A method and system for identifying integrity and defect categories of rigid composite piles

Abstract

The present invention relates to a method and system for identifying the integrity and defect categories of rigid composite piles. The method comprises performing low-strain reflection wave detection on multiple rigid composite piles to obtain a low-strain reflection wave detection signal from the core pile of each rigid composite pile; performing multi-resolution wavelet decomposition to extract the power mean, variance, and energy ratio of each wavelet component layer as the first eigenvector of the corresponding rigid composite pile; simultaneously, the diameter and length of the core pile are used as the second eigenvector of the rigid composite pile; a training sample data set is formed using the first eigenvector and the second eigenvector; a BP neural network is trained to obtain a defect discrimination model; and identification results are obtained through the output of the defect discrimination model. Compared with the prior art, the present invention reduces the interference caused by the multi-medium and multi-interface characteristics of the rigid composite piles on the analysis of the detection results, thereby improving the accuracy of the integrity and defect identification of the rigid composite piles.

Description

Technical Field

[0001] The present invention relates to the technical field of pile foundation detection, and in particular to a method and system for identifying integrity and defect categories of a rigid composite pile. Background Art

[0002] Due to its low strength, large compression, and poor permeability, soft soil geology is usually constructed using reinforced composite piles consisting of cement-soil mixing piles (outer core) and precast concrete piles (inner core). Reinforced composite piles can not only take advantage of the high bearing capacity of precast concrete piles, but also significantly reduce project costs, thus becoming increasingly widely used. However, since the construction of reinforced composite piles is a hidden underground project, it is often affected by factors such as immature construction technology, complex geological conditions, poor construction team skills, and lax quality control, leading to quality problems such as necking, diameter expansion, cracks, mud inclusions, sediment, and even broken piles during the pile foundation construction process. These problems will inevitably affect the bearing capacity of the pile foundation and thus the safety of the superstructure.

[0003] Existing defect identification methods for rigid composite piles typically employ the low-strain reflection wave method used in conventional pile foundation testing. This method combines changes in physical quantities such as the waveform and wave velocity of the received signal with mathematical statistics and personal experience to make qualitative or empirical judgments about the location and extent of defects. However, this method is significantly influenced by human experience, and its measurement accuracy is highly dependent on the level of the inspector. Clearly, it no longer meets the needs of modern engineering inspections. Furthermore, as a new type of composite pile, the multi-media and multi-interface characteristics of concrete, cement soil, and surrounding soil can interfere with the analysis of the low-strain reflection wave method's test results when core pile integrity testing is performed. This can lead to deviations between the analysis results and the actual core pile integrity, further increasing the difficulty of defect assessment and affecting the accuracy of defect identification. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method and system for identifying the integrity and defect types of rigid composite piles.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A method for identifying integrity and defect categories of a reinforced composite pile comprises the following steps:

[0007] S1. Performing a low-strain reflection wave test on a plurality of rigid composite piles to obtain a low-strain reflection wave test signal of the core pile of each rigid composite pile;

[0008] S2. Perform multi-resolution wavelet decomposition on the acquired reflected wave detection signal, extract the power mean, variance, and energy ratio of each wavelet component of each layer as the first eigenvector of the corresponding stiffened composite pile; at the same time, use the diameter and length of the core pile as the second eigenvector of the stiffened composite pile; and form a training sample data set with the first eigenvector and the second eigenvector;

[0009] S3, training the BP neural network through the training sample data set to obtain a defect discrimination model;

[0010] S4, performing a low-strain reflection wave test on the rigid composite pile to be tested, obtaining a low-strain reflection wave test signal of the core pile, and simultaneously obtaining the diameter and length of the core pile of the rigid composite pile to be tested;

[0011] S5. Extract the first eigenvector from the low-strain reflected wave detection signal of the rigid composite pile to be detected, and use it together with the diameter and length of the core pile as input data of the defect discrimination model, and obtain the recognition result from the output of the defect discrimination model.

[0012] Furthermore, the power mean P ai The expression is:

[0013]

[0014] Where N is the number of signal samples, x N is the amplitude corresponding to each layer signal sample, and i is the number of wavelet decomposition layers.

[0015] Furthermore, the variance S i 2 The expression is:

[0016]

[0017] Where N is the number of signal samples, x N is the amplitude of the corresponding signal sample at each layer, i is the number of wavelet decomposition layers, u i is the mean signal amplitude of each layer.

