Multi-source data fusion long protector extraction leakage risk identification method and device and medium

Abstract

The application provides a multi-source data fusion long casing pull-out leakage risk identification method, device and medium, relates to the technical field of data analysis, and obtains active ultrasonic echo energy and passive acoustic emission event energy of each sampling point of the long casing in the pull-out process; abnormal candidate points are marked by screening from the sampling points; a sliding window is constructed to traverse all the sampling points, the energy fluctuation trend and acoustic emission cumulative energy integral of all the abnormal candidate points in the sliding window are calculated based on the depth distribution law of the annular gap and the mud skin, and the leakage risk probability is comprehensively judged in combination with the density of the abnormal candidate points. Through the difference in acoustic characteristics between the annular gap and the weak layer of the mud skin, the formation position and severity of the annular gap and the weak layer of the mud skin are sensed in real time through the combination of the active ultrasonic emission and the passive acoustic emission monitoring, and the leakage risk is found in time.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a method, device, and medium for identifying leakage risks during the removal of long casings using multi-source data fusion. Background Technology

[0002] In the construction of cast-in-place pile retaining structures for shafts, tunnels, underground caverns, and deep foundation pits, the long casing follow-through drilling and grouting technology is widely used for drilling and support in complex geological conditions (such as soft plastic silty clay layers, backfilled construction waste layers, etc.). It includes three key stages: drilling → concrete pouring → casing removal. During the casing removal stage, hidden leakage channels at the borehole-soil interface are easily formed. This is because when the casing is pulled upwards, relative movement occurs between the outer wall of the casing and the surrounding soil, while the pile concrete is still in a plastic or semi-solid state and cannot instantly and completely fill the space left after casing removal. If the casing removal speed is too fast or the concrete lacks fluidity, a ring-shaped gap will form between the outer wall of the casing and the soil. Simultaneously, the mud remaining on the outer wall of the casing during the drilling stage is dragged to the borehole-soil interface during casing removal, forming a weak mud cake layer. The annular gap and the mud cake weak layer are formed at the pore-soil interface. After the concrete solidifies and forms the pore, conventional testing methods cannot directly reach them. If not dealt with in time, the annular gap and the mud cake weak layer are very likely to become preferential lateral seepage channels for groundwater during the operation period, which can lead to serious consequences such as leakage, water accumulation behind the tunnel lining, and attenuation of the bearing capacity of the pile foundation.

[0003] Current technologies for monitoring the casing removal stage only record operational parameters such as removal speed and force, and cannot directly detect whether annular gaps are formed or mud cake is produced. Therefore, there is an urgent need for a method that can perceive the state of the borehole-soil interface in real time and automatically identify leakage risks during casing removal. Summary of the Invention

[0004] The purpose of this invention is to provide a method, device, and medium for identifying leakage risks during the removal of long casings using multi-source data fusion. By pre-embedding an array of piezoelectric sensor networks on the outer wall of the long casing, active ultrasonic echo signals and passive acoustic emission signals are simultaneously collected during the removal of the long casing. The probability of leakage risk is comprehensively judged by fusion of multi-source data, energy fluctuation trends, and cumulative energy integration from acoustic emission, combined with the density of abnormal candidate points within the sliding window.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following solution: On the one hand, this application provides a method for identifying leakage risks during the removal of long casings using multi-source data fusion, specifically including the following steps: S1. Upon receiving the casing removal start signal, obtain the starting depth position of the long casing removal and the abnormal depth range determined based on historical geological survey data and historical construction data; S2. During the removal process, the real-time depth position of the casing is obtained. Based on the relative positional relationship between the real-time depth position and the abnormal depth range, the depth step size and detection parameters of the sensor are dynamically configured during data sampling. Multi-source sensing data of each sampling point is collected. The multi-source sensing data includes active ultrasonic echo energy and passive acoustic emission event energy. S3. Based on the relative positional relationship between the real-time depth position of the sampling point and the abnormal depth interval, as well as the corresponding multi-source sensing data, each sampling point is screened and marked to obtain abnormal candidate point markings. The marking types include annular gap anomalies and mud skin anomalies. S4. Construct a sliding window whose size changes with position to traverse all sampling points. Based on the depth distribution law of the annular gap and mud skin, perform piecewise linear regression calculation on the real-time depth position and ultrasonic echo energy of all abnormal candidate points in the sliding window to obtain the energy fluctuation trend. S5. Normalize the passive acoustic emission event energy of all abnormal candidate points within the sliding window to obtain the cumulative acoustic emission energy integral. Based on the energy fluctuation trend, the cumulative acoustic emission energy integral, and the density of abnormal candidate points with different marker types within the sliding window, comprehensively judge the leakage risk probability.

[0006] In some specific technical solutions, the detection parameters include the acquisition frequency, active ultrasonic emission power, and passive acoustic emission gain. The specific process of dynamically configuring the depth step size and detection parameters during data sampling is as follows: When the real-time depth position is within the abnormal depth range, the depth step size is set to the first step size, and the detection parameter is set to the first detection parameter. When the real-time depth position is outside the abnormal depth range, the depth step size is set to the second step size, and the detection parameter is set to the second detection parameter; wherein, the first step size is smaller than the second step size, and the first detection parameter is larger than the second detection parameter.

