A casing segment-level risk early warning method based on CCL signal sliding window aggregation

By identifying the coupling response to determine the casing segment boundary, and constructing sliding window samples within each casing segment for risk scoring and sorting aggregation, the problem of CCL logging signals being difficult to directly use for casing segment-level risk early warning is solved. This enables the screening of high-risk casing segments and the location of abnormal windows, improving the pertinence and efficiency of on-site verification.

CN123109630APending Publication Date: 2026-11-17SOUTHWEST PETROLEUM UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611590977.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing CCL logging signals are difficult to use directly for casing section-level risk early warning. Local anomaly responses are easily diluted in the overall average evaluation, and on-site verification lacks clear priorities and anomaly window locations.

Method used

By identifying the coupling response to determine the casing segment boundary, a sliding window sample is constructed within each casing segment, and risk scoring and ranking aggregation are performed to generate casing segment-level risk warning results.

Benefits of technology

It enables the sorting and screening of high-risk casing sections and the location of abnormal windows, improving the pertinence and efficiency of on-site verification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123109630A_ABST
    Figure CN123109630A_ABST
Patent Text Reader

Abstract

This invention relates to a casing segment-level risk early warning method based on CCL signal sliding window aggregation, belonging to the technical field of oil and gas well casing integrity detection and risk early warning. The method acquires CCL logging signals, depth data, and logging velocity data of the target well; identifies the coupling center location and establishes the original casing segment boundary; performs velocity normalization, coupling response area masking, and local anomaly enhancement processing on the CCL signal; constructs sliding window samples within each original casing segment and obtains window-level risk scores; sorts the window-level risk scores within the same casing segment, selects the top-ranked scores, and aggregates them to obtain segment-level risk scores; outputs the high-risk casing segment ranking, anomaly window locations, and verification markers based on the segment-level risk scores. This invention can highlight local anomaly responses within the casing segment, reduce risk dilution caused by averaging across the entire segment, and improve the targeting of on-site verification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of oil and gas well casing integrity detection and risk warning technology, specifically involving a casing segment-level risk warning method based on CCL signal sliding window aggregation. Background Technology

[0002] Oil and gas well casing is a crucial downhole string for maintaining wellbore stability, isolating formations, and ensuring safe oil and gas production. As oil and gas wells age, casing can be affected by factors such as formation creep, changes in injection and production pressures, corrosion and wear, cementing quality variations, and downhole disturbances, leading to anomalies such as reduced diameter, compression, bending, breakage, or deformation. These anomalies affect the string's throughput, wellbore integrity assessment, and the selection of subsequent well workover measures. Therefore, it is essential to promptly screen, prioritize, and review casing sections with potential anomalies.

[0003] Current casing anomaly detection typically relies on logging data such as boom caliper, downhole television, electromagnetic testing, and eddy current testing. Boom caliper and downhole television can visually reflect changes in wellbore shape or inner diameter, but in some well conditions, they present challenges such as complex construction, high operational risks, high retesting costs, or difficulty in continuous and efficient screening. Electromagnetic testing and eddy current testing can reflect changes in electromagnetic response caused by damage, corrosion, or deformation of the casing, but their interpretation usually depends on instrument response characteristics, charts, model samples, or human experience.

[0004] Existing technologies include methods for detecting and evaluating casing damage, which typically involve methods such as selecting optimal logging instruments, creating model tubing response characteristic charts, and partitioning or interpreting response curves to detect and evaluate casing damage. While these methods can provide a basis for interpreting casing damage, their focus is usually on extracting logging response features, creating charts, or evaluating damage types, and they do not establish a risk ranking and early warning process for casing sections based on CCL continuous logging signals.

[0005] Existing technologies also include methods for identifying downhole casing deformation types, which typically rely on electromagnetic response detection data, deformation type samples, or feature set comparison results to identify the casing deformation type. While these methods are helpful in determining casing deformation types, they usually require specific detection data or deformation type samples as a basis, making it difficult to directly utilize conventional CCL logging signals to output casing section-level risk ranking results for on-site verification.

[0006] CCL magnetic positioning signals are commonly used to identify casing couplings and assist in depth positioning. Related coupling detection or magnetic positioning technologies mainly focus on coupling signal acquisition, coupling positioning sensitivity, or improvements in the structure of magnetic positioning instruments. They are usually not aimed at casing section-level risk early warning, nor do they make full use of the natural segmentation relationship formed by adjacent coupling positions to convert continuous CCL logging signals into casing section-level risk assessment objects.

