In-service structure cross-domain prestress identification method based on ultrasonic guided waves

Abstract

The invention provides an in-service structure cross-domain prestress recognition method based on ultrasonic guided waves. The method comprises the steps that ultrasonic guided signals of steel strands under the in-service condition are collected; the ultrasonic guided signal of the steel strand under the in-service condition is input into a target machine learning model, the prestress predicted value of the steel strand under the in-service condition is obtained, the target machine learning model is obtained by taking the mapped source domain ultrasonic guided signal feature matrix and the corresponding prestress as samples for training, and the prestress predicted value of the steel strand under the in-service condition is obtained. The source domain ultrasonic guided-wave signals are ultrasonic guided-wave signals of the steel strand under different prestressing forces in an experimental state, and the mapping process is to perform mapping by taking the distribution difference minimization of the characteristic parameters of the source domain ultrasonic guided-wave signals and the characteristic parameters of the ultrasonic guided-wave signals of the steel strand under the in-service condition as a target. By implementing the method, the problem of feature distribution difference between the source domain and the target domain can be effectively solved, the demand of a target domain sample can be reduced, the data acquisition difficulty is reduced, and meanwhile, the training efficiency of the model is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and in particular relates to a method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves. Background Art

[0002] Prestressed steel strands play an important role in modern civil engineering structures. However, with the increase of service life, the loss of prestress in steel strands is inevitable due to factors such as material aging, stress relaxation, and environmental corrosion. Therefore, how to accurately identify the effective prestress of the structure is a key issue to ensure structural safety and needs to be solved.

[0003] Ultrasonic guided wave technology has the characteristics of long propagation distance and small signal attenuation, and has been widely used in structural health monitoring and non-destructive testing. In recent years, ultrasonic guided wave technology has also been used in the effective prestress identification of steel strands, including characteristic parameter-based methods and machine learning-based methods. However, whether it is a characteristic parameter-based method or a machine learning-based method, it is necessary to obtain ultrasonic guided wave signals under different prestress states in advance for parameter fitting and model training. For in-service structures in actual engineering, it is unrealistic to obtain ultrasonic guided wave signals of steel strands under different prestress states. The above problem is also a major problem that currently limits the further application of ultrasonic guided waves in the identification of prestress of steel strands. Therefore, how to determine the prestress of in-service structures is an urgent problem to be solved. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method and device for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves, so as to meet the demand for effectively predicting the prestress of in-service structures.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] According to a first aspect, the present invention provides a method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves, comprising: collecting ultrasonic guided wave signals of steel strands when in service; inputting the ultrasonic guided wave signals of the steel strands when in service into a target machine learning model to obtain a predicted value of the prestress of the steel strands when in service, wherein the target machine learning model is trained based on the mapped source domain ultrasonic guided wave signals and the corresponding prestress as samples, the source domain ultrasonic guided wave signals are ultrasonic guided wave signals of steel strands under different prestresses in an experimental state, and the mapping process is to minimize the distribution difference between the characteristic parameters of the source domain ultrasonic guided wave signals and the characteristic parameters of the ultrasonic guided wave signals of the steel strands when in service.

[0007] Optionally, the sample determination process of the target machine learning model includes: collecting ultrasonic waveguide signals of steel strands under different prestresses in an experimental state as source domain ultrasonic waveguide signals, and collecting ultrasonic waveguide signals of steel strands under service as target domain ultrasonic waveguide signals; respectively extracting multiple target feature parameters of the source domain ultrasonic waveguide signals and the target domain ultrasonic waveguide signals to obtain source domain ultrasonic waveguide signal feature parameters and target domain ultrasonic waveguide signal feature parameters; mapping the source domain ultrasonic waveguide signal feature parameters with the goal of minimizing the distribution difference between the source domain ultrasonic waveguide signal feature parameters and the target domain ultrasonic waveguide signal feature parameters to obtain a mapped source domain ultrasonic waveguide signal feature matrix; inputting the mapped source domain ultrasonic waveguide signal feature matrix and the corresponding prestress into the machine learning model to be trained for training to obtain a target machine learning model.

