Foundation pile defect intelligent identification method, system, equipment and medium

By using the progressive projection operator in the BP neural network model to update the weight parameter, the problems of computing power and memory space limitations in foundation pile defect recognition are solved, and efficient foundation pile defect recognition is achieved.

CN119089279AInactive Publication Date: 2024-12-06CHINA RAILWAY SOUTHWEST SCI RES INST CO LTD

Patent Information

Application Number
CN202411571065.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-06
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Due to the limitations of computing power and memory space, the calculation of foundation pile defect recognition is slow or difficult to complete.

Method used

Introducing an asymmetry projection operator to project the weight parameters of the BP neural network model, prune the connection gradient during iterative optimization, reduce unnecessary gradients, obtain the optimal gradient, and improve model stability and recognition accuracy.

Benefits of technology

A foundation pile defect recognition model with high stability, high recognition accuracy and fast computing processing speed is realized, solving the problem of slow identification caused by equipment computing capabilities and memory space limitations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089279A_ABST
    Figure CN119089279A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent foundation pile defect identification method, system and equipment and a medium, and particularly relates to the technical field of defect identification, and the technical key points are as follows: obtaining stress reflection waves of foundation pile detection, performing wavelet packet decomposition on the stress reflection waves to obtain reflection wave components of different frequency bands; frequency domain parameters and time domain parameters in the reflected wave components of different frequency bands are extracted respectively, multi-dimensional feature vectors are constructed by using the frequency domain parameters and the time domain parameters and combining corresponding reflected wave component energy coefficients, and a training set and a test set are generated; the BP neural network model is trained based on the multi-dimensional feature vectors in the training set, iterative optimization is conducted on the trained BP neural network model through the multi-dimensional feature vectors in the test set, the optimal gradient of the BP neural network model is generated in combination with a progressive projection operator, and an optimal foundation pile defect recognition model is obtained; and decomposing the stress reflection waves to be identified, and inputting the decomposed stress reflection waves into the foundation pile defect identification model to obtain foundation pile defect types.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of defect identification, and in particular to a method, system, equipment and medium for intelligently identifying foundation pile defects. Background Art

[0002] In the construction of engineering projects, pile foundation engineering is the most widely used foundation form in construction projects. With the continuous increase in the height of high-rise buildings, the complexity of building structures, the huge difference in the number of floors and heights, and the development of underground structures and spaces, pile foundation engineering is particularly important. The quality of pile foundation engineering is directly related to the structure and force safety of the building, and even more to people’s lives, property and social security.

[0003] In the process of detecting pile defects or integrity, a low-strain reflection wave method for pile foundation non-destructive testing is often used. It observes the propagation of longitudinal elastic waves and analyzes the types of pile defects based on experience. Therefore, there is a large human factor in the judgment of the defect type, resulting in low accuracy in defect identification. To solve this problem, researchers have proposed a neural network-based pile defect identification method in the prior art. The neural network model is introduced to identify the reflected wave test signal obtained by the detection to detect the defects of the pile. However, since pile detection often requires the deployment of deep upgrade models in mobile devices, and the computing processing units and memory space of these devices are very limited, this method often has problems such as slow calculation and identification or difficulty in completing calculation and identification during actual application.

[0004] Therefore, the present invention aims to provide a method, system, device and medium for intelligent identification of pile defects to solve the above-mentioned related problems. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the existing neural network model is limited by the computing power and memory space of the device, which leads to the related problems of slow calculation and recognition or difficulty in completing the calculation and recognition. The purpose is to provide a method, system, device and medium for intelligent identification of foundation pile defects. By introducing a progressive projection operator to update the weight parameters of the BP neural network model, in the iterative optimization process of the BP neural network model, the connection gradients between the layers of the BP neural network model are pruned, thereby reducing the unnecessary gradients of the layers of the BP neural network model, obtaining the optimal gradients of the layers of the BP neural network model, reducing the oscillation and instability in the training of the BP neural network, thereby obtaining a foundation pile defect identification model with high stability, high recognition accuracy and fast calculation processing speed, and solving the related problems that the existing neural network model is limited by the computing power and memory space of the device, which leads to slow calculation and recognition or difficulty in completing the calculation and recognition.

