Single-hole multi-segment ground stress fusion quality control method and system

Abstract

This invention provides a single-well multi-segment geostress fusion quality control method and system. The method converts the stress relief inversion results of each segment into a single-segment inversion result package. After invalid segment identification, quality rating, and correlation correction, candidate segment combinations and a set of retained segments are constructed. Tensor-level robust fusion is performed using a six-component stress vector, and cross-validation is conducted using the retained segments. The retained fusion results after validation are evaluated for principal stress magnitude dispersion and principal stress direction axial tensor statistics. Finally, the far-field stress tensor, principal stress magnitude and direction, quality grade, and supplementary measurement suggestions are output, improving the reliability and engineering acceptance of single-well multi-segment geostress results.

Description

Technical Field

[0001] This invention relates to the field of data processing technology for in-situ geostress testing and intelligent detection equipment for underground engineering, specifically to a single-hole multi-segment geostress fusion quality control method and system. Background Technology

[0002] In-situ stress is a crucial fundamental parameter for underground engineering design, surrounding rock stability evaluation, and support parameter determination. The in-situ stress relief method, which uses inclusion strain gauges or equivalent strain measurement devices to obtain stress-release strain and combines it with rock elastic parameters to invert the far-field stress at the measuring point, is a commonly used in-situ stress testing method for hard rock underground engineering.

[0003] In engineering sites, to improve the representativeness of tests, multiple test sections are often set up at different depths in the same borehole. Existing processing methods mostly focus on single-section inversion, abnormal strain elimination, eccentricity or elastic mode error correction, or simple statistical analysis of multiple results. These methods do not fully handle the spatial correlation between adjacent test sections in the same borehole, and also lack an external verification mechanism for candidate fusion results for test sections not involved in the test.

[0004] Adjacent test sections within the same borehole may be controlled by the same lithological segment, the same structural plane, or the same construction disturbance. Treating them simply as independent samples can easily overestimate the reliability of the results. On the other hand, directly merging samples across obvious lithological interfaces, fracture zones, or areas of abrupt tectonic stress change may misinterpret actual spatial stress variations as test anomalies. Furthermore, principal stress directions have axial equivalence; directly averaging over strike or dip angles can easily introduce directional statistical bias. Summary of the Invention

[0005] To address the problems in the prior art, the first aspect of this invention provides a single-well multi-segment geostress fusion quality control method. Without altering the mature single-segment stress relief inversion method, this method standardizes, rates, corrects correlations, combines constrained candidate combinations, performs tensor-level robust fusion, retains segment cross-validation, statistically analyzes principal stress magnitudes, statistically analyzes principal stress axial directions, and provides supplementary measurement feedback on the results from multiple segments. This improves the stability, interpretability, and engineering reliability of the single-well multi-segment geostress results.

[0006] This fusion quality control method includes the following steps:

[0007] Acquire stress relief test data from multiple test sections within the same borehole, and perform strain-stress inversion independently on each test section to form a corresponding single test section inversion result package;

[0008] Each measurement segment is evaluated for quality and assigned a quality control weight based on the inversion result package of a single measurement segment.

[0009] Determine the segment correlation coefficient between any two segments, and form a correlation correction parameter based on the segment correlation coefficient;

[0010] Based on the quality rating, quality control weight, and correlation correction parameters of the test segments, candidate test segment combinations and a set of retained test segments corresponding one-to-one with the candidate test segment combinations are constructed.

[0011] The far-field stress tensor of each segment within the candidate segment combination is converted into a six-component stress vector, and tensor-level robust fusion is performed in the six-component stress vector space to obtain the candidate fused far-field stress tensor.

[0012] External cross-validation is performed on the candidate fused far-field stress tensor using the far-field stress tensor of the retained measurement segment set, and the candidate fused far-field stress tensor is screened based on the cross-validation error.

[0013] Based on the selected candidate fusion far-field stress tensors, the magnitude and dispersion of principal stresses are statistically analyzed and evaluated. The direction of principal stresses and the direction concentration are determined by the axial direction tensor statistics.

[0014] Based on at least three of the following factors: the quality rating of the measurement segment, the quality control weight, the number of valid samples after correlation correction, the dispersion of principal stress magnitude, the concentration of principal stress direction, and the cross-validation error, the geostress inversion quality level and supplementary measurement recommendations are output.

[0015] Furthermore, the single-segment inversion result package includes: far-field stress tensor, inversion quality parameters, segment spatial information, field process information, principal stress magnitude, and principal stress direction; wherein, the segment spatial information includes: segment depth, borehole azimuth, borehole dip angle, and lithological logging information; the inversion quality parameters include: single-segment inversion residual, inversion condition number, and uncertainty parameters; the field process information includes: borehole wall cleaning, core integrity record, inclusion installation record, cementation and solidification record, strain channel integrity record, and strain release curve.

[0016] Furthermore, before assigning quality ratings and quality control weights to each measurement segment, invalid measurement segment identification is also included. This includes: when a measurement segment has missing key strain channels, a lack of stable platforms in the strain release curve, inclusion debonding, cementation failure, single-segment inversion non-convergence, insufficient rank of the inversion matrix, or severely broken core that cannot represent any of the original rock states, the measurement segment is identified as a discarded measurement segment.

[0017] Furthermore, the quality rating includes: determining the quality level of a measurement segment based on field process information, strain release curves, inversion quality parameters, multi-segment consistency quality, and geological interpretation consistency; the quality level of a measurement segment includes reliable measurement segments, usable measurement segments, suspected anomaly measurement segments, and eliminated measurement segments; wherein, reliable measurement segments are assigned a first quality control weight, usable measurement segments are assigned a second quality control weight, suspected anomaly measurement segments are assigned a third quality control weight, and eliminated measurement segments have a quality control weight of zero.

[0018] Furthermore, determining the correlation coefficient between any two measurement segments includes: obtaining the segment spacing, far-field stress tensor difference, and geological zoning information of any two measurement segments, and determining the correlation coefficient between the two measurement segments based on the segment spacing, far-field stress tensor difference, and geological zoning information.

[0019] When the correlation coefficient of two measurement segments is greater than the preset correlation threshold, or when at least two of the following conditions are met simultaneously: the distance between measurement segments is less than the preset correlation length, the difference in far-field stress tensor is less than the preset difference threshold, or they are located in the same local structural control area, the two measurement segments are marked as correlation correction measurement segments.

[0020] Furthermore, when two segments are marked as correlation-corrected segments, the quality control weight of at least one of the segments is updated based on the correlation correction parameter, or one segment is assigned to the candidate segment combination and the other segment is assigned to the reserved segment set.

[0021] Furthermore, the tensor-level robust fusion includes: […]. Far-field stress tensor of each measurement segment Mapped to a six-component stress vector and for the first Candidate segment combinations The six-component stress vector within is robustly estimated as follows:

[0022]

[0023] in, For the first The candidate fused far-field stress six-component vector corresponding to each candidate test segment combination The candidate fused far-field stress six-component vector is to be determined. For known measurement sections The six-component stress vector, To comprehensively consider the segment fusion weight after considering the segment quality weight and the correlation correction parameter, For a robust loss function, The normalized matrix or the inverse square root of the covariance matrix. This is the scale parameter.

