Kilometer-level geostress and wave velocity combined testing method and system

Abstract

The application provides a kind of kilometer geostress and wave speed combined test method and system, it is related to deep borehole combined test technical field.Acquire the pressure response sequence, acoustic response sequence and operation event time sequence of the same measuring section, and input MATE to extract pressure modal characteristics, acoustic modal characteristics and operation time sequence characteristics;The modal characteristic group is shared-unique decomposition, and the true state shared feature is formed;In the state attribution link, POT is introduced and a limited true state prototype set is established to form the true state prototype attribution result;According to the attribution result, form the state consistent measuring section identifier or the state mismatch measuring section identifier, and output the combined measuring section result or the mismatch measuring section result, solve the problem that the pressure response and acoustic response of different true stratum states are misjudged as a combined interpretation object under the condition of time-sharing test.The application is suitable for the combined interpretation of deep well section pressure and acoustic wave information.

Description

Technical Field

[0001] This invention relates to the field of deep borehole joint testing technology, and in particular to a method and system for joint testing of kilometer-level ground stress and wave velocity. Background Technology

[0002] Stress assessment in kilometer-deep boreholes typically relies on methods such as packer-stage pressurization, hydraulic fracturing, depressurization and rebound, and in-hole acoustic logging or wave velocity testing. Pressure response reflects stress changes during the pressurization, stabilization, and depressurization processes of the test section, while acoustic response (first wave arrival, amplitude attenuation, and wave velocity changes) characterizes rock mass structure, fracture development, and mechanical state. With increasing demand for assessment in deep mines, deep geothermal wells, unconventional oil and gas wells, and underground energy storage wells, the comprehensive utilization of pressure and acoustic information for the same test section has become an important technical direction for the geomechanical interpretation and mutual verification of test results in deep wells.

[0003] During field operations, packer expansion, pressure stabilization, pressure release, probe switching, and probe rod reset often occur sequentially along the same well section. Different testing methods are implemented in segments in time but point to the same testing segment in space. Therefore, how to map ground stress information and wave velocity information to the same formation state has become a problem of continuous concern in this field.

[0004] Existing technologies typically follow two paths. The first path revolves around direct testing of in-hole stress, as exemplified by patents CN116816337A, CN115467660A, and CN109357794A. This approach generally employs upper and lower packers to isolate the test section, injecting fluid or pressurizing it, and continuously recording downhole pressure changes. The in-hole stress parameters are then calculated by combining this data with fracturing pressure, pressure stabilization processes, or repeated pressurization processes. The second path revolves around in-hole acoustic testing, as exemplified by patents CN210742194U and CN215444043U. This involves arranging acoustic transceivers, water injection, or borehole sealing coupling structures within the borehole. By acquiring the arrival time of the first wave, amplitude attenuation, and propagation characteristics, the wave velocity or acoustic response of the test section can be obtained. In international literature, Prioul et al. combined imaging logging with sonic logging in 2007 to distinguish between fracture effects and stress effects, serving fracturing design. Another type of literature reconstructs formation stress using shear wave velocity distribution near the wellbore and interprets it using sonic logging. Samnejad et al. estimated the maximum horizontal principal stress using conventional logging and sonic logging in 2017, demonstrating that sonic information can already participate in stress-related interpretations. Regarding the joint processing of heterogeneous data, CVPR 2023 proposed the idea of ​​shared features and modal-specific feature decomposition, ACMMM 2024 proposed a prototype alignment method based on optimal transmission, and CVPR 2025 proposed a similarity-aware prototype optimal transmission allocation method, indicating that existing data processing technologies already possess the technical foundation to map heterogeneous responses to shared representations or finite prototype units.

[0005] In time-sharing operations at kilometer-level deep boreholes, existing technologies typically treat packer pressurization tests and in-hole acoustic tests as two directly corresponding information streams based on depth, or interpret the two results separately and then perform a post-comparison. Existing technologies lack dedicated processing for the timing of operational events within the same drilling segment, making it difficult to stably separate the pressure-side disturbances caused by packer expansion and contraction from the acoustic disturbances caused by probe switching, transducer contact changes, and probe resetting from the actual formation state components. Furthermore, existing technologies lack a unified attribution mechanism for finite real-state units within the same drilling segment. Therefore, under conditions of state transitions, boundary segments, and disturbance segments, it is easy to misjudge two responses that do not correspond to the same real-state as jointly interpretable objects.

[0006] Stress assessment in kilometer-deep boreholes typically relies on methods such as packer-stage pressurization, hydraulic fracturing, depressurization and rebound, and in-hole acoustic logging or wave velocity testing. Pressure response reflects stress changes during the pressurization, stabilization, and depressurization processes of the test section. Acoustic response (first wave arrival, amplitude attenuation, and wave velocity changes) characterizes rock mass structure, fracture development, and mechanical state. With the increasing demand for assessment in deep mines, deep geothermal wells, unconventional oil and gas wells, and underground energy storage wells, the synergistic use of pressure and acoustic responses within the same measurement section has become a key technological trend for mutually verifying geomechanical analysis and testing results in deep wells.

[0007] Existing solutions typically treat sealing pressure tests and borehole acoustic tests separately, or only correspond them according to the depth of the test section. This makes it difficult to distinguish between changes in the actual formation state and local anomalies caused by the operation process. As a result, under conditions of state jumps, boundary sections, and disturbance sections, pressure responses and acoustic responses that are not the same actual formation state are easily misjudged as objects that can be jointly interpreted. Summary of the Invention

[0008] The technical problem solved by this invention: This invention provides a method and apparatus for joint testing of ground stress and wave velocity at the kilometer level, which solves the problem of misjudging pressure response and acoustic response of different real strata as objects that can be jointly interpreted under time-sharing testing conditions.

[0009] The technical solution adopted by this invention to solve the above-mentioned technical problems is a kilometer-level ground stress and wave velocity joint testing method, comprising the following steps:

[0010] S1. Obtain the pressure response sequence of the same test section during the sealing pressurization test, the acoustic response sequence during the in-hole acoustic test, and the timing of the operation events to form the test section test data.

[0011] S2. Input the test data of the test section into MATE, and extract the pressure modal features corresponding to the pressure response sequence, the acoustic modal features corresponding to the acoustic response sequence, and the operation time sequence features corresponding to the operation event time sequence to form a modal feature group.

[0012] S3. Perform shared and unique decomposition on the modal feature group, and extract the real state shared features, which include pressure-side shared state features and acoustic-side shared state features.

[0013] S4. In the state attribution process of shared features of real state, POT is introduced to establish a finite set of real state prototypes. Using the finite set of real state prototypes as a common reference, similarity-aware optimal transmission allocation is performed on the shared state features of the pressure side and the shared state features of the sound wave side to form the real state prototype attribution result.

[0014] S5. Based on the results of the assignment of the real state prototype, determine the consistency of the state. When the shared state features of the pressure side and the shared state features of the acoustic side are assigned to the same real state prototype, a state-consistent measurement segment identifier is formed. When the shared state features of the pressure side and the shared state features of the acoustic side are not assigned to the same real state prototype, a state-mismatch measurement segment identifier is formed.

[0015] S6. Correlate the pressure response sequence and acoustic response sequence from the test data of the test section. Under the condition that the test section identifiers correspond to the same state, form a joint test section result. Under the condition that the test section identifiers correspond to the mismatched state, form a mismatched test section result.

[0016] S7. Output the joint test segment results or mismatch test segment results corresponding to the test segment.

[0017] Furthermore, in S1, the test data for the test segment is generated, specifically including:

[0018] During the sealing and pressurization test, multiple pressure sampling points are continuously collected around the same test section, and the pressure sampling points are arranged in chronological order of the pressure rise stage, pressure stabilization stage and pressure relief stage to form a pressure response sequence.

[0019] During the acoustic wave test inside the hole, multiple acoustic wave sampling points are continuously collected around the same test section, and the acoustic wave sampling points are arranged according to the arrival time of the first wave and the amplitude decay time to form an acoustic wave response sequence.

[0020] Based on the operation time points corresponding to packer expansion, pressure stabilization start, pressure stabilization end, pressure release start, probe switching and probe reset, the operation events are arranged in order of occurrence to form an operation event sequence.

[0021] The pressure response sequence, acoustic response sequence, and operation event time sequence are combined according to the corresponding relationship of the same test segment to form the test segment data.

[0022] Furthermore, in S2, modal feature sets are formed, specifically including:

[0023] The pressure response sequence from the test data of the test section is input into MATE. The pressure response sequence is processed by two layers of convolution and one layer of fully connected mapping according to the sampling order. Pressure modal features reflecting the changes in the pressure rise stage, pressure stabilization stage and pressure relief stage are extracted from the pressure response sequence.

[0024] The acoustic response sequence from the test data of the section is input into MATE. The acoustic response sequence is subjected to two layers of convolution calculation and one layer of fully connected mapping in the order of sampling. The acoustic modal features reflecting the changes in the arrival of the first wave and the changes in the amplitude attenuation are extracted from the acoustic response sequence.

[0025] Input the timing sequence of the operation events in the test data of the section into MATE, perform a two-layer fully connected mapping on the timing sequence of the operation events according to the order of event occurrence, form the timing sequence feature of the operation, and preserve the time position relationship of the packer expansion and contraction unloading, probe switching and probe reset in the timing sequence feature of the operation.

[0026] Based on the temporal position relationship in the operation sequence characteristics, the stage change position in the pressure modal characteristics is correspondingly calibrated, so that the pressure change in the pressure modal characteristics is associated with the time position corresponding to the packer expansion and contraction unloading, thus forming a pressure modal characteristic corresponding to the operation event sequence.

[0027] Based on the temporal position relationship in the operation sequence characteristics, the propagation change position in the acoustic modal characteristics is correspondingly calibrated, so that the propagation change in the acoustic modal characteristics is associated with the time position corresponding to probe switching and probe reset, thus forming acoustic modal characteristics corresponding to the operation event sequence.

[0028] The pressure modal features, acoustic modal features, and operation time sequence features corresponding to the operation event time sequence are collected according to the correspondence of the same measurement segment to form a modal feature group.

