Intelligent detection method for pile body integrity based on multi-source data fusion

Abstract

The present application relates to the technical field of pile body detection, in particular to a pile body integrity intelligent detection method based on multi-source data fusion, comprising: collecting a frequency spectrum corresponding to a stress wave through time sequence, taking standard reference data as a benchmark to perform time sequence analysis on the stress wave data under the current scene, collecting effective observation pairs in each detection point in combination with the constraint conditions of the current scene; taking the stress wave data of the effective observation pairs as input, outputting the defect type and defect area of the pile body at different depths through inversion calculation; determining the correlation coefficient between each defect area at different time points for the obtained defect area, obtaining the depth distribution characteristics between each defect area; determining the abnormality judgment standard at each detection point based on the depth distribution characteristics, and integrating the defect area into the output pile body detection result according to the abnormality judgment standard. The accuracy and efficiency of pile body detection are realized.

Description

Technical Field

[0001] This invention relates to the field of pile body detection technology, specifically to an intelligent detection method for pile body integrity based on multi-source data fusion. Background Technology

[0002] Due to the uncertainty of geological conditions and the complexity of the construction process, pile foundations often exhibit integrity defects during the forming process, such as diameter enlargement, necking, segregation, mud inclusion, and pile breakage. Currently, the main methods for detecting pile foundation integrity include low (high) strain testing, acoustic wave transmission, thermal integrity pile foundation testing, and fiber optic grating methods. Different testing methods will exhibit different characteristics depending on the soil properties, requiring the analysis of the characteristic distribution patterns during the pile body testing process to output the pile integrity defects.

[0003] For example, Chinese Patent Publication No. CN115128163A discloses a method for detecting the integrity of bridge pile foundations based on small offset scattering wave imaging. The method involves acquiring hammer impact signals through a geophone and a seismic acquisition host, performing forward numerical simulation of acoustic waves using a superposition velocity model of P-waves, and performing cross-correlation imaging using the forward and reverse extrapolation fields of P-waves to obtain the final scattering wave migration imaging results.

[0004] For example, Chinese Patent Publication No. CN120522287A discloses a method for detecting the integrity of pile foundations, including: applying a variable amplitude acoustic excitation sequence to the pile foundation structure, collecting acoustic response signals under different excitation amplitudes, and obtaining nonlinear dynamic response data reflecting the rheological properties of the material; performing dispersion analysis and hysteresis characteristic analysis on the nonlinear dynamic response data, extracting the stress memory dissipation coefficient and nonlinear memory capacity index, and establishing a time-varying characteristic curve; determining the location and type of defects based on the abnormal change range and memory capacity decay rate of the time-varying characteristic curve, and obtaining the evolution state parameters of the defects; calculating the material deterioration rate and structural damage development rate based on the historical change trend of the evolution state parameters, predicting the health status index of the pile foundation structure, and forming a full life cycle performance evolution curve.

[0005] Existing technologies calculate forward and reverse extrapolation fields using hammer impact signals, and extract pile foundation inspection results under waveform offset by utilizing the cross-correlation image of the two extrapolation fields; or extract pile foundation inspection results under material damage by using dispersion analysis and hysteresis characteristic analysis, considering soil damping changes and linear stress changes. However, if data acquisition conditions under different seismic source scenarios are considered, it is also necessary to consider the location and distribution identification of pile foundation defects to avoid interference from pile foundation defects in various scenarios, thereby improving the accuracy of pile foundation defect location and processing efficiency. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a pile integrity intelligent detection method based on multi-source data fusion, including: configuring detection points according to the location of the pile, using the stress wave data obtained from each detection point as the main body of analysis, collecting the spectrum diagram corresponding to the stress wave through time series acquisition, and determining the scene information corresponding to each detection point.

[0007] Based on standard reference data, time-series analysis is performed on stress wave data under the current scenario. Combined with the constraints of the current scenario, valid observation pairs are collected from each detection point.

[0008] Using effective observation data of stress waves as input, the inversion calculation outputs the defect types and defect areas of the pile body at different depths.