[0006] The present invention is achieved through the following technical solutions: A method for intelligently identifying pile defects, the method comprising: Obtain stress reflection waves from pile foundation detection, perform wavelet packet decomposition on the stress reflection waves to obtain reflection wave components of different frequency bands, and respectively extract frequency domain parameters and time domain parameters from the reflection wave components of different frequency bands, use frequency domain parameters, time domain parameters and corresponding reflection wave component energy coefficients to construct a multidimensional feature vector, and generate training sets and test sets; The BP neural network model is trained based on the multidimensional feature vectors in the training set, and the trained BP neural network model is iteratively optimized using the multidimensional feature vectors in the test set, and the optimal gradient of the BP neural network model is generated in combination with the progressive projection operator to obtain the optimal pile defect recognition model; The stress reflection wave to be identified is decomposed and input into the optimal pile defect identification model to output the pile defect type.

[0007] Furthermore, the optimal gradient of the BP neural network model is obtained by combining the progressive projection operator, specifically: Input the multidimensional feature vector in the training set into the trained BP neural network model for forward propagation to obtain the output feature vector of each layer of the BP neural network model; The loss function is used to calculate the loss value between the output feature vector of each layer of the BP neural network model and the true value, and the loss function gradient is obtained; Then, the model parameters of the BP neural network model are updated according to back propagation to obtain the gradients of each layer of the BP neural network model; The weight parameters of the BP neural network model are updated based on the progressive projection operator, the gradient of each layer of the BP neural network model is reduced, and the optimal gradient of each layer of the BP neural network model is obtained.

[0008] Furthermore, based on the progressive projection operator, the weight parameters of the BP neural network model are projected and updated to reduce the gradient of each layer of the BP neural network model, and the optimal gradient of each layer of the BP neural network model is obtained, which is specifically: Define the weight parameters of the BP neural network model, where the weight parameters include the weight tensors of each layer; The progressive projection operator is used to adjust the weight tensor of each layer of the BP neural network model to reduce the gradient of each layer of the BP neural network model and obtain the optimal gradient of each layer of the BP neural network model.

[0009] Furthermore, the stress reflection wave of the pile detection is obtained, and the stress reflection wave is decomposed by wavelet packet to obtain the reflection wave components of different frequency bands, and the frequency domain parameters and time domain parameters of the reflection wave components of different frequency bands are respectively extracted, specifically: The pile defect detection device is used to obtain the stress reflection wave of the pile detection. Daubechies Wavelet packet decomposition obtains eight layers of reflected wave components corresponding to different frequency bands; The time domain parameters of the reflected wave components in different frequency bands are extracted respectively, and the frequency domain parameters of the reflected wave components in different frequency bands are obtained by fast Fourier transform calculation.

[0010] Furthermore, the time domain parameters include the speed of sound, the first wave amplitude, the shape factor, the peak factor, the pulse factor, the margin factor and the clearance factor; the frequency domain parameters include the center of gravity frequency, the average frequency, the root mean square frequency and the frequency standard deviation.

[0011] Furthermore, the types of pile defects include normal piles, segregated piles, mud-enclosed piles, hollow piles and broken piles.

[0012] The present invention also provides a system for intelligently identifying foundation pile defects, which is used in any one of the above-mentioned methods for intelligently identifying foundation pile defects, and the system comprises: The first module is used to obtain stress reflection waves of pile foundation detection, perform wavelet packet decomposition on stress reflection waves to obtain reflection wave components of different frequency bands, and respectively extract frequency domain parameters and time domain parameters from reflection wave components of different frequency bands, construct multidimensional feature vectors using frequency domain parameters, time domain parameters and corresponding reflection wave component energy coefficients, and generate training sets and test sets; The second module is used to train the BP neural network model based on the multidimensional feature vectors in the training set, and iteratively optimize the trained BP neural network model using the multidimensional feature vectors in the test set, and generate the optimal gradient of the BP neural network model in combination with the progressive projection operator to obtain the optimal pile defect recognition model; The third module is used to decompose the stress reflection wave to be identified and input it into the optimal pile defect identification model to output the pile defect type.

[0013] Furthermore, the optimal gradient of the BP neural network model is obtained by combining the progressive projection operator, which is: Input the multidimensional feature vector in the training set into the trained BP neural network model for forward propagation to obtain the output feature vector of each layer of the BP neural network model; The loss function is used to calculate the loss value between the output feature vector of each layer of the BP neural network model and the true value, and the loss function gradient is obtained; Then, the model parameters of the BP neural network model are updated according to back propagation to obtain the gradients of each layer of the BP neural network model; The weight parameters of the BP neural network model are updated based on the progressive projection operator, the gradient of each layer of the BP neural network model is reduced, and the optimal gradient of each layer of the BP neural network model is obtained.

