Foundation pile low strain detection signal automatic classification method and system

Abstract

The invention discloses a foundation pile low strain detection signal automatic classification method and system, and relates to the technical field of foundation pile detection. The method comprises the following steps: acquiring a low-strain signal of the foundation pile; a low-strain detection technology is adopted to collect a low-strain signal of the foundation pile; the low-strain signals of the foundation piles are preprocessed; and the abnormal value is removed according to the standard deviation, and random noise in the low-strain signals of the foundation piles is reduced. And the extracted low-strain signals of the foundation pile are smoothed through a moving average filter. Extracting low-strain signal characteristics of the foundation pile; extracting preprocessed signal features including kurtosis, waveform indexes, pulse indexes and the like; an LSS-CM classification model is constructed; different weights are distributed to the characteristics of the preprocessed low-strain signals of the foundation piles, and an optimal classification threshold value is determined through a self-adaptive threshold value method. Outputting a classification result; and inputting the low-strain signals of the foundation piles into the LSS-CM classification model to obtain a final classification result. The method has high classification efficiency and accuracy, and can effectively classify unknown foundation pile low-strain signals.

Description

Technical Field

[0001] The present invention relates to the technical field of pile foundation detection, and in particular to an automatic classification method and system for pile foundation low-strain detection signals. Background Art

[0002] With the large-scale development of infrastructure construction, pile foundations are widely used in various construction projects, such as high-rise buildings, bridges, ports, etc. The quality of pile foundations is directly related to the safety and stability of the entire engineering structure. As a common and important pile foundation integrity detection method, low-strain detection of pile foundations plays a key role in engineering practice.

[0003] The basic principle of low-strain detection of pile foundation is to apply a transient impact force to the top of the pile to generate stress waves in the pile body, and use the sensor installed on the top of the pile to receive the signal reflected by the pile body. By analyzing the reflected wave signal, the integrity of the pile body, the location and type of defects, and the length of the pile body can be inferred. However, the analysis of traditional low-strain detection results of pile foundation mainly relies on the professional experience and manual classification of the inspectors.

[0004] There are many limitations in the manual classification of low-strain detection signals of pile foundations. First, the inspectors need to have deep professional knowledge and rich practical experience to accurately interpret the complex reflected wave signals. Different inspectors may have different understandings and judgments of the low-strain signals of the same pile foundation, which leads to subjectivity and uncertainty in the test results. Secondly, the manual analysis process is cumbersome and time-consuming, requiring careful observation, measurement and calculation of a large amount of data and waveforms. Especially when facing large-scale pile foundation inspection projects, inefficiency becomes a prominent problem. Furthermore, manual classification is easily affected by factors such as fatigue and emotions, and may result in misjudgment or omission, thus affecting the accuracy of engineering quality assessment. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for automatically classifying low-strain detection signals of pile foundations with high efficiency and high accuracy.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for automatically classifying low-strain detection signals of pile foundations, comprising the following steps:

[0007] S1, low-strain signal acquisition of foundation piles: adopt low-strain detection method, use low-energy transient to excite at the top of the pile, collect the velocity time history curve of the top of the measured pile, and save it in data point format;

[0008] S2, pre-processing the low strain signal of the pile foundation: the collected low strain signal of the pile foundation is processed according to the standard deviation σ sThe abnormal values ​​are corrected to remove the noise interference in the low strain signal of the pile foundation, and the low strain signal of the pile foundation is smoothed by using a sliding average filter to obtain the pre-processed low strain signal of the pile foundation;

[0009] S3, feature extraction: extracting characteristic parameters related to pile integrity from the pre-processed low-strain signal of the pile;

[0010] S4, constructing LSS-CM classification model: constructing the classification model LSS-CM, assigning weights to the extracted low-strain signal features of the pile foundation, batch analyzing the data, and using the adaptive threshold method to determine the optimal classification threshold;

[0011] S5, output classification results: after the low strain signal of the pile to be detected is processed through steps S1-S3, it is input into the constructed LSS-CM classification model, and the integrity evaluation of the pile is given according to the integrity status of the low strain signal of the pile.