[0009] For the acquired defect areas, multiple defect areas are arranged sequentially according to the acquisition time, and the correlation coefficient between each defect area at different time points is determined to obtain the depth distribution characteristics between each defect area.

[0010] Based on the depth distribution characteristics, the anomaly judgment criteria at each detection point are determined, and the defect area is integrated into the output pile body detection results according to the anomaly judgment criteria.

[0011] The beneficial effects of this invention are as follows: First, this invention configures detection points at equal intervals according to the pile body region boundary and seismic source type, and evenly distributes the detection points along the seismic source test direction, simultaneously collecting stress wave data under different seismic source types. This avoids the data bias caused by a single seismic source, provides a data foundation for subsequent data fusion analysis, and improves the accuracy of current pile integrity defect identification.

[0012] II. This invention employs constrained sorting and time-series synchronization verification. Observation pairs are sorted based on constraints such as frequency range, signal-to-noise ratio, and data distortion. Data synchronization is achieved through the binding of amplitude value, frequency value, and detection point. Subsequently, through multi-effect separation, exponential correction, and multi-map collaborative positioning, three types of effect data—site, path, and source—are extracted and interference is removed through exponential correction. Then, the wave impedance distribution curve is derived using the reflection coefficient. Combining the source / site effect separation map with the wave impedance curve, the defect type is determined from multiple dimensions, including frequency deviation, wave impedance deviation, and site effect value. Finally, defect areas are sorted according to integrity level. This reduces the deviation that occurs in the current data calculation and further clarifies the distribution pattern of each defect through data inversion, avoiding data calculation omissions caused by different sources in the current soil conditions, thereby improving the accuracy and efficiency of subsequent processing.

[0013] Third, this invention uses a correlation clustering + temporal integration approach. It calculates the correlation of defect areas and clusters them using the Pearson correlation coefficient, and then integrates the defect areas before and after the time sequence according to the principle of prioritizing intervals of the same depth and supplementing with the same defect type. It quantifies the temporal effectiveness under unified multi-source data, emphasizes the distribution pattern of concentrated and longitudinally extended defects, and provides sufficient feature basis for pile detection. Finally, it optimizes by defect type frequency and integrates multiple pile features, integrating the common features under batch processing and the differential features of individual piles, thereby improving the visibility and comprehensibility of the detection results. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1 This is a flowchart illustrating an intelligent detection method for pile integrity based on multi-source data fusion.

[0016] Figure 2 This is a site effect separation diagram of an intelligent detection method for pile integrity based on multi-source data fusion.

[0017] Figure 3 This is a source effect separation diagram of an intelligent pile integrity detection method based on multi-source data fusion.

[0018] Figure 4 This is a flowchart illustrating step S3 of the intelligent detection method for pile integrity based on multi-source data fusion.

[0019] Figure 5 This is a flowchart illustrating step S4 of the intelligent detection method for pile integrity based on multi-source data fusion. Detailed Implementation

[0020] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0021] See Figure 1 The intelligent detection method for pile integrity based on multi-source data fusion includes: S1, configuring detection points according to the location of the pile, using the stress wave data obtained from each detection point as the main body of analysis, acquiring the spectrum diagram corresponding to the stress wave through time series acquisition, and determining the scene information corresponding to each detection point.

[0022] S2, based on standard reference data, performs time-series analysis on stress wave data in the current scenario, and collects valid observation pairs from each detection point in combination with the constraints of the current scenario.

[0023] S3 takes the stress wave data of the effective observation pair as input, and outputs the defect type and defect area of ​​the pile body at different depths through inversion calculation.

[0024] S4. For the acquired defect areas, arrange multiple defect areas in sequence according to the acquisition time, determine the correlation coefficient between each defect area at different time points, and obtain the depth distribution characteristics between each defect area.

[0025] S5, based on the depth distribution characteristics, determines the anomaly judgment criteria at each detection point, and integrates the defect area into the output pile body detection results according to the anomaly judgment criteria.