[0024] Furthermore, the robust loss function adopts the Huber loss function, the Tukey double-weighted loss function, the weighted least squares loss function, or the iterative reweighted loss function; after obtaining the candidate fused far-field stress six-component vector, it is mapped to a second-order symmetric candidate fused far-field stress tensor.

[0025] Furthermore, the external cross-validation includes: calculating the tensor difference between the candidate fused far-field stress tensor and the far-field stress tensor of each segment in the corresponding retained segment set, and determining the cross-validation error based on the median, mean, quantile, or weighted statistic of the tensor difference; when the cross-validation error exceeds a preset validation threshold, reducing the weight or eliminating the corresponding candidate fused far-field stress tensor; when the cross-validation error does not exceed the preset validation threshold, retaining the corresponding candidate fused far-field stress tensor and assigning candidate fusion weights.

[0026] Furthermore, the statistical analysis and dispersion evaluation of the principal stress magnitudes includes: [the following is a partial sentence fragment:] ...for the selected... Candidate fused far-field stress tensor Perform eigenvalue decomposition to obtain the corresponding eigenvalue. The magnitude of each principal stress ,in According to the first The candidate fusion weights are determined by the cross-validation error of the candidate far-field stress tensor, the quality of the candidate segment combination, and the correlation correction results. The first one is determined according to the following formula. Weighted average of the principal stresses:

[0027]

[0028] And determine the first according to the following formula Weighted dispersion of the principal stress magnitudes:

[0029]

[0030] in, For the first Statistical results of the magnitude of each principal stress. For the first The dispersion of the magnitude of each principal stress.

[0031] Furthermore, the principal stress direction is determined statistically using the axial direction tensor; for each of the screened candidate fusion far-field stress tensors, unit vectors for the principal stress directions of the same order are obtained, and a second-order direction tensor is constructed based on the unit vectors and the corresponding candidate fusion weights:

[0032]

[0033] in, For the first The second-order direction tensor corresponding to each principal stress direction For the first The candidate fused far-field stress tensor corresponding to the first Unit vector in the direction of principal stress. For the corresponding candidate fusion weights; the principal eigenvector of the second-order direction tensor is taken as the average principal stress direction, and the concentration of principal stress directions is determined according to the eigenvalue distribution of the second-order direction tensor; wherein, the direction unit vector and Processed according to the equivalent axial direction.

[0034] A second aspect of this invention provides a single-well, multi-segment geostress fusion quality control system, comprising a field acquisition terminal and a data processing terminal; the field acquisition terminal includes a strain acquisition module, a segment spatial recording module, a field process recording module, and a lithology logging module; the data processing terminal is configured to receive and process data from the strain acquisition module, the segment spatial recording module, the field process recording module, and the lithology logging module, forming a single-segment inversion result package, and completing segment quality rating, correlation correction, candidate segment combination and retained segment set construction, tensor-level robust fusion, retained segment cross-validation, principal stress magnitude and direction statistics, and outputting segment quality level, candidate fusion far-field stress tensor, retained segment verification error, principal stress magnitude and direction statistics results, and supplementary measurement suggestions.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] 1. This invention uses stress relief test data from multiple sections of the same borehole as its object. It standardizes the inversion results of a single section into a result package containing far-field stress tensor, inversion quality parameters, field process records, and spatial information of the section. Before fusion, it introduces invalid section identification, quality rating, and section correlation correction. This method can identify low-quality sections such as missing key strain channels, abnormal release curves, inclusion debonding, and core fragmentation, preventing invalid data from entering the fusion process. Simultaneously, it corrects for duplicate contributions from adjacent sections based on section spacing, tensor differences, and geological zoning relationships, preventing sections with the same borehole height from being mistakenly identified as multiple independent samples, thus improving the reliability and statistical rationality of multi-section geostress results.

[0037] 2. This invention achieves separation of fusion data and external verification data through dynamic partitioning of candidate test segment combinations and retained test segment sets; it employs a six-component stress vector for tensor-level robust fusion, avoiding distortion of tensor relationships caused by directly averaging the magnitude and direction of principal stresses; and it evaluates the dispersion of principal stress magnitudes and performs axial tensor statistics on the verified candidate fusion results. Thus, it not only outputs the final far-field stress tensor, principal stress magnitudes, and directions, but also simultaneously provides the number of valid samples, cross-validation error, quality level, and supplementary testing suggestions, forming a closed loop of "testing—evaluation—fusion—verification—feedback" for engineering acceptance. Attached Figure Description

[0038] Figure 1 This is a flowchart of the single-hole multi-segment geostress fusion quality control method of the present invention;

[0039] Figure 2 This is a schematic diagram of the data structure of the single-segment inversion result package of the present invention;

[0040] Figure 3 This is a flowchart of the section quality rating and weight assignment process of the present invention;

[0041] Figure 4 This is a schematic diagram of the segment correlation correction of the present invention;

[0042] Figure 5 This is a schematic diagram illustrating the construction of the candidate segment combination and the reserved segment set of the present invention;

[0043] Figure 6 This is a schematic diagram of the tensor-level robust fusion and segment-retaining cross-validation of the present invention;

[0044] Figure 7 This is a schematic diagram illustrating the statistical analysis of principal stress magnitudes, principal stress direction tensors, and engineering acceptance levels of the present invention.

[0045] In the diagram: 100, borehole; 101, section; 102, retained section; 103, geological-stress evaluation interval; 104, removed section. Detailed Implementation

[0046] To facilitate understanding of the present invention, 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.

[0047] It is understandable that the following embodiments are used to explain the technical solutions of the present invention and are not used to limit the protection scope of the present invention. For the number of measurement segments, the spacing between measurement segments, the quality grading threshold, the robust fusion function, the form of the correlation coefficient, the verification error threshold, and the quality grade criterion, those skilled in the art can make appropriate adjustments according to the engineering geological conditions, the accuracy of the test instrument, and the purpose of engineering evaluation. As long as they still adopt the single measurement segment inversion result package, measurement segment quality rating, correlation correction, constrained candidate combination, magnitude-level robust fusion, retained measurement segment cross-validation, and supplementary measurement feedback logic described in the present invention, they should fall within the protection scope of the present invention.

[0048] A single-hole multi-measurement segment in-situ stress fusion quality control method provided by some embodiments of the present invention includes the following steps:

[0049] S1. Obtain the stress relief test data of multiple measurement segments in the same borehole, and independently perform strain-stress inversion on each measurement segment to form corresponding single measurement segment inversion result packages;

[0050] Specifically, set measurement segments 101 with different depths in the same borehole 100, where . Each measurement segment 101 uses stress relief testing to obtain released strain data. The stress relief test can be hollow inclusion overcoring stress relief, strain gauge overcoring stress relief at the bottom of the hole, or other equivalent test methods that can obtain three-dimensional stress relief strain data.