[0012] Correspondingly, the present invention also discloses an automatic classification system for low-strain detection signals of pile foundations, comprising:

[0013] Pile foundation low strain signal acquisition module: used to adopt low strain detection method, use low energy transient to excite at the pile top, collect the velocity time history curve of the measured pile top, and save it in data point format;

[0014] Pile foundation low strain signal preprocessing module: used to collect the pile foundation low strain signal according to the standard deviation σ s The abnormal values ​​are corrected to remove the noise interference in the low strain signal of the pile foundation, and the low strain signal of the pile foundation is smoothed by using a sliding average filter to obtain the pre-processed low strain signal of the pile foundation;

[0015] Feature extraction module: extracts characteristic parameters related to pile integrity from the pre-processed low-strain signal of the pile;

[0016] LSS-CM classification model building module: used to build the classification model LSS-CM, assign weights to the extracted low-strain signal features of the pile foundation, batch analyze the data, and use the adaptive threshold method to determine the optimal classification threshold;

[0017] Classification result output module: After the low strain signal of the pile to be detected is processed through steps S1-S3, it is input into the constructed LSS-CM classification model, and the integrity evaluation of the pile is given according to the integrity status of the low strain signal of the pile.

[0018] The beneficial effect of adopting the above technical solution is that the method and system utilize advanced signal processing methods and classification models to automatically and efficiently and accurately classify the low-strain detection results of pile foundations, thereby reducing the subjectivity and uncertainty of manual classification, improving detection efficiency, and ensuring the reliability of pile foundation quality detection, which helps to improve the quality and safety of the entire pile foundation project and even related construction projects, while reducing labor costs and time costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0020] Figure 1 is a main flow chart of the method described in the embodiment of the present invention;

[0021] Figure 2 is a signal diagram of the method according to an embodiment of the present invention, the foundation category of which is 1;

[0022] Figure 3 is a classification result diagram of the pile category 1 of the method according to the embodiment of the present invention;

[0023] Figure 4 is a signal diagram of the base type 2 of the method according to an embodiment of the present invention;

[0024] Figure 5 is a classification result diagram of pile category 2 according to the method of an embodiment of the present invention;

[0025] Figure 6 is a signal diagram of the base type 3 of the method according to an embodiment of the present invention;

[0026] Figure 7 is a classification result diagram of pile category 3 according to the method of an embodiment of the present invention;

[0027] Figure 8 is a signal diagram of the base type 4 of the method according to an embodiment of the present invention;

[0028] Fig. 9 is a classification result diagram of pile category 4 according to the method of an embodiment of the present invention;

[0029] Fig.10 It is a principle block diagram of the system described in the embodiment of the present invention. DETAILED DESCRIPTION

[0030] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] like Figure 1 As shown, an embodiment of the present invention discloses a method for automatically classifying low-strain detection signals of pile foundations, comprising the following steps:

[0033] S1 pile low strain signal acquisition: adopt low strain detection technology, use low energy transient to excite at the pile top, collect the velocity time history curve at the top of the measured pile, and save it in data point format.

[0034] S2 pre-processes the low strain signal of the pile foundation: the collected low strain signal of the pile foundation is processed according to the standard deviation σ s The outliers are corrected to remove the noise interference in the low strain signal of the pile foundation, and the low strain signal of the pile foundation is smoothed by using a sliding average filter to obtain the preprocessed low strain signal of the pile foundation.

[0035] S3 feature extraction: Extract characteristic parameters related to pile integrity from the preprocessed low-strain signal of the pile, including the sample kurtosis K, waveform index CI, pulse index PI, short-time average amplitude A of the low-strain signal of the pile STA And other features.

[0036] S4 constructs LSS-CM classification model: constructs the classification model LSS-CM (Low Strain Signal-Classification Model), assigns weights to the extracted low strain signal features of the pile foundation, batch analyzes the data, and uses the adaptive threshold method to determine the optimal classification threshold.