[0014] The present invention also provides a computer device, comprising a system memory and a processor, wherein the system memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any one of the methods described above are implemented.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: In the present invention, a progressive projection operator is introduced to project and update the weight parameters of the BP neural network model. During the iterative optimization process of the BP neural network model, the connection gradients between the layers of the BP neural network model are pruned, thereby reducing the unnecessary gradients of the layers of the BP neural network model, obtaining the optimal gradients of the layers of the BP neural network model, and reducing the oscillation and instability in the BP neural network training, thereby obtaining a foundation pile defect recognition model with high stability, high recognition accuracy, and fast calculation processing speed, which solves the related problems that the existing neural network model is limited by the computing power and memory space of the equipment, resulting in slow calculation and recognition or difficulty in completing calculation and recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a schematic diagram of a method flow of a method for intelligently identifying pile defects in this embodiment; Figure 2 This is a schematic diagram of module connections of a system for intelligent identification of pile defects in this embodiment; Figure 3 It is a structural schematic diagram of a computer device in this embodiment. DETAILED DESCRIPTION

[0018] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0019] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0020] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0021] As described in the background, in the prior art, some researchers have proposed a method for identifying pile defects based on a neural network, which introduces a neural network model to identify the reflected wave test signal obtained by detection, so as to detect the defects of the piles. However, since pile detection often requires the deployment of deep upgrade models in mobile devices, and the computing processing units and memory space of these devices are very limited, this method often encounters problems such as slow calculation and recognition or difficulty in completing calculation and recognition during actual application.

[0022] Therefore, the present invention provides a method, system, device and medium for intelligent identification of foundation pile defects, which project and update the weight parameters of the BP neural network model by introducing a progressive projection operator. In the iterative optimization process of the BP neural network model, the connection gradients between the layers of the BP neural network model are pruned, thereby reducing the unnecessary gradients of the layers of the BP neural network model, obtaining the optimal gradients of the layers of the BP neural network model, reducing the oscillation and instability in the BP neural network training, and thus obtaining a foundation pile defect identification model with high stability, high identification accuracy and fast calculation processing speed, solving the related problems that the existing neural network model is limited by the computing power and memory space of the equipment, resulting in slow calculation and identification or difficulty in completing calculation and identification.

[0023] Example 1

[0024] See also Figure 1As shown, this embodiment provides a method for intelligently identifying pile defects, the method comprising: S1: Obtain stress reflection waves from pile detection, perform wavelet packet decomposition on stress reflection waves to obtain reflection wave components of different frequency bands, and extract frequency domain parameters and time domain parameters from reflection wave components of different frequency bands respectively. Use frequency domain parameters, time domain parameters and corresponding reflection wave component energy coefficients to construct a multidimensional feature vector, and generate training and test sets. It should be noted that, in the present embodiment, the stress reflection wave is obtained by detecting the foundation pile using the low strain stress wave reflection method. Stress is applied to the top surface of the foundation pile by a striking device, and then the stress reflection wave signal from the pile body is received by a sensor fixed on the top surface of the pile. This technical solution is a conventional technical solution in the field, and those skilled in the art are all aware of this technical solution, so it will not be elaborated here.

[0025] At the same time, it should be noted that the wavelet packet decomposition method is used to decompose the high-frequency and low-frequency parts of the stress reflection wave to obtain reflection wave components of different frequency bands. The reflection wave components of each frequency band carry the characteristic information of the frequency band. Wavelet packet decomposition can help avoid the occurrence of problems such as spectrum leakage and spectrum fluctuation existing in fast Fourier transform when extracting frequency domain parameters; at the same time, the energy coefficients of the reflection wave components of different frequency bands are obtained by wavelet packet transform. This technical solution is a conventional technical solution in this field. Those skilled in the art are aware of this technical solution and will not be elaborated here.

[0026] S2: Train the BP neural network model based on the multidimensional feature vectors in the training set, iteratively optimize the trained BP neural network model using the multidimensional feature vectors in the test set, and generate the optimal gradient of the BP neural network model in combination with the progressive projection operator to obtain the optimal pile defect recognition model; S3: Decompose the stress reflection wave to be identified and input it into the optimal pile defect identification model to output the pile defect type.

