Frozen well inner layer collapsible well wall monitoring system and monitoring early warning method

Abstract

The application provides a frozen well inner layer collapsible well wall monitoring system and a monitoring and early warning method, relates to the technical field of well wall safety monitoring, and collects well wall displacement, temperature gradient and wall pressure, and maps corresponding characteristic quantities through deviation elimination and range proportional scaling; evolution mutation characteristics of displacement and temperature characteristic quantities are extracted, coupled gradient is determined through vector synthesis and dynamic gain, the coupled gradient is mapped as early warning excitation strength, adjustment damping is determined in combination with the convergence degree of displacement relative to preset limit characteristic quantities, time delay sensitivity coefficients are obtained through interactive analysis, a stress saturation term is constructed based on pressure characteristic quantities to generate a feedback adjustment factor, the sensitivity coefficient is combined to correct a static reference to generate a dynamic alarm characteristic threshold, and finally the deviation degree of displacement from the dynamic threshold is mapped to a safety redundancy space to solve a danger index and determine a danger level, so that nonlinear instability boundary quantification and early warning of a frozen well wall under time-varying disturbance are realized.

Description

Technical Field

[0001] This invention relates to the field of wellbore safety monitoring technology, specifically to a wellbore monitoring system and monitoring and early warning method for the shrinkable inner layer of a frozen well. Background Technology

[0002] As mineral resources are mined into deeper strata, the freezing method faces the dual challenges of unstable high ground pressure and alternating temperature fields. The inner collapsible wellbore, as a key support structure in deep frozen wells, is subject to multiple coupling effects from surrounding rock displacement convergence, frozen wall temperature field fluctuations, and alternating inter-wall pressure, making it highly susceptible to structural instability and even rupture. Accurate monitoring of the wellbore's evolution is a physical prerequisite for ensuring the safe operation of deep wells. However, wellbore structural damage is highly concealed and exhibits nonlinear evolution characteristics. Traditional periodic manual inspections or simple sensor monitoring methods often fail to accurately capture the nonlinear distortion of early micro-strain when subjected to alternating load impacts.

[0003] Existing methods for monitoring the walls of frozen wells typically involve embedding displacement, strain, or pressure sensors within the structure to assess the stress state of the well wall by collecting a single-dimensional physical quantity. These methods largely rely on converting the raw signal into instantaneous displacement or pressure values ​​after simple range scaling. In terms of early warning logic, current technologies generally employ fixed safety thresholds set based on expert experience, triggering alarms by determining whether the monitored value exceeds a preset threshold. When processing data, such solutions often treat displacement, temperature, and pressure as independent discrete parameters, focusing on monitoring the absolute magnitude of the values.

[0004] In actual deep environments, there is a complex nonlinear time-delay coupling relationship between wellbore displacement evolution and temperature disturbances. The linear accumulation of a single parameter cannot accurately reflect deep-seated anomalies in structural stress. However, existing monitoring methods lack vector synthesis and dynamic gain logic for multivariable physical characteristics, making it difficult to decouple the interference relationships between parameters, resulting in significant physical lag in the assessment results. Furthermore, deep formation pressure evolves in real time with the state of the frozen wall, and traditional fixed thresholds are unable to adapt to the dynamic shift of the wellbore bearing limit, easily causing early warning delays or frequent false alarms, making it difficult to accurately quantify the true instability boundary of the wellbore under multi-source disturbances.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a wellbore wall monitoring system and monitoring and early warning method for the shrinkable inner layer of frozen wells, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A wellbore monitoring system for the shrinkable inner layer of a frozen well, specifically comprising: The feature construction module is used to collect well wall displacement, temperature gradient and interwall pressure of the well wall to be monitored within a preset sampling window, and map them into corresponding displacement feature quantities, temperature feature quantities and pressure feature quantities after deviation elimination and range scaling. The feature quantities at each sampling time are combined in time sequence to construct a feature parameter matrix. The coupling gradient module is used to extract the evolutionary abrupt features of the displacement and temperature features in the feature parameter matrix, obtain a perturbation vector through vector synthesis, and dynamically gain the perturbation vector according to the instantaneous distribution ratio between the displacement and temperature features to obtain the coupling gradient that reflects the degree of multivariate coupling influence. The time delay analysis module is used to convert the coupled gradient into the warning excitation intensity using a nonlinear mapping function, and to determine the adjustment damping based on the convergence of the displacement feature quantity at the current sampling time relative to the preset displacement limit feature quantity. The time delay sensitivity coefficient is obtained through the interaction operation of the warning excitation intensity and the adjustment damping and the normalized mapping constraint analysis. The dynamic threshold module is used to construct a force saturation term based on the pressure characteristics at the current sampling time, and generate a feedback adjustment factor using the saturation constraint function. The correlation between the feedback adjustment factor and the time delay sensitivity coefficient is used to correct the preset static benchmark feature threshold and generate a dynamic alarm feature threshold. The hazard determination module is used to map the deviation between the displacement feature quantity at the current sampling time and the dynamic alarm feature threshold to a safety redundancy space defined by the preset displacement limit feature quantity for evaluation, generate a hazard index and determine the hazard level.

[0008] Furthermore, within the preset sampling window, the raw data of wellbore displacement, temperature gradient, and interwall pressure at each sampling time are obtained; The sampling window is defined as a continuous window that extends backward from the current sampling time, with the current sampling time as the endpoint. A sliding window at each preceding sampling time, where It is a preset positive integer; The collected raw data undergoes deviation elimination and range scaling mapping. Specifically, for any sampling time, the corresponding preset reference value is subtracted from each raw data to obtain the corresponding displacement deviation value, temperature deviation value, and pressure deviation value. Then, using the preset displacement range, preset temperature range, and preset pressure range, the displacement deviation value, temperature deviation value, and pressure deviation value are linearly normalized and mapped to obtain the displacement characteristic quantity, temperature characteristic quantity, and pressure characteristic quantity at the corresponding sampling time.

[0009] Furthermore, the specific steps for constructing the feature parameter matrix include: The displacement, temperature, and pressure features at the same sampling time are encapsulated into an instantaneous feature vector; The instantaneous feature vectors within the preset sampling window are stacked vertically in chronological order of sampling time to generate a feature parameter matrix with temporal dimension.

[0010] Furthermore, the specific steps for obtaining the coupling gradient are as follows: Extract the displacement feature sequence and temperature feature sequence of the current sampling time and the two consecutive sampling times before it from the feature parameter matrix, and use the discrete numerical difference algorithm to calculate the second-order displacement difference value and the second-order temperature difference value of the current sampling time, so as to serve as the second-order displacement change feature and the second-order temperature change feature of the current sampling time. The displacement second-order abrupt change feature and the temperature second-order abrupt change feature are squared respectively, and then the summation and square root are taken to obtain the disturbance vector reflecting the intensity of multi-source disturbance; Extract the displacement and temperature features at the current sampling time, sum the temperature features with a preset minimum positive bias constant, calculate the ratio of the displacement features to the sum, and perform an absolute value operation on the ratio as a dynamic gain reflecting the dominant physical evolution weight. The dynamic gain is multiplied and weighted with the disturbance vector to obtain the coupling gradient at the current sampling time, which reflects the degree of multivariate coupling influence. The specific calculation formula is as follows: In the formula, Sampling time The coupling gradient, It is a second-order discrete difference operator. Sampling time The corresponding displacement characteristic quantity, Sampling time The corresponding temperature characteristic quantity, The sampling time number. This is the preset minimum positive bias constant.

[0011] Furthermore, the specific steps for obtaining the time delay sensitivity coefficient are as follows: Using the natural constant as the base, the coupling gradient at the current sampling time is exponentially calculated to obtain the warning excitation intensity; Calculate the ratio of the displacement characteristic quantity at the current sampling time to the preset displacement limit characteristic quantity, and perform an absolute value operation on the ratio to obtain the deformation convergence. Summing the deformation convergence with the natural constant and then performing a natural logarithmic operation, the adjustable damping is obtained. The initial sensitivity is obtained by calculating the ratio of the warning excitation intensity to the adjustment damping. A preset safety upper limit coefficient is then used to truncate the initial sensitivity, and the smaller of the two values ​​is taken as the time-delay sensitivity coefficient. The specific calculation formula is as follows: In the formula, Sampling time The time delay sensitivity coefficient, The preset safety upper limit coefficient, Sampling time The coupling gradient, Sampling time The corresponding displacement characteristic quantity, For the preset displacement limit characteristic quantity, This is the sampling time sequence number.