[0037] S5 outputs the classification result: after the low strain signal of the pile to be detected has been subjected to the above-mentioned preprocessing and feature extraction steps, it is input into the constructed classification model (Low Strain Signal-Classification Model), and the integrity evaluation of the pile is given according to the integrity status of the low strain signal of the pile.

[0038] Further, the low-strain signal acquisition of the basic pile in step S1 further includes:

[0039] Apply a dynamic force F(t) to the pile top. The pile-soil system generates a dynamic response under the action of the dynamic force. Assume the wave impedance of the pile shaft is Z1 and the wave impedance of the pile bottom is Z2. According to the wave theory, when the stress wave propagates at the interface of different wave impedances, the reflection coefficient

[0040] When the wave impedance of the pile bottom is weaker than that of the pile shaft, i.e., Z2 < Z1, the reflection coefficient R < 0 at this time, and multiple reflections occur at the pile bottom position. The reflected wave at the pile bottom is in the same direction as the incident wave.

[0041] When the wave impedance of the pile bottom is stronger than that of the pile shaft, i.e., Z2 > Z1, the reflection coefficient R > 0, and multiple reflections occur at the pile bottom position. The odd-numbered reflected waves are in the opposite direction to the incident wave, and the even-numbered reflected waves are in the same direction as the incident wave. Moreover, the greater the change in the wave impedance between the pile bottom and the pile shaft, i.e., the greater |Z2 - Z1|, the greater the absolute value of the reflection coefficient formula, and the more obvious the reflected wave at the pile bottom.

[0042] For the same basic pile, the four waveform curves formed by four hammer strikes should be basically the same in shape, amplitude, and phase, and the data collection is considered qualified. In this method, one of the waveform curves formed by a hammer strike is used for judgment.

[0043] Further, the preprocessing of the low-strain signal of the basic pile in S2 specifically includes the following steps:

[0044] Let the acquired low-strain signal sequence of the basic pile be s(t i ), i = 1, 2, …, N, where N = 1024 in this method;

[0045] Calculate the mean value of the acquired low-strain signal of the basic pile. The mean value calculation formula is:

[0046]

[0047] In the above formula, s(t i ) represents the value of the i-th sampling point in the low-strain signal sequence of the basic pile, and N is the total number of sampling points;

[0048] Calculate the standard deviation σ s of the acquired low-strain signal of the basic pile. The standard deviation σ s calculation formula is:

[0049]

[0050] Set the outlier judgment threshold to 3σ s . If then s(t i) is marked as an outlier and corrected by linear interpolation of adjacent normal points, that is, s(t i ) If it is an abnormal value, the correction value is:

[0051]

[0052] The sliding average filter is used to smooth the low strain signal of the pile foundation, and the window length L of the sliding average filter is determined according to the frequency characteristics and noise level of the low strain signal of the pile foundation.

[0053] Determine the main frequency component f of the low strain signal of the pile dominant , then according to the sampling frequency f s Calculate the time resolution of the low strain signal of the pile foundation using the following formula:

[0054] Δt=1 / f s ,

[0055] The window length L is generally taken as a value related to the period of the main frequency component. In this method, L=10.

[0056] Smoothed low strain signal of pile s smoOth (n) The calculation formula is:

[0057]

[0058] In the above formula: Indicates rounding down.

[0059] For the boundary points at both ends of the pile low strain signal sequence, the boundary processing method of mirror symmetric expansion is adopted, that is, for the starting point t1, t0=2t1-t2 is added, and for the end point t n , supplement t n+1 =2t n -t n-1 , and then perform sliding average calculation.

[0060] Furthermore, the S3 feature extraction specifically includes the following steps:

[0061] The characteristic parameters related to the pile integrity are extracted from the preprocessed low strain signal s(n) of the pile.

[0062] Get the kurtosis of the low strain signal s(n) of the pile: The calculation formula of the kurtosis K is:

[0063]

[0064] In the above formula, μ s is the mean value of the low strain signal s(n) of the pile.

