Method, system and equipment for identifying internal defects of steel billet and medium
By using a joint model of a bidirectional long and short-term memory network and a support vector machine in the internal defect identification of special steel billets for electroslag remelting, the problem of relying on manual experience in the prior art is solved, and efficient, accurate and automated defect identification is achieved.
Patent Information
- Application Number
- CN202510068881.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-16
- Publication Date
- 2025-05-16
AI Technical Summary
The prior art relies on manual experience in the identification of internal defects of special steel billets for electroslag remelting, resulting in a highly subjective judgment result and a lack of automation and intelligence.
Using a joint model based on a bidirectional long and short-term memory network (Bi-LSTM) and a support vector machine (SVM), data is collected and preprocessed through flaw detection equipment, and the model is trained to identify internal defects of the billet.
It improves detection efficiency and accuracy, reduces interference from human factors, and realizes automated and intelligent detection, which is highly adaptable and scalable.
Smart Images
Figure CN120014329A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of machine learning technology, and in particular to a method, system, device and medium for identifying internal defects of steel billets. Background Art
[0002] Special steel for electroslag remelting is a special steel produced by electroslag remelting process, which is widely used in aerospace, nuclear industry, large forging manufacturing, precision mold manufacturing and special alloy preparation. As an intermediate product, the internal quality of special steel billet for electroslag remelting directly affects its final service performance. The internal defects of special steel billet for electroslag remelting mainly include looseness, shrinkage, inclusion aggregation and macro segregation. The law of defect heredity is mainly reflected in the retention of raw material defects and the amplification and expansion of defects: in terms of the retention of raw material defects, if the defects in the raw material are small and evenly distributed, these defects may be partially or completely eliminated due to the slag washing effect of the slag during the electroslag remelting process. However, for larger defects or unevenly distributed defects, electroslag remelting may not be able to completely eliminate them, resulting in these defects being retained in the billet; in terms of defect amplification and expansion, in some cases, small defects in the raw material may be amplified or expanded during the electroslag remelting process. This may be due to the high temperature, high pressure and complex physical and chemical effects during the electroslag remelting process. For example, if there are tiny cracks in the raw material, these defects may be enlarged or form new cracks during the electroslag remelting process.
[0003] Measures to reduce the heritability of defects Improving the quality of raw materials and selecting high-quality raw materials are fundamental measures to reduce the heritability of defects. Therefore, timely discovering defects inside the ingot through reasonable flaw detection methods is particularly important for ensuring the quality of the finished product. At present, the main method for non-destructive testing of electroslag ingots is traditional handheld flaw detection equipment. This method mainly relies on manual experience to determine defects, and the result is highly subjective. Therefore, it is necessary to establish a method for identifying internal defects of special steel billets for electroslag remelting based on support vector machine classifiers. Summary of the invention
[0004] The present application provides a method, system, equipment and medium for identifying internal defects of steel billets to solve the above-mentioned problems.
[0005] In one aspect, the present application provides a method for identifying internal defects of a steel billet, the method comprising the following steps:
[0006] Step S1: collecting detection data of internal defects of the steel billet through flaw detection equipment;
[0007] Step S2: inputting the data into a pre-trained steel billet internal defect recognition model for recognition; wherein the steel billet internal defect recognition model is constructed based on a joint model consisting of a bidirectional long short-term memory network and a support vector machine;
[0008] Step S3: Output the classification results of internal defects of the steel billet.
[0009] In one implementation of the present application, the step S1 further includes: preprocessing the collected data to ensure data quality and improve the generalization ability of the model; wherein the preprocessing includes: denoising, normalization, and data enhancement.
[0010] In one implementation of the present application, the model training process is specifically as follows:
[0011] Use a bidirectional long short-term memory network to process the detection data and obtain feature data;
[0012] The feature data is input into a support vector machine classifier, and a joint model consisting of a bidirectional long short-term memory network and a support vector machine is trained using a training data set.
[0013] In one implementation of the present application, the method further includes: adjusting model parameters to optimize the performance of the model; wherein the parameter adjustment includes: learning rate, number of hidden layer units, and kernel function.
[0014] In one implementation of the present application, the method further includes:
[0015] Use the validation data set to validate the trained model and evaluate the accuracy and false positive rate of the model;
[0016] Optimize the model according to the model optimization results.
