A method for identifying pile foundation defects based on matching low-strain curves with characteristic curves

Abstract

The present invention discloses a method for identifying pile foundation defects based on feature curve matching of low-strain curves, including: Step 1: Generate corresponding defect feature curves for each pile foundation defect type; Step 2: Convert the defect feature curves of each different pile foundation defect type into corresponding text files; Step 3: Slide and intercept the data on the actual low-strain curve, calculate the matching accuracy between the intercepted data and the defect feature curve to construct a matching accuracy curve, and find the maximum value of the matching accuracy exceeding the preset threshold in advance on the matching accuracy curve; Step 4: Calculate the classification probability of the defect type on the actual low-strain defect according to the matching accuracy; Step 5: Determine the level of the defect type. The present invention can realize the automatic identification of pile foundation defects in low-strain curves, greatly improve the efficiency and accuracy of pile foundation detection, and reduce the probability of misjudgment and missed judgment of pile foundation defects caused by manual judgment of low-strain curves.

Description

Technical Field

[0001] The present invention relates to the technical field of pile foundation defect identification, and specifically, to a pile foundation defect identification method based on feature curve matching of low strain curves. Background Art

[0002] Pile foundations (including prestressed pipe piles, cast-in-place piles, steel piles, etc.) are one of the foundation forms widely used in the field of engineering construction. Since pile foundations are important engineering structures with concealed characteristics, once the quality control is not strict during the construction process and the defects cannot be identified in time during the construction or building process, serious engineering hazards will be caused. Among the existing methods for pile foundation defect identification or detection, they mainly include high strain, low strain, core drilling and other detection methods. Among these existing methods, each identification method has its own advantages and disadvantages. Among them, since the low strain detection most commonly used is the non-destructive detection of pile foundations, and its operation process is simple and fast, therefore, the low strain detection has become an increasingly important means for pile foundation defect identification.

[0003] At present, the pile foundation defect identification based on the low strain method often completely or mostly relies on manual identification. To judge the defect type, size and location of the pile foundation, etc. all need to be completed by manual work, which is time-consuming and laborious, and is prone to missed detection (high missed detection rate) and false detection. There are also some methods for realizing pile foundation defect identification through neural networks, but there are also corresponding deficiencies. For example, the Chinese invention application with the publication number CN112418266A discloses a pile foundation integrity classification and identification method based on a convolutional neural network. In the method disclosed in this invention application, it is necessary to draw the low strain curve into an image and compress it into a picture with a size of 64 pixels × 64 pixels, and then use a deep learning neural network to learn the picture, so as to classify the pile foundation defects. In the process of compressing the image, the information of the low strain curve will be lost, especially for pile foundations with a long pile length, such as when the pile length is greater than 30m. Since the defect features on the curve only occupy a very small part of the range, in the process of compressing the image, the defect information on the low strain curve may be lost. In addition, when drawing the low strain curve, the existence of the coordinate axis and grid lines will also interfere with the convolutional calculation. The above reasons will all make the accuracy of using the convolutional neural network for the low strain curve of the pile foundation relatively low. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a pile foundation defect identification method based on feature curve matching of low strain curves, which can solve the problems described in the background art.

[0005] The technical solution for realizing the purpose of the present invention is: a pile foundation defect identification method based on feature curve matching of low strain curves, including the following steps:

[0006] Step 1: Generate corresponding defect characteristic curves for each type of pile foundation defect, and the defect characteristic curves form a set of defect characteristic curves;

[0007] Step 2: Generate a sliding window with the same length as the defect characteristic curve, move the sliding window on the actual low-strain curve according to a preset sliding step, intercept data from the actual low-strain curve, and use the distance of the sliding window from the origin on the actual low-strain curve as the independent variable and the matching accuracy between the intercepted data and the defect characteristic curve as the dependent variable to form a matching accuracy curve. Extract the upper peak value from the matching accuracy curve, and use the upper peak value greater than the preset threshold as the maximum matching accuracy value of the current matching accuracy curve. The maximum matching accuracy value characterizes the degree to which the current actual low-strain curve has the defect type reflected by the defect characteristic curve at the upper peak position.