[0065] Obtain the waveform index of the pile low strain signal s(n): The waveform index CI calculation formula is:

[0066]

[0067] In the above formula,

[0068] Obtain the pulse index of the low strain signal s(n) of the pile foundation: The pulse index PI calculation formula is:

[0069]

[0070] In the above formula It is the maximum value in the low strain signal sequence of the pile.

[0071] Obtain the short-time average amplitude of the low strain signal s(n) of the pile: First, determine the short-time analysis window length L STA , which is determined according to the fluctuation period characteristics of the low strain signal of the pile. In this method, L STA =256. Then calculate the short-time average amplitude A STA :

[0072]

[0073] In the above formula, j is the sampling point number corresponding to the current analysis moment, and a boundary processing method similar to the sliding average filter is used for the boundary points.

[0074] The extracted features of the low-strain signal of the pile are marked as feature vectors. For the k-th low-strain signal of the pile (k = 1, 2, ..., N), its feature vector is X k =[X k1 ,X k2 ,…,X kn ], where n represents the number of features extracted from each signal.

[0075] Furthermore, the S4 constructing the LSS-CM classification model specifically includes the following steps:

[0076] Different weights are assigned according to the importance of the low strain signal characteristics of the pile foundation: the weight vector of each feature is determined by the hierarchical analysis method:

[0077] W=[w1,w2,…,w n ],

[0078] Construct judgment matrix A = (a ij ), where the judgment matrix element a ij Represents the importance ratio of the i-th feature to the j-th feature, and satisfies

[0079] Calculate the maximum eigenvalue λ of the matrix max and its corresponding eigenvector V = [v1,v2,…,v p ].

[0080] The feature vector is normalized to obtain the weight vector W, where:

[0081]

[0082] Through consistency indicators and consistency ratio Among them, RI is the random consistency index, and a consistency test is performed. When CR<0.1, the weight vector is considered reasonable and valid.

[0083] The extracted k-th pile low strain signal feature vector X k Multiply it with the weight vector W to get the intermediate vector Y k , which is calculated as:

[0084] Y kj =X kj ×W j ,(j=1,2,…,n),

[0085] Then, for the intermediate vector Y k Sum all the elements of to get a comprehensive eigenvalue Z k , and its calculation formula is:

[0086]

[0087] Adaptive threshold method is used to determine the optimal classification threshold. Let the initial threshold vector be T0 = [t1, t2, t3, t4], these thresholds will be used to divide different category intervals.

[0088] Use the validation set data to calculate the classification accuracy under the current threshold. For the jth set of validation data, the model output is y j , according to the threshold vector T, the classification is determined to obtain the predicted category Compare with the true category. The classification accuracy calculation formula is:

[0089]

[0090] In the above formula, I(·) is an indicator function that returns 1 when the condition in the brackets is met, otherwise it returns 0.

[0091] The grid search method is used to search for each threshold component within a certain range with a fixed step size. Assume that the threshold adjustment step size vector is ΔT = [Δt1, Δt2, Δt3, Δt4]. In each iteration:

[0092] T k+1 =T k +ΔT k

[0093] In the above formula, ΔT k It is the step size vector at the kth iteration, which is dynamically adjusted according to the iteration situation.

[0094] For each new set of threshold vectors T k+1 , recalculate the classification accuracy Accuracy k+1 Repeat this process until the stopping condition is met, such as the accuracy no longer improves or the maximum number of iterations is reached. The threshold vector T that gives the highest classification accuracy * is the optimal classification threshold vector.

[0095] Finally, when the calculated Z k When the score is less than t1, it is determined to be category 1; when the score is greater than t1 and less than t2, it is determined to be category 2; when the score is greater than t2 and less than t3, it is determined to be category 3. When the score is greater than t4, it is determined to be category 4.