[0007] Furthermore, existing technologies also include schemes for casing anomaly identification using CCL signals, such as identification methods based on local change rates, anomaly thresholds, or anomaly levels in the CCL signals. These methods can reflect local abnormal changes in the CCL signals, but they typically focus on identifying anomaly points or windows, without further sorting and aggregating multiple window-level risk results within the same original casing segment. They also struggle to directly output the ranking of high-risk casing segments, the location of abnormal windows, and verification markers.

[0008] In casing risk screening scenarios, casing deformation or local structural anomalies often only produce abnormal responses within a short depth range within the casing segment. If the entire signal is directly averaged, local anomalies are easily diluted by normal intervals, resulting in a lack of prominence in the ranking of high-risk casing segments and a lack of clear priority and anomaly window locations during on-site verification. Therefore, it is necessary to propose a casing risk early warning method based on casing segment division using couplings, constructing sliding window samples within the casing segment, and using window-level risk scores for ranking and aggregation of the top-ranked segments. Summary of the Invention

[0009] This invention provides a casing segment-level risk early warning method based on CCL signal sliding window aggregation to solve the problems that existing CCL logging signals are mainly used for coupling positioning and are difficult to directly form casing segment-level risk early warning results, as well as the problem that local casing anomaly responses are easily diluted in the overall average evaluation. This invention aims to achieve the purpose of sorting and screening high-risk casing segments, locating anomaly windows, and providing on-site verification guidance.

[0010] This invention is achieved through the following technical solution:

[0011] A casing segment-level risk early warning method based on CCL signal sliding window aggregation includes:

[0012] Step S1: Acquire the CCL logging signal, depth data, and logging velocity data of the target well;

[0013] Step S2: Identify the center position of the coupling based on the coupling response in the CCL logging signal, and establish the original casing segment boundary based on the center positions of adjacent couplings to obtain multiple original casing segments;

[0014] Step S3: Preprocess the CCL logging signal to obtain an abnormally enhanced CCL signal;

[0015] Step S4: Construct multiple sliding window samples within each original casing segment;

[0016] Step S5: Perform risk scoring on the sliding window samples to obtain window-level risk scores;

[0017] Step S6: Sort the window-level risk scores within the same original casing segment, select the top-ranked window-level risk scores for aggregation, and obtain the segment-level risk scores.

[0018] Step S7: Generate casing segment-level risk ranking results based on the segment-level risk scores of each original casing segment, and output casing segment-level risk warning results.

[0019] To address the problem that existing CCL signals are difficult to use directly for casing section-level risk screening, this invention first uses the coupling response in the CCL signal to determine the coupling center position, and uses the depth interval between adjacent coupling center positions as the original casing section, so that the continuous logging signal is converted into a section-level analysis object corresponding to the actual casing structure.

[0020] To address the issues that logging velocity fluctuations, strong coupling response, and local noise may affect anomaly identification, this invention performs velocity normalization, coupling response region masking, and local anomaly enhancement processing on CCL logging signals, making the local anomaly response at non-coupling locations within the casing section more prominent.

[0021] To address the issue that casing anomalies often manifest as local responses within a segment rather than overall responses across the entire segment, this invention constructs multiple sliding window samples within each original casing segment and performs window-level risk scoring on the sliding window samples, thereby obtaining risk characterizations for multiple local depth intervals within the same original casing segment.

[0022] To address the issue that average evaluation of the entire casing segment can easily dilute local high-risk windows, this invention sorts the window-level risk scores within the same original casing segment and selects the top-ranked window-level risk scores for segment-level aggregation, so that the segment-level risk scores can reflect the contribution of local abnormal responses within the casing segment.

[0023] To address the lack of priority and precise depth range in on-site verification, this invention generates casing segment-level risk ranking results based on segment-level risk scores, and outputs the casing segment number, casing segment start and end depths, segment-level risk score, risk ranking, abnormal window location, and verification mark, providing a basis for on-site verification and subsequent measure formulation.

[0024] Furthermore, the method for identifying the coupling center position in step S2 by obtaining the coupling response from the CCL logging signal includes: determining candidate coupling positions based on local extrema, positive and negative peak combinations, peak amplitude, peak width, and coupling spacing constraints in the CCL logging signal; verifying the candidate coupling positions; and determining the retained position after verification as the coupling center position, thus forming a coupling center depth sequence. This improves the stability of coupling center position identification.

[0025] Furthermore, step S2, which establishes the boundary of the original casing segment based on the center positions of adjacent couplings, includes: defining the depth interval between the center positions of two adjacent couplings as an original casing segment, and determining the starting depth, ending depth, center depth, and length of each original casing segment. This ensures that the risk assessment object corresponds to the actual casing structure.