[0008] Optionally, with the goal of minimizing the distribution difference between the source domain ultrasonic guided signal characteristic parameters and the target domain ultrasonic guided signal characteristic parameters, the source domain ultrasonic guided signal characteristic parameters are mapped to obtain a mapped source domain ultrasonic guided signal characteristic matrix, including: constructing a source domain ultrasonic guided signal characteristic matrix according to the source domain ultrasonic guided signal characteristic parameters, and constructing a target domain characteristic matrix according to the target domain ultrasonic guided signal characteristic parameters; determining a source domain characteristic mean vector according to the source domain ultrasonic guided signal characteristic parameters, and determining a target domain characteristic mean vector according to the target domain ultrasonic guided signal characteristic parameters; constructing a source domain characteristic covariance matrix according to the source domain ultrasonic guided signal characteristic matrix and the source domain characteristic mean vector, and constructing a target domain characteristic covariance matrix according to the target domain characteristic matrix and the target domain characteristic mean vector; determining a feature conversion matrix with the goal of minimizing the Frobenius norm between the converted source domain characteristic covariance matrix and the target domain characteristic covariance matrix; and obtaining a mapped source domain ultrasonic guided signal characteristic matrix according to the feature conversion matrix and the source domain ultrasonic guided signal characteristic matrix.

[0009] Optionally, a source domain feature covariance matrix and a target domain feature covariance matrix are constructed according to the source domain ultrasonic guide signal feature matrix, the target domain feature matrix, the source domain feature mean vector, and the target domain feature mean vector, including:

[0010]

[0011] Among them, C s and C t Represent the source domain covariance matrix and the target domain covariance matrix respectively, n s and n t Represent the total number of samples in the source domain and the target domain respectively, X s and X t They represent the source domain ultrasonic waveguide signal feature matrix and the target domain feature matrix, and their sizes are (ns ,d) and (n t ,d),d represents the feature dimension of each sample,μ s and μ t Represent the source domain feature mean vector and the target domain feature mean vector, respectively. x s,i and x t,i denote the i-th sample in the source domain and the target domain respectively.

[0012] Optionally, the characteristic parameters include multiple ones of ultrasonic guided wave group velocity, arrival time, peak value, peak-to-peak value, kurtosis factor, shape factor, peak factor, pulse factor, margin factor, peak frequency, frequency domain maximum value, center frequency, and root mean square frequency.

[0013] Optionally, an ultrasonic guided wave acquisition system is used to collect ultrasonic guided wave signals of the steel strands in service and ultrasonic guided wave signals of the steel strands under different prestresses in an experimental state, wherein the ultrasonic guided wave acquisition system is composed of a piezoelectric ceramic sensor, a signal generator, a power amplifier and an oscilloscope.

[0014] Optionally, the target machine learning model adopts an extreme gradient boosting model.

[0015] According to a second aspect, the present invention provides an in-service structure cross-domain prestress identification device based on ultrasonic guided waves, comprising: a signal acquisition module, used to collect ultrasonic guided wave signals of steel strands in service; a prediction module, used to input the ultrasonic guided wave signals of steel strands in service into a target machine learning model to obtain a predicted value of prestress of the steel strands in service, wherein the target machine learning model is trained based on the mapped source domain ultrasonic guided wave signals and the corresponding prestress as samples, the source domain ultrasonic guided wave signals are ultrasonic guided wave signals of steel strands under different prestresses in an experimental state, and the mapping process is to minimize the distribution difference between the characteristic parameters of the source domain ultrasonic guided wave signals and the characteristic parameters of the ultrasonic guided wave signals of the steel strands in service.

[0016] According to the third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves as described in the first aspect or any embodiment of the first aspect.

[0017] According to the fourth aspect, an embodiment of the present invention provides a computer storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves as described in the first aspect or any embodiment of the first aspect.