[0096] The automatic classification method for pile foundation low strain detection signals, wherein: step S5 outputting the classification results further comprises:

[0097] S51, after the preprocessing and feature extraction steps, the low strain signal of the pile to be detected is input into the constructed LSS-CM classification model, and the integrity evaluation of the pile is given according to the integrity state of the low strain signal of the pile.

[0098] S52, for the integrity evaluation of LSS-CM model classification type 1, the detection wave waveform has no abnormal reflection, the wave velocity is normal, the pile body is intact, and it belongs to a complete pile (related waveforms such as Figure 2-Figure 3 shown).

[0099] S53, for the integrity evaluation of LSS-CM model classification type 2, the detection wave waveform has small distortion, the wave velocity is basically normal, the pile body has slight defects, and there is no impact on the use of the pile. It is a basically complete pile (related waveforms such as Figure 4-Figure 5 shown).

[0100] S54, for the integrity evaluation of LSS-CM model classification type 3, the detection wave waveform has abnormal reflection, the wave velocity is low, the pile body has obvious defects, and it has a certain impact on the use of the pile. It is an obvious defective pile (related waveforms such as Figure 6-Figure 7 shown).

[0101] S55, for the integrity evaluation of LSS-CM model classification type 4, the detection wave waveform is severely distorted, the pile body has serious defects or fractures, and it belongs to serious defective piles or broken piles (related waveforms such as Figure 8-Figure 9shown).

[0102] Corresponding to the method, Fig.10 As shown, the embodiment of the present invention also discloses an automatic classification system for pile foundation low strain detection signals, comprising:

[0103] The low-strain signal acquisition module 101 of the foundation pile is used to adopt a low-strain detection method, use a low-energy transient to excite the pile top, collect the velocity time history curve of the measured pile top, and save it in a data point format;

[0104] The pile low strain signal preprocessing module 102 is used to process the collected pile low strain signal according to the standard deviation σ s The abnormal values ​​are corrected to remove the noise interference in the low strain signal of the pile foundation, and the low strain signal of the pile foundation is smoothed by using a sliding average filter to obtain the pre-processed low strain signal of the pile foundation;

[0105] Feature extraction module 103: extracting characteristic parameters related to pile integrity from the pre-processed pile low strain signal;

[0106] LSS-CM classification model building module 104: used to build a classification model LSS-CM, assign weights to the extracted low-strain signal features of the pile foundation, batch analyze the data, and use an adaptive threshold method to determine the optimal classification threshold;

[0107] Classification result output module 105: after the low strain signal of the pile to be detected is processed in steps S1-S3, it is input into the constructed LSS-CM classification model, and the integrity evaluation of the pile is given according to the integrity state of the low strain signal of the pile.

[0108] It should be noted that the implementation method of the relevant modules in the system described in the present application can refer to the specific steps in the aforementioned method and will not be repeated here.

[0109] The system and method can quickly and accurately extract key feature information from the reflected wave signal, and compare and match it with a preset pile integrity classification standard or model, thereby realizing automatic classification and evaluation of pile integrity, providing a more reliable and efficient pile quality detection method for engineering construction, helping to ensure the overall quality and safety performance of construction projects, and promoting the development of pile detection technology towards intelligence and automation.

Claims

1. A method for automatically classifying low-strain detection signals of pile foundations, characterized in that The steps include: S1, low-strain signal acquisition of foundation piles: adopt low-strain detection method, use low-energy transient to excite at the top of the pile, collect the velocity time history curve of the top of the measured pile, and save it in data point format; S2, pre-processing the low strain signal of the pile foundation: the collected low strain signal of the pile foundation is processed according to the standard deviation σ s The abnormal values ​​are corrected to remove the noise interference in the low strain signal of the pile foundation, and the low strain signal of the pile foundation is smoothed by using a sliding average filter to obtain the pre-processed low strain signal of the pile foundation; S3, feature extraction: extracting characteristic parameters related to pile integrity from the pre-processed low-strain signal of the pile; S4, constructing LSS-CM classification model: constructing the classification model LSS-CM, assigning weights to the extracted low-strain signal features of the pile foundation, batch analyzing the data, and using the adaptive threshold method to determine the optimal classification threshold; S5, output classification results: after the low strain signal of the pile to be detected is processed through steps S1-S3, it is input into the constructed LSS-CM classification model, and the integrity evaluation of the pile is given according to the integrity status of the low strain signal of the pile.