[0012] Furthermore, the specific steps for generating the dynamic alarm feature threshold are as follows: The absolute value operation is performed on the pressure feature at the current sampling time to construct the force saturation term, and the hyperbolic tangent function operation is performed on the force saturation term to obtain the feedback adjustment factor; Calculate the product of the time delay sensitivity coefficient, the feedback adjustment factor, and the preset sensitivity coefficient compensation factor, and label the product as a dynamic reduction term. The value range of the sensitivity coefficient compensation factor is greater than a constant zero and less than the reciprocal of the preset theoretical maximum value of the time delay sensitivity coefficient. A dynamic adjustment gain is constructed using the difference between a constant and the dynamic reduction term. This dynamic adjustment gain is then multiplied by a preset static baseline characteristic threshold to generate a dynamic alarm characteristic threshold. The specific calculation formula is as follows: In the formula, Sampling time Dynamic alarm feature threshold, The preset static baseline feature threshold, The preset sensitivity coefficient compensation factor is, where, The range of values ​​is , The theoretical maximum value of the time delay sensitivity coefficient is preset. Sampling time The time delay sensitivity coefficient, Sampling time The pressure characteristic quantity, This is the sampling time sequence number.

[0013] Furthermore, the preset displacement limit feature quantity is used as the upper limit of judgment, and the dynamic alarm feature threshold is used as the lower limit of judgment. A safety redundancy space is constructed based on the algebraic difference between the preset displacement limit feature quantity and the dynamic alarm feature threshold, wherein the preset displacement limit feature quantity is always greater than the dynamic alarm feature threshold. The specific steps for obtaining the risk index are as follows: Calculate the algebraic difference between the displacement feature at the current sampling time and the corresponding dynamic alarm feature threshold, and calibrate the algebraic difference as the over-limit deviation; The ratio of the out-of-limit deviation to the safety redundancy space is calculated as the danger index corresponding to the current sampling time.

[0014] Furthermore, the specific steps for determining the hazard level are as follows: The danger index obtained at the current sampling time is compared with the preset level range. The preset level range includes a normal safety range, a primary warning range, a high danger range and a limit destruction range, which are progressively defined by a constant zero, a preset warning threshold and a constant one. The preset warning threshold is a value that is greater than constant zero and less than constant one. When the danger index is less than or equal to a constant zero, it falls into the normal safe range, and the danger level of the well wall to be monitored is determined to be the safe level. When the danger index is greater than a constant zero and less than or equal to the preset warning threshold, it falls into the primary warning range, and the danger level of the well wall to be monitored is determined to be the primary warning level. When the danger index is greater than the preset warning threshold and less than a constant, it falls into the high-risk range, and the danger level of the well wall to be monitored is determined to be high-risk. When the danger index is greater than or equal to a constant, it falls into the limit damage range, and the danger level of the well wall to be monitored is determined to be the limit damage level.

[0015] The present invention also provides a method for monitoring and early warning of the retractable wellbore of the inner layer of a frozen well. This method is used to implement the aforementioned monitoring system for the retractable wellbore of the inner layer of a frozen well, comprising: Step 1: Collect the well wall displacement, temperature gradient and inter-wall pressure of the well wall to be monitored within the preset sampling window, and map them into the corresponding displacement characteristic quantity, temperature characteristic quantity and pressure characteristic quantity after deviation elimination and range scaling. Combine the characteristic quantities of each sampling time in time sequence to construct the characteristic parameter matrix. Step 2: Extract the evolution and abrupt change features of the displacement and temperature features in the feature parameter matrix, obtain the perturbation vector through vector synthesis, and dynamically gain the perturbation vector according to the instantaneous distribution ratio between the displacement and temperature features to obtain the coupling gradient that reflects the degree of multivariate coupling influence. Step 3: Use a nonlinear mapping function to convert the coupled gradient into a warning excitation intensity, and determine the adjustment damping based on the convergence of the displacement feature quantity at the current sampling time relative to the preset displacement limit feature quantity. The time delay sensitivity coefficient is obtained through the interaction operation of the warning excitation intensity and the adjustment damping and the normalized mapping constraint analysis. Step 4: Construct a force saturation term based on the pressure characteristics at the current sampling time, and generate a feedback adjustment factor using the saturation constraint function. Use the correlation between the feedback adjustment factor and the time delay sensitivity coefficient to lower and correct the preset static benchmark characteristic threshold, and generate a dynamic alarm characteristic threshold. Step 5: Map the deviation between the displacement feature quantity at the current sampling time and the dynamic alarm feature threshold to the safety redundancy space defined by the preset displacement limit feature quantity for evaluation, generate a hazard index and determine the hazard level.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention performs vector synthesis based on the evolutionary abrupt changes of extracted displacement and temperature features, and constructs a coupled gradient by executing dynamic gain based on instantaneous allocation ratio. Subsequently, it analyzes the time-delay sensitivity coefficient by combining the interactive mapping logic of early warning excitation intensity and damping adjustment. This invention overcomes the defect of physical lag in evaluation results caused by the linear accumulation of a single parameter in traditional methods. By deeply physical integrating wellbore displacement convergence with alternating temperature field fluctuations, it captures and amplifies the stress anomaly response caused by hidden micro-strains inside the structure, thereby improving the sensitivity and foresight of wellbore instability monitoring in deep extreme environments. This invention also quantifies the stress saturation state through multi-dimensional mapping of pressure characteristics and uses a feedback adjustment factor in conjunction with the sensitivity coefficient to dynamically adjust the static benchmark threshold. This allows the alarm trigger point to automatically decrease as the wellbore load increases, rather than mechanically waiting for the monitored value to cross a fixed threshold, thus solving the problem of delayed early warning caused by the evolution of deep formation bearing capacity over time. By introducing feedback adjustment logic, transient local load fluctuations can be effectively identified and eliminated, avoiding unnecessary frequent false alarms. This threshold adjustment method transforms simple numerical comparison into state trend prediction, achieving early locking before the wellbore enters the critical danger zone, significantly improving the reliability of the early warning response under complex alternating loads. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall system modules of the present invention; Figure 2 A graph showing the relationship between the threshold sinking of dynamic alarm features based on multi-source disturbances and risk evolution; Figure 3 This is a schematic diagram of the overall method flow of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example: Please see Figures 1-2 The present invention provides a technical solution: A wellbore monitoring system for the shrinkable inner layer of a frozen well, specifically comprising: The feature construction module is used to collect well wall displacement, temperature gradient and interwall pressure of the well wall to be monitored within a preset sampling window, and map them into corresponding displacement feature quantities, temperature feature quantities and pressure feature quantities after deviation elimination and range scaling. The feature quantities at each sampling time are combined in time sequence to construct a feature parameter matrix.

[0021] In the complex environment of deep frozen wells, the inner compressible wellbore is subjected to multiple coupled interferences from the lateral pressure of the deep surrounding rock, the release of cold energy from the frozen wall, and its own shrinkage deformation. Discrete observation of a single physical quantity is insufficient to resolve the deep-seated driving forces of the disaster. Therefore, this embodiment constructs multiple monitoring sections laterally at key strata (deep high-stress rock layers or weak areas where the frozen wall intersects) of the wellbore to be monitored, based on the stratigraphic sequence distribution. Within each monitoring section, a multi-point data acquisition array is deployed at equal angles along the circumferential direction of the wellbore. Using this spatial monitoring topology, raw data on wellbore displacement, temperature gradient, and interwall pressure are simultaneously extracted within a preset sampling window. Wellbore displacement is collected to capture the macroscopic geometric convergence of each acquisition point on the monitoring section under external compression; the temperature gradient is used to reveal the non-uniform stress distribution generated by heat exchange between the inner and outer sides of the wellbore; and the interwall pressure, through the interlayer normal pressure fed back from the measuring points, objectively characterizes the actual load transfer state between structural layers. This further demonstrates how initial temperature fluctuations induce a surge in load within the structure.

[0022] To capture the trend evolution of wellbore conditions, this embodiment introduces a method that uses the current time as the endpoint and backtracks backward, including continuous... A sliding window at each preceding sampling time, where This is a preset positive integer. Considering the slow accumulation of temperature stress in the frozen wall and the instantaneous slippage due to micro-cracks in the surrounding rock, and the need to balance the recording of cumulative effects with the sensitive capture of abrupt changes in the time parameter setting, this embodiment calibrates the sampling frequency by measuring the seismic wave frequency band during tunneling. Based on the rock and soil wave velocity and frozen wall thickness during the geological exploration stage, the thickness is divided by the wave velocity to estimate the hysteresis time of stress transmission to the well wall. This time is then multiplied by the sampling frequency to calculate the window length. Based on this method, this embodiment preferably sets the sampling frequency to [value missing]. (That is, data is collected every 5 seconds), and the window length is set to 30 (i.e., a span of 150 seconds). Using... The sampling frequency can avoid the interference of high-frequency vibrations of tens of hertz generated by deep tunneling machinery or blasting operations, while being sufficient to capture the initial distortion of micro-strain induced by the release of cold energy in the frozen wall; and setting the sampling window length to 30 means that each time a data content spanning 150 seconds is captured. The 150-second sampling window can match the typical physical response cycle of the frozen surrounding rock after experiencing transient extreme load impact, where the internal stress is redistributed and completely transmitted to the inner well wall, thereby stripping out the real signal distortion caused by the surge of ultimate load.