[0018] The embodiment of the present invention provides a method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves. By mapping the characteristic parameters of ultrasonic guided waves in the source domain (experimental state) with the characteristic parameters of the target domain (in-service state), the problem of feature distribution differences between the source domain and the target domain can be effectively solved. In the present invention, source domain data is usually easier to obtain. The features of the source domain data are aligned with the target domain through feature mapping. After the feature space is aligned, the two domains have similar distributions in the feature space. The mapped source domain features retain the relationship with the prestress and are distributed close to the target domain. Therefore, the supervision information (prestress label) on the source domain can be effectively migrated to the target domain. The trained model can predict the prestress of the target domain, thereby reducing the demand for samples in the target domain, solving the problem that it is difficult to determine the actual prestress of the steel strands inside the in-service structure, resulting in the inability to provide effective samples, reducing the difficulty of data collection, and improving the training efficiency of the model.

[0019] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art may be taught from the practice of the present invention. The objectives and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0021] Figure 1 It is a specific example flow chart of a method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves in the present invention;

[0022] Figure 2 It is a specific example flow chart of a training sample determination process of a target machine learning model in a method for cross-domain prestress identification of an in-service structure based on ultrasonic guided waves of the present invention;

[0023] Figure 3 is one of the feature parameter distributions of the source domain and the target domain before feature mapping in the present invention;

[0024] Figure 4 is a feature parameter distribution of one of the source domain and the target domain after feature mapping is performed in the present invention;

[0025] Figure 5 It is the target domain prestress recognition result output by the machine learning model trained with the initial source domain feature parameters in the present invention;

[0026] Figure 6 The target domain prestress recognition result output by the machine learning model trained with the new source domain feature parameters after performing feature mapping in the present invention;

[0027] Figure 7 Comparison of target domain prestress prediction errors output by the machine learning model before and after feature mapping in the present invention;

[0028] Figure 8 It is a principle block diagram of a specific example of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. 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.

[0030] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0031] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0032] The embodiment of the present invention provides a method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves. Figure 1 As shown, including:

[0033] S101, collecting ultrasonic waveguide signals of steel strands in service;

[0034] S102, inputting the ultrasonic guided wave signal of the steel strand in service into the target machine learning model to obtain the prestress prediction value of the steel strand in service, wherein the target machine learning model is trained based on the mapped source domain ultrasonic guided wave signal and the corresponding prestress as samples, the source domain ultrasonic guided wave signal is the ultrasonic guided wave signal of the steel strand under different prestresses in the experimental state, and the mapping process is to minimize the distribution difference between the characteristic parameters of the source domain ultrasonic guided wave signal and the characteristic parameters of the ultrasonic guided wave signal of the steel strand in service.

[0035] Exemplarily, the steel strand in service is a steel strand in an in-service structure in an actual project, and the ultrasonic guided wave acquisition system can be used to collect the ultrasonic guided wave signal of the steel strand, and the ultrasonic guided wave acquisition system can be composed of an electric ceramic sensor, a signal generator, a power amplifier and an oscilloscope. In the following description of this embodiment, the ultrasonic guided wave signal of the steel strand under different prestresses in the experimental state is defined as the source domain ultrasonic guided wave signal, and the ultrasonic guided wave signal of the steel strand in service is defined as the target domain ultrasonic guided wave signal.

