced from 9 to 3), the ratio of the three principal stresses is directly obtained. Finally, the absolute value of the principal stress is calculated by data fusion, which solves the limitation of traditional DSA in being unable to independently determine the value. This method is simple to operate, low in cost, and applicable to deep "three-high" environments, providing reliable parameter support for the stability analysis of underground engineering. The measured results deviate from the hydraulic fracturing method by only 3°–17°, verifying its engineering reliability.

[0052] Example 2: Example 1 above only uses static images and single measurement data, without considering the temporal changes in stress release during the coring process. This example uses a multi-source heterogeneous data intelligent fusion framework to integrate multi-dimensional information such as temporal deformation data, environmental parameters, equipment status and geological characteristics, and achieves high-precision dynamic correction of the principal stress direction through machine learning algorithms.

[0053] S201 collects time-series deformation, environmental parameters, equipment status and geological feature data, constructs sub-networks based on the collected data, and merges the sub-networks to obtain a fused feature vector;

[0054] Specifically, a laser ranging system was used to collect radial deformation data, recording the rotation angle and timestamp. Sensors were used to collect environmental data: a PT100 platinum resistance thermometer for temperature, a capacitive humidity sensor for humidity, and a triaxial accelerometer for vibration. The environmental data collection frequency was the same as that for the radial deformation data. Geological features were extracted by calculating fracture density, using the following formula: Where F is the fracture density, N C L represents the number of cracks. t To determine the core length, outliers were removed from the collected data. The time-series deformation, environmental parameters, equipment status, and geological feature data were standardized to eliminate the influence of dimensions. Wavelet transform was performed on the time-series deformation using Daubechies 4, with the decomposition level set to 3, decomposing the signal into wavelet coefficients of different frequencies.

[0055] The system employs a temporal feature extraction subnetwork (LSTM), a two-layer LSTM network, to extract filtered deformed data as input and output a temporal feature vector. An environmental compensation network (MLP) is constructed based on environmental parameters. The MLP is a multilayer perceptron with a three-layer fully connected structure, taking environmental parameters as input and outputting an environmental impact vector. A geological prior network (CNN) is constructed based on geological feature data. The CNN's input layers are two-channel images of fracture density and bedding direction, with two 3×3 convolutional layers (32-64 channels) and a stride of 2. The fully connected layers output the geological feature vector. An attention mechanism is used to weight and fuse the temporal features, environmental compensation, and geological prior features, outputting a predicted value for the principal stress direction. The formula for calculating the attention weights is as follows: Where, α i For attention weights, f i f is the eigenvector. q Let W be the query vector and W be the weight matrix. Multiply the feature vectors of each feature source with their corresponding attention weights to obtain a weighted feature vector. Then sum the weights to obtain the fused feature vector.

[0056] S202, construct a prediction model based on the fused feature vector, predict the principal stress direction based on the prediction model, compare the predicted principal stress direction with the actual measured principal stress direction, and if the error exceeds the preset threshold, correct the predicted principal stress direction.

[0057] Furthermore, a Seq2Seq (sequence-to-sequence) architecture is used to construct a prediction model. The constructed prediction model is trained by dividing the collected multi-source heterogeneous data into training, validation, and test sets. The deep neural network fusion model is trained using the training set, and the backpropagation algorithm and gradient descent optimizer are used to update the model parameters. The trained model is validated using the validation set, and its performance metrics, such as directional accuracy (MAE) and stability (variance ratio), are evaluated. Based on the model's performance on the validation set, hyperparameters, such as learning rate, batch size, and number of network layers, are adjusted to improve model performance. The fused feature vector is used as input to the prediction model. The prediction model outputs the estimated value of the principal stress direction at the next time step, i.e., the predicted value of the principal stress direction, based on the input and internal algorithm. The predicted value of the principal stress direction is compared with the real-time observation value, and a threshold is set. If the error obtained from the comparison exceeds the preset threshold, the parameters of the prediction model are adjusted and optimized.