[0023] During the stress evolution of the compressible wellwall within a frozen well, significant differences exist across physical dimensions and absolute values ​​in the wellwall contraction displacement (typically on the millimeter level), the frozen wall temperature gradient (typically on the degree Celsius / meter level), and the external frost heave pressure (typically on the megapascal level). If multidimensional numerical calculations are performed directly, the pressure parameter will be absolutely dominant, completely masking the crucial minute anomalies in the displacement parameter during the early stages of instability. Given the asynchronous hardware response of different types of sensors, this solution introduces a unified global hardware clock as a trigger benchmark. At each preset sampling point, the latest valid physical values ​​returned by the three sensors are forcibly latched and extracted, thus overcoming the underlying communication delay and ensuring forced clock phase alignment of data from all dimensions at the same sampling moment. Therefore, before performing mapping, this embodiment first pre-removes outliers from the original data: using physical extreme value criteria, jump points exceeding the sensor's linear range or logic threshold (negative pressure or displacement exceeding 100 millimeters) are removed; and combining the displacement changes between two adjacent samples, instantaneous spikes exceeding the physical limits of soil rheology are removed. If the current sampling point is determined to be an outlier, smoothing is performed using the arithmetic mean of three consecutive valid observations within the preceding window to maintain the integrity of the data on the time axis. Specifically, if there are fewer than three valid preceding observations, the arithmetic mean of all existing valid preceding observations is used for filling; if there are no valid preceding observations (i.e., the first sampling point is determined to be outlier), the preset baseline value corresponding to that dimension is used directly for filling. Through the above processing, the collected raw data undergoes deviation elimination and range scaling mapping. Specifically, for any sampling time within the preset sampling window, the raw data of wellbore displacement, temperature gradient, and interwall pressure are acquired. Each raw data point is subtracted from its corresponding preset baseline value to remove static background interference under the initial equilibrium state, extracting displacement, temperature, and pressure deviation values ​​purely caused by disturbances. Subsequently, the displacement, temperature, and pressure deviation values ​​are linearly normalized and mapped according to the preset displacement, temperature, and pressure ranges corresponding to the wellbore structure to obtain the displacement, temperature, and pressure characteristic quantities at the corresponding sampling time. Under extreme conditions in deep formations, irreversible and drastic deformation of the wellbore structure can cause the calculated results of characteristic quantities to far exceed the preset rated range, easily leading to numerical overflow. This embodiment applies truncation constraints to the mapped characteristic quantities, limiting their numerical range to... Within this range, the operation is ensured to always be anchored within physically sensitive active regions, rather than operating in invalid regions where values ​​diverge. By setting the upper limit to 1.2 instead of 1.0, this scheme reserves space for actual physical evolution. The over-limit capture redundancy enables the early warning algorithm to identify and quantify extreme destructive states that exceed the design limits, reducing the risk of early warning response failure caused by numerical divergence; at the same time, the lower limit is hard-truncated to 0 to ensure that the early warning calculation always focuses on the damaging positive compressive load.

[0024] The specific calculation formula for the displacement characteristic quantity is as follows: In the formula, Sampling time The corresponding displacement characteristic quantity, Sampling time The raw data of wellbore displacement, This is the preset reference value corresponding to the wellbore displacement. To preset the displacement range, This is the sampling time sequence number.

[0025] The specific formula for calculating temperature characteristic quantities is as follows: In the formula, Sampling time The corresponding temperature characteristic quantity, Sampling time The raw data of the temperature gradient, This is a preset reference value corresponding to the temperature gradient. To preset the temperature range, This is the sampling time sequence number.

[0026] The specific formula for calculating the pressure characteristic quantity is as follows: In the formula, Sampling time The corresponding pressure characteristic quantity, Sampling time The raw data of the inter-wall pressure, This is the preset reference value corresponding to the inter-wall pressure. To preset the pressure range, This is the sampling time sequence number.

[0027] Among them, the displacement characteristic quantity represents the real-time proportion of the actual shrinkage of the well wall relative to the pressure relief limit of the compressible interlayer; it is positively driven by the measured well wall displacement, and the static geometric offset after demolding is obtained by subtracting the preset displacement benchmark value, and the preset displacement range is used as the scale damping to reflect the physical avoidance boundary; when this characteristic quantity is large, the buffer space is about to be exhausted and the inner layer faces the risk of rigid compression, while a small value indicates that it is still in the safe elastic pressure relief period. The temperature characteristic quantity represents the degree to which the thermal driving strength of the frozen wall approaches the design allowable limit; it increases positively with the measured temperature gradient, and the temperature of the space is filtered out by subtracting the preset temperature benchmark value, and is negatively constrained by the preset temperature range; a large value means that the cold energy is released rapidly, forming a steep thermal stress field of easily tearable concrete, while a small value indicates that the internal and external heat exchange is slow. The pressure characteristic quantity represents the stress saturation degree of the external frost heave load relative to the ultimate compressive strength of the inner concrete layer. It is positively driven by the measured inter-wall pressure, subtracting the preset pressure benchmark value to unload the original hydrostatic soil pressure base, and mapping the bearing difference of the structure from plastic yield to plastic yield using the preset pressure range. When this characteristic quantity is large, the frost heave compressive stress is rapidly consuming the bearing margin, and physical fracture is imminent; when it is small, it indicates that the structure is under stress within the elastic allowable range. Using the preset range as a common denominator scale factor, the larger the value, the deeper the deformation absorption capacity, thermal inertia, and mechanical resistance margin of the structure, and the more gradual the growth of the corresponding characteristic quantity. By transforming discrete absolute physical quantities into dimensionless indicators that reflect the degree of approach to actual danger, the destructive energy is accurately separated and the judgment error caused by local environmental disturbances is avoided.

[0028] In this embodiment, the preset benchmark values ​​corresponding to the aforementioned raw data represent the objective background state of the well wall under initial equilibrium conditions. This is achieved by completing the inner well wall casting and ensuring the frozen wall is in a stable, closed-loop state (i.e., the temperature of the main surface shown by the peripheral temperature measuring holes drops to a certain level). For the initial observation period (within a continuous 72-hour period) where the data remains stable for several consecutive days, the average value of the raw data from each monitoring session is taken as the preset baseline value. The preset baseline value corresponding to the wellbore displacement is set as follows: (That is, taking the cast surface as the initial zero point), the preset reference value corresponding to the temperature gradient is set according to the initial geothermal field. The preset reference value for the inter-wall pressure is set based on the hydrostatic earth pressure. The preset ranges for the aforementioned wellbore structure represent the physical tolerance limits of the wellbore structure before it enters a failure state. These ranges are directly extracted and set based on the wellbore design drawings and the mechanical parameters of the compressible interlayer material. The preset displacement range is set to the maximum allowable compression of the compressible plate (…). The preset temperature range is set to the design limit temperature gradient of the frozen wall. The preset pressure range is set to the ultimate compressive strength bearing value of the inner well wall concrete. ).

[0029] Finally, in this embodiment, the displacement, temperature, and pressure features at the same sampling time are encapsulated into an instantaneous feature vector, and the vectors are arranged in reverse order of sampling time within a preset sampling window. The instantaneous feature vectors are stacked vertically to generate a sample of size 1. The feature parameter matrix is ​​constructed. Each column of this matrix carries the evolution trajectory of a single physical attribute within a historical window, while each row maps the instantaneous coupling state of the three types of features at a given historical moment. To ensure the continuity of the matrix's evolution along the time axis, this scheme introduces a sliding recursion mechanism: when new sampled data enters the window, the matrix automatically removes the oldest row vector and simultaneously pushes in the latest instantaneous feature vector. This provides a standardized computational basis for subsequent vector synthesis and capturing of evolutionary mutation features.

[0030] The coupling gradient module is used to extract the evolutionary abrupt features of the displacement and temperature features in the feature parameter matrix, obtain a perturbation vector through vector synthesis, and dynamically gain the perturbation vector according to the instantaneous distribution ratio between the displacement and temperature features to obtain the coupling gradient that reflects the degree of multivariate coupling influence.