[0026] The current scheme uses concrete piles, steel pipe piles, precast piles and other structures as the inspection objects. By using wave impedance distribution, it identifies cracks, necking, segregation and other conditions under integrity analysis, and quantifies the location and depth of defects.

[0027] After measuring stress wave data under different seismic sources, the equipment performs Fourier transform processing and combines the measurement data from each pile body to separate the site effect, path effect, and source effect. This allows the determination of the received waveform characteristics at each pile body location, and further identification of the location and corresponding depth of defects.

[0028] Generally, transient pulse force from hammer impact, impact force from free-falling hammer, periodic dynamic force generated by electromagnetic or hydraulically driven vibrators, and impact force from explosive detonation can be used as the seismic source for the current test. The stress waves generated are then processed as part of the current scenario verification. Since the underground medium exhibits relatively uniform medium continuity within a 2km range during waveform transmission, the test piles can be grouped into units of 1km each. Piezoelectric accelerometers, velocity detectors, and distributed fiber optic sensors are used to measure and acquire stress wave data.

[0029] In step S1, when configuring detection points according to the location of the pile, the implementation method further includes: S11, determining the regional boundary of the pile distribution area based on the current location of the pile and the gap between each pile; the regional distribution boundary is used to describe the regional boundary of the current pile when it is distributed.

[0030] S12 utilizes the seismic source type used at the regional boundary, and sets up detection points at equal intervals along the direction of the current seismic source test to collect stress wave data under different seismic source types. Detection points can be placed on the pile top, and set at equal intervals along the line according to the direction of the current seismic source test. The time-domain velocity signal of the transient excitation response at the pile top is collected, and a spectrum diagram with time as the abscissa and amplitude as the ordinate is generated to obtain the waveform of the pile body measurement. Simultaneously, the subsequent selection of frequency for observation is used to determine the relative relationship between frequency and signal energy, to help explain issues such as frequency distortion that occur during pile body detection.

[0031] It should be noted that the scene information at each detection point will also record the current position coordinates of the pile, the distribution of piles, the type of seismic source, the arrival time of stress waves, and the composition of the geological medium, which characterize the scene of pile integrity measurement.

[0032] In step S2, the reference event method will be used to perform preliminary time-series analysis on the currently acquired stress wave data and other data. Combined with the data of the corresponding scenario, data that can be effectively measured under the corresponding seismic source will be selected and regarded as the valid observation pair for output. Each valid observation pair consists of 1 seismic source + 1 detection point (pile top geophone) + 1 frequency point (0.5-30Hz). Then, based on the stress wave data measured under the corresponding combination, the effective extraction of data will be completed.

[0033] The constraints in the current scenario are mostly expressed as the signal-to-noise ratio, frequency range, and constraints that the data has no obvious distortion. Further check whether the detection point data is output synchronously and output the corresponding data.

[0034] The implementation method for collecting effective observation pairs from each detection point in step S2 includes: S21, introducing the frequency range of the current stress wave data, using frequency range constraints, signal-to-noise ratio constraints, and data distortion constraints as constraints under the current scenario, and sorting the collected stress wave data in sequence to form multiple observation pairs; the signal-to-noise ratio constraint requires that the signal-to-noise ratio be at least greater than or equal to 3 to prevent the acquisition of data with excessive noise, which would reduce the accuracy of subsequent inversion processing; the frequency range constraint will use a specific range for processing, such as the range of 0.5-30Hz, and process the data collected within this range to identify defects in the pile body; finally, the data distortion constraint requires that the data be positive, not infinite, and that the waveform has no sudden truncation or abnormal peaks in the spectrum.

[0035] S22, For the sorted observation pairs, perform a time synchronization check on each observation pair, and regard the checked observation pairs as valid observation pairs for output.

[0036] The time synchronization test involves aligning the measured data of all observation pairs in time and determining that each observation pair consists of one seismic source, one detection point, and one frequency point to obtain the output data combination.