[0026] Furthermore, the method for preprocessing the CCL logging signal in step S3 to obtain the anomalously enhanced CCL signal includes: performing velocity normalization processing based on the logging velocity data, masking the collar strong response region, and performing local anomaly enhancement processing. This reduces the interference of logging velocity changes and collar strong response on anomaly identification.

[0027] Furthermore, the method for constructing multiple sliding window samples within each original casing segment in step S4 includes: generating multiple sliding windows within the depth range corresponding to each original casing segment according to a preset window length and a preset sliding step size, and extracting signal data or feature data within the sliding windows as sliding window samples. This preserves local signal variations at different depth positions within the same original casing segment.

[0028] Furthermore, the method for risk scoring of the sliding window samples in step S5 includes: inputting the sliding window samples into a risk scoring model to obtain the corresponding window-level risk score; the risk scoring model is a preset scoring rule, a classification model, a regression model, an anomaly detection model, or a combination thereof. This converts continuous CCL logging signals into sortable and aggregateable window-level risk scores.

[0029] Furthermore, the method in step S6 for sorting window-level risk scores within the same original casing segment and selecting a preset number of top-ranked window-level risk scores for aggregation includes: Let the first... The original casing segment contains A sliding window of samples, with the preset number of samples prioritized. and The smaller of the two values ​​is determined as the number of windows participating in the aggregation. And according to the sorted first The window-level risk score is calculated. The segment-level risk score of each original casing segment is calculated. This reduces the dilution effect of a large number of low-risk windows on local high-risk anomalies.

[0030] Furthermore, the method for generating a casing segment-level risk ranking result based on the segment-level risk score of each original casing segment and outputting the casing segment-level risk warning result in step S7 includes: sorting multiple original casing segments from high to low according to the segment-level risk score, identifying high-risk casing segments, outputting the casing segment-level risk warning result, and generating a verification mark. This improves the interpretability and usability of the casing segment-level risk warning result.

[0031] By utilizing the coupling response in the CCL logging signal to establish the original casing segment boundary, this invention enables the risk assessment object to correspond with the actual casing structure, avoiding the casing segment boundary offset problem that may occur when dividing only according to fixed depth intervals.

[0032] By constructing multiple sliding window samples within each original casing segment, this invention can identify local anomalies at different depths within the same casing segment, enabling the risk assessment results to retain both the overall expression at the casing segment level and reflect the distribution of local anomalies within the segment.

[0033] By performing window-level risk scoring on sliding window samples, this invention can convert continuous CCL logging signals into sortable and aggregateable window-level risk scores, providing a quantitative basis for subsequent casing section-level risk early warning.

[0034] By aggregating a predetermined number of window-level risk scores that rank highest within the same original casing segment, this invention can enhance the contribution of high-risk windows within the segment to the segment-level risk results and reduce the dilution effect of a large number of low-risk windows on local high-risk anomalies.

[0035] By outputting the casing section-level risk ranking results, abnormal window locations, and verification markers, this invention can provide a basis for on-site verification, screening of key well sections, and subsequent handling, thereby improving the interpretability and usability of casing section-level risk warning results.

[0036] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0037] 1. The present invention provides a casing segment-level risk early warning method based on CCL signal sliding window aggregation. It utilizes the coupling response in the CCL signal to establish the original casing segment boundary, thereby converting continuous logging signals into casing segment-level analysis objects with engineering significance.

[0038] 2. The present invention provides a casing section-level risk early warning method based on CCL signal sliding window aggregation. By normalizing velocity, masking the coupling response area, and enhancing local anomalies, the interference of logging velocity fluctuations and strong coupling responses on casing anomaly identification is reduced.

[0039] 3. The present invention provides a casing segment-level risk warning method based on CCL signal sliding window aggregation. By using sliding window risk scoring and prioritizing segment aggregation, it highlights the contribution of local abnormal windows within a segment to the segment-level risk score and reduces the risk dilution caused by averaging across the entire segment.

[0040] 4. The present invention provides a casing segment-level risk early warning method based on CCL signal sliding window aggregation, which outputs high-risk casing segment ranking, abnormal window position and review mark, thereby improving the pertinence of on-site review and the screening efficiency of high-risk casing segments. Attached Figure Description

[0041] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0042] Figure 1 This is a schematic diagram of the casing segment-level risk warning method based on CCL signal sliding window aggregation according to a specific embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of coupling identification and sleeve segment division in a specific embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the sliding window sample construction in a specific embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of risk aggregation at the top level in a specific embodiment of the present invention;

[0046] Figure 5 This is a schematic diagram of the output of early warning results for high-risk casing sections in a specific embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are for illustrative purposes only and are not intended to limit the invention.