[0051] The method can be executed by a data processing terminal, a field acquisition host or a server supporting the in-situ stress test equipment. The data processing terminal receives the data of the strain acquisition module, the measurement segment space recording module, the field process recording module, and the lithology cataloging module, and outputs the measurement segment quality grade, the candidate fusion far-field stress tensor, the retained measurement segment verification error, the statistical results of the principal stress magnitude and direction, and the supplementary measurement suggestions.

[0052] Each measurement segment 101 independently obtains the far-field stress tensor according to the strain-stress inversion model corresponding to its strain measurement device, and forms a single measurement segment inversion result package :

[0053]

[0054] Among them, ; is the far-field stress tensor independently inverted by the measurement segment; are the principal stress magnitude and principal stress direction of this measurement segment; is the single measurement segment inversion residual or fitting quality index; is the uncertainty, covariance or inversion condition number information; Record the process on site; This includes information on the depth of the borehole section, borehole azimuth, borehole dip angle, and the lithological zone in which it is located.

[0055] In this embodiment, the single-segment inversion result package refers to the standardized data set independently obtained after each segment completes the stress relief test, according to the strain-stress inversion model adapted to its strain measurement device. This embodiment does not limit the specific instrument model used in a single hole stress relief test; hollow ingots, hole bottom strain gauges, and other equivalent testing devices capable of acquiring three-dimensional stress relief strain data can all be used as single-segment data sources.

[0056] The single-segment inversion results package includes: far-field stress tensor, inversion quality parameters, segment spatial information, field process information, principal stress magnitude, and principal stress direction. The segment spatial information includes: segment depth, borehole azimuth, borehole dip angle, and lithological logging information; the inversion quality parameters include: single-segment inversion residuals, inversion condition number, and uncertainty parameters; the field process information includes: borehole wall cleaning, core integrity record, inclusion installation record, cementation and solidification record, strain channel integrity record, and strain release curve.

[0057] S2. Based on the inversion result package of a single measurement segment, perform quality rating and assign quality control weights to each measurement segment;

[0058] Specifically, before assigning quality ratings and quality control weights to each measurement segment, an invalid segment determination process is also included. If the inversion quality parameters or field process information are unqualified, the measurement segment is determined to be segment 104 and will not participate in subsequent tensor-level robust fusion. Unqualified inversion quality parameters or field process information include missing key strain channels, lack of stable plateaus in strain release curves, inclusion debonding, cementation failure, single-segment inversion non-convergence, insufficient rank of the inversion matrix, and severely fragmented core that cannot represent the original rock state.

[0059] For valid measurement segments, a quality score is awarded based on the quality of the field process, the quality of the strain release curve, the quality of single-segment inversion, the consistency quality of multiple measurement segments, and the consistency of geological interpretation.

[0060]

[0061] in, For the first The overall quality score of each test section; It indicates the quality of the on-site process and is used to evaluate the cleanliness of the borehole wall, the installation position of the inclusions, the bonding and curing status, the process of removing the casing, and the integrity of the core. It indicates the quality of the strain release curve and is used to evaluate the initial stable section, the release of change section, the final stable plateau, signal jumps, and channel integrity. It represents the inversion quality of a single measurement segment and is used to evaluate inversion residuals, goodness of fit, condition number, and convergence status; It represents the consistency quality of multiple measurement segments and is used to evaluate the consistency of the far-field stress tensor of the measurement segment with the overall trend of adjacent measurement segments or the same evaluation interval. It indicates the consistency of geological interpretation and is used to determine whether anomalies in a section can be explained by structural planes, lithological abrupt changes, fracture zones, or local stress disturbances. to These are the weighting coefficients.

[0062] Based on the invalid segment identification and quality scoring results, segments are divided into reliable segments, usable segments, suspected abnormal segments, and eliminated segments. Reliable segments are assigned the first quality control weight, usable segments are assigned the second quality control weight, suspected abnormal segments are assigned the third quality control weight or are prioritized as retained verification segments, and eliminated segments have a weight of zero. The first quality control weight is greater than the second quality control weight, and the second quality control weight is greater than the third quality control weight.

[0063] S3. Determine the correlation coefficient between any two measurement segments, and form a correlation correction parameter based on the correlation coefficient.

[0064] Specifically, determining the correlation coefficient between any two measurement segments includes: obtaining the segment spacing, far-field stress tensor difference, and geological zoning information of any two measurement segments, and determining the correlation coefficient between the two measurement segments based on the segment spacing, far-field stress tensor difference, and geological zoning information.

[0065] The first Far-field stress tensor of each measurement segment Convert to a six-component stress vector The far-field stress tensor can be expressed as:

[0066]

[0067] in, , , For the first The three normal stress components obtained from the inversion of each measurement segment; , , For the first The three shear stress components were obtained from the inversion of the test segment.

[0068] The corresponding six-component stress vector is:

[0069]

[0070] For any two test segments and Define normalized tensor differences for:

[0071]

[0072] in, and The first The first test segment and the first The six-component stress vector corresponding to each test segment; and It is a normalized matrix or the inverse square root of the covariance matrix; To prevent positive numbers with a denominator of zero; Used to represent the degree of difference between the inversion stress results of two measurement segments in tensor space.

[0073] set up For the first The first test segment and the first The axial depth spacing between each measuring segment represents the physical spatial distance along the drilling direction; let... This is a geological zoning item used to indicate whether two survey sections are located in the same lithological continuum, the same structural control zone, or the same tectonic disturbance zone; [The following is a list of geological zoning items:] ... Let the correlation coefficient be the segment correlation coefficient, then:

[0074]

[0075] In one specific implementation, It can be in the following form:

[0076]

[0077] in, To preset the relevant length, Preset tensor difference scale. Segment spacing. The smaller the tensor difference The smaller and geological zoning item The larger the value, the higher the correlation coefficient of the measurement segment. The larger.

[0078] When the correlation coefficient between two measurement sections The correlation is greater than the preset correlation threshold, or at least two of the following conditions are met simultaneously: segment spacing Less than the preset correlation length, far-field stress tensor normalization difference If the difference is less than the preset difference threshold and the two segments are located in the same local structure control area, the two segments are marked as correlation correction segments.

[0079] In this embodiment, the initially allocated quality control weight refers to the weight assigned to the measurement segment based on its quality level before correlation correction. The correlation correction parameter refers to the weight assigned based on the measurement segment's correlation coefficient. Intersection spacing Tensor differences and geological zoning items The obtained parameters are used to correct the segment fusion weights, effective sample size, or confidence assessment.

[0080] When two segments are marked as correlation-corrected segments and simultaneously enter the same candidate segment combination, they are not treated as completely independent high-contribution samples. Instead, the segment fusion weight of at least one segment is relevance-corrected. For example, the segment fusion weight of the first segment can be adjusted accordingly. Initial quality control weights for each test segment Revised to:

[0081]

[0082] in, The corrected segment fusion weights, For the first Each test segment is initially assigned a quality control weight based on its quality level. For the first The first test segment and the first The correlation coefficient between the measurement segments This is the correlation reduction factor, and .

[0083] In another embodiment, the following modified form may also be adopted:

[0084]

[0085] The above method is used to avoid two highly correlated test segments participating in the fusion with the initially assigned quality control weights in the same candidate test segment combination, thereby preventing adjacent test segments in the same hole from being counted as completely independent samples repeatedly.

