Aluminum alloy deformation mechanism acoustic emission semi-supervised identification method and device and electronic equipment
By employing a semi-supervised machine learning method and utilizing spectral feature vector similarity and cross-entropy loss function, the problem of accurate identification of crack and dislocation signals during aluminum alloy deformation was solved, and efficient identification of acoustic emission signals during aluminum alloy deformation was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XI AN JIAOTONG UNIV
- Filing Date
- 2024-05-08
- Publication Date
- 2026-06-09
Smart Images
Figure CN118332385B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to acoustic emission data processing methods, specifically to a semi-supervised identification method, apparatus, and electronic equipment for acoustic emission of aluminum alloy deformation mechanisms. Background Technology
[0002] Aluminum alloys are widely used in aerospace and other fields. A significant cause of their failure is damage under complex stress environments, including second-phase fracture, cracking at the interface between the second phase and the matrix, and the propagation of these two types of cracks from the crack initiation point into the matrix. Traditional destructive ex-situ testing cannot perform non-destructive online identification and monitoring of these crack damage behaviors. Acoustic emission technology, as a commonly used non-destructive testing method, has become one of the important methods for studying material fracture mechanisms and detecting material damage.
[0003] Due to the influence of second-phase strengthening particles, the acoustic emission signals during aluminum alloy deformation exhibit a complex interplay of dislocation and crack signals, making sudden fracture failure highly likely. However, identifying and labeling crack signals during aluminum alloy deformation is extremely difficult, making it impossible to use supervised learning methods like neural networks, which require massive amounts of label information, for identification and classification. Furthermore, there is a data imbalance between crack and dislocation signals, with the number of dislocation signals far exceeding the number of crack signals. This makes unsupervised learning clustering methods, such as K-means and Gaussian mixture models, which are heavily influenced by initial values, prone to local optima, and have poor confidence levels, unsuitable. Therefore, finding effective machine learning methods to improve the accuracy and effectiveness of identifying complex acoustic emission signals, including cracks, during aluminum alloy deformation is a pressing technical challenge. Summary of the Invention
[0004] To address the aforementioned deficiencies in existing technologies, this invention provides a semi-supervised method, apparatus, and electronic device for identifying acoustic emission mechanisms in aluminum alloy deformation. By employing a semi-supervised machine learning method, it can fully utilize unlabeled data and a small amount of labeled data to obtain a machine learning model for identifying acoustic emission signals and deformation mechanisms. This improves the accuracy and effectiveness of identifying complex acoustic emission signals, such as cracks, during aluminum alloy deformation.
[0005] This invention is achieved through the following technical solution:
[0006] According to one aspect of the present invention, a semi-supervised method for acoustic emission identification of aluminum alloy deformation mechanisms is provided, comprising:
[0007] Acoustic emission signals and characteristic parameter data were collected during the deformation process of aluminum alloys;
[0008] The acoustic emission signals are preprocessed based on the characteristic parameter data to obtain the feature vector representing each signal;
[0009] Calculate the similarity of acoustic emission signals in the feature space based on the feature vectors;
[0010] The unsupervised loss function term in semi-supervised learning is determined based on the similarity of acoustic emission signals in the feature space.
[0011] Based on the unsupervised loss function in the semi-supervised learning, a complete semi-supervised loss function is determined, and the machine learning model is trained using the signal feature vector. The trained machine learning model is then applied to identify the deformation mechanism of the acoustic emission signal of the aluminum alloy.
[0012] Preferably, the acoustic emission signals and characteristic parameter data of the aluminum alloy material deformation are collected, including the occurrence time, maximum amplitude, ring count, duration, energy parameters and rise time of the acoustic emission signals of the aluminum alloy material.
[0013] Preferably, the acoustic emission signals are preprocessed based on the characteristic parameter data, including filtering the acoustic emission signals based on the parameters corresponding to each acoustic emission signal, and removing acoustic emission signals with low ring count and low energy based on the ring count and energy parameters of each acoustic emission signal;
[0014] The feature vector of each acoustic emission signal is obtained, and the spectral response or power spectral density of each acoustic emission signal is calculated using Fourier transform or Welch method as the spectral feature vector of the signal.