[0037] When performing time-series synchronization verification for each observation pair, if the target is the spectrum corresponding to the stress wave data, the implementation method also includes: using the amplitude values ​​collected at each time sequence in the spectrum, binding the amplitude value range with each detection point, synchronously recording the frequency value of each detection point at the corresponding time, and taking the frequency value corresponding to the data and the amplitude value at the corresponding time sequence as the data of the current output observation pair.

[0038] In one embodiment of the present invention, in step S3, the stress wave data of the effective observation pair is inverted, and the wave impedance distribution of the pile body is inverted, and the data contribution of each detection point is introduced, thereby completing the output of the wave impedance distribution curve.

[0039] like Figure 4 As shown, when outputting the defect type and defect area of ​​the pile body at different depths in step S3, the implementation method also includes: S31, for the same batch of valid observation point pairs, based on the standard reference data, extract the data items corresponding to the site effect, path effect and source effect respectively through the inversion model.

[0040] S32, for the data terms of site effect, path effect and source effect, is corrected by exponential transformation to obtain the corrected data terms of site effect, path effect and source effect.

[0041] S33 recombines the data items corrected for site effect, path effect and source effect, and determines the reflection coefficient of each detection point based on the amplitude of the combined spectrum. Based on the reflection coefficient, the wave impedance at each depth is derived to obtain the output wave impedance distribution curve.

[0042] The inversion model will be calculated based on the Fourier spectral decomposition model. If the observed Fourier spectrum is used, the implementation formula is as follows: ,in, This represents the amplitude spectrum of the time series of ground motion (acceleration / velocity) recorded at detection point j, originating from source i, after Fourier transform. Here, i represents the index number of the source, j represents the index number of the detection point, and f represents the frequency. Indicates the magnitude of earthquake source i, used to interpret the specific description under the current vibration test; This represents the equivalent distance from the seismic source to the detection point. At this point, the integrity of the current pile body is judged based on the waveform characteristics of the seismic source in the corresponding scenario. This is to verify the modal characteristics under slight vibration waveforms, thereby identifying the feature distribution under different seismic sources and extracting the data characteristics under each scenario. The source spectrum of source i represents the frequency characteristics of the energy radiated by the source itself, and is used to explain the source effect in the current scenario. The site response coefficient of detection point j is used to explain the amplification / filtering effect of local geological conditions at detection point j on seismic waves and to explain the corresponding site effect. This represents the path attenuation term, which describes the energy decay during the propagation of seismic waves. In the current scenario, this path attenuation term is only used to distinguish the path effects of different seismic sources during testing. This represents the random residual term recorded from detection point j to earthquake source i, used to describe the random fluctuations of the model.

[0043] When separating the site effect, path effect, and source effect, the above three items are separated, and their values ​​are checked by logarithmic calculation.

[0044] At this point, standard reference data can be introduced to gradually extract data items under the three effects. These data items are then uniformly transformed into logarithmic vectors using a base-10 logarithm. Singular value decomposition is then used to solve for the parameters, and threshold settings are truncated. Only significantly non-zero singular values ​​are retained, and noise is filtered out. The portion of the singular value decomposition greater than the product of the maximum singular value and the truncation threshold is selected. This process is iterated continuously, projecting the corresponding singular values ​​onto the principal component directions of the left singular vector space. This transforms the singular value decomposition calculation process to the data items under each effect, resulting in relatively pure singular value decomposition data. These data are then subjected to an exponential transformation with a base-10 exponent to restore them to their original format. The values ​​of the three effects after removing correlation effects are then interpreted according to the three effects. The reflection coefficient at the corresponding location can be derived by recombining the values ​​of the three effects and using the ratio of the amplitude of the combined frequency spectrum to the reference amplitude. The reference amplitude represents the initial amplitude of the stress wave.

[0045] Generally, wave impedance is used to derive its impedance abrupt change interface. Once its impedance interface is known, the corresponding depth can be determined by the incident and reflected waves during stress wave acquisition and the time difference of the reflected waves. This allows us to know the wave impedance distribution at different impedance interfaces and thus explain the current wave impedance variation with depth.