[0008] Among them, the matching accuracy between the defect characteristic curve and the data intercepted by the sliding window on the actual low-strain curve is calculated according to formula ①:

[0009]

[0010] In the formula, C i represents the matching accuracy between the defect characteristic curve corresponding to the i-th type of defect and the data intercepted by the sliding window on the actual low-strain curve, f i represents the defect characteristic curve corresponding to the i-th type of defect, g i represents the data intercepted by the sliding window on the actual low-strain curve, and M is the data length of the defect characteristic curve, that is, the number of data points on the defect characteristic curve.

[0011] Step 3: Calculate the classification probability of the defect type on the actual low-strain defect according to the matching accuracy.

[0012] Furthermore, in Step 1, the time length of the defect characteristic curve is taken as 0.01 ms - 1 ms, and the number of discrete data points on the defect characteristic curve is between 50 and 1000.

[0013] Furthermore, the pile foundation defect types include local necking, local bulging, and bending.

[0014] Furthermore, the preset sliding step of the sliding window is 1 - 10.

[0015] Furthermore, after Step 1 and before Step 2, the following steps are also included:

[0016] Convert the defect characteristic curves of each different pile foundation defect type into corresponding text files, denoted as defect characteristic texts, and intercept data on the defect characteristic texts.

[0017] Further, the specific implementation of step 3 includes the following steps:

[0018] Step 4-1: Take the matching accuracy between all the defect feature curves of the same defect type in the defect feature curve set and the actual low-strain curve of the corresponding defect type as the matching accuracy of this defect type, so as to obtain the matching accuracy of the actual low-strain curve in each defect type.

[0019] Among them, the j-th matching accuracy of the actual low-strain curve under the i-th defect type is denoted as C i,j , therefore, the set of all matching accuracies of the actual low-strain curve under the i-th defect type is [C i,1 , C i,2 , …, C i,j , … C i,Q , Q is the total number of maximum values of the matching accuracy, that is, the total number of upper peaks exceeding the preset value.

[0020] Calculate the classification probability P i of the defect type in the actual low-strain curve being classified as the i-th defect type A i,j according to formula ②:

[0021]

[0022] In the formula, e represents the natural logarithm base.

[0023] Step 4-2: Take the maximum value in the classification probability set [P i,1 , P i,2 , …, P i,Q as the final probability that the defect type in the actual low-strain curve is classified as the i-th defect type A i .

[0024] Further, after step 3, it also includes determining the grade of the defect type.

[0025] Further, the specific implementation of determining the grade of the defect type includes:

[0026] Scale the sampling frequency and amplitude of the defect feature curves of each defect type to obtain the scaled defect feature curves, calculate the matching accuracy between the scaled defect feature curves and the actual low-strain curve, and output the corresponding defect grade according to the size of the matching accuracy.

[0027] Further, for defect type A iScale the sampling frequency and amplitude of the defect characteristic curve, then recalculate the classification probability of the actual low-strain curve in each scaled defect characteristic curve according to the above method, and gradually scale the amplitude until the classification probability of the defect type in the actual low-strain defect is maximized, and determine the scaling value of the current amplitude. Then, stratify the defect degree into several levels according to the scaling size ranking.

[0028] Further, the scaling range of the sampling frequency is 0 - 10.

[0029] The beneficial effects of the present invention are as follows: The present invention can realize the automatic identification of pile foundation defects in low-strain curves, greatly improve the efficiency and accuracy of pile foundation detection, and reduce the probability of misjudgment and missed judgment of pile foundation defects caused by manual judgment of low-strain curves. Description of the Drawings

[0030] Figure 1 is a schematic flow chart of the method of the present invention;

[0031] Figure 2 are the defect characteristic curves of the pile break defect type and the enlarged diameter defect type;

[0032] Figure 3 is a schematic diagram of the matching accuracy curve calculated by the sliding step length of the defect characteristic curve on the actual low-strain curve;

[0033] Figure 4 is a schematic diagram of the position of the defect type located according to the position of the maximum matching accuracy. Specific Embodiment

[0034] Next, in combination with the drawings and specific embodiments, the present invention will be further described:

[0035] As Figures 1-4 shown, a method for identifying pile foundation defects based on feature curve matching of low-strain curves includes the following steps:

[0036] Step 1: Generate corresponding defect characteristic curves for each pile foundation defect type. The pile foundation defect types include local necking, local enlargement, and bending, etc. Each defect characteristic curve constitutes a defect characteristic curve set. Among them, the defect type set is denoted as {A1, A2, …, A i , … A N}, where A i represents the i-th defect type.

