Method for detecting semi-buried metamaterial pile based on multi-source data fusion

Abstract

The application relates to the technical field of data processing, in particular to a semi-buried metamaterial pile detection method based on multi-source data fusion, which comprises the following steps: acquiring various monitoring parameters of a pile to be detected, collecting all kinds of monitoring parameters of different piles in a historical period, and forming a historical set; calculating the response difference of various monitoring parameters of each pile; calculating an interference heterogeneity representing the difference in disturbed states of any two piles under multiple interferences, clustering all piles in the historical set according to the interference heterogeneity, performing sample balancing processing on the various monitoring parameters of all piles in different clusters, and training a neural network model to obtain a noise reduction network model; and performing noise reduction processing on the various monitoring parameters of the pile to be detected, so as to judge and evaluate the health state of the pile to be detected. The application improves the detection accuracy and reliability of the health state of the semi-buried metamaterial pile.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method for detecting semi-embedded metamaterial piles based on multi-source data fusion. Background Technology

[0002] Semi-embedded metamaterial piles are a new type of seismic isolation and damping component that combines the local resonance mechanism of metamaterials with pile foundation structures. They are widely used in seismic isolation and support projects in vibration-sensitive areas such as rail transit and building construction. Because their service life spans both above-ground and underground media environments, it is necessary to detect the displacement, deformation, and other characteristics of the piles to prevent the loss of seismic isolation gaps due to structural damage, which would lead to the failure of protective capabilities and ensure the seismic safety of engineering sites and buildings.

[0003] In actual testing, due to the complexity of the sensor acquisition environment, the data is easily affected by multiple factors such as soil inhomogeneity, interface wave disturbance, pile-soil coupling, and external environmental dynamic loads, resulting in a large amount of complex noise interference. This leads to severe distortion and falsification of the monitoring data. At the same time, due to the strong local heterogeneity of geological interference faced by different piles, the severe imbalance of data samples under different interference conditions causes the noise reduction model to overfit to the conventional interference of large samples, and fails to accurately filter out the complex and strong interference in local severe areas. This results in poor noise reduction effect of the data, and cannot effectively suppress the data deviation caused by the superposition of multi-source noise, affecting the accuracy of the detection of the health status of the pile. Summary of the Invention

[0004] To address the aforementioned technical issues, a method for detecting semi-embedded metamaterial piles based on multi-source data fusion is provided to resolve existing problems.

[0005] The solution to the technical problem presented in this application is to provide a method for detecting semi-embedded metamaterial piles based on multi-source data fusion, comprising the following steps: Multiple monitoring parameters of the pile to be tested are obtained, and all kinds of monitoring parameters of different piles in historical periods are collected to form a historical set; For the historical set, the spatial distribution characteristics of two adjacent piles and their differences under different monitoring parameters are analyzed to characterize the data distortion of the piles under multiple interferences, and the response difference of each pile under various monitoring parameters is calculated. The deviation characteristics and discreteness of the response difference of monitoring parameters between any two piles in the historical set are evaluated. The interference heterogeneity characterizing the difference of the disturbed state of any two piles under multiple disturbances is calculated. Based on this, all piles in the historical set are clustered. The various monitoring parameters of all piles in different clusters are sample balanced. The balanced data is used to train the neural network model to obtain the noise reduction network model. Multiple monitoring parameters of the pile to be tested are input into the noise reduction network model for noise reduction processing. The health status of the pile to be tested is judged and evaluated using the noise-reduced monitoring parameters.

[0006] Preferably, the monitoring parameters include strain data, tilt angle, relative displacement, horizontal displacement, and vertical displacement.

[0007] Preferably, the calculation of the response variability of various monitoring parameters for each pile includes: For each pile in the historical set, multiple piles adjacent to it are defined as neighboring piles; the data of each monitoring parameter of each pile at all times are used to form various parameter sequences, and the difference between each pile and each neighboring pile for the same parameter sequence is calculated as the relative difference. Based on the spatial distance between each pile and its nearest neighboring piles, the association weight between each pile and its nearest neighboring piles is determined. Based on the association weight, the relative differences of the same monitoring parameter between each pile and all its neighboring piles are weighted and summed to serve as the response difference of each pile for various monitoring parameters.