[0086] S4. Based on the quality rating, quality control weight and correlation correction parameters of the test segment, construct candidate test segment combinations and a set of retained test segments that correspond one-to-one with the candidate test segment combinations;

[0087] Specifically, based on lithological continuity, structural plane development, borehole logging, core integrity, expected stress gradient, and engineering evaluation objectives, the same borehole is divided into one or more geological-stress evaluation intervals 103. These geological-stress evaluation intervals refer to a set of sections that are comparable in terms of lithology, structural plane development, stress gradient, and engineering evaluation objectives.

[0088] Candidate segment combinations should be constructed within the same geological-stress evaluation interval. For segments that cross obvious lithological boundaries, fault fracture zones, tectonic stress abrupt change zones, or strongly heterogeneous zones, they should not be directly included in the same candidate segment combination for fusion; they can be fused in sections first, or a stress gradient correction model that varies with depth can be introduced before fusion.

[0089] Dynamically construct the first effective measurement segment Candidate segment combinations And set a corresponding set of reserved segments for the candidate segment combination. The candidate test segment combination Used to participate in the current candidate fusion; the reserved segment set It does not participate in the candidate fusion, but is used for external cross-validation of the candidate fusion result.

[0090] Candidate segment combination and the set of reserved test sections The segment can be dynamically changed using the leave-one-out method, group retention method, correlation-based hierarchical retention method, or random retention method. A segment can be used as a fusion segment in one candidate combination and as a reserved segment in another candidate combination; however, for the same candidate segment combination, segments in the reserved segment set do not participate in the current fusion calculation of that candidate segment combination.

[0091] In some embodiments, candidate segment combinations At least the following constraints must be satisfied:

[0092] The number of valid measurement segments shall not be less than the preset number;

[0093] The proportion or weight of trusted test segments and available test segments shall not be less than the preset proportion;

[0094] When two correlation-corrected segments are selected into the same candidate segment combination, the segment fusion weight of at least one of the correlation-corrected segments is updated or reduced based on the corresponding correlation correction parameter.

[0095] In addition to the candidate test segment combinations, at least one test segment is reserved as a reserved test segment set for external cross-validation.

[0096] In some embodiments, the candidate segment combination shall at least satisfy one of the following constraints: depth coverage constraint within the same geological-stress evaluation interval, proportion constraint of suspected abnormal segments, and influence coefficient constraint of single segment.

[0097] S5. Convert the far-field stress tensor of each segment in the candidate segment combination into a six-component stress vector, and perform tensor-level robust fusion in the six-component stress vector space to obtain the candidate fused far-field stress tensor.

[0098] Specifically, regarding the first Candidate segment combinations The far-field stress tensors of each measurement segment are robustly fused at the tensor level. To avoid directly averaging the principal stress magnitudes, azimuth angles, or dip angles, which would disrupt the physical consistency of the stress tensor, this embodiment first fuses the far-field stress tensors of each measurement segment. Mapped to a six-component stress vector Then, robust fusion is performed in the six-component stress vector space.

[0099] For candidate segment combinations Candidate fused far-field stress six-component vector Determine using the following formula:

[0100]

[0101] in, For the first The candidate fused far-field stress six-component vector corresponding to each candidate segment combination; The candidate fusion far-field stress six-component vector to be determined; For the first The known six-component stress vector obtained by independent inversion of each test segment; For the first The comprehensive fusion weight of each test segment; A robust loss function; It is a normalized matrix or the inverse square root of the covariance matrix; This is the scale parameter.

[0102] The physical meaning of the above formula is: to find a candidate fused stress vector within the same geological-stress evaluation interval. Combine it with candidate test segments Known stress vectors for each internal measurement section The weighted robust residuals are minimized overall. Unlike direct averaging, this formula can achieve the following: Expressing the results of segment quality differences and correlation corrections, through Suppressing suspected abnormal segments, by and Eliminate differences in scale and uncertainty of different stress components.

[0103] In some embodiments, when the weighted least squares loss function is used, the residuals of each segment are calculated in square form; when the Huber loss function is used, small residuals are processed in square form and large residuals are processed in linear form to reduce the impact of abnormal segments; when the Tukey double-weight loss function is used, abnormal residuals exceeding the set scale can be strongly reduced in weight or even approximately eliminated.

[0104] Seek Then, it is mapped to the candidate fused far-field stress tensor according to the correspondence between the six components and the second-order symmetric tensor. :

[0105]

[0106] in, , , These correspond to the three normal stress components of the candidate fusion far-field stress, respectively. , , These correspond to the three shear stress components of the candidate fusion far-field stress.

[0107] S6. Using the far-field stress tensor of the measured segments in the retained measured segment set, perform external cross-validation on the candidate fused far-field stress tensor.

[0108] Specifically, for the first Candidate segment combinations The obtained candidate fused far-field stress tensor Using its corresponding set of preserved test segments External cross-validation is performed. The reserved test segment set... The test segment in the middle does not participate in the first The robust fusion calculation of each candidate segment combination is only used to verify whether the candidate fusion result can explain the results of segments that did not participate in the fusion.

[0109] No. The external cross-validation error of each candidate fusion result can be expressed as:

[0110]

[0111] in, For the first External cross-validation error of each candidate fusion result; For the first One candidate fused far-field stress six-component vector; To preserve the set of test segments The Middle The known six-component stress vector of the retained test section; For the first The normalized matrix or inverse square root of covariance matrix corresponding to each retained measurement segment; For scale parameters; The statistical unit can be represented by the mean, median, quantile, maximum value, or weighted statistical value.

[0112] when Not exceeding the preset verification threshold At that time, it is explained that the candidate fused far-field stress tensor If the retained segments that did not participate in the fusion have acceptable interpretability, the candidate fusion result is retained, and a candidate fusion weight is assigned. .

[0113] when Exceeding the preset verification threshold If the candidate fusion result is negative, it indicates that the combination may have internal consistency but insufficient external explanatory power, and the candidate fusion result should be downweighted or eliminated.

[0114] Candidate fusion weights It can be determined in the following way:

[0115]

[0116] in, For the first Candidate fusion weights for each candidate fusion result; For the first The combination quality coefficient of each candidate segment combination is used to comprehensively characterize the proportion of reliable segments, the proportion of available segments, the proportion of suspected abnormal segments, the depth of coverage, and the degree of correlation correction within the candidate combination. To prevent positive numbers with a denominator of zero, it is also possible to... Set to with Other forms of functions that are monotonically decreasing.

[0117] Using the above method, only candidate fusion results that simultaneously satisfy internal robustness and external interpretability will be included in the final far-field stress tensor, principal stress magnitude statistics, and principal stress direction statistics with higher weights.

[0118] S7. Determine the final far-field stress tensor based on external cross-validation, perform principal stress magnitude statistics and dispersion evaluation, and determine the principal stress direction and direction concentration through axial direction tensor statistics.

[0119] Specifically, the six-component far-field stress vectors of multiple candidates that have passed external cross-validation and are retained are analyzed. Weighted fusion is performed according to the corresponding candidate fusion weights to obtain the final six-component vector of far-field stress. :

[0120]

[0121] in, This represents the final six-component vector of far-field stress. For the first Candidate fusion weights for each candidate fusion result.