2. The automatic classification method for pile foundation low strain detection signals according to claim 1, characterized in that: The S1 pile low strain signal acquisition includes the following steps: A dynamic force F(t) is applied to the pile top. The pile-soil system produces a dynamic response under the action of the dynamic force. Assuming the wave impedance of the pile body is Z1 and the wave impedance of the pile bottom is Z2, according to the wave theory, when the stress wave propagates at different wave impedance interfaces, the reflection coefficient 3. The automatic classification method for pile foundation low strain detection signals according to claim 1, characterized in that: The method for pre-treating the pile foundation low strain in S2 comprises the following steps: First, calculate the standard deviation σ of the low strain signal of the collected pile s , assuming that the collected low strain signal sequence of the pile is s(t i ),i=1,2,…,, the mean calculation formula is: Where s(t i ) represents the value of the i-th sampling point in the pile low strain signal sequence, and N is the total number of sampling points; Standard Deviation Among them, the outlier judgment threshold is set to 3σ s ,like Then s(t i ) is marked as an outlier and corrected by linear interpolation of adjacent normal points, that is, s(t i ) If it is an abnormal value, correct the value 4. The automatic classification method for pile foundation low strain detection signals according to claim 1, characterized in that: In S2: The sliding average filter is used to smooth the low strain signal of the pile foundation. The window length L of the sliding average filter is determined according to the frequency characteristics and noise level of the signal. The main frequency component f of the low strain signal of the pile foundation is determined. dominant , then according to the sampling frequency f s The time resolution of calculating the low strain signal of the pile is Δt=1 / f s The window length L is generally taken as a value related to the period of the main frequency component. In this method, L = 10. The smoothed low strain signal of the pile s smooth (n) The calculation formula is: in Indicates rounding down; for the boundary points at both ends of the pile low strain signal sequence, the boundary processing method of mirror symmetric expansion is adopted, that is, for the starting point t1, t0 = 2t1-t2 is added, and for the end point t n , supplement t n+1 =2t n -t n-1 , and then perform sliding average calculation.

5. The automatic classification method for pile foundation low strain detection signals according to claim 1, characterized in that: The S3 extracts features of the low strain signal of the pile foundation including the following steps: Extract characteristic parameters related to pile integrity from the preprocessed pile low strain signal s(n), including: Kurtosis: The calculation formula of Kurtosis K is: where μ s is the mean value of the low strain signal s(n) of the pile; Waveform indicator: The calculation formula of the waveform indicator CI is: in Pulse index: The calculation formula of pulse index PI is: in It is the maximum value in the low strain signal sequence of the pile; Short-time average amplitude: First determine the short-time analysis window length L STA , the length is determined according to the fluctuation period characteristics of the pile signal, and then the short-time average amplitude is calculated Where j is the sampling point number corresponding to the current analysis moment, and a boundary processing method similar to the sliding average filter is used for the boundary points; The features of the extracted pile low strain signal are marked as feature vectors. For the kth pile low strain signal, k = 1, 2, ..., N, its feature vector is X k =[X k1 , X k2 , …, X kn ], where n represents the number of features extracted from each signal.