[0046] For example ;in, Indicates the reflection coefficient; This represents the wave impedance value at the h-th impedance interface. The wave impedance here can be obtained by multiplying the concrete density, cross-sectional area, and wave velocity. When a defect or abrupt change occurs at any interface, the change in cross-sectional area at the defect location can be understood based on the constant value represented by the concrete density and the relative constant value of the wave velocity, thereby understanding the integrity of the pile body at the corresponding location. This represents the reference impedance value, indicating the wave impedance value when the pile is intact. It is typically set using the value of a complete section located 2-3 meters below the pile top; further depth can be determined by... ;in, This represents the depth value of the h-th impedance interface. The wave velocity value represents the baseline condition. This represents the reflected wave time difference at the h-th impedance interface. The reflected wave time difference is obtained by subtracting the reflected wave time from the incident wave time.

[0047] S34: For the data items corrected for site effect and source effect, extract the site effect separation map and source effect separation map when the effect is separated, and locate the defect area at each detection point by combining the value range of the wave impedance distribution curve.

[0048] Since the path effect exhibits path attenuation during the overall processing, it is difficult to verify the occurrence of defects in the corresponding scenario based solely on the changes in the path. Therefore, auxiliary processing can be carried out based on the site effect and the source effect.

[0049] For the site effect separation map and the source effect separation map, waveform images with frequency (in Hertz) as the horizontal axis and signal energy (in decibels or other arbitrary units, representing the signal energy at a specific frequency) as the vertical axis are used respectively. This method tends to use frequency + signal energy to identify the absorption characteristics of different frequency components and the identified spectral distortion, which helps to identify impedance changes in the corresponding layer. Abnormal energy peaks or valleys in the spectrum are used to verify the abnormalities identified by the waveform amplitude, thereby matching and locating the defect location.

[0050] like Figures 2-3 As shown in the figure, the site response amplitude and the source spectrum amplitude are used as the signal energy values ​​in the current scenario. Combined with the frequency value, the response of the pile body to different sources and sites is highlighted. These data will be used as reference data. The signal energy values ​​in the current scenario will be cross-validated with the original data to serve as an auxiliary basis for the processing of defect areas. This will ensure that the fluctuation of the signal energy can be associated with the unknown defects, types and degrees, thereby improving the accuracy of pile body integrity detection in the corresponding scenario.

[0051] The method for locating the defect area at each detection point in step S34 also includes: extracting all frequency points with signal energy deviations greater than a preset threshold for the source effect separation map; for each frequency point, finding the depth corresponding to the frequency point, and configuring the defect type of the corresponding detection point according to its depth value.

[0052] At this point, the preset threshold will be set according to the signal energy deviation between the current source effect separation map and the complete pile. When setting the unit in decibels, you can choose 5dB as the threshold, or use other methods to select the normalized value in the representation of the current image.

[0053] Only when the depth at a given frequency point is within the actual length of the pile is a defect type set, such as a depth-related defect. This defect type is then labeled according to the corresponding depth value to verify the signal energy deviation at that depth. Otherwise, it is considered an invalid difference, such as an error caused by uneven source excitation. Here, data within the pile is locked, and subsequent wave impedance deviation, reflection coefficient, and site effect values ​​are used for cross-validation.

[0054] The purpose of the source effect separation map is to remove the interference of uneven excitation of the source, restore the spectral characteristics of the source itself, and avoid the wave impedance inversion deviation caused by the instability of the source. The method of selecting frequency points for its frequency range is to determine the defect situation at the corresponding depth by the reflection instability.

[0055] For the wave impedance distribution curve, the defect type corresponding to the detection point is configured based on the wave impedance deviation value and the reflection coefficient.

[0056] The wave impedance deviation value will identify whether there is a change in the pile structure at the corresponding depth by checking the currently locked depth value and the wave impedance value in the surrounding ±0.2m range. When the wave impedance deviation value is >15%, it can be regarded as a change in the pile structure, and the corresponding defect type will be set; otherwise, it is regarded as a source of seismic interference at that location.