[0008] Preferably, the calculation process of the association weight is as follows: negatively map the spatial distance between each pile and each of its nearest neighboring piles, normalize it, and use it as the association weight between each pile and each of its nearest neighboring piles.

[0009] Preferably, the calculation process for the interference heterogeneity is as follows: Calculate the difference in response to the same monitoring parameter between any two piles, perform a positive mapping on it, and use it as the abnormal deviation amount; A positive mapping is performed on the dispersion of the response differences of all piles under various monitoring parameters, which serves as the interference sensitivity of various monitoring parameters; The interference sensitivity and abnormal deviation are positively fused to determine the interference heterogeneity of any two piles.

[0010] Preferably, the specific process of the forward fusion is as follows: using the interference sensitivity of all kinds of monitoring parameters as weights, the abnormal deviation of all monitoring parameters between any two piles is weighted and summed to obtain the interference heterogeneity of any two piles.

[0011] Preferably, the step of performing sample balancing processing on multiple monitoring parameters of all piles within different clusters includes: forming a sample set by combining all monitoring parameters of all piles within the same cluster; performing sample balancing processing on all sample sets to obtain all balanced sample sets.

[0012] Preferably, training the neural network model to obtain the noise reduction network model includes: acquiring multiple monitoring parameters of the pile body under an interference-free environment, and using the multiple monitoring parameters of all pile bodies in the same cluster under an interference-free environment as sample labels; using all balanced sample sets and their sample labels to train the neural network model, and using the trained neural network model as the noise reduction network model.

[0013] Preferably, the process of obtaining various monitoring parameters of the pile body under an interference-free environment is as follows: a theoretical analysis model of the semi-embedded metamaterial pile body is established in advance using finite element simulation software, and a preset standard excitation load is applied to it to output various monitoring parameters of the pile body under an interference-free environment.

[0014] Preferably, the step of judging and evaluating the health status of the pile to be tested includes: comparing and analyzing various monitoring parameters after noise reduction with corresponding preset thresholds to obtain the judgment results of each monitoring parameter; taking the judgment results of all monitoring parameters as independent evidence sources, setting multiple health levels as identification frameworks, using DS evidence theory to fuse and calculate all evidence sources, and outputting the health level to which the pile to be tested belongs.

[0015] This application has at least the following beneficial effects: This application calculates the response variability of various monitoring parameters for each pile. Its advantages include fully considering the spatial local correlation of engineering geological interference. By analyzing the spatial distribution of adjacent piles and their differences under multiple parameters, it can accurately quantify the measurement distortion caused by multiple interferences to a single pile, effectively identifying the difference between local abnormal responses and global noise distribution. It also calculates the interference heterogeneity characterizing the difference in the disturbed state of any two piles under multiple interferences. Its advantages include accurately quantifying the difference in the disturbed state of two piles under the superposition of multiple interferences by evaluating the deviation characteristics and discreteness of the response variability, effectively characterizing the spatial heterogeneity of the interference environment between different piles. Furthermore, it clusters all piles in the historical set. Its advantages include accurately classifying piles with similar interference characteristics and homogeneous noise distributions through clustering, laying the foundation for sample equalization and noise reduction model training. Finally, it performs sample equalization processing on various monitoring parameters of all piles within different clusters. Using the equalized data, it trains a neural network model to obtain a noise reduction network model. Its advantages include effectively eliminating the significant difference in the number of samples due to different interference types by introducing sample equalization processing. This approach addresses model training bias by avoiding oversmoothing or underfitting caused by majority-class samples during training. By utilizing balanced data, the neural network can learn signal patterns under various interference characteristics fairly and effectively. The trained denoising network model exhibits stronger adaptability and better generalization performance against multi-source interference in complex geological environments. Multiple monitoring parameters of the pile under test are input into the denoising network model for noise reduction. The denoised parameters are then used to judge and evaluate the health status of the pile. The beneficial effects include the effective filtering of complex noise and distortion components in the data, restoring the true trend of monitoring parameters, and eliminating underlying measurement bias caused by multi-source interference. Furthermore, by using high-quality denoised data for health status assessment, the system achieves comprehensive, high-precision, and robust intelligent detection and health assessment of semi-embedded metamaterial piles. This fundamentally avoids the decision-making risk of detection bias caused by input distortion, improves the effective correction and noise suppression of data in complex geological environments, and enhances the accuracy and reliability of detecting the health status of semi-embedded metamaterial piles. Attached Figure Description