[0027] Specifically, in this embodiment, a progressive projection operator is introduced to project and update the weight parameters of the BP neural network model. During the iterative optimization process of the BP neural network model, the connection gradients between the layers of the BP neural network model are pruned, thereby reducing the unnecessary gradients of the layers of the BP neural network model and obtaining the optimal gradients of the layers of the BP neural network model. Thus, a pile defect recognition model with high stability, high recognition accuracy and fast calculation and processing speed is obtained, which solves the problem that the existing neural network model is limited by the computing power and memory space of the equipment, resulting in slow calculation and recognition or difficulty in completing calculation and recognition.

[0028] Furthermore, the optimal gradient of the BP neural network model is obtained by combining the progressive projection operator, specifically: Input the multidimensional feature vector in the training set into the trained BP neural network model for forward propagation to obtain the output feature vector of each layer of the BP neural network model; The loss function is used to calculate the loss value between the output feature vector of each layer of the BP neural network model and the true value, and the loss function gradient is obtained; It should be noted that in this embodiment, the loss function used is the global Softmin The loss function is: ;

[0029] in, Represents the index of the output feature vector; Indicates The distance between the output feature vector and its corresponding true value; For control softmin Function output representation of the parameters, when hour, softmin The output of the function tends to be standard min function; Indicates The output feature vector is The distance to the true value.

[0030] Then, the model parameters of the BP neural network model are updated according to back propagation to obtain the gradients of each layer of the BP neural network model; It should be noted that, in this embodiment, the gradient is calculated by back propagation using the chain rule. , the gradient calculation is as follows: ;

[0031] in, represents partial derivative; Indicates that globally softmin The loss function The probability that an output feature vector belongs to its corresponding true value.

[0032] for , the gradient calculation is as follows:

[0033] in, Indicates that globally softmin The loss function The output feature vector belongs to The probability of a true value.

[0034] The weight parameters of the BP neural network model are updated based on the progressive projection operator, the gradient of each layer of the BP neural network model is reduced, and the optimal gradient of each layer of the BP neural network model is obtained.

[0035] Furthermore, based on the progressive projection operator, the weight parameters of the BP neural network model are projected and updated to reduce the gradient of each layer of the BP neural network model, and the optimal gradient of each layer of the BP neural network model is obtained, which is specifically: Define the weight parameters of the BP neural network model, where the weight parameters include the weight tensors of each layer; Specifically, in this embodiment, the weight parameters of the BP neural network model are defined as ,in Represents the weight tensor of the first hidden layer; Represents the weight tensor of the second hidden layer; A weight tensor representing the output layer.

[0036] The progressive projection operator is used to adjust the weight tensor of each layer of the BP neural network model to reduce the gradient of each layer of the BP neural network model and obtain the optimal gradient of each layer of the BP neural network model.

[0037] Specifically, in this embodiment, the progressive projection operator is: ; ;

[0038] In the formula, represents the progressive projection operator; Represented as a symbolic function, used to determine the weight parameter The symbol of ; mapping function g: R → R ; Mapping function h:R→R ; and These are all learnable parameters.

[0039] Furthermore, the stress reflection wave of the pile detection is obtained, and the stress reflection wave is decomposed by wavelet packet to obtain the reflection wave components of different frequency bands, and the frequency domain parameters and time domain parameters of the reflection wave components of different frequency bands are respectively extracted, specifically: The pile defect detection device is used to obtain the stress reflection wave of the pile detection. Daubechies Wavelet packet decomposition obtains eight layers of reflected wave components corresponding to different frequency bands; It should be noted that, in this embodiment, the pile defect detection device can adopt a low-strain detector, which includes an excitation device such as a hand hammer or a force rod and a velocity sensor. The excitation device is used to generate stress at the top of the pile, and the velocity sensor is used to receive the reflected stress reflection wave. In other embodiments, a pile dynamic measuring instrument or an integrated pile detector and other instruments and equipment that can obtain stress reflection waves can also be used.

[0040] At the same time, it should be noted that in this embodiment, the selected Daubechies Wavelet packet decomposition is based on the idea of ​​multi-resolution analysis. It decomposes the stress reflection wave signal into sub-bands of different frequencies. Each sub-band contains the information of the stress reflection wave signal within a specific frequency range, so that different frequency components of the signal can be analyzed independently. Daubechies The wavelet function can represent the signal at different scales. In the present embodiment, the number of decomposition layers is set to 8, and other numbers of layers may also be set in other embodiments.

