Concrete structure nondestructive testing damage identification method

Abstract

The invention belongs to the technical field of non-destructive testing, and particularly relates to a concrete structure non-destructive testing damage identification method which comprises the following steps: step 1, arranging monitoring points of a concrete structure, and selecting the monitoring points at the maximum load or easily damaged nodes of the concrete structure; 2, adopting different monitoring modes for the monitoring points, wherein the monitoring modes comprise active monitoring and passive monitoring; and step 3, carrying an acoustic emission probe in the active monitoring. According to the method, the acoustic emission signal classification model based on machine learning is established and combined with various characteristic parameters of the acoustic emission signals, so that extraction and classification of data information are comprehensively guaranteed, a nondestructive testing and damage analysis method based on the decision tree is provided, a decision tree classification model can be established, the accuracy is high, and the method is suitable for popularization and application. A normal acceleration measuring device and a noise collecting device are placed at the position of a selected measuring point on the surface of a concrete structure and near the concrete structure, so that lossless and non-contact knotting detection technology is realized.

Description

Technical Field

[0001] The invention relates to a concrete structure damage identification method, in particular to a concrete structure nondestructive detection damage identification method. Background Art

[0002] In the maintenance and management of engineering projects, non-destructive testing of large concrete structures is a very important technical means, which can detect and promptly warn of damage caused to the structure during operation. However, the current non-destructive testing technology has not yet been fully developed; acoustic emission technology has attracted a lot of attention in recent years due to its characteristics such as no need for transmitting transducers, overall detection, and suitability for dynamic monitoring, especially in environments that require long-term health observation.

[0003] At present, many projects try to use acoustic emission technology to detect internal damage of large concrete structures; however, due to the complexity of the internal destruction process of concrete, a large number of acoustic emission signals will be generated during its deterioration process. Excessive data volume will bring difficulties to data analysis. Although traditional parameter analysis methods such as amplitude and ringing count are simple and easy to implement, they are easily interfered and cannot accurately reflect the damage state of the structure; the frequency domain analysis method has strong anti-noise ability, but it also has the defect of poor consistency, which may cause deviations in the understanding of information and thus form wrong judgments.

[0004] In view of this, a method is needed that can effectively analyze nondestructive testing data for damage identification. Summary of the invention

[0005] The main purpose of the present disclosure is to provide a method for nondestructive testing and damage identification of concrete structures, so as to effectively solve the problems raised by the inventor in the above-mentioned background technology.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A nondestructive detection damage identification method for concrete structure comprises the following steps:

[0008] Step 1: Arrange monitoring points of the concrete structure, and select the monitoring points at the maximum load or the nodes prone to damage of the concrete structure; specifically: arrange vibration response measurement points on the upper surface of the concrete structure, arrange noise measurement points 20 cm away from the normal position of the vibration response measurement points, install accelerometers at each vibration response measurement to measure the normal acceleration, and install microphones at the noise measurement points to measure the overall noise;

[0009] Step 2: Adopt different monitoring methods for monitoring points, including active monitoring and passive monitoring;

[0010] Step 3: The active monitoring is to carry an acoustic emission probe and make the acoustic emission probe contact with the surface of the concrete structure to form acoustic emission data points, and cover the acoustic emission probe with a sound insulation device;

[0011] Step 4: Collect the signal emitted by the acoustic emission probe, use a computer to calculate and analyze to obtain the acoustic emission signal parameters, reorganize the data of the acoustic emission parameters, and screen out reliable acoustic emission signal parameters, wherein the acoustic emission signal parameters include rise time RT, ring count C, energy E, duration TD, amplitude PA, RA value, and center frequency FG;

[0012] Step 5: Analyze the linear correlation between different acoustic emission parameter data, and remove the parameter with the highest correlation. Repeat steps 1 to 3 so that the highest parameter is removed from the data of each repeated test process, leaving three parameters with the lowest correlation.

[0013] Step 6: Use the CFSFDP machine learning algorithm to perform cluster analysis on the data in step 4;

[0014] Step 7: Compare the cluster analysis results with the concrete damage characteristics;

[0015] Step 8: Using a 10-fold cross validation method, construct the data in step 4 into training samples, build a decision tree classification model, and input the training samples into the decision tree classification model for training;

[0016] Step 9: Conduct repeated experiments to verify the conclusions and avoid accidental results.