[0048] Example 1:

[0049] like Figure 1As shown, this embodiment provides a casing segment-level risk early warning method based on CCL signal sliding window aggregation. This method takes the target well's CCL logging signal, depth data, and logging velocity data as input, and outputs the casing segment number, casing segment start and end depths, segment-level risk score, risk ranking, abnormal window location, and verification marker. The core idea of ​​this method is: first, to establish the original casing segment boundary using the coupling response in the CCL signal; then, to construct multiple sliding window samples within each original casing segment; and finally, to aggregate the top-ranked segments based on the window-level risk score ranking, thereby converting continuous logging signals into casing segment-level risk early warning results that can be verified on-site.

[0050] In the data acquisition step, the CCL logging signal of the target well, the depth data corresponding to the CCL logging signal, and the logging velocity data are acquired. The depth data is arranged according to depth points. organize, Representing depth point CCL signal value at that location Representing depth point The logging velocity corresponding to the location or its adjacent sampling interval. The CCL logging signal can be the original sampling signal, or a signal after removing obvious bad spots, depth alignment, or resampling.

[0051] In one specific implementation, the CCL logging signal and depth data are first checked to ensure a one-to-one correspondence. If the number of sampling points for the CCL logging signal and depth data is inconsistent, they are aligned based on timestamps, sampling sequence numbers, or depth interpolation relationships. For duplicate depth points, the sampled values ​​with higher signal quality or the average value can be retained. For missing depth points, nearest neighbor interpolation, linear interpolation, or resampling at fixed depth intervals can be used to fill in the gaps. After the above processing, a data sequence is formed, arranged in ascending or descending order of depth.

[0052] Perform a basic quality check on the CCL logging signals. This quality check includes determining whether there are long-term constant values, abnormal jump values, depth reversal segments, velocity missing segments, or signal points significantly exceeding the instrument's range. Points that do not meet the quality requirements can be removed, marked, or replaced by interpolation. This step does not necessarily require changing the original signal value; it can also simply generate quality markers for use in subsequent risk scoring or result verification.

[0053] like Figure 2 As shown, the coupling response is identified on the CCL logging signal after basic quality inspection. The standard coupling response typically exhibits a combination of negative peaks on both sides and a positive peak in the middle. In this embodiment, candidate positive and negative peaks can be determined first based on local extremum detection, and then, based on the depth interval, amplitude relationship, and peak width relationship between the positive peak and the two negative peaks, it can be determined whether a candidate coupling response is constituted.

[0054] In one specific implementation, for each depth point, it is determined whether it is a local maximum or a local minimum. If a local maximum has local minimums on both its left and right sides, and the depth distance between the local maximum and the two local minimums is within a preset peak spacing range, then the depth where the local maximum is located is taken as a candidate coupling center position. Further, if the difference between the amplitude of the local maximum and the amplitudes of the two local minimums is greater than a preset amplitude threshold, then the candidate coupling center position is retained.

[0055] The spacing between the center positions of the candidate couplings is verified. Let the depth spacing between the center positions of two adjacent candidate couplings be... ,Will Compare with the preset coupling spacing range. When When the distance between two candidate coupling positions is less than the preset lower limit, it is considered that at least one of the two candidate coupling positions may be noise or an abnormal response. In this case, candidate positions with larger response amplitudes, more complete positive and negative peak combinations, or more consistent with the spacing between the front and rear couplings can be retained. When the depth exceeds the preset upper limit of the coupling spacing, the depth range is marked as a suspected missing coupling range or an abnormal coupling range to be reviewed.

[0056] The candidate coupling center positions are then subjected to response pattern verification. This verification includes determining whether a standard coupling response pattern with negative peaks on both sides and a positive peak in the middle exists near the candidate coupling center position; determining whether there is a preset relative amplitude relationship between the positive and negative peaks; and determining whether the peak width of the coupling response is within a preset range. After amplitude verification, spacing verification, and response pattern verification, the retained candidate coupling center positions constitute a coupling center depth sequence.

[0057] The original casing segment boundary is established based on the coupling center depth sequence. Let the coupling center depth sequence be... ,in The number of identified couplings. For the centers of two adjacent couplings... and , depth range Determined as the number One original casing section The initial depth of the original casing segment is Ending depth is The center depth is the midpoint between the two, and the length of the casing section is... for and The depth difference between them.

[0058] The original casing segments are marked with their status. The length of each original casing segment is compared with the preset standard casing length. If the length deviation is within the preset allowable range, it is marked as a normal casing segment; if the length is significantly shorter than the preset standard casing length, it is marked as a short segment; if the length is significantly longer than the preset standard casing length, it is marked as a suspected missing coupling segment; if the coupling response pattern is unstable or the reliability of the coupling center position is low, it is marked as an abnormal coupling segment awaiting verification. The above status markings are not directly equivalent to the casing risk level, but can serve as auxiliary information for risk interpretation and on-site verification.