[0015] The spectral eigenvectors of each signal are normalized.
[0016] Preferably, the similarity of the acoustic emission signals in the feature space is calculated based on the feature vectors, and the similarity is calculated by the distance between the spectral feature vectors.
[0017] Preferably, the complete semi-supervised loss function is determined based on the unsupervised loss function in semi-supervised learning. The supervised learning loss term in semi-supervised learning is the cross-entropy loss, and the total loss is the weighted sum of the supervised learning loss term and the unsupervised learning loss term.
[0018] Preferably, the machine learning model is trained using signal feature vectors based on the unsupervised loss function in semi-supervised learning, including:
[0019] The machine learning model is trained by minimizing the loss function of semi-supervised learning, and the trainable parameters in the classifier are optimized by using the gradient descent algorithm.
[0020] The model's performance is evaluated by scoring its recognition results on labeled data.
[0021] The score is the harmonic mean of the model's precision and recall.
[0022] Precision is the proportion of correctly predicted samples out of all samples predicted as positive; recall is the proportion of correctly predicted samples out of all samples that were actually positive.
[0023] Preferably, the deformation mechanism of acoustic emission signals of aluminum alloys is identified using a trained machine learning model, including:
[0024] By acquiring real-time acoustic emission signals and characteristic parameter data of aluminum alloys, and obtaining the feature vector representing each signal, data acquisition and processing for new experiments and applications can be achieved.
[0025] The signal's feature vector is used as the input to the model, and the forward computation function in the model is used to predict the signal's deformation mechanism.
[0026] The output of the machine learning model is collected, and the deformation mechanism category with the highest output probability is the result identified in real time.
[0027] According to one aspect of the present invention, a semi-supervised identification device for acoustic emission of aluminum alloy deformation mechanism is provided, comprising:
[0028] The acquisition module is used to acquire acoustic emission signals and characteristic parameter data of aluminum alloy materials in real time during deformation.
[0029] The preprocessing module is used to preprocess the acoustic emission signal based on the feature parameter data to obtain the feature vector representing each signal;
[0030] The calculation module is used to calculate the similarity of acoustic emission signals in the feature space based on the feature parameter data;
[0031] The determination module is used to determine the unsupervised loss function term in semi-supervised learning based on the similarity of acoustic emission signals in the feature space;
[0032] The training module is used to train a semi-supervised acoustic emission machine learning classifier model using the feature vectors of the signal;
[0033] The recognition module stores the trained model and identifies the deformation mechanism of the acoustic emission signal of aluminum alloy material online based on the feature vector of the new acoustic emission signal.
[0034] According to another aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the described semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism.
[0035] A computer-readable storage medium storing a computer program and machine learning model data, wherein the computer program, when executed by a processor, implements the steps of the semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism.
[0036] The present invention, by adopting the above technical solution, has the following beneficial effects:
[0037] 1. This invention uses a semi-supervised learning method to identify the deformation mechanism of acoustic emission signals during the deformation process of aluminum alloys. By involving supervised and unsupervised learning loss terms in the semi-supervised learning loss, it makes full use of unlabeled data and a small amount of labeled data, thus solving the problem of poor identification accuracy caused by the lack of labeled data and the difficulty in utilizing unlabeled data in the prior art.
[0038] 2. The unsupervised loss term in the semi-supervised learning method used in this invention involves calculating the similarity of acoustic emission signals in the feature space using the cosine distance between spectral feature vectors. This method has low computational requirements, high immediacy, and is suitable for signals of individual acoustic emission events with different durations. By using cosine distance, the influence of signal energy magnitude on similarity is masked, focusing more on the overall energy distribution differences of the acoustic emission signals, resulting in more accurate similarity estimation.
