Acoustic emission microcrack type prediction method and device, storage medium and electronic equipment

Abstract

The invention discloses an acoustic emission microcrack type prediction method and device, a storage medium and electronic equipment, and belongs to the technical field of civil engineering artificial intelligence. The method comprises the steps that all components of moment tensor decomposition of a microcrack formed by a to-be-recognized test sample, fracture surface parameter data and waveform data of the microcrack of the to-be-recognized test sample are obtained, the to-be-recognized test sample is subjected to an acoustic emission test, the microcrack is formed in the to-be-recognized test sample, and the waveform data of the microcrack of the to-be-recognized test sample are obtained. The microcrack is an intermediate-form microcrack when the acoustic emission test is not completed; and inputting the waveform data of the microcracks into a deep learning model for identification to obtain a microcrack prediction result of the test sample to be identified. The device, the storage medium and the electronic equipment can be used for realizing the method. According to the method, the applicability of a moment tensor analysis result can be improved, and the micro-crack type predicted through waveform data is explored.

Description

Technical Field

[0001] The present invention relates to the field of civil engineering artificial intelligence technology, and in particular to an acoustic emission microcrack type prediction method, device, storage medium and electronic equipment. Background Art

[0002] With the development of underground resources and the utilization of underground space, engineering safety issues are becoming increasingly prominent. As a powerful non-destructive testing method, acoustic emission technology can play an important role in ensuring the safety of underground projects. The essence of the failure of quasi-brittle materials such as rocks in underground projects is the process of specimen destruction caused by the initiation, expansion, and penetration of internal microcracks to form macro cracks. The formation and expansion of microcracks, as an irreversible energy dissipation process, will release the surface energy of the crack in the form of transient elastic waves. This phenomenon is called acoustic emission. The acoustic emission signal is actually a transient elastic wave, which is generated by the internal damage or deformation of the material itself. Therefore, the acoustic emission signal itself contains rich information about the damage process of brittle materials. By processing and analyzing the acoustic emission signal, many scholars have achieved rich results in acoustic emission positioning, failure mechanism, AE characteristics of rock instability precursors, and rock damage evolution laws. These results are undoubtedly of great significance to the prevention and control of rock engineering disasters and the study of the mechanical properties of brittle materials such as rocks. Summary of the invention

[0003] In view of this, the present invention provides an acoustic emission microcrack type prediction method, device, storage medium and electronic device, which can improve the applicability of moment tensor analysis results and explore the prediction of microcrack types through waveform data, thereby being more suitable for practical use.

[0004] In order to achieve the above first object, the technical solution of the acoustic emission microcrack type prediction method provided by the present invention is as follows:

[0005] The acoustic emission microcrack type prediction method provided by the present invention comprises the following steps:

[0006] Obtaining components of moment tensor decomposition of microcracks formed on the test sample to be identified, fracture surface parameter data, and waveform data of microcracks of the test sample to be identified, wherein the test sample to be identified undergoes an acoustic emission test to form microcracks on the test sample to be identified, and the microcracks are intermediate morphological microcracks when the acoustic emission test is not completed;

[0007] The waveform data of the microcracks is input into a deep learning model for identification to obtain a prediction result of the microcracks of the test sample to be identified.

[0008] The acoustic emission microcrack type prediction method provided by the present invention can be further implemented by adopting the following technical measures.

[0009] Preferably, the fracture surface parameter data of the microcracks formed in the test sample to be identified includes the fracture surface parameter trend The inclination angle δ, the sliding direction ζ and the inclination angle θ.

[0010] Preferably, the waveform data of the microcracks of the test sample to be identified is obtained by screening the timestamp of the acoustic emission positioning data.

[0011] Preferably, the method for establishing the deep learning model comprises the following steps:

[0012] Obtaining the components of moment tensor decomposition, fracture surface parameter data, and microcrack waveform data of each sample in the test sample training set, wherein each test sample in the test sample training set undergoes an acoustic emission test to form microcracks on each test sample, and the microcracks are the final morphology of microcracks when the acoustic emission test is completed;

[0013] The deep learning model is obtained based on the components of the moment tensor decomposition of each sample, the fracture surface parameter data, the microcrack waveform data of each sample, and the final morphological microcracks when the acoustic emission test is completed.