[0017] Preferably, the training of the decision tree classification model in step eight comprises the following steps:

[0018] Establish the root node of the model; calculate the information gain of the candidate parameters of the training sample, and mark the first attribute with the largest information gain as the new root node, where the information gain calculation formula is as follows:

[0019]

[0020]

[0021] I(P,Y)=H(P)-H(P|Y i );

[0022] In the above formula: H(P) is the information entropy, X is the damage type represented by the acoustic emission data, and p i is the frequency of the signal generated by the corresponding damage, H(P|Y i ) is the conditional entropy, Y i is the data feature i of the corresponding parameter of acoustic emission, p(y i) is the parameter y i Among the characteristics, the damage type p i The probability of occurrence, I(P, Y) is the information gain;

[0023] A new branch is generated at the new root node, and the branch is defined as a subset of the acoustic emission samples of the training sample set; so after the subtree is created, the final decision tree is used as the decision tree classification type.

[0024] Preferably, in step eight, the hyperparameters are further adjusted to optimize the decision tree classification model; based on the hyperparameters, a proxy model is constructed using the Bayesian optimization strategy, and the proxy model is used to predict indicators from the hyperparameter configuration, and the selection is repeated until a preset number of iterations is reached or a preset constraint value is exceeded, and finally the parameters are approached to the optimal value, thereby completing the optimization and parameter adjustment of the model; based on the decision tree classification model, the data to be classified is input into the model that completes the parameter adjustment, and through calculation, the classification model marks each acoustic emission data with a category label and outputs it to complete the classification of the data points. According to the characteristics of concrete structure damage, the damage type is divided into four types: matrix damage, aggregate damage, interface friction, and fiber fracture. According to the classification results corresponding to the damage type, damage pattern recognition is performed. When the sum of the signal ratios of aggregate damage and fiber fracture exceeds the sum of the signal ratios of matrix damage and interface friction, it is marked as potential severe damage, and the information is fed back to the external information center.

[0025] Preferably, the method in step five is: introducing orthogonal linear transformation, using principal component analysis algorithm to convert related multivariate data into linearly unrelated multivariate data; then using Laplace coefficients and related parameters to evaluate the informativeness and credibility of different characteristic parameters of each acoustic emission data point; thereby screening out reliable acoustic emission signal parameters.

[0026] Preferably, in step seven, the concrete damage types include fiber / aggregate fracture, fiber or aggregate pull-out, matrix shear fracture and matrix tensile fracture.

[0027] Preferably, in step 2, the passive monitoring method is: using a sliding smoothing method to eliminate trend items from the measured signal of the piezoelectric ceramic sensor; using an empirical mode decomposition method to reduce signal noise to obtain a near-baseline, noise-free preprocessed signal; extracting the free vibration response from the signal obtained in the above steps, and identifying the modal parameters of the free vibration data; finally, establishing a damage indication index to determine the occurrence of non-destructive damage or classify the degree of damage.

[0028] Preferably, in step eight, the optimization of hyperparameters adopts a Bayesian optimization method.

[0029] Preferably, the hyperparameters include: minimum leaf size, minimum number of leaf nodes, maximum number of splits and maximum number of branch nodes.

[0030] Preferably, the process of identifying the degree of structural damage in the passive monitoring is: selecting a signal transmitter to transmit a waveform signal, and identifying the degree of structural damage by the amplitude of the received signal piezoelectric ceramic sensor; if the amplitude of the received signal is smaller than the amplitude under non-destructive conditions, it is determined that there is damage in the monitored area; if the amplitude of the received signal is equal to the amplitude under non-destructive conditions, it is determined that there is no damage in the monitored area.

[0031] In view of this, compared with the prior art, the beneficial effects of the present invention are:

[0032] (I) In the present application, an acoustic emission signal classification model based on machine learning is established in combination with a variety of characteristic parameters of acoustic emission signals to comprehensively ensure the extraction and classification of data information, and a non-destructive testing and damage analysis method based on a decision tree is proposed. A decision tree classification model can be established with high accuracy. The type of structural damage can also be mined from acoustic emission data, abnormal signals in the data can be identified, and it can be determined which signal is more likely to represent the failure of the structure.