[0039] 3. This invention uses cross-entropy loss as the supervised learning loss term in semi-supervised learning, which makes the model focus on the misidentified labeled data during the training process, thereby further improving the model's recognition accuracy.
[0040] 4. The semi-supervised learning loss of this invention introduces a parameter that controls the importance of supervised learning and unsupervised learning. Adjusting this parameter balances the importance of unlabeled data and a small amount of labeled data. Therefore, the model has high recognition accuracy and effectiveness and can be applied to various situations where the amount of labeled data is not fixed. Attached Figure Description
[0041] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, do not constitute an undue limitation of the invention. In the drawings:
[0042] Figure 1 This is a technical flowchart of the acoustic emission semi-supervised identification method for aluminum alloy deformation mechanism of the present invention;
[0043] Figure 2 This is a schematic diagram of the semi-supervised learning framework and the determination of the loss function of this invention;
[0044] Figure 3 This invention provides a semi-supervised identification device for acoustic emission of aluminum alloy deformation mechanism, as illustrated in an embodiment of the present invention.
[0045] Figure 4This is a schematic diagram of the electronic device structure shown in an embodiment of the present invention;
[0046] Figure 5 This is a diagram showing the identification results of dislocation and crack acoustic emission signals during the aluminum alloy stretching process according to an embodiment of the present invention. Detailed Implementation
[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention.
[0048] As a specific embodiment of the present invention, combined with Figure 1 The present invention provides a semi-supervised method for acoustic emission identification of aluminum alloy deformation mechanism, which specifically includes the following steps:
[0049] S1. Real-time acquisition of acoustic emission signals and characteristic parameter data during the deformation process of aluminum alloy.
[0050] In this embodiment, acoustic emission signals and characteristic parameter data of aluminum alloy material deformation process are collected in real time by acoustic emission detection equipment, including the occurrence time, maximum amplitude, ring count, duration, energy parameters and rise time of acoustic emission signals of porous metal materials.
[0051] Acoustic emission (AE) technology captures high-frequency elastic waves released from within materials due to structural changes or defect evolution, enabling the perception of material deformation and damage over a continuous time period. It boasts a high sampling rate, resulting in high sensitivity and high temporal resolution. When acquiring AE data from aluminum alloys, an AE sensor is attached to the surface of the material under test, and an external force is applied. The high-frequency elastic waves released by the material are converted into electrical signals by a piezoelectric sensor, amplified by a preamplifier, and then processed by a data acquisition card. The data acquisition card is responsible for converting the AE electrical signals to digital signals, filtering and acquiring valid AE signals, and calculating and determining characteristic parameters such as energy and duration. Based on set amplitude threshold values, impact discrimination time, and other parameters, a set of valid AE signals and corresponding characteristic parameters are acquired.
[0052] S2. Preprocess the acoustic emission signal based on the characteristic parameter data to obtain the feature vector representing each signal.
[0053] The acoustic emission signals are filtered based on the parameters corresponding to each acoustic emission signal. According to the ring count and energy parameters of each acoustic emission signal, acoustic emission signals with low ring count and low energy are eliminated. In this embodiment, acoustic emission signals with a ring count of less than 3 are eliminated.
[0054] The feature vector representing each acoustic emission signal is obtained as follows:
[0055] 21) The power spectral density (PSD) of each acoustic emission signal is calculated using the Welch method as the spectral feature vector of the signal. Since the sensor’s sensitive frequency band is 95-850kHz, the frequency components above 1000kHz are close to 0, so the spectral feature data above 1000kHz are removed.
[0056] 22) Normalize the spectral feature vector of each of the above signals using the following formula:
[0057]
[0058] In the formula: G(f) is the original spectral eigenvector; f min f max These are the lowest and highest frequencies corresponding to the spectral feature vectors, namely 0 and 1000kHz, respectively.
[0059] S3. Calculate the similarity of acoustic emission signals in the feature space based on the feature vectors.