[0029] Furthermore, in S3, shared features of the real state are extracted, specifically including:

[0030] The pressure modal features in the modal feature group are combined with the operation time sequence features to locate the change part in the pressure modal features that is consistent with the time of packer expansion and contraction unloading, forming the pressure side candidate disturbance part and the pressure side candidate shared part.

[0031] The acoustic modal features in the modal feature group are combined with the operation time sequence features to locate the part of the acoustic modal features that is consistent with the time of the change in transducer contact, forming the acoustic side candidate disturbance part and the acoustic side candidate shared part.

[0032] Shared and unique decompositions are performed on the candidate shared portions of the pressure side and the candidate shared portions of the acoustic wave side. The common change components between the candidate shared portions of the pressure side and the candidate shared portions of the acoustic wave side are calculated, and the common change components are identified as shared state features. Based on the time position corresponding to the packer expansion and contraction unloading in the operation time sequence features, unique components are extracted from the candidate disturbance portion of the pressure side. The change portion driven by the packer expansion and contraction unloading is identified as the pressure side disturbance portion, and the pressure side disturbance portion is identified as part of the modal unique features.

[0033] Based on the time position corresponding to the change in transducer contact in the operation time sequence characteristics, the unique component is extracted from the candidate disturbance part on the acoustic side. The part driven by the change in transducer contact is identified as the acoustic side disturbance part, and the acoustic side disturbance part is identified as another part of the modal unique features.

[0034] The shared state characteristics are simultaneously preserved from the pressure mode characteristics and the acoustic mode characteristics, and the pressure-side disturbance part and the acoustic-side disturbance part are not included in the shared state characteristics, thus forming a true state shared characteristic that represents the true formation state of the same section.

[0035] Furthermore, in S4, the results of assigning the real-state prototype are formed, specifically including:

[0036] The real state shared features are input into the state attribution link of MATE. POT is introduced into the state attribution link of MATE, and a finite real state prototype set is established based on the distribution relationship of the real state shared features that jointly reflect the real stratigraphic state of the same section.

[0037] The pressure-side shared state features are calculated one by one with the finite set of real state prototypes to obtain the similarity between the pressure-side shared state features and each real state prototype in the finite set of real state prototypes, and the state attribution calculation input of the pressure-side shared state features is formed.

[0038] The acoustic wave-side shared state features are calculated one by one with the finite set of real state prototypes to obtain the similarity between the acoustic wave-side shared state features and each real state prototype in the finite set of real state prototypes, and the state attribution calculation input of the acoustic wave-side shared state features is formed.

[0039] Using a finite set of real state prototypes as a common reference, the state assignment calculation inputs of the pressure-side shared state features and the sound-wave-side shared state features are subjected to similarity-aware optimal transmission allocation within the POT, so that the state assignment updates of the pressure-side shared state features and the sound-wave-side shared state features are completed within the same finite set of real state prototypes.

[0040] Based on the updated pressure-side shared state characteristics and the updated acoustic-side shared state characteristics, the true state prototype assignment results corresponding to the pressure-side shared state characteristics and the true state prototype assignment results corresponding to the acoustic-side shared state characteristics are determined respectively.

[0041] The results of assigning the true state prototypes corresponding to the shared state characteristics on the pressure side and the results of assigning the true state prototypes corresponding to the shared state characteristics on the acoustic side are aggregated according to the correspondence of the same measurement segment to form the true state prototype assignment results.

[0042] Furthermore, in S5, state consistency determination specifically includes:

[0043] Compare the assignment values ​​of each real state prototype corresponding to the shared state feature on the pressure side in the real state prototype assignment results, determine the real state prototype with the largest assignment value, and form the real state prototype identifier on the pressure side.

[0044] The assignment values ​​of each real state prototype corresponding to the shared state feature on the acoustic side in the real state prototype assignment results are compared, and the real state prototype with the largest assignment value is determined to form the real state prototype identifier on the acoustic side.

[0045] The pressure side real state prototype identifier is compared with the sound wave side real state prototype identifier. If the pressure side real state prototype identifier and the sound wave side real state prototype identifier are the same, a state consistent measurement segment identifier is formed.

[0046] When the prototype identifier of the actual state on the pressure side differs from that on the acoustic side, a state mismatch test segment identifier is formed.

[0047] Furthermore, in S6, the joint measurement results and the mismatched measurement results are formed, specifically including:

[0048] Based on the consistent test segment identifier and the mismatch test segment identifier, the test segment data is filtered by conditions to form consistent test segment data and mismatch test segment data.

[0049] Under the condition of consistent measurement segment identification, the pressure response sequence and acoustic response sequence corresponding to the same measurement segment are extracted from the consistent measurement segment data, and the pressure response sequence and acoustic response sequence are associated according to the correspondence of the same measurement segment to form a joint measurement segment result;

[0050] Under the condition of state mismatch test segment identification, extract the pressure response sequence and acoustic response sequence corresponding to the same test segment from the mismatch test segment data, and determine the pressure response sequence and acoustic response sequence that do not meet the correspondence relationship of the same real state prototype as the mismatch test segment result.

[0051] Furthermore, in S7, the output of the joint test segment results or mismatched test segment results for the corresponding test segment specifically includes:

[0052] Based on the results of the real state prototype attribution, extract the real state prototype identifier corresponding to the same test segment.

[0053] Based on the consistent state segment identifier, condition judgment is performed, and the joint segment result corresponding to the same segment is called under the condition corresponding to the consistent state segment identifier.

[0054] Based on the state mismatch test segment identifier, condition judgment is performed, and the mismatch test segment result corresponding to the same test segment is called under the condition corresponding to the state mismatch test segment identifier;

[0055] The consistent state test segment identifier and its corresponding joint test segment result, and the mismatch state test segment identifier and its corresponding mismatch test segment result are combined according to the same test segment correspondence to form the output result.

[0056] This invention also provides a kilometer-level ground stress and wave velocity joint testing system to realize the kilometer-level ground stress and wave velocity joint testing method as described above. The system includes a test segment data acquisition module, a modal feature group extraction module, a sharing and unique decomposition module, a real state prototype attribution module, a state consistency determination module, a test segment adaptation module, and an output module.

[0057] The segment test data acquisition module is used to acquire the pressure response sequence of the same test segment in the sealing pressurization test, the acoustic response sequence in the in-hole acoustic wave test, and the operation event sequence to form the test segment test data.

[0058] The modal feature group extraction module is used to input the test data of the test section into MATE, and extract the pressure modal features corresponding to the pressure response sequence, the acoustic modal features corresponding to the acoustic response sequence, and the operation time sequence features corresponding to the operation event time sequence to form a modal feature group.

[0059] The shared and unique decomposition module is used to perform shared and unique decomposition on the modal feature group to obtain pressure-side shared state features, acoustic wave-side shared state features, pressure-side modal unique features, and acoustic wave-side modal unique features.

[0060] The real state prototype attribution module is used to introduce POT in the state attribution process of real state shared features, establish a finite real state prototype set, and use the finite real state prototype set as a common reference to perform similarity-aware optimal transmission allocation on pressure-side shared state features and sound-wave-side shared state features to form real state prototype attribution results.

[0061] The state consistency determination module is used to determine the state consistency based on the real state prototype attribution result. When the pressure side shared state features and the sound wave side shared state features are assigned to the same real state prototype, a state consistency measurement segment identifier is formed. When the pressure side shared state features and the sound wave side shared state features are not assigned to the same real state prototype, a state mismatch measurement segment identifier is formed.

[0062] The test segment adaptation module is used to associate the pressure response sequence and the acoustic response sequence from the test segment test data, and form a joint test segment result under the condition that the test segment identifiers are consistent in state, and form a mismatched test segment result under the condition that the test segment identifiers are mismatched in state.

[0063] The output module is used to output the joint test segment results or mismatch test segment results corresponding to the test segment.

[0064] The beneficial effects of this invention are as follows: This invention provides a method and system for joint testing of ground stress and wave velocity at the kilometer level. It acquires segment test data, inputs the segment test data into MATE (Mechanical Analysis and Testing), extracts modal feature groups, performs shared and unique decomposition on the modal feature groups, extracts shared features of the true state, introduces POT (Position of Oriented Transmission) in the state attribution stage of the shared features of the true state, establishes a finite set of true state prototypes, and uses the finite set of true state prototypes as a common reference to perform similarity-aware optimal transmission allocation on the pressure-side shared state features and the acoustic-side shared state features, forming true state prototype attribution results. Based on the true state prototype attribution results, state consistency is determined, forming a state-consistent segment identifier or a state-mismatched segment identifier. Under the condition corresponding to the state-consistent segment identifier, a joint segment result is formed; under the condition corresponding to the state-mismatched segment identifier, a mismatched segment result is formed. The joint segment result or mismatched segment result corresponding to the segment is output. This solves the problem of misjudging pressure response and acoustic response of different true strata states as jointly interpretable objects under time-division testing conditions.

[0065] The present invention has the following advantages:

[0066] 1. Accurately solve the state matching problem in time-sharing testing: By introducing the MATE model to extract multimodal features and combining it with POT for similarity-aware optimal transmission allocation, the misjudgment of pressure and acoustic response under different real strata conditions such as state jump, boundary segment and disturbance segment is effectively avoided, and the reliability of joint test results is significantly improved.

[0067] 2. Deep fusion and decomposition of multimodal features: Through shared and unique decomposition techniques, shared features of the real stratigraphic state and unique features brought about by operational disturbances are separated, ensuring accurate capture of the real stratigraphic state and providing a clean feature basis for joint interpretation.

[0068] 3. Significant engineering application value: For kilometer-deep strata testing scenarios, this method can stably output joint measurement results with consistent conditions, providing high-precision joint data support of ground stress and wave velocity for deep resource exploration and underground engineering stability assessment, thereby reducing engineering decision-making risks.

[0069] 4. Enhanced automation and intelligence: The algorithm-driven judgment throughout the entire process replaces the traditional state matching method that relies on human experience, reducing human error and improving testing efficiency and result consistency, making it suitable for large-scale, multi-stage engineering testing needs. Attached Figure Description

[0070] Figure 1 This is a flowchart illustrating the kilometer-level ground stress and wave velocity joint testing method provided by the present invention.