6. The automatic classification method for pile foundation low strain detection signals according to claim 1, characterized in that: In step S4, the method for constructing the LSS-CM classification model includes the following steps: Different weights are assigned according to the importance of the low strain signal characteristics of the pile foundation: the weight vector W = [w1, w2, ..., w n ], construct the judgment matrix A = (a ij ), where the judgment matrix element a ij Represents the importance ratio of the i-th feature to the j-th feature, and satisfies Calculate the maximum eigenvalue λ of the matrix max and its corresponding eigenvector V = [v1, v2, …, v p ], the feature vector is normalized to obtain the weight vector W, that is, And through the consistency index and consistency ratio A consistency check is performed. When CR < 0.1, the weight vector is considered reasonable and valid, where RI is the random consistency index; The extracted k-th pile low strain signal feature vector X k Multiply it with the weight vector W to get the intermediate vector Y k , which is calculated as follows: kj =X kj ×W j , (j=1, 2, ..., n), then, for the intermediate vector Y k Sum all the elements of to get a comprehensive eigenvalue Z k , and its calculation formula is: Adaptive threshold method is used to determine the optimal classification threshold. The initial threshold vector is set to T0 = [t1, t2, t3, t4]. The threshold is used to divide different category intervals. Use the validation set data to calculate the classification accuracy under the current threshold. For the jth set of validation data, the model output is y j , according to the threshold vector T, the classification is determined to obtain the predicted category Compared with the true category, the classification accuracy calculation formula is: Where I(·) is an indicator function that returns 1 when the condition in the brackets is true, otherwise it returns 0; The grid search method is used to search for each threshold component within a certain range with a fixed step size. The threshold adjustment step size vector is set to ΔT = [Δt1, Δt2, Δt3, Δt4]. In each iteration: T k+1 =T k +ΔT k ; Where ΔT k is the step vector at the kth iteration, which is dynamically adjusted according to the iteration situation; For each new set of threshold vectors T k+1 , recalculate the classification accuracy Accuracy k+1 Repeat this process until the stopping condition is met, such as the accuracy no longer improves or the maximum number of iterations is reached, so that the threshold vector T with the highest classification accuracy is * That is the optimal classification threshold vector; When calculating Z k When the score is less than t1, it is judged as category 1; when the score is greater than t1 and less than t2, it is judged as category 2; when the score is greater than t2 and less than t3, it is judged as category 3; when the score is greater than t4, it is judged as category 4.

7. The automatic classification method for pile foundation low strain detection signals according to claim 1, characterized in that: The method for outputting the classification result in S5 comprises the following steps: For the LSS-CM classification model with a classification type of 1, the integrity evaluation is that the detection wave waveform has no abnormal reflection, the wave velocity is normal, and the pile body is intact, which is a complete pile; For the LSS-CM classification model with classification type 2, the integrity evaluation is that the detection wave waveform has small distortion, the wave velocity is basically normal, the pile body has slight defects, and there is no impact on the use of the pile, which is a basically complete pile; For the LSS-CM classification model classification type 3, the integrity evaluation is that the detection wave waveform has abnormal reflection, the wave velocity is low, the pile body has obvious defects, and there is a certain impact on the use of the pile, which is an obvious defect pile; For the LSS-CM classification model classification type 4, the integrity evaluation is that the detection wave waveform is severely distorted and the pile body has serious defects or fractures, which are serious defective piles or broken piles.

8. An automatic classification system for pile foundation low strain detection signals, characterized in that include: Pile foundation low strain signal acquisition module: used to adopt low strain detection method, use low energy transient to excite at the pile top, collect the velocity time history curve of the measured pile top, and save it in data point format; Pile foundation low strain signal preprocessing module: used to collect the pile foundation low strain signal according to the standard deviation σ s The abnormal values ​​are corrected to remove the noise interference in the low strain signal of the pile foundation, and the low strain signal of the pile foundation is smoothed by using a sliding average filter to obtain the pre-processed low strain signal of the pile foundation; Feature extraction module: extracts characteristic parameters related to pile integrity from the pre-processed low-strain signal of the pile; LSS-CM classification model building module: used to build the classification model LSS-CM, assign weights to the extracted low-strain signal features of the pile foundation, batch analyze the data, and use the adaptive threshold method to determine the optimal classification threshold; Classification result output module: After the low strain signal of the pile to be detected is processed through steps S1-S3, it is input into the constructed LSS-CM classification model, and the integrity evaluation of the pile is given according to the integrity status of the low strain signal of the pile.

Images