[0007] In some specific technical solutions, multi-source sensing data also includes travel time. The specific process of filtering and labeling each sampling point is as follows: Calculate the neighborhood depth interval corresponding to the sampling point based on the speed of the long casing removal, analyze the passive acoustic emission event energy distribution of all sampling points in the neighborhood depth interval, and extract the local acoustic emission features of each sampling point in the neighborhood depth space. The ratios of the active ultrasonic echo energy and travel time at the sampling points to the preset reference acoustic parameters are calculated to obtain the echo energy ratio and travel time reduction rate. For each sampling point, the local acoustic emission characteristics, echo energy ratio, and travel time shortening rate are comprehensively analyzed to determine if there are any anomalies. If anomalies are found, the sampling point is designated as anomaly candidate point, and each anomaly candidate point is marked with the corresponding anomaly type.

[0008] In some specific technical solutions, the local acoustic emission characteristics include the presence of sudden spikes and continuous low-amplitude background. When the echo energy ratio collected by the sampling point in the sampling point area is greater than the first preset energy threshold and the travel time shortening rate is greater than the first preset travel time threshold; Continue to determine whether there are sudden spikes and continuous low-amplitude backgrounds. When only sudden spikes exist at the sampling point, mark the sampling point as an abnormal candidate point for annular gap anomaly. When a sampling point simultaneously exhibits both a sudden spike and a continuous low-amplitude background, the sampling point is marked as a candidate point for a composite anomaly of annular gap anomaly and mud skin anomaly. If the echo energy ratio collected by the sampling point is less than the first preset energy threshold and the travel time shortening rate is greater than the second preset travel time threshold, and if the sampling point only has a continuous low-amplitude background, the sampling point will be marked as an abnormal candidate point for mud skin anomaly.

[0009] In some specific technical solutions, the process of extracting the local acoustic emission features of each sampling point in the neighborhood depth space is as follows: The passive acoustic emission event energy signal segment within the neighborhood depth range is obtained, and after removing low-frequency construction noise by inputting it into a bandpass filter, a threshold detection is performed to obtain the local acoustic emission signal. The local acoustic emission signal is subjected to sudden spike detection. When a signal segment with an amplitude exceeding the upper limit threshold and lasting for a preset time length is detected in the local acoustic emission signal, the local acoustic emission feature is determined to have a sudden spike. Continuous low-amplitude background detection is performed on the local acoustic emission signal. The root mean square amplitude and standard deviation of the local acoustic emission signal are calculated. When the root mean square amplitude and standard deviation are both within the preset range, the local acoustic emission feature is judged to have continuous low-amplitude background. When there are sudden spikes in the local acoustic emission signal, the sudden spikes need to be removed from the local acoustic emission signal before continuous low-amplitude background detection is performed.

[0010] In some specific technical solutions, the specific process of obtaining the energy fluctuation trend is as follows: The ultrasonic echo energies of all abnormal candidate points within the sliding window are arranged into an energy sequence according to the acquisition order. The number of local extreme points, the standard deviation of energy distribution, and the arithmetic mean of echo energy are calculated for the energy sequence. When the number of local extreme points exceeds the maximum threshold and the standard deviation of energy distribution is greater than a preset multiple of the arithmetic mean of echo energy, the energy fluctuation trend is judged to be fluctuating. When the number of local extreme points is less than the minimum threshold and the standard deviation of energy distribution is less than a preset multiple of the arithmetic mean of echo energy, the energy fluctuation trend is judged to be stable.

[0011] In some specific technical solutions, the process of comprehensively assessing the probability of leakage risk is as follows: Different weights are assigned to anomaly candidate points of different label types within the sliding window, and the anomaly candidate point density ρ within the sliding window is calculated based on the weight of each label type. Different confidence levels are assigned to the energy fluctuation trend, the cumulative energy integral of acoustic emission, and the density ρ, respectively, to obtain the instantaneous probability within the current sliding window. The instantaneous probability of the previous sliding window is obtained, and the instantaneous probability of adjacent sliding windows is calculated by using an exponentially weighted moving average to obtain the leakage risk probability.

[0012] In some specific technical solutions, the process of determining the probability of leakage risk also includes the following steps: All sampling points within the sliding window are detected. When there are abnormal candidate points marked as composite abnormal candidate points, the active ultrasonic echo energy distribution and passive acoustic emission event energy distribution of all sampling points within the entire sliding window are further analyzed to determine the energy change trend within the sliding window. When the energy change trend simultaneously exhibits fluctuation, stability, and acoustic emission dual modes, the probability of leakage risk due to composite defects within the sliding window increases. When the energy change trend only has fluctuations and sudden spikes, the probability of leakage risk due to annular gap defects within the sliding window is increased. When the energy change trend only has a stable period and there is a continuous low-amplitude background, the probability of leakage risk due to mud skin defects within the sliding window is increased. When there is no energy change trend within the sliding window, the composite anomaly candidate point is determined to be an isolated noise point, and the label type of the composite anomaly candidate point is removed.