[0031] In deep, multi-field, strong interference environments, the evolution rates and destructive contributions of different physical quantities vary significantly. The frozen wall temperature field (thermal source) and the wellbore displacement field (deformation response) constitute the high-frequency dynamic coupling axis in the time domain, while the inter-wall pressure is a macroscopic static load exhibiting highly gradual variation. Based on this physical hierarchical logic, this embodiment uses displacement and temperature characteristics with high-frequency response as excitation sources for abrupt change capture, constructing a coupling gradient reflecting transient disturbances. The gradually varying pressure characteristics are then directionally assigned to the subsequent dynamic threshold module. By constructing a stress saturation term, the bearing saturation state of the wellbore structure is quantified, thereby performing a subsidence correction on the static baseline characteristic threshold. This hierarchical processing strategy, through decoupling extraction and nonlinear multiplication operations, achieves coordinated interactive modulation of transient disturbances and static background loads in the threshold correction stage.

[0032] First, the displacement and temperature feature sequences of the current sampling time and its two preceding consecutive sampling times are extracted from the feature parameter matrix. In the initial stage, the feature parameter matrix needs to be vertically inserted with at least three consecutive valid sampling vectors along the time dimension to ensure that process interruptions due to missing historical data are avoided during the initial stage. Since first-order features represent absolute deformation states, and first-order differences represent the slow rheological or creep rate of the structure; while second-order differences are equivalent to evolutionary acceleration, this embodiment utilizes a backward discrete numerical difference algorithm to retrieve the current sampling time... Previous sampling time and the first two sampling times The characteristic values ​​are calculated by subtracting twice the characteristic value from the previous sampling time and adding the characteristic values ​​from the two previous sampling times. This yields the second-order displacement difference and the second-order temperature difference values ​​for the current sampling time, which are then used as the second-order displacement abrupt change characteristics and the second-order temperature abrupt change characteristics for the current sampling time, respectively. This further filters out the extremely slow linear contraction background interference of the wellbore during the safe period. Since a fixed sampling cycle with equal intervals has been set, the discrete difference operator in this embodiment directly uses numerical difference, which can maintain the physical purity of the abrupt change characteristics in the dimensionless system and avoid the risk of numerical overflow caused by the introduction of the time dimension.

[0033] Subsequently, because the distortion of mechanical displacement (mechanical energy release) and the distortion of temperature gradient (thermal energy shock) are orthogonal in physical properties, and often alternate between positive and negative in real-world conditions (such as expansion and contraction, heating and cooling), simple algebraic linear addition can easily lead to cancellation of positive and negative values, masking the true destructive force. Since the wellbore displacement and temperature gradient have been mapped using preset reference values ​​and preset ranges, The dimensionless scalar is used in this embodiment to sum and take the square root of the squares of the second-order displacement abrupt change characteristic and the second-order temperature abrupt change characteristic to obtain the disturbance vector reflecting the intensity of multi-source disturbance. The specific calculation formula is as follows: In the formula, Sampling time The perturbation vector, It is a second-order discrete difference operator. Sampling time Second-order abrupt shift characteristics, Sampling time The corresponding displacement characteristic quantity, Sampling time Temperature second-order abrupt change characteristics, Sampling time The corresponding temperature characteristic quantity, This is the sampling time sequence number.

[0034] The disturbance vector characterizes the absolute energy level of the sudden, combined thermodynamic impact currently experienced by the wellbore. A large disturbance vector indicates a violent release of cold energy or the application of strong transient destructive kinetic energy by micro-crack displacement, easily triggering brittle cracking. A smaller disturbance vector indicates only slow rheology or steady-state heat transfer, without abrupt destructive energy tearing the matrix. The second-order displacement and temperature abrupt changes map the mechanical acceleration of increased deformation and the thermodynamic acceleration of heat exchange, respectively. Both are positively correlated with the disturbance vector. In this formula, the release of mechanical energy and the thermal stress impact are physically orthogonal evolution dimensions. Simple algebraic addition would lead to the alternation of positive and negative physical quantities in real-world conditions (e.g., positive for contraction, negative for heating), causing the destructive energy to cancel each other out. This formula first transforms the directional abrupt change into a positive scalar through squaring, eliminating sign reversal interference and performing nonlinear amplification on the distorted signal; then, it reduces the dimension to a linear scale reflecting the absolute destructive level through square root reduction, achieving pure stripping and highly sensitive capture of the combined destructive kinetic energy under orthogonal evolution.

[0035] Next, the displacement and temperature features at the current sampling time are extracted. The temperature feature is summed with a preset minimum positive bias constant. The ratio of the displacement feature to this sum is calculated, and the absolute value of this ratio is used as the dynamic gain reflecting the dominant physical evolution weight. The specific calculation formula is as follows: In the formula, Sampling time The value of dynamic gain. Sampling time The corresponding displacement characteristic quantity, Sampling time The corresponding temperature characteristic quantity, The sampling time number. This is the preset minimum positive bias constant.

[0036] The dynamic gain characterizes the weight of the dominant physical evolution of the wellbore at the current moment and its physical vulnerability. A large dynamic gain indicates that the mechanical deformation response far exceeds the thermal field fluctuations, the structure has lost its deformation resistance, and has entered a high-risk stage dominated by mechanical fracture. A small dynamic gain indicates that pure thermal fluctuations dominate, and the structure is still in the elastic resistance period of thermal expansion and contraction. The displacement characteristic quantity maps to the macroscopic geometric deformation consumption and is positively correlated with the dynamic gain; the temperature characteristic quantity maps to the thermal driving intensity and is inversely correlated with the dynamic gain, effectively suppressing the gain value. The preset minimum positive bias constant characterizes the inherent thermal inertia noise of thick-walled concrete, and is summed with the temperature characteristic quantity, filtering out minute thermal fluctuations while avoiding the calculation singularity of zero denominator. In this formula, under deep physical interference, the difference in displacement response caused by the same thermal impact can directly reflect the true physical damage degree of the wellbore concrete. By taking the absolute value of the ratio of the two factors, a very high penalty weight is assigned to the deformation-dominated malignant disturbance, while physical damping is applied to the low-risk disturbance of pure geothermal alternation, thus quantifying the dynamic vulnerability response of the internal structural resistance to external impact energy. The absolute value calculation ensures that whether the wellbore experiences rapid compression or violent unloading and rebound, any high-frequency deformation deviating from the baseline state is equally considered a malignant depletion of structural stability, preventing the loss of penalty weight due to directional sign.

[0037] In this embodiment, the minimum positive bias constant is set to During the resting observation period after the inner well wall is cast and not subjected to external frost heave, the maximum baseline drift amplitude of the temperature sensor under natural environmental alternation is extracted. This amplitude is then compared with the maximum allowable thermal stress span of the well wall component without observable micro-strain. The maximum value of these two values ​​is then defined as the minimum positive bias constant. The minimum positive bias constant not only avoids the calculation singularity of zero denominator but also characterizes the inherent thermal inertia noise of thick-walled structures, filtering out extremely small thermal fluctuations. In real deep working conditions, if the well wall is subjected to pure external mechanical compression, resulting in a very large displacement characteristic, while the temperature remains unchanged, causing the temperature characteristic to approach zero, the ratio of this dynamic gain will exhibit an exponentially abnormal surge due to the influence of the minimum denominator. To address this, an upper limit for engineering cutoff is set (limiting the maximum gain to 100). This involves extracting the critical failure displacement of the retractable interlayer within the monitoring wellbore and calculating its ratio to the extreme value of unit thermal stress disturbance allowed during the frozen wall design phase. This ratio is then rounded to calibrate the upper cutoff. When the calculated dynamic gain exceeds this threshold, the upper cutoff value is forcibly output. This prevents distortion of the overall early warning feedback calculation caused by extreme off-center loading in a single dimension.

[0038] Finally, the dynamic gain is multiplied and weighted by the disturbance vector to obtain the coupling gradient at the current sampling time, which reflects the degree of multivariate coupling influence. The specific calculation formula is as follows: In the formula, Sampling time The coupling gradient, Sampling time The perturbation vector, Sampling time Dynamic gain, It is a second-order discrete difference operator. Sampling time The corresponding displacement characteristic quantity, Sampling time The corresponding temperature characteristic quantity, The sampling time number. This is the preset minimum positive bias constant.

[0039] The coupling gradient characterizes the ultimate destructive potential caused by malignant interference from deep multiphysics fields. When the coupling gradient is large, the high-intensity abrupt shock has completely breached the thermodynamic defenses of the structure and transformed into severe mechanical deformation, bringing the wellbore to the brink of instability and failure. When the coupling gradient is small, it indicates that the structure can safely absorb the impact due to its own strength. The disturbance vector maps the absolute destructive kinetic energy base applied by external multi-source strong interference, while the dynamic gain maps the physical vulnerability of the current wellbore structure and the deformation-dominant weight. Both are positively correlated with the coupling gradient; a surge in either parameter will increase the hazard rating. The simple disturbance vector only represents the intensity of the external impact, and the simple dynamic gain only reflects the attenuation state of internal resistance. By combining the two through a product-weighted operation, a highly sensitive dimensionality reduction analysis of the actual catastrophic risk of the wellbore under combined inducements is achieved.