[0071] Figure 2 This is a schematic diagram of the process for generating test data in the kilometer-level ground stress and wave velocity joint testing method provided by the present invention;

[0072] Figure 3 This is a schematic diagram of the process for forming modal characteristic groups in the kilometer-level ground stress and wave velocity joint testing method provided by the present invention;

[0073] Figure 4 This is a schematic diagram of the process for forming real-state shared features in the kilometer-level ground stress and wave velocity joint testing method provided by the present invention;

[0074] Figure 5 This is a flowchart illustrating the process of generating the real-state prototype attribution results in the kilometer-level ground stress and wave velocity joint testing method provided by this invention.

[0075] Figure 6 This is a flowchart illustrating the state consistency determination process in the kilometer-level ground stress and wave velocity joint testing method provided by the present invention.

[0076] Figure 7 This is a schematic diagram illustrating the process of forming the results of the joint test section and the results of the mismatched test section in the kilometer-level ground stress and wave velocity joint testing method provided by the present invention. Detailed Implementation

[0077] To address the issue that pressure and acoustic responses under time-sharing testing conditions are easily affected by operational disturbances such as packer expansion and contraction, probe switching, and probe resetting, making it difficult to confirm whether they correspond to the same actual geological conditions, this invention provides a kilometer-level joint testing method for ground stress and wave velocity. Figure 1 As shown, it includes the following steps:

[0078] S1. Obtain the pressure response sequence of the same test section during the sealing pressurization test, the acoustic response sequence during the in-hole acoustic wave test, and the timing of the operation events to form the test section test data.

[0079] S2. Input the test data of the test section into MATE, and extract the pressure modal features corresponding to the pressure response sequence, the acoustic modal features corresponding to the acoustic response sequence, and the operation time sequence features corresponding to the operation event time sequence to form a modal feature group.

[0080] S3. Perform shared and unique decomposition on the modal feature group to extract real state shared features, which include pressure-side shared state features and sound wave-side shared state features.

[0081] S4. In the state attribution process of shared features in real state, POT is introduced to establish a finite set of real state prototypes. Using the finite set of real state prototypes as a common reference, similarity-aware optimal transmission allocation is performed on the shared state features on the pressure side and the shared state features on the sound wave side to form the real state prototype attribution result.

[0082] S5. Based on the results of the assignment of the real state prototype, determine the consistency of the state. When the shared state features of the pressure side and the shared state features of the acoustic side are assigned to the same real state prototype, a state-consistent measurement segment identifier is formed. When the shared state features of the pressure side and the shared state features of the acoustic side are not assigned to the same real state prototype, a state-mismatch measurement segment identifier is formed.

[0083] S6. Associate the pressure response sequence and acoustic response sequence from the test data of the test section. Under the condition that the test section identifiers correspond to the same state, form a joint test section result. Under the condition that the test section identifiers correspond to the mismatched state, form a mismatched test section result.

[0084] S7. Output the joint test segment results or mismatch test segment results corresponding to the test segment.

[0085] In one optional implementation, in S1, the test data for the test segment is generated, specifically including:

[0086] During the sealing and pressurization test, multiple pressure sampling points are continuously collected around the same test section, and the pressure sampling points are arranged in chronological order of the pressure rise stage, pressure stabilization stage and pressure relief stage to form a pressure response sequence.

[0087] During the acoustic wave test inside the hole, multiple acoustic wave sampling points are continuously collected around the same test section, and the acoustic wave sampling points are arranged according to the arrival time of the first wave and the amplitude decay time to form an acoustic wave response sequence.

[0088] Based on the operation time points corresponding to packer expansion, pressure stabilization start, pressure stabilization end, pressure release start, probe switching and probe reset, the operation events are arranged in order of occurrence to form an operation event sequence.

[0089] The pressure response sequence, acoustic response sequence, and operation event time sequence are combined according to the corresponding relationship of the same test segment to form the test segment data.

[0090] For example, with Taking the first test segment as an example, focusing on the first... The three inputs of each test section into the MATE, and the original pressure sampling record formed by the isolation pressurization test, are written as follows: ,in Indicates the first Sampling time at each pressure sampling point Indicates the first One pressure value, , This indicates the sampling time of the last pressure sampling point; the original acoustic wave sampling record formed by the in-orifice acoustic wave test is written as... ,in Indicates the first Sampling time of each acoustic sampling point Indicates the first Each amplitude value, , This indicates the sampling time of the last acoustic wave sampling point; the corresponding operation time point for the same round of testing is written as... , among which fixed position Corresponding to the packer expansion, fixed position Corresponding to the start of voltage stabilization, fixed position The voltage stabilization is complete, and the position is fixed. Corresponding to the start of pressure relief, fixed position Corresponding probe switching, fixed position Reset the corresponding measuring rod. Measuring section number. It is only used to establish the correspondence of the same measurement segment and is not included in the numerical sequence content.

[0091] right according to Sort the samples in ascending order to obtain a sequence of pressure sampling points arranged continuously according to sampling time. The corresponding packer inflation time is used as the starting boundary of the pressure response, and the last pressure sampling time after sorting is used as the ending boundary of the pressure response. , , The pressure sampling points are divided into three stages: pressure rise, pressure stabilization, and pressure release. Pressure sampling points for the pressure rise stage are located in the first segment, those for the pressure stabilization stage in the middle segment, and those for the pressure release stage in the last segment. To meet the fixed-length input requirement of MATE, the sorted pressure sampling points are mapped to discrete index axes. Forming a pressure response sequence When an index position corresponds to multiple pressure sampling points, the arithmetic mean of the multiple pressure values ​​is taken; when an index position does not correspond to a pressure sampling point, linear interpolation is used if adjacent pressure sampling points exist, and boundary preservation is used if adjacent sampling points are missing at the beginning or end of the sequence. Pressure Response Sequence The dimension is Each index position contains a pressure value.

[0092] right according to Sort by size from smallest to largest, and select probe switching time. With the probe reset time The acoustic sampling points between the points are designated as the effective interval for acoustic wave testing within the borehole. The acoustic sampling points within the effective interval are arranged according to sampling time, so that the amplitude change corresponding to the arrival of the first wave is located in the first segment, and the amplitude change corresponding to amplitude attenuation is distributed along time in the later segment. The acoustic sampling points within the effective interval are then mapped to discrete index axes. Forming a sound wave response sequence When an index position corresponds to multiple acoustic sampling points, the arithmetic mean of the multiple amplitude values ​​is taken; when an index position does not correspond to an acoustic sampling point, linear interpolation is used if adjacent preceding and following acoustic sampling points exist, and boundary preservation is used if adjacent sampling points are missing at the beginning or end of the sequence. Acoustic response sequence The dimension is Each index position contains an amplitude value.

[0093] Will Arrange the work events in a fixed order to form a time sequence. To eliminate the difference in absolute clock start point, the packer is inflated to correspond to the time. As a unified time reference, each operation time point is converted into a relative time offset to obtain... Job event sequence The first position corresponds to the relative time offset of packer inflation, the second position to the relative time offset of the start of pressure stabilization, the third position to the relative time offset of the end of pressure stabilization, the fourth position to the relative time offset of the start of pressure relief, the fifth position to the relative time offset of probe switching, and the sixth position to the relative time offset of probe reset. (Operational event sequence) The dimension is The index order of the six positions simultaneously preserves the occurrence order and temporal position relationship of the six types of job events.

[0094] The rule for establishing the correspondence within the same measurement segment is: pressure response sequence. Source: The acoustic response sequence of the sealed pressure test of each test section. Source: In-hole acoustic wave testing of each section, operation event timing. Source: Coverage The test data represents the same round of work records for each test segment. Specifically, the test data is represented by three inputs bound according to the corresponding relationship of the same test segment, where the pressure response sequence is written as... The acoustic response sequence is written as The sequence of job events is written as follows The input content received by MATE is... , and ,in Characterizes the pressure change process during the pressure rise, pressure stabilization, and pressure relief phases. Characterizing the propagation process of the first wave arrival and amplitude attenuation, Characterizes the time-position relationship of packer expansion, pressure stabilization start, pressure stabilization end, pressure relief start, probe switching, and probe reset.

[0095] Information on pressure response sequence, acoustic response sequence, operational event timing, and test data of the measurement segment is shown in Table 1.

[0096] Table 1 Data Information Table

[0097]

[0098] In one alternative implementation, in S2, a modal feature set is formed, such as... Figure 2 As shown, it specifically includes:

[0099] The pressure response sequence from the test data of the test section is input into MATE. The pressure response sequence is processed by two layers of convolution and one layer of fully connected mapping according to the sampling order. Pressure modal features reflecting the changes in the pressure rise stage, pressure stabilization stage and pressure relief stage are extracted from the pressure response sequence.

[0100] The acoustic response sequence from the test data of the section is input into MATE. The acoustic response sequence is subjected to two layers of convolution calculation and one layer of fully connected mapping in the order of sampling. The acoustic modal features reflecting the changes in the arrival of the first wave and the changes in the amplitude attenuation are extracted from the acoustic response sequence.

[0101] Input the timing sequence of the operation events in the test data of the section into MATE, perform a two-layer fully connected mapping on the timing sequence of the operation events according to the order of event occurrence, form the timing sequence feature of the operation, and preserve the time position relationship of the packer expansion and contraction unloading, probe switching and probe reset in the timing sequence feature of the operation.

[0102] Based on the temporal position relationship in the operation sequence characteristics, the stage change position in the pressure modal characteristics is correspondingly calibrated, so that the pressure change in the pressure modal characteristics is associated with the time position corresponding to the packer expansion and contraction unloading, thus forming a pressure modal characteristic corresponding to the operation event sequence.

[0103] Based on the temporal position relationship in the operation sequence characteristics, the propagation change position in the acoustic modal characteristics is correspondingly calibrated, so that the propagation change in the acoustic modal characteristics is associated with the time position corresponding to probe switching and probe reset, thus forming acoustic modal characteristics corresponding to the operation event sequence.

[0104] The pressure modal features, acoustic modal features, and operation time sequence features corresponding to the operation event time sequence are collected according to the correspondence of the same measurement segment to form a modal feature group.