[0014] Preferably, the acoustic emission microcrack type prediction method further comprises the following steps:

[0015] Obtaining the final microcracks of the test sample to be identified when the acoustic emission test is completed;

[0016] The components of moment tensor decomposition of microcracks of the test sample to be identified during the acoustic emission test, fracture surface parameter data, waveform data of microcracks, and the final morphological microcracks when the acoustic emission test is completed are added to the test sample training set.

[0017] Preferably, the deep learning model includes MolienetV2, InceptionV3, Resnet50 and VGG16, and the acoustic emission microcrack type prediction method also includes selecting the optimal model by applying the test set test specimen, wherein the test method includes the following steps:

[0018] Obtaining the components of moment tensor decomposition, fracture surface parameter data, and microcrack waveform data of each sample in the test sample test set, wherein each test sample in the test sample training set undergoes an acoustic emission test to form microcracks on each test sample, and the microcracks include intermediate morphology microcracks when the acoustic emission test is not completed, and final morphology microcracks when the acoustic emission test is completed;

[0019] Inputting the microcrack waveform data of each sample corresponding to the intermediate form of each sample when the acoustic emission test is not completed into different deep learning models for identification, and obtaining the microcrack prediction result of each sample in the test sample test set;

[0020] The predicted microcrack results of each sample are compared with the final morphological microcracks when the acoustic emission test is completed to obtain a comparison result, and the deep learning model with the best degree of consistency between the comparison results is used as the optimal model.

[0021] Preferably, the acoustic emission microcrack type prediction method further comprises performing data processing on the waveform of the microcrack, and the data processing comprises the following steps:

[0022] De-noising the waveform of the microcracks to obtain a waveform diagram of the microcracks after de-noising;

[0023] Performing continuous wavelet transform on the denoised microcrack waveform to obtain a CWT waveform diagram of the microcrack;

[0024] Performing wavelet packet transform on the denoised microcrack waveform to obtain a WPT waveform diagram of the microcrack;

[0025] The waveform of the microcracks after data processing is obtained by sorting out the denoised waveform of the microcracks, the CWT waveform of the microcracks, and the WPT waveform of the microcracks.

[0026] In order to achieve the above second purpose, the technical solution of the acoustic emission microcrack type prediction device provided by the present invention is as follows:

[0027] The acoustic emission microcrack type prediction device provided by the present invention comprises:

[0028] A data acquisition module, used to acquire each component of moment tensor decomposition of microcracks formed in the test sample to be identified, fracture surface parameter data, and waveform data of microcracks of the test sample to be identified, wherein the test sample to be identified undergoes an acoustic emission test to form microcracks on the test sample to be identified, and the microcracks are intermediate morphological microcracks when the acoustic emission test is not completed;

[0029] The microcrack type prediction module is used to input the waveform data of the microcracks into the deep learning model for identification, so as to obtain the microcrack prediction result of the test sample to be identified.

[0030] In order to achieve the third objective above, the technical solution of the computer-readable storage medium provided by the present invention is as follows:

[0031] The computer-readable storage medium provided by the present invention stores an acoustic emission microcrack type prediction program, and when the acoustic emission microcrack type prediction program is executed by a processor, the steps of the acoustic emission microcrack type prediction method provided by the present invention are implemented.

[0032] In order to achieve the fourth objective, the technical solution of the electronic device provided by the present invention is as follows:

[0033] The electronic device provided by the present invention comprises a memory and a processor, wherein the memory stores an acoustic emission microcrack type prediction program, and when the acoustic emission microcrack type prediction program is executed by the processor, the steps of the acoustic emission microcrack type prediction method provided by the present invention are implemented.

[0034] The acoustic emission crack type prediction method, device, storage medium and electronic device provided by the present invention can quantitatively calculate the fracture surface parameters and qualitatively distinguish the micro-fracture type through moment tensor analysis; by using the denoised image data as the input of the deep learning model, the damage in the micro-fracture information transmission process in the acoustic emission signal can be avoided as much as possible; by comparing the training results of each model to obtain the best combination of image input and training model type, a high-precision micro-fracture type recognition model with an image as input can be saved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0036] Figure 1 A flowchart of the steps of the method for predicting microcrack types using acoustic emission provided by an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of the signal flow relationship between the functional modules in the acoustic emission microcrack type prediction device provided by an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of the structure of an acoustic emission microcrack type prediction device in a hardware operating environment provided by an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of a deep composite structure rock sample model involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0040] Figure 5a Schematic diagram of a numerical simulation model of a deep composite structure rock mass involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention (tensile);