[0036] The target machine learning model can be a trained extreme gradient boosting (XGBoost) model. Since the model training requires a large number of samples, however, in practical applications, only the target domain ultrasonic guided wave signal can be obtained, and it is difficult to determine the actual prestress of the steel strand inside the in-service structure. In this case, it is impossible to provide an effective sample label for the model. However, the source domain ultrasonic guided wave signal is obtained by tensioning a bare steel strand step by step in the laboratory and collecting it using an ultrasonic guided wave acquisition system. While collecting the source domain ultrasonic guided wave signal, the corresponding prestress of each source domain ultrasonic guided wave signal can be recorded, that is, the source domain ultrasonic guided wave signal and the corresponding prestress are both known quantities. Therefore, the sample for training the target machine learning model provided in this embodiment is the source domain ultrasonic guided wave signal characteristic parameter mapped with the goal of minimizing the distribution difference between the source domain ultrasonic guided wave signal characteristic parameter and the target domain ultrasonic guided wave signal characteristic parameter. That is, the mapped source domain ultrasonic guided wave signal characteristic parameter is used as a sample, and the prestress corresponding to the source domain ultrasonic guided wave signal is used as a sample label, which is input into the target machine learning model for training. The target machine learning model is supervised training, and its training process is a prior art, which will not be described in detail here.

[0037] The embodiment of the present invention provides a method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves. By mapping the characteristic parameters of ultrasonic guided waves in the source domain (experimental state) with the characteristic parameters of the target domain (in-service state), the problem of feature distribution differences between the source domain and the target domain can be effectively solved. In the present invention, source domain data is usually easier to obtain. The features of the source domain data are aligned with the target domain through feature mapping. After the feature space is aligned, the two domains have similar distributions in the feature space. The mapped source domain features retain the relationship with the prestress and are distributed close to the target domain. Therefore, the supervision information (prestress label) on the source domain can be effectively migrated to the target domain. The trained model can predict the prestress of the target domain, thereby reducing the demand for samples in the target domain, solving the problem that it is difficult to determine the actual prestress of the steel strands inside the in-service structure, resulting in the inability to provide effective samples, reducing the difficulty of data collection, and improving the training efficiency of the model.

[0038] As an optional implementation, the sample determination process of the target machine learning model, such as Figure 2As shown, including:

[0039] S201, collecting ultrasonic waveguide signals of steel strands under different prestresses in an experimental state as source domain ultrasonic waveguide signals, and collecting ultrasonic waveguide signals of steel strands in service as target domain ultrasonic waveguide signals;

[0040] S202, respectively extracting a plurality of target characteristic parameters of the source domain ultrasonic waveguide signal and the target domain ultrasonic waveguide signal to obtain characteristic parameters of the source domain ultrasonic waveguide signal and characteristic parameters of the target domain ultrasonic waveguide signal;

[0041] S203, with the goal of minimizing the distribution difference between the source domain ultrasonic conductive signal characteristic parameters and the target domain ultrasonic conductive signal characteristic parameters, mapping the source domain ultrasonic conductive signal characteristic parameters to obtain a mapped source domain ultrasonic conductive signal characteristic matrix;

[0042] S204, inputting the mapped source domain ultrasonic waveguide signal feature matrix and the corresponding prestress into the machine learning model to be trained to obtain a target machine learning model.

[0043] Exemplarily, the target characteristic parameters include ultrasonic guided wave group velocity, arrival time, peak value, peak-to-peak value, kurtosis factor, shape factor, peak factor, pulse factor, margin factor, peak frequency, frequency domain maximum value, center frequency, and root mean square frequency. The method of extracting multiple target characteristic parameters of the source domain ultrasonic guided wave signal and the target domain ultrasonic guided wave signal can be that when the target characteristic parameter is a peak value, the maximum amplitude of the signal within a period of time is calculated; when the target characteristic parameter is a peak-to-peak value, the difference between the maximum value and the minimum value of the signal within a period of time can be calculated; when the target characteristic parameter is a kurtosis factor, the fourth-order normalized moment of the signal can be calculated to reflect the sharpness of the signal distribution, etc. This embodiment only gives some examples of target characteristic parameter extraction, and the extraction process is all prior art and will not be repeated.