[0059] The CCL logging signal is preprocessed to obtain anomaly-enhanced CCL signals. Preprocessing includes velocity normalization, collar response region masking, and local anomaly enhancement. These three processes can be performed sequentially, or one or more can be selected based on data quality. The preferred execution order is: velocity normalization first, then collar response region masking, and finally local anomaly enhancement.

[0060] In the velocity normalization process, the logging velocity corresponding to each depth point is first determined based on the logging velocity data. And calculate the reference speed. The reference speed This can be the mean, median, or average velocity of the logging rate for the entire well section, or the average velocity for the target stable section. For depth points... CCL signal value at the location According to and The relationship can be used to correct the signal amplitude, or it can be based on... Resample the sampling points at equal depth intervals. This step reduces amplitude fluctuations and sampling density differences caused by variations in instrument speed.

[0061] In the processing of the coupling response region mask, a preset depth range is extended upwards and downwards from the center of each coupling to form a strong coupling response region. The preset depth range can be determined based on the coupling response peak width, sampling interval, or instrument resolution. Signals within the strong coupling response region can be processed using methods such as zeroing, interpolation replacement, weight reduction, or simply marking without replacement. If interpolation replacement is used, linear or smooth interpolation is performed using signals from the non-coupling regions on either side of the strong coupling response region.

[0062] In local anomaly enhancement processing, the target depth point Establish a local statistical window centered on the sample. The length of the local statistical window can be determined based on the sampling interval, the expected anomaly length, or the sleeve segment length. Calculate the median of the CCL signal within the local statistical window. and median absolute deviation And based on the target depth point signal value and The difference between them yields an abnormal enhancement value. As one implementation method, the abnormal enhancement value can be represented as: ,in To prevent positive numbers with a denominator of zero.

[0063] After preprocessing, an abnormally enhanced CCL signal is obtained. This abnormally enhanced CCL signal can retain the depth index of the original signal, or it can be combined with the velocity-normalized signal, coupling mask markers, and casing segment numbers to form the input data for subsequent sliding window samples. For cases where strong coupling response regions are marked and masked but not replaced, mask features can be added to the window samples to enable the risk scoring model to distinguish between coupling response regions and non-coupling response regions.

[0064] like Figure 3 As shown, multiple sliding window samples are constructed within each original casing segment. For the first... One original casing section Let its initial depth be Ending depth is According to the preset window length and preset sliding step size from Start by generating sliding windows sequentially. Each sliding window corresponds to a window start depth, a window end depth, and a window center depth.

[0065] When the length of a certain original casing segment is greater than or equal to the preset window length, multiple sliding windows with overlapping or non-overlapping relationships are generated according to the preset sliding step size. When the length of a certain original casing segment is less than the preset window length, the entire original casing segment can be used as a sliding window sample, or an adaptive window length can be used to generate samples. The above short segment processing method can be determined according to the on-site verification requirements for short segments or abnormal coupling segments.

[0066] Each sliding window sample includes at least one of the following: the original CCL signal sequence within the window, the velocity-normalized CCL signal sequence, the anomalously enhanced CCL signal sequence, the depth sequence, the coupling mask marker, statistical features, and the normalized positional features of the window relative to its original casing segment. The normalized positional features may include the positional ratio of the window center depth to the starting depth of the original casing segment, and the positional ratio of the window start and end depths to the length of the original casing segment.

[0067] Statistical features may include at least one of the following: mean, variance, standard deviation, maximum value, minimum value, peak-to-peak value, energy, mean absolute value, number of local extrema, maximum value of abnormal enhancement, mean value of abnormal enhancement, and number of sampling points exceeding a preset abnormality threshold. By combining sequence features and statistical features, the risk scoring model can simultaneously utilize the temporal morphology and local anomaly intensity information of the CCL signal.

[0068] Multiple sliding window samples belonging to the same original casing segment are combined into a multi-window sample package corresponding to that original casing segment. Let... Indicates the first The first original casing segment before sorting A number of sliding window samples, among which This refers to the sample sequence number of the sliding window. The multi-window sample package retains the window number, start and end depths, center depth, original casing segment number, and corresponding input features for each sliding window sample. This multi-window sample package is used for subsequent window-level risk scoring and segment-level sorting of top-ranked windows.