[0041] Figure 5b A schematic diagram of a numerical simulation model of a deep composite structure rock mass involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention (mixed);

[0042] Figure 5c Schematic diagram of a numerical simulation model of a deep composite structure rock mass involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention (shear);

[0043] Figure 6a A numerical simulation model diagram (original waveform) of a deep composite structure rock mass after adding cementation involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0044] Figure 6b A numerical simulation model diagram of a deep composite structure rock mass after adding cementation (denoised waveform) involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0045] Figure 6c A numerical simulation model diagram (CWT diagram) of a deep composite structure rock mass after adding cementation involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0046] Figure 6d A numerical simulation model diagram (spectrum diagram) of a deep composite structure rock mass after adding cementation involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0047] Figure 7a A typical deep composite structure numerical simulation model diagram (waveform diagram) involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0048] Figure 7b A typical deep composite structure numerical simulation model diagram (CWT diagram) involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0049] Figure 7c A typical deep composite structure numerical simulation model diagram (spectrum diagram) involved in the acoustic emission microcrack type prediction method provided in an embodiment of the present invention;

[0050] Among them, Figure 7a , Figure 7b , Figure 7c In statistics, the F1 value is an indicator used to measure the accuracy of a binary classification model. It takes into account both the precision and recall of the classification model. The F1 value can be regarded as a harmonic mean of the model's precision and recall, with a maximum value of 1 and a minimum value of 0.

[0051] MolienetV2 neural network: It is an efficient and lightweight convolutional neural network architecture launched by the Google research team in 2018. It is an upgraded version of MolienetV1. Its main goal is to reduce the model size and computational cost while maintaining high accuracy. It is particularly suitable for running on mobile devices and resource-constrained environments.

[0052] InceptionV3 network: It is the third generation of Inception network developed by Google. Its main features are the use of multi-branch structure for parallel calculation and the reduction of calculation amount and number of parameters by decomposing large convolution kernels.

[0053] ResNet50 network: It is a deep convolutional neural network based on the residual network architecture, proposed by Kaiming He et al. of Microsoft Research in 2015. The core innovation of ResNet50 is the introduction of residual blocks, which solves the gradient vanishing and degradation problems in deep networks through skip connections, allowing the network to reach deeper levels without overfitting or gradient problems;

[0054] VGG16 network: It is a deep convolutional neural network architecture proposed by the Visual Geometry Group of the University of Oxford in 2014. VGG16 achieved excellent results in the 2014 ImageNet (ImageNet is a landmark large-scale image dataset widely used in deep learning research and computer vision competitions for visual recognition tasks) image classification competition, proving the powerful ability of deep convolutional neural networks in image classification tasks. DETAILED DESCRIPTION

[0055] In order to solve the problems existing in the prior art, the present invention provides an acoustic emission microcrack type prediction method, device, storage medium and electronic equipment, which can improve the applicability of moment tensor analysis results and explore the prediction of microcrack types through waveform data, thereby being more suitable for practical use.

[0056] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the acoustic emission microcrack type prediction method, device, storage medium and electronic device proposed by the present invention, its specific implementation, structure, features and effects in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "embodiment" does not necessarily refer to the same embodiment. In addition, the features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0057] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B. Specifically, it is understood that: A and B may be included at the same time, A may exist alone, or B may exist alone, and any of the above three situations may be met.

[0058] Acoustic Emission Microcrack Type Prediction Method

[0059] See attached Figure 1 The acoustic emission microcrack type prediction method provided by the present invention comprises the following steps:

[0060] Step S1: Obtain the components of the moment tensor decomposition of the microcracks formed on the test sample to be identified, the fracture surface parameter data, and the waveform data of the microcracks of the test sample to be identified, wherein the test sample to be identified undergoes an acoustic emission test to form microcracks on the test sample to be identified, and the microcracks are intermediate morphological microcracks when the acoustic emission test is not completed.