[0017] The present application also provides a billet internal defect recognition system, the system comprising:
[0018] A data acquisition module is used to collect detection data of internal defects of steel billets through flaw detection equipment;
[0019] A prediction module is used to input data into a pre-trained steel billet internal defect recognition model for recognition; wherein the steel billet internal defect recognition model is constructed based on a joint model consisting of a bidirectional long short-term memory network and a support vector machine;
[0020] The result output module is used to output the classification results of internal defects of the steel billet.
[0021] The present application also provides a steel billet internal defect identification device, the device comprising:
[0022] at least one processor; and,
[0023] a memory communicatively connected to the at least one processor; wherein,
[0024] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can complete the aforementioned method for identifying internal defects of a steel billet.
[0025] The present application also provides a non-volatile computer storage medium for identifying internal defects of steel billets, which stores computer executable instructions. The computer executable instructions are executed by a processor to implement the aforementioned method for identifying internal defects of steel billets.
[0026] The present application provides a method, system, device and medium for identifying internal defects of steel billets, which have the following features:
[0027] Beneficial effects:
[0028] (1) Improve detection efficiency and accuracy. In terms of improving detection efficiency, the bidirectional long short-term memory network adds bidirectionality to the traditional long short-term memory network, that is, it considers both the forward and backward information of the input data. This enables the bidirectional long short-term memory network to capture the information in the time series data more comprehensively. In terms of improving detection accuracy, the bidirectional long short-term memory network model is divided into two independent long short-term memory networks, and the input sequence is input into these two long short-term memory network models in forward and reverse order for feature extraction. Then, the two output vectors are concatenated, and the formed word vector is used as the final feature expression of the word. Therefore, the underlying dimension of the bidirectional long short-term memory network is twice the hidden layer dimension of the ordinary long short-term memory network;
[0029] (2) Reduce interference from human factors. Manual experience judgment is often affected by multiple factors such as the operator's skill level, experience, and fatigue level, resulting in instability and uncertainty in the judgment results. The deep learning model is not affected by these human factors and can maintain consistent judgment standards and results. The support vector machine classifier (SVM) selected in this method maps the data to a higher-dimensional space by introducing a kernel function, thereby solving the problem of linear inseparability in the original space. In the classification of ultrasonic signals of internal defects in metals, appropriate kernel functions can be selected to adapt to different signal characteristics and classification requirements. The diversity of this kernel function provides SVM with more powerful classification capabilities and flexibility, which helps to reduce the impact of human errors on detection results and improve the reliability and stability of detection.
[0030] (3) Realize automated and intelligent testing. Deep learning technology can be combined with non-destructive testing equipment to achieve automated and intelligent testing. By training the deep learning model, it can automatically identify and process different types of defects, thereby reducing dependence on manual operations. This automated and intelligent testing method helps improve production efficiency, reduce labor costs, and make non-destructive testing more suitable for large-scale production and quality control.
[0031] (4) Strong adaptability and good scalability. Deep learning models have strong adaptability and scalability. With the continuous emergence of new materials and new processes, the field of non-destructive testing is facing more and more challenges. Deep learning models can adapt to new testing needs and challenges through continuous learning and optimization. In addition, deep learning models can also be combined with other technologies, such as sensor technology and big data analysis, to further improve the accuracy and efficiency of non-destructive testing. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0033] Figure 1 A flow chart of a method for identifying internal defects of a steel billet provided in an embodiment of the present application;
[0034] Figure 2 This is a sample diagram of a loose low-magnification round billet of the XTD electroslag series Ф650 specification provided in the embodiment of the present application;
[0035] Figure 3 This is a sample diagram of a XTD electroslag series Ф650 specification shrinkage cavity low-magnification round billet provided in the embodiment of the present application;
[0036] Figure 4 A network model diagram for classification and identification of internal defects of steel billets provided in an embodiment of the present application;
[0037] Figure 5 A scatter plot of the Bi-LSTM-SVM classification effect based on the mean value and standard deviation feature values provided in the embodiment of the present application;
[0038] Figure 6 A scatter plot of the Bi-LSTM-SVM classification effect based on the root mean square and skewness eigenvalues provided in the embodiment of the present application;
[0039] Figure 7 A scatter plot of the Bi-LSTM-SVM classification effect based on the margin factor and the amplitude factor eigenvalues provided in the embodiment of the present application;
[0040] Figure 8The model confusion matrix provided in the embodiment of the present application;
[0041] Fig. 9 The SVM parallel coordinate diagram provided by the embodiment of the present application;
[0042] Fig.10 The quantized conjugate gradient method optimization algorithm curve provided in the embodiment of the present application;
[0043] Fig.11 The gradient state and outlier change curve provided in the embodiment of the present application;
[0044] Fig.12 The error rates of the training set, validation set, and test set provided in the embodiments of the present application;
[0045] Fig.13 Fitting curves of the training set, validation set and test set provided in the embodiments of the present application;
[0046] Fig.14 A composition diagram of a steel billet internal defect recognition system provided in an embodiment of the present application;
[0047] Fig.15 A schematic diagram of a steel billet internal defect identification device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0048] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0049] The embodiments of the present application provide a method, system, device and medium for identifying internal defects of a steel billet. The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.