[0018] Furthermore, the energy ratio of each layer of wavelet is D i The expression is:

[0019]

[0020]

[0021]

[0022] Where N is the number of signal samples, x N is the amplitude of the signal sample at each layer, i is the number of wavelet decomposition layers, Ei is the energy of each layer, E t is the total energy.

[0023] Furthermore, the BP neural network includes input layer nodes, output layer nodes and hidden layer nodes. The elements of the feature vector are loaded into the corresponding input layer nodes as input quantities, and the output nodes generate classification results, i.e., output quantities. A nonlinear relationship between the input quantity and the output quantity is obtained through a certain amount of training sample data, thereby obtaining a fixed defect discrimination model.

[0024] Furthermore, the defect discrimination model outputs recognition results including six categories, namely: complete pile, reduced diameter pile, expanded diameter pile, broken pile, segregated pile and excessive sediment thickness.

[0025] A system for identifying integrity and defect categories of rigid composite piles, comprising:

[0026] The model training module is used to build a defect identification model, including:

[0027] A data acquisition unit is used to perform low-strain reflection wave detection on multiple rigid composite piles to obtain a low-strain reflection wave detection signal of the core pile of each rigid composite pile;

[0028] The sample combination unit performs multi-resolution wavelet decomposition on the acquired reflected wave detection signal, extracts the power mean, variance and energy ratio of each wavelet component of each layer as the first eigenvector of the corresponding stiff composite pile; at the same time, the diameter and length of the core pile of the stiff composite pile are used as the second eigenvector of the stiff composite pile; the first eigenvector and the second eigenvector constitute a training sample data set;

[0029] A training unit, used to train a BP neural network through a training sample data set to obtain a defect discrimination model;

[0030] The identification module is used to identify the integrity and defects of the rigid composite pile to be tested, including:

[0031] The first processing unit is used to perform low-strain reflection wave detection on the rigid composite pile to be detected, obtain the low-strain reflection wave detection signal of the core pile, and simultaneously obtain the diameter and length of the core pile of the rigid composite pile to be detected;

[0032] The second processing unit is used to extract the first eigenvector from the low-strain reflected wave detection signal of the rigid composite pile to be detected, and use it together with the diameter and length of its core pile as input data of the defect discrimination model, and the identification result is obtained by the output of the defect discrimination model.

[0033] Furthermore, the BP neural network includes input layer nodes, output layer nodes and hidden layer nodes. The elements of the feature vector are loaded into the corresponding input layer nodes as input quantities, and the output nodes generate classification results, i.e., output quantities. A nonlinear relationship between the input quantity and the output quantity is obtained through a certain amount of training sample data, thereby obtaining a fixed defect discrimination model.

[0034] Furthermore, the defect discrimination model outputs identification results including six categories, namely: complete piles, reduced diameter piles, expanded diameter piles, broken piles, segregated piles and excessively thick sediment.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. This invention reduces the potential for interference from the surrounding media surrounding the core of a rigid composite pile during the analysis of the reflected wave detection signal by performing multi-resolution wavelet decomposition on the detection signal from the low-strain reflection wave method. Furthermore, the invention combines a BP neural network algorithm to automatically determine the integrity of the rigid composite pile foundation and analyze defect types, thereby improving the accuracy and efficiency of defect analysis. Users can then take appropriate measures to address defective piles, reducing the potential for engineering accidents. This is of great significance for the promotion and engineering application of rigid composite piles.

[0037] 2. After the wavelet decomposition of the detection signal, the present invention selects the power mean, variance and the energy ratio of each layer of wavelet as the first eigenvector, and the diameter and length of the core pile as the second eigenvector. The training sample data set determined in this way can summarize and collect the differences in the detection signals of the low-strain reflection wave method of each stiff core pile to a large extent, and consider the influence of pile type differences such as pile length and pile diameter on the low-strain detection signal, thereby improving the recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the process of the present invention.

[0039] Figure 2 Schematic diagram of wavelet multi-resolution decomposition.

[0040] Figure 3 Schematic diagram of integrity and defect identification of rigid composite piles based on BP neural network. DETAILED DESCRIPTION

[0041] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0042] Example 1

[0043] This embodiment provides a method for identifying the integrity and defect categories of rigid composite piles, including performing multi-resolution wavelet decomposition on the low-strain reflection wave detection signal of the rigid composite pile, extracting feature vectors from each obtained wavelet component to form a set of data samples, using the corresponding data samples and the pile body integrity and defect categories as the input and output of the BP neural network, respectively. After training and verifying multiple sets of data samples, a defect discrimination model is obtained. Figure 1 The specific steps are as follows:

[0044] Step S1 : performing low-strain reflection wave detection on a plurality of rigid composite piles to obtain a low-strain reflection wave detection signal of the core pile of each rigid composite pile.