[0013] Secondly, this application provides a multi-source data fusion long casing removal leakage risk identification device, comprising: Memory is used to store computer-readable instructions non-transiently; A processor for executing computer-readable instructions, which, when executed by the processor, implement a first aspect of a multi-source data fusion method for identifying leakage risks during long casing removal.

[0014] Thirdly, this application provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, they implement the multi-source data fusion method for identifying leakage risks during long sleeve removal as described in the first aspect.

[0015] The beneficial effects of this invention are: This invention provides a method, device, and medium for identifying leakage risks during the removal of long casings using multi-source data fusion. In order to perceive the formation location and severity of the annular gap and the weak mud layer in real time during the removal of long casings and to detect leakage risks in a timely manner, the invention utilizes the differences in acoustic characteristics between the annular gap and the weak mud layer, and achieves real-time perception of the state of the pore-soil interface by combining active ultrasonic emission and passive acoustic emission monitoring.

[0016] This invention utilizes dynamic differentiated sampling, adaptively adjusting the step size and detection parameters based on the abnormal depth range to improve data density and signal-to-noise ratio in key areas, balancing detection accuracy and data efficiency. By combining active echo (energy, travel time) and passive acoustic emission (sudden spikes, continuous low amplitudes), it distinguishes different anomaly types with a low false positive rate. Based on the causal chain of "sudden increase in echo energy, shortened travel time, and synchronous acoustic emission" during the formation of annular gaps, and the stable trend characteristics during the formation of weak mud layers, it quantifies the energy change trend through the number of extreme points and the standard deviation / mean ratio, scientifically reflecting the defect morphology. By comprehensively considering the density of anomaly candidate points, energy fluctuation trends, and the cumulative integral of acoustic emission, it outputs the leakage risk probability. Simultaneously, through confirmation of the overall energy distribution within the window and dual-modal acoustic emission, it further improves the reliability of identification in complex scenarios, providing a direct basis for subsequent adaptive control (such as deceleration and grouting). Attached Figure Description

[0017] Figure 1 This invention provides a flowchart for identifying leakage risks during the removal of long casings using multi-source data fusion. Figure 2 This is a schematic diagram of active ultrasonic echoes of abnormal candidate points of different marker types provided in an embodiment of the present invention; in, Figure 2 (a) in the diagram represents the waveform under normal conditions. Figure 2 (b) in the diagram represents the waveform diagram corresponding to the candidate point of the annular gap anomaly. Figure 2 (c) in the figure represents the waveform diagram corresponding to the candidate point of mud skin anomaly; Figure 3 This is a schematic diagram of the passive acoustic emission waveforms of abnormal candidate points of different marker types provided in the embodiments of the present invention.

[0018] in, Figure 3 (a) in the diagram represents the waveform diagram corresponding to the candidate points of mud skin anomalies. Figure 3 (b) in the figure represents the waveform diagram corresponding to the candidate point of the anomaly. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: Example 1 like Figure 1As shown in this embodiment, a multi-source data fusion method for identifying leakage risks during long casing removal is proposed, which specifically includes the following steps: S1. Upon receiving the casing removal start signal, obtain the starting depth position of the long casing removal and the abnormal depth range determined based on historical geological survey data and historical construction data; It is understandable that, during the long casing removal stage, when an annular gap or a weak mud cake layer is formed between the outer wall of the casing and the soil around the pile, the two defects have different acoustic characteristics. Therefore, this embodiment utilizes the difference in acoustic characteristics between the annular gap and the weak mud cake layer to achieve real-time quantitative assessment of the state of the hole-soil interface through dual-mode fusion of active ultrasonic waves and passive acoustic emission.

[0020] Annular gaps typically form when the long casing is removed too quickly or the concrete lacks fluidity. The gap usually extends 0.2-0.5m. It is filled with air or groundwater, and its acoustic impedance differs significantly from that of the steel casing and the surrounding soil, causing strong sound wave reflection at this interface. This results in a wave pattern (energy spike → gradual decrease → another spike) on the active echo energy-depth curve. Simultaneously, the formation of the gap is accompanied by concrete tearing or soil fracturing, releasing transient acoustic emission events, which manifest as sudden spikes in the acoustic emission signal (amplitude > 80dB, duration < 10ms). The mud skin weak layer, on the other hand, is formed by the dragging of residual mud during drilling, typically covering a depth of several meters. Its acoustic impedance is between that of the steel casing and the soil. It reflects sound waves with moderate energy, exhibiting continuous and blurred reflections, showing a stable trend on the energy-depth curve (moderate energy, no drastic fluctuations). Its formation process is not accompanied by obvious fracturing events, but it may generate continuous, low-amplitude acoustic emission signals due to shear dragging. The differences in acoustic wave reflection and fracture acoustic emission characteristics between these two types of defects form the physical basis for distinguishing them. Furthermore, the formation of annular gaps often leads to the tearing or dragging of the original mud skin; therefore, the typical characteristics of composite defects (annular gap + mud skin) are: local waveform fluctuations and spikes appearing against a large, stable mud skin background, and the acoustic emission simultaneously includes sudden spikes and continuous low-amplitude backgrounds. The annular gap region (0.2~0.5m) is smaller than the mud skin region (several meters), and the energy curve of the composite defect exhibits both fluctuating and stable segments.