[0122] Will Mapped to the final far-field stress tensor :

[0123]

[0124] For the final far-field stress tensor Perform eigenvalue decomposition:

[0125]

[0126] in:

[0127]

[0128] , , These are the final maximum principal stress, intermediate principal stress, and minimum principal stress, respectively. It is an orthogonal matrix composed of unit vectors of the three principal stress directions.

[0129] To evaluate the stability of principal stress magnitudes across different candidate fusion results, the far-field stress tensor of each retained candidate fusion after screening was used. Perform eigenvalue decomposition:

[0130]

[0131] in:

[0132]

[0133] For the The principal stresses, and their weighted statistical values ​​are:

[0134]

[0135] Its weighted dispersion is:

[0136]

[0137] in, Indicates the first Statistical results of the magnitude of each principal stress; Indicates the first The degree of dispersion of principal stress magnitudes among candidate fusion results retained after multiple screenings. The smaller the value, the more stable the principal stress. The larger the value, the more significantly the principal stress is affected by candidate combinations, local anomaly segments, or geological zones.

[0138] For the principal stress directions, the arithmetic mean of the azimuth or inclination angles cannot be directly calculated; instead, axial direction tensor statistics should be used. Let the first... Among the candidate fused far-field stress tensors, the first one is... The unit vectors in the principal stress directions are: The corresponding candidate fusion weights are Then the first The second-order direction tensor corresponding to each principal stress direction is:

[0139]

[0140] in, For the first The second-order direction tensor corresponding to each principal stress direction; It is the outer product of the unit vectors in the direction. Because and For the same principal stress axis, the two have the same outer product. Therefore, the above second-order direction tensor can eliminate the angle averaging error caused by the equivalence of positive and negative principal stress directions.

[0141] right Perform eigenvalue decomposition and take the eigenvector corresponding to its largest eigenvalue as the eigenvector of the eigenvalue. The average axial direction of the principal stresses .

[0142] like for The three eigenvalues ​​can be used to characterize the directional concentration in the following form:

[0143]

[0144] in, For the first Concentration of each principal stress direction; The larger it is, the more likely it is to be the first The more concentrated the principal stresses are in the direction of the stress; The smaller the value, the higher the degree of dispersion in the principal stress direction, and the acceptance level of the results should be reduced or supplementary testing and verification should be recommended.

[0145] S8. Based on at least three of the following factors: the quality rating of the measurement section, the quality control weight, the number of valid samples after correlation correction, the dispersion of principal stress magnitude, the concentration of principal stress direction, and the cross-validation error, output the geostress inversion quality level and supplementary measurement recommendations.

[0146] Specifically, let the fusion weight of the measurement segments participating in the statistics be... The correlation coefficient of the measurement segment is The number of effective samples after correlation correction It can be represented as:

[0147]

[0148] To avoid the number of valid samples exceeding the actual number of measured segments, it can be further limited to:

[0149]

[0150] in, This refers to the number of valid measurement segments included in the statistics. The amount of statistically independent information is not equivalent to the actual number of test segments in the field. When multiple test segments are highly correlated, even if the individual test quality of each segment is high, they should not be treated as multiple completely independent samples to excessively narrow the confidence interval.

[0151] The final geostress inversion quality grade is determined based on at least three of the following indicators: number of valid segments, percentage of reliable and usable segments, number of segments removed, percentage of suspected anomaly segments, and number of valid samples after correlation correction. Discreteness of candidate fused far-field stress tensor, and external cross-validation error Dispersion of principal stress magnitude Concentration of principal stress directions And the influence coefficient of a single measurement section.

[0152] In some embodiments, if the number of valid test segments is sufficient and the proportion of reliable and usable test segments is high, then... Meets preset requirements; external cross-validation error Smaller, with less dispersion in the magnitudes of the three principal stresses. , , None of them exceeded the preset threshold, and the concentration of the three principal stress directions was... , , If all preset requirements are met, the geostress results can be rated as a high quality level.

[0153] If the number of effective measurement segments is insufficient, low-quality measurement segments are concentrated within a continuous depth range, external cross-validation error exceeds the limit, principal stress magnitude dispersion exceeds the limit, principal stress direction concentration is lower than the preset threshold, the proportion of highly correlated measurement segments is too high, the influence coefficient of a single measurement segment exceeds the preset upper limit, or the core elastic parameters do not match the lithology of the measurement segment, a supplementary measurement suggestion will be output. When the number of effective samples after correlation correction is lower than the preset effective sample number threshold, the dispersion of any principal stress magnitude exceeds the corresponding dispersion threshold, the concentration of any principal stress direction is lower than the corresponding concentration threshold, or the external cross-validation error exceeds the preset validation threshold, the data processing terminal generates a supplementary measurement suggestion; the supplementary measurement suggestion includes the supplementary measurement depth range, the reason for supplementary measurement, the location of the local structural surface to be avoided, and the recommended supplementary measurement method.

[0154] The depth range for supplementary measurements can be determined based on the concentrated locations of low-quality measurement segments, suspected abnormal measurement segments, locations with large verification errors in retained measurement segments, and locations of structural planes or lithological abrupt changes. Supplementary measurement methods may include adding measurement segments at adjacent depths of the original borehole, resetting measurement segments to avoid local joints, arranging verification measurement segments in adjacent boreholes, re-acquiring core elastic parameters, or replacing the strain measurement device for re-measurement. The in-situ stress inversion quality level can be classified into Level I, Level II, Level III, and Level IV. Level I indicates high-quality test data, sufficient number of valid samples, small cross-validation error, low dispersion of principal stress magnitude, and high concentration of principal stress direction; the results can be directly used as engineering design parameters. Level II indicates that the test data is generally reliable, but there are a few low-quality test sections or local dispersion; the results can be used as a reference for engineering design and verified in conjunction with geological conditions. Level III indicates insufficient number of valid samples, dispersion of candidate fusion results, or low concentration of direction; the results need to be supplemented by testing or verified in conjunction with adjacent boreholes before use. Level IV indicates a high proportion of invalid test sections, cross-validation error exceeding the limit, or instability in both the magnitude and direction of principal stress; the results are not recommended for direct acceptance, and test sections should be rearranged or supplementary tests should be conducted.

[0155] The following describes a specific embodiment of the single-hole multi-segment geostress fusion quality control method of the present invention:

[0156] A geostress testing borehole 100 was drilled in the surrounding rock of an underground cavern. After the borehole entered the stable section of the surrounding rock, eight testing sections 101 at different depths were arranged along the borehole axis, designated as Section 1 to Section 8. Each section was tested using the hollow inclusion strain gauge stress relief method. After each section underwent inclusion installation, cementation and curing, initial reading stabilization, strain gauge release, and strain acquisition, a single-section inversion was performed according to the strain-stress inversion model corresponding to that hollow inclusion strain gauge. This yielded the far-field stress tensor, principal stress magnitude and direction, inversion residuals, release curve characteristics, and field process records for that section.