[0037] In this step, based on the pile foundation low-strain theory, artificial generation is used to generate corresponding defect characteristic curves for different pile foundation defect types, that is, each pile foundation defect corresponds to a defect characteristic curve, and this defect characteristic curve is a low-strain curve.

[0038] Among them, the time length of the defect characteristic curve can be taken within the range of 0.01 ms - 1 ms (millisecond), and the number of discrete data points on the defect characteristic curve is between 50 and 1000.

[0039] According to the low-strain theory of pile foundations, when there are certain types of defects in the pile foundation (such as local necking, local enlargement), the reflected waves measured at the pile top will exhibit corresponding characteristics. For example, when there is local necking in the pile foundation, a sine curve with an initial phase of 0 will appear on the defect characteristic curve; when there is local enlargement in the pile foundation, a sine curve with an initial phase of π will appear on the defect characteristic curve.

[0040] Step 2: Convert the defect characteristic curves of each different pile foundation defect type into corresponding text files, and record them as defect characteristic texts.

[0041] Step 3: Generate a sliding window with the same length as the defect characteristic curve, and the sliding step size is 1 - 10. Move the sliding window on the actual low-strain curve according to the sliding step size, and intercept data from the actual low-strain curve, that is, corresponding data can be directly intercepted on the defect characteristic text. And take the distance of the sliding window from the origin on the actual low-strain curve as the independent variable and the matching accuracy between the intercepted data and the defect characteristic curve as the dependent variable to form a matching accuracy curve. The matching accuracy is also the matching degree between the actual low-strain curve and the defect characteristic curve, and the value on the matching accuracy curve is the matching accuracy. Extract the position and value of the upper peak value (i.e., the peak-to-peak value) from the matching accuracy curve, and take the upper peak value greater than the preset threshold as the matching accuracy maximum value of the current matching accuracy curve. The matching accuracy maximum value characterizes the degree to which the current actual low-strain curve has the defect type reflected by the defect characteristic curve at the upper peak position.

[0042] Among them, the matching accuracy between the defect characteristic curve and the data intercepted by the sliding window on the actual low-strain curve is calculated according to formula ①:

[0043]

[0044] In the formula, C i represents the matching accuracy between the defect characteristic curve corresponding to the i-th defect type and the data intercepted by the sliding window on the actual low-strain curve, f i represents the defect characteristic curve corresponding to the i-th defect type, g i represents the data intercepted by the sliding window on the actual low-strain curve, and M is the data length of the defect characteristic curve, that is, the number of data points on the defect characteristic curve.

[0045] In the above formula ①, f i and g iare one-dimensional vectors with the same dimension, so formula ① is actually f i and g i to do a dot product.

[0046] The actual low-strain curve is the low-strain curve of the pile foundation defect obtained by actual measurement. Its length is much greater than the defect characteristic curve, generally dozens to hundreds of times the length of the defect characteristic curve. An actual low-strain curve may include one or more defect types. Therefore, the defect characteristic curve corresponding to each pile foundation defect type needs to be matched and calculated with the data intercepted by the sliding window of the same actual low-strain curve, and the matching accuracy curve of different defect types is obtained.

[0047] Step 4: Calculate the classification probability of the defect type on the actual low-strain curve according to the matching accuracy. Its specific implementation includes:

[0048] Step 4-1: Take the matching accuracy between all the defect characteristic curves of the same defect type in the defect characteristic curve set and the actual low-strain curve of the corresponding defect type as the matching accuracy of this defect type, so as to obtain the matching accuracy of the actual low-strain curve in each defect type.