[0069] The sliding window samples are input into the risk scoring model to obtain a window-level risk score. The risk scoring model is used to output a window-level risk score based on the temporal, statistical, or locational characteristics of the sliding window samples. The window-level risk score can be a normalized value between 0 and 1, or a risk score value after monotonic transformation. The higher the window-level risk score, the higher the probability or degree of casing structure abnormality risk within the corresponding sliding window.

[0070] Risk scoring models can employ classification models, regression models, anomaly detection models, or combinations thereof. Classification models output the probability of risk categories, regression models output continuous risk levels, and anomaly detection models output anomaly scores that deviate from the normal CCL response pattern. In practice, supervised learning models, semi-supervised models, or unsupervised anomaly detection models can be selected based on whether historical labeled data is available.

[0071] During the model training phase, historical CCL logging data with casing anomaly interpretation results are acquired. These casing anomaly interpretation results include at least one of the following: caliper interpretation section, manual casing variation interpretation section, or field verification results. The historical CCL logging data undergoes the same coupling identification, original casing segment division, preprocessing, and sliding window construction steps as the target well to obtain historical sliding window samples.

[0072] Risk labels are determined based on the overlap of depth intervals between historical sliding window samples and casing anomaly interpretation results. If the depth interval of the historical sliding window is... The interpretation range corresponding to the casing anomaly interpretation result is: Then the overlap length of the two It can be represented as: .when Greater than the preset overlap length threshold, or When the ratio of the length of the historical sliding window to the length of the historical sliding window is greater than a preset overlap ratio threshold, the historical sliding window is identified as a risk sample.

[0073] Historical sliding windows that do not overlap with casing anomaly interpretation results can be identified as non-risk or low-risk samples. Historical sliding windows with only minor overlap or insufficient credibility of interpretation results can be marked as samples to be reviewed, or assigned lower weights during training. In this way, manual interpretation results and caliper interpretation results can be converted into training samples usable by the risk scoring model.

[0074] like Figure 4 As shown, window-level risk scores within the same original casing segment are sorted and aggregated at the top segment level. Let the first segment be... One original casing section Contains The sliding window samples are sorted by window-level risk scores from highest to lowest as follows: .in, Indicates the sorted order of the first... Each window-level risk score, This represents the window-level risk score that is ranked first.

[0075] Determine the number of windows to participate in the aggregation. If the preset number of windows to be sorted first is... Then the first Number of windows for aggregation of each original casing segment It can be represented as: .in, This indicates the number of items that appear first in the default sorting. Indicates the first The number of windows in which each original casing segment actually participates in aggregation. When Greater than or equal to When selecting, choose the one that appears earlier in the sorted list. Each window-level risk score participates in aggregation; when Less than At that time, all window-level risk scores within the original casing segment are selected for aggregation.

[0076] In one specific implementation, the number of windows participating in the aggregation Based on the total number of sliding windows within the original casing segment Adaptively determined. Specifically, the base window size. ,For example ;when Less than or equal to At that time, take That is, all windows within the original casing segment participate in the aggregation; when Greater than At that time, it can be taken , or take ,in To preset the participation ratio, for example ,and No more than and preset upper limit value ,For example Through the above adaptive approach, all window information can be fully utilized for short casing segments, while the number of windows participating in aggregation can be controlled for long casing segments to avoid diluting the segment-level risk score with a large number of low-risk windows.

[0077] In another specific implementation, an inflection point method is used for adaptive determination. Value: After sorting the window-level risk scores within the same original casing segment from highest to lowest, calculate the difference between two adjacent window-level risk scores to obtain a difference sequence; determine the position number corresponding to the maximum value in the difference sequence as... This method involves selecting the window preceding the point of most drastic score decline for aggregation. It automatically determines the number of aggregation windows based on the actual distribution of risk scores within the window, eliminating the need for a pre-defined fixed number. value.

[0078] According to the sorted order The window-level risk score is calculated. The segment-level risk score of each original casing segment. As one implementation method, the segment-level risk score... It can be the front The average of the risk scores at each window level, i.e. As another implementation method, segment-level risk scores. A weighted method can be used to determine this, that is... ,in Indicates the first Segment-level risk score of each original casing segment Indicates the first The number of windows that actually participate in the aggregation for each original casing segment. Indicates the first The first original casing segment after sorting Each window-level risk score, For sorting number The weights corresponding to the window-level risk scores.

[0079] When determining segment-level risk scores using a weighted method, the weight corresponding to a window-level risk score that ranks higher should not be less than the weight corresponding to a window-level risk score that ranks lower. ,in, Indicates the sorted order of the first... The weights corresponding to the window-level risk scores Indicates the sorted order of the first... The weights corresponding to the window-level risk scores.