[0016] The following section provides a more detailed description of the method for detecting semi-embedded metamaterial piles based on multi-source data fusion, in conjunction with the accompanying drawings.

[0017] Figure 1 A flowchart illustrating the steps of a method for detecting semi-embedded metamaterial piles based on multi-source data fusion provided in this application embodiment; Figure 2 A flowchart illustrating the steps of a method for obtaining response difference in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of the method for detecting semi-embedded metamaterial piles based on multi-source data fusion, in conjunction with the accompanying drawings and implementation examples, is provided. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for detecting semi-embedded metamaterial piles based on multi-source data fusion according to an embodiment of this application. The method includes the following steps: Step 1: Obtain various monitoring parameters of the pile to be tested, and collect all kinds of monitoring parameters of different piles in historical periods to form a historical set.

[0021] During their service life, semi-embedded metamaterial piles not only perform conventional load-bearing and deformation control functions, but also, through internal resonant units and elastic structures, regulate the propagation of elastic waves, thereby serving specific objectives such as vibration reduction, seismic isolation, or wave field manipulation. With increasing service time, under soil loads, external dynamic loads, and long-term pile-soil interactions, the pile may experience displacement, structural deformation, and relative movement of internal mass blocks. If these changes are not identified promptly and accurately, they will directly threaten the structural safety of the pile itself and the stability of its metamaterial functions, and may even lead to the failure of the overall vibration reduction or seismic isolation system. Therefore, in order to fully understand the working status of the pile during its service life and accurately assess the safety of its overall structure and the effectiveness of its internal metamaterial functions, a comprehensive perception and quantitative assessment of the service status of semi-embedded metamaterial piles is necessary.

[0022] Based on the above analysis, sensor devices deployed on the piles are used to collect multiple monitoring parameters of each pile in real time, specifically: In this embodiment, a semi-buried metamaterial pile is used as the detection object. The piles are arranged in a periodic array, with part of them buried in the soil and the other part exposed. On the exposed part of the pile, monitoring points are selected at equal intervals from the top of the pile to the ground surface along the height direction of the pile. Fiber optic grating detectors are attached to each monitoring point to collect strain data at each monitoring point in real time. The maximum strain data of all monitoring points at the same time is extracted as the strain data of each pile at each time to monitor the structural deformation of the pile. An inclination monitoring device is installed at the top of the pile to collect the inclination angle of the pile at various times in real time in order to monitor the overall inclination posture of the pile. A wire-type displacement meter is installed between the concentrated mass block of the resonant unit inside the pile body and the matrix concrete. The shell of the wire-type displacement meter is fixed to the inner wall of the matrix concrete, and the wire end is fixed to the concentrated mass block. By measuring the change in the extension length of the wire in real time, the relative displacement of the mass block relative to the pile body is continuously collected to evaluate the relative motion state of the internal unit of the metamaterial. A GNSS receiver is deployed on the top of the pile to collect the horizontal displacement between the pile and the soil, in order to monitor the horizontal slip characteristics under the pile-soil coupling state. A static level is arranged vertically along the pile to measure the vertical displacement of the pile, i.e., vertical displacement, in order to monitor the settlement of the pile. It should be noted that the acquisition frequency for strain data, tilt angle, relative displacement, and horizontal displacement is 100Hz, and the acquisition frequency for vertical displacement is 1Hz. As for other implementation methods, the implementer can set the frequency according to the actual situation.