[0049] Among them, the j-th matching accuracy of the actual low-strain curve under the i-th defect type is denoted as C i,j , so the set of all matching accuracies of the actual low-strain curve under the i-th defect type is [C i,1 , C i,2 , …, C i,j , … C i,Q , Q is the total number of maximum values of the matching accuracy, that is, the total number of upper peaks exceeding the preset value.

[0050] The classification probability P i that the defect type in the actual low-strain curve is classified as the i-th defect type A i,j is calculated according to formula ②:

[0051]

[0052] In the formula, e represents the natural logarithm.

[0053] The classification probability is calculated by formula ② mainly for three reasons: First, the classification probability calculated for the defect type with the maximum matching accuracy is the largest; Second, the sum of the classification probabilities of all defect types is 1; The classification probabilities of each defect type are all greater than 0.

[0054] Through formula ②, the defect type with the maximum matching accuracy will obtain the largest classification probability. At the same time, the sum of the corresponding classification probabilities of the same actual low-strain curve in all defect types is 1.

[0055] Step 4-2: Among the classification probability set [P i,1 , P i,2 , …, P i,Q , take the maximum value as the defect type in the actual low-strain curve and classify it as the i-th defect type A i 's final probability.

[0056] Step 5: Determine the grade of the defect type.

[0057] Perform sampling frequency scaling and amplitude scaling on the defect characteristic curves of each defect type to obtain the scaled defect characteristic curves, calculate the matching accuracy between the scaled defect characteristic curves and the actual low-strain curve, and output the corresponding defect grade according to the size of the matching accuracy.

[0058] In the low-strain detection method, the larger the defect it represents, the larger the reflection amplitude in the low-strain curve. Therefore, the defect grade can be determined by performing sampling frequency scaling and amplitude scaling on the defect characteristic curves.

[0059] Among them, perform sampling frequency and amplitude scaling on the defect characteristic curve of defect type A i . The scaling range (i.e., the scaling factor) of the sampling frequency is 0 - 10. Then, recalculate the classification probability of the actual low-strain curve in each scaled defect characteristic curve according to the above method for the scaled defect characteristic curve. And gradually scale the amplitude until the classification probability of the defect type in the actual low-strain defect is maximized, determine the scaling value of the current amplitude, and then stratify the defect degree into several grades according to the size of the scaling value. For example, divide the defect degree into 4 grades from small to large (grades Ⅰ, Ⅱ, Ⅲ, and Ⅳ respectively), and different scaling values correspond to different grades.

[0060] Reference Figures 2-4 , select a prestressed pipe pile as the pile foundation, with a pile length of 7m (meters). The soil layer is mainly plain fill, silty clay, and silty clay. The actual low-strain curve of this pile foundation is detected on-site using the low-strain detection method, and then the defect type is identified by using the matching accuracy obtained from the matching between the defect characteristic curve and the actual low-strain curve. Figure 2 The left half is the defect characteristic curve of the broken pile defect type, and the right half is the defect characteristic curve of the enlarged diameter defect type. Figure 3 is a schematic diagram of the matching accuracy curve calculated by the sliding step of the defect characteristic curve on the actual low-strain curve. The preset threshold is 0.05, and the position and value of the maximum matching accuracy on the output matching accuracy curve that is greater than 0.05 are output. Figure 4 is a schematic diagram of the position of the defect type located according to the position of the maximum matching accuracy. Figure 4"Defect 1" in it represents a certain type of defect.

[0061] The present invention can realize the automatic identification of pile foundation defects in low-strain curves, greatly improve the efficiency and accuracy of pile foundation detection, and reduce the probability of misjudgment and missed judgment of pile foundation defects caused by manual judgment of low-strain curves.