[0105] For example, in this stage, MATE receives three inputs, which are pressure response sequences. Acoustic response sequence and task event sequence . Indicates the first One pressure value, Indicates the first Each amplitude value, to This sequentially represents the time positions corresponding to packer inflation, pressure stabilization start, pressure stabilization end, pressure release start, probe switching, and probe reset. MATE consists of a pressure response sequence encoding path, an acoustic response sequence encoding path, and a work event timing encoding path. The pressure response sequence encoding path only receives the pressure response sequence. The acoustic response sequence encoding path only receives acoustic response sequences. The job event timing encoding path only receives job event timing. The three paths do not overlap in this step.

[0106] Pressure response sequence When encoding, The inputs are placed into the first 1D convolutional layer in indexed order. This first 1D convolutional layer contains 32 convolutional neurons, each locally connected to its five adjacent pressure sampling points. To preserve the sequential order and boundary meaning of the pressure response sequence, boundary preservation is used when the convolutional window exceeds the boundary of the pressure response sequence; the pressure value from the nearest boundary position is copied to the missing position. The first 1D convolutional layer... After extracting the local pressure variation pattern, the first layer of intermediate pressure characteristics is formed by the linear rectification unit. . This represents the set of intermediate features output sequentially by 32 convolutional neurons along the stress response sequence. Then... The input is a second one-dimensional convolutional layer. This second one-dimensional convolutional layer contains 64 convolutional neurons, each of which is locally connected to three adjacent locations in the previous layer. This connection is used to further extract the continuously changing structures in the pressure rise, stabilization, and depressurization stages. The second one-dimensional convolutional layer is then processed by a linear rectifier unit to form the second layer of intermediate pressure features. .Will After being expanded in index order, the data is fed into a fully connected layer containing 64 neurons, which outputs stress modality features. Each All of them are obtained by mapping multiple adjacent local windows in the pressure response sequence through two layers of convolution and one layer of fully connected layer. The 64 positions jointly characterize the changes in the pressure rise stage, pressure stabilization stage and pressure relief stage.

[0107] Acoustic response sequence When encoding, The first 1D convolutional layer is input in indexed order. This layer contains 32 convolutional neurons, each locally connected to five adjacent acoustic sampling points. Boundary preservation is used even when the convolution window exceeds the acoustic response sequence boundary. The first 1D convolutional layer locally extracts waveform changes near the arrival of the first wave and initial amplitude attenuation changes, which are then processed by a linear rectifier unit to form the first layer of intermediate acoustic features. .Will The input is a second one-dimensional convolutional layer containing 64 convolutional neurons. Each convolutional neuron is locally connected to three adjacent locations in the previous layer. This connection is used to further extract the amplitude attenuation variation structure unfolding along the propagation process. This structure is then processed by a linear rectifier unit to form the intermediate features of the second layer of acoustic waves. .Will After being expanded in index order, the data is fed into a fully connected layer containing 64 neurons, which outputs acoustic modal features. Each All of them are obtained by mapping multiple adjacent local windows in the acoustic response sequence through two layers of convolution and one layer of fully connected layer. The 64 positions jointly characterize the changes in the arrival of the first wave and the changes in the amplitude attenuation.

[0108] Sequence of job events When encoding, The data is input into the first fully connected layer in a fixed event sequence. The first fully connected layer contains 16 neurons, each connected to all 6 time points, and the intermediate features of the first layer are formed through a linear rectifier unit. . Each position is a combination of the values ​​from the six job time positions after connection weights and bias mapping. The input is a second fully connected layer, which contains 32 neurons. Each neuron is connected to all 16 intermediate features from the first layer, and these features are then processed by a linear rectifier unit to form the job time sequence features. . The 32 positions together carry the sequence of occurrence and corresponding positional relationship of packer expansion and contraction unloading, probe switching and probe reset in the operation event time sequence. Instead of each of the 32 positions being associated with a single operation event, the 32 positions are used as a complete time position encoding vector for calibration.

[0109] After obtaining the pressure mode characteristics and job timing characteristics Next, the positions of the stage changes in the pressure modal characteristics were calibrated accordingly. The packer inflation time position was used as the reference. As a starting point, the corresponding time position after depressurization begins As the calibration endpoint for pressure change, the time position corresponding to the start of pressure stabilization will be used. The corresponding time position after voltage stabilization ends Corresponding time position to the start of depressurization Projecting onto the 64 pressure mode feature locations yields three-stage boundary indices. , and The stage boundary index uses: To be confirmed.

[0110] according to , and ,Will The corresponding relationships between the location of each feature and the pressure rise, stabilization, and depressurization stages are established, and the pressure modal features are updated to correspond to the pressure modal features corresponding to the time sequence of the operation events. . The value remains The encoding result, The positional meaning is determined based on the stage boundary index, thereby establishing a correlation between pressure changes and the time position corresponding to the packer expansion and contraction unloading.

[0111] After obtaining the acoustic modal characteristics and job timing characteristics Next, the locations of propagation changes in the acoustic modal characteristics were calibrated accordingly. The probe was then switched to the corresponding time position. As the starting point for sound wave propagation calibration, the probe is reset to the corresponding time position. As the endpoint for acoustic wave propagation calibration, the 64 acoustic wave modal feature positions are mapped sequentially to the propagation interval between probe switching and probe reset. This ensures that the earlier feature positions correspond to changes in the arrival of the first wave, the feature positions continuing along the index direction correspond to changes in wave amplitude attenuation, and the final feature positions correspond to the propagation end position near probe reset. Based on this mapping relationship, the acoustic wave modal features are updated to correspond to the acoustic wave modal features of the operational event sequence. . The value remains The encoding result, The positional meaning is determined by the probe switching and probe reset, thereby establishing a correlation between the changes in sound wave propagation and the time position corresponding to the probe switching and probe reset.

[0112] After completing the three-way encoding and two-way corresponding calibration, , and They are aggregated according to the correspondence of the same measurement segment to form modal feature groups. . This is a 64-dimensional pressure mode feature vector corresponding to the time sequence of work events. This is a 64-dimensional acoustic modal feature vector corresponding to the timing of the operation events. This is a 32-dimensional temporal feature vector. In this stage, MATE does not generate state determination results, perform shared-specific decomposition, or perform state attribution; it only completes input encoding, time position retention, and corresponding calibration, thus enabling the modal feature group... It becomes the direct input for the shared-specific decomposition calculation process.

[0113] The timing information of the job events is shown in Table 2.

[0114] Table 2. Job Event Sequence Information Table

[0115]

[0116] In one alternative implementation, in S3, shared features of the real state are extracted, such as... Figure 3 As shown, it specifically includes:

[0117] The pressure modal features in the modal feature group are combined with the operation time sequence features to locate the change part in the pressure modal features that is consistent with the time of packer expansion and contraction unloading, forming the pressure side candidate disturbance part and the pressure side candidate shared part.

[0118] The acoustic modal features in the modal feature group are combined with the operation time sequence features to locate the part of the acoustic modal features that is consistent with the time of the change in transducer contact, forming the acoustic side candidate disturbance part and the acoustic side candidate shared part.

[0119] Shared and unique decompositions are performed on the candidate shared portions of the pressure side and the candidate shared portions of the acoustic wave side. The common change components between the candidate shared portions of the pressure side and the candidate shared portions of the acoustic wave side are calculated, and the common change components are identified as shared state features. Based on the time position corresponding to the packer expansion and contraction unloading in the operation time sequence features, unique components are extracted from the candidate disturbance portion of the pressure side. The change portion driven by the packer expansion and contraction unloading is identified as the pressure side disturbance portion, and the pressure side disturbance portion is identified as part of the modal unique features.

[0120] Based on the time position corresponding to the change in transducer contact in the operation time sequence characteristics, the unique component is extracted from the candidate disturbance part on the acoustic side. The change part driven by the change in transducer contact is identified as the acoustic side disturbance part, and the acoustic side disturbance part is identified as another part of the modal unique features.

[0121] The shared state characteristics are simultaneously preserved from the pressure mode characteristics and the acoustic mode characteristics, and the pressure-side disturbance part and the acoustic-side disturbance part are not included in the shared state characteristics, thus forming a true state shared characteristic that represents the true formation state of the same section.

[0122] For example, MATE receives modal feature sets in this stage. . This represents the pressure modal characteristics corresponding to the timing of the work events, with 64 locations corresponding to the pressure change responses at the completed time locations; The acoustic modal characteristics corresponding to the timing of the operation event are represented, with 64 positions corresponding to the propagation change response that has been calibrated at the time and position. This represents the timing characteristics of the operation. The 32 locations jointly encode the order and corresponding positional relationship of packer expansion and contraction unloading, probe switching, and probe reset in the operation event timing. MATE performs shared-specific decomposition in this stage, aiming to retain the changes jointly determined by the actual formation state of the same section as shared state characteristics, while retaining the pressure-side changes driven by packer expansion and contraction unloading and the acoustic-side changes driven by transducer contact changes as modal-specific characteristics.

[0123] First, the pressure modal characteristics are located. and Concatenate them in a fixed order to form the pressure path input vector. This results in a 96-dimensional input. According to... The time positions corresponding to packer inflation, pressure stabilization start, pressure stabilization end, and pressure relief start, which have already been retained, are used to determine the set of pressure-side disturbance location intervals at 64 pressure mode characteristic locations. The pressure modal characteristic positions between the packer inflation position and the pressure stabilization start position are defined as the inflation drive range, and the pressure modal characteristic positions between the pressure stabilization end position and the pressure relief start position are defined as the unloading drive range. The inflation drive range and the unloading drive range together constitute the pressure stabilization drive range. .fall into of It was identified as a candidate perturbation on the pressure side and did not fall into the range. of The pressure-side candidate shared portion was identified. The original indices of the 64 pressure modal feature locations remained unchanged, ensuring that each pressure modal feature location continued to retain its temporal location meaning consistent with the operational timing characteristics.