[0157] Each measurement segment generates the following single-segment inversion result package:

[0158]

[0159] in, ; For the first The far-field stress tensor obtained by independent inversion of each measurement segment; The magnitude and direction of the principal stresses in this measured section; These are the inversion residuals or fitting quality parameters for a single measurement segment; This is for inversion condition number, covariance, or uncertainty information; Record the process on site, including borehole wall cleaning, inclusion insertion depth, cementation curing time, strain channel connection status, whether the release process was smooth, and core integrity; This is to measure the depth of the section, borehole azimuth, borehole dip angle, lithological range, and development of structural surfaces.

[0160] After generating the results package, a hard threshold is applied to each measurement segment. If a segment exhibits issues such as missing key strain channels, no final stable plateau in the strain release curve, inclusion debonding, failed cementation, single-segment inversion non-convergence, insufficient rank of the inversion matrix, or severely fragmented core that cannot represent the original rock state, then that segment is classified as segment 104 and will not participate in subsequent tensor-level robust fusion. The hard threshold is used to exclude data that clearly cannot reflect the in-situ stress release process, preventing such segments from amplifying errors in subsequent statistics.

[0161] For sections that are not eliminated, a soft-weight quality rating is further performed. The section quality score can be calculated using the following formula:

[0162]

[0163] in, It indicates the quality of the on-site process and is used to evaluate the cleanliness of the borehole wall, the installation position of the inclusions, the degree of cementation and curing, and the integrity of the core. It indicates the quality of the strain release curve and is used to evaluate the initial stable section, the release of change section, the final stable plateau, signal jumps, and channel integrity. It represents the inversion quality of a single measurement segment and is used to evaluate inversion residuals, goodness of fit, condition number, and convergence status; It represents the consistency quality of multiple measurement segments and is used to evaluate the consistency of the far-field stress tensor of the measurement segment with the overall trend of adjacent measurement segments or the same evaluation interval. It indicates the consistency of geological interpretation and is used to determine whether anomalies in a section can be explained by structural planes, lithological abrupt changes, fracture zones, or local stress disturbances. to These are the weighting coefficients.

[0164] Specifically, test segments can be divided into four categories: trusted test segments, usable test segments, suspected anomaly test segments, and eliminated test segments. Trusted test segments are used for high-weight participation in candidate combinations; usable test segments are used for medium-weight participation in candidate combinations; suspected anomaly test segments generally participate with low weight, or are preferentially reserved as test segments for external verification; eliminated test segments do not participate in fusion.

[0165] For example, in this embodiment, the release curve of segment 1 is stable, the core is intact, and the inversion residual is small, so it is rated as a reliable segment; the release curves of segments 2 and 3 are both relatively stable, and the inversion results of individual segments are similar, so they are also rated as reliable segments, but the correlation between the two needs to be corrected later; the release curve of segment 4 has a local jump, and a set of relatively developed joints can be seen in the core, and the inversion result deviates from the adjacent segments, so it is rated as a suspected abnormal segment; the release curve of segment 5 is basically stable, but the inversion residual is slightly high, so it is rated as a usable segment; segment 6 did not form a stable platform after release, and the core was severely broken, so it was rated as segment 104 to be removed; segment 7 is of good quality and is rated as a reliable segment; segment 8 has an intact core, but the curve tail fluctuates slightly, so it is rated as a usable segment. The above rating results can be summarized in the following form:

[0166] Test section 1 The curve is stable, the core is intact, and the residual is small. Trusted segment High weighting in fusion Test Section 2 The curve is stable and close to the results of the third measurement segment. Trusted segment High weighting, but relevance correction is needed. Section 3 The curve is stable and close to the results of the second measurement segment. Reliable segment High weighting, but relevance correction is needed. Section 4 Curve jumps, local joints, result deviations Suspected abnormal segment Low weight or as a verification test segment Section 5 The curve is basically stable, but the residual is slightly high. Available measurement section Medium weighting participates in fusion Section 6 No stable platform, core fragmentation Eliminate test sections Not participating in integration Section 7 The curve is stable and the residual is small. Reliable segment High weighting in the integration Section 8 The tail section showed slightly larger fluctuations, but the core sample remained intact. Available measurement section Medium weighting participates in fusion

[0167] This embodiment demonstrates that the present invention does not simply average all individual measurement results, but rather performs a graded processing of each measurement segment before fusion, considering factors such as field process, release curves, inversion quality, segment consistency, and geological interpretation. This step allows for the removal of invalid data and the reduction of weight for low-quality data before errors enter the fusion calculation.

[0168] After completing the quality rating of the borehole sections, the same borehole is divided into geological-stress evaluation intervals (103) based on borehole logging, core observation, structural plane development, and lithological continuity. If multiple sections are located within intervals with continuous lithology, similar structural plane development, relatively gentle expected stress gradients, and the same engineering evaluation purpose, they are assigned to the same geological-stress evaluation interval. If a section crosses a significant lithological boundary, fault fracture zone, tectonic stress abrupt change zone, or strongly heterogeneous zone, it is not directly merged into the same interval with sections on either side.

[0169] In this embodiment, segments 1 to 5, 7, and 8 are located within the same continuous hard rock segment and can serve as candidate data sources within the same geological-stress evaluation interval. Segment 6 was deemed discarded (segment 104) because it failed to form a stable platform after release and had severely fragmented core material, and therefore did not participate in candidate combinations, robust fusion, or retention verification. Segment 4 was rated as a suspected anomalous segment due to local jumps in the release curve and the presence of well-developed joints in the core material. It was not included in the fusion of some candidate combinations but was retained as a segment to examine the impact of local anomalies on the candidate fusion results.

[0170] The far-field stress tensor of each effective measurement segment Mapped to a six-component stress vector For the second and third measurement sections, the axial depth distance between them is... Smaller, far-field stress tensor normalization difference The diameter is relatively small, and borehole logging shows that both are located in the same lithological continuum, therefore the geological zoning item... High. According to the following formula:

[0171]

[0172] Or adopt a specific form:

[0173]

[0174] It can be determined that the second and third test segments are highly correlated test segments. For such test segments, this invention does not treat them as two completely independent samples and directly participate in the same candidate combination with the initially assigned quality control weights. Instead, it adopts a correlation correction strategy: in one candidate combination, the second test segment is allowed to participate in the fusion, and the third test segment is used as a reserved test segment for verification; in another candidate combination, the third test segment is allowed to participate in the fusion, and the second test segment is used as a reserved test segment for verification; or both can appear simultaneously, but at least the test segment fusion weight of one of the test segments is updated or reduced based on the correlation correction parameter.

[0175] Specifically, the following dynamic candidate segment combinations and reserved segment set are constructed:

[0176]

[0177]

[0178]

[0179]

[0180] in, Indicates the first A combination of candidate test segments, Indicates the relationship with the first The set of retained test segments corresponding to each candidate test segment combination. Candidate combination The test segments in the middle participate in the current candidate fusion and retain the set. The test segments in the test do not participate in the current candidate fusion; they are only used for external cross-validation. Because... and The cross-validation process dynamically changes the partitioning. A certain segment can be a reserved segment in one combination and a segment participating in fusion in another combination; however, for the same candidate combination, the reserved segment does not participate in the fusion of that combination.