[0023] Therefore, for the pile body to be tested, data of each monitoring parameter of the pile body to be tested at each time are collected according to the above process. The monitoring parameters include strain data, tilt angle, relative displacement, horizontal displacement and vertical displacement. It also collects all monitoring parameters of different piles during historical periods to form a historical dataset; Time alignment processing is performed on the data of all monitoring parameters in the historical set. Specifically, using the time axis of the high acquisition frequency as the reference, a linear interpolation method is used to interpolate between two adjacent vertical displacements, so that the time resolution of the vertical displacement is improved to the level of the high acquisition frequency. The linear interpolation method is a well-known technique and will not be described in detail here.

[0024] At this point, various monitoring parameters and historical data of the pile to be tested are obtained.

[0025] Step 2: For the historical set, analyze the spatial distribution characteristics of two adjacent piles and their differences under different monitoring parameters to characterize the data distortion of the piles under multiple interference effects, and calculate the response difference of various monitoring parameters for each pile.

[0026] Furthermore, since the semi-embedded metamaterial piles are located in complex geological environments, the data collected during the detection process is easily affected by multiple factors such as pile-soil coupling, non-uniform soil settlement, interface wave scattering, local resonance effect, and complex external dynamic loads. This results in a low signal-to-noise ratio of the collected multi-source data, which deviates significantly from the true mechanical state of the pile. Therefore, in order to extract effective structural response characteristics and achieve accurate health assessment, it is necessary to reduce noise in the data to suppress detection errors caused by multiple noise interferences.

[0027] Secondly, the piles are arranged in a periodic array, with multiple piles linked, partially embedded, and staggered heights. Under the influence of multiple disturbances, due to differences in their layout and local soil conditions, the dynamic responses of piles at different spatial locations exhibit significant spatial heterogeneity under the same external excitation or environmental disturbance. By analyzing the differences in the same monitoring parameter between each pile and the surrounding piles, the response difference degree is calculated. The flowchart of the method for obtaining the response difference degree provided in this application embodiment is as follows: Figure 2 As shown, it specifically includes: For each stake in the history set, multiple stakes adjacent to it are defined as nearest neighbor stakes; In this embodiment, with each pile as the center, the K nearest neighbor algorithm is used to obtain the K nearest piles to each pile as the nearest neighbor piles. Here, K is set to 7. As for other implementation methods, the implementer can set it according to the actual situation. The K nearest neighbor algorithm is a well-known technology and will not be described in detail here.

[0028] The data of each monitoring parameter of each pile at all times are used to form various parameter sequences. The difference between each pile and each of its nearest neighboring piles for the same parameter sequence is calculated as the relative difference. In this embodiment, the DTW distance between each pile and each of its nearest neighboring piles for the same parameter sequence is calculated as a relative difference. The DTW distance is a well-known technique and will not be described in detail here.

[0029] It should be noted that the greater the relative difference, the more significant the response deviation between the two piles caused by multiple disturbances to the monitoring parameter, and the higher the degree of distortion of the pile state.

[0030] Negative mapping is performed on the spatial distance between each pile and its nearest neighbor piles, and the normalized distance is used as the association weight between each pile and its nearest neighbor piles. In this embodiment, spatial distance is measured by calculating the Euclidean distance between each pile and its nearest neighbor piles. Euclidean distance is a well-known technique and will not be described in detail here. The specific process of negative mapping is as follows: the reciprocal of the spatial distance is used as the negative mapping result. The specific process of normalization is as follows: the sum of the negative mapping results between each pile and all its nearest neighbor piles is calculated, and the ratio of the negative mapping result to the sum is used as the association weight. In order to avoid the denominator being 0 when calculating the reciprocal, a parameter tuning factor is added to the denominator. In this embodiment, the parameter tuning factor is set to 0.001. In other implementation methods, the implementer can set it according to the actual situation.