[0021] Because a single detection method cannot simultaneously meet the identification requirements of annular gaps and weak mud cake layers, active ultrasonic echo can accurately measure echo energy and travel time, directly reflecting the difference in interfacial acoustic impedance and determining the existence and width of annular gaps. However, it can only acquire the macroscopic characteristics of reflected echoes and cannot distinguish the composite defect of "gap-dominant + mud cake associated" (because gap reflection completely masks mud cake characteristics). Passive acoustic emission can capture transient signals of material fracture, with frequency bands separated from construction noise, and the casing, as a steel structure, provides a low-attenuation waveguide path, reflecting the characteristics of mud cake tearing / dragging, but it cannot accurately locate the defect depth. Therefore, combining the two methods—using active echo for gap determination and passive acoustic emission for mud cake determination—creates a complete identification of annular gaps, mud cake, and composite defects at the data fusion level.

[0022] Therefore, in order to collect different acoustic features, this embodiment uses multiple piezoelectric sensor arrays (8 arrays in this embodiment) evenly distributed circumferentially along the outer wall of the long protective casing. These arrays have dual functions of active ultrasonic wave transmission / reception and passive acoustic emission sensing. Each array is equidistant along the axial direction of the casing (0.5 meters apart), forming a rectangular sensor network covering the entire circumference of the casing. Each sensing unit is installed in a rotatable detection mounting tube within a radially protruding mounting slot on the outer wall of the casing, allowing it to switch between active detection and passive sensing modes before the long protective casing is removed. A signal acquisition and processing system is also provided, comprising an active ultrasonic excitation module (generating high-voltage pulses to drive the sensing unit to emit ultrasonic waves) and a passive acoustic emission receiving module (pre-amplifying, filtering, and digitizing the acoustic emission signal). The same sensing array is time-division multiplexed using an analog switch matrix.

[0023] Before the removal of the long casing begins, the depth range where leakage is likely to occur is pre-determined to provide initial prior information for subsequent sliding window detection, reducing blind searches and improving identification efficiency. Geological survey data and historical construction experience are transformed into quantifiable depth ranges, serving as the basis for subsequent control systems to proactively intensify monitoring and lower trigger thresholds. When the casing removal start signal is received, the system obtains the starting depth position of the long casing removal. h 0 (usually the initial burial depth of the bottom of the casing, set to 0). At the same time, read the abnormal depth range Φ from the locally stored geological exploration data and historical construction data, that is, the depth range in which the stratum is prone to form annular gaps or weak mud skin layers in history (such as alternating soft and hard strata, groundwater level fluctuation zones, and areas with concentrated defects detected by historical low strain).

[0024] S2. During the removal process, the real-time depth position of the casing is obtained. Based on the relative positional relationship between the real-time depth position and the abnormal depth range, the depth step size and detection parameters of the sensor are dynamically configured during data sampling. Multi-source sensing data of each sampling point is collected. The multi-source sensing data includes active ultrasonic echo energy and passive acoustic emission event energy. The detection parameters include the acquisition frequency, active ultrasonic emission power, and passive acoustic emission gain. The specific process for dynamically configuring the depth step size and detection parameters during data sampling is as follows: When the real-time depth position is within the abnormal depth range, the depth step size is set to the first step size, and the detection parameter is set to the first detection parameter. When the real-time depth position is outside the abnormal depth range, the depth step size is set to the second step size, and the detection parameter is set to the second detection parameter; wherein, the first step size is smaller than the second step size, and the first detection parameter is larger than the second detection parameter.

[0025] Specifically, by comparing whether the real-time depth position h is within the abnormal depth range Φ, the sampling step size and detection parameters are dynamically adjusted to achieve encrypted data collection in high-risk areas and routine data collection in low-risk areas. This balances data accuracy and computational efficiency, ensuring that sufficient sampling points can still be captured even when the defect extension range is small (0.2~0.5m). For example, the parameters can be set as follows: If h∉Φ, the depth step size is set to the second step size Δh2=0.10 m, and the detection parameters are the second detection parameters (normal values). If h∈Φ, the depth step size is set to the first step size Δh1=0.05 m, and the detection parameters are the first detection parameters (doubling the acquisition frequency, increasing the active ultrasonic emission power by 20%, and increasing the passive acoustic emission gain by 6dB). At each sampling point depth h, the system synchronously acquires: active ultrasonic echo energy E(h) (mV), travel time T(h) (μs), passive acoustic emission event energy signal (continuous sampling, sampling rate 1MHz, with timestamp), and tube removal speed v(h) (m / min). Simultaneously, the depth h of each sampling point and its corresponding time t(h) are recorded.