[0040] The time-delay analysis module is used to convert the coupled gradient into the warning excitation intensity using a nonlinear mapping function, and to determine the adjustment damping based on the convergence of the displacement feature quantity at the current sampling time relative to the preset displacement limit feature quantity. The time-delay sensitivity coefficient is obtained through the interaction operation of the warning excitation intensity and the adjustment damping and the normalized mapping constraint analysis.

[0041] During the disaster-causing evolution of deep frozen wells, the destructive energy released from the initiation to the completion of microfractures in the rock mass exhibits an exponential increase. Based on this physical law, this embodiment uses the natural constant as the base and performs an exponential calculation on the coupling gradient at the current sampling time to obtain the early warning excitation intensity. The specific calculation formula is as follows: In the formula, Sampling time The intensity of early warning incentives, Sampling time The coupling gradient, This is the sampling time sequence number.

[0042] The early warning excitation intensity characterizes the absolute initial driving force that propels the bottom-level computation to execute the early warning defense line at the current moment. When the early warning excitation intensity is high, the destructive energy inside the wellbore structure has exceeded the critical threshold, and the catastrophic process is in an extremely active and high-risk state, requiring absolute priority response. When it is low, it indicates that the multi-field interference has not yet produced enough abrupt kinetic energy to tear the matrix. The ultimate destructive potential caused by the malignant interference of deep multi-physics fields mapped by the coupled gradient is exponentially strongly positively correlated with the early warning excitation intensity. As the value of the coupled gradient increases, the early warning excitation intensity also increases. Since the accumulation and release of destructive energy inside the rock mass microfractures from initial initiation, local expansion to macroscopic through fracture, is not a gradual linear process, but exhibits typical exponential avalanche surge characteristics. It is ensured that the output remains stable when the destructive potential is in a low resting period, and once the energy exceeds the critical point, an explosively amplified excitation signal is forcibly output, thereby giving absolute priority response to high-risk energy mutations.

[0043] Meanwhile, as the wellbore deformation approaches the geometric limit, the compressible interlayer exhibits significant deformation hardening and plastic energy dissipation characteristics, and the structure's physical response to external high-frequency impacts generates resistance inertia. This embodiment extracts the displacement characteristic quantity at the current sampling moment and the preset displacement limit characteristic quantity, calculates their ratio, and performs an absolute value operation to obtain the deformation convergence. Specifically, to objectively define the physical dead line of the inner wellbore structure's buffer space, this embodiment obtains the initial laying thickness of the compressible interlayer at the monitoring layer and extracts the critical compaction strain rate allowed by the material of the interlayer in the engineering mechanics specifications, calculating their product as the physical limit compressibility. This embodiment divides the physical limit compressibility by the preset displacement range to obtain a dimensionless multiplier constant, which is then calibrated as the preset displacement limit characteristic quantity. The preset displacement limit characteristic quantity is set to 1.6 to ensure that the deformation convergence calculation is anchored to the actual depletion degree of the remaining pressure space. Then, the deformation convergence is summed with the natural constant and subjected to a natural logarithmic operation to analytically obtain the adjustable damping. The specific calculation formula is as follows: In the formula, Sampling time The value corresponding to the adjustable damping, Sampling time The corresponding displacement characteristic quantity, For the preset displacement limit characteristic quantity, This is the sampling time sequence number.

[0044] The adjustable damping characterizes the physical sluggishness and strain hardening resistance generated when the wellbore structure approaches the geometric yield limit. A larger adjustable damping indicates strong resistance to transient impacts; a smaller damping indicates sufficient yield space and weaker damping restraint. The displacement characteristic at the current sampling moment maps to the actual macroscopic geometric consumption, while the preset displacement limit characteristic maps to the irreversible physical dead line. The deformation convergence is obtained based on the absolute value of the ratio between the two, which is positively correlated with the adjustable damping. As the displacement characteristic approaches the preset displacement limit characteristic, the deformation convergence increases, further improving the physical damping. The natural constant and the deformation convergence are summed to forcibly raise the operator base. As the structure shrinks towards its limit, the mechanical response changes from elastic sensitivity to plastic hardening. The outer nested natural logarithm operator, with its decreasing rate of increase, serves as an important step in constraining aggressive feedback. At the same time, the summation term introduces a natural constant to ensure that the argument of the natural logarithm operation is always greater than or equal to the constant itself, thereby forcing the output of the damped adjustment to be always greater than or equal to one, avoiding the abnormal amplification singularity caused by a denominator of zero or less than one.

[0045] Finally, to objectively quantify the urgency of the early warning trigger and to prevent extreme geological faults from causing the early warning excitation intensity to approach infinity, this embodiment calculates the ratio of the early warning excitation intensity to the adjustment damping, and analyzes to obtain the preliminary sensitivity. Subsequently, a preset safety upper limit coefficient is introduced, and a minimum value comparison operator is used to perform truncation constraint processing on the preliminary sensitivity, that is, extracting the smaller value between the preliminary sensitivity and the safety upper limit coefficient, which is then used as the final time-delay sensitivity coefficient. The specific calculation formula is as follows: In the formula, Sampling time The time delay sensitivity coefficient, The preset safety upper limit coefficient, Sampling time The coupling gradient, Sampling time The value corresponding to the adjustable damping, Sampling time The corresponding displacement characteristic quantity, For the preset displacement limit characteristic quantity, This is the sampling time sequence number.

[0046] The time-delay sensitivity coefficient characterizes the physical driving weight and urgency of the current wellbore damage state in influencing the dynamic adjustment of the early warning defense line. A large time-delay sensitivity coefficient indicates that the wellbore has experienced a high-intensity burst of damage and has not yet entered the plastic energy dissipation period, leading to a rapid deterioration of the disaster process. A smaller coefficient indicates that the structure is relatively quiescent or that macroscopic plastic deformation has absorbed some of the impact kinetic energy, thus maintaining the stability of the early warning defense line. The early warning excitation intensity reflects the absolute boost effect of the transient destructive impact and is positively correlated with the time-delay sensitivity coefficient. Adjusting the damping quantifies the sluggish response effect of the wellbore entering the ultimate convergence period and is strongly negatively correlated with the time-delay sensitivity coefficient. By using the minimum value comparison operator to extract the smaller value between the initial sensitivity and the safety upper limit coefficient, forced truncation is performed to further prevent memory overflow and response deadlock caused by the divergence of computational data due to extreme geological fractures.

[0047] In this embodiment, since the final issuance of deep disaster early warnings must be limited by the physical execution limits of the underlying hardware, this embodiment extracts the absolute sum of the inherent communication flow delay and relay mechanical action delay of the monitoring and control hardware, and uses this as the minimum physical response period of the underlying hardware. It then retrieves the theoretically shortest disintegration time window from the critical yielding to macroscopic collapse of the rock mass from the geological survey report; divides this theoretically shortest disintegration time window directly by the minimum physical response period of the underlying hardware, and rounds the resulting quotient down to obtain the safety upper limit coefficient. This embodiment sets the preset safety upper limit coefficient to 50, further preventing the instantaneous breach of the early warning defense and logical breaks caused by infinitely amplified sensitivity, ensuring the continuity of dynamic down-probing operations and the steady-state triggering of the alarm relays.

[0048] The dynamic threshold module is used to construct a force saturation term based on the pressure characteristics at the current sampling time, and generate a feedback adjustment factor using the saturation constraint function. The correlation between the feedback adjustment factor and the time delay sensitivity coefficient is used to lower and correct the preset static benchmark characteristic threshold, thereby generating a dynamic alarm characteristic threshold.

[0049] In this embodiment, to isolate the load directionality, the absolute physical load representing the pure compressive strength of the surrounding rock is obtained. An absolute value operation is performed on the pressure characteristic at the current sampling moment to construct a stress saturation term. Subsequently, to demonstrate that in the early stages of compression, a small increase in pressure leads to a sharp increase in structural hazard, while in the later stages of compression, the structure is extremely compacted, and the function output is smooth and approximates the asymptote, a hyperbolic tangent function operation is performed on the stress saturation term to obtain a feedback adjustment factor, preventing the divergence of subsequent reduction operations due to an excessively large single static load. The specific calculation formula is as follows: In the formula, Sampling time Feedback adjustment factor, Sampling time The pressure characteristic quantity, This is the sampling time sequence number.