[0041] The time domain parameters of the reflected wave components in different frequency bands are extracted respectively, and the frequency domain parameters of the reflected wave components in different frequency bands are obtained by fast Fourier transform calculation.

[0042] Furthermore, the time domain parameters include the speed of sound, the first wave amplitude, the shape factor, the peak factor, the pulse factor, the margin factor and the clearance factor; the frequency domain parameters include the center of gravity frequency, the average frequency, the root mean square frequency and the frequency standard deviation.

[0043] Furthermore, the types of pile defects include normal piles, segregated piles, mud-enclosed piles, hollow piles and broken piles.

[0044] Example 2

[0045] See also Figure 2 As shown, the present invention also provides a system for intelligently identifying foundation pile defects, which is used in any one of the above-mentioned methods for intelligently identifying foundation pile defects, and the system includes: The first module 100 is used to obtain stress reflection waves of pile foundation detection, perform wavelet packet decomposition on the stress reflection waves to obtain reflection wave components of different frequency bands, and respectively extract frequency domain parameters and time domain parameters from the reflection wave components of different frequency bands, construct a multidimensional feature vector using the frequency domain parameters, time domain parameters and corresponding reflection wave component energy coefficients, and generate a training set and a test set; The second module 200 is used to train the BP neural network model based on the multidimensional feature vectors in the training set, iteratively optimize the trained BP neural network model using the multidimensional feature vectors in the test set, and generate the optimal gradient of the BP neural network model in combination with the progressive projection operator to obtain the optimal pile defect recognition model; The third module 300 is used to decompose the stress reflection wave to be identified and input it into the optimal pile defect identification model to output the pile defect type.

[0046] Furthermore, the optimal gradient of the BP neural network model is obtained by combining the progressive projection operator, which is: Input the multidimensional feature vector in the training set into the trained BP neural network model for forward propagation to obtain the output feature vector of each layer of the BP neural network model; The loss function is used to calculate the loss value between the output feature vector of each layer of the BP neural network model and the true value, and the loss function gradient is obtained; Then, the model parameters of the BP neural network model are updated according to back propagation to obtain the gradients of each layer of the BP neural network model; The weight parameters of the BP neural network model are updated based on the progressive projection operator, the gradient of each layer of the BP neural network model is reduced, and the optimal gradient of each layer of the BP neural network model is obtained.

[0047] It should be noted that the modules in the system of Example 2 correspond to the steps in the method of Example 1. The steps in the method of Example 1 have been described in detail in Example 1. In this Example 2, the contents of the modules in the system will not be described in detail.

[0048] Example 3

[0049] See also Figure 3 As shown, this embodiment further provides a computer device, including a system memory 1005 and a processor 1001, wherein the system memory 1005 stores a computer program, and the processor 1001 implements the steps of any of the above methods when executing the computer program.

[0050] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, the processor 1001 implements the functions of each module / unit in the above system / device embodiments when executing the computer program.

[0051] Specifically, in this embodiment, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the system memory 1005, and are executed by the processor 1001 to complete the present application. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0052] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will appreciate that this does not constitute a limitation on the terminal device, and may include more or less components than shown in the figure, or combine certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0053] The processor 1001 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0054] The system memory 1005 may be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device. The system memory 1005 may also be a storage device 1004 of the terminal device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the terminal device. Further, the system memory 1005 may also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store computer programs and other programs and data required by the terminal device. The system memory 1005 may also be used to temporarily store data that has been output or is to be output.

[0055] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0056] Example 4

[0057] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above methods are implemented.

[0058] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, system or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art.

[0059] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In an embodiment of the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system, or device.

[0060] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent identification of pile defects, characterized in that: Methods include: Obtain stress reflection waves from pile foundation detection, perform wavelet packet decomposition on the stress reflection waves to obtain reflection wave components of different frequency bands, and respectively extract frequency domain parameters and time domain parameters from the reflection wave components of different frequency bands, use frequency domain parameters, time domain parameters and corresponding reflection wave component energy coefficients to construct a multidimensional feature vector, and generate training sets and test sets; The BP neural network model is trained based on the multidimensional feature vectors in the training set, and the trained BP neural network model is iteratively optimized using the multidimensional feature vectors in the test set, and the optimal gradient of the BP neural network model is generated in combination with the progressive projection operator to obtain the optimal pile defect recognition model; The stress reflection wave to be identified is decomposed and input into the optimal pile defect identification model to output the pile defect type.