[0033] (ii) In the present application, the electrical signals monitored by the piezoelectric ceramic sensor can be analyzed and processed to effectively identify the damage of the concrete structure, especially the crack damage, that is, it can accurately determine whether the damage has occurred, the location of the damage and the extent of the damage. By selecting the measuring points on the surface of the concrete structure and its vicinity, placing the normal acceleration measurement device and the noise collection device, the looseness detection technology can be realized non-destructively and non-contact, providing a new identification method for the safety assessment technology of mixed soil structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The figure shows a schematic flow chart of the nondestructive testing damage identification method for concrete structures provided by the present invention;

[0035] Figure 2 Shown is a schematic diagram of the passive monitoring experimental system. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] See also Figure 1-2, the present invention provides the following embodiments:

[0038] A nondestructive detection damage identification method for concrete structure comprises the following steps:

[0039] Step 1: Arrange monitoring points of the concrete structure, and select the monitoring points at the maximum load or the nodes prone to damage of the concrete structure; specifically: arrange vibration response measurement points on the upper surface of the concrete structure, arrange noise measurement points 20 cm away from the normal position of the vibration response measurement points, install accelerometers at each vibration response measurement to measure the normal acceleration, and install microphones at the noise measurement points to measure the overall noise;

[0040] Step 2: Adopt different monitoring methods for monitoring points, including active monitoring and passive monitoring;

[0041] Step 3: Active monitoring is to carry an acoustic emission probe and make it contact with the surface of the concrete structure to form acoustic emission data points, and cover the acoustic emission probe with a sound insulation device;

[0042] Step 4: Collect the signal emitted by the acoustic emission probe, use computer calculation and analysis to obtain the acoustic emission signal parameters, reorganize the data of the acoustic emission parameters, and screen out reliable acoustic emission signal parameters. The acoustic emission signal parameters include rise time RT, ring count C, energy E, duration TD, amplitude PA, RA value, and center frequency FG;

[0043] Step 5: Analyze the linear correlation between different acoustic emission parameter data, and remove the parameter with the highest correlation. Repeat steps 1 to 3 so that the highest parameter is removed from the data of each repeated test process, leaving three parameters with the lowest correlation.

[0044] Step 6: Use the CFSFDP machine learning algorithm to perform cluster analysis on the data in step 4;

[0045] Step 7: Compare the cluster analysis results with the concrete damage characteristics;

[0046] Step 8: Use the 10-fold cross validation method to construct the data in step 4 into training samples, build a decision tree classification model, and input the training samples into the decision tree classification model for training;

[0047] Step 9: Conduct repeated experiments to verify the conclusions and avoid accidental results.

[0048] Specifically, the training of the decision tree classification model in step eight includes the following steps:

[0049] Establish the root node of the model; calculate the information gain of the candidate parameters of the training sample, and mark the first attribute with the largest information gain as the new root node. The information gain calculation formula is as follows:

[0050]

[0051] I(P,Y)=H(P)-H(P|Y i );

[0052] In the above formula: H(P) is the information entropy, X is the damage type represented by the acoustic emission data, and p i is the frequency of the signal generated by the corresponding damage, H(P|Y i ) is the conditional entropy, Y i is the data feature i of the corresponding parameter of acoustic emission, p(y i ) is the parameter y i Among the characteristics, the damage type p i The probability of occurrence, I(P, Y) is the information gain;

[0053] A new branch is generated at the new root node, and the branch is defined as a subset of the acoustic emission samples of the training sample set; so after the subtree is created, the final decision tree is used as the decision tree classification type.

[0054] Specifically, in step eight, the hyperparameters are further adjusted to optimize the decision tree classification model; based on the hyperparameters, a proxy model is constructed using the Bayesian optimization strategy, and the proxy model is used to predict indicators from the hyperparameter configuration, and the selection is repeated until the preset number of iterations is reached or the preset constraint value is exceeded, and finally the parameters are approached to the optimal value, thereby completing the optimization and parameter adjustment of the model; based on the decision tree classification model, the data to be classified is input into the model that has completed the parameter adjustment. Through calculation, the classification model marks each acoustic emission data with a category label and outputs it to complete the classification of the data points. According to the characteristics of concrete structure damage, the damage type is divided into four types: matrix damage, aggregate damage, interface friction, and fiber fracture. According to the classification results corresponding to the damage type, damage pattern recognition is performed. When the sum of the signal ratios of aggregate damage and fiber fracture exceeds the sum of the signal ratios of matrix damage and interface friction, it is marked as potential serious damage and the information is fed back to the external information center.