[0057] As for the reflection coefficient, we can check the range of values ​​for the reflection coefficient. When the reflection coefficient is not in the range of ±0.1, it is considered that a defect has occurred at the corresponding position. It is known that a positive reflection coefficient indicates diameter expansion and a negative reflection coefficient indicates diameter contraction. Based on its value range, we can set a defect type label for the corresponding position.

[0058] For the site effect separation map, extract the site effect value corresponding to each detection point, and configure the defect type corresponding to the detection point based on the site effect value.

[0059] When checking the site effect separation diagram, the judgment can be made based on the values ​​of the complete pile foundation. For example, the normal range of site effect values ​​is 0.8 to 1.2. If it does not fall within this range, it means that there is a deviation in the site effect. If it exceeds the corresponding range multiple times, it is considered to have a defect, and the relevant defect type is marked according to its value.

[0060] The purpose of the site effect separation map is to remove the interference of the soil layer, restore the response characteristics of the site itself, help distinguish the signal anomalies caused by site interference from those caused by pile defects, and summarize the data of the defective part according to the value range extracted from the image, together with the aforementioned extracted defect types, to verify the specific regional distribution of the pile defects.

[0061] Summarize the defect types of all detection points, arrange them according to the location corresponding to each defect type, and regard the arranged data as the output defect area.

[0062] In the current scenario, the defect types marked at all detection points will be arranged one by one according to their location. The depth coordinates of each defect type will be assigned to the depth range of ±0.2m. After all defect types are arranged according to the length of the depth range, the area corresponding to the depth range is regarded as the output defect area to explain the defect situation that exists in the pile body at multiple depths.

[0063] In scenarios where the above formula applies, the implementation method of arranging the positions corresponding to each defect type also includes: extracting the impedance interface corresponding to the reflection coefficient, using the wave impedance value and depth value at each impedance interface to determine the level of pile integrity, and identifying at least one sample data containing the expected level of pile integrity; the expected level of pile integrity will be set according to the ratio of the wave impedance value when the pile is intact at the corresponding depth to the current wave impedance value, and the ratio will be normalized to explain the current pile integrity ratio, and can be divided into five levels according to the interval of the ratio to explain the setting of the current expected level.

[0064] The current pile position is checked against the depth value at each impedance interface, the defect areas at each pile are synchronized, and the output order of the defect areas is determined according to the expected level of pile integrity in each defect area.

[0065] Each output defect area will be sorted according to the expected level of pile integrity, and finally the relevant areas will be checked and processed.

[0066] In one embodiment of the present invention, in step S4, the correlation between defect areas is used to further verify whether the pile defects are common to all, thereby reflecting the modal differences of different seismic sources under the test scenario.

[0067] like Figure 5 As shown, the implementation method of obtaining the depth distribution characteristics between each defect region in step S4 also includes: S41, for the defect regions at the same time point, clarifying the depth value range and spectrum range corresponding to each defect region.

[0068] S42, using the depth range and spectral range, calculate the correlation coefficient between any two defect regions according to the Pearson correlation coefficient; when using the Pearson correlation coefficient, calculate the Pearson correlation coefficient for each defect region based on the corresponding depth value and the corresponding spectral value, and then consider the average of the two as the correlation between the two defect regions.

[0069] It should be noted that the spectrum range represents the relevant data of the current defect area in the spectrum diagram.

[0070] S43, based on the correlation coefficient between any two defect regions, uses the correlation coefficient as the clustering index to cluster the defect regions, and regards the clustered defect regions as the output depth distribution features.

[0071] During clustering, the K-means algorithm can be used to set the k value according to the elbow rule. 0.6 can be selected as the threshold for the current clustering. Highly correlated defect areas at the corresponding depths will be classified to explain the relatively consistent defect types at different depths in the current scenario. These defect types will serve as the subject of the current pile detection and be integrated into the output data report.