[0124] Then, the acoustic modal characteristics are located. and Concatenate them in a fixed order to form the sound wave path input vector. This results in a 96-dimensional input. According to... The time and position corresponding to probe switching and probe reset, which have already been retained, are used to determine the set of acoustic side disturbance location intervals at 64 acoustic modal characteristic positions. Expand the preset window length forward and backward, centered on the position corresponding to the probe switch. This forms the first acoustic disturbance positioning range; the preset window length is then extended forward and backward from the position corresponding to the probe reset. This forms the second acoustic side disturbance localization range. In a feasible fixed example, Take 4. The first acoustic side disturbance location interval and the second acoustic side disturbance location interval together constitute... .fall into of The candidate perturbation component on the acoustic side was identified but did not fall into the range. of The candidate shared section on the acoustic side has been identified. No misalignment was found between probe switching and probe reset. The propagation changes in position remain in the candidate shared portion on the acoustic side.

[0125] After completing the positioning, Input pressure path sharing—a unique decomposition fully connected layer, which will Input acoustic path shared-specific decomposition fully connected layer. Both the pressure path shared-specific decomposition fully connected layer and the acoustic path shared-specific decomposition fully connected layer have 96 neurons, each fully connected to the 96 input values ​​of the corresponding path, and each containing 96 connection weights and 1 bias. The first 64 neurons of each shared-specific decomposition fully connected layer output the shared portion, and the last 32 neurons output the specific portion. The pressure path obtains 64 shared output values. and 32 pressure-side modal characteristics The acoustic wave path yields 64 shared output values. and 32 unique features of acoustic side modes . and The output values ​​at the same index position correspond to the same shared position, which is used to carry the change response that may be caused by the actual stratum state of the same section.

[0126] When determining the common changing components, do not directly... and All outputs with the same index are treated as shared state features, and are instead distributed at each shared location. The set of pressure-side disturbance positioning intervals Set of acoustic side disturbance location intervals Pressure path shared output value Shared output value with acoustic wave path Calculate the numerical values ​​of common variation components :

[0127] ,

[0128] In the formula, Indicates the first The common change component values ​​of the shared locations; Indicates a shared location index; Represents the set of pressure-side disturbance location intervals. In position The function that retrieves values ​​on the position fall into Time takes 1, position Not fallen Take 0 at the time; Represents the set of acoustic side disturbance location intervals In position The function that retrieves values ​​on the position fall into Time takes 1, position Not fallen Take 0 at the time; This represents the set of pressure-side disturbance location intervals; This represents the set of acoustic disturbance location intervals; Indicates pressure path shared output value The Middle The value at each position; Indicates the shared output value of the sound wave path. The Middle The value at each position; This represents the absolute value operation; This indicates a fixed positive number to prevent the denominator from being zero, taken from a fixed set of feasible examples. In this calculation method, and Having the same dimensions, the ratio is a dimensionless quantity. Since it is a dimensionless numerical value, there is no issue of dimension mismatch.

[0129] According to the numerical values ​​of common changing components Perform shared state feature retention. In a fixed, implementable example, set a retention threshold for common variable components. .when At that time, The Middle Shared output value Retained in the shared state characteristics on the pressure side, and The Middle Shared output value Retained in the shared state characteristics on the acoustic side; when At that time, and The Middle The shared location is synchronized to zero, preventing the first one from being zeroed out. Each shared location enters the shared state feature. After processing by the retention rule, the pressure-side shared state feature is obtained. Shared state characteristics with acoustic wave side . The 64 locations consist of shared output values ​​from pressure paths that satisfy the condition for retaining common changing components. The 64 positions consist of shared output values ​​from acoustic paths that satisfy the condition for preserving common variation components. and They jointly characterize the changes determined by the actual stratigraphic state of the same section.

[0130] When determining modality-specific features, based on and Interpretation of positioning results and The source of. The output portion dominated by the expansion-driven region and the unloading-driven region is identified as the pressure-side disturbance portion and is retained as part of the modal-specific features; The output portion dominated by the areas near the probe switching position and the probe reset position is identified as the acoustic side disturbance component and retained as another part of the modal-specific characteristics. Therefore, and True state shared features are formed in pairs. The total dimension is The first 64 dimensions are shared state features on the pressure side, and the last 64 dimensions are shared state features on the sound wave side. Represents modality-specific features, with a total dimension of The first 32 dimensions represent the modal-specific features corresponding to the pressure-side disturbance, while the latter 32 dimensions represent the modal-specific features corresponding to the acoustic-side disturbance. In this stage, MATE does not perform state attribution calculations; it only locates and decomposes the pressure-side candidate disturbance, pressure-side candidate shared, acoustic-side candidate disturbance, and acoustic-side candidate shared parts, and then assigns the true state shared features. Retain it as the input object for the state attribution process.

[0131] Information on each path and step is shown in Table 3.

[0132] Table 3 Information on each path and step

[0133]

[0134] In one alternative implementation, in S4, the real state prototype attribution result is formed, such as... Figure 4 As shown, it specifically includes:

[0135] The real state shared features are input into the state attribution link of MATE. POT is introduced into the state attribution link of MATE, and a finite real state prototype set is established based on the distribution relationship of the real state shared features that jointly reflect the real stratigraphic state of the same section.

[0136] The pressure-side shared state features are calculated one by one with the finite set of real state prototypes to obtain the similarity between the pressure-side shared state features and each real state prototype in the finite set of real state prototypes, and the state attribution calculation input of the pressure-side shared state features is formed.

[0137] The acoustic wave-side shared state features are calculated one by one with the finite set of real state prototypes to obtain the similarity between the acoustic wave-side shared state features and each real state prototype in the finite set of real state prototypes, and the state attribution calculation input of the acoustic wave-side shared state features is formed.

[0138] Using a finite set of real state prototypes as a common reference, the state assignment calculation inputs of the pressure-side shared state features and the sound-wave-side shared state features are subjected to similarity-aware optimal transmission allocation within the POT, so that the state assignment updates of the pressure-side shared state features and the sound-wave-side shared state features are completed within the same finite set of real state prototypes.

[0139] Based on the updated pressure-side shared state characteristics and the updated acoustic-side shared state characteristics, the true state prototype assignment results corresponding to the pressure-side shared state characteristics and the true state prototype assignment results corresponding to the acoustic-side shared state characteristics are determined respectively.

[0140] The results of assigning the true state prototypes corresponding to the shared state characteristics on the pressure side and the results of assigning the true state prototypes corresponding to the shared state characteristics on the acoustic side are aggregated according to the correspondence of the same measurement segment to form the true state prototype assignment results.

[0141] For example, the state attribution link of MATE receives the real state sharing feature. . This represents the shared state characteristics of the pressure side, with 64 positions corresponding to the shared output values ​​of the pressure side retained after the shared-specific decomposition. This represents the shared state characteristics on the acoustic side, with 64 positions corresponding to the shared output values ​​on the acoustic side retained after the shared-specific decomposition. The state attribution process within MATE consists of a prototype comparison unit and a POT allocation unit. The prototype comparison unit has 8 prototype neurons, which together form a finite set of real state prototypes. Each real-state prototype This is a 64-dimensional vector, where each location corresponds to the connection weights of a prototype neuron on the shared feature space. It represents the prototype distribution of a real-world state across 64 shared locations. (Finite set of real-world prototypes) The method for obtaining the MATE parameters is as follows: During the determination of MATE parameters, the pressure-side shared state features and acoustic-side shared state features corresponding to the same measurement segment are continuously input into 8 prototype neurons, so that the 64 connection weights of the 8 prototype neurons converge to the stable distribution of different real formation states; after the connection weights of the 8 prototype neurons stabilize, the 8 sets of 64-dimensional connection weights that have stabilized are fixed as a finite set of real state prototypes. .

[0142] For the When assigning the execution status of each test segment, the prototype comparison unit first checks... and With a finite set of real-state prototypes Perform a one-to-one correspondence calculation. Specifically, in the first... A real-state prototype Up, click on shared location digit-by-digit comparison and Differences, and comparisons digit by digit. and The differences; after completing the bit-by-bit comparison of 64 shared locations, the result is obtained. A set of pressure-side comparison results and a set of acoustic-side comparison results corresponding to a real-state prototype; according to The prototype indexes are arranged in order to form eight sets of pressure-side comparison results, which form the state attribution calculation input for the shared state characteristics of the pressure sides, according to... The prototype indexes of the eight sets of acoustic-side comparison results are arranged sequentially to form the state assignment calculation input for the shared state features of the acoustic side. The POT allocation unit uses the eight real state prototypes as eight common allocation columns, placing the pressure-side shared state features and the acoustic-side shared state features within the same prototype column system to perform similarity-aware optimal transmission allocation. The POT allocation unit does not change the finite set of real state prototypes. Instead of adding independent branches, the number of prototypes is not increased. Instead, the degree of support of the pressure-side shared state features and the sound-wave-side shared state features for the same real state prototype is examined simultaneously on the same prototype column, so that the two shared state paths can complete the unified assignment update within the same finite set of real state prototypes.

[0143] To implement similarity-aware optimal transmission allocation as an executable computation, within the POT allocation unit, for any mode... In the State attribution update value on a real-state prototype Calculate using the following formula:

[0144] ,

[0145] ,

[0146] in, Indicates the first Each test segment in modal The next corresponding number The state assignment update value of a real state prototype; Indicates the segment index; Indicates the prototype index of the actual state; This represents the index of the prototype column of the true state in the normalized denominator; Indicates modal marker, Corresponding to the shared state characteristics on the pressure side, Corresponding to the shared state characteristics on the acoustic side; This represents the location index among 64 shared locations; Indicates the shared state characteristics of the pressure side. In the The value at each position; Indicates the shared state characteristics on the acoustic side In the The value at each position; Representing modes In the The value at each position, when Time to take ,when Time to take ; Indicates the first A real-state prototype In the The prototype value at each position; Indicates the first The real-state prototype in the first... The prototype value at each position; This represents the absolute value operation; This indicates a fixed positive number to prevent the denominator from being zero, taken from a fixed set of feasible examples. ; This indicates that summation is performed based on the index range. All variables involved in the comparison are included in the formula. , , , and All are direct numerical values ​​in the same coordinate system of the shared feature space. The difference term in the numerator and denominator has the same dimension as the amplitude term, and the polynomial is a dimensionless assigned value.