[0181] With candidate combinations For example, the six-component stress vectors of the 1st, 2nd, 5th and 7th measurement segments are taken respectively. , , , Based on the segment quality level, segments 1, 2, and 7 are considered reliable segments and assigned higher initial quality control weights; segment 5 is considered usable and assigned a medium initial quality control weight. Since segments 2 and 3 are highly correlated segments, and segment 3 is placed in the reserved segment set in this portfolio... The second test section is in When a segment participates in the fusion process, its segment fusion weight can either be retained as the higher weight after correction, or it can be determined according to... The number of samples is reduced; its contribution to the effective sample size will be discussed later. Correlation correction is performed during the process.

[0182] Candidate Combinations The robust fusion calculation is as follows:

[0183]

[0184] During the calculation, the weighted average of the reliable segments within the candidate combination is first used as the initial estimate. Then, the residual of each segment relative to the current estimate is calculated. If the residual of a certain segment is significantly larger than that of other segments, its influence is reduced by using the Huber loss function, the Tukey double-weighted loss function, or an iterative reweighting method. When the change in the candidate fusion vector between two adjacent iterations is less than the preset convergence threshold, the candidate fusion six-component stress vector is obtained. And further mapped to candidate fused far-field stress tensor .

[0185] right , , Repeat the above process to obtain the following results: , , In this way, different candidate combinations can express the far-field stress state within the same geological-stress evaluation interval from the perspective of different subsets of measurement segments, while avoiding excessive control of the final results by a single measurement segment or a single highly correlated measurement segment group.

[0186] This embodiment corresponds to Figure 6 The technical process from left to middle: The far-field stress tensor of each segment in the candidate segment combination 500 is first converted into a six-component stress vector, and then enters the tensor-level robust fusion module 600; this module simultaneously considers mass weight, correlation correction and robust cost function, and finally forms the candidate fused far-field stress tensor.

[0187] After obtaining multiple candidate fused far-field stress tensors, external cross-validation is performed using the corresponding retained measurement segment set. (The candidate combination is then used.) For example, its corresponding set of retained test segments is Sections 3 and 8 do not participate. The robust fusion is only used to test the explanatory power of the candidate fusion result for segments not involved in the test.

[0188] For candidate fusion results The verification error is:

[0189]

[0190] like The result did not exceed the preset verification threshold, indicating that the candidate far-field stress state formed by the fusion of segments 1, 2, 5, and 7 can well explain segments 3 and 8, therefore it is retained. And assign higher fusion weights to the candidates. .like If the result exceeds the preset verification threshold, it indicates that the candidate fusion result is insufficient for interpreting segments not involved in the test. Therefore, appropriate action should be taken. Reduce or remove from the list.

[0191] Similarly, for , , The retained measurement sections were verified separately to obtain the corresponding verification errors. , , And based on this, determine the candidate fusion weights. , , In this embodiment, if The set of reserves The verification error in the fourth section was significantly larger than expected. Furthermore, the fourth section contained abrupt changes in the release curve and records of local joint development. Therefore, it can be concluded that there are testing quality issues or local geological disturbances in the depth range near the fourth section, and the depth should be reduced accordingly. The candidate fusion weights of the candidate fusion results are determined, and the depth near the fourth measurement segment is included in the supplementary measurement suggestion.

[0192] After cross-validation, the retained candidate fusion results are assumed to be:

[0193]

[0194] The corresponding candidate fusion weights are:

[0195]

[0196] Its final far-field stress six-component vector can be determined by the following formula:

[0197]

[0198] Will Mapped to the final far-field stress tensor Then, eigenvalue decomposition is performed on it to obtain the final principal stress magnitude and principal stress direction:

[0199]

[0200]

[0201] To further evaluate the stability of the principal stress magnitudes, eigenvalue decomposition is performed on each retained candidate fusion tensor. For example:

[0202]

[0203]

[0204]

[0205] For the maximum principal stress, its statistical value and dispersion are as follows:

[0206]

[0207]

[0208] Statistical values ​​of intermediate principal stress and minimum principal stress , and dispersion , Calculate in the same way. Therefore, Figure 7 The “principal stress magnitude” no longer represents only a single value obtained from the final eigenvalue decomposition, but further includes the statistical values ​​of principal stress magnitudes and the evaluation of dispersion among multiple retained candidate fusion results.

[0209] For the principal stress directions, taking the direction of maximum principal stress as an example, let the unit vectors of the maximum principal stress directions corresponding to the three retained candidate fusion results be as follows:

[0210]

[0211] The second-order direction tensor corresponding to the direction of maximum principal stress is:

[0212]

[0213] Pick The eigenvector corresponding to the largest eigenvalue is used as the direction of the average maximum principal stress. The intermediate principal stress direction and the minimum principal stress direction are constructed respectively. , And obtain the average direction , .

[0214] like The three characteristic values ​​are , , ,and Then the following can be adopted:

[0215]

[0216] As the first The directional concentration of each principal stress direction ( ).like The larger value indicates that the first result given by the fusion of multiple candidate results is... The principal stresses are relatively consistent in direction; if The smaller value indicates a higher degree of dispersion in the principal stress direction, suggesting that the acceptance level of the results should be lowered or a recommendation for supplementary testing and verification should be made.

[0217] The final output includes: the final far-field stress tensor. Statistical values ​​of principal stress magnitude , , Dispersion of principal stress magnitude , , Mean principal stress direction , , directional concentration , , Cross-validation error The number of effective samples after correlation correction Quality level and recommendations for supplementary testing.

[0218] If segment 6 is removed, segment 4 is listed as a suspected anomaly, and segments 2 and 3 are marked as highly correlated segments, and the cross-validation error of the final candidate fusion results is small, then... , , None of them exceeded the preset threshold for the dispersion of principal stress magnitude. , , If all the preset directional concentration requirements are met, the geostress results can be rated as high-level. If the area near the 4th measurement section continues to cause large verification errors, or if the magnitude dispersion of a certain principal stress is significantly large, or the directional concentration of a certain principal stress is significantly low, then the depth range near the 4th measurement section will be listed as a supplementary measurement range, and it is recommended to avoid local joints and reset the measurement section, or to arrange a verification measurement section in an adjacent borehole.

[0219] This embodiment corresponds to Figure 6 right side and Figure 7The content shown is as follows: The candidate fusion far-field stress tensor is cross-validated by the retained measurement segment set 102. The validation results determine whether the candidate fusion results are retained, downweighted, or eliminated. The retained candidate fusion far-field stress tensor enters the principal stress statistics module. In this module, the principal stress magnitude statistics and dispersion evaluation, and the principal stress axial direction tensor statistics are performed simultaneously. Finally, the quality level and supplementary measurement suggestions are output by the acceptance and supplementary measurement feedback module.