[0044] After extracting the characteristic parameters of the source domain ultrasonic guided signal and the target domain ultrasonic guided signal, it is necessary to perform feature mapping on the characteristic parameters of the source domain ultrasonic guided signal so that their distribution is as close as possible to the characteristic parameters of the target domain ultrasonic guided signal. Specifically, a source domain ultrasonic guided signal feature matrix is ​​constructed according to the characteristic parameters of the source domain ultrasonic guided signal, and a target domain feature matrix is ​​constructed according to the characteristic parameters of the target domain ultrasonic guided signal; a source domain feature mean vector is determined according to the characteristic parameters of the source domain ultrasonic guided signal, and a target domain feature mean vector is determined according to the characteristic parameters of the target domain ultrasonic guided signal; a source domain feature covariance matrix is ​​constructed according to the source domain ultrasonic guided signal feature matrix and the source domain feature mean vector, and a target domain feature covariance matrix is ​​constructed according to the target domain feature matrix and the target domain feature mean vector; a feature conversion matrix is ​​determined with the goal of minimizing the Frobenius norm between the converted source domain feature covariance matrix and the target domain feature covariance matrix; a mapped source domain ultrasonic guided signal feature matrix is ​​obtained according to the feature conversion matrix and the source domain ultrasonic guided signal feature matrix.

[0045] Among them, according to the source domain ultrasonic guide signal feature matrix, the target domain feature matrix, the source domain feature mean vector and the target domain feature mean vector, the source domain feature covariance matrix and the target domain feature covariance matrix are constructed, including:

[0046]

[0047] Among them, C s and C t Represent the source domain covariance matrix and the target domain covariance matrix respectively, n s and n t Represent the total number of samples in the source domain and the target domain respectively, X s and X t They represent the source domain ultrasonic waveguide signal feature matrix and the target domain feature matrix, and their sizes are (n s ,d) and (n t ,d),d represents the feature dimension of each sample,μ s and μ t Represent the source domain feature mean vector and the target domain feature mean vector, respectively. x s,i and x t,i denote the i-th sample in the source domain and the target domain respectively.

[0048] Minimizing the feature distribution of the source domain and the target domain can be transformed into the following problem, that is, finding a feature transformation matrix A so that the Frobenius norm between the transformed source domain feature covariance matrix and the target domain feature covariance matrix is ​​minimized:

[0049]

[0050] in, represents the Frobenius norm.

[0051] After obtaining the feature conversion matrix, the mapped source domain ultrasonic waveguide signal feature matrix can be obtained. The mapped source domain ultrasonic waveguide signal feature matrix and the corresponding prestress are input into the machine learning model to be trained to obtain the target machine learning model.

[0052] In this embodiment, before and after feature mapping is performed, the distribution of one of the source domain ultrasonic waveguide signal feature parameters and the target domain ultrasonic waveguide signal feature parameters is as follows: Figure 3 and Figure 4 As shown in . For one of the characteristic parameters of the source domain ultrasonic waveguide signal, such as Figure 3 As shown in , the features are mainly concentrated in the range of 0-0.005, while in the distribution of one of the characteristic parameters of the target domain ultrasonic waveguide signal, its main distribution interval is 0.004-0.012. For the mapped features, such as Figure 4 As shown, the distributions of the characteristic parameters of the ultrasonic waveguide signal in the source domain and the characteristic parameters of the ultrasonic waveguide signal in the target domain are very close.

[0053] In this embodiment, the target domain ultrasonic wave guide signal prestress recognition result output by the machine learning model trained with the source domain ultrasonic wave guide signal characteristic parameters is as follows: Figure 5 As shown in the figure, the dotted line represents the ideal situation where the actual value of prestress is exactly the same as the predicted value. In this case, the deviation between the predicted value and the actual value is very large. The target domain prestress recognition result output by the machine learning model trained with the characteristic parameters of the source domain ultrasonic guide signal after feature mapping is shown in Figure 6 As shown in Figure 2, in this case, the prestress prediction value is significantly closer to the true value. The comparison of the target domain prestress prediction error output by the machine learning model before and after feature mapping is shown in Figure 2. Figure 7 As shown, after executing feature mapping, the prediction error of prestress in the target domain is only 47% of that in the case where feature mapping is not executed. It can be seen that the method provided by the present invention can effectively predict the cross-domain prestress of in-service structures.