[0080] As another implementation method, the front can also be The maximum value among the window-level risk scores is used as the segment-level risk score, or the average, weighted average, and maximum value are combined to obtain the segment-level risk score. This top-ranking aggregation method prevents a small number of high-risk windows within the original casing segment from being diluted by a large number of low-risk windows.

[0081] Multiple original casing segments are ranked by risk based on their segment-level risk scores. A predetermined number of original casing segments can be selected as high-risk segments based on their segment-level risk scores, or segments with segment-level risk scores exceeding a predetermined risk threshold can be designated as high-risk segments. The predetermined number and risk threshold can be determined based on on-site verification capabilities, well section length, historical risk distribution, or risk screening sensitivity.

[0082] Furthermore, based on the quantile or threshold range of the segment-level risk score, the original casing segments are divided into high-risk warning segments, medium-risk monitoring segments, low-risk review segments, and routine segments. For example, original casing segments with segment-level risk scores within a pre-defined proportion range are classified as high-risk warning segments, those with segment-level risk scores in the middle range are classified as medium-risk monitoring segments, and those with segment-level risk scores below the low-risk threshold are classified as routine segments.

[0083] like Figure 5 As shown, the output displays the early warning results for high-risk casing sections. These results include at least one of the following: casing section number, starting and ending depths, section-level risk score, risk ranking, abnormal window location, casing section status marker, risk score of the highest-ranked window, and review marker. The output results can be sorted from highest to lowest section-level risk score to allow on-site personnel to prioritize viewing high-risk casing sections.

[0084] The location of the anomaly window is determined based on the windows that participate in the aggregation of the top-ranked windows. The anomaly window location can include the start and end depths of the highest-risk window, the center depth of the highest-risk window, the start and end depths of the top-ranked windows, and the relative position of the top-ranked windows within their respective original casing segments. By specifying the anomaly window location, the casing segment-level risk results can be further refined to a specific depth range, facilitating on-site verification.

[0085] High-risk casing segments are matched with caliper interpretation segments or manual casing variation interpretation segments for interval overlap. If a caliper interpretation segment or manual casing variation interpretation segment overlaps with a certain original casing segment in depth interval, the original casing segment is marked as having a spatial correspondence with the corresponding interpretation segment. When an interpretation segment spans multiple original casing segments, all original casing segments that overlap with the interpretation segment are retained, and the main response casing segment is determined based on the overlap length, overlap ratio, or segment-level risk score.

[0086] Verification markers are generated based on the interval overlap matching results. Verification markers can include states such as matched, pending verification, verified, or unmatched. For original casing segments with high segment-level risk scores but not yet matched with the caliper interpretation segment or manual casing variation interpretation segment, they can be marked as pending verification; for original casing segments that have already matched with the interpretation segment, they can be marked as matched; for original casing segments that have completed field verification, they can be marked as verified.

[0087] The output of this embodiment can be used for on-site verification plan development. On-site personnel can determine the casing sections to be verified based on risk ranking, determine the verification depth range based on the location of abnormal windows, determine whether there are short segments or suspected missing coupling segments based on the casing section status markers, and determine whether verification needs to be combined with wellbore interpretation data, manual casing variation interpretation data, or other logging data based on the verification markers.

[0088] This embodiment establishes the original casing segment boundary through the coupling response, converting continuous CCL logging signals into segment-level analysis objects organized according to the casing structure; it reduces logging velocity fluctuations and interference from strong coupling responses through velocity normalization, coupling response area masking, and local anomaly enhancement processing; it obtains window-level risk scores through sliding window samples and risk scoring models; and it highlights the contribution of local anomaly windows to the segment-level risk scores by aggregating the top-ranked windows, thereby improving the screening efficiency of high-risk casing segments and the targeting of on-site verification.

Claims

1. A casing segment-level risk early warning method based on CCL signal sliding window aggregation, characterized in that, include: Step S1: Acquire the CCL logging signal, depth data, and logging velocity data of the target well; Step S2: Identify the center position of the coupling based on the coupling response in the CCL logging signal, and establish the original casing segment boundary based on the center positions of adjacent couplings to obtain multiple original casing segments; Step S3: Preprocess the CCL logging signal to obtain an abnormally enhanced CCL signal; Step S4: Construct multiple sliding window samples within each original casing segment; Step S5: Perform risk scoring on the sliding window samples to obtain window-level risk scores; Step S6: Sort the window-level risk scores within the same original casing segment, and aggregate the top-ranked window-level risk scores to obtain the segment-level risk score; Step S7: Generate casing segment-level risk ranking results based on the segment-level risk scores of each original casing segment, and output casing segment-level risk warning results.