[0062] The embodiments disclosed in this specification are only an illustration of the unilateral features of the present invention. The protection scope of the present invention is not limited to this embodiment, and any other functionally equivalent embodiments fall within the protection scope of the present invention. For those skilled in the art, various corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A method for identifying pile foundation defects based on feature curve matching of low strain curves, characterized in that It includes the following steps: Step 1: Generate corresponding defect characteristic curves for each type of pile foundation defect, and the individual defect characteristic curves form a set of defect characteristic curves; Step 2: Generate a sliding window of the same length as the defect characteristic curve. Move the sliding window on the actual low-strain curve at a preset sliding step size, intercept data from the actual low-strain curve, and use the distance of the sliding window from the origin on the actual low-strain curve as the independent variable and the matching accuracy between the intercepted data and the defect characteristic curve as the dependent variable to form a matching accuracy curve. Extract the upper peak value from the matching accuracy curve, and take the upper peak value greater than the preset threshold as the maximum matching accuracy value of the current matching accuracy curve. The maximum matching accuracy value characterizes the degree to which the current actual low-strain curve has the defect type reflected by the defect characteristic curve at the upper peak position; Among them, the matching accuracy between the defect characteristic curve and the data intercepted by the sliding window on the actual low-strain curve is calculated according to formula ①: where C i represents the matching accuracy between the defect feature curve corresponding to the i-th type of defect and the data intercepted by the sliding window on the actual low-strain curve, f i represents the defect feature curve corresponding to the i-th type of defect, g i represents the data intercepted by the sliding window on the actual low-strain curve, M is the data length of the defect feature curve, that is, the number of data points on the defect feature curve Step 3: Calculate the classification probability of the defect type on the actual low-strain defect according to the matching accuracy; The specific implementation of step 3 includes the following steps: Step 4-1: Take the matching accuracy between all the defect characteristic curves of the same defect type in the set of defect characteristic curves and the actual low-strain curve of the corresponding defect type as the matching accuracy of this defect type, so as to obtain the matching accuracy of the actual low-strain curve in each defect type; Among them, the matching accuracy of the actual low-strain curve under the i-th defect type at the j-th is denoted as C i,j , therefore, the set of all matching accuracies of the actual low-strain curve under the i-th defect type is [C i,1 , C i,2 , …, C i,j , … C i,Q , where Q is the total number of maximum values of the matching accuracy, that is, the total number of upper peaks exceeding the preset value. The defect type in the actual low-strain curve calculated according to formula ② is classified as the i-th defect type A i with a classification probability P i,j : In the formula, e represents the natural logarithm base; Step 4-2: In the classification probability set [P i,1 , P i,2 , …, P i,Q , the maximum value is used as the classification of the defect type in the actual low strain curve and is classified as the i-th defect type A i 's final probability.

2. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 1, wherein In step 1, the time length of the defect characteristic curve is taken as 0.01 ms - 1 ms, and the number of discrete data points on the defect characteristic curve is between 50 and 1000.

3. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 1, wherein The pile foundation defect types include local necking, local enlargement, and bending.

4. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 1, wherein, The preset sliding step size of the sliding window is 1 - 10.

5. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 1, wherein, Before step 2 and after step 1, the following steps are also included: Convert the defect characteristic curves of each different pile foundation defect type into corresponding text files, denoted as defect characteristic texts, and intercept data on the defect characteristic texts.

6. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 1, wherein, After step 3, it also includes determining the grade of the defect type.

7. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 1, characterized in that, The specific implementation of determining the grade of the defect type includes: Perform sampling frequency scaling and amplitude scaling on the defect characteristic curves of each defect type to obtain the scaled defect characteristic curves, calculate the matching accuracy between the scaled defect characteristic curves and the actual low-strain curve, and output the corresponding defect grade according to the size of the matching accuracy.

8. The method for identifying pile foundation defects based on matching low-strain curves with characteristic curves according to claim 7, wherein, Scale the sampling frequency and amplitude of the defect characteristic curve of defect type A i , then recalculate the classification probability of the actual low-strain curve in each scaled defect characteristic curve according to the above method, and gradually scale the amplitude until the classification probability of the defect type in the actual low-strain defect is maximized, and determine the scaling value of the current amplitude. Then, stratify the defect degree into several levels according to the scaling value size ranking.

9. The pile foundation defect identification method based on the matching of low-strain curves with characteristic curves according to claim 8, characterized in that, The scaling range of the sampling frequency is 0 - 10.

Images