[0054] The embodiment of the present invention provides a device for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves, comprising:

[0055] A signal acquisition module, used to collect ultrasonic waveguide signals of steel strands in service;

[0056] The prediction module is used to input the ultrasonic guided wave signal of the steel strand in service into the target machine learning model to obtain the prestress prediction value of the steel strand in service, wherein the target machine learning model is trained based on the mapped source domain ultrasonic guided wave signal and the corresponding prestress as samples, the source domain ultrasonic guided wave signal is the ultrasonic guided wave signal of the steel strand under different prestresses in the experimental state, and the mapping process is to minimize the distribution difference between the characteristic parameters of the source domain ultrasonic guided wave signal and the characteristic parameters of the ultrasonic guided wave signal of the steel strand in service.

[0057] The present application also provides an electronic device, such as Figure 8 As shown, a processor 501 and a memory 502, wherein the processor 501 and the memory 502 may be connected via a bus or other means.

[0058] The processor 501 may be a central processing unit (CPU). The processor 501 may also be 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, or a combination of the above chips.

[0059] The memory 502, as a non-transient computer-readable storage medium, can be used to store non-transient software programs, non-transient computer executable programs and modules, such as program instructions / modules corresponding to the cross-domain prestress identification method for in-service structures based on ultrasonic guided waves in an embodiment of the present invention. The processor executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory.

[0060] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 502 may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0061] The one or more modules are stored in the memory 502, and when executed by the processor 501, the following is performed: Figure 1 The cross-domain prestress identification method of an in-service structure based on ultrasonic guided waves in the illustrated embodiment.

[0062] For details of the above electronic equipment, please refer to Figure 1 The corresponding related descriptions and effects in the illustrated embodiments can be understood and will not be repeated here.

[0063] This embodiment also provides a computer storage medium, which stores computer executable instructions, and the computer executable instructions can execute the cross-domain prestress identification method of in-service structures based on ultrasonic guided waves in any of the above method embodiments. Wherein, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0064] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A method for identifying cross-domain prestress in an in-service structure based on ultrasonic guided waves, characterized in that: include: Collect ultrasonic waveguide signals of steel strands in service; The ultrasonic guided wave signal of the steel strand in service is input into the target machine learning model to obtain the prestress prediction value of the steel strand in service, wherein the target machine learning model is trained based on the mapped source domain ultrasonic guided wave signal and the corresponding prestress as samples. The source domain ultrasonic guided wave signal is the ultrasonic guided wave signal of the steel strand under different prestresses in the experimental state. The mapping process is to minimize the distribution difference between the characteristic parameters of the source domain ultrasonic guided wave signal and the characteristic parameters of the ultrasonic guided wave signal of the steel strand in service.

2. According to claim 1, a method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves is characterized in that: The sample determination process of the target machine learning model includes: The ultrasonic waveguide signals of the steel strands under different prestress in the experimental state are collected as the source domain ultrasonic waveguide signals, and the ultrasonic waveguide signals of the steel strands under service are collected as the target domain ultrasonic waveguide signals; Extracting multiple target characteristic parameters of the source domain ultrasonic waveguide signal and the target domain ultrasonic waveguide signal respectively, and obtaining the source domain ultrasonic waveguide signal characteristic parameters and the target domain ultrasonic waveguide signal characteristic parameters; Taking the distribution difference between the characteristic parameters of the ultrasonic conductive signal in the source domain and the characteristic parameters of the ultrasonic conductive signal in the target domain as the goal, the characteristic parameters of the ultrasonic conductive signal in the source domain are mapped to obtain a characteristic matrix of the ultrasonic conductive signal in the source domain after mapping; The mapped source domain ultrasonic waveguide signal feature matrix and the corresponding prestress are input into the machine learning model to be trained to obtain the target machine learning model.