[0031] It should be noted that, since the geological conditions and environmental disturbances of two piles that are closer together are more similar, the correlation between environmental and soil disturbances is higher; conversely, the farther apart two piles are in space, the lower the correlation between environmental and soil disturbances. This makes the correlation weight between two piles that are closer in space greater, so that neighboring piles with strong spatial correlation contribute more to the assessment of relative differences, thus more accurately highlighting the abnormal deviation and measurement distortion of piles in locally similar environments.

[0032] Based on the association weight, the relative differences of the same monitoring parameter between each pile and all its nearest neighbor piles in the historical set are weighted and summed to serve as the response difference of each pile in the historical set for various monitoring parameters. In this embodiment, the formula for calculating the response difference is: in, For the first The first pile The degree of response difference of the monitoring parameters For the first The pile body and the first Between the nearest piles, the first The relative difference of the monitoring parameters. For the first The pile body and the first The correlation weight between neighboring piles For the first The number of all neighboring piles corresponding to each pile; It should be noted that the greater the response difference, the greater the difference between the pile and the piles that are closer to it in space under the monitoring parameter, reflecting that the pile is more significantly affected by multiple interferences and local anomalies under the monitoring parameter, and the more the data of the monitoring parameter collected from the pile deviates from the true state; conversely, it indicates that the state of the pile is more consistent with that of the surrounding piles, and the reliability of the data is higher.

[0033] Thus, the response differences of various monitoring parameters for each pile in the historical data set are obtained.

[0034] Step 3: Evaluate the deviation characteristics of the response difference of the monitoring parameters between any two piles in the historical set, as well as the discreteness of the response difference. Calculate the interference heterogeneity, which characterizes the difference in the disturbed state of any two piles under multiple disturbances. Based on this, cluster all piles in the historical set. Perform sample equalization processing on the multiple monitoring parameters of all piles in different clusters. Use the equalized data to train the neural network model to obtain the noise reduction network model.

[0035] In complex geological environments, the superimposed effects of multiple interference sources on semi-embedded metamaterial piles exhibit heterogeneity and multidimensional coupling, leading to significant differences in the degree of distortion across different dimensions such as displacement, deformation, and relative motion among different piles. In other words, the response differences between piles vary. To quantify the anomalies in pile response under the superimposed effects of different interferences, further comparative analysis of the response differences between piles is necessary. Based on this, the deviation of the response differences between two piles for different monitoring parameters is analyzed, and the interference difference is calculated to assess the overall response deviation of the two piles under the superimposed effects of multiple interferences. Specifically: Calculate the difference in response to the same monitoring parameter between any two piles in the historical set, and perform a positive mapping on it as an abnormal deviation quantity; In this embodiment, the absolute value of the difference in response difference between any two piles under the same monitoring parameter is calculated. The positive mapping process is as follows: select the maximum response difference of all piles under various monitoring parameters; take the ratio of the absolute value of the difference to the maximum response difference under the monitoring parameter as the abnormal deviation amount. If the maximum response difference is 0, then the abnormal deviation amount is directly set to 0.

[0036] It should be noted that if the maximum response difference is 0, then the abnormal deviation amount is set to 0.

[0037] A positive mapping is performed on the dispersion of the response differences of all piles under various monitoring parameters, which serves as the interference sensitivity of various monitoring parameters; In this embodiment, the degree of dispersion is measured by calculating the variance of the response differences of all piles under various monitoring parameters; secondly, the positive mapping process is as follows: calculate the sum of the degrees of dispersion of all kinds of monitoring parameters, and take the ratio of the degree of dispersion of various monitoring parameters to the sum as the interference sensitivity; the normalization of the degree of dispersion is achieved through the positive mapping process.