[0061] Specifically, the acoustic emission microcrack type prediction method provided by the embodiment of the present invention predicts the final form of microcracks when the acoustic emission test is completed based on the intermediate form of microcracks when the acoustic emission test is not completed. Its significance lies in that microcracks may occur in quasi-brittle materials such as rocks in underground projects. In the early stage of the microcracks, no substantial damage or engineering accidents are caused. Some cracks will not cause substantial damage or engineering accidents even if they develop to the final state, while some cracks will cause substantial damage or engineering accidents after continuing to develop. Therefore, in the process of microcrack development, it is particularly important to predict the final state of microcracks, which can provide an important theoretical basis for actively taking repair measures.

[0062] Step S2: Input the waveform data of the microcracks into the deep learning model for identification, and obtain the prediction results of the microcracks of the test sample to be identified.

[0063] Specifically, in the process of using deep learning models to predict the types of microcracks that have occurred in the test specimens to be identified, the microcrack waveform data are all feature vectors. Based on these feature vectors and the deep learning model trained with training samples, the final state of the ongoing microcracks can be obtained. In this case, the more training samples there are and the more types there are, the higher the accuracy of the prediction.

[0064] The acoustic emission crack type prediction method provided by the present invention can quantitatively calculate the fracture surface parameters and qualitatively distinguish the micro-fracture type through moment tensor analysis; by using the denoised image data as the input of the deep learning model, the damage in the micro-fracture information transmission process in the acoustic emission signal can be avoided as much as possible; by comparing the training results of each model to obtain the best combination of image input and training model type, a high-precision micro-fracture type recognition model with an image as input can be saved.

[0065] Among them, the fracture surface parameter data of the microcracks formed in the test sample to be identified includes the fracture surface parameter trend The inclination angle δ, the sliding direction ζ and the inclination angle θ.

[0066] The waveform data of the microcracks of the test sample to be identified is obtained by screening the timestamp of the acoustic emission positioning data.

[0067] The method for establishing a deep learning model includes the following steps:

[0068] Obtaining the components of moment tensor decomposition, fracture surface parameter data, and microcrack waveform data of each sample in the test sample training set, wherein each test sample in the test sample training set is subjected to an acoustic emission test to form microcracks on each test sample, and the microcracks are the final morphology of microcracks when the acoustic emission test is completed;

[0069] A deep learning model is obtained based on the components of the moment tensor decomposition of each sample, the fracture surface parameter data, the microcrack waveform data of each sample, and the final morphological microcracks when the acoustic emission test is completed.

[0070] In this case, as the number of samples in the test specimen training set increases, the characteristic vector values ​​of the microcracks of different samples become more diverse. When predicting the microcracks of the test specimens to be identified, the reference samples that can be used for comparison and prediction are also more diverse. Therefore, the prediction results of the final state of the cracks are also more accurate.

[0071] The acoustic emission microcrack type prediction method further includes the following steps:

[0072] Obtain the final microcracks of the test specimen to be identified after the acoustic emission test is completed;

[0073] The components of the moment tensor decomposition of the microcracks of the test sample to be identified during the acoustic emission test, the fracture surface parameter data, the waveform data of the microcracks, and the final morphological microcracks when the acoustic emission test is completed are added to the test sample training set.

[0074] In this case, each identified test sample can be added to the test sample training set to form an increased training sample. Therefore, the test sample training set of the acoustic emission crack type prediction method provided by the present invention has the ability to expand samples. Therefore, as the number of identifications of the test sample to be identified increases, the identification accuracy of future test samples to be identified can be further increased.

[0075] Among them, the deep learning models include MolienetV2, InceptionV3, Resnet50 and VGG16, and the acoustic emission microcrack type prediction method also includes selecting the optimal model by applying the test set test specimens, wherein the test method includes the following steps:

[0076] Obtaining the components of moment tensor decomposition, fracture surface parameter data, and microcrack waveform data of each sample in the test sample test set, wherein each test sample in the test sample training set is subjected to an acoustic emission test to form microcracks on each test sample, and the microcracks include intermediate microcracks when the acoustic emission test is not completed, and final microcracks when the acoustic emission test is completed;

[0077] The microcrack waveform data of each sample corresponding to the intermediate form of each sample before the acoustic emission test is completed are input into different deep learning models for recognition, and the microcrack prediction results of each sample in the test sample test set are obtained;

[0078] The predicted microcrack results of each sample are compared with the final morphological microcracks when the acoustic emission test is completed to obtain the comparison results, and the deep learning model with the best degree of consistency in the comparison results is selected as the optimal model.