[0060] Similarity is calculated by the distance between spectral feature vectors:
[0061]
[0062] In the formula, D cos Let y be the cosine distance between the eigenvectors, and let x and y be the spectral eigenvectors of the two acoustic emission signals, respectively.
[0063] S4. Determine the unsupervised loss function term in semi-supervised learning based on the similarity of acoustic emission signals in the feature space, such as... Figure 2 As shown, specifically:
[0064] The unsupervised learning loss term in semi-supervised learning can be expressed as:
[0065]
[0066] In the formula, x i Let μ be the spectral eigenvector of the i-th acoustic emission signal. k Let be the average spectral feature vector of the acoustic emission signal identified as the kth deformation mechanism, N be the total number of acoustic emission signals, K be the number of existing deformation mechanisms, and D be the cosine distance between feature vectors.
[0067] S5. Based on the unsupervised loss function term, determine the complete semi-supervised learning loss function, and train the machine learning model using the signal feature vector.
[0068] In semi-supervised learning, the supervised learning loss term is the cross-entropy loss, which can be expressed as:
[0069]
[0070] In the formula, y ik , This describes whether the i-th acoustic emission signal belongs to the k-th deformation mechanism in both the actual situation and the model prediction. The complete semi-supervised learning loss function can be expressed as:
[0071]
[0072] In the formula, These represent the total loss value for semi-supervised learning, the loss term for supervised learning, and the loss term for unsupervised learning, respectively. β is a parameter that controls the importance of supervised and unsupervised learning, which is set to 0.05 in this example.
[0073] A machine learning model can be represented as:
[0074] y = f(x, Θ)
[0075] In the formula, y is the output of the machine learning classifier, f is the computation function of the model, x is the acoustic emission signal feature vector that serves as the input to the classifier, and Θ is the set of trainable parameters in the candidate model of the classifier.
[0076] The machine learning model is trained by minimizing the loss function of semi-supervised learning, and the trainable parameters in the classifier are optimized using the gradient descent algorithm. The gradient descent process can be represented as:
[0077]
[0078]
[0079] In the formula, γ is the learning rate.
[0080] After the machine learning model is trained, the F1 score of the model's recognition results on labeled data is calculated to evaluate the model's performance. The F1 score is the harmonic mean of the model's precision and recall, and can be expressed as:
[0081]
[0082] Taking the prediction of positive samples as an example, precision is the proportion of correctly predicted samples out of all samples predicted as positive, while recall is the proportion of correctly predicted samples out of all samples that are actually positive.
[0083] S6. Apply the trained machine learning model to identify the deformation mechanism of acoustic emission signals of aluminum alloy materials.
[0084] Specifically, it includes:
[0085] 61) Based on the real-time acquisition of acoustic emission signals and characteristic parameter data of aluminum alloy material deformation and the acquisition of feature vectors representing each signal, data acquisition and processing for new experiments and applications can be realized;
[0086] 62) The feature vector of the signal is used as the input of the model, and the forward calculation function in the model is used to predict the deformation mechanism of the signal;
[0087] 63) Collect the output of the machine learning model, and the deformation mechanism category with the highest output probability is the result identified in real time.
[0088] The following is a specific embodiment to more clearly explain the semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism according to the present invention:
[0089] Semi-supervised identification of acoustic emission signals of dislocations and cracks during the stretching process of aluminum alloys.
[0090] The parameters set in this embodiment are as follows: to retain as much detail as possible in the acquired acoustic emission signal while reducing the acquired background noise signal, the acoustic emission sampling frequency is set to 20MHz, and the preamplifier gain is 40dB. The acquired acoustic emission parameter data includes: occurrence time, maximum amplitude, duration, energy, rise time, and ring count.
[0091] Acoustic emission signals with a ring count less than 10 were discarded based on parameters. The power spectral density (PSD) of each signal was obtained using the "Welch" method, with each acoustic emission signal represented by a 51-dimensional power spectral density vector describing the frequency band from 0 to 1000 kHz at 20 kHz intervals. To eliminate the influence of energy differences between different signals, the power spectral density vector was normalized. This normalized vector is the characteristic vector of the signal.