[0050] The feedback adjustment factor characterizes the absolute physical compaction state and resistance overdraft caused by the current macroscopic off-center load on the wellbore. A large value indicates severe compression of the structure, indicating that it has entered a state of plastic yielding or complete compaction; a small value indicates slight compression of the surrounding rock, indicating that it is still in the elastic stage and has sufficient residual resistance. The pressure characteristic quantity maps the external macroscopic static slowly varying load. An absolute value operation is performed on it to construct a stress saturation term, aiming to completely remove the directionality of the original stress. This stress saturation term is positively correlated with the dependent variable, and the feedback adjustment factor increases nonlinearly as the stress saturation term intensifies. Due to the strong nonlinear characteristics of the energy consumption evolution of the wellbore under continuous high pressure, a slight increase in stress in the initial stage of pressure leads to a sharp increase in the degree of danger; in the later stage of stress, the wellbore is extremely compacted, the rate of deterioration slows down, and the function output smoothly approaches the upper limit asymptote.

[0051] Next, since the time-delay sensitivity coefficient represents the structural vulnerability to high-frequency abrupt shocks, while the feedback adjustment factor represents the degree of resistance overextension under long-term static heavy loads, their product achieves cross-frequency domain physical interference between high-frequency dynamic damage and low-frequency static losses. Therefore, this embodiment calculates the product of the time-delay sensitivity coefficient, the feedback adjustment factor, and a preset sensitivity coefficient compensation factor, and calibrates this product as a dynamic reduction term. The specific calculation formula is as follows: In the formula, Sampling time The dynamic reduction item, The preset sensitivity coefficient compensation factor is, where, The range of values ​​is , The theoretical maximum value of the time delay sensitivity coefficient is preset. Sampling time The time delay sensitivity coefficient, Sampling time The pressure characteristic quantity, This is the sampling time sequence number.

[0052] The dynamic reduction term represents the percentage of overall physical resistance of the wellbore structure that has been overdrawn under combined severe working conditions. A large dynamic reduction term indicates that the structure has suffered severe interference from multiple sources, resulting in serious internal damage and near exhaustion of resistance; a small dynamic reduction term indicates that the structure has been slightly disturbed and has sufficient residual resistance. The time-delay sensitivity coefficient maps the high-frequency physical vulnerability of the structure after a transient impact, while the feedback adjustment factor maps the low-frequency resistance overdraft state under long-term static heavy pressure. Both are positively correlated with the dynamic reduction term. The preset sensitivity coefficient compensation factor acts as a safety valve, forcibly constraining the combined amplification factor of the two. Real deep disasters are cross-frequency domain physical interferences of high-frequency kinetic energy abrupt changes and low-frequency static load compression. Multiplying the time-delay sensitivity coefficient and the feedback adjustment factor can replicate the malignant destruction mechanism of the synergistic superposition of dynamic and static disaster-causing factors; at the same time, the preset sensitivity coefficient compensation factor is introduced for multiplication to further smooth out the series differences of different physical dimensions, ensuring that the generated dynamic reduction term is limited to a reasonable range of physical losses.

[0053] In this embodiment, the value range of the sensitivity coefficient compensation factor is limited to being greater than a constant zero and less than the reciprocal of the preset theoretical maximum value of the time-delay sensitivity coefficient. The preset theoretical maximum value of the time-delay sensitivity coefficient is equivalent to a preset safety upper limit coefficient, which is set to 50. This equivalent setting constructs a globally consistent boundary from parameter parsing to threshold sinking, further preventing out-of-bounds divergence caused by conflicts between parameter definitions.

[0054] Given that the extreme value of the feedback adjustment factor approaches one, and the upper limit of the time-delay sensitivity coefficient is locked by this theoretical maximum value, if both reach their peak simultaneously under extreme conditions, the unconstrained multiplication will cause the subsequent deduction to exceed the constant one, thus causing the safety red line to fall below zero. This reciprocal constraint ensures that the generated dynamic reduction term is a constant decimal between zero and one, preventing computational collapse due to the negative warning line. To reserve sufficient buffer redundancy for extreme conditions, this embodiment sets the preset sensitivity coefficient compensation factor to 0.01. The reciprocal of the preset theoretical maximum value of the time-delay sensitivity coefficient (i.e., 0.02) is extracted and calibrated by multiplying it with the preset engineering standard tolerance coefficient (set to 0.5). In this embodiment, based on the historical statistical data of the strength decay of similar underground components under long-term service, the upper and lower boundaries of the natural fluctuation of its bearing capacity are determined, and the midpoint of the fluctuation range is directly extracted to calibrate the engineering standard tolerance coefficient. In this embodiment, the engineering standard tolerance coefficient is set to 0.5, which aims to effectively filter out numerical false alarms caused by local micro-stress concentration through an equally divided compromise scale, and ensure the steady state of boundary determination.

[0055] Finally, a dynamic adjustment gain is constructed using the difference between a constant and the dynamic reduction term. This dynamic adjustment gain is then multiplied by a preset static baseline characteristic threshold to perform a sinking correction, ultimately generating a dynamic alarm characteristic threshold. The specific calculation formula is as follows: In the formula, Sampling time Dynamic alarm feature threshold, The preset static baseline feature threshold, Sampling time The dynamic reduction item, The preset sensitivity coefficient compensation factor is, where, The range of values ​​is , The theoretical maximum value of the time delay sensitivity coefficient is preset. Sampling time The time delay sensitivity coefficient, Sampling time Feedback adjustment factor, Sampling time The pressure characteristic quantity, This is the sampling time sequence number.

[0056] The dynamic alarm characteristic threshold represents the true limit of safety of the well wall under its current damaged state. A larger value indicates a healthy structure with minimal disturbance and a high upper limit of allowable resistance; a smaller value indicates an extremely vulnerable structure. The static benchmark characteristic threshold establishes the absolute physical bearing capacity under undamaged conditions and is positively correlated with the dependent variable. In this embodiment, the ultimate strength of the well wall material is multiplied by the actual well wall thickness to obtain the maximum theoretical bearing pressure peak when the structure is on the verge of fracture. Subsequently, this maximum theoretical bearing pressure peak is divided by the safety factor mandated by engineering specifications to calculate the allowable net resistance value. To ensure that this physical index operates in the same dimension as the characteristic matrix, the allowable net resistance value is proportionally mapped to a preset pressure range to obtain a dimensionless static benchmark characteristic threshold. In this embodiment, the static benchmark characteristic threshold is set to 0.6. This is used to re-concretize the mathematical scaling ratio as the physical stress red line on the actual monitoring platform, so that the final output can directly connect with on-site emergency response decisions. The time-delay sensitivity coefficient and the pressure characteristic quantity respectively map the destructive degree of transient high-frequency impact and the resistance overdraft degree of macroscopic low-frequency off-center load. Both the absolute value of the time-delay sensitivity coefficient and the pressure characteristic quantity processed by hyperbolic tangent are inversely correlated with the dependent variable; the preset sensitivity coefficient compensation factor constrains the unbounded divergence of the multiplicative amplification. Using a constant to represent the initial perfect resistance, after subtracting the dynamic reduction term constructed jointly by the sensitivity coefficient compensation factor, the time-delay sensitivity coefficient, and the processed pressure characteristic quantity, and then multiplying it by the static benchmark characteristic threshold, a downward correction is achieved where the more vulnerable the component, the more stringent the warning line becomes.

[0057] The hazard determination module is used to map the deviation between the displacement feature quantity at the current sampling time and the dynamic alarm feature threshold to a safety redundancy space defined by the preset displacement limit feature quantity for evaluation, generate a hazard index and determine the hazard level.

[0058] In this embodiment, the construction environment of the frozen well is affected by multiple factors, including formation pressure disturbances and the migration of the frozen wall temperature field. The absolute displacement of the well wall is affected by non-structural damaging deformations such as thermal expansion and contraction and initial natural rheology. If an alarm is triggered directly based on a single absolute displacement threshold, it is highly likely to cause frequent false alarms due to normal physical background fluctuations. Therefore, this embodiment uses the preset displacement limit characteristic quantity as the upper limit of judgment and the dynamic alarm characteristic threshold as the lower limit of judgment. A safety redundancy space is constructed based on the algebraic difference between the preset displacement limit characteristic quantity and the dynamic alarm characteristic threshold, further quantifying the elastic buffer capacity remaining before the substantial collapse of the well wall structure under the current specific temperature and pressure conditions. Among them, the preset displacement limit characteristic quantity is always greater than the dynamic alarm characteristic threshold. The preset displacement limit characteristic quantity anchors the physical boundary of the inner compressible structure becoming unstable and fractured, and is the absolute red line for judging the complete failure of the structure; while the dynamic alarm characteristic threshold is only the warning starting line for triggering the defense response. If this relationship is not satisfied, the warning signal will lag behind the occurrence of physical damage. Secondly, the positive algebraic difference ensures that the wellbore retains a certain elastic buffer depth to dissipate formation disturbance energy after the warning line is breached.