2. The method for intelligent identification of pile defects according to claim 1, characterized in that: Combined with the progressive projection operator, the optimal gradient of the BP neural network model is generated, specifically: Input the multidimensional feature vector in the training set into the trained BP neural network model for forward propagation to obtain the output feature vector of each layer of the BP neural network model; The loss function is used to calculate the loss value between the output feature vector of each layer of the BP neural network model and the true value, and the loss function gradient is obtained; Then, the model parameters of the BP neural network model are updated according to back propagation to obtain the gradients of each layer of the BP neural network model; The weight parameters of the BP neural network model are updated based on the progressive projection operator, the gradients of each layer of the BP neural network model are reduced, and the optimal gradients of each layer of the BP neural network model are generated.

3. The method for intelligently identifying pile defects according to claim 2, characterized in that: Based on the progressive projection operator, the weight parameters of the BP neural network model are projected and updated to reduce the gradients of each layer of the BP neural network model and generate the optimal gradients of each layer of the BP neural network model. Specifically: Define the weight parameters of the BP neural network model, where the weight parameters include the weight tensors of each layer; The progressive projection operator is used to adjust the weight tensor of each layer of the BP neural network model to reduce the gradient of each layer of the BP neural network model and obtain the optimal gradient of each layer of the BP neural network model.

4. The method for intelligently identifying pile defects according to claim 1, characterized in that: Obtain the stress reflection wave of the pile detection, perform wavelet packet decomposition on the stress reflection wave to obtain reflection wave components of different frequency bands, and extract the frequency domain parameters and time domain parameters of the reflection wave components of different frequency bands respectively, specifically: The pile defect detection device is used to obtain the stress reflection wave of the pile detection. Daubechies Wavelet packet decomposition obtains eight layers of reflected wave components corresponding to different frequency bands; The time domain parameters of the reflected wave components in different frequency bands are extracted respectively, and the frequency domain parameters of the reflected wave components in different frequency bands are obtained by fast Fourier transform calculation.

5. The method for intelligently identifying pile defects according to claim 1, characterized in that: The time domain parameters include sound velocity, first wave amplitude, shape factor, peak factor, pulse factor, margin factor and clearance factor; the frequency domain parameters include center of gravity frequency, average frequency, root mean square frequency and frequency standard deviation.

6. The method for intelligent identification of pile defects according to claim 1, characterized in that: The types of pile defects include normal piles, segregated piles, mud-enclosed piles, hollow piles and broken piles.

7. A pile defect intelligent identification system, characterized in that: The system is used in a method for intelligently identifying pile defects according to any one of claims 1 to 6, and the system comprises: The first module is used to obtain stress reflection waves of pile foundation detection, perform wavelet packet decomposition on stress reflection waves to obtain reflection wave components of different frequency bands, and respectively extract frequency domain parameters and time domain parameters from reflection wave components of different frequency bands, construct multidimensional feature vectors using frequency domain parameters, time domain parameters and corresponding reflection wave component energy coefficients, and generate training sets and test sets; The second module is used to train the BP neural network model based on the multidimensional feature vectors in the training set, and iteratively optimize the trained BP neural network model using the multidimensional feature vectors in the test set, and generate the optimal gradient of the BP neural network model in combination with the progressive projection operator to obtain the optimal pile defect recognition model; The third module is used to decompose the stress reflection wave to be identified and input it into the optimal pile defect identification model to output the pile defect type.

8. The intelligent identification system for pile defects according to claim 7 is characterized in that: Combined with the progressive projection operator, the optimal gradient of the BP neural network model is obtained, specifically: Input the multidimensional feature vector in the training set into the trained BP neural network model for forward propagation to obtain the output feature vector of each layer of the BP neural network model; The loss function is used to calculate the loss value between the output feature vector of each layer of the BP neural network model and the true value, and the loss function gradient is obtained; Then, the model parameters of the BP neural network model are updated according to back propagation to obtain the gradients of each layer of the BP neural network model; The weight parameters of the BP neural network model are updated based on the progressive projection operator, the gradient of each layer of the BP neural network model is reduced, and the optimal gradient of each layer of the BP neural network model is obtained.

9. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Method and system for identifying integrity and defect category of stiff composite pile

    CN113962271A

  • Pile foundation integrity evaluation method based on deep learning and ultrasonic reflection

    CN117951581A

Cited By

  • Foundation pile defect intelligent identification method and system based on KPCA-LightGBM-SADE

    CN122045714A