[0038] Using the interference sensitivity of all monitoring parameters as weights, the abnormal deviations of all monitoring parameters between any two piles are weighted and summed to obtain the interference heterogeneity of any two piles. It should be noted that the larger the abnormal deviation, the greater the difference in the distortion caused by interference with the two piles on this monitoring parameter, reflecting that they face different local abnormal interference or different structural response characteristics in this dimension. The greater the interference sensitivity, the more drastic the distortion performance of this monitoring parameter fluctuates and the higher the dispersion among all piles in the world. The more obvious the heterogeneity of the state between piles captured by this monitoring parameter, the more sensitive this monitoring parameter is to local environmental interference, with outstanding identification and discrimination, and can effectively expose local anomalies. It should be given higher weight in the comprehensive evaluation, accurately reflecting the key characteristics of local structural anomalies and highlighting the response differences caused by local interference factors. The greater the interference heterogeneity, the more completely different the response characteristics exhibited by the two piles under the superposition of multiple interferences. The more significant the influence of differences in environmental interference, soil state, and local coupling conditions, the more significantly the two piles are in completely different interference superposition environments or states. Conversely, the more consistent the local geological interference patterns and noise superposition laws are between the two piles.

[0039] Furthermore, in complex geological environments with multiple superimposed disturbances, the types and intensities of disturbances experienced by different piles vary significantly, resulting in drastically different noise distribution characteristics in the data from different piles. In practical engineering, the amount of data from minor, normal disturbances is often extremely large, while the amount of data from severe, sporadic disturbances (such as extreme local resonance or severe settlement) is extremely sparse. If all collected data are directly mixed and fed into a single denoising network model for training, this severe data imbalance will lead to serious bias and overfitting in the neural network. That is, the model will overlearn high-frequency normal noise features while ignoring low-frequency but crucial severe disturbance features for damage assessment, resulting in severe oversmoothing or false filtering, causing the denoising model to fail. To address the problem of uneven distribution of multiple disturbance features in the sample space and ensure that a single denoising model can capture global disturbance patterns without bias, piles are clustered, grouping piles with similar disturbance characteristics into the same cluster, specifically: Using the interference heterogeneity between any two piles as a distance metric, all piles in the historical set are clustered to obtain multiple clusters. In this embodiment, agglomerative hierarchical clustering algorithm is used for clustering. The agglomerative hierarchical clustering algorithm is a well-known technology and will not be described in detail here.

[0040] A sample set is formed by combining all monitoring parameters of all piles within the same cluster. It should be noted that because the number of piles deployed in special / severe geological areas is far less than in conventional areas, there is a severe imbalance in the data size of the sample sets corresponding to different clusters. To avoid bias towards the noise features of large sample sets during model training, data balancing processing is required for the different sample sets, specifically as follows: Perform sample balancing on all sample sets to obtain the balanced sample sets. In this embodiment, a combination of SMOTE oversampling and random undersampling algorithms is used for sample balancing. After balancing, the amount of data in all sample sets tends to be balanced. The combination of SMOTE oversampling and random undersampling algorithms is a well-known technique and will not be described in detail here.

[0041] A theoretical analysis model of the semi-embedded metamaterial pile was pre-established using finite element simulation software, and a preset standard excitation load was applied to it. Multiple monitoring parameters of the pile under undisturbed conditions were output. The multiple monitoring parameters of all piles in the same cluster under undisturbed conditions were used as sample labels. In this embodiment, the establishment process of the theoretical analysis model specifically includes: obtaining the structural design parameters and geological survey parameters of the semi-embedded metamaterial pile, including pile size, concrete strength grade, and soil physical and mechanical parameters obtained from on-site surveys; constructing a finite element simulation model using finite element analysis software, and setting corresponding boundary constraints and interface contact properties in the finite element simulation model to obtain a theoretical analysis model characterizing the ideal stress state of the pile; secondly, eliminating local abnormal interference factors such as interface wave scattering and non-uniform soil settlement in the theoretical analysis model, and setting the preset standard excitation load to apply only the reference environmental static load and standard frequency band dynamic excitation that match the actual service conditions of the pile, thereby obtaining various monitoring parameters of different piles in an interference-free environment.