[0050] Figure 1 A flow chart of a method for identifying internal defects of a steel billet provided in an embodiment of the present application. Figure 1 As shown, the method mainly includes the following steps:
[0051] Step S1: collecting detection data of internal defects of the steel billet through flaw detection equipment;
[0052] Step S2: inputting the data into a pre-trained steel billet internal defect recognition model for recognition; wherein the steel billet internal defect recognition model is constructed based on a joint model consisting of a bidirectional long short-term memory network and a support vector machine;
[0053] Step S3: Output the classification results of internal defects of the steel billet.
[0054] The implementation process of the embodiment of the present application is described below in conjunction with specific applications.
[0055] Taking the XTD electroslag series Ф650 low-multiple round billets of a steel plant as the sample source, Figure 2 For loose defect samples, Figure 3 For shrinkage defect samples, CSV file samples of the detection data are collected by taking points around the handheld flaw detector, and the data training samples are formed by this. Core looseness and shrinkage are used as sample labels. There are 505 loose samples and 504 shrinkage samples, totaling 1009 samples. The sample composition of the training set, validation set, and test set is shown in Table 1.
[0056] Table 1 Sample composition of training set, validation set and test set
[0057]
[0058] The sample features are defined according to Table 2, with a total of 13 eigenvalues.
[0059] Table 2 Ultrasonic signal feature calculation method
[0060]
[0061] where s i is the amplitude of the sampled data points of the signal, and N is the number of sampled data points for each sample.
[0062] The data samples are normalized using the Min-Max method to convert the data into the range of [0,1]. This helps eliminate the dimensional differences between different data sets and improve the stability and accuracy of the model.
[0063] Enter the data Figure 4The network model shown in the figure has 8 layers and 484 learnable parameters, including feature input layer (Sequence feather input), batch normalization layer (BN), bidirectional memory network layer (BiLSTM), DropOut layer, activation function layer (Relu), fully connected layer (Fc), SoftMax (classification layer) and output layer (classout). Among them, the feature input layer is used to input feature values, the BN layer is used for data normalization and can accelerate the convergence speed, and the 45Bi-LSTM layer means that the Bi hidden size is 45, which is used for training and learning time-frequency feature information from the extracted features. In the case of "Drop Out 0.4", 40% of the probability samples in each training batch are discarded (that is, they do not participate in the training and calculation of the current batch). Relu and SoftMax are defined as the activation functions of the input layer and the output layer, and the SVM classifier is used to classify the features of the Bi-LSTM layer into 2 different types of sample features.
[0064] Import the file into the classification model and use MATLAB software to train the model, and the model accuracy is 99.8%. Figure 5 , Figure 6 , Figure 7 Shown is a scatter plot of the Bi-LSTM-SVM classification effect with different feature values.
[0065] Figure 8 The confusion matrix of the true value and the predicted value is shown, where the labels 1 and 2 represent porosity and shrinkage, respectively. The confusion matrix shows that the classification accuracy of sample 1, i.e., 505 loose samples, is 100%, but 2 samples of the 504 shrinkage samples are predicted incorrectly.
[0066] Fig. 9 The diagram shows a parallel coordinate graph, in which the data points of the SVM classifier are represented as broken lines. Each broken line represents a data sample, and its inflection point is located on a different parallel coordinate axis. Each axis represents a feature dimension. The position of the broken line on a certain axis represents the value of the sample on that dimension. This representation method can intuitively observe and analyze the distribution and change trend of data samples on different dimensions.
[0067] The model is trained using the quantized conjugate gradient optimization algorithm. The curves of the training set, validation set, and test set are as follows: Fig.10 As shown in Figure 2, it can be seen that the three curves are similar, which means that the effectiveness of the model is good and the selected algorithm is superior.