[0026] S3. Based on the relative positional relationship between the real-time depth position of the sampling point and the abnormal depth interval, as well as the corresponding multi-source sensing data, each sampling point is screened and marked to obtain abnormal candidate point markings. The marking types include annular gap anomalies and mud skin anomalies. Specifically, multi-source sensing data also includes travel time, and the specific process of filtering and labeling each sampling point is as follows: S31. Based on the speed of removing the long casing v ( h Calculate the neighborhood depth interval corresponding to the sampling point, for example, it can be [h−w,h+w], with a neighborhood depth interval radius w=0.1 m. Analyze the passive acoustic emission event energy distribution of all sampling points within the neighborhood depth interval, and extract the local acoustic emission features of each sampling point in the neighborhood depth space. Specifically, the process of extracting the local acoustic emission features of each sampling point in the neighborhood depth space is as follows: S311. Obtain the passive acoustic emission event energy signal segment within the neighborhood depth interval, input it into the bandpass filter to remove construction low-frequency noise, and then perform threshold detection to obtain the local acoustic emission signal as the local acoustic emission signal corresponding to all sampling points within the neighborhood depth interval. S312. Perform sudden spike detection on the local acoustic emission signal. When a signal segment with an amplitude exceeding the upper limit threshold (e.g., 80dB) and a duration of a preset time (e.g., <10ms) is detected in the local acoustic emission signal, it is determined that the local acoustic emission feature has a sudden spike. S313. Before performing continuous low-amplitude background detection on the local acoustic emission signal, if there are sudden spikes in the local acoustic emission signal, the sudden spikes need to be removed from the local acoustic emission signal first. The root mean square amplitude and standard deviation of the local acoustic emission signal are calculated. When the root mean square amplitude and standard deviation are both within the preset range, the local acoustic emission feature is judged to be the presence of continuous low-amplitude background.

[0027] S32. Calculate the active ultrasonic echo energy at each sampling point. E ( h ) and time of travel T ( h ) and preset reference acoustic parameters (reference) E 0( h ) T 0( h The ratio of the echo energy ratio is obtained by taking the ratio of the echo energy ... RE ( h and the rate of decrease in running time Δ T ( h ): RE ( h )= E 0( h ) / E ( h ), Δ T ( h )=[ T 0( h ) − T ( h )] / T 0( h ) in E 0( h The ) represents the initial reference echo energy at depth h. T 0( h () represents the initial calibration travel time of the first wave at depth h; S33. For each sampling point, analyze the local acoustic emission characteristics, echo energy ratio, and travel time shortening rate to determine if there are any anomalies. If there are anomalies, the sampling point will be used as anomaly candidate point, and each anomaly candidate point will be marked with the corresponding anomaly type.

[0028] Among them, the local acoustic emission characteristics include the presence of sudden spikes and continuous low-amplitude backgrounds. Based on acoustic reflection theory and fracture mechanics, the annular gap exhibits a large difference in acoustic impedance, high echo energy, and shortened travel time, and generates sudden acoustic emission upon formation. In contrast, the mud crust has moderate acoustic impedance, moderate echo energy, and slight changes in travel time, generating continuous low-amplitude acoustic emission upon tearing. The specific marking process is as follows: S331. When the echo energy ratio collected by the sampling point in the sampling point area is greater than the first preset energy threshold and the travel time reduction rate is greater than the first preset travel time threshold; Continue to determine whether there are sudden spikes and continuous low-amplitude backgrounds. When only sudden spikes exist at the sampling point, mark the sampling point as an abnormal candidate point for annular gap anomaly. When a sampling point simultaneously exhibits both a sudden spike and a continuous low-amplitude background, the sampling point is marked as a candidate point for a composite anomaly of annular gap anomaly and mud skin anomaly. S332. If the echo energy ratio collected by the sampling point is less than the first preset energy threshold and the travel time shortening rate is greater than the second preset travel time threshold, and if the sampling point only has a continuous low-amplitude background, the sampling point will be marked as an abnormal candidate point of mud skin anomaly.

[0029] The first preset energy threshold, the first preset travel time threshold, and the second preset travel time threshold are related to whether the real-time depth position h of the sampling point belongs to the abnormal depth range and the type of defect. For example, when determining whether there is an annular gap defect, if h∈Φ, the first preset energy threshold is 1.8 and the first preset travel time threshold is 0.15; otherwise, the first preset energy threshold is 2.0 and the first preset travel time threshold is 0.20. When determining whether there is a mud skin defect, the lower limit of the first preset energy threshold is 1.2, the upper limit is set according to whether it belongs to the abnormal depth range, and the second preset travel time threshold is 0.1.

[0030] S333. If neither of the conditions in S331 and S332 is met, then the sampling point is marked as a normal sampling point without any abnormalities.

[0031] It is understandable that the active ultrasonic echo waveform and passive acoustic emission waveform differ at different abnormal sampling points, such as... Figure 2 As shown, when the length of the casing is 15m, the abnormal depth range Φ={[3,6],[12,14]}, and the real-time depth position is 12.3m, which is within the abnormal depth range, when the sampling point is normal, it behaves as follows. Figure 2 In (a), the waveform is smooth, and when the sampling point is a candidate point for an annular gap anomaly, it behaves as follows: Figure 2 As shown in (b), the waveform is sharp and the travel time is significantly shortened. When the sampling point is a candidate point for mud skin anomaly, the behavior is as follows. Figure 2As shown in (c), the waveform exhibits moderately blurred reflections; as... Figure 3 As shown, the passive acoustic emission waveform is characterized as follows when the sampling point is a candidate point for mud skin anomaly: Figure 3 (a) Continuous low amplitude, when the sampling point is a composite anomaly candidate point, manifests as follows: Figure 3 In (b), there is a sudden spike followed by a continuous low-amplitude background.