[0079] In this case, since there are many types of deep learning models, the models provided in this embodiment include MolienetV2, InceptionV3, Resnet50 and VGG16. The internal logic of different deep learning models is not exactly the same. Therefore, the final morphological microcracks obtained by predicting the intermediate morphological microcracks of the same unfinished acoustic emission test are not exactly the same. Therefore, by using the test sample test set, a model that better matches the sample results in the test sample test set can be selected from the above-mentioned models MolienetV2, InceptionV3, Resnet50 and VGG16, thereby making the model that better matches the test sample to be identified, and making the model that better fits the prediction results of the test sample to be identified.

[0080] The acoustic emission microcrack type prediction method further includes data processing of the microcrack waveform, and the data processing includes the following steps:

[0081] De-noising the waveform of the micro-cracks to obtain a denoised waveform diagram of the micro-cracks;

[0082] Perform continuous wavelet transform on the denoised microcrack waveform to obtain the CWT waveform of the microcrack;

[0083] Perform wavelet packet transform on the denoised microcrack waveform to obtain the WPT waveform of the microcrack;

[0084] The waveform of the microcracks after data processing is obtained by sorting out the denoised waveform of the microcracks, the CWT waveform of the microcracks, and the WPT waveform of the microcracks.

[0085] In this case, by sorting out the denoised microcrack waveform, the CWT waveform of the microcrack, and the WPT waveform of the microcrack, data that is more friendly to the deep learning model can be obtained. Therefore, more uncertain factors can be eliminated from the prediction of the acoustic emission crack type, making the prediction results more accurate.

[0086] Acoustic Emission Microcrack Type Prediction Device

[0087] The acoustic emission microcrack type prediction device provided by the present invention comprises:

[0088] The data acquisition module is used to obtain the components of the moment tensor decomposition of the microcracks formed in the test sample to be identified, the fracture surface parameter data, and the waveform data of the microcracks of the test sample to be identified, wherein the test sample to be identified undergoes an acoustic emission test to form microcracks on the test sample to be identified, and the microcracks are intermediate morphology microcracks when the acoustic emission test is not completed.

[0089] Specifically, the acoustic emission microcrack type prediction method provided by the embodiment of the present invention predicts the final form of microcracks when the acoustic emission test is completed based on the intermediate form of microcracks when the acoustic emission test is not completed. Its significance lies in that microcracks may occur in quasi-brittle materials such as rocks in underground projects. In the early stage of the microcracks, no substantial damage or engineering accidents are caused. Some cracks will not cause substantial damage or engineering accidents even if they develop to the final state, while some cracks will cause substantial damage or engineering accidents after continuing to develop. Therefore, in the process of microcrack development, it is particularly important to predict the final state of microcracks, which can provide an important theoretical basis for actively taking repair measures.

[0090] The microcrack type prediction module is used to input the waveform data of the microcracks into the deep learning model for identification, and obtain the microcrack prediction results of the test sample to be identified.

[0091] Specifically, in the process of using the deep learning model to predict the type of microcracks that have occurred in the test specimen to be identified, the components of the moment tensor decomposition of the microcracks, the fracture surface parameter data, and the waveform data of the microcracks are all feature vectors. Based on these feature vectors and the deep learning model trained with training samples, the final state of the ongoing microcracks can be obtained. In this case, the more training samples and the more types they have, the higher the accuracy of the prediction.

[0092] The acoustic emission crack type prediction device provided by the present invention can quantitatively calculate the fracture surface parameters and qualitatively distinguish the micro-fracture type through moment tensor analysis; by using the denoised image data as the input of the deep learning model, the damage in the micro-fracture information transmission process in the acoustic emission signal can be avoided as much as possible; by comparing the training results of each model to obtain the best combination of image input and training model type, a high-precision micro-fracture type recognition model with an image as input can be saved.

[0093] Computer readable storage medium

[0094] The computer-readable storage medium provided by the present invention stores an acoustic emission microcrack type prediction program, and when the acoustic emission microcrack type prediction program is executed by a processor, the steps of the acoustic emission microcrack type prediction method provided by the present invention are implemented.