2. The casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for identifying the center position of the coupling based on the coupling response in the CCL logging signal in step S2 includes: Candidate coupling locations are determined based on local extrema, positive and negative peak combinations, peak amplitude, peak width, and coupling spacing constraints in the CCL logging signal. The candidate coupling positions are verified by amplitude, spacing and response shape, and candidate coupling positions that do not meet the coupling response constraints are eliminated. The candidate coupling positions retained after verification are determined as the coupling center positions, forming a coupling center depth sequence.

3. The casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for establishing the original casing segment boundary based on the center position of adjacent couplings in step S2 includes: The depth interval between the center positions of two adjacent couplings is defined as an original sleeve segment; Calculate the starting depth, ending depth, center depth, and length of each original casing segment; The original casing segment is marked with a status based on the deviation between the casing segment length and the preset standard casing length; The status markers include at least one of the following: normal casing segment, short segment, suspected missing coupling segment, and abnormal coupling segment awaiting verification.

4. The casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for preprocessing the CCL logging signal in step S3 includes: The amplitude, sampling interval, or depth sequence of the CCL logging signal are normalized and corrected based on the logging velocity data to obtain the velocity-normalized CCL signal. The strong response region of the coupling is determined with the center position of the coupling as the center, and the CCL signal in the strong response region of the coupling is zeroed, interpolated and replaced, weighted down or marked and masked. A local statistical window is established centered on the target depth point, and the CCL signal at the target depth point is subjected to local anomaly enhancement processing based on the statistical characteristics of the CCL signal within the local statistical window.

5. A casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 4, characterized in that, The method for performing local anomaly enhancement processing on the CCL signal of the target depth point in step S3 includes: Calculate the difference between the CCL signal value at the target depth point and the median of the CCL signal values ​​within the local statistical window; The abnormal enhancement value corresponding to the target depth point is obtained by the ratio between the difference and the absolute deviation of the median within the local statistical window.

6. The casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for constructing multiple sliding window samples within each original cannula segment in step S4 includes: Multiple sliding windows are generated within the depth range corresponding to each original casing segment, according to the preset window length and preset sliding step size. Extract the signal data or feature data within each sliding window as a sliding window sample; Multiple sliding window samples belonging to the same original casing segment are combined into a multi-window sample package corresponding to that original casing segment. The sliding window sample includes at least one of the following: the original CCL signal sequence within the window, the abnormally enhanced CCL signal sequence, the depth sequence, statistical features, and the normalized position features of the window relative to its original casing segment.

7. The casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for risk scoring of the sliding window samples in step S5 includes: Input the sliding window samples into the risk scoring model to obtain the corresponding window-level risk score; The risk scoring model is a pre-set scoring rule, a machine learning-based classification model, a regression model, an anomaly detection model, or a combination thereof; The risk scoring model is trained using historical CCL logging data with casing anomaly interpretation results, which include at least one of the following: caliper interpretation section, manual casing variation interpretation section, or field verification results.

8. The casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for sorting window-level risk scores within the same original casing segment and aggregating the top-ranked windows in step S6 includes: Let the first The original casing segment contains A sliding window sample, the Each sliding window sample corresponds to Each window-level risk score; The window-level risk scores are sorted from high to low to obtain a sorted sequence of window-level risk scores. Quantity ranked first in the preset sorting and The smaller of the two values ​​is determined as the number of windows participating in the aggregation. ; According to the sorted order The window-level risk score is calculated as follows: The segment-level risk score of each original casing segment.

9. A casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 8, characterized in that, The segment-level risk scores mentioned in step S6 are based on the sorted top... The average, weighted average, or maximum value of each window-level risk score is determined; where, when using the weighted average to determine the segment-level risk score, the weight corresponding to the window-level risk score that ranks higher is not less than the weight corresponding to the window-level risk score that ranks lower.

10. A casing segment-level risk early warning method based on CCL signal sliding window aggregation according to claim 1, characterized in that, The method for generating casing segment-level risk ranking results and outputting casing segment-level risk warning results based on the segment-level risk scores of each original casing segment in step S7 includes: The original casing segments were sorted from highest to lowest according to their segment-level risk scores; Select the pre-defined number of original casing segments that rank highest as high-risk casing segments, or identify the original casing segments whose segment-level risk scores are higher than the pre-defined risk threshold as high-risk casing segments. Output casing segment-level risk warning results, which include at least one of the following: casing segment number, casing segment start and end depths, segment-level risk score, risk ranking, abnormal window location, and review mark; The verification markers are generated based on the depth interval overlap matching results between high-risk casing sections and wellbore instrument interpretation sections, manual casing variation interpretation sections, or on-site verification results.