[0045] Step S2, performing multi-resolution wavelet decomposition on the acquired reflected wave detection signal, extracting the power mean, variance and energy ratio of each layer of wavelet components as the first eigenvector of the corresponding stiffening composite pile; at the same time, the diameter d and length l of the core pile are used as the second eigenvector of the stiffening composite pile; the first eigenvector and the second eigenvector constitute a training sample data set.

[0046] In step S3, a BP (back propagation) neural network is trained to obtain a defect discrimination model by using the training sample data set as an input and the defect category of the rigid composite pile as an output.

[0047] Step S4 , performing a low-strain reflection wave detection on the rigid composite pile to be detected, obtaining a low-strain reflection wave detection signal of the core pile, and simultaneously obtaining the diameter and length of the core pile of the rigid composite pile to be detected.

[0048] Step S5: extracting the first eigenvector from the low-strain reflected wave detection signal of the rigid composite pile to be detected, and using it together with the diameter and length of the core pile as input data of the defect discrimination model, and obtaining the recognition result from the output of the defect discrimination model.

[0049] like Figure 2 As shown in the figure, the specific resolution wavelet decomposition is as follows: the low-strain reflected wave signal is subjected to N-layer wavelet multi-resolution analysis to obtain the characteristic information A and D of the signal in different frequency bands, that is, the original signal is decomposed into N+1 orthogonal frequency bands. The decomposed signal in each frequency band represents the characteristic information of the original signal in this frequency range. The following information is selected as the first eigenvector.

[0050] 1) Power average

[0051] The power mean P of each layer obtained after the discrete signal is decomposed by wavelet ai for:

[0052]

[0053] Where N is the number of signal samples, x N is the amplitude corresponding to each layer signal sample, and i is the number of wavelet decomposition layers.

[0054] 2) Variance

[0055] The power variance S of each layer obtained after the discrete signal is decomposed by wavelet i 2 for:

[0056]

[0057] Where N is the number of signal samples, x N is the amplitude of the corresponding signal sample at each layer, i is the number of wavelet decomposition layers, u i is the mean signal amplitude of each layer.

[0058] 3) Energy distribution

[0059] After the wavelet is decomposed into i layers, the energy ratio of each layer of wavelet is D i The expression is:

[0060]

[0061]

[0062]

[0063] Where N is the number of signal samples, x N is the amplitude of the signal sample at each layer, i is the number of wavelet decomposition layers, E i is the energy of each layer, E t is the total energy.

[0064] like Figure 3 As shown in the figure, BP neural network is a multi-layer feedforward neural network trained according to the error back propagation algorithm. It is a widely used neural network. BP neural network consists of input layer nodes, output layer nodes and hidden layer nodes. The hidden layer can be one or more layers. The elements x1~x n As input, it is loaded into the corresponding input layer node, and the output node generates the classification results y1~y n Through a certain amount of training sample data, the nonlinear relationship between input and output is obtained, thereby fixing the defect discrimination model.

[0065] Since the defects of engineering pile bodies are divided into: diameter reduction, diameter expansion, fracture, segregation and excessive sediment, these five defect types are represented by 1, 2, 3...5 as output results respectively; if it is a complete pile, 0 is used as the output result. Therefore, the defect discrimination model outputs the identification results including six categories, namely 0 complete pile, 1 diameter reduction pile, 2 diameter expansion pile, 3 broken pile, 4 segregation pile and 5 excessive sediment.

[0066] In summary, this embodiment uses a multi-resolution wavelet decomposition method to decompose test data, combined with a BP neural network algorithm. This reduces the interference caused by the multi-media and multi-interface characteristics of rigid composite piles on the analysis of test results, and improves the accuracy of integrity assessment and defect identification of rigid composite piles. This has significant implications for the promotion and engineering application of rigid composite piles.