[0072] Furthermore, if the correlation between defective regions at different points in time is considered, defective regions in the preceding and following time sequences can be integrated in a continuous time extension manner.

[0073] Therefore, the implementation of step S4 also includes: S44, for the defect areas at different time points, based on the clustered defect areas, combined with the integration results at different time points, to form the output depth distribution features.

[0074] The integration results at different time points also include: determining whether the defect areas of the preceding and following time series are within the same depth range; if they are within the same depth range, then integrating the defect areas of the preceding and following time series.

[0075] If they are not within the same depth range, the defect areas are compared in the manner of continuous depth ranges, and defect areas with the same defect type are integrated. Continuous depth ranges can be defined by the depth difference between adjacent defect areas being less than 0.3m, indicating that the defect areas appear continuously.

[0076] The remaining defect areas represent areas that are neither adjacent nor of the same type. These can be directly divided into corresponding defect areas and output separately, thus completing the integration of all defect areas. The output depth distribution characteristics will represent the representation of pile defects at different depths.

[0077] The anomaly judgment criteria for each detection point represent the calculation process and specific feature values ​​used when extracting depth distribution characteristics, which in turn represent the processing flow for the corresponding defects. By combining the processing flow with the defect area output, the distribution patterns and commonly used judgment parts in the pile body inspection process can be clearly identified, as well as the sensitivity of each detection point to different features.

[0078] In step S5, when integrating the defect area into the output pile body detection result, the implementation method also includes: according to the anomaly judgment criteria of each detection point, selecting the depth distribution feature at any pile body, optimizing the depth distribution feature at multiple pile bodies according to the frequency of occurrence of the defect type at the current detection time, and taking the optimized data as the output pile body detection result.

[0079] When performing feature optimization, features are examined based on the frequency of defect types at different depths to identify frequently occurring pile defects. Defect types can be prioritized based on their frequency ratio, or data search can be used to combine defect type occurrence batches with other data to create a dataset based on specific frequencies. This dataset is then output as multiple frequency-related datasets to improve the visibility of pile inspection results.

[0080] At this point, based on the proportion of batches that appear, we can output the common features that appear more frequently. Then, based on the features that appear less frequently, we compare these less frequent data with the common data, and thus we can know the difference data of a single pile body relative to other pile bodies. We then regard these data as the data after optimization.

[0081] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for intelligent detection of pile integrity based on multi-source data fusion, characterized in that, include: Based on the location of the pile, detection points are configured, and the stress wave data obtained from each detection point is used as the main body of analysis. The stress wave spectrum corresponding to the time series is collected, and the scene information corresponding to each detection point is determined. Based on standard reference data, time-series analysis of stress wave data under the current scenario is performed, and effective observation pairs are collected from each detection point in combination with the constraints of the current scenario. Using effective observation data of stress waves as input, the inversion calculation outputs the defect types and defect areas of the pile body at different depths; For the acquired defect areas, multiple defect areas are arranged sequentially according to the acquisition time, and the correlation coefficient between each defect area at different time points is determined to obtain the depth distribution characteristics between each defect area. Based on the depth distribution characteristics, the anomaly judgment criteria at each detection point are determined, and the defect area is integrated into the output pile body detection results according to the anomaly judgment criteria. The methods for collecting valid observation pairs from each detection point include: The frequency range of the current stress wave data is introduced, and the frequency range constraint, signal-to-noise ratio constraint, and data distortion constraint are used as constraints under the current scenario. The collected stress wave data are sorted in turn to form multiple observation pairs. For each sorted observation pair, a time synchronization check is performed, and the checked observation pair is considered as a valid output observation pair. When outputting the defect types and defect areas at different depths of the pile body, the implementation methods also include: For the same batch of valid observation point pairs, based on the standard reference data, the data items corresponding to the site effect, path effect and source effect are extracted by the inversion model respectively. For the data items of site effect, path effect and source effect, an exponential transformation is used to correct them, resulting in the corrected data items of site effect, path effect and source effect. The data items corrected for site effect, path effect and source effect are recombined, and the reflection coefficient of each detection point is determined by the amplitude of the combined spectrum. Based on the reflection coefficient, the wave impedance at each depth is derived to obtain the output wave impedance distribution curve. For the data items corrected for site effect and source effect, extract the site effect separation map and source effect separation map when the effect is separated, and locate the defect area at each detection point by combining the value range of the wave impedance distribution curve. Other methods for locating defect areas at each detection point include: For the source effect separation map, extract all frequency points where the signal energy deviation is greater than a preset threshold; for each frequency point, find the depth corresponding to that frequency point, and configure the defect type of the corresponding detection point according to its depth value; For the wave impedance distribution curve, the defect type corresponding to the detection point is configured based on the wave impedance deviation value and the reflection coefficient, respectively. For the site effect separation map, extract the site effect value corresponding to each detection point, and configure the defect type corresponding to the detection point based on the site effect value. Summarize the defect types of all detection points, arrange them according to the location corresponding to each defect type, and regard the arranged data as the output defect area.