3. The method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves according to claim 2 is characterized in that: With the goal of minimizing the distribution difference between the characteristic parameters of the ultrasonic conductive signal in the source domain and the characteristic parameters of the ultrasonic conductive signal in the target domain, the characteristic parameters of the ultrasonic conductive signal in the source domain are mapped to obtain the characteristic matrix of the ultrasonic conductive signal in the source domain after mapping, including: constructing a source domain ultrasonic waveguide signal feature matrix according to the source domain ultrasonic waveguide signal feature parameters, and constructing a target domain feature matrix according to the target domain ultrasonic waveguide signal feature parameters; Determine a source domain feature mean vector according to the source domain ultrasonic wave guide signal feature parameters, and determine a target domain feature mean vector according to the target domain ultrasonic wave guide signal feature parameters; A source domain feature covariance matrix is ​​constructed according to a source domain ultrasonic guide signal feature matrix and a source domain feature mean vector, and a target domain feature covariance matrix is ​​constructed according to a target domain feature matrix and a target domain feature mean vector; The feature conversion matrix is ​​determined with the goal of minimizing the Frobenius norm between the converted source domain feature covariance matrix and the target domain feature covariance matrix; According to the characteristic conversion matrix and the characteristic matrix of the source domain ultrasonic waveguide signal, the mapped characteristic matrix of the source domain ultrasonic waveguide signal is obtained.

4. The method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves according to claim 3 is characterized in that: According to the source domain ultrasonic guide signal feature matrix, the target domain feature matrix, the source domain feature mean vector and the target domain feature mean vector, the source domain feature covariance matrix and the target domain feature covariance matrix are constructed, including: Among them, C s and C t Represent the source domain covariance matrix and the target domain covariance matrix respectively, n s and n t Represent the total number of samples in the source domain and the target domain respectively, X s and X t They represent the source domain ultrasonic waveguide signal feature matrix and the target domain feature matrix, and their sizes are (n s ,d) and (n t ,d),d represents the feature dimension of each sample,μ s and μ t Represent the source domain feature mean vector and the target domain feature mean vector, respectively. x s,i and x t,i denote the i-th sample in the source domain and the target domain respectively.

5. A method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves according to any one of claims 1 to 4, characterized in that: The characteristic parameters include multiple ones of ultrasonic guided wave group velocity, arrival time, peak value, peak-to-peak value, kurtosis factor, shape factor, peak factor, pulse factor, margin factor, peak frequency, frequency domain maximum value, center frequency, and root mean square frequency.

6. A method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves according to any one of claims 1 to 4, characterized in that: An ultrasonic guided wave acquisition system is used to collect ultrasonic guided wave signals of steel strands in service and ultrasonic guided wave signals of steel strands under different prestresses in an experimental state. The ultrasonic guided wave acquisition system consists of a piezoelectric ceramic sensor, a signal generator, a power amplifier and an oscilloscope.

7. A method for identifying cross-domain prestress of an in-service structure based on ultrasonic guided waves according to any one of claims 1 to 4, characterized in that: The target machine learning model adopts the extreme gradient boosting model.

8. A device for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves, characterized in that: include: A signal acquisition module, used to collect ultrasonic waveguide signals of steel strands in service; The prediction module is used to input the ultrasonic guided wave signal of the steel strand in service into the target machine learning model to obtain the prestress prediction value of the steel strand in service, wherein the target machine learning model is trained based on the mapped source domain ultrasonic guided wave signal and the corresponding prestress as samples, the source domain ultrasonic guided wave signal is the ultrasonic guided wave signal of the steel strand under different prestresses in the experimental state, and the mapping process is to minimize the distribution difference between the characteristic parameters of the source domain ultrasonic guided wave signal and the characteristic parameters of the ultrasonic guided wave signal of the steel strand in service.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves as described in any one of claims 1 to 7.

10. A computer storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by the processor, the steps of the method for identifying cross-domain prestress of in-service structures based on ultrasonic guided waves as described in any one of claims 1-7 are implemented.

Images