[0067] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A method for identifying integrity and defect categories of reinforced composite piles, characterized in that: The following steps are involved: S1. Performing a low-strain reflection wave test on a plurality of rigid composite piles to obtain a low-strain reflection wave test signal of the core pile of each rigid composite pile; S2. Perform multi-resolution wavelet decomposition on the acquired reflected wave detection signal, extract the power mean, variance, and energy ratio of each wavelet component of each layer as the first eigenvector corresponding to the rigid composite pile; at the same time, use the diameter and length of the core pile as the second eigenvector corresponding to the rigid composite pile; and form a training sample data set with the first eigenvector and the second eigenvector; S3, using the training sample data set as input to train the BP neural network to obtain a defect discrimination model; S4, performing a low-strain reflection wave test on the rigid composite pile to be tested, obtaining a low-strain reflection wave test signal of the core pile, and simultaneously obtaining the diameter and length of the core pile of the rigid composite pile to be tested; S5. Extract the first eigenvector from the low-strain reflected wave detection signal of the rigid composite pile to be detected, and use it together with the diameter and length of the core pile as input data of the defect discrimination model, and obtain the recognition result from the output of the defect discrimination model.

2. A method for identifying integrity and defect categories of reinforced composite piles according to claim 1, characterized in that: The power mean value P ai The expression is: Where N is the number of signal samples, x N is the amplitude corresponding to each layer signal sample, and i is the number of wavelet decomposition layers.

3. The method for identifying integrity and defect types of reinforced composite piles according to claim 1, characterized in that: The variance S i 2 The expression is: Where N is the number of signal samples, x N is the amplitude of the corresponding signal sample at each layer, i is the number of wavelet decomposition layers, u i is the mean signal amplitude of each layer.

4. A method for identifying integrity and defect categories of reinforced composite piles according to claim 1, characterized in that: The energy ratio D of each layer of wavelet i The expression is: Where N is the number of signal samples, x N is the amplitude of the signal sample at each layer, i is the number of wavelet decomposition layers, E i is the energy of each layer, E t is the total energy.

5. The method for identifying integrity and defect types of reinforced composite piles according to claim 1, characterized in that: The BP neural network includes input layer nodes, output layer nodes and hidden layer nodes. The elements of the feature vector are loaded into the corresponding input layer nodes as input quantities, and the output nodes generate classification results, i.e., output quantities. A nonlinear relationship between the input quantities and the output quantities is obtained through a certain amount of training sample data, thereby obtaining a fixed defect discrimination model.

6. The method for identifying integrity and defect types of reinforced composite piles according to claim 1, characterized in that: The defect discrimination model outputs identification results including six categories: complete pile, reduced diameter pile, expanded diameter pile, broken pile, segregated pile and excessive sediment thickness.

7. A system for identifying integrity and defect types of reinforced composite piles, characterized in that: include: The model training module is used to build a defect identification model, including: A data acquisition unit is used to perform low-strain reflection wave detection on multiple rigid composite piles to obtain a low-strain reflection wave detection signal of the core pile of each rigid composite pile; The sample combination unit performs multi-resolution wavelet decomposition on the acquired reflected wave detection signal, extracts the power mean, variance and energy ratio of each wavelet component of each layer as the first eigenvector corresponding to the rigid composite pile; at the same time, the diameter and length of the core pile of the rigid composite pile are used as the second eigenvector corresponding to the rigid composite pile; the first eigenvector and the second eigenvector constitute a training sample data set; A training unit, used to train a BP neural network using a training sample data set as an input to obtain a defect discrimination model; The identification module is used to identify the integrity and defects of the rigid composite pile to be tested, including: The first processing unit is used to perform low-strain reflection wave detection on the rigid composite pile to be detected, obtain the low-strain reflection wave detection signal of the core pile, and simultaneously obtain the diameter and length of the core pile of the rigid composite pile to be detected; The second processing unit is used to extract the first eigenvector from the low-strain reflected wave detection signal of the rigid composite pile to be detected, and use it together with the diameter and length of its core pile as input data of the defect discrimination model, and the identification result is obtained by the output of the defect discrimination model.

8. A system for identifying integrity and defect types of reinforced composite piles according to claim 7, characterized in that: The BP neural network includes input layer nodes, output layer nodes and hidden layer nodes. The elements of the feature vector are loaded into the corresponding input layer nodes as input quantities, and the output nodes generate classification results, i.e., output quantities. A nonlinear relationship between the input quantities and the output quantities is obtained through a certain amount of training sample data, thereby obtaining a fixed defect discrimination model.

9. The system for identifying integrity and defect types of reinforced composite piles according to claim 7, characterized in that: The defect discrimination model outputs identification results including six categories, namely: complete pile, reduced diameter pile, expanded diameter pile, broken pile, segregated pile and excessive sediment thickness.

Images