2. The intelligent pile integrity detection method based on multi-source data fusion according to claim 1, characterized in that, When configuring detection points based on the location of the pile, the implementation methods also include: Based on the current location of the pile, the regional boundary of the pile distribution area is determined by the gaps between the piles; the regional distribution boundary is used to describe the regional boundary when the current pile is distributed. By utilizing the type of seismic source used at the regional boundary, detection points are configured at equal intervals along the direction of the current seismic source test to collect stress wave data under different seismic source types.

3. The intelligent detection method for pile integrity based on multi-source data fusion according to claim 1, characterized in that, When performing time-series synchronization checks on each observation pair, the implementation methods also include: The amplitude values ​​collected at each time step in the spectrum are bound to each detection point according to the range of amplitude values. The frequency values ​​of each detection point at the corresponding time step are recorded synchronously. The data corresponding to the frequency values ​​and the amplitude values ​​at the corresponding time steps are regarded as the data of the current output observation pair.

4. The intelligent detection method for pile integrity based on multi-source data fusion according to claim 1, characterized in that, The implementation method of arranging according to the position corresponding to each defect type also includes: The impedance interface corresponding to the reflection coefficient is extracted. The wave impedance value and depth value at each impedance interface are used to determine the level of pile integrity and identify at least one sample data containing the expected level of pile integrity. The current pile position is checked against the depth value at each impedance interface, the defect areas at each pile are synchronized, and the output order of the defect areas is determined according to the expected level of pile integrity in each defect area.

5. The intelligent detection method for pile integrity based on multi-source data fusion according to claim 1, characterized in that, Other methods for obtaining the depth distribution characteristics between different defect regions include: For defect areas at the same point in time, clarify the depth range and spectral range corresponding to each defect area; By taking into account the depth range and the spectral range, the correlation coefficient between any two defect regions is calculated according to the Pearson correlation coefficient. Based on the correlation coefficient between any two defect regions, the defect regions are clustered using the correlation coefficient as the clustering index, and the clustered defect regions are regarded as the output depth distribution features. For defect regions at different time points, based on the clustered defect regions and combined with the integration results at different time points, the output depth distribution characteristics are formed.

6. The intelligent detection method for pile integrity based on multi-source data fusion according to claim 5, characterized in that, The methods for achieving integration results at different points in time also include: Determine whether the defect regions of the preceding and following time sequences are within the same depth range. If they are within the same depth range, then integrate the defect regions of the preceding and following time sequences. If they are not within the same depth range, compare the defect areas in the manner of continuous depth ranges and integrate the defect areas with the same defect type.

7. The intelligent detection method for pile integrity based on multi-source data fusion according to claim 1, characterized in that, When integrating defect areas into the output pile body inspection results, the implementation methods also include: Based on the anomaly judgment criteria for each detection point, select the depth distribution characteristics at any pile body, optimize the depth distribution characteristics at multiple pile bodies according to the frequency of occurrence of the defect type at the current detection time, and take the optimized data as the output pile body detection result.

Images