[0095] The computer-readable storage medium provided by the present invention can quantitatively calculate the fracture surface parameters and qualitatively distinguish the microfracture type through moment tensor analysis; by using the denoised image data as the input of the deep learning model, the damage in the process of microfracture information transmission in the acoustic emission signal can be avoided as much as possible; by comparing the training results of each model to obtain the best combination of image input and training model type, a high-precision microfracture type recognition model with an image as input can be saved.

[0096] Electronic devices

[0097] The electronic device provided by the present invention comprises a memory and a processor. The memory stores an acoustic emission microcrack type prediction program. When the acoustic emission microcrack type prediction program is executed by the processor, the steps of the acoustic emission microcrack type prediction method provided by the present invention are implemented.

[0098] The electronic device provided by the present invention can quantitatively calculate the fracture surface parameters and qualitatively distinguish the type of microfracture through moment tensor analysis; by using the denoised image data as the input of the deep learning model, the damage in the process of microfracture information transmission in the acoustic emission signal can be avoided as much as possible; by comparing the training results of each model to obtain the best combination of image input and training model type, a high-precision microfracture type recognition model with an image as input can be saved.

[0099] Reference Figure 3 , Figure 3 It is a schematic diagram of the structure of an acoustic emission microcrack type prediction device in the hardware operating environment involved in the embodiment of the present invention.

[0100] like Figure 3 As shown, the acoustic emission microcrack type prediction device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) memory, or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0101] Those skilled in the art will understand that Figure 3 The structure shown in the figure does not constitute a limitation on the acoustic emission microcrack type prediction device, and may include more or less components than those shown in the figure, or combine certain components, or arrange the components differently.

[0102] like Figure 3 As shown, the memory 1005 as a storage medium may include an operating system, a data storage module, a network communication module, a user interface module, and an acoustic emission microcrack type prediction program.

[0103] exist Figure 3In the acoustic emission microcrack type prediction device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the acoustic emission microcrack type prediction device of the present invention can be set in the acoustic emission microcrack type prediction device, and the acoustic emission microcrack type prediction device calls the acoustic emission microcrack type prediction program stored in the memory 1005 through the processor 1001, and executes the acoustic emission microcrack type prediction method provided by the embodiment of the present invention.

[0104] Example

[0105] This application provides an acoustic emission microcrack type identification method that integrates moment tensor and deep learning, such as Figure 1 As shown, the following steps are included:

[0106] S1 Prepare indoor test specimens according to the designed test and determine the specimen size and the sound wave velocity in the specimen;

[0107] Specifically, the indoor test sample is prepared according to the designed test, and the sample size and the sound wave velocity and wave speed in the sample are determined, including:

[0108] S1.1 Prepare standard size specimens according to the test materials and size requirements. The uniaxial compression test size is 50 mm in diameter and 100 mm in height.

[0109] S1.2 Determine the weight, size and sound wave velocity of the specimen by using a balance, vernier caliper and ultrasonic velocimeter to prepare for subsequent acoustic emission tests, such as Figure 2 shown.

[0110] S2 conducts indoor acoustic emission tests and collects acoustic emission monitoring data;

[0111] Specifically, the indoor acoustic emission test and the collection of acoustic emission monitoring data include:

[0112] S2.1 Design the coordinates of the acoustic emission sensors in the test, connect the sensors to the amplifier and collector through wiring, and lay the foundation for acoustic emission monitoring during the loading test;

[0113] S2.2 sets the relevant parameters and acquisition channels in the PAC system, acquisition targets, etc., and the sensor coordinates. In the experiment, the sampling frequency is set to 1MHz, the noise threshold is set to 45dB, and the waveform length is set to 1024μs; the impact lockout time (HLT), the time interval for closing the measurement circuit to avoid interference from reflected waves or late waves, is set to 2000μs;

[0114] S2.3 Perform a lead-breaking test to calibrate the amplification / reduction factor of the sensor receiving amplitude, such as Figure 3 As shown;

[0115] S2.4 Conduct acoustic emission mechanical tests, monitor acoustic emission during the mechanical loading test, and obtain relevant acoustic emission positioning, parameters and waveform data, such as Figure 4 As shown;

[0116] S2.5 exports parameters and waveform data via ASCII output and ASCII waveform output respectively.