[0147] According to the above calculation method, Perform status updates column by column. The first summation term in the numerator is used simultaneously. and For the The real-state prototypes are subject to common constraints, such that a large common support can only be obtained when the shared state characteristics on both the pressure side and the acoustic side are relatively small compared to the real-state prototype; the second summation term in the molecule preserves the mode. In the The unilateral assignment basis is based on eight real-state prototypes; the denominator is uniformly normalized across the eight real-state prototype columns, ensuring that the assignment updates for the same mode across the eight real-state prototypes form comparable allocation results. After completing the calculations for all eight columns, the real-state prototype assignment results corresponding to the shared state features on the pressure side are obtained. The result of attribution of the real state prototype corresponding to the shared state features of the acoustic side . It is an 8-dimensional vector, with 8 positions corresponding to the pressure-side attribution update values ​​of 8 real-state prototypes; This is an 8-dimensional vector, with 8 positions corresponding to the acoustic side attribution update values ​​of 8 real-state prototypes. Then... and The results are compiled according to the correspondence of the same measurement segment to form the actual state prototype attribution results. . The total dimension is The first eight dimensions correspond to the assignment results of the pressure-side shared state features relative to eight real-state prototypes within the finite set of real-state prototypes, while the latter eight dimensions correspond to the assignment results of the acoustic-side shared state features relative to eight real-state prototypes within the finite set of real-state prototypes. The input to MATE in this stage is the real-state shared features. The output of MATE in this stage is the result of the real-state prototype attribution. The recursive method completed by the POT allocation unit is based on the prototype column. Calculate column by column and And complete the state attribution update within the same finite set of real state prototypes.

[0148] In one optional implementation, in S5, state consistency determination, such as... Figure 5 As shown, it specifically includes:

[0149] Compare the assignment values ​​of each real state prototype corresponding to the shared state feature on the pressure side in the real state prototype assignment results, determine the real state prototype with the largest assignment value, and form the real state prototype identifier on the pressure side.

[0150] The assignment values ​​of each real state prototype corresponding to the shared state feature on the acoustic side in the real state prototype assignment results are compared, and the real state prototype with the largest assignment value is determined to form the real state prototype identifier on the acoustic side.

[0151] The pressure side real state prototype identifier is compared with the sound wave side real state prototype identifier. If the pressure side real state prototype identifier and the sound wave side real state prototype identifier are the same, a state consistent measurement segment identifier is formed.

[0152] When the prototype identifier of the actual state on the pressure side differs from that on the acoustic side, a state mismatch test segment identifier is formed.

[0153] For example, the output layer of MATE only receives the results of the real-state prototype attribution. . This indicates the attribution result of the shared state features on the pressure side relative to the eight true state prototypes within the finite set of true state prototypes. Indicates the first The first test segment corresponds to the first The stress-side attribution value of a real-state prototype. ; This indicates the attribution result of the shared state features on the acoustic side relative to the eight real state prototypes within the finite set of real state prototypes. Indicates the first The first test segment corresponds to the first The acoustic side attribution value of a real-state prototype. The output layer of MATE consists of a real-state prototype identifier generation unit and a state consistency determination layer. The real-state prototype identifier generation unit includes a pressure-side comparison channel and an acoustic side comparison channel. The input of the pressure-side comparison channel is... vector The output is a prototype identifier of the actual state on the pressure side. The input to the acoustic side comparison channel is... vector The output is a prototype identifier of the real state on the acoustic side. The input to the state consistency decision layer is a scalar. and scalar The output of the state consistency determination layer is a scalar state consistency measurement segment identifier. Scalar state mismatch measurement segment identifier In this stage, MATE no longer calls the pressure response sequence, acoustic response sequence, job event timing, pressure modal characteristics, acoustic modal characteristics, and job timing characteristics; it only uses... and Complete the comparison and judgment.

[0154] Prototype identification of the actual state on the pressure side The generation process employs a recursive comparison method based on the prototype index of the actual state. The pressure-side comparison channel first reads... The first attribution value and will As the current maximum attribution value, the first true state prototype is selected as the current candidate true state prototype. The pressure-side comparison channel continues... Read them one by one in order Each time a new pressure-side attribution value is read, it is compared with the current maximum attribution value: when If the value is greater than the current maximum ownership value, update the current maximum ownership value to [value]. And update the current candidate real state prototype to the first one. A real-state prototype; when When the value is less than the current maximum attribution value, keep the current maximum attribution value and the current candidate true state prototype unchanged; when... When the value equals the current maximum attribution value, a fixed disambiguation rule is used to retain the real state prototype with the smaller index as the current candidate real state prototype. After completing the sequential comparison of the eight pressure-side attribution values, the index of the retained current candidate real state prototype is determined as the pressure-side real state prototype identifier. . For the value to belong to The scalar represents the real state prototype number with the largest shared state characteristic attribution value on the pressure side.

[0155] True prototype identification of the sound wave side The generation process maintains the same structure and recursive rules as the pressure-side comparison channel. The acoustic-side comparison channel first reads... The first attribution value and will As the current maximum attribution value, the first true state prototype is selected as the current candidate true state prototype. The acoustic side comparison channel continues... Read them one by one in order and will Compare successively with the current maximum value. When If the value is greater than the current maximum attribution value, update the current maximum attribution value and update the current candidate true state prototype to the [number]th [value]. A real-state prototype; when When it is less than the current maximum attribution value, it remains unchanged; when If the value equals the current maximum assigned value, the fixed disambiguation rule of prioritizing the smaller real-state prototype index is still applied. After comparing the eight acoustic-side assigned values, the remaining current candidate real-state prototype index is determined as the acoustic-side real-state prototype identifier. . For the value to belong to The scalar represents the real state prototype number with the largest shared state feature attribution value on the acoustic side.

[0156] exist and After generation, the state consistency determination layer... and Perform a one-to-one consistency comparison. The state consistency determination layer first reads the scalar. and scalar Then perform an equality check. When At that time, generate a consistent test segment identifier. and will The value is assigned to 1, and a state mismatch test segment identifier is generated simultaneously. and will Assign a value of 0; when At that time, generate a state mismatch test segment identifier. and will The value is assigned to 1, and the state-consistent test segment is marked. The value is assigned to 0. Therefore... and The results are mutually exclusive binary judgments, and any segment corresponds to only one segment state in the same judgment. Indicates the segment identifier that is in a consistent state. Indicates the state mismatch test segment identifier. and These represent the prototype identifiers for the pressure side and the sound wave side, respectively.

[0157] After the above processing, the results generated by MATE in this stage include the prototype identification of the actual state on the pressure side. Sound wave side real state prototype identification Consistent status segment identifier State mismatch test segment identifier .in, and Number the prototypes of the two scalar real states. and This results in two scalar binary judgments. The execution path of MATE in this stage is clearly defined as: [to determine the result of the binary judgment]. Input pressure side comparison channel to generate ,Will Input acoustic side comparison channel to generate Then and Input state consistency determination layer to generate and This processing path converts continuous assignment values ​​in the real state prototype assignment results into discrete real state prototype identifiers, and then converts the discrete real state prototype identifiers into binary determination identifiers that can be directly used for segment screening, thus fully supporting real state prototype comparison and state consistency determination.

[0158] In an optional implementation, in S6, a joint measurement result and a mismatch measurement result are formed, such as... Figure 6 As shown, it specifically includes:

[0159] Based on the consistent test segment identifier and the mismatch test segment identifier, the test segment data is filtered by conditions to form consistent test segment data and mismatch test segment data.

[0160] Under the condition of consistent measurement segment identification, the pressure response sequence and acoustic response sequence corresponding to the same measurement segment are extracted from the consistent measurement segment data, and the pressure response sequence and acoustic response sequence are associated according to the correspondence of the same measurement segment to form a joint measurement segment result;

[0161] Under the condition of state mismatch test segment identification, extract the pressure response sequence and acoustic response sequence corresponding to the same test segment from the mismatch test segment data, and determine the pressure response sequence and acoustic response sequence that do not meet the correspondence relationship of the same real state prototype as the mismatch test segment result.

[0162] For example, the first The data actually used in this stage for each test segment is the test segment identifier with consistent status. State mismatch test segment identifier Pressure response sequence Acoustic response sequence and task event sequence . For the value to belong to binary identifier, For the value to belong to The binary identifier, using Indicates the first Sampling time at each pressure sampling point Indicates the first The pressure values ​​at each pressure sampling point are used Indicates the first Sampling time of each acoustic sampling point Indicates the first The amplitude value of each sound wave sampling point to These sequentially represent the operational time points corresponding to packer inflation, pressure stabilization start, pressure stabilization end, pressure release start, probe switching, and probe reset. Consistent status measurement section identifiers. Mismatch with state measurement segment identifier Mutually exclusive use, when Perform consistent filtering at the same time, when Mismatch screening is performed at that time.

[0163] when First, sequence the task events. Directly related to stress testing , , and Sampling time axis mapped to pressure response sequence Above. The mapping rule is executed using the nearest neighbor positioning method: for each job time point. The sampling time with the smallest absolute time difference among 128 pressure sampling times was identified. And record the corresponding sampling index as . , , and These represent the corresponding sampling positions in the pressure response sequence for packer inflation, pressure stabilization start, pressure stabilization end, and pressure relief start, respectively. Only when... Only then is the pressure response sequence considered to satisfy the stage sequence corresponding to the operation event timing. Based on , , and Each of the 128 pressure sampling points was assigned a stage position mark. .when At that time, A value of 0 indicates the pressure rise stage; when At that time, A value of 1 indicates the voltage stabilization stage; when At that time, A value of 2 is assigned to indicate the depressurization stage; the remaining positions maintain their original sampling order and are recorded as positions outside the boundary. This forms a time-location matched pressure response sequence. ,in The three fields represent the pressure sampling location index, pressure value, and stage location marker, respectively.