[0042] The neural network model is trained using all the balanced sample sets and sample labels, and the trained neural network model is used as the denoising network model. In this embodiment, a Convolutional Long Short-Term Memory (CNN-LSTM) fusion model is used for training. The CNN-LSTM fusion model is a well-known technique and will not be described in detail here. The loss function of the CNN-LSTM fusion model is the mean squared error loss function, and the optimizer is the Adam optimizer. The model is trained by dividing all sample sets into training, testing, and validation sets, and calculating the difference between the model's predicted output and the sample labels using the mean squared error loss function during training. This difference is then used to update the network weights of the CNN-LSTM fusion model. The training of the neural network model is also a well-known technique and will not be described in detail here.

[0043] Thus, the noise reduction network model is obtained.

[0044] Step 4: Input various monitoring parameters of the pile to be tested into the noise reduction network model for noise reduction processing, and use the various monitoring parameters after noise reduction to judge and evaluate the health status of the pile to be tested.

[0045] Furthermore, a noise reduction network model is used to denoise the monitoring parameters of the pile under test in order to assess and determine the health status of the pile under test. Specifically: For the various monitoring parameters of the pile body to be tested, a noise reduction network model is used to perform noise reduction processing to obtain the various monitoring parameters after noise reduction. The various monitoring parameters after noise reduction are compared and analyzed with the corresponding preset thresholds to obtain the judgment results of each monitoring parameter; In this embodiment, the maximum values ​​of various monitoring parameters after noise reduction are extracted and compared for analysis, specifically: For strain data, the maximum strain data after noise reduction is obtained and compared with the preset allowable stress of concrete to determine whether the strain of the pile body under test exceeds the allowable stress of concrete, and the result of deformation damage determination is obtained. For the tilt angle, the maximum tilt angle after noise reduction is obtained, and it is compared with the preset stable range to determine whether the tilt angle of the pile under test exceeds the stable range, thus obtaining the tilt degree determination result. For relative displacement, the maximum relative displacement after noise reduction is obtained and compared with the preset design offset to determine whether the motion state of the mass block inside the pile to be tested exceeds the design offset, and the determination result of the relative motion state is obtained. For horizontal and vertical displacements, the maximum horizontal and vertical displacements after noise reduction are obtained and compared with preset displacement limits to determine whether the horizontal and vertical displacements of the pile under test exceed the displacement limits, thus obtaining the displacement status determination result. The preset allowable stress of concrete is set to 0.8 times the design value of concrete axial compressive strength. The preset stability range is set to 0°~1.5°. The preset design offset can be determined by the design parameters of metamaterial structure, the allowable deformation value of rubber connector and the calibration of vibration table test, and is set to 10mm. The preset displacement limit is set to 2% of the pile length. These data are determined by engineering design parameters, on-site calibration test and relevant structural testing specifications. As other implementation methods, implementers can set them according to the actual situation.

[0046] To eliminate the bias of a single judgment result under complex interference and achieve complementary multi-source information and comprehensive decision-making, the Dempster-Shafer Evidence Theory is used to perform decision-level fusion of all the above judgment results, specifically as follows: The judgment results of all monitoring parameters are taken as independent evidence sources. Multiple health levels are set as identification frameworks. The DS evidence theory is used to fuse and calculate all evidence sources and output the health level of the pile to be tested. It should be noted that the various health levels refer to normal state, minor damage state, moderate damage state, severe damage state, and failure state. Secondly, based on the degree of deviation of each monitoring parameter from its corresponding preset threshold, the basic probability allocation value of the monitoring parameter belonging to each health level within the identification framework is calculated using a Gaussian membership function or a trapezoidal membership function. Then, the Dempster combination rule is used for fusion calculation. Among these, the DS evidence theory is a well-known technique and will not be elaborated here.

[0047] Based on the health level of the pile to be tested, a test report is generated or a corresponding early warning mechanism is triggered, realizing comprehensive anti-interference testing and intelligent health status assessment of semi-embedded metamaterial piles in complex environments.

[0048] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0049] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0050] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application, without departing from the content of the technical solution of this application, shall fall within the protection scope of the technical solution of this application.