[0059] For the displacement characteristic quantity acquired at the current sampling moment, the algebraic difference between it and the dynamic alarm characteristic threshold is calculated, and this difference is calibrated as the over-limit deviation, ensuring that each unit of characteristic quantity deviation directly corresponds to the actual increase in structural hidden danger exceeding the safety bottom line. Subsequently, the ratio of this over-limit deviation to the safety redundancy space is calculated to solve the danger index at the current moment. The microscopic deformation data is smoothly transformed into a macroscopic risk state description. In order to achieve precise engineering intervention for the entire process of well wall deterioration, this embodiment compares the obtained danger index at the current sampling moment with a preset level interval. The preset level interval includes a progressively advancing normal safety interval, a primary warning interval, a high-risk interval, and a limit failure interval, and each interval is defined by a constant zero, a preset warning threshold, and a constant one in sequence. The preset warning threshold is a value greater than constant zero and less than constant one.

[0060] In this embodiment, the constant zero represents the zero deviation between the actual displacement characteristic quantity and the dynamic alarm characteristic threshold, i.e., the trigger zero point of the early warning response; the constant one corresponds to the saturation state where the over-limit deviation completely fills the safety redundancy space. Based on the displacement change characteristics of the well wall material under increasing load, when the observed deformation rate of the material changes from the initial stable contraction to a sharp increase, marking the material's formal transition from the elastic bearing stage to the plastic yielding stage, the proportion of the compressive displacement at this physical inflection point to the total displacement of ultimate failure is extracted and calibrated as the preset warning threshold. In this embodiment, the preset warning threshold is set to 0.6, which is based on the nonlinear characteristics of the damage evolution of deep frozen well wall structures. When the proportion of well wall displacement to redundancy space reaches 0.6, the structure often transitions from the initial stable micro-damage state to the accelerated deterioration plastic failure stage. On the one hand, it effectively accommodates non-damaging deformation caused by environmental fluctuations within the range below 0.6; on the other hand, when the index exceeds 0.6, it reserves space for emergency response. The displacement redundancy ensures that the reinforcement measures can be effective before the structure reaches the ultimate failure boundary (i.e., the danger index is 1.0).

[0061] When the danger index is less than or equal to a constant zero, it falls into the normal safe range, and the danger level of the well wall to be monitored is determined to be the safe level. When the danger index is greater than a constant zero and less than or equal to the preset warning threshold, it falls into the primary warning range, and the danger level of the well wall to be monitored is determined to be the primary warning level. When the danger index is greater than the preset warning threshold and less than a constant, it falls into the high-risk range, and the danger level of the well wall to be monitored is determined to be high-risk. When the danger index is greater than or equal to a constant, it falls into the limit damage range, and the danger level of the well wall to be monitored is determined to be the limit damage level.

[0062] Table 1 is an example table of comprehensive data on the multi-source nonlinear disturbance characteristics and risk evolution across levels of the shrinkable well wall of the inner layer of the frozen well under test at 25 sampling times.

[0063] Table 1: Comprehensive Data Table of Multi-Source Nonlinear Disturbance Characteristics and Risk Trans-Level Evolution Table 1 shows the data demonstrating the accurate quantification and dynamic capture of the steady-state rheological changes to nonlinear instability and degradation of the inner collapsible wellwall under the complex alternating disturbance environment of deep frozen wells using the proposed method. (Sampling number) Although the displacement characteristic quantity shows a slow linear increase, due to the lack of sudden multi-field malignant interference, the coupling gradient value is extremely small, resulting in the analytically obtained time-delay sensitivity coefficient remaining at a low level (below 2.0). At this time, the dynamic alarm characteristic threshold hardly decreased (stabilizing above 0.59), and the corresponding danger index is always less than a constant zero. This reflects the normal mechanical response of the wellbore under normal frost heave rheology or natural weak ground pressure oscillations during the safe pressure relief period, and the structure has not undergone substantial micro-crack penetration evolution; sampling sequence number During sampling, the displacement characteristic exhibits nonlinear accelerated distortion, leading to a surge in the coupling gradient. Simultaneously, the time-delay sensitivity coefficient also experiences an explosive increase (triggering a cutoff of the upper limit of 50 at sampling sequence number 16). Due to this transient strong disturbance, the dynamic alarm characteristic threshold actively and significantly decreases downward (from 0.586 to 0.360), causing the danger index to smoothly cross zero and approach the high-level warning red line. Sampling sequence number The mid-displacement characteristic quantity is extremely close to the preset displacement limit, and the danger index exceeds 0.6 and eventually stabilizes at around 0.86, indicating that the compressibility of the inner layer has completely exhausted the compressive space, and the structure has suffered irreversible macroscopic plastic failure. This table proves that traditional monitoring methods based on fixed absolute displacement thresholds are prone to causing early warning delays, and that nonlinear coupling of multi-source data can filter out false interference from the initial elastic background.

[0064] Please see Figure 3 The present invention also provides a method for monitoring and early warning of the retractable wellbore of the inner layer of a frozen well. This method is used to implement the aforementioned monitoring system for the retractable wellbore of the inner layer of a frozen well, comprising: Step 1: Collect the well wall displacement, temperature gradient and inter-wall pressure of the well wall to be monitored within the preset sampling window, and map them into the corresponding displacement characteristic quantity, temperature characteristic quantity and pressure characteristic quantity after deviation elimination and range scaling. Combine the characteristic quantities of each sampling time in time sequence to construct the characteristic parameter matrix. Step 2: Extract the evolution and abrupt change features of the displacement and temperature features in the feature parameter matrix, obtain the perturbation vector through vector synthesis, and dynamically gain the perturbation vector according to the instantaneous distribution ratio between the displacement and temperature features to obtain the coupling gradient that reflects the degree of multivariate coupling influence. Step 3: Use a nonlinear mapping function to convert the coupled gradient into a warning excitation intensity, and determine the adjustment damping based on the convergence of the displacement feature quantity at the current sampling time relative to the preset displacement limit feature quantity. The time delay sensitivity coefficient is obtained through the interaction operation of the warning excitation intensity and the adjustment damping and the normalized mapping constraint analysis. Step 4: Construct a force saturation term based on the pressure characteristics at the current sampling time, and generate a feedback adjustment factor using the saturation constraint function. Use the correlation between the feedback adjustment factor and the time delay sensitivity coefficient to lower and correct the preset static benchmark characteristic threshold, and generate a dynamic alarm characteristic threshold. Step 5: Map the deviation between the displacement feature quantity at the current sampling time and the dynamic alarm feature threshold to the safety redundancy space defined by the preset displacement limit feature quantity for evaluation, generate a hazard index and determine the hazard level.

[0065] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0066] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0067] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0068] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A wellbore monitoring system for the shrinkable inner layer of a frozen well, characterized in that, Specifically, it includes: The feature construction module is used to collect well wall displacement, temperature gradient and interwall pressure of the well wall to be monitored within a preset sampling window, and map them into corresponding displacement feature quantities, temperature feature quantities and pressure feature quantities after deviation elimination and range scaling. The feature quantities at each sampling time are combined in time sequence to construct a feature parameter matrix. The coupling gradient module is used to extract the evolutionary abrupt features of the displacement and temperature features in the feature parameter matrix, obtain a perturbation vector through vector synthesis, and dynamically gain the perturbation vector according to the instantaneous distribution ratio between the displacement and temperature features to obtain the coupling gradient that reflects the degree of multivariate coupling influence. The time delay analysis module is used to convert the coupled gradient into the warning excitation intensity using a nonlinear mapping function, and to determine the adjustment damping based on the convergence of the displacement feature quantity at the current sampling time relative to the preset displacement limit feature quantity. The time delay sensitivity coefficient is obtained through the interaction operation of the warning excitation intensity and the adjustment damping and the normalized mapping constraint analysis. The dynamic threshold module is used to construct a force saturation term based on the pressure characteristics at the current sampling time, and generate a feedback adjustment factor using the saturation constraint function. The correlation between the feedback adjustment factor and the time delay sensitivity coefficient is used to correct the preset static benchmark feature threshold and generate a dynamic alarm feature threshold. The hazard determination module is used to map the deviation between the displacement feature quantity at the current sampling time and the dynamic alarm feature threshold to a safety redundancy space defined by the preset displacement limit feature quantity for evaluation, generate a hazard index and determine the hazard level.