[0068] Fig.11 It shows the state of the model during the training process, mainly a change curve of the gradient state and outliers.
[0069] Fig.12 It is the statistics of error, where the center line represents no error. Theoretically, the closer the data is to the center line, the better the model calculation effect is. It can be seen from the figure that most of the model calculation errors are close to the center line.
[0070] Fig.13 Represents the fitting curves of the training set, validation set, and test set. The closer the curve is to 1, the better the fitting performance is, indicating that the true value of the sample and the predicted value are highly consistent, thus achieving a better classification effect.
[0071] The above is a method for identifying internal defects of a steel billet provided in an embodiment of the present application. Based on the same inventive concept, an embodiment of the present application also provides a system for identifying internal defects of a steel billet. Fig.14 A composition diagram of a steel billet internal defect recognition system provided in an embodiment of the present application, such as Fig.14 As shown, the system mainly includes: a data acquisition module 1401, which is used to collect detection data of internal defects of steel billets through flaw detection equipment;
[0072] Prediction module 1402, used for inputting data into a pre-trained steel billet internal defect recognition model for recognition; wherein the steel billet internal defect recognition model is constructed based on a joint model consisting of a bidirectional long short-term memory network and a support vector machine;
[0073] The result output module 1403 is used to output the classification results of internal defects of the steel billet.
[0074] The above is a steel billet internal defect recognition system provided by the embodiment of the present application. Based on the same inventive concept, the embodiment of the present application also provides a steel billet internal defect recognition device. Fig.15 A schematic diagram of a steel billet internal defect identification device provided in an embodiment of the present application, such as Fig.15 As shown, the device mainly includes: at least one processor 1501; and a memory 1502 that is communicatively connected to the at least one processor; wherein the memory 1502 stores instructions that can be executed by the at least one processor 1501, and the instructions are executed by the at least one processor 1501 so that the at least one processor 1501 can complete the aforementioned method for identifying internal defects of a steel billet.
[0075] In addition, an embodiment of the present application further provides a non-volatile computer storage medium for identifying internal defects of steel billets, which stores computer executable instructions, and the computer executable instructions are executed by a processor to implement the aforementioned method for identifying internal defects of steel billets.
[0076] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0077] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0079] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0080] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0081] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0082] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A method for identifying internal defects of a steel billet, characterized in that: The method comprises the following steps: Step S1: collecting detection data of internal defects of the steel billet through flaw detection equipment; Step S2: inputting the data into a pre-trained steel billet internal defect recognition model for recognition; wherein the steel billet internal defect recognition model is constructed based on a joint model consisting of a bidirectional long short-term memory network and a support vector machine; Step S3: Output the classification results of internal defects of the steel billet.
2. A method for identifying internal defects of a steel billet according to claim 1, characterized in that: The step S1 also includes: preprocessing the collected data to ensure data quality and improve the generalization ability of the model; wherein the preprocessing includes: denoising, normalization, and data enhancement.
3. A method for identifying internal defects of a steel billet according to claim 1, characterized in that: The model training process is as follows: Use a bidirectional long short-term memory network to process the detection data and obtain feature data; The feature data is input into a support vector machine classifier, and a joint model consisting of a bidirectional long short-term memory network and a support vector machine is trained using a training data set.
4. A method for identifying internal defects of a steel billet according to claim 3, characterized in that: The method also includes: adjusting model parameters to optimize model performance; wherein the parameter adjustment includes: learning rate, number of hidden layer units, and kernel function.
5. A method for identifying internal defects of a steel billet according to claim 3, characterized in that: The method further comprises: Use the validation data set to validate the trained model and evaluate the accuracy and false positive rate of the model; Optimize the model according to the model optimization results.
6. A steel billet internal defect recognition system, characterized in that: The system comprises: A data acquisition module is used to collect detection data of internal defects of steel billets through flaw detection equipment; A prediction module is used to input data into a pre-trained steel billet internal defect recognition model for recognition; wherein the steel billet internal defect recognition model is constructed based on a joint model consisting of a bidirectional long short-term memory network and a support vector machine; The result output module is used to output the classification results of internal defects of the steel billet.
7. A steel billet internal defect identification device, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can complete the method for identifying internal defects of a steel billet as described in any one of claims 1-5.
8. A non-volatile computer storage medium for identifying internal defects of steel billets, storing computer executable instructions, characterized in that: The computer executable instructions are executed by a processor to implement a method for identifying internal defects of a steel billet as described in any one of claims 1 to 5.