[0032] S4. Construct a sliding window whose size changes with position to traverse all sampling points. Based on the depth distribution law of the annular gap and mud skin, perform piecewise linear regression calculation on the real-time depth position and ultrasonic echo energy of all abnormal candidate points in the sliding window to obtain the energy fluctuation trend. Specifically, the process of obtaining the energy fluctuation trend is as follows: S41. Arrange the ultrasonic echo energies of all abnormal candidate points within the sliding window into an energy sequence according to the acquisition order, and calculate the number of local extreme points, the standard deviation of energy distribution, and the arithmetic mean of echo energy in the energy sequence. S42. When the number of local extreme points exceeds the maximum threshold and the standard deviation of energy distribution is greater than a preset multiple of the arithmetic mean of echo energy, the energy fluctuation trend is judged to be fluctuating. S43. When the number of local extreme points is less than the minimum quantity threshold and the standard deviation of energy distribution is less than a preset multiple of the arithmetic mean of echo energy, the energy fluctuation trend is judged to be stable.

[0033] Moving from single-point anomaly detection based on discrete sampling to regional identification, transient interference such as debris blockage and sensor noise can be filtered out. Subsequent actions are only triggered when multiple consecutive depth points exhibit an abnormal trend. By analyzing the number of extreme points and the standard deviation / mean ratio of the energy sequence within the window, waveform fluctuations or stationary trends are extracted, reflecting the depth continuity characteristics of the defect. The extension patterns of annular gaps (fluctuations) and mud cake (stationarity) are utilized to effectively filter isolated noise. The size of the sliding window changes with depth; when the center depth of the sliding window is within the abnormal depth interval Φ, the sliding window size can be: Nwin =10 (corresponding to 0.45m). When located outside Φ, due to the change in step size, Nwin is also taken as 10 (corresponding to 0.9m). The real-time depth position of the sampling point at the latest moment in the sliding window is used. hcur With the endpoint as the starting point, take N consecutive sampling points forward from all sampling points to form a window H.

[0034] S5. Normalize the passive acoustic emission event energy of all abnormal candidate points within the sliding window to obtain the cumulative acoustic emission energy integral. IAEBased on the energy fluctuation trend, the cumulative energy integral of acoustic emission, and the density of abnormal candidate points of different marker types within the sliding window, the probability of leakage risk is comprehensively judged.

[0035] The specific process for comprehensively assessing the probability of leakage risk is as follows: S51. Assign different weights to anomaly candidate points of different label types within the sliding window, and calculate the anomaly candidate point density ρ within the sliding window based on the weight of each label type. S52. Assign different confidence levels to the energy fluctuation trend, the cumulative energy integral of acoustic emission, and the density ρ, respectively, to obtain the instantaneous probability within the current sliding window, obtain the instantaneous probability of the previous sliding window, and use the exponentially weighted moving average to comprehensively calculate the instantaneous probabilities of adjacent sliding windows to obtain the leakage risk probability.

[0036] To avoid isolated noise, the process of determining the probability of leakage risk also includes the following steps: All sampling points within the sliding window are detected. When there are abnormal candidate points marked as composite abnormal candidate points, the active ultrasonic echo energy distribution and passive acoustic emission event energy distribution of all sampling points within the entire sliding window are further analyzed to determine the energy change trend within the sliding window. When the energy change trend simultaneously exhibits fluctuation, stability, and acoustic emission dual modes, the probability of leakage risk due to composite defects within the sliding window increases; acoustic emission dual modes refer to the simultaneous presence of sudden spikes and continuous low-amplitude backgrounds. When the energy change trend only has fluctuations and sudden spikes, the probability of leakage risk due to annular gap defects within the sliding window is increased. When the energy change trend only has a stable period and there is a continuous low-amplitude background, the probability of leakage risk due to mud skin defects within the sliding window is increased. When there is no energy change trend within the sliding window, the composite anomaly candidate point is determined to be an isolated noise point, the label type of the composite anomaly candidate point is removed, and it is downgraded to a normal sampling point without anomalies.

[0037] To promptly reduce the probability of leakage risk, the leakage risk probability is compared with a trigger threshold (set to 0.55 within the abnormal range and 0.65 outside the abnormal range). When the leakage risk probability exceeds the trigger threshold, active verification is performed, and the following dynamic response characteristics are dynamically acquired: 1. Pause the removal of the long casing for 5 seconds, continuously collect the active ultrasonic echo energy over a period of time, fit the energy change curve, and calculate the linear fitting slope accordingly. kEt ; 2. The percentage decrease in acoustic emission count rate during the 2nd to 5th second of the long casing removal process relative to the last second before the pause, and the percentage decrease in removal force after the long casing removal process is paused relative to the period before the pause.