2. The wellbore monitoring system for the retractable inner layer of a frozen well according to claim 1, characterized in that: Within the preset sampling window, acquire the raw data of wellbore displacement, temperature gradient and interwall pressure at each sampling time. The sampling window is defined as a continuous window that extends backward from the current sampling time, with the current sampling time as the endpoint. A sliding window at each preceding sampling time, where It is a preset positive integer; The collected raw data undergoes deviation elimination and range scaling mapping. Specifically, for any sampling time, the corresponding preset reference value is subtracted from each raw data to obtain the corresponding displacement deviation value, temperature deviation value, and pressure deviation value. Then, using the preset displacement range, preset temperature range, and preset pressure range, the displacement deviation value, temperature deviation value, and pressure deviation value are linearly normalized and mapped to obtain the displacement characteristic quantity, temperature characteristic quantity, and pressure characteristic quantity at the corresponding sampling time.

3. The wellbore monitoring system for the shrinkable inner layer of a frozen well according to claim 2, characterized in that: The specific steps for constructing the feature parameter matrix include: The displacement, temperature, and pressure features at the same sampling time are encapsulated into an instantaneous feature vector; The instantaneous feature vectors within the preset sampling window are stacked vertically in chronological order of sampling time to generate a feature parameter matrix with temporal dimension.

4. The wellbore monitoring system for the retractable inner layer of a frozen well according to claim 1, characterized in that: The specific steps for obtaining the coupling gradient are as follows: Extract the displacement feature sequence and temperature feature sequence of the current sampling time and the two consecutive sampling times before it from the feature parameter matrix, and use the discrete numerical difference algorithm to calculate the second-order displacement difference value and the second-order temperature difference value of the current sampling time, so as to serve as the second-order displacement change feature and the second-order temperature change feature of the current sampling time. The displacement second-order abrupt change feature and the temperature second-order abrupt change feature are squared respectively, and then the summation and square root are taken to obtain the disturbance vector reflecting the intensity of multi-source disturbance; Extract the displacement and temperature features at the current sampling time, sum the temperature features with a preset minimum positive bias constant, calculate the ratio of the displacement features to the sum, and perform an absolute value operation on the ratio as a dynamic gain reflecting the dominant physical evolution weight. The dynamic gain is multiplied and weighted with the disturbance vector to obtain the coupling gradient at the current sampling time, which reflects the degree of multivariate coupling influence. The specific calculation formula is as follows: In the formula, Sampling time The coupling gradient, It is a second-order discrete difference operator. Sampling time The corresponding displacement characteristic quantity, Sampling time The corresponding temperature characteristic quantity, The sampling time number. This is the preset minimum positive bias constant.

5. A wellbore monitoring system for the retractable inner layer of a frozen well according to claim 1, characterized in that: The specific steps for obtaining the time delay sensitivity coefficient are as follows: Using the natural constant as the base, the coupling gradient at the current sampling time is exponentially calculated to obtain the warning excitation intensity; Calculate the ratio of the displacement characteristic quantity at the current sampling time to the preset displacement limit characteristic quantity, and perform an absolute value operation on the ratio to obtain the deformation convergence. Summing the deformation convergence with the natural constant and then performing a natural logarithmic operation, the adjustable damping is obtained. The initial sensitivity is obtained by calculating the ratio of the warning excitation intensity to the adjustment damping. A preset safety upper limit coefficient is then used to truncate the initial sensitivity, and the smaller of the two values ​​is taken as the time-delay sensitivity coefficient. The specific calculation formula is as follows: In the formula, Sampling time The time delay sensitivity coefficient, The preset safety upper limit coefficient, Sampling time The coupling gradient, Sampling time The corresponding displacement characteristic quantity, For the preset displacement limit characteristic quantity, This is the sampling time sequence number.

6. A wellbore monitoring system for the retractable inner layer of a frozen well according to claim 1, characterized in that: The specific steps for generating the dynamic alarm feature threshold are as follows: The absolute value operation is performed on the pressure feature at the current sampling time to construct the force saturation term, and the hyperbolic tangent function operation is performed on the force saturation term to obtain the feedback adjustment factor; Calculate the product of the time delay sensitivity coefficient, the feedback adjustment factor, and the preset sensitivity coefficient compensation factor, and label the product as a dynamic reduction term. The value range of the sensitivity coefficient compensation factor is greater than a constant zero and less than the reciprocal of the preset theoretical maximum value of the time delay sensitivity coefficient. A dynamic adjustment gain is constructed using the difference between a constant and the dynamic reduction term. This dynamic adjustment gain is then multiplied by a preset static baseline characteristic threshold to generate a dynamic alarm characteristic threshold. The specific calculation formula is as follows: In the formula, Sampling time Dynamic alarm feature threshold, The preset static baseline feature threshold, The preset sensitivity coefficient compensation factor is, where, The range of values ​​is , The theoretical maximum value of the time delay sensitivity coefficient is preset. Sampling time The time delay sensitivity coefficient, Sampling time The pressure characteristic quantity, This is the sampling time sequence number.

7. A wellbore monitoring system for the retractable inner layer of a frozen well according to claim 1, characterized in that: Using the preset displacement limit feature quantity as the upper limit of judgment and the dynamic alarm feature threshold as the lower limit of judgment, a safety redundancy space is constructed based on the algebraic difference between the preset displacement limit feature quantity and the dynamic alarm feature threshold, wherein the preset displacement limit feature quantity is always greater than the dynamic alarm feature threshold. The specific steps for obtaining the risk index are as follows: Calculate the algebraic difference between the displacement feature at the current sampling time and the corresponding dynamic alarm feature threshold, and calibrate the algebraic difference as the over-limit deviation; The ratio of the out-of-limit deviation to the safety redundancy space is calculated as the danger index corresponding to the current sampling time.

8. A wellbore monitoring system for the retractable inner layer of a frozen well according to claim 7, characterized in that: The specific steps for determining the risk level are as follows: The danger index obtained at the current sampling time is compared with the preset level range. The preset level range includes a normal safety range, a primary warning range, a high danger range and a limit destruction range, which are progressively defined by a constant zero, a preset warning threshold and a constant one. The preset warning threshold is a value that is greater than constant zero and less than constant one. When the danger index is less than or equal to a constant zero, it falls into the normal safe range, and the danger level of the well wall to be monitored is determined to be the safe level. When the danger index is greater than a constant zero and less than or equal to the preset warning threshold, it falls into the primary warning range, and the danger level of the well wall to be monitored is determined to be the primary warning level. When the danger index is greater than the preset warning threshold and less than a constant, it falls into the high-risk range, and the danger level of the well wall to be monitored is determined to be high-risk. When the danger index is greater than or equal to a constant, it falls into the limit damage range, and the danger level of the well wall to be monitored is determined to be the limit damage level.

9. A method for monitoring and early warning of the shrinkable wellbore wall in frozen wells, characterized in that: The aforementioned method for monitoring and early warning of the retractable wellbore of the inner layer of a frozen well is used to implement the retractable wellbore monitoring system of the inner layer of a frozen well as described in any one of claims 1-8, comprising: Within a preset sampling window, the well wall displacement, temperature gradient, and interwall pressure of the well wall to be monitored are collected and mapped to the corresponding displacement characteristic quantity, temperature characteristic quantity, and pressure characteristic quantity after deviation elimination and range scaling. The characteristic quantities at each sampling time are combined in time sequence to construct a characteristic parameter matrix. The evolutionary abrupt changes of displacement and temperature features in the feature parameter matrix are extracted, and a perturbation vector is obtained through vector synthesis. The perturbation vector is dynamically amplified according to the instantaneous distribution ratio between the displacement and temperature features to obtain a coupling gradient that reflects the degree of multivariate coupling influence. The coupled gradient is converted into an early warning excitation intensity using a nonlinear mapping function, and the adjustment damping is determined based on the convergence of the displacement feature quantity at the current sampling time relative to the preset displacement limit feature quantity. The time delay sensitivity coefficient is obtained through the interaction operation of the early warning excitation intensity and the adjustment damping and the normalized mapping constraint analysis. Based on the pressure characteristics at the current sampling time, a force saturation term is constructed, and a feedback adjustment factor is generated using the saturation constraint function. The correlation between the feedback adjustment factor and the time delay sensitivity coefficient is used to lower and correct the preset static benchmark characteristic threshold, thereby generating a dynamic alarm characteristic threshold. The deviation between the displacement feature quantity at the current sampling time and the dynamic alarm feature threshold is mapped to a safety redundancy space defined by the preset displacement limit feature quantity for evaluation, generating a hazard index and determining the hazard level.

Images