[0220] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for quality control of geostress fusion across multiple measurement sections in a single borehole, characterized in that, The method is executed by a data processing terminal and includes the following steps: Acquire stress relief test data from multiple sections within the same borehole, and perform strain-stress inversion independently on each section to form a corresponding single-section inversion result package; Each measurement segment is evaluated and its quality control weight is assigned based on the inversion result package of a single measurement segment. Determine the segment correlation coefficient between any two segments, and form a correlation correction parameter based on the segment correlation coefficient; Based on the quality rating, quality control weight, and correlation correction parameters of the test segments, candidate test segment combinations and a set of retained test segments corresponding one-to-one with the candidate test segment combinations are constructed. The far-field stress tensor of each segment within the candidate segment combination is converted into a six-component stress vector, and tensor-level robust fusion is performed in the six-component stress vector space to obtain the candidate fused far-field stress tensor. External cross-validation is performed on the candidate fused far-field stress tensor using the far-field stress tensor of the retained measurement segment set, and the candidate fused far-field stress tensor is screened based on the cross-validation error. Based on the selected candidate fusion far-field stress tensors, the magnitude and dispersion of principal stresses are statistically analyzed and evaluated. The direction and concentration of principal stresses are determined by the axial direction tensor statistics. Based on at least three of the following factors: the quality rating of the measurement segment, the quality control weight, the number of valid samples after correlation correction, the dispersion of principal stress magnitude, the concentration of principal stress direction, and the cross-validation error, the geostress inversion quality level and supplementary measurement recommendations are output.

2. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, The single-segment inversion results package includes: far-field stress tensor, inversion quality parameters, segment spatial information, field process information, principal stress magnitude, and principal stress direction; wherein, the segment spatial information includes: segment depth, borehole azimuth, borehole dip angle, and lithological logging information; the inversion quality parameters include: single-segment inversion residual, inversion condition number, and uncertainty parameters; the field process information includes: borehole wall cleaning, core integrity record, inclusion installation record, cementation and solidification record, strain channel integrity record, and strain release curve.

3. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, Before assigning quality ratings and quality control weights to each measurement segment, invalid measurement segment identification is also included. This includes: when a measurement segment has missing key strain channels, a lack of stable plateaus in the strain release curve, inclusion debonding, cementation failure, single-segment inversion non-convergence, insufficient rank of the inversion matrix, or severely broken core that cannot represent any of the original rock states, the measurement segment is identified as a discarded measurement segment.

4. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, The quality rating includes: determining the quality level of a measurement segment based on field process information, strain release curves, inversion quality parameters, multi-segment consistency quality, and geological interpretation consistency; the quality level of a measurement segment includes reliable segments, usable segments, suspected anomaly segments, and eliminated segments; wherein, reliable segments are assigned a first quality control weight, usable segments are assigned a second quality control weight, suspected anomaly segments are assigned a third quality control weight, and eliminated segments have a quality control weight of zero.

5. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, Determining the correlation coefficient between any two measurement segments includes: obtaining the measurement segment spacing, far-field stress tensor difference, and geological zoning information of any two measurement segments, and determining the correlation coefficient between the two measurement segments based on the measurement segment spacing, far-field stress tensor difference, and geological zoning information; When the correlation coefficient of two measurement segments is greater than the preset correlation threshold, or when at least two of the following conditions are met simultaneously: the distance between measurement segments is less than the preset correlation length, the difference in far-field stress tensor is less than the preset difference threshold, or they are located in the same local structural control area, the two measurement segments are marked as correlation correction measurement segments.

6. The single-hole multi-segment geostress fusion quality control method according to claim 5, characterized in that, When two segments are marked as correlation-corrected segments, the quality control weight of at least one of the segments is updated based on the correlation correction parameter, or one segment is assigned to the candidate segment combination and the other segment is assigned to the reserved segment set.

7. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, The tensor-level robust fusion includes: [the following is a list of components] Far-field stress tensor of each measurement segment Mapped to a six-component stress vector and for the first Candidate segment combinations The six-component stress vector within is robustly estimated as follows: in, For the first The candidate fused far-field stress six-component vector corresponding to each candidate test segment combination The candidate fused far-field stress six-component vector is to be determined. For known measurement sections The six-component stress vector, To comprehensively consider the segment fusion weight after considering the segment quality weight and the correlation correction parameter, For a robust loss function, The normalized matrix or the inverse square root of the covariance matrix. This is the scale parameter.

8. The single-hole multi-segment geostress fusion quality control method according to claim 7, characterized in that, The robust loss function adopts the Huber loss function, Tukey double-weighted loss function, weighted least squares loss function or iterative reweighted loss function; after obtaining the candidate fused far-field stress six-component vector, it is mapped to a second-order symmetric candidate fused far-field stress tensor.

9. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, The external cross-validation includes: calculating the tensor difference between the candidate fused far-field stress tensor and the far-field stress tensor of each segment in the corresponding retained segment set, and determining the cross-validation error based on the median, mean, quantile, or weighted statistic of the tensor difference; when the cross-validation error exceeds a preset validation threshold, reducing the weight or eliminating the corresponding candidate fused far-field stress tensor; when the cross-validation error does not exceed the preset validation threshold, retaining the corresponding candidate fused far-field stress tensor and assigning candidate fusion weights.

10. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, The statistical analysis and dispersion evaluation of the principal stress magnitudes include: [the following is a summary of the selected principal stresses]. Candidate fused far-field stress tensor Perform eigenvalue decomposition to obtain the corresponding eigenvalue. The magnitude of each principal stress ,in According to the first The candidate fusion weights are determined by the cross-validation error of the candidate far-field stress tensor, the quality of the candidate segment combination, and the correlation correction results. The first one is determined according to the following formula. Weighted average of the principal stresses: And determine the first according to the following formula Weighted dispersion of the principal stress magnitudes: in, For the first Statistical results of the magnitude of each principal stress. For the first The dispersion of the magnitude of each principal stress.

11. The single-hole multi-segment geostress fusion quality control method according to claim 1, characterized in that, The principal stress directions are determined statistically using axial direction tensors. For the selected candidate fusion far-field stress tensors, unit vectors for the principal stress directions of the same order are obtained, and second-order direction tensors are constructed based on these unit vectors and the corresponding candidate fusion weights. in, For the first The second-order direction tensor corresponding to each principal stress direction For the first The candidate fused far-field stress tensor corresponding to the first Unit vector in the direction of principal stress. For the corresponding candidate fusion weights; the principal eigenvector of the second-order direction tensor is taken as the average principal stress direction, and the concentration of principal stress directions is determined according to the eigenvalue distribution of the second-order direction tensor; wherein, the direction unit vector and Processed according to the equivalent axial direction.

12. A single-well multi-segment geostress fusion quality control system, characterized in that, It includes a field acquisition terminal and a data processing terminal; the field acquisition terminal includes a strain acquisition module, a section spatial recording module, a field process recording module, and a lithology logging module; the data processing terminal is configured to receive and process data from the strain acquisition module, the section spatial recording module, the field process recording module, and the lithology logging module, forming a single section inversion result package, and completing section quality rating, correlation correction, candidate section combination and retained section set construction, tensor-level robust fusion, retained section cross-validation, principal stress magnitude and direction statistics, and outputting section quality level, candidate fusion far-field stress tensor, retained section verification error, principal stress magnitude and direction statistics results, and supplementary measurement suggestions.

Images