[0038] when kEt If the acoustic emission count rate decreases by more than 50% and the pull-out force decreases by less than 20% (<−0.1, normalized), it indicates a self-healing annular gap. The gap exists, but plastic concrete can slowly flow in and fill it. In this case, it is only necessary to reduce the pull-out speed of the long casing and use micro-vibration assistance. For example, gradually reduce the pull-out speed of the long casing from the current value (e.g., 1.5 m / min) to 0.5~1.0 m / min, with a reduction rate ≤0.1 m / min², to avoid abrupt changes causing secondary disturbances. Start the micro-vibration mode of the vibratory hammer, with an amplitude of 2~5 mm and a frequency of 20~30 Hz, and continue until the echo energy recovers to within 1.2 times the normal value. Micro-vibration promotes concrete flow and accelerates gap filling. After reducing the speed, continuously monitor the echo energy, calculating the echo energy ratio every 0.1 m depth. When the echo energy ratio of three consecutive sampling points is below 1.5, the gap is considered closed, and the pull-out speed can be gradually restored. when kEt If the acoustic emission count rate decreases by less than or equal to 30%, and the pull-out force decreases by more than or equal to 20%, it indicates a non-self-healing annular gap. This gap is either too large or the concrete has already begun to set and cannot self-fill. In this case, cement grout needs to be injected under high pressure into the borehole-soil interface through a pre-embedded pile-side grouting pipe. The grouting pressure is 0.5~2.0 MPa, the water-cement ratio is 0.6~1.0, and the grouting volume is estimated based on the gap volume. During the grouting process, the echo energy is monitored in real time. When the echo energy ratio drops below 1.5 and remains stable for more than 10 seconds, grouting is stopped. If it does not recover after 5 minutes, the grouting pressure is increased. After grouting, the area is left to stand for 3~5 minutes to allow the grout to begin to set. Then, the pull-out process is resumed at a low speed of 0.3~0.5 m / min, with continuous monitoring.

[0039] when kEt If the acoustic emission is approximately 0, and the acoustic emission is continuous with low amplitude (without sudden spikes) and the echo energy is moderate, it indicates that a weak mud skin layer has appeared. In this case, it is only necessary to record the depth range. After the long casing is completely pulled out, low-pressure, low-flow grouting is used to solidify the pile side according to the recorded depth range.

[0040] Example 2 This embodiment provides a multi-source data fusion device for identifying leakage risks during long casing removal, including: Memory is used to store computer-readable instructions non-transiently; A processor is configured to execute the computer-readable instructions, which, when executed by the processor, implement the multi-source data fusion method for identifying leakage risks during long casing removal as described in Embodiment 1.

[0041] Example 3 This embodiment provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, they implement a multi-source data fusion method for identifying leakage risks during long casing removal as described in Embodiment 1.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for identifying leakage risks during the removal of long casings using multi-source data fusion, characterized in that, Specifically, the following steps are included: S1. Upon receiving the casing removal start signal, obtain the starting depth position of the long casing removal and the abnormal depth range determined based on historical geological survey data and historical construction data; S2. During the removal process, the real-time depth position of the casing is obtained. Based on the relative positional relationship between the real-time depth position and the abnormal depth range, the depth step size and detection parameters of the sensor are dynamically configured during data sampling. Multi-source sensing data of each sampling point is collected. The multi-source sensing data includes active ultrasonic echo energy and passive acoustic emission event energy. S3. Based on the relative positional relationship between the real-time depth position of the sampling point and the abnormal depth interval, as well as the corresponding multi-source sensing data, each sampling point is screened and marked to obtain abnormal candidate point markings. The marking types include annular gap anomalies and mud skin anomalies. S4. Construct a sliding window whose size changes with position to traverse all sampling points. Based on the depth distribution law of the annular gap and mud skin, perform piecewise linear regression calculation on the real-time depth position and ultrasonic echo energy of all abnormal candidate points in the sliding window to obtain the energy fluctuation trend. S5. Normalize the passive acoustic emission event energy of all abnormal candidate points within the sliding window to obtain the cumulative acoustic emission energy integral. Based on the energy fluctuation trend, the cumulative acoustic emission energy integral, and the density of abnormal candidate points with different marker types within the sliding window, comprehensively judge the leakage risk probability.

2. The method for identifying leakage risks during long casing removal using multi-source data fusion as described in claim 1, characterized in that, The detection parameters include the acquisition frequency, active ultrasonic emission power, and passive acoustic emission gain. The specific process for dynamically configuring the depth step size and detection parameters during data sampling is as follows: When the real-time depth position is within the abnormal depth range, the depth step size is set to the first step size, and the detection parameter is set to the first detection parameter. When the real-time depth position is outside the abnormal depth range, the depth step size is set to the second step size, and the detection parameter is set to the second detection parameter; wherein, the first step size is smaller than the second step size, and the first detection parameter is larger than the second detection parameter.

3. The method for identifying leakage risks during long casing removal using multi-source data fusion as described in claim 1, characterized in that, Multi-source sensing data also includes travel time. The specific process of filtering and labeling each sampling point is as follows: Calculate the neighborhood depth interval corresponding to the sampling point based on the speed of the long casing removal, analyze the passive acoustic emission event energy distribution of all sampling points in the neighborhood depth interval, and extract the local acoustic emission features of each sampling point in the neighborhood depth space. The ratios of the active ultrasonic echo energy and travel time at the sampling points to the preset reference acoustic parameters are calculated to obtain the echo energy ratio and travel time reduction rate. For each sampling point, the local acoustic emission characteristics, echo energy ratio, and travel time shortening rate are comprehensively analyzed to determine if there are any anomalies. If anomalies are found, the sampling point is designated as anomaly candidate point, and each anomaly candidate point is marked with the corresponding anomaly type.