[0117] S3 sets the moment tensor model through sensor layout and positioning point data to calculate the fracture surface parameter trend The inclination angle δ, sliding direction ζ and inclination angle θ are used to identify the microcrack type by the advantage discrimination method proposed by Ohtsu et al.

[0118] Specifically, the moment tensor model is set by sensor layout and positioning point data to calculate the fracture surface parameter trend. The inclination angle δ, sliding direction ζ and inclination angle θ are used to identify the microcrack type by the advantage discrimination method proposed by Ohtsu et al., including:

[0119] S3.1 screens micro-rupture location points and searches for corresponding waveform data based on the timestamp of each location point;

[0120] S3.2 By combining the sensor layout data, the coordinates of each positioning point in S3.1 and the corresponding multiple acoustic emission parameter data, etc., moment tensor analysis and calculation are performed to obtain the components of moment tensor decomposition and the direction of fracture surface parameters. The inclination angle δ, the sliding direction ζ and the inclination angle θ;

[0121] S3.3 The shear component ratio X is normalized by using the dominance discrimination method for the moment tensor decomposition result in S3.2 to obtain a micro-fracture discrimination result X≤0.4, which is a tensile type; 0.4≤X≤0.6, which is a mixed type; X≥0.6, which is a shear type, as shown in FIG5 ;

[0122] S3.4 organizes the micro-fracture identification results in S3.3 to form a database of moment tensor analysis results as the output of subsequent deep learning network training.

[0123] S4 filters the waveform data by the timestamp of the acoustic emission positioning data, and then processes the waveform data to obtain corresponding image data;

[0124] Specifically, the filtering of waveform data by the timestamp of the acoustic emission positioning data and then processing the waveform data to obtain corresponding image data includes:

[0125] S4.1 searches for the corresponding waveform data based on the timestamp of each positioning point. One positioning point corresponds to 6 waveform data.

[0126] S4.2 denoises the waveform data, and then performs continuous wavelet transform (CWT) and wavelet packet transform (WPT) on the waveform to obtain three types of image data: denoised waveform image, CWT image and WPT spectrum image. The original waveform and the processed image data are shown in Figure 6;

[0127] S4.3 organizes the image data in S4.2 to form a database of image data obtained by waveform processing as input for subsequent deep learning network training.

[0128] S5 establishes a deep learning model based on the microcrack type identification results obtained in S3 and the corresponding image data obtained in S4, and explores the method of predicting microcrack types through waveform data;

[0129] Specifically, the deep learning model is established based on the microcrack type identification result obtained in S3 and the corresponding image data obtained in S4, and the method of predicting the microcrack type through waveform data is explored, including:

[0130] S5.1 builds 4 deep learning models (MolienetV2, InceptionV3, Resnet50 and VGG16) based on the results of S3 and S4, and splits the database into training set: test set = 4:1 for training and verification;

[0131] S5.2 comprehensively considers the priority and magnitude, processes and combines the evaluation indicators, designs a practical multi-indicator ranking system, evaluates the model training effect, and obtains the input-output combination model with the best effect. The ranking result is shown in Figure 7;

[0132] S5.3 saves the model with the best training effect, and can identify the micro-fracture type result by inputting an image data.

[0133] The acoustic emission microcrack type identification method of the present application integrates moment tensor and deep learning. The acoustic emission microcrack type identification method of the present application can quantitatively calculate the fracture surface parameters and qualitatively distinguish the microcrack type through moment tensor analysis; by using the denoised image data as the input of the deep learning model, the damage in the process of microcrack information transmission in the acoustic emission signal can be avoided as much as possible; by comparing the training results of each model to obtain the best combination of image input and training model type, a high-precision microcrack type identification model with an image as input can be saved.

[0134] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0135] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for predicting microcrack types using acoustic emission, characterized in that: The following steps are involved: Obtaining components of moment tensor decomposition of microcracks formed on the test sample to be identified, fracture surface parameter data, and waveform data of microcracks of the test sample to be identified, wherein the test sample to be identified undergoes an acoustic emission test to form microcracks on the test sample to be identified, and the microcracks are intermediate morphological microcracks when the acoustic emission test is not completed; The waveform data of the microcracks is input into a deep learning model for identification to obtain a prediction result of the microcracks of the test sample to be identified.