Claims

1. A method for detecting semi-embedded metamaterial piles based on multi-source data fusion, characterized in that, The method includes the following steps: Multiple monitoring parameters of the pile to be tested are obtained, and all kinds of monitoring parameters of different piles in historical periods are collected to form a historical set; For the historical set, the spatial distribution characteristics of two adjacent piles and their differences under different monitoring parameters are analyzed to characterize the data distortion of the piles under multiple interferences, and the response difference of various monitoring parameters for each pile is calculated. The deviation characteristics and discreteness of the response difference of monitoring parameters between any two piles in the historical set are evaluated. The interference heterogeneity characterizing the difference of the disturbed state of any two piles under multiple interferences is calculated. Based on this, all piles in the historical set are clustered. The various monitoring parameters of all piles in different clusters are sample balanced. The balanced data is used to train the neural network model to obtain the noise reduction network model. Multiple monitoring parameters of the pile to be tested are input into the noise reduction network model for noise reduction processing. The health status of the pile to be tested is judged and evaluated using the noise-reduced monitoring parameters. The calculation of the response variability of various monitoring parameters for each pile includes: For each pile in the historical set, multiple piles adjacent to it are defined as neighboring piles; the data of each monitoring parameter of each pile at all times are used to form various parameter sequences, and the difference between each pile and each neighboring pile for the same parameter sequence is calculated as the relative difference. Based on the spatial distance between each pile and its nearest neighboring piles, the association weight between each pile and its nearest neighboring piles is determined. Based on the association weight, the relative differences of the same monitoring parameter between each pile and all its neighboring piles are weighted and summed to serve as the response difference of each pile for various monitoring parameters. The calculation process for the interference heterogeneity is as follows: Calculate the difference in response to the same monitoring parameter between any two piles, perform a positive mapping on it, and use it as the abnormal deviation amount; A positive mapping is performed on the dispersion of the response differences of all piles under various monitoring parameters, which serves as the interference sensitivity of various monitoring parameters; The interference sensitivity and abnormal deviation are positively fused to determine the interference heterogeneity of any two piles.

2. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 1, characterized in that, The monitoring parameters include strain data, tilt angle, relative displacement, horizontal displacement, and vertical displacement.

3. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 1, characterized in that, The calculation process of the association weight is as follows: negatively map the spatial distance between each pile and each of its nearest neighboring piles, normalize it, and use it as the association weight between each pile and each of its nearest neighboring piles.

4. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 1, characterized in that, The specific process of the forward fusion is as follows: using the interference sensitivity of all kinds of monitoring parameters as weights, the abnormal deviation of all monitoring parameters between any two piles is weighted and summed to obtain the interference heterogeneity of any two piles.

5. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 1, characterized in that, The process of balancing multiple monitoring parameters of all piles within different clusters includes: forming a sample set by combining all monitoring parameters of all piles within the same cluster; and balancing all sample sets to obtain a balanced sample set.

6. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 5, characterized in that, The process of training the neural network model to obtain a noise reduction network model includes: acquiring multiple monitoring parameters of the pile body under an interference-free environment, and using the multiple monitoring parameters of all pile bodies in the same cluster under an interference-free environment as sample labels; using all balanced sample sets and their sample labels to train the neural network model, and using the trained neural network model as the noise reduction network model.

7. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 6, characterized in that, The process of obtaining various monitoring parameters of the pile body under an interference-free environment is as follows: a theoretical analysis model of the semi-embedded metamaterial pile body is established in advance using finite element simulation software, and a preset standard excitation load is applied to it to output various monitoring parameters of the pile body under an interference-free environment.

8. The method for detecting semi-embedded metamaterial piles based on multi-source data fusion as described in claim 1, characterized in that, The process of judging and evaluating the health status of the pile to be tested includes: comparing and analyzing various noise-reduced monitoring parameters with corresponding preset thresholds to obtain the judgment results of each monitoring parameter; taking the judgment results of all monitoring parameters as independent sources of evidence, setting multiple health levels as identification frameworks, using DS evidence theory to fuse and calculate all sources of evidence, and outputting the health level to which the pile to be tested belongs.

Images