4. The method for identifying leakage risks during the removal of long casings using multi-source data fusion as described in claim 3, characterized in that, Local acoustic emission characteristics include the presence of sudden spikes and continuous low-amplitude background. When the echo energy ratio collected by the sampling point in the sampling point area is greater than the first preset energy threshold and the travel time shortening rate is greater than the first preset travel time threshold; Continue to determine whether there are sudden spikes and continuous low-amplitude backgrounds. When only sudden spikes exist at the sampling point, mark the sampling point as an abnormal candidate point for annular gap anomaly. When a sampling point simultaneously exhibits both a sudden spike and a continuous low-amplitude background, the sampling point is marked as a candidate point for a composite anomaly of annular gap anomaly and mud skin anomaly. If the echo energy ratio collected by the sampling point is less than the first preset energy threshold and the travel time shortening rate is greater than the second preset travel time threshold, and if the sampling point only has a continuous low-amplitude background, the sampling point will be marked as an abnormal candidate point for mud skin anomaly.

5. The method for identifying leakage risks during the removal of long casings using multi-source data fusion as described in claim 3, characterized in that, The process of extracting the local acoustic emission features of each sampling point in the neighborhood depth space is as follows: The passive acoustic emission event energy signal segment within the neighborhood depth range is obtained, and after removing low-frequency construction noise by inputting it into a bandpass filter, a threshold detection is performed to obtain the local acoustic emission signal. The local acoustic emission signal is subjected to sudden spike detection. When a signal segment with an amplitude exceeding the upper limit threshold and lasting for a preset time length is detected in the local acoustic emission signal, the local acoustic emission feature is determined to have a sudden spike. Continuous low-amplitude background detection is performed on the local acoustic emission signal. The root mean square amplitude and standard deviation of the local acoustic emission signal are calculated. When the root mean square amplitude and standard deviation are both within the preset range, the local acoustic emission feature is judged to have continuous low-amplitude background. When there are sudden spikes in the local acoustic emission signal, the sudden spikes need to be removed from the local acoustic emission signal before continuous low-amplitude background detection is performed.

6. The method for identifying leakage risks during long casing removal using multi-source data fusion as described in claim 4, characterized in that, The specific process for obtaining the energy fluctuation trend is as follows: The ultrasonic echo energies of all abnormal candidate points within the sliding window are arranged into an energy sequence according to the acquisition order. The number of local extreme points, the standard deviation of energy distribution, and the arithmetic mean of echo energy are calculated for the energy sequence. When the number of local extreme points exceeds the maximum threshold and the standard deviation of energy distribution is greater than a preset multiple of the arithmetic mean of echo energy, the energy fluctuation trend is judged to be fluctuating. When the number of local extreme points is less than the minimum threshold and the standard deviation of energy distribution is less than a preset multiple of the arithmetic mean of echo energy, the energy fluctuation trend is judged to be stable.

7. The method for identifying leakage risks during the removal of long casings using multi-source data fusion as described in claim 6, characterized in that, The specific process for comprehensively assessing the probability of leakage risk is as follows: Different weights are assigned to anomaly candidate points of different label types within the sliding window, and the anomaly candidate point density ρ within the sliding window is calculated based on the weight of each label type. Different confidence levels are assigned to the energy fluctuation trend, the cumulative energy integral of acoustic emission, and the density ρ, respectively, to obtain the instantaneous probability within the current sliding window. The instantaneous probability of the previous sliding window is obtained, and the instantaneous probability of adjacent sliding windows is calculated by using an exponentially weighted moving average to obtain the leakage risk probability.

8. The method for identifying leakage risks during long casing removal using multi-source data fusion as described in claim 3, characterized in that, The steps involved in assessing the probability of leakage risk are as follows: All sampling points within the sliding window are detected. When there are abnormal candidate points marked as composite abnormal candidate points, the active ultrasonic echo energy distribution and passive acoustic emission event energy distribution of all sampling points within the entire sliding window are further analyzed to determine the energy change trend within the sliding window. When the energy change trend simultaneously exhibits fluctuation, stability, and acoustic emission dual modes, the probability of leakage risk due to composite defects within the sliding window increases. When the energy change trend only has fluctuations and sudden spikes, the probability of leakage risk due to annular gap defects within the sliding window is increased. When the energy change trend only has a stable period and there is a continuous low-amplitude background, the probability of leakage risk due to mud skin defects within the sliding window is increased. When there is no energy change trend within the sliding window, the composite anomaly candidate point is determined to be an isolated noise point, and the label type of the composite anomaly candidate point is removed.

9. A multi-source data fusion device for identifying leakage risks during long casing removal, characterized in that, include: Memory is used to store computer-readable instructions non-transiently; A processor for executing the computer-readable instructions, which, when executed by the processor, implement the multi-source data fusion method for identifying leakage risks during long casing removal as described in any one of claims 1 to 8.

10. A non-transitory computer-readable storage medium, wherein, The non-transitory computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the multi-source data fusion method for identifying leakage risks during long casing removal as described in any one of claims 1 to 8.

Images