[0058] The technical solutions described in the above-mentioned embodiments of this application have at least the following technical effects or advantages: by using multi-dimensional information such as time-series deformation data, environmental parameters, equipment status and geological characteristics, and through machine learning algorithms, the complex behavior of rock cores during stress release is reflected more accurately, thereby significantly improving the measurement accuracy of principal stress direction. In the process of deep resource development, the geostress direction of underground rock mass can be accurately determined to optimize roadway support design and mining scheme, and ensure the safety and efficiency of mining.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for measuring in-situ stress based on the fusion of core radial deformation and differential strain, characterized in that, include: S101, acquire borehole wall images and core surface images, identify common features based on borehole wall images and core surface images, and map the orientation information of the common features in the borehole wall images to the core surface images; S102, Measure the radial deformation data of the core, calculate the radial deformation curve based on the measurement data, and determine the direction of the principal stress based on the position of the peak and trough of the radial deformation curve; S103. Cut the core according to the radial deformation data, mark the end face along the crest and trough direction. The three faces of the cut core are parallel to the maximum horizontal principal stress, the minimum horizontal principal stress and the vertical direction, respectively. Strain gauges are set on the three faces to measure the triaxial strain. Calculate the principal strain ratio based on the triaxial strain. Calculate the absolute values ​​of the three principal stresses based on the radial deformation of the core and the principal strain ratio.

2. The in-situ stress measurement method based on the fusion of core radial deformation and differential strain according to claim 1, characterized in that, The steps to map the azimuth information of common features of borehole wall images to core surface images are as follows: mark the position and azimuth of fracture peaks in the borehole wall images; mark the position of the peaks of the same fracture in the core surface images and draw feature marker lines; map the azimuth information in the borehole images to the core surface images, draw orientation reference lines and mark the geographical orientation.

3. The in-situ stress measurement method based on the fusion of core radial deformation and differential strain according to claim 1, characterized in that, The calculation formula of the radial deformation curve is: Δ rθ = A sin (θ - φ) + B, wherein Δ rθ is the radial deformation amount of the core at the rotation angle θ, that is, the radial deformation curve, A is the deformation amplitude, θ is the rotation angle, φ is the phase angle, and α is the average deformation amount.

4. The in-situ stress measurement method based on the fusion of core radial deformation and differential strain according to claim 1, characterized in that, The formula for calculating the principal strain is: Wherein, ε1 to ε9 are 9 strain values measured in the experiment, ε xx , ε yy , and ε zz are normal strains in x, y, and z directions respectively, and the stress measurement test by the differential strain method is changed from 9 strain components to be measured to 3 positions to be measured, and the formula for three normal strain measurement results to satisfy the proportional relationship is: ε xx : ε yy : ε zz =a:b:c, wherein a, b, and c are actual measurement ratios.

5. The geostress measurement method based on the fusion of core radial deformation and differential strain as described in claim 1, characterized in that, S201 collects time-series deformation, environmental parameters, equipment status and geological feature data, constructs sub-networks based on the collected data, and merges the sub-networks to obtain a fused feature vector; S202, construct a prediction model based on the fused feature vector, predict the principal stress direction based on the prediction model, compare the predicted principal stress direction with the actual measured principal stress direction, and if the error exceeds the preset threshold, correct the predicted principal stress direction.

6. The in-situ stress measurement method based on the fusion of core radial deformation and differential strain according to claim 5, characterized in that, The subnetworks include a long short-term memory network, an environmental compensation network, and a geological prior network.

7. The in-situ stress measurement method based on the fusion of core radial deformation and differential strain according to claim 5, characterized in that, The extraction of the geological features is achieved by calculating the fracture density, and a calculation formula of the fracture density is as follows: Wherein, F is the fracture density, N C is the fracture quantity, and L t is the core length.

8. The in-situ stress measurement method based on the fusion of core radial deformation and differential strain according to claim 7, characterized in that, Attention weights are used to weight and fuse temporal features, environmental compensation, and geological prior features. The formula for calculating the attention weights is as follows: Where, α i For attention weights, f i Let f be the feature vector, representing the vectors of temporal features, environmental compensation, and geological prior features, respectively. q Let W be the query vector and W be the weight matrix.

Images