[0092] Based on the analysis of the microscopic deformation mechanism of aluminum alloys, it is necessary to identify the acoustic emission signals induced by dislocation transport and crack propagation during the deformation process. The loss function is determined according to the method of this invention.
[0093] The model training is guided by a loss function, and the model employs a two-layer feedforward neural network. The extracted feature vectors are used as the model's input, and the loss function... Connected to the model's output, the neural network calculates the loss function through a forward pass and optimizes the parameters using backpropagation. The training process uses the Adam optimizer with L2 regularization applied and is trained for 10,000 iterations with an initial learning rate of 0.001. The trained model is evaluated and achieves an F1 score of 0.923.
[0094] The trained neural network model identifies the tensile process of the new aluminum alloy and predicts the deformation mechanism of acoustic emission signals in real time. The identification results of the acoustic emission signals during the complete tensile process are shown in [link to data]. Figure 3 As shown, the total acoustic emission signal is identified as two sets: dislocation signal and crack signal. According to an exemplary embodiment of the present invention, as... Figure 4 As shown, an acoustic emission semi-supervised identification device 100 for aluminum alloy deformation mechanism is used to implement the method, comprising:
[0095] Acquisition module 110 is used to acquire acoustic emission signals and characteristic parameter data of aluminum alloy deformation in real time;
[0096] The preprocessing module 120 is used to preprocess the acoustic emission signal according to the feature parameter data to obtain the feature vector representing each signal;
[0097] The calculation module 130 is used to calculate the similarity of acoustic emission signals in the feature space based on the feature parameter data;
[0098] The determination module 140 is used to determine the unsupervised loss function term in semi-supervised learning based on the similarity of acoustic emission signals in the feature space;
[0099] Training module 150 is used to train an acoustic emission machine learning classifier model using the feature vectors of the signal;
[0100] The recognition module 160 is used to store the trained model and to identify the deformation mechanism of the acoustic emission signal of aluminum alloy online based on the feature vector of the new acoustic emission signal.
[0101] According to exemplary embodiments of the present invention, such as Figure 5 As shown, the present invention provides an electronic device 200 for implementing a semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism, comprising: a data acquisition card 210, a memory 220, a communication bus 230, and a processor 240.
[0102] In one embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used to implement online identification of acoustic emissions from aluminum alloy deformation mechanisms.
[0103] In one embodiment of the present invention, a semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanisms, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data.
[0104] The computer storage medium can be any available medium or data storage device that a computer can access, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs)), optical storage (e.g., CDs, DVDs, BDs, HVDs), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).
[0105] This invention is not limited to the above embodiments. Based on the technical solutions disclosed in this invention, those skilled in the art can make some substitutions and modifications to some of the technical features without creative effort, and all such substitutions and modifications are within the protection scope of this invention.
Claims
1. A semi-supervised method for acoustic emission identification of aluminum alloy deformation mechanism, characterized in that, include: Acoustic emission signals and characteristic parameter data were collected during the deformation process of aluminum alloys; The acoustic emission signal is preprocessed based on the feature parameter data to obtain a feature vector representing each signal; Based on the feature vector, the similarity of the acoustic emission signals in the feature space is calculated; The unsupervised loss function term in semi-supervised learning is determined based on the similarity of the acoustic emission signals in the feature space. Based on the unsupervised loss function in the semi-supervised learning, determine the complete semi-supervised loss function, and train the machine learning model using the signal feature vector; A trained machine learning model was used to identify the deformation mechanism of acoustic emission signals from aluminum alloys. The similarity of acoustic emission signals in the feature space is calculated by the distance between their spectral eigenvectors: In the formula, D cos The cosine distance between the eigenvectors. x , y These are the spectral feature vectors of the two acoustic emission signals, respectively; The unsupervised loss function term in semi-supervised learning is determined based on the similarity of acoustic emission signals in the feature space, including: The unsupervised learning loss term in semi-supervised learning is represented as: In the formula, x i For the first i The spectral feature vector of an acoustic emission signal, μ k To be identified as the first k The average spectral eigenvector of the acoustic emission signal of the deformation mechanism, N The total number of acoustic emission signals. K The number of existing deformation mechanisms, D The cosine distance between the eigenvectors; Based on the unsupervised loss function in the semi-supervised learning, the complete semi-supervised loss function is determined. The supervised learning loss term in the semi-supervised learning is the cross-entropy loss, expressed as: In the formula, The actual situation and the model prediction are described respectively. i Is the acoustic emission signal the first one? k A kind of deformation mechanism; The complete semi-supervised loss function is expressed as: In the formula, These represent the total loss value for semi-supervised learning, the loss term for supervised learning, and the loss term for unsupervised learning, respectively. β Parameters used to control the importance of supervised and unsupervised learning.