[0055] Specifically, the method in step five is: introduce orthogonal linear transformation, use principal component analysis algorithm to transform related multivariate data into linearly unrelated multivariate data; then use Laplace coefficients and related parameters to evaluate the informativeness and credibility of different characteristic parameters of each acoustic emission data point; thereby screening out reliable acoustic emission signal parameters.

[0056] Specifically, in step seven, the concrete damage types include fiber / aggregate fracture, fiber or aggregate pull-out, matrix shear fracture, and matrix tensile fracture.

[0057] Specifically, in step 2, the passive monitoring method is: use the sliding smoothing method to eliminate the trend item of the measured signal of the piezoelectric ceramic sensor; use the empirical mode decomposition method to reduce the noise of the signal to obtain a near-baseline, noise-free preprocessed signal; extract the free vibration response from the signal obtained in the above steps, and identify the modal parameters of the free vibration data; finally, establish a damage indication index to judge the occurrence of non-destructive damage or classify the degree of damage.

[0058] Specifically, in step eight, the optimization of hyperparameters adopts the Bayesian optimization method.

[0059] Specifically, the hyperparameters include: minimum leaf size, minimum number of leaf nodes, maximum number of splits, and maximum number of branch nodes.

[0060] Specifically, the process of identifying the degree of structural damage in passive monitoring is: select a signal transmitter to transmit a waveform signal, and identify the degree of structural damage by the amplitude of the received signal piezoelectric ceramic sensor; if the amplitude of the received signal is smaller than the amplitude under non-destructive conditions, it is determined that there is damage in the monitored area; if the amplitude of the received signal is equal to the amplitude under non-destructive conditions, it is determined that there is no damage in the monitored area.

[0061] The specific implementation method of this embodiment is as follows: by establishing an acoustic emission signal classification model based on machine learning combined with multiple characteristic parameters of the acoustic emission signal, the extraction and classification of data information is fully guaranteed, and a non-destructive testing and damage analysis method based on a decision tree is proposed. A decision tree classification model can be established with high accuracy, and the type of structural damage can be mined from the acoustic emission data, and abnormal signals in the data can be identified, and it can be determined which signal is more likely to represent the failure of the structure; the electrical signal monitored by the piezoelectric ceramic sensor can be effectively identified from it after analysis and processing, especially crack damage, that is, it can accurately determine whether damage occurs, the location of the damage and the degree of the damage, and the normal acceleration measurement device and the noise collection device are placed on the surface of the concrete structure and the vicinity of the selected measuring point position, so that the looseness detection technology can be realized non-destructively and non-contact, providing a new identification method for the safety assessment technology of mixed soil structure.

[0062] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0063] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for nondestructive testing and damage identification of concrete structures, characterized in that: The following steps are involved: Step 1: Arrange monitoring points of the concrete structure, and select the monitoring points at the maximum load or the nodes prone to damage of the concrete structure; specifically: arrange vibration response measurement points on the upper surface of the concrete structure, arrange noise measurement points 20 cm away from the normal position of the vibration response measurement points, install accelerometers at each vibration response measurement to measure the normal acceleration, and install microphones at the noise measurement points to measure the overall noise; Step 2: Adopt different monitoring methods for monitoring points, including active monitoring and passive monitoring; Step 3: The active monitoring is to carry an acoustic emission probe and make the acoustic emission probe contact with the surface of the concrete structure to form acoustic emission data points, and cover the acoustic emission probe with a sound insulation device; Step 4: Collect the signal emitted by the acoustic emission probe, use a computer to calculate and analyze to obtain the acoustic emission signal parameters, reorganize the data of the acoustic emission parameters, and screen out reliable acoustic emission signal parameters, wherein the acoustic emission signal parameters include rise time RT, ring count C, energy E, duration TD, amplitude PA, RA value, and center frequency FG; Step 5: Analyze the linear correlation between different acoustic emission parameter data, and remove the parameter with the highest correlation. Repeat steps 1 to 3 so that the highest parameter is removed from the data of each repeated test process, leaving three parameters with the lowest correlation. Step 6: Use the CFSFDP machine learning algorithm to perform cluster analysis on the data in step 4; Step 7: Compare the cluster analysis results with the concrete damage characteristics; Step 8: Using a 10-fold cross validation method, construct the data in step 4 into training samples, build a decision tree classification model, and input the training samples into the decision tree classification model for training; Step 9: Conduct repeated experiments to verify the conclusions and avoid accidental results.