[0164] Sequence of job events Directly related to sound wave testing and Sampling time axis mapped to the acoustic response sequence Above. The mapping rules are also executed using the nearest neighbor location method: for and Find the sampling time with the smallest absolute time difference among the 256 sound wave sampling times. And record the corresponding sampling index as and . This indicates the corresponding sampling position in the acoustic response sequence where the probe is switched. This indicates the corresponding sampling position of the probe reset in the acoustic response sequence. Only when... Only then is it considered that the acoustic response sequence satisfies the propagation order corresponding to the timing of the operation event. Based on and Position markers were assigned to each of the 256 acoustic sampling points. .when At that time, A value of 0 indicates the position of the probe before switching; when At that time, A value of 1 indicates that the probe has switched to a position between probe rod reset and position; when At that time, A value of 2 is assigned to represent the position of the probe after reset. This forms the acoustic response sequence after time-position matching. ,in The three fields represent the acoustic sampling location index, amplitude value, and location marker, respectively.

[0165] Only when , and Only when both conditions are met will the association under the same measurement segment correspondence be executed. The association method is: indexed by the same measurement segment. As a unique binding key, the pressure response sequence after matching the time position. Acoustic response sequence matched with time and location Corresponding data is retained, cross-segment splicing is not performed, and the sampling order within the two sequences is not changed. The resulting joint segment result is obtained from... and Together they constitute, among which A sequence consisting of 128 three-field records. It is a sequence consisting of 256 three-field records.

[0166] when At times, or although but Not valid, or If the condition is not met, the joint association is not performed. In this case, the original pressure response sequence is preserved. and the original acoustic response sequence The original pressure response sequence and the original acoustic response sequence were indexed according to the same measurement segment. Parallel determination of the results of the mismatched test segments indicates that the pressure response sequence and acoustic response sequence do not meet the correspondence of the same real-state prototype, or do not meet the time position relationship of the operation event, and do not include the joint correlation structure after time position matching.

[0167] After the above screening and matching processes, the newly generated or updated data in this stage only includes the pressure response sequence after time-location matching. Acoustic response sequence after time-location matching The results include joint measurement segment results and mismatched measurement segment results. Joint measurement segment results correspond to measurement segments where the state-consistent measurement segment identifier is valid, while mismatched measurement segment results correspond to measurement segments where the state-mismatched measurement segment identifier is valid or the time-position sequence constraint is invalid. The entire processing revolves around the same measurement segment index. This allows for the condition screening, time and location matching, and result differentiation of the consistent state measurement segment identifier, the mismatched state measurement segment identifier, the pressure response sequence, the acoustic response sequence, and the operation event time sequence within the same measurement segment range.

[0168] In one optional implementation, in S7, the joint measurement result or mismatched measurement result corresponding to the measurement segment is output, such as... Figure 7 As shown, it specifically includes:

[0169] Based on the results of the real state prototype attribution, extract the real state prototype identifier corresponding to the same test segment.

[0170] Based on the consistent state segment identifier, condition judgment is performed, and the joint segment result corresponding to the same segment is called under the condition corresponding to the consistent state segment identifier.

[0171] Based on the state mismatch test segment identifier, condition judgment is performed, and the mismatch test segment result corresponding to the same test segment is called under the condition corresponding to the state mismatch test segment identifier;

[0172] The consistent state test segment identifier and its corresponding joint test segment result, and the mismatch state test segment identifier and its corresponding mismatch test segment result are combined according to the same test segment correspondence to form the output result.

[0173] For example, this step belongs to the result construction step. The data actually called in this step is the section number. Pressure-side real-state prototype identification Sound wave side real state prototype identification Consistent status segment identifier State mismatch test segment identifier Pressure response sequence after time-location matching Acoustic response sequence after time-location matching Original pressure response sequence and the original acoustic response sequence . Indicates the first The section number of each measurement segment; Indicates the first Prototype identifier of the actual pressure side state of each test section; Indicates the first The prototype identifier of the actual state of the acoustic side of each measurement segment; Indicates the first The status of each measurement segment is consistent with the measurement segment identifier; Indicates the first Individual test segment status mismatch test segment identifier; Indicates the first time position after matching Each pressure record contains three fields: pressure sampling location index, pressure value, and stage location marker, respectively. Indicates the first time position after matching Each sound wave record has three fields: sound wave sampling location index, amplitude value, and location marker.

[0174] First, based on the prototype attribution results of the actual state, the results have already been extracted. and Generate prototype identifiers for the actual state of the same measurement segment. To ensure a fixed output field structure, [the following will be implemented]. and Combined in order . for The identifier vector stores the prototype identifier of the actual state on the pressure side in the first position and the prototype identifier of the actual state on the sound wave side in the second position. When... hour, The two positions in the middle have the same value; when hour, Two positions retain their original judgment values ​​to preserve the difference information of the true state prototype in the case of state mismatch. Through this processing, the discrete judgment results in the true state prototype attribution results are organized into true state prototype identifiers that can be directly called within the same measurement segment.

[0175] After the prototype identifiers for the actual state are organized, they are identified according to the state-consistent segment identifiers. Mismatch with state measurement segment identifier Execution condition judgment. This step only processes two valid states: the first is... and The second type is and .when and At that time, call the same measurement section number The corresponding time location matched pressure response sequence Acoustic response sequence matched with time and location The call process does not change. The order of the 128 pressure records remains unchanged. The arrangement order of the 256 sound wave records is only... , , , and Write the same result record using a fixed field order. The result content written at this time corresponds to the combined measurement segment result, which is directly derived from the pressure response sequence after time-location matching. Acoustic response sequence matched with time and location The data is no longer encapsulated into a new dataset object.

[0176] when and At that time, call the same measurement section number The corresponding original pressure response sequence and the original acoustic response sequence Maintain during the call. The sampling time order of the 128 pressure sampling records remains unchanged, maintaining... The sampling time order of the 256 acoustic wave sampling records remains unchanged, only the sampling time order is changed. , , , and Write the same result record in the same order of fields that correspond to the state. The result content written at this time is the result content corresponding to the mismatched test segment result, which is directly derived from the original pressure response sequence. and the original acoustic response sequence It consists of no additional joint or related structures.

[0177] Therefore, the order of the fields in the result record generated in this step remains fixed: the first field is the section number. The second field is the real-state prototype identifier. The third field is the identifier for the consistent test segment. Fields 4 and 5 contain the measurement results. When At that time, the fourth field is written with the pressure response sequence after the time position is matched. The fifth field contains the acoustic response sequence after time and location matching. ;when At that time, the fourth field is written with the original pressure response sequence. The fifth field contains the original acoustic response sequence. The entire construction process revolved around the same section number. Expand the results so that the segment number, the real state prototype identifier, the state-consistent segment identifier, and the joint segment result or mismatched segment result are combined in the same result record.

[0178] The relevant information for the determination is shown in Table 4.

[0179] Table 4. Relevant Information for Judgment

[0180]

[0181] In one alternative implementation, by simultaneously expressing attribute values ​​and uncertainties in the same graph, the joint interpretation results and the degree of credibility are presented synchronously, making it easier for subsequent interpreters to identify reliable and risky segments.

[0182] This invention also provides a kilometer-level ground stress and wave velocity joint testing system to realize the kilometer-level ground stress and wave velocity joint testing method as described above. The system includes a test segment data acquisition module, a modal feature group extraction module, a sharing and unique decomposition module, a real state prototype attribution module, a state consistency determination module, a test segment adaptation module, and an output module.

[0183] The segment test data acquisition module is used to acquire the pressure response sequence of the same test segment in the sealing pressurization test, the acoustic response sequence in the in-hole acoustic wave test, and the operation event sequence to form the test segment test data.

[0184] The modal feature group extraction module is used to input the test data of the test section into MATE, and extract the pressure modal features corresponding to the pressure response sequence, the acoustic modal features corresponding to the acoustic response sequence, and the operation time sequence features corresponding to the operation event time sequence to form a modal feature group.

[0185] The shared and unique decomposition module is used to perform shared and unique decomposition on the modal feature group to obtain pressure-side shared state features, acoustic wave-side shared state features, pressure-side modal unique features, and acoustic wave-side modal unique features.

[0186] The real state prototype attribution module is used to introduce POT in the state attribution process of real state shared features, establish a finite real state prototype set, and use the finite real state prototype set as a common reference to perform similarity-aware optimal transmission allocation on pressure-side shared state features and sound-wave-side shared state features to form real state prototype attribution results.

[0187] The state consistency determination module is used to determine the state consistency based on the real state prototype attribution result. When the pressure side shared state features and the sound wave side shared state features are assigned to the same real state prototype, a state consistency measurement segment identifier is formed. When the pressure side shared state features and the sound wave side shared state features are not assigned to the same real state prototype, a state mismatch measurement segment identifier is formed.

[0188] The test segment adaptation module is used to associate the pressure response sequence and the acoustic response sequence from the test segment test data, and form a joint test segment result under the condition that the test segment identifiers are consistent in state, and form a mismatched test segment result under the condition that the test segment identifiers are mismatched in state.

[0189] The output module is used to output the joint test segment results or mismatch test segment results corresponding to the test segment.

Claims

1. A method for joint testing of ground stress and wave velocity at the kilometer level, characterized in that, Includes the following steps: S1. Obtain the pressure response sequence of the same test section during the sealing pressurization test, the acoustic response sequence during the in-hole acoustic test, and the timing of the operation events to form the test section test data. S2. Input the test data of the test section into MATE, and extract the pressure modal features corresponding to the pressure response sequence, the acoustic modal features corresponding to the acoustic response sequence, and the operation time sequence features corresponding to the operation event time sequence to form a modal feature group. S3. Perform shared and unique decomposition on the modal feature group, and extract the real state shared features, which include pressure-side shared state features and acoustic-side shared state features. S4. In the state attribution process of shared features of real state, POT is introduced to establish a finite set of real state prototypes. Using the finite set of real state prototypes as a common reference, similarity-aware optimal transmission allocation is performed on the shared state features of the pressure side and the shared state features of the sound wave side to form the real state prototype attribution result. S5. Based on the results of the assignment of the real state prototype, determine the consistency of the state. When the shared state features of the pressure side and the shared state features of the acoustic side are assigned to the same real state prototype, a state-consistent measurement segment identifier is formed. When the shared state features of the pressure side and the shared state features of the acoustic side are not assigned to the same real state prototype, a state-mismatch measurement segment identifier is formed. S6. Correlate the pressure response sequence and acoustic response sequence from the test data of the test section. Under the condition that the test section identifiers correspond to the same state, form a joint test section result. Under the condition that the test section identifiers correspond to the mismatched state, form a mismatched test section result. S7. Output the joint test segment results or mismatch test segment results corresponding to the test segment.

2. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S1, the test data for the measurement segment is generated, specifically including: During the sealing and pressurization test, multiple pressure sampling points are continuously collected around the same test section, and the pressure sampling points are arranged in chronological order of the pressure rise stage, pressure stabilization stage and pressure relief stage to form a pressure response sequence. During the acoustic wave test inside the hole, multiple acoustic wave sampling points are continuously collected around the same test section, and the acoustic wave sampling points are arranged according to the arrival time of the first wave and the amplitude decay time to form an acoustic wave response sequence. Based on the operation time points corresponding to packer expansion, pressure stabilization start, pressure stabilization end, pressure release start, probe switching and probe reset, the operation events are arranged in order of occurrence to form an operation event sequence. The pressure response sequence, acoustic response sequence, and operation event time sequence are combined according to the corresponding relationship of the same test segment to form the test segment data.

3. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S2, modal feature sets are formed, specifically including: The pressure response sequence from the test data of the test section is input into MATE. The pressure response sequence is processed by two layers of convolution and one layer of fully connected mapping according to the sampling order. Pressure modal features reflecting the changes in the pressure rise stage, pressure stabilization stage and pressure relief stage are extracted from the pressure response sequence. The acoustic response sequence from the test data of the section is input into MATE. The acoustic response sequence is subjected to two layers of convolution calculation and one layer of fully connected mapping in the order of sampling. The acoustic modal features reflecting the changes in the arrival of the first wave and the changes in the amplitude attenuation are extracted from the acoustic response sequence. Input the timing sequence of the operation events in the test data of the section into MATE, perform a two-layer fully connected mapping on the timing sequence of the operation events according to the order of event occurrence, form the timing sequence feature of the operation, and preserve the time position relationship of the packer expansion and contraction unloading, probe switching and probe reset in the timing sequence feature of the operation. Based on the temporal position relationship in the operation sequence characteristics, the stage change position in the pressure modal characteristics is correspondingly calibrated, so that the pressure change in the pressure modal characteristics is associated with the time position corresponding to the packer expansion and contraction unloading, thus forming a pressure modal characteristic corresponding to the operation event sequence. Based on the temporal position relationship in the operation sequence characteristics, the propagation change position in the acoustic modal characteristics is correspondingly calibrated, so that the propagation change in the acoustic modal characteristics is associated with the time position corresponding to probe switching and probe reset, thus forming acoustic modal characteristics corresponding to the operation event sequence. The pressure modal features, acoustic modal features, and operation time sequence features corresponding to the operation event time sequence are collected according to the correspondence of the same measurement segment to form a modal feature group.

4. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S3, shared features of the real state are extracted, specifically including: The pressure modal features in the modal feature group are combined with the operation time sequence features to locate the change part in the pressure modal features that is consistent with the time of packer expansion and contraction unloading, forming the pressure side candidate disturbance part and the pressure side candidate shared part. The acoustic modal features in the modal feature group are combined with the operation time sequence features to locate the part of the acoustic modal features that is consistent with the time of the change in transducer contact, forming the acoustic side candidate disturbance part and the acoustic side candidate shared part. Perform sharing and unique decomposition on the candidate shared parts of the pressure side and the candidate shared parts of the acoustic wave side, calculate the common change components between the candidate shared parts of the pressure side and the candidate shared parts of the acoustic wave side, and determine the common change components as the shared state features; Based on the time position corresponding to the packer expansion and contraction unloading in the operation time sequence characteristics, the unique component extraction is performed on the candidate disturbance part of the pressure side. The change part driven by the packer expansion and contraction unloading is identified as the pressure side disturbance part, and the pressure side disturbance part is identified as part of the modal unique features. Based on the time position corresponding to the change in transducer contact in the operation time sequence characteristics, the unique component is extracted from the candidate disturbance part on the acoustic side. The part driven by the change in transducer contact is identified as the acoustic side disturbance part, and the acoustic side disturbance part is identified as another part of the modal unique features. The shared state characteristics are simultaneously preserved from the pressure mode characteristics and the acoustic mode characteristics, and the pressure-side disturbance part and the acoustic-side disturbance part are not included in the shared state characteristics, thus forming a true state shared characteristic that represents the true formation state of the same section.

5. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S4, the results of the real state prototype attribution are formed, specifically including: The real state shared features are input into the state attribution link of MATE. POT is introduced into the state attribution link of MATE, and a finite real state prototype set is established based on the distribution relationship of the real state shared features that jointly reflect the real stratigraphic state of the same section. The pressure-side shared state features are calculated one by one with the finite set of real state prototypes to obtain the similarity between the pressure-side shared state features and each real state prototype in the finite set of real state prototypes, and the state attribution calculation input of the pressure-side shared state features is formed. The acoustic wave-side shared state features are calculated one by one with the finite set of real state prototypes to obtain the similarity between the acoustic wave-side shared state features and each real state prototype in the finite set of real state prototypes, and the state attribution calculation input of the acoustic wave-side shared state features is formed. Using a finite set of real state prototypes as a common reference, the state assignment calculation inputs of the pressure-side shared state features and the sound-wave-side shared state features are subjected to similarity-aware optimal transmission allocation within the POT, so that the state assignment updates of the pressure-side shared state features and the sound-wave-side shared state features are completed within the same finite set of real state prototypes. Based on the updated pressure-side shared state characteristics and the updated acoustic-side shared state characteristics, the true state prototype assignment results corresponding to the pressure-side shared state characteristics and the true state prototype assignment results corresponding to the acoustic-side shared state characteristics are determined respectively. The results of assigning the true state prototypes corresponding to the shared state characteristics on the pressure side and the results of assigning the true state prototypes corresponding to the shared state characteristics on the acoustic side are aggregated according to the correspondence of the same measurement segment to form the true state prototype assignment results.

6. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S5, state consistency determination specifically includes: Compare the assignment values ​​of each real state prototype corresponding to the shared state feature on the pressure side in the real state prototype assignment results, determine the real state prototype with the largest assignment value, and form the real state prototype identifier on the pressure side. The assignment values ​​of each real state prototype corresponding to the shared state feature on the acoustic side in the real state prototype assignment results are compared, and the real state prototype with the largest assignment value is determined to form the real state prototype identifier on the acoustic side. The pressure side real state prototype identifier is compared with the sound wave side real state prototype identifier. If the pressure side real state prototype identifier and the sound wave side real state prototype identifier are the same, a state consistent measurement segment identifier is formed. When the prototype identifier of the actual state on the pressure side differs from that on the acoustic side, a state mismatch test segment identifier is formed.

7. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S6, the joint measurement results and the mismatched measurement results are generated, specifically including: Based on the consistent test segment identifier and the mismatch test segment identifier, the test segment data is filtered by conditions to form consistent test segment data and mismatch test segment data. Under the condition of consistent measurement segment identification, the pressure response sequence and acoustic response sequence corresponding to the same measurement segment are extracted from the consistent measurement segment data, and the pressure response sequence and acoustic response sequence are associated according to the correspondence of the same measurement segment to form a joint measurement segment result; Under the condition of state mismatch test segment identification, extract the pressure response sequence and acoustic response sequence corresponding to the same test segment from the mismatch test segment data, and determine the pressure response sequence and acoustic response sequence that do not meet the correspondence relationship of the same real state prototype as the mismatch test segment result.

8. The method for joint testing of kilometer-level ground stress and wave velocity according to claim 1, characterized in that, In S7, the output of the joint measurement segment results or mismatched measurement segment results for the corresponding measurement segment includes: Based on the results of the real state prototype attribution, extract the real state prototype identifier corresponding to the same test segment. Based on the consistent state segment identifier, condition judgment is performed, and the joint segment result corresponding to the same segment is called under the condition corresponding to the consistent state segment identifier. Based on the state mismatch test segment identifier, condition judgment is performed, and the mismatch test segment result corresponding to the same test segment is called under the condition corresponding to the state mismatch test segment identifier; The consistent state test segment identifier and its corresponding joint test segment result, and the mismatch state test segment identifier and its corresponding mismatch test segment result are combined according to the same test segment correspondence to form the output result.

9. A kilometer-level combined ground stress and wave velocity testing system, characterized in that, To implement the kilometer-level joint testing method for ground stress and wave velocity as described in any one of claims 1-8, the system includes a test segment data acquisition module, a modal feature group extraction module, a sharing and unique decomposition module, a real state prototype attribution module, a state consistency determination module, a test segment adaptation module, and an output module. The test data acquisition module is used to acquire the pressure response sequence of the same test section in the sealing pressurization test, the acoustic response sequence in the in-hole acoustic test, and the timing of operation events to form test data of the test section. The modal feature group extraction module is used to input the test data of the test section into MATE, and extract the pressure modal features corresponding to the pressure response sequence, the acoustic modal features corresponding to the acoustic response sequence, and the operation time sequence features corresponding to the operation event time sequence to form a modal feature group. The shared and unique decomposition module is used to perform shared and unique decomposition on the modal feature group to obtain pressure-side shared state features, acoustic wave-side shared state features, pressure-side modal unique features, and acoustic wave-side modal unique features. The real state prototype attribution module is used to introduce POT in the state attribution process of real state shared features, establish a finite real state prototype set, and use the finite real state prototype set as a common reference to perform similarity-aware optimal transmission allocation on pressure-side shared state features and sound-wave-side shared state features to form real state prototype attribution results. The state consistency determination module is used to determine the state consistency based on the real state prototype attribution result. When the pressure side shared state features and the sound wave side shared state features are assigned to the same real state prototype, a state consistency measurement segment identifier is formed. When the pressure side shared state features and the sound wave side shared state features are not assigned to the same real state prototype, a state mismatch measurement segment identifier is formed. The test segment adaptation module is used to associate the pressure response sequence and the acoustic response sequence from the test segment test data, and form a joint test segment result under the condition that the test segment identifiers are consistent in state, and form a mismatched test segment result under the condition that the test segment identifiers are mismatched in state. The output module is used to output the joint test segment results or mismatch test segment results corresponding to the test segment.

Images