2. The semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism according to claim 1, characterized in that, Acquire acoustic emission signals and characteristic parameter data of aluminum alloy deformation, including the occurrence time, maximum amplitude, ring count, duration, energy parameters, and rise time of the acoustic emission signals of aluminum alloy deformation.
3. The semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism according to claim 1, characterized in that, The acoustic emission signal is preprocessed based on the characteristic parameter data, including: Based on the parameters corresponding to each acoustic emission signal, the acoustic emission signals are filtered out. According to the ring count and energy parameters of each acoustic emission signal, acoustic emission signals with low ring count and low energy are eliminated. The feature vector of each acoustic emission signal is obtained, and the spectral response or power spectral density of each acoustic emission signal is calculated using Fourier transform or Welch method as the spectral feature vector of the signal. The spectral eigenvectors of each signal are normalized.
4. The semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism according to claim 1, characterized in that, Based on the unsupervised loss function in the semi-supervised learning, a machine learning model is trained using signal feature vectors, including: The machine learning model is trained by minimizing the loss function of semi-supervised learning, and the trainable parameters in the classifier are optimized by using the gradient descent algorithm. The model's performance is evaluated by scoring its recognition results on labeled data. The score is the harmonic mean of the model's precision and recall. Precision is the proportion of correctly predicted samples out of all samples predicted as positive; recall is the proportion of correctly predicted samples out of all samples that were actually positive.
5. The semi-supervised identification method for acoustic emission of aluminum alloy deformation mechanism according to claim 1, characterized in that, The deformation mechanism of acoustic emission signals from aluminum alloys is identified using a pre-trained machine learning model, including: By acquiring real-time acoustic emission signals and characteristic parameter data of aluminum alloys, and obtaining the feature vector representing each signal, data acquisition and processing for new experiments and applications can be achieved. The signal's feature vector is used as the input to the model, and the forward computation function in the model is used to predict the signal's deformation mechanism. The output of the machine learning model is collected, and the deformation mechanism category with the highest output probability is the result identified in real time.
6. An apparatus for a semi-supervised identification method of acoustic emission of aluminum alloy deformation mechanism as described in any one of claims 1-5, characterized in that, include: The acquisition module is used to acquire acoustic emission signals and characteristic parameter data of aluminum alloy deformation in real time; The preprocessing module is used to preprocess the acoustic emission signal according to the feature parameter data to obtain the feature vector representing each signal; The calculation module is used to calculate the similarity of acoustic emission signals in the feature space based on the feature parameter data; The determination module is used to determine the unsupervised loss function term in semi-supervised learning based on the similarity of acoustic emission signals in the feature space; The training module is used to train the acoustic emission machine learning classifier model using the feature vectors of the signal; The identification module is used to store the trained model and identify the deformation mechanism of the acoustic emission signal of aluminum alloy online based on the feature vector of the new acoustic emission signal.
7. An electronic device, characterized in that, Including the processor and memory; The memory is used to store computer programs, the computer programs including program instructions; The processor is used to invoke a program stored in the memory to execute the method as described in any one of claims 1-5.