2. A method for nondestructive testing and damage identification of concrete structures according to claim 1, characterized in that: The training of the decision tree classification model in step eight includes the following steps: Establish the root node of the model; calculate the information gain of the candidate parameters of the training sample, and mark the first attribute with the largest information gain as the new root node, where the information gain calculation formula is as follows: I(P,Y)=H(P)-H(P|Y i ); In the above formula: H(P) is the information entropy, X is the damage type represented by the acoustic emission data, and p i is the frequency of the signal generated by the corresponding damage, H(P|Y i ) is the conditional entropy, Y i is the data feature i of the corresponding parameter of acoustic emission, p(y i ) is the parameter y i Among the characteristics, the damage type p i The probability of occurrence, I(P, Y) is the information gain; A new branch is generated at the new root node, and the branch is defined as a subset of the acoustic emission samples of the training sample set; so after the subtree is created, the final decision tree is used as the decision tree classification type.

3. A method for nondestructive testing and damage identification of concrete structures according to claim 2, characterized in that: In the step eight, the hyperparameters are further adjusted to optimize the decision tree classification model; based on the hyperparameters, a proxy model is constructed using the Bayesian optimization strategy, and the proxy model is used to predict indicators from the hyperparameter configuration, and the selection is repeated until the preset number of iterations is reached or the preset constraint value is exceeded, and finally the parameters are approached to the optimal value, thereby completing the optimization and parameter adjustment of the model; based on the decision tree classification model, the data to be classified is input into the model that has completed the parameter adjustment, and through calculation, the classification model marks each acoustic emission data with a category label and outputs it to complete the classification of the data points. According to the characteristics of concrete structure damage, the damage type is divided into four types: matrix damage, aggregate damage, interface friction, and fiber fracture. According to the classification results corresponding to the damage type, damage pattern recognition is performed. When the sum of the signal ratios of aggregate damage and fiber fracture exceeds the sum of the signal ratios of matrix damage and interface friction, it is marked as potential serious damage, and the information is fed back to the external information center.

4. A method for nondestructive testing and damage identification of concrete structures according to claim 1, characterized in that: The method in step five is: introducing orthogonal linear transformation, using principal component analysis algorithm to transform related multivariate data into linearly unrelated multivariate data; then using Laplace coefficients and related parameters to evaluate the informativeness and credibility of different characteristic parameters of each acoustic emission data point; thereby screening out reliable acoustic emission signal parameters.

5. The method for nondestructive testing and damage identification of concrete structure according to claim 1, characterized in that: In step seven, the concrete damage types include fiber / aggregate fracture, fiber or aggregate pull-out, matrix shear fracture and matrix tensile fracture.

6. A method for nondestructive testing and damage identification of concrete structures according to claim 1, characterized in that: In the step 2, the passive monitoring method is: using a sliding smoothing method to eliminate the trend term of the measured signal of the piezoelectric ceramic sensor; using an empirical mode decomposition method to reduce signal noise to obtain a near-baseline, noise-free preprocessed signal; The free vibration response is extracted from the signal obtained in the above steps, and the modal parameters of the free vibration data are identified; finally, a damage indication index is established to determine whether lossless damage has occurred or to classify the degree of damage.

7. A method for nondestructive testing and damage identification of concrete structures according to claim 3, characterized in that: In step eight, the hyperparameters are optimized using a Bayesian optimization method.

8. A method for nondestructive testing and damage identification of concrete structures according to claim 7, characterized in that: The hyperparameters include: minimum leaf size, minimum number of leaf nodes, maximum number of splits, and maximum number of branch nodes.

9. A method for nondestructive testing and damage identification of concrete structures according to claim 6, characterized in that: The process of identifying the degree of structural damage in the passive monitoring is as follows: a signal transmitter is selected to transmit a waveform signal, and the degree of structural damage is identified by the amplitude of the received signal piezoelectric ceramic sensor; if the amplitude of the received signal is smaller than the amplitude under non-destructive conditions, it is determined that there is damage in the monitored area; if the amplitude of the received signal is equal to the amplitude under non-destructive conditions, it is determined that there is no damage in the monitored area.

Images