2. The acoustic emission microcrack type prediction method according to claim 1, characterized in that: The fracture surface parameter data of the microcracks formed in the test sample to be identified includes the fracture surface parameter trend The inclination angle δ, the sliding direction ζ and the inclination angle θ.

3. The acoustic emission microcrack type prediction method according to claim 1, characterized in that: The waveform data of the microcracks of the test sample to be identified is obtained by screening the timestamp of the acoustic emission positioning data.

4. The acoustic emission microcrack type prediction method according to claim 1, characterized in that: The method for establishing the deep learning model comprises the following steps: Obtaining the components of moment tensor decomposition, fracture surface parameter data, and microcrack waveform data of each sample in the test sample training set, wherein each test sample in the test sample training set undergoes an acoustic emission test to form microcracks on each test sample, and the microcracks are the final morphology of microcracks when the acoustic emission test is completed; The deep learning model is obtained based on the components of the moment tensor decomposition of each sample, the fracture surface parameter data, the microcrack waveform data of each sample, and the final morphological microcracks when the acoustic emission test is completed.

5. The acoustic emission microcrack type prediction method according to claim 4, characterized in that: The following steps are also included: Obtaining the final microcracks of the test sample to be identified when the acoustic emission test is completed; The components of moment tensor decomposition of microcracks of the test sample to be identified during the acoustic emission test, fracture surface parameter data, waveform data of microcracks, and the final morphological microcracks when the acoustic emission test is completed are added to the test sample training set.

6. The acoustic emission microcrack type prediction method according to claim 4, characterized in that: The deep learning models include MolienetV2, InceptionV3, Resnet50 and VGG16, and the acoustic emission microcrack type prediction method also includes selecting the optimal model by applying the test set test specimens, wherein the test method includes the following steps: Obtaining the components of moment tensor decomposition, fracture surface parameter data, and microcrack waveform data of each sample in the test sample test set, wherein each test sample in the test sample training set undergoes an acoustic emission test to form microcracks on each test sample, and the microcracks include intermediate morphology microcracks when the acoustic emission test is not completed, and final morphology microcracks when the acoustic emission test is completed; Inputting the microcrack waveform data of each sample corresponding to the intermediate form of each sample when the acoustic emission test is not completed into different deep learning models for identification, and obtaining the microcrack prediction result of each sample in the test sample test set; The predicted microcrack results of each sample are compared with the final morphological microcracks when the acoustic emission test is completed to obtain a comparison result, and the deep learning model with the best degree of consistency between the comparison results is used as the optimal model.

7. The method for predicting microcrack types using acoustic emission according to any one of claims 1, 4 and 6, characterized in that: The invention also includes performing data processing on the waveform of the microcracks, wherein the data processing includes the following steps: De-noising the waveform of the microcracks to obtain a waveform diagram of the microcracks after de-noising; Performing continuous wavelet transform on the denoised microcrack waveform to obtain a CWT waveform diagram of the microcrack; Performing wavelet packet transform on the denoised microcrack waveform to obtain a WPT waveform diagram of the microcrack; The waveform of the microcracks after data processing is obtained by sorting out the denoised waveform of the microcracks, the CWT waveform of the microcracks, and the WPT waveform of the microcracks.

8. An acoustic emission microcrack type prediction device, characterized in that: include: A data acquisition module, used to acquire each component of moment tensor decomposition of microcracks formed in the test sample to be identified, fracture surface parameter data, and waveform data of microcracks of the test sample to be identified, wherein the test sample to be identified undergoes an acoustic emission test to form microcracks on the test sample to be identified, and the microcracks are intermediate morphological microcracks when the acoustic emission test is not completed; The microcrack type prediction module is used to input the waveform data of the microcracks into the deep learning model for identification, so as to obtain the microcrack prediction result of the test sample to be identified.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an acoustic emission microcrack type prediction program, and when the acoustic emission microcrack type prediction program is executed by the processor, the steps of the acoustic emission microcrack type prediction method described in any one of claims 1-7 are implemented.

10. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores an acoustic emission microcrack type prediction program, and when the acoustic emission microcrack type prediction program is executed by the processor, the steps of the acoustic emission microcrack type prediction method described in any one of claims 1-7 are implemented.

Images